1.1 Purpose

Every commercial medium used in the IVF laboratory is a hypothesis about what a gamete or an embryo needs, expressed as a list of concentrations. This page sets out those constituents, the physiological or physicochemical reason each one is there, and the evidence behind that reason. It covers embryo culture media (sequential and single-step), buffered handling media, fertilization and insemination media, sperm preparation and cryopreservation media, embryo transfer media, the oil overlay, and vitrification and warming solutions. It then follows what happens to a medium once it is in use: ammonium generation, pH and osmolality drift, substrate depletion, peroxide and photoproduct formation, and the storage instabilities that matter to a quality system.

Some media used in the laboratory are outside this page's scope and are not evaluated here: oocyte maturation media, both conventional IVM and the two-step capacitation (CAPA) IVM system, in which maturation of cumulus-oocyte complexes and rescue maturation of denuded immature oocytes are different questions and in which a prospective study allocating sibling oocytes to maturation medium with or without amphiregulin has tested that supplement [1]; follicular flushing media; artificial oocyte activation solutions, which the ESHRE add-on recommendations address [2]; pharmacological sperm motility enhancers; and the media for ovarian and testicular tissue cryopreservation. Conclusions drawn here about embryo culture media should not be carried over to those categories.

The page is written for laboratory directors, senior embryologists, andrologists and research scientists. It assumes familiarity with the biology and concentrates on mechanism, evidence and consequence. Where the literature disagrees, the disagreement is stated rather than smoothed over.

1.2 How the evidence is presented

Numbered citations are placed after the statements they support; a citation at the end of a passage covers the findings summarised in it, not every inference drawn from them. Where a sentence describes common laboratory or commercial practice rather than a study result, it is this page's description of practice and carries no citation. Citations are numbered in order of first appearance, the reference list is published as a companion page because of the size limit of a single web page (Section 18), and each entry in it leads to a journal record (PubMed and DOI where they exist), an official document, or a manufacturer document, so that any statement can be checked at source; where a cited paper carries an indexed erratum the reference entry says so and whether the correction was reviewed. Where a result drives a conclusion, the text gives the design, the unit of allocation (patient, cycle, oocyte or sibling embryo), the primary endpoint and the counts or confidence interval where the source reports them; percentages are quoted as reported, sibling designs are identified because their units are not independent patients, and a non-significant result is not read as equivalence, a subgroup as an indication, or a comparison of two products as a class effect.

This is a narrative reference rather than a systematic review: primary papers are used for mechanism and history, and current systematic reviews and professional guidance for clinical conclusions. The reference base is anchored in the work of Patrick Quinn (Human Tubal Fluid medium and its glucose- and phosphate-free derivatives), David Gardner and Michelle Lane (amino acids, ammonium and sequential media), John Biggers and colleagues (simplex optimisation, KSOM and the single-medium argument), Henry Leese (embryo metabolism), and the composition and stability analyses of Dean Morbeck, Jason Swain and others. From that anchor set the literature was expanded by a citation-network search run through PubMed, Europe PMC and CrossRef, last run on 19 September 2026: the reference lists of the anchor papers, the reference lists of the papers most often cited among them, and the papers published since 2019 that cite them, ranked by the number of anchor references they share, together with topic searches for each section; candidates were screened against the text of the section they would support, and every numerical statement taken from an added paper was checked against its abstract. A search built outward from a set of investigators reaches historical depth more reliably than it reaches independent clinical trials, so statements that a finding is the only one, the first, or has not been tested are bounded to the studies reviewed here rather than to the literature as a whole.

Commercial products are named where they anchor the science, for example HTF, P-1, KSOM, G-1 and G-2, Global, CSCM, G-TL, EmbryoGlue, SSS and the Cryotop solutions. Naming a product is not an endorsement. Formulations change, and manufacturers do not always publish full compositions, so any lot-specific claim must be taken from the current instructions for use and certificate of analysis rather than from this page.

1.3 Three ideas that organise everything else

Three ideas recur throughout. First, the preimplantation embryo changes its physiology as it develops: before compaction it relies mainly on pyruvate and lactate, uses glucose sparingly and is highly sensitive to its environment; after compaction it forms a transporting epithelium, its glucose consumption rises and it becomes more capable of regulating its own interior, with the substrate preferences overlapping rather than switching outright [3–6]. Second, media are not inert. At 37 °C they generate ammonium, lose CO₂ the moment they leave the incubator, concentrate as water evaporates, and accumulate the embryo's own metabolic products [7–9]. Third, the culture environment can leave a mark that outlasts the culture period: in randomised human trials, birthweight and early growth differed between commercial media, and in mouse models imprinted gene methylation differed between media, although a 2024 network meta-analysis found no significant overall media effect on birthweight, so the human evidence is mixed [10–13]. A laboratory that understands its media is better placed to recognise the changes that occur in its incubators unseen.

2.1 Mouse embryo culture and the discovery of the energy substrates

Modern IVF media descend directly from mouse embryo culture in the 1950s and 1960s. Whitten cultured eight-cell mouse embryos to blastocysts in a simple bicarbonate-buffered salt solution with glucose and albumin, and then extended culture back to the two-cell stage by adding lactate [14, 15]. Brinster introduced the microdrop-under-oil method that the field still uses, and in a series of systematic studies defined the osmolality and pH optima for two-cell mouse embryos and showed that lactate, pyruvate, oxaloacetate and phosphoenolpyruvate could each supply energy for development, whereas glucose and a long list of other intermediates could not [16–18]. Biggers, Whittingham and Donahue then showed that the mouse oocyte and zygote require pyruvate specifically, establishing the pattern of energy metabolism that underlies every cleavage-stage medium since [19]. Whittingham's M16 medium formalised this knowledge into a bicarbonate-buffered formulation with pyruvate, lactate and glucose that became the laboratory standard for the next two decades [20].

Two further lessons from this period still shape formulation. Chelation of divalent cations with EDTA was found to allow one-cell embryos from otherwise blocking mouse strains to develop in a chemically defined medium [21], and it was recognised that glucose and phosphate together could arrest hamster embryos at the two-cell stage by driving premature glycolysis [22]. Removal of glucose, addition of glutamine and EDTA, and a raised lactate to pyruvate ratio gave the CZB medium that carried random-bred mouse zygotes through the two-cell block [23]. Before the simplex work, Lawitts and Biggers had shown that the block itself could be overcome in a conventional medium by lowering the standard concentrations of sodium chloride, potassium chloride, potassium phosphate, glucose or pyruvate, singly or in combination, with the effect of the salts depending on the bicarbonate concentration, although blastocyst yields stayed low [24]. The benefit of glutamine is strain-dependent: in F1 hybrid embryos 1 mM glutamine did not change the proportion reaching the blastocyst but reduced blastocyst cell number through trophectoderm cell death, and renewing the medium at 48 h did not reverse it, so that glutamine-free KSOM gave cell numbers approaching in vivo values [25].

2.2 The first human media

Human IVF began with somatic-cell culture media and simple salt solutions. Edwards, Steptoe and Purdy fertilised and cleaved human oocytes in vitro in 1970 using media of that era [26], and the first birth followed in 1978 [27]. Ménézo formulated the B2 medium in 1976 as a defined medium closer to genital-tract fluids than to serum, and it supported gamete survival, oocyte maturation and cleavage in several species [28]. Early clinical experience showed that the details of medium handling mattered as much as the formulation: Quinn and colleagues associated adequate gassing and equilibration, double rinsing of pipettes and catheters, HEPES-buffered medium for manipulation outside the incubator, and control of temperature near the embryo with higher and more consistent pregnancy rates [29]. Ménézo and Testart also demonstrated that serum was not necessary for human fertilization, culture or transfer when albumin was substituted, opening the route to serum-free systems [30].

2.3 Quinn's Human Tubal Fluid medium (1985) and its glucose- and phosphate-free derivative

In 1985 Quinn, Kerin and Warnes formulated a medium on the measured composition of human tubal fluid, drawing on the in vivo collections and analyses of Lippes and colleagues and of Borland and colleagues, notably its high potassium concentration relative to serum [31, 32]. In a randomised 2 × 2 trial comparing HTF with the modified Tyrode's medium T6, and culture under oil with culture in loosely capped tubes, HTF gave a clinical pregnancy rate of 30% of 60 transfers against 11% of 53 with T6. Reducing the potassium content of HTF to that of T6 significantly reduced mouse zygote development, and the authors concluded that the higher potassium was the likely main contributor, while allowing that other compositional differences could have played a part [33]. HTF contains, in mM, sodium chloride 101.6, potassium chloride 4.69, magnesium sulfate 0.2, potassium dihydrogen phosphate 0.37, calcium chloride 2.04, sodium bicarbonate 25.0, glucose 2.78, sodium pyruvate 0.33 and sodium lactate 21.4, with penicillin, streptomycin and phenol red [33]. The elevated potassium and chloride and relatively low calcium mirror the elemental analysis of human oviduct fluid reported by Borland and colleagues [32].

A decade later Quinn revisited the formulation in light of the glucose and phosphate findings in animal embryos. A modified HTF containing EDTA and glutamine but devoid of glucose and phosphate ("Basal XI HTF", later commercialised as P-1) improved development of both F1 hybrid and blocking-strain mouse zygotes and gave high fertilization, cleavage, implantation and blastocyst rates in human IVF [34, 35]. An independent randomised comparison in 109 unselected patients with sibling oocytes found fertilization of 58.6% in HTF against 62.5% in P1, embryo quality of 68.7% in P1 where sibling HTF quality was 15.4%, and implantation after 48 h transfer of 12.2% in P1 against 6.8% in HTF [36]; the one large randomised trial of glucose omission itself, in 741 patients, found better embryo quality but similar pregnancy rates and called reduced glucose prudent rather than proven (Section 5.1) [37]. Quinn was careful to note that the physiological phosphate level in the female tract, the true phosphate requirement, and the specificity of amino acid requirements before and after compaction remained open questions [38]. His later reviews credit these formulation changes with a substantial share of the improvement in clinical outcomes over the 1990s (the 2000 review carries an indexed erratum that could not be retrieved; nothing here depends on a specific value from it) [39, 40].

2.4 Simplex optimisation: SOM, KSOM and KSOM^AA

Lawitts and Biggers took an empirical rather than a physiological approach. Using sequential simplex optimisation, which adjusts several components simultaneously and follows the response, they identified sodium chloride, pyruvate, potassium phosphate and glucose as the components whose high concentrations were most detrimental to pronuclear mouse embryos, and produced Simplex Optimised Medium (SOM); removal of phosphate gave the largest single gain [41]. In SOM, glucose at 5 mM had no inhibitory effect, raising sodium chloride was progressively inhibitory, and glutamine protected against high salt by acting as an organic osmolyte [42]; betaine had the same protective action, and electron-probe measurements showed that osmolytes limited the rise in intracellular sodium and fall in potassium caused by high sodium chloride [43]. Raising the potassium concentration to 2.5 mM gave KSOM, which carried zygotes from outbred CF1 females past the two-cell block to blastocysts at 88% with full compaction [44, 45]. Adding Eagle's essential and non-essential amino acids (KSOM^AA) further increased blastocyst formation, hatching and cell number, and produced blastocysts whose expression of a panel of metabolic and transcriptional genes was indistinguishable from in vivo controls [46]; genome-wide, KSOM^AA blastocysts later proved closer to in vivo blastocysts than those from Whitten's medium (29 against 114 genes altered) but not identical to them [47]. In protein-free KSOM the amino acids supported hatching, inner cell mass proliferation and basement membrane formation, demonstrating that a chemically defined, protein-free medium could carry the mouse zygote to the late blastocyst [48]. KSOM has an osmolality near 255 mOsm/kg, reflecting the low-sodium optimum found by simplex [45, 49].

2.5 Gardner and Lane: amino acids, ammonium and sequential media

Gardner and Lane's 1993 paper is the point from which modern amino acid formulation follows. Eagle's 20 amino acids increased mouse blastocyst formation, cell number and hatching, the benefit was attributable mainly to the non-essential group, and continuous culture in amino acids slowed cleavage after about 72 h unless the medium was renewed, which the authors attributed to accumulation of an inhibitory product, most likely ammonium [50]. The following year they showed in transfer experiments that the embryo's nitrogen requirement switches during the preimplantation period: non-essential amino acids and glutamine supported the highest implantation rates from 48 h cultures, all 20 amino acids gave the highest implantation after 93 h, and essential amino acids increased fetal development per implantation. Ammonium in the medium retarded fetal growth and induced the neural tube defect exencephaly in a time- and concentration-dependent manner [51]. Differential regulation was then mapped in detail: non-essential amino acids and glutamine drive cleavage rates to the eight-cell stage, after which essential amino acids increase cleavage, inner cell mass development and fetal development after transfer [52].

The clinical corollary was sequential media. Gardner and Lane argued that the changing physiology and nutrient requirements of the embryo call for more than one medium across the preimplantation period, and that the conditions optimal for the zygote are not those optimal for the blastocyst [3, 53]. The first medium (G1) was formulated on oviduct fluid, with low glucose, high lactate, pyruvate, non-essential amino acids, glutamine and EDTA; the second (G2) on uterine fluid, with higher glucose, lower lactate, all 20 amino acids, vitamins and no EDTA [54, 55]. In a private-practice programme, blastocysts cultured in these media implanted at twice the rate of day 3 embryos and fewer embryos were required per transfer [56]. The same laboratory later showed that blastocyst score predicts implantation, opening the path to single blastocyst transfer [57].

2.6 Two philosophies: "back to nature" and "let the embryo choose"

Summers and Biggers named the two approaches to formulation: the "back to nature" approach, which sets concentrations to match the female tract and changes them as the embryo moves from oviduct to uterus, and the "let the embryo choose" approach, which optimises a single medium empirically and supplies a broad set of nutrients throughout, trusting the embryo's own transport and metabolic regulation [58]. Biggers and Racowsky demonstrated that a one-step protocol in KSOM^AA supported human blastocyst formation at rates not significantly different from a sequential P-1 to CCM system, with births following transfer of those blastocysts [59]. Biggers and Summers subsequently reviewed the arguments for two-step protocols, resting on glucose, EDTA, glutamine and amino acids, and concluded that each was equivocal [60]. Gardner's position is that amino acids and specific macromolecules are the key components and that temporal dynamics are an important medium characteristic [61]. Both camps agree on more than they dispute: reduced oxygen, amino acids, minimal ammonium exposure, and a rigorous quality system. Section 7 returns to the clinical evidence on sequential versus single-step culture, which shows no consistent superiority of either design without demonstrating that all formulations are equivalent [62, 63].

2.7 The parallel history in domestic species

Human media did not develop in isolation. Tervit, Whittingham and Rowson cultured sheep and cattle ova to the blastocyst stage in 1972 in a medium formulated on Restall and Wales's chemical analysis of sheep oviduct fluid, and Synthetic Oviduct Fluid (SOF) became the base of ruminant embryo culture in the same way that HTF later became a base for human culture [64, 65]. Reduced oxygen, not somatic co-culture, was what first carried ruminant embryos through the block in SOF: nuclei per embryo peaked at 8% oxygen in sheep and at 4 to 8% in cattle and were lowest at 20% [66], and in a simple SOF-like medium glucose was unnecessary and deleterious to bovine morula development, lactate served as well as pyruvate, and the twenty Eagle's amino acids dramatically improved morula development and cell number whereas vitamins did not [67]. The ruminant embryo is more demanding than the mouse: the transition from maternal to embryonic control falls at the 8- to 16-cell stage in sheep and cattle, against the 2-cell stage in the mouse and the 4- to 8-cell stage in the human, and it is there that ruminant embryos arrest in inadequate media [68]. Several of the components now standard in human media were first shown to matter at that block. In sheep zygotes cultured in SOF with albumin, Eagle's amino acids did not reduce arrest at the 8- to 16-cell stage unless the medium was renewed every 48 h to relieve ammonium, in which case arrest fell to 6%, blastocyst formation rose from 29% to 67% and cell number from 52 to 105; culture in groups of four gave 173 cells, equal to in vivo blastocysts, and sheep blastocysts produced more ammonium from amino acids than mouse blastocysts did, which the authors read as greater amino acid use by the ruminant embryo. These are the origins of the amino acid, ammonium and group-culture principles carried into the G1 and G2 system (Sections 2.5, 7.3 and 14.1) [69]. The stage-dependent amino acid switch that Lane and Gardner documented in the mouse (Section 2.5) was shown in bovine zygotes too: non-essential amino acids with glutamine in the first 72 h increased development to the 8- to 16-cell stage and to blastocyst, essential amino acids in that window had no effect, and all twenty amino acids after day 4 increased blastocyst development and cell number [70]. Holm and colleagues then showed that a fully defined bovine system, SOF with amino acids, citrate and myo-inositol and polyvinyl alcohol in place of protein, could match serum, albumin and co-culture for blastocyst development (42% against 36% in co-culture) and produce normal calves, although serum was still needed during maturation and fertilization for normal fertilization in their hands [71]; the citrate in that system has its own history, having been identified as the low-molecular-weight embryotrophic contaminant of commercial bovine serum albumin [72].

The domestic species also delivered the warning that removed serum from embryo culture. Lambs from ovine embryos cultured in SOF with 20% human serum weighed 4.2 kg against 3.4 kg for controls and 3.5 kg for embryos cultured in serum-free SOF with albumin and amino acids, with a longer gestation, and the serum-cultured embryos carried abundant lipid droplets [73]. Young, Sinclair and Wilmut named the large offspring syndrome of cattle and sheep, with its organ defects, difficult parturition and fetal and neonatal loss, and noted that in vitro culture was one of four situations that produced it, that its programming was unpredictable, and that the responsible factors and mechanisms had not been identified [74]. Section 15 takes up the human counterpart of that question. Bovine in vitro production has since become a large commercial technology in North America, South America and Europe, combined with sexed semen and genomic selection, and its remaining limitations, lower cryotolerance and lower pregnancy rates of in vitro produced embryos relative to in vivo embryos, are a reminder that a culture system can support development without reproducing it [75]. The bovine embryo has also become a candidate release test for human media, discussed in Section 16.1.

3.1 Carbohydrates change along the tract and across the cycle

The "back to nature" formulations rest on direct measurement of human tract fluids. Gardner, Lane, Calderon and Leeton collected oviduct and uterine fluid across the menstrual cycle and measured pyruvate, lactate and glucose by microfluorimetry. Oviduct pyruvate did not vary with cycle day (mean 0.24 mM); oviduct lactate rose from 4.87 mM in the follicular phase to 10.50 mM at ovulation, and oviduct glucose fell from 3.11 mM in the follicular phase to 0.50 mM at mid-cycle before rising again to 2.32 mM in the luteal phase. Uterine fluid was constant across the cycle at 0.10 mM pyruvate, 5.87 mM lactate and 3.15 mM glucose, all significantly different from mid-cycle oviduct values. Cumulus cells consumed glucose avidly and released lactate, so the fertilising oocyte and cleavage-stage embryo sit in a low-glucose, high-lactate microenvironment [55]. The mouse precedent was Gardner and Leese's microanalysis of oviduct fluid, which found pyruvate 0.37, glucose 3.40 and lactate 4.79 mM near the cumulus mass, pyruvate falling to 0.14 mM and glucose rising to 5.19 mM without cumulus cells, and which showed that a medium matched to those values supported blastocysts as well as M16 while culture in either raised the fraction of glucose converted to lactate from 44% in fresh blastocysts to 73% and 91% [76]. An independent human dataset from vascularly perfused Fallopian tubes gave glucose 1.11, pyruvate 0.14 and lactate 5.4 mM, with arginine, alanine and glutamate the most abundant of 17 amino acids, all below perfusate levels [77]; the difference from the mid-cycle value of 0.50 mM above reflects both collection method and cycle phase, since the perfused tubes were from the proliferative and early secretory phases, and is one reason tract fluid is a benchmark rather than a recipe. The G1 and G2 carbohydrate profiles were formulated on these measurements: G1 with glucose 0.5 mM and lactate 10.5 mM, and G2 with glucose 3.15 mM, pyruvate 0.10 mM and lactate 5.87 mM, while G1 pyruvate was set at 0.32 mM, above the 0.24 mM mean measured in the oviduct [53, 54].

3.2 Ions, protein and osmolality

Electron-probe analysis of relatively undisturbed human oviduct fluid found high potassium and chloride and low calcium relative to serum, with sodium and magnesium similar to serum [32]. Undiluted human uterine fluid has the same osmolality as serum but lower total cation and albumin concentrations, with potassium high and sodium and calcium low relative to serum, and potassium and calcium varying cyclically [78]. Tubal fluid is actively secreted by the oviduct epithelium and its ion and nutrient composition is regulated rather than a simple serum transudate [79]. Leese and colleagues frame the oviduct and endometrial epithelia as the final link in a supply line from maternal diet to embryo, which makes tract fluids a plausible route for developmental programming and a benchmark, not a blueprint, for culture media [80]. The most recent attempt to build a medium directly on human oviduct fluid, with 31 components set to values measured in laparoscopic samples from 28 patients, found that the tract carries lower essential and higher non-essential amino acid concentrations than current media, and in a multicentre randomised trial of 9633 zygotes from 1435 patients reported higher good-quality and utilised blastocyst rates than the control medium, significantly so in older women [81].

3.3 Oxygen

Direct polarographic measurement in rhesus monkeys, hamsters and rabbits found oviduct and uterine oxygen tensions far below atmospheric, from about 60 mmHg (8.7% O₂) in the rabbit oviduct to 11 to 14 mmHg (about 1.5%) in the monkey uterus throughout the cycle, with intrauterine oxygen falling further around the time of blastocyst formation and implantation in hamsters and rabbits [82]. Fibre-optic measurement at the human endometrial surface at mid-cycle gave a mean of 11.8% of air saturation, which corresponds to roughly 2.5% oxygen when air saturation is referenced to atmospheric air at about 21% oxygen, with wide inter-patient variation and minute-scale oscillations [83]. The rhesus, hamster and rabbit values above are given as the authors reported them, as partial pressures with their own percentage conversions [82]. These measurements are the physiological basis for the reduced-oxygen culture discussed in Section 4.5. A systematic review of in vivo oxygen, pH and temperature measurements in the female tract of any species, covering 1946 to 2015, is the citable synthesis of this literature [84].

4.1 Water, endotoxin and the regulatory definition of a medium

Water is the constituent present at the highest concentration and the one most easily contaminated. Under 21 CFR 884.6180 the FDA defines reproductive media and supplements as products in direct physical contact with human gametes or embryos, explicitly including water, acid solutions, rinsing solutions, sperm separation media, supplements and the oil used to cover media; supplements are reagents added to enhance specific properties, such as proteins, sera and antibiotics. The category is Class II with special controls that include mouse embryo assay information, endotoxin testing, sterilisation validation, design specifications, labelling, biocompatibility testing and clinical testing; phosphate-buffered saline for washing, culture oil used as an overlay, and water for assisted reproduction are exempt from premarket notification within stated limits [85].

Endotoxin deserves its place among the special controls. Dumoulin and colleagues found endotoxin in all five commercial albumin preparations they tested, at varying concentrations, and showed that human sperm survival and mouse IVF and culture were unaffected by relatively high endotoxin levels, so those assays cannot be relied on to detect it [86]. Nagata and Shirakawa, using the Limulus amoebocyte lysate test across 163 human IVF cycles, found clinical pregnancy and live birth rates of 26.1% and 20.7% with media below 1 pg/mL endotoxin against 9.9% and 5.6% at or above 1 pg/mL, with no gestational sac or fetal heart in media above 2 pg/mL, and proposed a standard below 1 pg/mL with an allowable limit of 2 pg/mL [87]. Those picogram values belong to that study's assay and era. Endotoxin is now reported on certificates of analysis in endotoxin units (EU/mL), a measure of biological activity, and there is no universal mass-to-EU conversion across endotoxin sources and assays, so the historical association should not be read as a current acceptance criterion; the current product specification and a locally set acceptance limit are what the receiving inspection checks [88, 89].

4.2 Osmolality and cell volume regulation

Brinster established that two-cell mouse embryos have a defined osmolality optimum, and that osmolality, pH and energy source can be varied independently, with an interaction only between energy source and pH [17, 18]. The reason osmolality matters so much to the early embryo was worked out by Baltz and colleagues. Mouse embryos are inhibited identically by raised osmolality produced with sodium chloride or with raffinose, so the effect is osmotic rather than ionic, and a set of organic osmolytes that are substrates of the glycine and beta amino acid transport systems (glycine, glutamine, betaine, proline, beta-alanine and hypotaurine) protect zygotes and allow development past the two-cell block, with glycine effective at an EC₅₀ of about 50 µM [90]. The glycine transporter GLYT1 is activated at ovulation and regulates steady-state intracellular glycine, and hence cell volume, through osmotically controlled uptake and efflux [91, 92]. The transporter that carries glycine as an osmolyte was identified as GLYT1, whose osmotically regulated activity is immediate and independent of protein synthesis, with osmotic stress and glycine's protection evident in vitro even at oviductal osmolarity [93], and human cleavage-stage embryos transport glycine through a saturable, sarcosine-inhibitable system consistent with GLYT1, with spare human eggs showing the same swelling-activated currents as mouse zygotes [94]. Taurine is released by human and mouse oocytes and embryos on hyposmotic challenge or Na⁺/K⁺-ATPase inhibition, acting as an osmolyte in the same way [95]. Raised osmolality or raised glucose and phosphate induce the two-cell block even in "non-blocking" F1 hybrid mouse embryos, only at a higher threshold, and glycine rescues the osmotic block [96].

The practical consequences are that human embryo media are formulated between roughly 255 and 295 mOsm/kg, that KSOM-type media sit at the low end and HTF-type media in the upper half of that range [49, 97], and that any process which concentrates the medium is a stressor. Swain and colleagues showed that small drop volume, warm stages and pipetting drops before overlay each raise osmolality within minutes, that airflow interacts with them to raise it further, together by as much as 40 mOsm/kg, and that mouse development is impaired above 310 mOsm/kg [98]. Evaporation during culture is treated in Section 14.3.

4.3 pH: the bicarbonate and CO₂ system and intracellular pH regulation

Culture media are buffered by sodium bicarbonate in equilibrium with CO₂ in the gas phase, so that medium pH is set jointly by the bicarbonate concentration, the CO₂ partial pressure and altitude. Most manufacturers recommend a pH of 7.2 to 7.4. The first randomised comparison across that range, a double-blind multicentre trial of 424 women stopped before its planned size, cultured embryos from ICSI to fresh day 3 transfer at a measured pH of 7.22 or 7.38 and found live birth per retrieval of 29.3% against 27.0% (OR 1.13, 95% CI 0.72 to 1.75), with no difference in fertilization, implantation, miscarriage or morphology [99], whereas a retrospective sibling-oocyte study of 604 oocytes cultured at 6.0% or 7.0% CO₂ (pH 7.37 against 7.30) found no difference in fertilization or blastulation but a higher euploidy rate at the higher pH (58.7% against 42.8%) [100]; the two address different endpoints in different designs and populations, a null result for live birth after day 3 transfer and an association with blastocyst euploidy, and neither establishes a universally optimal pH or a causal effect of pH on euploidy. An optimal external pH has therefore not been isolated experimentally from CO₂ and bicarbonate, and because components such as lactate and amino acids alter intracellular pH, two media at the same external pH can produce different intracellular values [101]. Quinn and Cooke showed that media formulated by the manufacturer with different bicarbonate concentrations to give the same pH under 5% or 6% CO₂ performed similarly in the mouse embryo assay and in human fertilization, development and clinical outcome [102]. The practical lesson is not that a laboratory should alter a commercial formulation, which falls outside the product's instructions for use, but that the manufacturer's stated CO₂ condition, the laboratory's set point and altitude, and the measured pH in the medium must be reconciled: choose the product version and gas set point that bring measured pH into the specified range, and re-measure whenever any of the three changes [88, 101]. Hentemann and colleagues found in a mouse model that development was improved when pH was about 7.30 before the pronuclear stage and lowered to 7.15 during cleavage, one argument for stage-specific pH in sequential systems [103].

Inside the embryo, pH is regulated by transport systems that, in the hamster, appear only at fertilization. The Na⁺/H⁺ antiporter that relieves acidosis is activated during egg activation in the hamster, under protein kinase C and calcium control [104], and the HCO₃⁻/Cl⁻ exchanger that relieves alkalosis is active in embryos but not oocytes, with a set point near pHi 7.24 [105]. Human cleavage-stage embryos regulate pHi at about 7.12 through the same two systems, the exchanger set between 7.2 and 7.3 and the antiporter below about 6.8, with a third Na⁺- and HCO₃⁻-dependent mechanism completing recovery from acidosis [106]. Human oocytes hold pHi near 7.4 through maturation and fertilization, aged and immature oocytes recover more slowly from alkaline shock, embryos up to the morula could not recover from acidosis in this study (in contrast to the Na⁺-dependent recovery reported by Phillips and colleagues, a disagreement the literature has not resolved) whereas blastocysts controlled both, and fertilization by conventional insemination was pH-sensitive whereas ICSI activated oocytes at every pH tested, which the authors attributed to the pH dependence of sperm-zona interaction; this speaks to activation, not to the pH sensitivity of subsequent development [107]. Lactic and pyruvic acids are weak acids, supplied in media as their sodium salts; in the mouse zygote lactate entered by facilitated transport with a smaller passive component, and at concentrations above 5 mM it lowered intracellular pH and reduced glycolysis, an effect the compacted embryo no longer showed [5]. Non-essential amino acids and glutamine buffer pHi in the mouse embryo up to the four-cell stage, one of several reasons they belong in cleavage media [108]. Forcing pHi up or down disrupts perinuclear mitochondrial organisation and microfilaments and reduces development [109]. The requirement for bicarbonate is not only about buffering. Bicarbonate/chloride exchange is active at every stage from zygote to blastocyst in the mouse, and inhibiting it disrupted intracellular pH and markedly reduced development at moderately raised external pH [110]; for the gametes, replacing the bicarbonate buffer with HEPES at identical pH abolished mouse fertilization and adding back 10 mM bicarbonate restored it, because the sperm need bicarbonate for the acrosome reaction rather than being inhibited by HEPES itself [111]. For the embryo, in bicarbonate-free HEPES or phosphate media without a CO₂ phase no mouse embryonic growth occurred, whereas HEPES medium containing bicarbonate supported development and live births not significantly different from T6 under CO₂ [112].

4.4 Temperature

Temperature is a property of the culture system rather than of the medium, but the medium is where the embryo experiences it, and the metaphase II spindle is the structure at risk. Cooling human oocytes to room temperature for 10 to 30 minutes reduced spindle size, disorganised or abolished microtubules and dispersed chromosomes, and fewer than half of spindles resembled controls after 1 to 4 h of rewarming [113]. Polarised light microscopy showed complete spindle disassembly within 5 min of cooling, recovery within 20 min if rewarming began promptly, and recovery in 5 of 5, 2 of 5 and 0 of 5 oocytes held for 10 min at 33, 28 and 25 °C respectively [114]. Overheating is equally damaging: spindles began to disassemble after 10 min at 39 °C or 1 min at 40 °C and did not fully reconstitute on return to 37 °C [115]. These data are why handling media are warmed and why heated stages, dish geometry and drop volume are qualified against a thermocouple in the drop rather than the instrument display [88]. Culture temperature itself has a dose response: in a time-lapse incubator with six chambers held from 35.0 to 37.5 °C in 0.5 °C steps, mouse blastocyst development was best at 37.0 °C and mitotic timings slowed at cooler settings and accelerated toward 37.5 °C [116]; the human randomised comparisons are small and have not established an exact optimum. Hong and colleagues split the mature oocytes of 52 couples between 36 and 37 °C from ICSI to transfer or vitrification and found more usable blastocysts per zygote at 37 °C (48.4% against 41.2%), with no difference in fertilization, aneuploidy or sustained implantation after paired euploid transfers [117]; the 2019 Cochrane review found three trials of 563 women comparing 37.0 or 37.1 °C with 36.0 to 36.6 °C, could not pool them, and was uncertain on very low-quality evidence whether the cooler setting changes clinical pregnancy (OR 1.08, 95% CI 0.73 to 1.60; one trial of 412 women) or miscarriage [118]; and a patient-randomised trial of 234 good-prognosis women cultured to the blastocyst stage at a stable 36.6 or 37.1 °C found clinical pregnancy of 38.3% against 38.6% by intention to treat (OR 0.98, 95% CI 0.56 to 1.73), with no embryological difference [119]. The distinction to keep is between acute cooling or overheating, which damages the spindle within minutes, and small stable differences in culture temperature, which the embryo appears to tolerate; the mouse optimum of 37.0 °C is not evidence of an exact human optimum.

4.5 Oxygen

Atmospheric oxygen (about 20%) is roughly two to thirteen times the tension measured in the tract (Section 3.3). Two randomised trials in human blastocyst programmes found benefit from 5% oxygen: Waldenström and colleagues reported a higher blastocyst rate (47.8% versus 42.1%), more cryopreserved blastocysts and a birth rate of 42% against 32% [120], and Meintjes and colleagues found live birth implantation of 42.9% versus 30.7% and live births of 57.4% versus 42.6% [121]. The largest early randomised study, 1380 consecutive treatments at 5% or 20% oxygen to day 2 or 3, found no difference in fertilization (60% against 61%), pregnancy (26.6% against 25.4%) or implantation, but surplus embryos formed blastocysts at a higher mean rate per cycle under 5% (25.8% against 20.4%) with more cells and fewer blastocysts under 25 cells, which is why the benefit is described for blastocyst programmes [122]; a later cohort of 871 patients followed for five to seven years including frozen transfers found that 5% oxygen after two days of culture gave more good-quality and cryopreserved embryos and a higher cumulative live birth rate per cycle, with similar birthweights [123]. The Cochrane review of seven trials with 2422 participants concluded in favour of low oxygen for culture [124]. Gardner has argued that reports of adverse perinatal outcomes after extended culture cluster in programmes using atmospheric oxygen, that atmospheric oxygen increases the embryo's sensitivity to other stressors, and that culture at 20% oxygen can no longer be justified [125]. Kelley and Gardner showed that individual culture and 20% oxygen each delay cleavage and reduce hatching and blastocyst cell number, and that the combination reduces cell number and the inner cell mass proportion further still [126]. Kaser and colleagues, randomising sibling bipronucleate and tripronucleate embryos donated for research rather than clinical embryos, found that switching to 2% oxygen from day 3 reduced cleavage arrest and increased blastulation, but with fewer cells per blastocyst and altered amino acid and redox metabolite profiles [127]; the clinical data on ultralow oxygen are limited and conflicting, and the oxygen schedules differ enough that they should not be pooled: a randomised oocyte-allocation study of 2298 oocytes from 152 conventional-IVF patients, published in a pharmacology journal, cultured at 2% or 5% from the day of retrieval and found no difference in day 3 quality, day 5 blastulation (62.8% against 61.9%) or clinical pregnancy after the first transfer (68.1% against 71.1%) [128]; a retrospective paired-cycle study of 120 couples in which the second cycle used 5% to day 3 and then 2% reported usable blastocyst rates of 32.8% against 21.8% and, in the 59 and 56 cycles that reached a morula or blastocyst transfer, cumulative live birth in 26 of 59 couples (44.1%) against 10 of 56 (17.9%), a design in which the biphasic cycle was always the second attempt [129]; and a prospective sibling-oocyte split of 658 complexes found that a gradient from 8% to 5% to 2% produced fewer clinically used blastocysts than static 5% (36.9% against 47.3%), with delayed blastulation and smaller blastocysts [130]. Ultralow oxygen therefore remains a research question, not a recommendation. Among the studies reviewed here, neonatal observations after it come only from that small paired-cycle series, which reported 30 live newborns after the biphasic schedule and 10 after 5% oxygen with no significant difference in gestational age, birthweight, length or sex [129], too little to establish safety, and no longer-term offspring follow-up has been reported. For 5 to 6% against 20%, the offspring data are reassuring but not uniform: a secondary analysis of a randomised trial of 1125 oocyte-donation cycles with day 3 transfer found no effect of oxygen on obstetric and neonatal outcomes [131], while a retrospective cohort of 13,831 fresh cleavage-stage singleton births found 5% oxygen associated with slightly longer gestation, 45 g lower birthweight, more small-for-gestational-age and fewer large-for-gestational-age infants [132]. The kinetics matter for handling: drops prepared in room air equilibrated to 5% oxygen with a half-life of 71 min under light oil and 116 min under heavy oil, and reoxygenated with a half-time of about 50 min under light oil once returned to atmospheric conditions, so a dish out of the incubator is a hidden oxygen exposure and a freshly prepared dish is not at its set point for hours [133]. Wale and Gardner's review of chemical and physical factors, and the Cairo consensus, treat reduced oxygen as an expected standard of care [88, 97].

5.1 Inorganic salts

Sodium and chloride set the bulk osmolality and the membrane potential. Simplex optimisation identified high sodium chloride as one of the most damaging variables for pronuclear mouse embryos, all embryos in a high-sodium-chloride medium blocked at the two-cell stage, and raising sodium chloride from 75 to 125 mM was progressively inhibitory in the absence of glutamine [41, 42]. Sodium is also the counter-ion for the Na⁺/H⁺ antiporter and the sodium-dependent amino acid and osmolyte transporters described above, and chloride for the HCO₃⁻/Cl⁻ exchanger [105, 106].

Potassium is the ion that most distinguishes tract-fluid media from serum-based ones. Human tubal fluid has a high potassium concentration [32], and the higher potassium of HTF relative to T6 was identified as the likely main contributor to its superiority in mouse development, with other differences not excluded [33]. In the simplex series, raising potassium from SOM to 2.5 mM in KSOM improved compaction and blastocyst yield [44]. The potassium story is not one-directional: in a simple mouse medium, potassium at the oviduct-fluid level inhibited cleavage and blastocyst formation, an inhibition partly relieved by removing phosphate and abolished by adding amino acids, vitamins, insulin, EGF and transferrin, which also brought glycolytic activity back toward in vivo values; in the same study four days in 4.79 mM lactate gave more fetuses after transfer than 23.3 mM [137].

Calcium is required for compaction, which depends on the calcium-dependent adhesion glycoprotein uvomorulin, later identified as E-cadherin. Compaction of the eight-cell mouse embryo involves calcium-dependent cell shape and membrane changes [138], Fab fragments against the 84 kDa surface glycoprotein prevent or reverse compaction [139], and E-cadherin null embryos compact only through residual maternal protein and then fail to form a trophectoderm epithelium or blastocoel [140]. Calcium is also required for sperm capacitation (Section 10.1). Human tract fluid has lower calcium than serum, and modifying the calcium concentration of HTF had no effect in Quinn's mouse zygote assay, which suggests some tolerance within the range tested rather than general insensitivity [32, 33].

In vivo two-cell hamster embryos hold intracellular magnesium near 0.37 mM and calcium near 130 nM; culture in 0.5 mM magnesium and 2.0 mM calcium tripled intracellular calcium, and raising medium magnesium to 2.0 mM lowered it again, increased blastocyst formation, cell number and inner cell mass, and doubled fetal potential after transfer [141]. Magnesium is the cofactor for the kinases of glycolysis and for ATP-dependent enzymes generally, which is the proposed basis of EDTA's effect in the mouse: EDTA lowered intracellular magnesium and thereby limited 3-phosphoglycerate kinase activity in the cleavage-stage embryo (Section 5.5) [142]. Lane and Gardner reviewed the embryo's regulation of intracellular calcium, magnesium and phosphate as a single homeostatic problem that culture conditions can either respect or disturb [143].

Phosphate is the most debated inorganic constituent. In the hamster, phosphate at 0.1 mM or above blocked development from the two-cell stage whether or not glucose was present, whereas glucose blocked only when phosphate was present, implicating premature stimulation of glycolysis and inefficient ATP production [22]; direct respirometry confirmed that glucose with phosphate cut the oxygen consumption of eight-cell hamster embryos two- to threefold, the Crabtree effect, and that Krebs-cycle substrates supported development only in their absence [144]. Removal of potassium phosphate produced the largest single improvement in the simplex optimisation of mouse media [41], and Quinn's phosphate- and glucose-free HTF improved mouse and human outcomes [34, 35], although the one large randomised trial of glucose omission in human culture, 741 patients cultured from the pronucleate stage to day 2 or 3 transfer, found better embryo quality without a difference in pregnancy rate and called reduced glucose prudent rather than proven [37]. Against this, Biggers and McGinnis found in a factorial study that phosphate up to 1.35 mM inhibited mouse blastocyst formation only slightly (at most 20%) and independently of glucose [145], and Quinn himself listed the physiological phosphate level and true requirement of the human embryo as unresolved [38]. Commercial media differ in phosphate content [134].

Bicarbonate is discussed in Section 4.3. It is a buffer, a substrate for the HCO₃⁻/Cl⁻ exchanger, the activator of soluble adenylyl cyclase in sperm (Section 10.1), and a requirement without which mouse embryos do not grow even when pH is held by another buffer [112, 146].

5.2 Energy substrates: pyruvate, lactate and glucose

Pyruvate is the preferred substrate of the oocyte and zygote and, under standard culture conditions, a required one. Biggers, Whittingham and Donahue showed that the mouse oocyte and zygote depend on pyruvate [19], and Brinster found an optimum near 0.3 mM for two-cell embryos [18]. Non-invasive microfluorimetry established that pyruvate uptake exceeds glucose uptake from the oocyte through the cleavage stages until the blastocyst, when glucose becomes the predominant substrate [147]. In human embryos pyruvate uptake rises from about 28 to 40 pmol per embryo per hour between days 2.5 and 4.5, and embryos that arrest take up significantly less pyruvate than those that blastulate [148]. Conaghan and colleagues found that pyruvate alone supported 59% blastocyst development from the two- to four-cell stage, that pyruvate-free medium arrested 84% of human embryos and that glucose could not compensate; uptake tracked the exogenous concentration with optimal development at the highest concentration tested, 0.47 mM [149]. Lane and Gardner later showed that the zygote's inability to use lactate is due to inactive malate-aspartate shuttle, limited by aspartate, and that supplying aspartate allows zygotes to develop to term on lactate without pyruvate, which reframes pyruvate dependence as a redox-shuttle limitation rather than an absolute substrate requirement [6].

Lactate is present at the highest concentration of any carbohydrate substrate in tract fluid and in the formulations of Table 1b, and it is not simply fuel. Increasing lactate reduced the proportion of pyruvate oxidised by the zygote, consistent with cytosolic NAD⁺ depletion, whereas in the blastocyst lactate increased pyruvate oxidation and was itself oxidised at three times the zygote rate; the lactate dehydrogenase isozyme was the same at both stages, so the difference lies in in situ redox regulation [150]. Supplied above 5 mM, lactate lowered the intracellular pH of the mouse zygote (Section 4.3) [5]. Only the L-isomer is metabolised; media that specify sodium L-lactate, or that report the D,L racemate, therefore differ in effective concentration, one reason lactate is among the most variable analytes across commercial products [134, 135]. The blastocyst itself produces lactate through aerobic glycolysis, which Gardner and Harvey propose serves implantation through endometrial tissue remodelling, angiogenesis and immune modulation [151].

Glucose has the most complicated history of the three. Before compaction, glucose in the presence of phosphate can drive premature glycolysis and developmental arrest [22], and removing it from the first 48 h of culture allowed random-bred mouse zygotes to pass the two-cell block, although glucose was then required from the four-cell stage onward for blastocyst formation [23]. Human embryos cultured without added glucose reached the eight-cell stage slightly more often and produced blastocysts with more cells than those in 1 mM glucose [149]. Yet glucose is essential later: mouse embryos deprived of glucose compensated with pyruvate but formed fewer, smaller blastocysts, and one-cell embryos deprived throughout degenerated after the morula [152]. The later requirement is not, or not only, an energy requirement. At the 8- to 16-cell transition in the mouse, glucose does not fuel mitochondrial ATP production and glycolysis is dispensable for blastocyst formation; glucose acts through the hexosamine pathway and the pentose phosphate pathway, with sphingolipid and mTOR signalling, to specify the trophectoderm, and has no role in inner cell mass specification [153]. Isotope tracing shows why the early embryo is so particular: cleavage-stage metabolism is rigid in its nutrient requirements, sensitive to reductive stress and runs a disequilibrated TCA cycle, and only with loss of maternal lactate dehydrogenase B and zygotic transcription do the shuttles and fatty acid oxidation that relax these constraints come into play, to the point that blastocysts can develop without external nutrients [154]. In the presence of non-essential amino acids and glutamine, removing glucose from mouse culture reduced blastocyst cell number and implantation, which is why G1 retains 0.5 mM rather than zero [155]. In the simplex media glucose at 5 mM had no inhibitory effect [42, 145]. The resolution is that the inhibitory effect of glucose depends on the presence of phosphate and on the absence of amino acids and EDTA, and that a low but non-zero concentration before compaction with a higher concentration after it reflects both the tract fluid gradient and the embryo's changing capacity to use it [55, 60]. After compaction glucose consumption was associated with viability in one study of 50 single blastocyst transfers: uptake on days 4 and 5 was significantly higher in embryos that produced a live birth, independent of morphology, and female embryos consumed 28% more on day 4 than males [156], consistent with the earlier mouse observation that glucose uptake before transfer predicted development [157]. These are research associations rather than validated selection tests. Two older observations frame them: in the mouse, blastocysts whose glycolytic activity, the fraction of glucose converted to lactate, was near in vivo values had higher viability after transfer than those with elevated glycolysis, and selecting on that criterion gave a fourfold increase in pregnancy rate, so that it is the flux, not the uptake, that marks the non-viable blastocyst [158]; and in human embryos, mean pyruvate uptake was lower on days 2 and 3 in embryos that implanted (22.9 against 27.1 pmol per embryo per hour on day 2), although the ranges overlapped and uptake alone could not predict implantation [159].

Two syntheses frame the field. Leese's "quiet embryo" hypothesis proposes that viable embryos have a lower overall metabolism than those that arrest and that damage raises nutrient turnover, from which Leese drew the suggestion that media should not supply more than the embryo needs [160]; that suggestion has since been tested in the mouse, where carbohydrates, amino acids and vitamins could be cut by half without detriment, a cut to a quarter impaired blastocyst development unless pyruvate and lactate were restored to half, blastocysts in the half-strength pyruvate and lactate version contained more inner cell mass cells and ATP than controls with altered metabolic activity, and a two-step reduced-nutrient design implanted and produced fetuses at the control rate [161]; amino acid turnover in human, bovine and porcine embryos correlated with DNA damage in support of this [162]. The higher glucose consumption of viable human blastocysts described above is not a contradiction so much as a reminder that different metabolic measures, at different stages and through different pathways, can point in different directions, and that uptake is not the same as ATP production; neither hypothesis yields a formulation rule. Leese's forty-year retrospective and Gardner and Harvey's account of blastocyst metabolism agree that pyruvate, lactate, glucose and amino acids have overlapping roles as energy sources, biosynthetic precursors, redox regulators and signals to the epigenome [4, 151].

5.3 Amino acids

The amino acid formulations in IVF media derive from Eagle's classification of the essential and non-essential amino acids for mammalian cells in culture [163]. The evidence that they belong in embryo media is set out in Section 2.5: non-essential amino acids and glutamine stimulate cleavage and blastocoel formation and buffer intracellular pH before compaction, essential amino acids stimulate cleavage and inner cell mass development after the eight-cell stage, and the combination in the right order maximises viability after transfer [50, 52, 108]. In protein-free KSOM, amino acids increased hatching and cell proliferation in both lineages and improved basement membrane formation [48]. In human embryos randomised between glucose-free Earle's medium with glutamine alone, the same medium with non-essential amino acids for cleavage and all 20 for blastocyst formation, and a commercial sequential medium, blastocyst development was similar but the amino acid mixtures increased total cell number from 61.8 to about 100 and reduced the dead cell index [164]. Even brief exposure matters: culturing in vivo-derived mouse blastocysts for only 3 to 6 h in a simple salt and carbohydrate medium doubled glycolysis, reduced pyruvate oxidation and reduced viability after transfer, and amino acids and vitamins prevented both the metabolic perturbation and the loss of viability [165]. Van Winkle's review of transport regulation covers the mechanisms by which these benefits are delivered [166]. Contemporary media carry the essential amino acids at a fraction of Eagle's concentrations because reducing them increased mouse blastocyst development and cell number and lowered ammonium accumulation, and the authors recommended reduced essential amino acid concentrations on that basis [167].

Amino acids also act as organic osmolytes (Section 4.2), as intracellular pH buffers, and as a candidate viability signature: in a research dataset of spare day 2 to 3 human embryos, the turnover of alanine, arginine, glutamine, methionine and asparagine predicted blastocyst formation with high accuracy in that dataset (reported as more than 95%), with leucine most consistently depleted and alanine most consistently produced (the paper's erratum, which corrects most of the plotted amino acid values, 12.5 times too high and arginine 1.25 times, and revises some P values, states that the conclusions are unaffected) [168], and in a retrospective analysis of 53 ICSI cycles the turnover of asparagine, glycine and leucine correlated with clinical pregnancy and live birth independent of age, FSH and morphology [169]. These are research associations; prospective validation across contemporary media and laboratories has not established amino acid profiling as a routine selection method.

The cost of amino acids is ammonium, which is the subject of Section 5.4 and Section 14.1. The amino acid profile is also among the most variable features of commercial media between products [134, 136].

5.4 Glutamine and its dipeptides

Glutamine is a preferred energy substrate of the early embryo and an osmolyte [23, 42], but free glutamine is chemically unstable. Tritsch and Moore showed in 1962 that glutamine decomposes spontaneously in culture media, a cyclisation in which the side-chain amide nitrogen is lost as ammonia and the residue becomes pyrrolidone carboxylic acid (pyroglutamate), and Ozturk and Palsson characterised the decomposition at 37 °C as first order, with a rate constant that increases with pH and is unaffected by serum [170, 171]. Ammonium also arises from the embryo's own metabolism of amino acids, a process distinct from the spontaneous breakdown of glutamine [172]. The response of the field has been to supply glutamine as a dipeptide. In hybridoma culture, alanyl-glutamine and glycyl-glutamine are hydrolysed extracellularly by a peptidase released by the cells and their use markedly lowers ammonia and lactate accumulation [173]. In a KSOM-type mouse medium, alanyl-glutamine had no significant effect on blastocyst formation, hatching or cell numbers compared with glutamine, whereas glycyl-glutamine significantly increased inner cell mass and trophectoderm cell numbers, and transfer experiments showed no gross fetal abnormality, leading Biggers and colleagues to suggest that glycyl-glutamine may be the better choice even though alanyl-glutamine had already been adopted in human media [174]. Many contemporary products use alanyl-glutamine, and the composition sheet states which; in the trial by Hardarson and colleagues, the ammonium concentration at the end of culture was significantly lower in a single-step time-lapse medium than in the corresponding sequential blastocyst medium [175]. Whether a product uses free glutamine, a dipeptide, or none is a question for the manufacturer's composition sheet, and Tarahomi and colleagues found that storage and sham culture changed the concentrations of multiple components across the fifteen media they analysed [136].

5.5 Chelators: EDTA

EDTA was first shown to allow one-cell mouse embryos to develop in a chemically defined medium by Abramczuk, Solter and Koprowski [21], and embryos cultured in 100 µM EDTA with amino acids yielded fetuses at a rate not significantly different from in vivo controls whereas cord serum was markedly inferior [176]. A mechanism was proposed in the mouse by Lane and Gardner: embryos cultured without EDTA show an abnormal rise in glycolysis; EDTA inhibits 3-phosphoglycerate kinase by chelating magnesium, an effect mimicked by the kinase inhibitor Cibacron blue and reversed by exogenous magnesium, and embryos cultured with EDTA have lower intracellular magnesium [142]. An earlier proposal came from Nasr-Esfahani, Winston and Johnson in 1992: over the first 48 h EDTA was the major beneficial factor for both blocking and non-blocking mouse strains, and they concluded that chelation of transition metals, which catalyse the generation of reactive oxygen species, was the most effective way to overcome the block [177]. The same property makes EDTA harmful after compaction, when the embryo's glucose metabolism rises (a rise that, in the mouse, serves trophectoderm specification through signalling pathways as much as energy supply, Section 5.2 [153]): in bovine embryos 100 µM EDTA increased cleavage during the first 72 h but reduced morula, blastocyst and inner cell mass development when continued, and the best outcome came from EDTA for 72 h followed by EDTA-free medium [178]. In CF1 mouse embryos EDTA acted in synergy with non-essential amino acids and glutamine during the first 48 h and was inhibitory during the second 48 h, with loss of viability [155]. This is one of the strongest mechanistic arguments for a sequential design or, in single-step media, for a low EDTA concentration throughout; Biggers and Summers regard the EDTA argument, like the others for two-step protocols, as equivocal in clinical practice [60].

5.6 Vitamins

Vitamins are absent from many cleavage-stage formulations and present in some blastocyst-stage media, G2 among them. Kane and Bavister showed that development of eight-cell hamster embryos to hatched blastocysts in protein-free medium was stimulated by the water-soluble vitamins of Ham's F-10, and that omission of inositol, pantothenate or choline reduced hatching, inositol most severely [179]; in the same system polyvinyl alcohol could serve as the only macromolecule when vitamins and amino acids were present, and certain bovine serum albumins reduced but did not eliminate the vitamin requirement, which is why vitamin needs differ between protein-supplemented and protein-free media [180]. In sheep, amino acids and vitamins together increased glucose uptake and lactate production by blastocysts [69], and in mouse blastocysts vitamins acted in synergy with amino acids to prevent culture-induced metabolic perturbation and loss of viability [165]. G2 was formulated with vitamins for this reason [54]. Riboflavin, where present, is also the principal photosensitiser responsible for peroxide generation in light-exposed media (Section 14.6) [181, 182].

5.7 Antibiotics

Antibiotics are included as insurance against the bacterial load of semen, follicular aspirates and handling. The evidence that they are benign is thin. In one study, Magli and colleagues cultured 196 human zygotes in conventional, half-strength and antibiotic-free versions of their medium and found significantly higher cleavage rates at every observation, including the blastocyst stage, in the antibiotic-free medium, with no improvement at half strength [183]. In hamster embryos, penicillin, streptomycin and gentamicin individually had no effect, but penicillin and streptomycin together reduced eight-cell and blastocyst development, and no antibiotic improved development; the authors concluded that antibiotics are unnecessary for up to 72 h with proper sterile technique under oil [184]. Kastrop and colleagues examined 95 contaminated culture dishes from one laboratory over several years: in that series infections occurred only after conventional IVF and never after ICSI, most were Escherichia coli or Candida, and 73% of the E. coli isolates were resistant to both penicillin and streptomycin in the medium [185]. Gentamicin has replaced penicillin and streptomycin in many commercial media; the choice and concentration should be read from the composition sheet, and whether to move to antibiotic-free culture is a laboratory policy decision to be weighed against the contamination data above and the laboratory's own sterile technique [184, 185].

5.8 Hyaluronan, polyvinyl alcohol and other macromolecules

Protein is treated in Section 6. Two non-protein macromolecules deserve mention here. Hyaluronan, the glycosaminoglycan abundant in tract fluid, was shown by Gardner, Rodriguez-Martinez and Lane to give the highest implantation and fetal development after mouse blastocyst transfer when it replaced albumin, with the benefit traced to its presence in the transfer medium; cell number and hatching were highest when both albumin and hyaluronan were present, so the authors proposed both in culture media and hyaluronan alone in transfer medium [186]. Later commercial versions of G2 list hyaluronan alongside albumin, whereas the published G2 of Table 1b did not contain it. The commercial transfer medium EmbryoGlue is not the albumin-free transfer medium of the mouse experiment: its package insert lists recombinant human albumin, hyaluronan and gentamicin in a bicarbonate-buffered medium [187]. The clinical evidence for hyaluronan in transfer media is reviewed in Section 11. CD44, the principal hyaluronan receptor, is expressed on human oocytes, cumulus, embryos and pre-hatching blastocysts [188]. Polyvinyl alcohol has been used as a synthetic surfactant and carrier in research media, and KSOM supplemented with amino acids supports mouse development to the late blastocyst without any protein at all, demonstrating that macromolecules are not obligatory for development, whatever their practical value for handling and viability [48, 186]; in the primary study, PVA alone in place of albumin gave blastocysts at a moderately reduced rate but severely impaired hatching, amino acids alone gave a high blastocyst rate, and adding PVA to KSOM with amino acids had no further effect [189].

5.9 Antioxidants

Reactive oxygen species arise from embryo metabolism, from atmospheric oxygen, from light, from peroxidised oil and from transition metals in protein supplements. Guérin, El Mouatassim and Ménézo reviewed the external protection provided in follicular and tubal fluid by hypotaurine, taurine and ascorbate and the internal enzymatic defences, and cautioned that supplementing media with antioxidants is a complex matter of maintaining the pro-oxidant and antioxidant balance [190]. Hypotaurine and taurine were long included in some formulations on this basis and for their osmolyte role, although adding taurine to a medium already containing non-essential amino acids had no effect on CF1 mouse development [155]. The modern evidence concerns a defined combination of acetyl-L-carnitine (10 µM), N-acetyl-L-cysteine (10 µM) and alpha-lipoic acid (5 µM). In mouse embryos the combination increased blastocyst cell number, maintained intracellular glutathione, accelerated development from the five-cell stage and improved fetal development after transfer, with the greatest benefit in 20% oxygen and individual culture [191]; in mouse IVF it accelerated development from the two-cell stage, increased blastocyst cell number and reduced intracellular hydrogen peroxide [192]. A prospective randomised sibling-oocyte comparison across 1563 metaphase II oocytes from 133 patients found more good-quality day 3 embryos (50.2% versus 40.7%) and, in an age subgroup of women aged 35 to 40, an implantation rate by fetal heart of 50.0% against 23.5% and ongoing pregnancy of 50% against 25.8%; sibling designs share patient-level factors and the subgroup was not the primary endpoint [193]. The largest test to date was a single-centre trial that randomised 1482 patients before oocyte collection, with patients and doctors blinded: clinical pregnancy per randomised patient from fresh single blastocyst transfer was not increased by the antioxidant media (22.9% versus 26.1%; RR 0.88, 95% CI 0.73 to 1.05), nor was per-protocol live birth (29.5% versus 32.4%), whereas the fertilization rate rose in ICSI cycles (68.3% versus 57.9%), failed-fertilization cycles fell from 8.0% to 3.7% and more blastocysts were utilised per patient (3.09 versus 2.70); cumulative outcomes were not reported, and the study disclosed manufacturer support [194]. A smaller randomised study of 127 cycles again found the laboratory and transfer benefits concentrated in women aged 35 to 40 [195]. The balance of evidence is therefore that this antioxidant combination improves some laboratory endpoints, that a clinical pregnancy benefit was not shown in the one adequately powered patient-randomised trial, and that the age-subgroup signal awaits a trial designed to test it. More embryos per patient is not the same as more live births. The mouse work that preceded the human trials had already shown the benefit to be age- and composition-dependent: a combination of alpha-lipoic acid, alpha-tocopherol, hypotaurine, N-acetylcysteine and a sirtuin activator raised blastocyst development from embryos of aged females from 72.7% to 87.6% with higher mitochondrial membrane potential and ATP, whereas in young females N-acetylcysteine with lipoic acid alone improved development and the full mixture did not [196].

5.10 Growth factors, insulin and cytokines

The preimplantation embryo expresses growth factor ligands and receptors, and the tract expresses their ligands, so autocrine, paracrine and endocrine signalling before implantation is well established [197]. Paria and Dey showed that mouse embryos cultured singly in 25 µL developed less well than those in groups of 5 or 10, that epidermal growth factor and transforming growth factors alpha and beta 1 rescued single embryos, and that doubling the volume to 50 µL made single culture worse [198]. O'Neill found that autocrine mediators, including platelet-activating factor, must act by the two-cell stage as survival factors rather than mitogens [199], that zygote development fell from 80% to 26% as embryo concentration was diluted from 1 to 0.001 per µL whereas two-cell embryos developed well at all concentrations, so the autocrine requirement lies before the two-cell stage [200], and an embryo-derived PAF-like factor was shown to act through the PAF receptor, since high density, conditioned medium or PAF advanced compaction and hatching and two PAF-receptor antagonists blocked the density effect [201]. Against this, a multi-strain mouse study found that PAF, EGF, IGF-1 and GM-CSF added in combination to individually cultured embryos improved neither blastocyst formation, hatching nor lineage cell numbers, reduced day 5 blastocyst formation in one strain and oocyte-source group, and did not change development after transfer [202], and an optimised combination of VEGF, PDGF, IGF-1, IGF-2 and G-CSF left morphokinetics unchanged but increased trophectoderm and total cell number, outgrowth, fetal weight, crown-rump length and morphological development, while the fetal liver and placental transcriptomes diverged further from in vivo controls than those of untreated cultured embryos, which the authors read as a reason for caution [203].

Insulin and the insulin-like growth factors. Insulin is not a declared constituent of most embryo media, but it is the best-studied of the growth-promoting hormones and the one most often proposed as a supplement. In the mouse, Harvey and Kaye showed that insulin at picomolar concentrations increased blastocyst cell number by 9%, entirely through a 23% increase in inner cell mass cells, with an EC₅₀ of 0.54 pM, and advanced blastocyst formation, consistent with insulin receptors expressed from about the eight-cell stage and exogenous insulin reaching the inner cell mass across the trophectoderm [204]; a receptor-blocking antibody then showed that protein synthesis was stimulated by both insulin and IGF-1 through the insulin receptor, whereas insulin's mitogenic effect on the inner cell mass went through its own receptor and IGF-1's through another [205]. In a simple medium, Gardner and Sakkas found that, once phosphate was removed, insulin, EGF and transferrin together with amino acids and vitamins abolished the inhibition produced by oviduct-level potassium and normalised glycolysis [137]. The human data come from Winston's group: IGF-I is expressed by the mid-cycle Fallopian tube and present in human tubal and uterine fluid at 8.0 and 10.9 nM, the embryo expresses the IGF-I receptor but not the ligand before implantation, and supplementation of donated embryos raised blastocyst formation from 35% to 60% and inner cell mass cell number by 59%, an effect blocked by an antibody to the receptor [206]; a later study found that 1.7 nM IGF-I raised blastocyst formation from 49% to 74% and halved the proportion of apoptotic nuclei (16.3% to 8.7%) without significantly changing total cell number, so that IGF-I acts in the human embryo as a survival factor [207]. Concentration matters in both directions: high IGF-1 (130 nM) or high insulin, used as a model of the elevated levels seen in polycystic ovary syndrome, triggered apoptosis of the mouse inner cell mass, accompanied for IGF-1 by down-regulation of the IGF-1 receptor and reduced insulin-stimulated glucose uptake [208], and a shortfall of insulin has its own consequence, since halving insulin together with the branched-chain amino acids in a defined mouse medium left blastocyst cell numbers unchanged but increased birthweight and early postnatal weight gain, with relative hypertension in male offspring [209]. Among the studies reviewed here, clinical use has been tested in one sibling-oocyte study: a split of 5142 oocytes from 360 patients in two centres, cultured from day 0 to day 5 or 6 in insulin-supplemented or control single-step medium, reported higher day 3 quality and compaction, more blastocysts of better quality, more cryopreserved, and higher clinical, ongoing and twin pregnancy rates in the insulin arm, with the authors themselves calling for a multicentre randomised trial in other populations and media before any conclusion [210]. Insulin, transferrin and selenium as a combined supplement, standard in somatic cell culture and in some oocyte maturation media, has been examined in human embryos only as a rescue of arrested embryos, where it reduced the arrest rate and raised cyclin A2 expression in a research setting that does not bear on clinical culture [211].

LIF, HB-EGF and gp130. Leukaemia inhibitory factor gave contradictory results in human embryos: 5 to 20 ng/mL recombinant LIF did not increase blastocyst formation of donated embryos in a serum-containing medium [212], whereas 1000 IU/mL raised blastocyst formation from 18.4% to 43.6% in a complex serum-free medium [213], and a later study found that the LIF co-receptor gp130, but not LIF itself, improved blastocyst formation of thawed embryos from 43% to 73% [214]. Heparin-binding EGF in the same serum-free system raised blastocyst development from 40.7% to 71.0% and hatching from 45.5% to 81.8% at 100 nM without changing cell number or substrate consumption [215], and the group's own summary warned that the normality of such blastocysts must be established before clinical use, since apparent improvements in culture have been detrimental to pregnancy outcome in other species [216].

The clinical trials. Two cytokine formulations have reached randomised trials in human media. Sjöblom, Wikland and Robertson reported that 2 ng/mL recombinant GM-CSF increased human blastocyst development from 30% to 76% with 35% more cells, mainly in the inner cell mass [217]. In a 14-centre placebo-controlled trial of 1332 randomised women (1149 receiving transfer), the prespecified primary endpoint, ongoing implantation rate at week 7, was 23.5% with GM-CSF at 2 ng/mL against 20.0% with control, a difference whose confidence interval included no effect (OR 1.26, 95% CI 0.91 to 1.75); the secondary outcomes at week 12 (23.0% versus 18.7%) and live birth (28.9% versus 24.1%) reached significance, the effect depended on the albumin concentration in the medium, and the previous-miscarriage finding came from an exploratory subgroup analysis [218]. A trial of 443 ICSI cycles randomised to a medium containing GM-CSF, HB-EGF and LIF together or to conventional medium reported ongoing pregnancy per randomised participant of 47% against 36% (absolute difference 12 percentage points, 95% CI 2.5 to 21), live birth of 101 of 224 against 71 of 219 (RR 1.39, 95% CI 1.09 to 1.77), cumulative live birth within a year of 132 of 224 against 97 of 219 (RR 1.32, 95% CI 1.09 to 1.61) and pregnancy loss of 27 of 124 against 37 of 103, with the authors noting that the long-term effect of cytokine enrichment is unknown; the percentages printed for cumulative live birth in the paper's abstract and Table 2, 60% and 44% with a 12-point difference, do not match those counts, which compute to 58.9% and 44.3%, so the counts and relative risks are the values used here and the discrepancy is left unreconciled [219]; it tests the combination, not the three factors separately, and is not among the trials pooled in the Cochrane review. The Cochrane review of five GM-CSF trials concluded that it is uncertain whether GM-CSF supplementation changes live birth (OR 1.19, 95% CI 0.93 to 1.52; low-quality evidence) [220], and ESHRE's 2023 good practice recommendations on add-ons concluded that evidence for both efficacy and safety is insufficient and that growth-factor-supplemented embryo culture medium is not recommended [2]. Species specificity is a further caution for mouse testing of cytokine media: mouse GM-CSF increased mouse blastocyst cell number through trophectoderm and primitive endoderm, not epiblast, whereas human GM-CSF did not increase mouse blastocyst cell number in either a human or a mouse medium, although both forms induced ectopic NANOG expression in the trophectoderm at increasing concentrations [221]. The interaction with albumin concentration is a separate warning that supplements cannot be evaluated in isolation from the rest of the formulation.

5.11 Indicators, and what is deliberately absent

Phenol red was included in HTF as a visual pH indicator [33] and has largely been removed from contemporary embryo media, in part because phenol red preparations used in tissue culture were shown to have weak oestrogenic activity [222], and in part because a pH indicator is no substitute for measurement; laboratories should measure pH under their own gas conditions rather than judge it by colour [8, 88]. Serum is absent for the reasons given in Section 6.5. In conventional embryo culture formulations, hormones and lipids are not declared as additions, although albumin carries both (Section 6) and Leese notes that lipids are beginning to receive the attention they deserve as nutrients [4]; the insulin-, growth-factor- and cytokine-supplemented formulations of Section 5.10 are the exceptions, and a statement about what a medium omits is only ever a statement about the products, lots and dates examined. Three claims should be kept apart: that a manufacturer declares no addition, that an analyte was not detected, and that the substance is absent. Copper and zinc, for example, were not detected in the media and lots analysed by Morbeck and colleagues, a finding bounded by the analytes, methods and detection limits used and not a statement of formulation intent for all products, whereas transition metals were found at high levels in several protein supplements (Section 6.2) [134, 223].

6.1 Why protein is added

Protein is added to embryo media mainly for its physicochemical roles; any nutritional contribution is secondary and not well quantified. Mouse embryos reach the late blastocyst in protein-free KSOM with amino acids, which shows that exogenous protein is not obligatory under those conditions, not that it contributes nothing [48], and the human demonstration is older still: Caro and Trounson randomised patients between T6 with 10% maternal serum and T6 with neither protein nor amino acids and found fertilization of 68% against 69%, normal embryos 96% against 97%, pregnancy 18% against 14% and birth 13% against 11%, none significantly different [229]. Albumin is a colloid that reduces the surface tension of the medium and keeps embryos from sticking to glass and plastic; it is a carrier of fatty acids, lipids and hormones; it binds heavy metals, peroxides and other lipophilic compounds; and it is the sterol acceptor that supports sperm capacitation in the systems in which this has been studied [230–232]. In the mouse, albumin increased blastocyst cell number in culture without improving outcome after transfer, whereas hyaluronan improved implantation and fetal development through its presence at transfer [186]. Blake and colleagues traced the history from whole crude plasma to genetically engineered human albumin [233]. The choice today is between plasma-derived human serum albumin, globulin-enriched serum substitutes, and recombinant human albumin, and each carries a different burden of undefined components.

6.2 Plasma-derived human serum albumin: what else is in the bottle

Human serum albumin for IVF is a pharmaceutical fraction, supplied at about 5 mg/mL in ready-to-use media or as a concentrate for the laboratory to add at 5 to 10 mg/mL [226, 234]. Because it is purified from pooled plasma, it carries a variable cargo. Proteomic analysis of eight commercial media, each supplemented by the manufacturer with 5 mg/mL purified HSA, identified 110 non-declared proteins, with HSA averaging 94% of total protein and individual non-declared proteins accounting for up to 4.7%; the pattern strongly suggested that the purified HSA was the source [226]. A 2025 proteomic study of 85 samples from 13 production lots of two manufacturers, using different methods, products and sampling and therefore not a controlled comparison of detection counts, found more than 700 undeclared human proteins in unconditioned media; the protein profile did not differ between media in which embryos reached the blastocyst and those in which they arrested, but correlated strongly with production lot, and spiking a protein-free base medium showed that serum-derived HSA is the source of the undeclared, batch-variable proteins, with recombinant HSA spiked as the comparator [235]. The lipid cargo differs between brands as well: fifteen commercial albumins contained from 0.07 to 16.77 µg of fatty acid per mg, media below 0.5 µg/mg supported better mouse blastocyst development than those above 1.4 µg/mg, and carnitine mitigated the difference [236]; in a direct comparison of two commercial HSA products and a recombinant albumin at 10% in G1/G2, the product with the highest fatty acid content, an eightfold excess of saturated fatty acids, gave reduced blastocyst development, more superoxide and neutral lipid, and lower implantation in the mouse [237]. Plasticiser travels with the protein: di(2-ethylhexyl)phthalate and its monoester were measured at up to 114 and 263 ng/mL in culture and sperm-washing media and up to 982 and 1840 ng/mL in protein sources, which were identified as the origin [238]. And the supplement is itself a source of ammonium: all fifteen protein-supplemented ready-to-use media accumulated ammonium over six weeks at 2 to 8 °C, and both media and the eight protein supplements accumulated it over four days at 37 °C (Section 14.1) [239]. Morbeck and colleagues measured 39 amino acids, organic acids, ions and metals in six supplements (recombinant albumin, two HSA preparations and three complex supplements). Recombinant albumin had the fewest undefined components and the lowest metal content and supported high blastocyst development at both 5% and 20% oxygen, whereas one HSA product (Buminate) and two complex supplements (LGPS and SPS) contained high levels of pro-oxidant transition metals, and blastocyst formation was reduced with Buminate, SPS and SSS, the last being high in amino acids rather than metals; pre-compaction development was delayed relative to recombinant albumin for every plasma-derived supplement, with an interaction between supplement and oxygen [223]. Endotoxin was present in all five commercial albumins tested by Dumoulin and colleagues [86].

Pharmaceutical albumin is stabilised for heat treatment with sodium octanoate (caprylate) and, in some products, N-acetyltryptophan, and an analytical survey of 138 and 159 pharmaceutical lots found the great majority within 20% of the labelled stabiliser content [240]. The embryotoxicity of the stabiliser was first shown by Leonard and colleagues: at 5% HSA all three manufacturers' products supported 70% or more mouse blastocysts, at 15% two of the three fell below 50%, desalting rescued one of them, mass spectrometry showed high octanoic acid in that product, and adding octanoic acid to the good product reproduced the toxicity, with the stabiliser content varying by lot [241]. Fredrickson, Krisher and Morbeck cultured mouse zygotes individually in recombinant albumin with 0, 400, 800 or 1200 µM octanoate and found that blastocyst development, hatching, developmental kinetics and cell number were reduced at concentrations in the range used in human IVF, and that fetal and placental weights were increased at 800 µM, leading them to recommend that octanoate exposure be monitored and reduced [227]. Albumin also carries lipids, and in Otsuki's experiments its presence in the medium increased the passage of a lipophilic tracer from peroxidised oil into the zona pellucida (Section 12), so the protein supplement is a proposed route by which oil quality reaches the embryo [232].

Lot-to-lot variation is an old problem with a modern face. Different lots of commercial bovine serum albumin either inhibited or stimulated hamster embryo development [242], in goat embryos both oxygen and the protein source mattered, since reduced oxygen raised expanded and hatched blastocysts from 29% to 80% with one BSA while early-cleavage cell numbers were less than half as high with another BSA or with goat or fetal calf serum [243], maternal sera that had supported human zygotes to term pregnancy, and fetal cord sera, did not permit mouse blastocyst formation even after dialysis, and only some HSA batches supported mouse blastocysts whereas all BSA sources did, a species discordance that limits the mouse assay as a screen for human protein lots [244], and in rabbit blastocysts one lot caused complete hatching and more than twice the cell number of another lot that caused none [245]. The clinical equivalents are the birthweight associations discussed in Section 6.6 and Section 15. This page's recommendation is that a protein lot change be treated as a culture-system change and qualified as such (Section 16).

6.3 Globulin-enriched serum substitutes (SSS, SPS and related products)

A retrospective comparison of 1019 cycles found clinical pregnancy of 34% with a plasma protein fraction containing albumin and mixed globulins (Plasmanate) against 24% with maternal serum, but the trend reversed in the prospective randomised trial the same group ran, a caution that applies to every retrospective advantage reported for a complex supplement [246]. Pool and Martin reported a continuing pregnancy rate of 38.7% per transfer across 103 retrievals using a plasma protein fraction containing albumin with significant alpha- and beta-globulins in insemination, growth and transfer media, and suggested that glycoprotein components of serum, or the general physicochemical properties of the fraction, support embryo growth [247]. Weathersbee, Pool and Ord then formulated Synthetic Serum Substitute, a plasma-derived product despite the word synthetic, from HSA (84% of protein) and human globulins (16%); an early lot containing a plasma lipoprotein fraction accelerated early growth but reduced hatching and formed a precipitate, so the lipoprotein was removed, and the final lipoprotein-free SSS accelerated early human and mouse embryo growth relative to the plasma protein fraction [234]. The one randomised clinical trial came from Meintjes and colleagues, who randomised 528 patients to HSA alone or HSA plus SSS from the pronuclear stage: implantation was 50.6% versus 43.8% and live birth 62.8% versus 54.2% in favour of the enriched supplement, with the caveat that additional human-derived protein demands rigorous quality control that not every laboratory can provide [228]. Complex supplements are also the ones in which Morbeck and colleagues found the highest amino acid concentrations (SSS) and high transition metal levels (SPS, LGPS), so the same fraction that carries the beneficial globulins carries the pro-oxidant load [223]. The GM-CSF trial found that the cytokine's effect depended on the albumin concentration in the medium, another reminder that protein is an active variable [218].

6.4 Recombinant human albumin

Recombinant human albumin removes the plasma pool and its cargo. In a prospective randomised study of 85 women, media with recombinant albumin gave fertilization, cleavage, blastocyst formation, implantation, pregnancy and early loss rates comparable to HSA, and the authors proposed it as a replacement that may reduce the risks of prion contamination and plasma-derived impurities; a study of that size observes comparable outcomes but cannot establish equivalence or rare-event safety [248]. In the compositional analysis it had the fewest undefined components and the lowest metal content [223], and Dyrlund and colleagues traced the non-declared plasma proteins in commercial media to the purified HSA supplement rather than to the base medium [226]. In bovine culture, recombinant albumin supported blastocyst development at rates comparable to BSA, adding citrate to it increased expansion and post-cryopreservation re-expansion and hatching, and hyaluronan increased cryosurvival with either protein, which shows what a defined supplement may need to add back [249]. Recombinant production removes the reliance on pooled human plasma; host-cell impurities, stabilisers and manufacturing controls still require assessment, and the product lacks the globulins of the complex supplements. A recombinant albumin is defined by what it lacks, not by what it adds.

6.5 Serum: why it left

Serum, including the patient's own, supplemented most early human media. Ménézo and Testart showed in 1984 that B3 medium with 1% pure human serum albumin performed as well as B2 with cord serum at every stage, that serum offered no positive effect, and that removing it improved the homogeneity of results and permitted metabolic analysis of spent media [30]. In sheep, culture in synthetic oviduct fluid with 20% human serum produced embryos with abundant lipid droplets and lambs that were significantly heavier (4.2 kg against 3.4 to 3.5 kg) with longer gestation than embryos cultured in a serum-free, albumin and amino acid medium [73], one of the four situations that produce the large offspring syndrome reviewed by Young, Sinclair and Wilmut [74]; a later 2 × 2 study in sheep located the sensitive window before compaction, since serum during the first two days of culture retarded early development yet increased placental, fetal, heart and liver weights and the proportion of fetuses more than three standard deviations above control, whereas serum in the last two days only increased blastocyst yield [250]. Performance and offspring were not the only reasons serum left: one batch of pooled, heat-inactivated human serum used as a medium supplement transmitted hepatitis B to 79 women in a single programme, 24 of whom were pregnant during the acute infection [251]. Gardner's serum-free sheep system, in which amino acids, vitamins and group culture replaced serum and somatic cells and yielded blastocysts with cell numbers equal to in vivo controls, and his general case against serum and co-culture, settled the case for defined media in embryo culture [69, 252]. Serum has left embryo culture. Some sperm cryopreservation media contain egg yolk or yolk-derived components (Section 10.4); that is a different material serving a different purpose and should not be confused with serum.

6.6 Protein and the offspring

The protein supplement is one of the few medium variables that has been isolated in a human birthweight analysis. Zhu and colleagues compared 1097 singletons born after culture in G1 v5 (HSA) and G1-PLUS v5 (a version with a different protein source) and found that gestational age- and sex-adjusted birthweight Z scores and the proportion of large-for-gestational-age infants were significantly higher with G1-PLUS, although absolute birthweight did not differ [253]. The octanoate data in mice point in the same direction, with increased fetal and placental weight at 800 µM [227]. The human comparison is retrospective and cannot establish causation. The mouse study is a controlled exposure and supports a causal effect of octanoate in that model, without establishing a human effect or its mechanism. Both justify treating protein source and concentration as a documented variable in the laboratory's culture-system record rather than an interchangeable consumable.

7.1 What actually differs

A sequential system supplies a cleavage medium to day 3 and a blastocyst medium from day 3 to day 5 or 6. The compositional differences that matter are the ones established in Section 5: low glucose, high lactate, non-essential amino acids and glutamine, and EDTA before compaction; higher glucose, lower lactate, all 20 amino acids, vitamins and no EDTA after it [53, 54]. A single-step medium supplies a single formulation throughout, either renewed on day 3 or left undisturbed, and rests on the argument that the embryo regulates its own uptake from a broad supply [58, 59]. The compositional analyses show that "single-step" is not a formulation class: the four products analysed by Morbeck and colleagues varied notably in pyruvate, lactate and amino acids, and two showed an interaction with oxygen in the mouse assay [135]. The pH set point can also differ between stages within a sequential system, which Hentemann and colleagues found beneficial in the mouse [103].

7.2 The clinical evidence

Systematic reviews of the media literature have been hampered by the fact that nearly every trial compares a different pair of products. Mantikou and colleagues found 22 randomised trials evaluating 31 comparisons, only four reporting live birth, and no superior medium on pooled analysis [254]; the 2015 Cochrane review of 32 studies reached the same conclusion and could not pool data [255]. Its 2026 update, with 26 full-article studies and cumulative live birth as a critical outcome, again found no comparisons that could be pooled; the one trial reporting cumulative live birth gave moderate-certainty evidence that G5 likely increases it relative to HTF (44.1% versus 37.9%; OR 1.29, 95% CI 0.98 to 1.70), and the evidence for live birth across the other comparisons was of very low certainty [256]. The primary trials behind the pooled estimates deserve naming. Macklon and colleagues randomised 158 women between five days of a single medium, the same medium renewed on day 3, and G1/G2, and found no difference in blastulation, implantation or pregnancy, the first trial to isolate day 3 renewal as its own arm [257]; Sepúlveda and colleagues, in a donor-oocyte randomised model of 47 donors and 79 recipients, found development on days 3 to 5 and implantation significantly higher in a single medium renewed on day 3 than in a sequential system [258]; Reed and colleagues split 893 sibling embryos from 80 cycles between uninterrupted single medium and sequential media and found no difference in day 3 quality but more blastocysts available and selected for day 5 transfer from the single medium [259]; Paternot and colleagues, randomising 147 women under 36, found comparable numbers of good-quality day 3 embryos but more day 1 cytoplasmic halos, more and more unequal blastomeres, and a higher embryo utilisation rate in the single medium (56% against 49%), a difference in morphology and in usable embryos for that product pair which does not establish a live-birth advantage or a class effect [260]; and a prospective comparison of 972 continuous-medium and 514 sequential-medium cycles powered for blastocyst rate found 37.9% against 33.4% per inseminated oocyte in favour of the continuous medium with similar cumulative delivery rates [261]. For the sequential versus single-step question specifically, Sfontouris and colleagues found no difference in ongoing pregnancy per randomised woman (RR 0.9, 95% CI 0.7 to 1.3; two trials, 246 women), a small increase in blastocyst formation per randomised oocyte or zygote with single media (risk difference +0.06, 95% CI +0.01 to +0.12; ten studies, 7455 oocytes or zygotes, I² = 83%) but no increase in top-quality blastocysts, and very low quality of evidence throughout [62]; Dieamant and colleagues likewise found no difference in clinical or ongoing pregnancy or miscarriage across four trials [63]. The paired randomised SuMMIT trial, in which 2257 sibling zygotes from 192 patients were split between a sequential system and a monophasic medium and euploid blastocysts from each arm were transferred together, found a higher usable blastocyst rate with the sequential system (55.2% versus 46.9%) but no difference in day of blastulation, aneuploidy or sustained implantation among the 126 patients who completed a paired transfer [262]. A non-inferiority sibling-zygote trial of 1356 zygotes from 128 patients found that a single-step time-lapse medium met its prespecified non-inferiority margin (8 percentage points; good-quality day 5 blastocysts 21.1% against 22.2%, reported mean difference −1.2, 95% CI −6.0 to 3.6) against G-1/G-2 for its primary endpoint, with fewer good-quality embryos on day 3, some altered morphokinetic parameters and lower end-of-culture ammonium; non-inferiority on that endpoint, in that comparison, is all the trial establishes [175]. In a 59-patient sibling-oocyte cohort in time-lapse incubators, single medium with or without renewal on day 3 gave blastocyst rates, morphokinetics, pregnancy and birthweight that did not differ significantly [263], and a randomised sibling-oocyte study of 3016 oocytes from 177 PGT-A patients in a dry benchtop incubator found no difference in usable, euploid or mosaic blastocysts per inseminated oocyte between day 3 and day 5 refreshment, with lower odds of day 5 blastulation (OR 0.80) and of high-quality blastulation (OR 0.82) when the day 3 refresh was omitted, which the authors describe as a slight reduction in morphologically high-quality blastocysts without loss of euploid blastocysts [264]. Late renewal does not rescue slow embryos: embryos randomised on day 6 to fresh or unchanged medium reached day 7 biopsy or freezing criteria at the same rate, with an adverse effect of refreshment in women aged 40 or over [265]. Ploidy is a further endpoint on which the two designs have been compared: a retrospective cohort of 347 sequential and 519 single-step PGT-A cycles found more blastocysts but a higher aneuploidy rate with the single-step medium (54.0% against 45.8%) and fewer euploid blastocysts per cycle in women under 38, with no difference in outcome after euploid transfer [266], a counter-signal to the SuMMIT finding that is retrospective and should be read alongside it.

The fair summary is that the available trials have not established consistent superiority of either approach for blastocyst yield or clinical outcome. Failure to detect a difference in heterogeneous and often imprecise studies is not proof of equivalence, and it does not show that all formulations or workflows perform alike. The mechanistic arguments for sequential media are real in animal models and equivocal in the clinic [60, 61], and the decision is usually made on workflow: uninterrupted time-lapse culture and reduced handling favour single-step media, while control of ammonium exposure and stage-specific composition favour sequential media or renewal. The individual product matters more than the design class, and Section 15 shows that products with different compositions have been associated with different birthweights in randomised and pseudo-randomised comparisons [10, 11]; a linked-registry cohort of 1058 single-step and 474 sequential fresh blastocyst singleton births found the single-step class itself associated with higher odds of large-for-gestational-age birth after adjustment for protein, oxygen and fertilization method [267].

7.3 Group culture, volume and density

Whatever the medium, the embryo conditions it. Lane and Gardner showed that reducing incubation volume from 320 to 20 µL and increasing embryo density from 1 to 16 per drop each increased mouse blastocyst cell number and viability after transfer [268]; Paria and Dey demonstrated the same cooperative effect and its rescue by growth factors [198]; and in sheep, culture in groups of four produced blastocysts with cell numbers equal to in vivo controls [69]. In human embryos, Moessner and Dodson found higher cleavage rates and embryo scores in grouped than in single culture [269], and Ebner and colleagues, splitting 936 zygotes between individual, contact-neighbour and group culture in 30 µL drops, found better compaction, blastulation and blastocyst quality with group culture and recommended reducing culture volume or increasing density [270]. The human evidence is not all one way. The earliest randomised test, with four arms from day 3, found blastocyst formation of 35%, 45%, 36% and 36% for group culture in small volume, single culture in small or large volume and group culture in large volume, with only the number of day 3 embryos predicting blastulation [271]; a randomised trial of 830 zygotes from 103 egg-donation cycles in 35 µL drops found blastocyst formation and usable-embryo rates higher with individual than with group culture, with no difference in implantation or pregnancy [272]; a retrospective cohort of 10,941 day 3 embryos cultured at 30, 15 or 10 µL per embryo found no difference in cleavage, morphology or implantation across densities [273]; and a PGT-A cohort found lower blastocyst formation with individual culture (43.5% against 48.5%) but no significant difference in sustained implantation after euploid single transfer (63.9% against 65.0%), so that the effect, where present, is on yield rather than on the competence of the blastocysts obtained [274]. In the other direction, a randomised trial of 532 cycles found clinical pregnancy of 50.8% against 40.6% and live birth of 41.5% against 32.9% with group culture in a microwell dish compared with individual culture, the one live-birth trial in favour of grouping and a design that reconciles group culture with individual identification [275]. In the mouse the remedies for single culture have been mapped: single culture before or after compaction reduced blastocyst cell number, reducing volume from 20 to 2 µL increased it under 5% oxygen but not under 20%, time-lapse microwells increased inner cell mass number, and embryo-conditioned medium raised hatching and cell number [276]. Group culture also masks toxicity in the mouse embryo assay, which is why individually cultured one-cell embryos are the most sensitive QC configuration (Section 16) [277, 278]. The trade-off in single culture is the loss of autocrine support; Kelley and Gardner's data on individual culture combined with atmospheric oxygen quantify how much (Section 4.5) [126].

8.1 Why a second buffer system exists

A bicarbonate-buffered medium holds its pH only while it is in equilibrium with the CO₂ of the incubator. On the bench, CO₂ leaves the medium and pH rises within minutes, a drift that is fastest in small volumes and open dishes [8]. Oocyte retrieval, denudation, ICSI, biopsy, vitrification and transfer preparation all happen in room air, and the oocyte, which in the hamster lacks the Na⁺/H⁺ antiporter and HCO₃⁻/Cl⁻ exchanger that appear at fertilization [104, 105], and which in the human recovers more slowly from an alkaline load when aged or immature [107], is the cell least able to cope with it. Handling media therefore replace most of the bicarbonate with a zwitterionic buffer whose pKa is near physiological pH and which is independent of the gas phase. Quinn and colleagues' 1984 observation that HEPES-buffered medium for manipulation outside a 5% CO₂ atmosphere was associated with better mouse development and more consistent pregnancy rates is the earliest clinical statement of this practice [29].

8.2 HEPES, MOPS and the temperature dependence of pKa

HEPES and MOPS belong to the family of zwitterionic buffers introduced by Good and colleagues in 1966 and extended by Ferguson and colleagues in 1980, selected for pKa values between 6 and 8, high water solubility, minimal membrane permeability, minimal metal binding and low toxicity [279, 280]. Their pKa values fall as temperature rises, by roughly 0.014 pH units per degree for HEPES, so a handling medium adjusted to pH 7.3 at 37 °C reads higher at room temperature and a medium adjusted at room temperature reads lower on a warm stage; HEPES sits near pKa 7.3 and MOPS near 7.0 at 37 °C [279–281]. Swain and Pool showed that combining HEPES, MOPS and DIPSO in different ratios shifts the effective buffering range, extends buffering across the working temperature range, and lowers the absolute concentration of any single buffer, and that all three, singly and in combination, support embryo development [281]. Will, Clark and Swain reviewed the buffers used in ART and concluded that not every biological buffer is appropriate for gametes and embryos, and that adverse effects can be buffer-, species-, cell-type- and concentration-dependent [282]. The primary data behind that conclusion come from other species: in bovine in vitro production, about 41 min of out-of-incubator exposure to 10 mM HEPES, TES or MOPS or to phosphate-buffered saline did not change cleavage, but day 7 to 9 blastocyst rates were higher after HEPES and MOPS than after PBS and TES, and PBS and TES altered the expression of stress and adhesion genes [283]; in porcine embryo culture HEPES did not affect development and 15 mM was adopted, whereas more than 35 mM induced fragmentation [284]; and in the guinea pig, capacitation, the acrosome reaction and zona penetration all occurred in bicarbonate-free MOPS, TES, HEPES, Tris or TAPSO media, but less efficiently than with bicarbonate, so bicarbonate acts as more than a buffer even where it is not absolutely essential [285].

Handling media retain some bicarbonate, in an amount that varies by product. Mouse embryos did not grow in HEPES medium without bicarbonate but developed and produced live young in HEPES medium containing bicarbonate without a CO₂ phase [112], bicarbonate is the substrate of the HCO₃⁻/Cl⁻ exchanger that relieves alkalosis in embryos [105, 106], and bicarbonate is required for efficient sperm capacitation through soluble adenylyl cyclase (Section 10.1) [146]. A handling medium is for handling: Swain's review describes both denuded oocytes and cryopreserved-thawed embryos as especially susceptible to external pH perturbation and argues for stabilising pH during every bench procedure rather than relying on speed [8].

8.3 Light, HEPES and peroxide

HEPES has one hazard that bench procedures amplify. Zigler and colleagues showed that adding 25 mM HEPES to a riboflavin-containing medium markedly increased the production of cytotoxic products on exposure to visible light, that only HEPES and riboflavin need be present, and that hydrogen peroxide is the principal cytotoxic agent [182]. Wang had earlier shown that riboflavin and tryptophan in light-exposed medium yield toxic photoproducts, and Wang and Nixon identified hydrogen peroxide as the lethal species [181, 286]. Handling media formulated without riboflavin remove the photosensitiser, but the general lesson stands: buffered media spend their working life under microscope illumination, and Ottosen and colleagues calculated that 95% of the radiation reaching embryos during IVF procedures comes from microscopes rather than room lighting [287]. Section 14.6 returns to light.

8.4 Composition summary

A handling medium is otherwise a culture medium: the same balanced salts, pyruvate, lactate and usually glucose, amino acids in many current products, albumin at culture concentrations, and antibiotic where the corresponding culture medium contains one. The design difference is the buffer system, and it dictates how the medium is prepared. There is no single rule: a MOPS- or HEPES-buffered product with minimal bicarbonate is warmed and used in air and must not be exposed to elevated CO₂, which would drive its pH below range, whereas a dual-buffered product that retains a working bicarbonate concentration must be equilibrated in a CO₂ incubator before use; one manufacturer gives G-MOPS as an example of the first class and G-GAMETE and ASP as examples of the second [288]. The instructions for use of the specific product govern, and a laboratory that changes handling medium must re-read them. What is general is that handling media are not for extended culture, where the bicarbonate and CO₂ system, the full amino acid complement and the equilibration of the culture dish are what the embryo needs [8, 282].

8.5 Buffered media during ICSI: what enters the oocyte

ICSI is the one procedure in which the handling medium does not stay outside the cell. The question is older than the transcriptome work: in 2006 Morgia and colleagues randomised 708 women to injection in a bicarbonate-buffered medium without HEPES (2204 oocytes) or in a HEPES-buffered medium (2168 oocytes) and reported more triploid and degenerated oocytes, more highly fragmented embryos, and lower pregnancy and implantation rates in the HEPES arm, with no difference in fertilization, cleavage or miscarriage, concluding that HEPES during microinjection should be avoided; the abstract gives no percentages, HEPES concentration or time on the stage, and the trial is single-centre with blinding not stated [289]. Two decades later, Mendola and colleagues showed by confocal microscopy, in 80 human metaphase II oocytes donated to research, that piercing the membranes during sham ICSI allowed unrestricted influx of the surrounding buffer into the oocyte, beyond the operator's control, and by ultra-low-input RNA sequencing of 40 oocytes that the transcriptome then differed by buffer: cytoskeletal transcripts fell in every pierced cohort, the bicarbonate cohort had higher IGF2 and G6PD, the HEPES and MOPS cohorts had higher MAF1, a stress-induced transcriptional repressor, and the HEPES cohort had lower SMC3 and higher GPX1 and LAMP1, which the authors read as oxidative and lysosomal stress; they concluded that a bicarbonate-buffered holding medium during ICSI may reduce the stress imposed on the oocyte [290]. A retrospective comparison of 200 ICSI cycles in one centre in 2023, with metaphase II oocytes allocated to a bicarbonate-buffered single-step culture medium or a HEPES-buffered flushing medium for the injection itself, reported higher fertilization (89.0% against 82.3%), more cleaved embryos and good-quality blastocysts, more embryos frozen, and higher clinical pregnancy and live birth in the bicarbonate group, with no difference in live birth from frozen embryos, miscarriage or multiple pregnancy; the design is retrospective and single-centre, and the abstract does not describe how oocytes were allocated [291]. Campos and Nel-Themaat's 2026 review places these findings where they belong: HEPES- and MOPS-buffered handling media have decades of clinical use behind them, most concerns about zwitterion effects beyond buffering come from experimental settings that do not represent clinical procedures and may reflect media composition rather than the zwitterion itself, the human oocyte evidence is limited and preliminary (the 2006 trial notwithstanding), and the case for bicarbonate-buffered ICSI media is one for further evaluation with standardised pH measurement and attention to oil and dish, not for an immediate change in practice [292]. There is also a practical trade: a bicarbonate-buffered medium used on the ICSI stage loses CO₂ and rises in pH from the moment it leaves the incubator (Section 14.2), so a laboratory that trials one exchanges a buffer question for a time and gas question, which is what the call for standardised pH_e measurement is about. The medium in the hours after injection acts on the oocyte too: four hours in either of two commercial human media after mouse ICSI gave fewer and less frequent calcium oscillations than KSOM, and female offspring from those groups were heavier with larger organs [293].

8.6 Procedure media for ICSI and biopsy: PVP, hyaluronate, hyaluronidase and calcium-free media

Sperm-slowing media. Polyvinylpyrrolidone was adopted in the first ICSI protocols to slow sperm for capture and to control fluid movement in the injection pipette. Two of the founding papers already described its downside: immobilisation by squeezing the tail damages the sperm membrane enough for thiol reagents to reach the nucleus, and PVP in the drop impeded that access, so that PVP may interfere with nuclear decondensation [294], and transmission electron microscopy showed PVP damaging the plasma, acrosomal and mitochondrial membranes with secondary deterioration of chromatin and axoneme [295]. A sibling-oocyte study in which the sperm was caught by its tail without PVP gave 74% normal fertilization against 62% with conventional PVP-ICSI at similar survival [296], a non-randomised series reported 84% against 58% [297], and a prospective three-arm study of 210 cycles found every outcome improve as the PVP concentration fell from 10% to 7% to 5%, with cleavage, good-quality cleavage-embryo and good-quality blastocyst rates all higher at 5% than at 10% [298]. Time in PVP matters as much as concentration: viability, morphology, chromatin quality and mitochondrial membrane potential fell and DNA fragmentation and acrosome reaction rose after 15 min in 10% PVP, the authors advising less than 15 min [299], and DNA fragmentation measured by TUNEL rose after 1 to 1.5 h in either PVP or hyaluronate, with similar increases in the two media, so that time, not medium, is the variable [300]. Animal work shows where the injected PVP goes: in cattle, 10% PVP increased acrosome-reacted sperm and ICSI fertilization but injected PVP remained detectable in 41% of embryos and suppressed cleavage, morula and blastocyst development, with effects that differed among three PVP suppliers [301], whereas sham injection of 2 to 3 pL of medium with or without PVP into bovine zygotes did not reduce fertilization, hatching or cell number, so the volume delivered is part of the question [302]. In a mouse ICSI model, sperm handled in PVP, hyaluronate or plain medium activated oocytes and supported development equally, and PVP-exposed sperm had a lower oxidation-reduction potential than hyaluronate-exposed sperm [303]. Hyaluronate was introduced as the alternative by Balaban and colleagues, whose hyaluronate product slowed sperm enough for capture, prevented sticking to plastic and glass, did not affect zygote development and gave clinical pregnancy rates similar to PVP [304]; in a randomised sibling-oocyte comparison of 206 oocytes, hyaluronate gave less abnormal fertilization (1.9% against 9.7%), a trend to higher normal fertilization (73.8% against 62.1%) and a longer injection time (2.5 against 2.1 min), with no difference in day 3 morphology or morphokinetics [305]. PVP has one deliberate oocyte-side use: after abnormal oolemma rupture in Piezo-ICSI, immediate exposure to 7% PVP reduced degeneration from 19.7% to 6.0% and increased normal fertilization [306]. A hyaluronate slowing medium should not be confused with hyaluronan-binding sperm selection (PICSI), which uses a hyaluronan-coated dish to select bound sperm: the HABSelect trial of 2752 couples found no significant difference in term live birth (27.4% against 25.2%) [307], its mechanistic arm attributed the equalisation of live birth rates between older and younger women in the PICSI arm to avoidance of DNA-damaged sperm [308], the earlier multicentre trial had found lower pregnancy loss in couples whose sperm bound hyaluronan poorly [309], a 2026 sibling-oocyte study found no effect on embryo euploidy in couples with mostly low sperm DNA fragmentation (52.1% against 54.7%) [310], and the 2026 Cochrane update, which removed four earlier trials over data trustworthiness, concluded that hyaluronan selection may make no difference to or slightly increase live birth (RR 1.09, 95% CI 0.97 to 1.24; one trial, low certainty) but probably slightly reduces miscarriage (RR 0.59, 95% CI 0.44 to 0.80; two trials, moderate certainty) [311].

Calcium in the injection medium. The calcium content of the medium at injection is a variable with a measurable effect. In hamster oocytes injected with human sperm, a calcium-free injection medium improved male pronucleus formation in post-ovulatory aged oocytes (49.5% against 32.3% at 16 h), where parthenogenetic activation is the hazard [312]; in human oocytes the ionomycin-induced calcium rise increased with the total calcium of the commercial medium, and in the mouse raising calcium three- or sixfold gave no benefit [313]; and in a retrospective paired-cycle study of 178 patients, adding 5 mM calcium chloride to the PVP drop raised fertilization from 28.8% to 49.7% and clinical pregnancy from 4.9% to 25.0% in the cohort with prior fertilization below 50%, with no fertilization gain in patients whose previous fertilization had been normal [314]. Vigorous aspiration of ooplasm during injection was itself shown to increase calcium influx and fertilization in the early ICSI literature [315].

Hyaluronidase. Cumulus removal before ICSI uses hyaluronidase followed by mechanical stripping, and both the enzyme's concentration and its source have been tested. Van de Velde and colleagues found that 10 IU/mL was as effective as 39 or 78 IU/mL when a pipette of at least 1000 µm inner diameter was used, with no effect on survival, fertilization, activation, development or pregnancy, and that the wider pipette shortened the exposure to the enzyme [316]; a small randomised sibling-oocyte study reported higher fertilization (92.3% against 80.6%) and better embryo quality with 8 IU/mL than with 80 IU/mL [317]. Bovine testicular hyaluronidase was the traditional source; a recombinant human enzyme proved non-inferior in a randomised sibling-oocyte trial of 135 patients (oocyte intactness 89.6% against 92.9%, fertilization 73.9% against 77.1%) [318], a retrospective series reported higher fertilization and less oocyte damage after switching to it [319], and a meta-analysis of three randomised trials with 2445 oocytes found no difference in fertilization, embryo quality or live birth between recombinant and bovine enzyme [320]. The concern about the enzyme is mechanistic as well as regulatory: in bovine oocytes, hyaluronidase treatment lowered blastocyst rates after ICSI (22.4% against 36.1%) and after IVF, with higher intracellular calcium in treated oocytes and no change in reactive oxygen species or glutathione [321]. In a large observational dataset with electronic witnessing of 7999 oocytes, denudation and ICSI timings and operators did not associate with the cumulative delivery rate once maternal age and oocyte number were accounted for, with only a mild association between trigger-to-denudation time and blastulation [322].

Biopsy media. The two biopsy stages use different media, and the distinction matters. For cleavage-stage (blastomere) biopsy, calcium- and magnesium-free medium loosens the calcium-dependent adhesion between blastomeres that begins with compaction (Section 5.1): Dumoulin and colleagues showed for day 3 biopsy that Ca²⁺/Mg²⁺-free medium allowed easier removal with less cell lysis and a shorter procedure, and that development to the blastocyst was unaffected even after 45 min of exposure, an observation that belongs to that stage and that experiment [323]. The ESHRE PGT Consortium and SIG Embryology recommendations describe cleavage-stage biopsy either directly in the Ca²⁺/Mg²⁺-free biopsy medium or in a HEPES-buffered medium after a period in the biopsy medium, whereas for trophectoderm biopsy they state that Ca²⁺/Mg²⁺-free medium should not be used and that the biopsy is performed in a buffered handling medium of the kind Section 8 describes, the trophectoderm cells being aspirated and excised by laser or mechanically rather than freed by loosening their adhesion [324]; the 2026 ESHRE laboratory recommendations refer to that document for the technique and recommend removing about 5 to 10 trophectoderm cells [89]. Trophectoderm biopsy protocols differ in more than the medium: in a matched retrospective comparison of 835 pairs of euploid frozen transfers, blastocysts biopsied with sequential hatching and biopsy at the blastocyst stage had higher survival, clinical pregnancy, ongoing implantation and live birth than those from a day 3 pre-hatching protocol, and no monozygotic twins against four [325], and biopsied embryos grown to the blastocyst stage survived cryopreservation and implanted as well as intact blastocysts in the slow-freezing era [326]. Beyond the omission of the two divalent cations at the cleavage stage, the buffer, protein and exposure limits of a biopsy medium are product-specific and taken from the instructions for use; the general principle, that the oocyte and embryo are outside the incubator in a buffered medium for these procedures, is the one Section 8 has set out.

9.1 What fertilization requires of a medium

Conventional insemination asks one medium to serve three cell types at once: capacitating sperm, a metaphase II oocyte inside its cumulus, and, within hours, a zygote. HTF was designed for this purpose, and its glucose (2.78 mM), bicarbonate (25 mM), calcium and albumin satisfy the sperm's requirements while pyruvate and lactate satisfy the oocyte's [33]. Sperm need bicarbonate and calcium to capacitate and albumin as a cholesterol acceptor (Section 10.1) [231, 327]; in the mouse, replacing the bicarbonate buffer with HEPES at identical pH abolished fertilization and 10 mM bicarbonate restored it, the requirement being for the acrosome reaction at the zona [111]. They need a glycolysable sugar: in the mouse, a glycolytic product is obligatory for initiation of the acrosome reaction and the whiplash motility required for fertilization, even though early capacitation events proceed without exogenous substrate [328], and in human sperm glucose or fructose is required for hyperactivated motility and glycolytic ATP for vigorous motility, with 5.56 mM glucose supporting the greatest number of acrosome reactions [329]. Dale and colleagues also found a pH-sensitive step in conventional fertilization that ICSI bypassed, which they attributed to sperm-zona interaction (Section 4.3); the experiment concerned activation under the conditions tested and does not make the injected oocyte or its later development insensitive to pH, but it is one reason a fertilization medium must be properly equilibrated at the time of insemination [107].

Fertilization media therefore keep glucose at HTF-like concentrations even when the paired cleavage medium is glucose-restricted. In Quinn's 1995 series the gametes were washed and inseminated in the glucose- and phosphate-free HTF derivative with high fertilization rates [34, 35], but in later commercial practice the derivative (P-1) was used from the zygote stage with insemination in a glucose-containing medium, a pairing consistent with the sperm's requirement for a glycolysable sugar [328, 329]; in a sibling comparison the same group later found embryos of better morphology after insemination in Basal XI HTF than in a glucose-containing HTF variant, with no fertilization difference [330]. Whether a dedicated fertilization medium is needed at all has, among the studies reviewed here, been tested in one comparison: insemination in a low-lactate continuous culture medium gave the same fertilization rate as a dedicated fertilization medium with a lower triploidy rate and more good-quality blastocysts, in a two-arm comparison not described as randomised [331]. A physiological component missing from HTF-type media has also been identified: creatine, present in follicular and oviductal fluid, prolonged mouse sperm capacitation and supported fertilization with very small sperm numbers when added to IVF medium [332]. For ICSI the requirements collapse to those of the oocyte and zygote, and in common practice the injected oocyte is placed directly into cleavage medium.

9.2 Duration of gamete co-incubation

Sperm in a fertilization dish are a source of reactive oxygen species and metabolic waste. Gianaroli and colleagues randomised 119 patients to 1 h or 16 h sperm-oocyte exposure and found a higher fertilization rate with 1 h (73% versus 66%), less polyspermy (1% versus 3%), more four- to five-cell embryos at 40 h, and implantation and pregnancy rates of 11% and 28% against 8% and 15%; fluorescently labelled sperm entered the cumulus within 15 min and appeared in the oocyte cortex at 4 h [333]. Catt and Henman reported that reducing oxygen during insemination, adding protective components to the medium and shortening insemination each increased success rates, while the age-related decline in outcome persisted [334]. The finding was confirmed in sibling gametes, where fertilization did not differ between 1 h and overnight co-incubation but embryo quality was better after 1 h [330], and ammonium measured in spent insemination medium from eight patients' oocyte cohorts was higher after 16 to 18 h of conventional co-incubation (about 27 µM) than after 2 h of brief co-incubation (about 21 µM) or in medium incubated without gametes (about 19 µM), the paper's assay range being 7 to 286 µM, and the same paper's retrospective series of 609 live-birth cycles found more preterm births after conventional co-incubation, an association driven by twins that does not establish ammonium as its cause [335]. The Cochrane review of eight trials found brief co-incubation associated with higher ongoing pregnancy (OR 2.42) and clinical pregnancy (OR 2.36) on low-quality evidence, with live birth unreported [336]. The live-birth question was then asked directly: a triple-blind trial randomised 320 women to 3 to 4 h or 20 h of co-incubation with live birth in the fresh cycle as its primary outcome and found 33.0% (53 of 160) against 36.8% (59 of 160) by intention to treat (RR 0.90, 95% CI 0.67 to 1.21), with no difference in clinical or ongoing pregnancy, miscarriage or cumulative live birth, at motile sperm concentrations of 0.3 to 1.2 million/mL [337], and a 2023 meta-analysis of eleven randomised trials found higher implantation (OR 1.97), ongoing pregnancy (OR 2.18) and top-quality embryo rates with brief co-incubation but no difference in live birth (OR 1.09, 95% CI 0.72 to 1.65) and a slightly lower normal fertilization rate (OR 0.89, 95% CI 0.80 to 0.99) [338]. The improved intermediate endpoints have therefore not been shown to translate into more live births, and the composition of the fertilization medium remains inseparable from how long the gametes are left in it.

10.1 Capacitation: bicarbonate, calcium and albumin

Capacitation is a medium-driven process. Visconti and colleagues showed that mouse sperm incubated in media lacking bovine serum albumin, calcium chloride or sodium bicarbonate fail to capacitate, as judged by chlortetracycline pattern, zona-induced acrosome reaction and fertilization, that the capacitation-associated protein tyrosine phosphorylation does not occur, and that each function is restored by adding back the missing component in a concentration-dependent manner; the bicarbonate requirement is not explained by a change in intracellular or medium pH [327]. The bicarbonate sensor is soluble adenylyl cyclase, which bicarbonate stimulates directly and pH-independently to raise cAMP [146], and bicarbonate through protein kinase A activates the phospholipid scramblase that disorders the sperm plasma membrane and prepares it for cholesterol efflux and the acrosome reaction [339]. In hamster sperm, as little as 2.9 mM bicarbonate under 0.6% CO₂ increased the motility index 2.7- to 3.6-fold, progressive and hyperactivated motility rose dose-dependently, and the half-maximal and maximal bicarbonate concentrations for the full set of responses were 9.2 and 34 mM; the late responses depended on bicarbonate being present early in capacitation [340]. Albumin acts as a sterol acceptor: delipidated albumin still supported mouse sperm capacitation and mediated a 20% fall in sperm sterol, and cholesterol-loaded albumin inhibited fertilization dose-dependently [231]; Cross reviewed cholesterol efflux as the membrane basis of capacitation [341]. Hyperactivated motility, the flagellar pattern needed to penetrate the cumulus and zona, is regulated by calcium signalling and reviewed by Suarez [342].

10.2 Energy substrates and antioxidants for sperm

Sperm media carry glucose at a higher concentration than cleavage media because sperm motility draws on glycolysis along the principal piece as well as on mitochondrial respiration in the midpiece, and the balance between them depends on species, substrate availability and the condition of the cells: capacitation increased flux through both glycolysis and oxidative phosphorylation in mouse and human sperm, and while epididymal mouse sperm depended on glucose, ejaculated mouse and human sperm could also use pyruvate and citrate [343], and pyruvate in the mouse stimulated hyperactivation not as a mitochondrial fuel but by supplying NAD⁺ to flagellar glycolysis through repression of lactate oxidation, which is a function for the pyruvate and lactate that wash media carry [344]. In the mouse, glycolysis supplied most of the ATP for flagellar movement and motility could not be sustained on respiratory substrates alone when glycolysis was inhibited [345]; in human sperm, glucose or fructose was required for hyperactivated motility and for vigorous motility over hours, and without a glycolysable sugar only 2 to 3% of sperm remained motile after 18 h [329]. Sperm are also uniquely exposed to their own oxidative chemistry. Aitken and Clarkson showed that centrifuging unselected sperm suspensions provokes a burst of reactive oxygen species from a subpopulation of defective cells that then damages the functional sperm around them, that separating motile sperm before centrifugation avoids this, and that vitamin E protects [346]. Albumin was the most effective inhibitor of lipid peroxidation and loss of motility in rabbit sperm, followed by hypotaurine, taurine, epinephrine, pyruvate and lactate, with hypotaurine acting as an intracellular superoxide scavenger [347]. Hypotaurine and taurine have therefore appeared in sperm and fertilization media, and albumin at 5 mg/mL or more does double duty as sterol acceptor and antioxidant.

10.3 Density gradient media and swim-up

Preparation removes seminal plasma, whose decapacitation factors hold sperm in the uncapacitated state, together with debris, leukocytes and the ROS-producing sperm described above [346, 348]. The method was described by Lopata and colleagues in 1976, who let sperm migrate from semen into a diluent for an hour, obtained plasma-free suspensions of higher motility than the ejaculate whereas repeated dilution and centrifugation lowered motility, and formulated a diluent matching the electrolyte composition of preovulatory human tubal fluid, which they had analysed for electrolytes, substrates and trace elements; in it sperm swam faster and deteriorated more slowly over 24 h, a tubal-fluid-modelled sperm medium a decade before HTF [349]. The buffer of the preparation medium has recently been shown to matter: in a split-ejaculate study of 54 men, swim-up in bicarbonate-buffered medium yielded sperm with higher motility and mitochondrial activity than swim-up in media containing HEPES or MOPS, alone or with bicarbonate, which gave reduced motility, poorer kinematics and lower mitochondrial potential [350], and in 108 asthenozoospermic ejaculates (from a series of 136, 28 of which were used to set a 2 h incubation time) incubation in a continuous single-step culture medium gave higher motility than three other media in routine use, without numerical values in the abstract [351]. Density gradient media consist of colloidal silica particles in a HEPES-buffered isotonic salt solution, prepared as layers of two or more densities by diluting the stock (commonly to about 40% and 80%, though the product-specific dilutions vary); the density comes from the silica, and whether and how much protein the solution contains is a separate, product-specific property stated in the instructions for use. Percoll, a polyvinylpyrrolidone-coated silica, was the original clinical gradient; after it was withdrawn from clinical use, silane-coated silica products (ISolate, PureSperm) and iodixanol (OptiPrep) were evaluated against it, and the silane-coated products gave recovery, vitality, motility, morphology and chromatin condensation not significantly different from Percoll, whereas iodixanol gave lower recovery [348, 352]. Mortimer's review remains the reference for the principles of gradient, swim-up and wash preparation and their selection according to sample quality [348]. The WHO laboratory manual, sixth edition, standardises the methods and their reference limits [353]. For intrauterine insemination, the Cochrane review found only very low-quality evidence and could not establish a difference between swim-up and gradient techniques in clinical pregnancy, with live birth unreported [354].

10.4 Sperm cryopreservation media

Sperm freezing predates every other cryopreservation in ART. Polge, Smith and Parkes discovered in 1949 that glycerol allowed fowl spermatozoa to survive freezing to low temperature [355], and Sherman's 1973 synopsis records the state of human semen banking with glycerol-based media after the first births from frozen semen in the 1950s [356]. The permeating cryoprotectant in the sperm freezing media cited here, and in the WHO manual's description of sperm cryopreservation, is glycerol [353]; an alternative free of permeating cryoprotectant for oligoasthenoteratozoospermic samples, in which 5% high-molecular-weight (360 kDa) PVP serves as a non-permeating protective additive with vapour cooling and plunging, gave fewer morphological abnormalities, more sperm with high mitochondrial membrane potential and lower DNA fragmentation than 5% glycerol in a paired comparison confined to that sperm population [357]. Mahadevan and Trounson found that a medium containing 15% glycerol, 0.05 M sucrose and 1% glycine preserved motility and vitality at least as well as egg yolk-citrate containing 15% glycerol and gave a higher insemination pregnancy rate; glycerol was superior to DMSO or ethylene glycol, 7.5% glycerol and 7.5% ethylene glycol did not differ, sucrose or raffinose improved survival, the optimal pH was 6.5 to 7, and diluent-to-semen ratios from 1:1 to 1:4 did not affect survival, which means the final glycerol concentration the sperm experienced ranged from about 7.5% down to 3% depending on the ratio [358]. Any recipe therefore needs three numbers, the stock concentration, the mixing ratio and the resulting final concentration, and the 1980s formulations cited here are historical reference points, not a description of current products, whose composition and mixing instructions come from the manufacturer. Weidel and Prins compared eight buffer systems and found that TES-Tris-citrate-egg yolk-glycerol (TEST-yolk) media maintained the highest post-thaw progressive motility, cervical mucus penetration and vitality, and that once the medium was optimal the container and thaw rate made little difference [359]. Egg yolk supplies lipoproteins that stabilise the sperm membrane during cooling, the rationale for the yolk in TEST-yolk media [353, 359]; yolk-free formulations rely on other components, as the glycerol, sucrose and glycine medium of Mahadevan and Trounson did [358], and the mechanisms of cryoinjury across the lethal intermediate zone between about −15 and −60 °C during cooling and again during warming are reviewed by Gao and Critser [360]. The WHO manual describes the glycerol-egg yolk-citrate and glycerol-based media in current use, together with the infectious-disease considerations of animal-derived yolk [353].

11.1 Rationale

The transfer medium is the last fluid the embryo sees before the endometrium, and the case for making it different from culture medium rests on the mouse experiments of Gardner, Rodriguez-Martinez and Lane: replacing albumin with hyaluronan gave the highest implantation and fetal development after blastocyst transfer, and the benefit was traced to hyaluronan's presence in the transfer medium rather than during culture [186]. Hyaluronan is a major glycosaminoglycan of the female tract at the time of implantation [186], its receptor CD44 is present on human embryos through the pre-hatching blastocyst [188], and the proposed mechanisms include receptor-mediated adhesion and a viscosity effect that limits embryo expulsion and dispersal after deposition [225]. The viscosity effect has been measured: a hyaluronan-enriched transfer medium had an estimated viscosity of 3.59 cP against 1.28 cP for a conventional transfer medium, 1.15 to 1.77 cP for complete culture media and 1.06 cP for water, so hyaluronan roughly triples viscosity while protein supplementation barely changes it [361]. The adhesion mechanism has been tested directly in a human model and not supported: human embryos expressed CD44, HMMR, the hyaluronan synthases and hyaluronidase genes, but 10 min of pre-treatment with a widely used hyaluronan transfer medium did not promote blastocyst attachment to endometrial epithelial cells over 1 to 48 h of co-culture [362]. Simon and colleagues then showed in a randomised double-blind trial that 0.5 mg/mL fermentation-derived hyaluronan could replace serum substitute entirely as the sole macromolecule in a human transfer medium, with clinical pregnancy of 62.5% versus 52% and implantation of 34% versus 26.8%, differences that did not reach significance in 80 patients but established that a blood-product-free transfer medium was feasible [363].

11.2 The clinical evidence

Urman and colleagues randomised 1282 consecutive fresh transfers to hyaluronan-enriched or conventional transfer medium and found clinical pregnancy of 54.6% versus 48.5% and implantation of 32% versus 25%, a number needed to treat of 17 overall and of 7 to 8 in women over 35, with previous failed cycles, or with only poor-quality embryos [364]. The 2020 Cochrane review of 26 studies with 6704 participants concluded on moderate-quality evidence that a functional hyaluronan concentration (0.5 mg/mL) probably increases live birth (RR 1.21, 95% CI 1.10 to 1.31; 10 trials, 4066 women), from an assumed 33% to between 37% and 44%, and clinical pregnancy (RR 1.16), with a probable increase in multiple pregnancy (RR 1.45) and only low-quality evidence for a small reduction in miscarriage that disappeared when only low-risk-of-bias trials were included [225]. That estimate has since weakened. A 2026 systematic review with a search to December 2024 pooled eleven randomised trials reporting live birth (803 of 2026 against 697 of 2037) and found RR 1.14 (95% CI 0.99 to 1.31, P = 0.07, I² = 54%), falling to RR 1.06 (0.85 to 1.31) when trials at high risk of bias were excluded, with the clinical pregnancy effect likewise disappearing on sensitivity analysis, most included trials at high risk of bias, and the overall certainty of the evidence rated very low [365]. A separate meta-analysis that split the trials by oocyte source found the benefit of a functional hyaluronan concentration on clinical pregnancy, live birth and multiple pregnancy in autologous-oocyte cycles but not in the few donor-oocyte trials (live birth RR 1.12, 95% CI 0.86 to 1.44 in two studies of 317 participants; clinical pregnancy RR 1.06, 95% CI 0.97 to 1.28 in three studies of 351), which its authors read as either an intrinsic difference or a lack of power [366]. The frozen-cycle trial on which the ESHRE statement rests randomised 550 women in two centres to a 0.5 mg/mL hyaluronan transfer medium or to a conventional blastocyst medium that itself contains 0.125 mg/mL hyaluronan, and found live birth of 25.5% against 25.8% (RR 0.99, 95% CI 0.74 to 1.31) with all other outcomes similar; the comparator in most of these trials is low-hyaluronan rather than hyaluronan-free [367]. Observational data for frozen transfers are divided: a retrospective cohort of 1221 single frozen euploid transfers found live birth of 59.1% with hyaluronan-enriched medium against 43.2% with a zwitterion-buffered medium (RR 1.37, 95% CI 1.22 to 1.54), which its authors suggested may reflect the zona opening made at trophectoderm biopsy, while warning of unmeasured confounding [368], whereas a stratified cohort of 549 single euploid transfers found live birth of 66.3% against 72.9% (not significant) with a higher positive hCG rate only for C-grade blastocysts [369], an earlier cohort of 1721 frozen cycles found lower implantation and clinical pregnancy with the enriched medium in first attempts and higher in third or later attempts [370], and a small series of 85 morphologically poor euploid blastocysts found clinical pregnancy of 47.4% against 21.5% [371]. On exposure time and safety, a before-and-after cohort of 3391 fresh transfers found the live-birth effect of a hyaluronan-rich medium irrespective of whether embryos were exposed for 2 to 4 h or 10 to 30 min, with no difference in gestational age, birthweight or sex ratio [372]. The single-embryo-transfer trial the Cochrane authors called for has been registered as a four-centre patient-blinded randomised trial of 858 women undergoing a single fresh or frozen blastocyst transfer, with live birth per transfer as the primary outcome and the sample size calculated for the overall comparison [373]. The Cochrane reviewers attributed the multiple pregnancy signal to transferring more than one embryo in the presence of hyaluronan; it is a clinical observation, not proof of a defined adhesion mechanism; the Cochrane authors call for trials of adherence compounds with single embryo transfer, and ESHRE's advice is to monitor the multiple pregnancy rate. ESHRE's 2023 add-on recommendations, drawing on the same review and on a subsequent randomised trial in frozen cycles, concluded that hyaluronic acid in transfer media increases live birth after fresh transfers without a significant effect on adverse outcomes, that no effect was seen after frozen transfers, and that its addition is recommended with continued monitoring of the multiple pregnancy rate [2]. That recommendation predates the 2026 review. Read together, the position is that the pooled live-birth benefit seen in the earlier trials, most of them at high risk of bias, is no longer statistically clear and rests on very-low-certainty evidence, that the one adequately sized frozen-cycle trial found no effect, and that neither a non-significant pooled estimate nor a guideline recommendation proves or disproves the biological effect of hyaluronan. The registered single-embryo-transfer trial includes both fresh and frozen transfers and should add to the live-birth evidence; what it can say about fresh transfers in particular will depend on recruitment, the prespecified analyses and the precision of that subgroup. Compositionally, transfer media are blastocyst-stage media, HEPES- or bicarbonate-buffered according to the intended handling, with hyaluronan added; the commercial product EmbryoGlue retains albumin in recombinant form [187].

12.1 What oil does

Brinster's 1963 method of culturing embryos in microdrops under paraffin oil solved three problems at once: evaporation, which concentrates the drop; gas exchange, which is slowed but not prevented so that pH and temperature change more gradually when a dish is moved; and access, since a drop under oil can be observed and manipulated without a lid [16]. Oil also buffers the drop against airborne contaminants, and, because it is hydrophobic, it takes lipophilic compounds out of the medium. Miller and Pursel showed that radiolabelled oestradiol, progesterone and androstenedione fell by 51%, 89% and 77% in medium under oil within 24 h, that up to 14% of oestradiol moved through the oil into a neighbouring drop, and that FSH, leucine, glucose, lactate, sodium and prostaglandins did not enter the oil [374]. Oil thus alters medium composition selectively, removing steroids and other lipophilic molecules, and it can carry them between drops; in pig oocyte maturation, progesterone, oestradiol and testosterone accumulated in medium without oil but did not reach high concentrations under mineral oil, and nuclear maturation and later blastocyst formation were better without the overlay, which shows the functional consequence of the sink for a hormone-dependent process [375]. The sink works in the other direction as well: a medium that supported no development uncovered supported 96% blastocysts under oil, so an overlay can remove an embryotoxic substance from the medium and can mask a toxic medium, another reason the exact configuration must be tested [376]. Gas exchange under oil has now been timed: drops prepared in room air reached 5% oxygen with a half-life of 71 min under light oil and 116 min under heavy oil, and reoxygenated with a half-time of about 50 min under light oil in an atmospheric incubator (Section 4.5) [133].

12.2 What oil is

Mineral and paraffin oils are complex mixtures of saturated hydrocarbons refined from petroleum, differing in chain length distribution, viscosity and residual unsaturated or aromatic content. Morbeck and Leonard describe oil as the least defined product used in IVF and review its history, chemistry, processing and optimal use [377]. The broadest direct comparison characterised 13 commercial oil brands for viscosity, density, peroxide value and oxidation potential, for their capacity to damp pH, osmolality and temperature fluctuation, and in a sensitised mouse embryo assay with cell counts: the brands differed in their stabilising capacity, higher viscosity tended to protect against fluctuation, two of the 13 marketed oils were embryotoxic in the sensitised assay, and total and inner cell mass cell numbers differed between oil groups, with a single lot tested per brand [378]. The human trials are few. A randomised comparison of four mineral oils in 500 women found similar day 2 quality but more top-quality day 3 embryos under one oil than under two others [379]; a randomised comparison of 1237 sibling oocytes under a paraffin or a mineral oil found no difference in fertilization, top-quality day 3 embryos (41.7% against 41.2%), utilisation or live birth per transfer (26.9% against 26.2%) [380]; and a two-centre cohort of 762 couples using a paraffin oil and a mineral oil in successive periods found live birth after the first transfer of 40.6% against 39.5% (RR 1.02, 95% CI 0.91 to 1.14) [381], so where oils differ it is product by product rather than paraffin against mineral as classes. Martinez and colleagues compared two commercial oils in a pig model, found that one reduced cleavage and blastocyst rates without any difference in oxidation state or element transfer, and traced the difference to the composition of volatile organic compounds in the oils and the transfer of straight-chain alkanes, pentanal and 1,3-diethylbenzene into the medium [382]. Culture oil used as an overlay is exempt from FDA premarket notification within the limits of the regulation, though it remains a regulated device subject to the manufacturer's quality system; in practice the manufacturer's certificate and the laboratory's receiving inspection are the controls the laboratory can act on [85].

12.3 Peroxidation

The characteristic failure of oil is peroxidation. Otsuki, Nagai and Chiba traced unexpected deterioration of human embryos to elevated peroxide levels in the oil, caused by oxygen exposure and unfavourable storage [383]. They then showed that the time to haemolysis of erythrocytes under oil tracked the oil's peroxide value, that albumin in the medium facilitated both haemolysis and the passage of a lipophilic tracer into the human zona pellucida, and that under peroxidised oil (peroxide value above 0.02 meq/kg) the tracer no longer entered the zona, which the authors interpreted as prior uptake of agents from the oil; they proposed albumin-mediated transfer as the route by which agents from peroxidised oil reach the embryo, a mechanism inferred from a tracer rather than measured for the peroxides themselves [232]. Peroxides are not the only contaminant class: one batch of silicone oil blocked one- and two-cell mouse development in KSOM while permitting high blastocyst rates in CZB, whose higher EDTA and albumin protected, with circumstantial evidence pointing to zinc, so that whether an oil lot reads as toxic depends on the medium under it and an assay in one medium does not qualify the oil for another [384]. Standard quality control assays underestimate the hazard: the one-cell mouse embryo assay was twice as sensitive as the two-cell assay and twenty times as sensitive as the human sperm motility assay to cumene hydroperoxide in oil, its sensitivity doubled with individual culture, and peroxide reduced blastocyst cell number dose-dependently [277]. Extending the mouse assay to 144 h or adding cell counts detected oil toxicity that the 96 h blastocyst rate missed, with fourfold greater sensitivity, and group culture had a protective effect that masks toxicity [278]. Time-lapse morphokinetics detected toxicity in two lots of oil that did not affect blastocyst rate [385]. The protein in the assay medium matters too: for low peroxide levels, 5 mg/mL HSA with or without globulins increased detection of toxicity compared with BSA, and group culture and certain micro-well dishes greatly reduced sensitivity [386]. In a mouse study of two contaminated lots, washing oil with water, medium or medium plus albumin was equally effective at reducing toxicity and the concentration of peroxides, aldehydes and alkenals in one lot, and reduced the surfactant Triton X-100 in the other by 25% [387]. That is evidence about which contaminants are water-extractable, not a release procedure: a 25% reduction of one contaminant does not show that every hazardous constituent has been removed, and a clinical lot that fails or is suspected should be quarantined, investigated with the manufacturer and released or rejected as a documented decision under the laboratory's quality system, the non-conformance handling that the laboratory guidance expects [88, 89].

12.4 Evaporation under oil

Oil reduces but does not abolish evaporation. In non-humidified benchtop incubators the osmolality of 50 and 200 µL drops under oil rose linearly over five days by up to 20 mOsm/kg, the increase depended on drop volume and possibly on the oil, and no change occurred in a humidified incubator [9]. The oil is now known to be a determinant: in a dry incubator osmolality rose progressively from day 3 to day 7 under both oils tested and not at all in a humidified incubator, and a single-step medium reached higher osmolality under a mineral than under a paraffin oil [388]; over six days in a dry incubator the rise was greatest under 3.0 mL of light oil and least under 4.5 mL of heavy oil, and yet human blastocyst formation, quality and cell count did not differ between those two extremes, a caution that the mouse thresholds of Section 14.3 are not validated human limits [389]. A randomised response-surface study of 107 factor combinations over seven days showed that the surface area-to-volume ratio of the medium, the oil density and the oil height interact quantitatively to set the rate of osmolality rise, and produced a predictive model [390]. Mestres and colleagues identified incubator humidity, oil volume, medium type and time-lapse dish design as the determinants of evaporation, found that protein supplementation made no difference, and warned that media with initially high osmolality run the greater risk of becoming hypertonic during culture [391]. In a randomised trial of 297 women, dry culture as implemented in that centre's benchtop system produced lower day 3 quality and compaction, lower blastocyst formation, quality and cryopreservation rates, and lower clinical and ongoing pregnancy than humidified culture (OR 0.57, 95% CI 0.36 to 0.91, and OR 0.54, 95% CI 0.34 to 0.85); the paper's 2021 correction relabels the dispersion statistic in its patient table and revises the chemical-pregnancy odds ratio and leaves the outcomes cited here unchanged, and because humidity interacts with oil depth, dish geometry, drop size, starting osmolality and duration, the finding describes that configuration rather than dry incubation in general [392]. A second clinical dataset exists beside that trial: 20 µL drops under oil in a dry incubator reached 308 mOsm after three days against 285 to 290 mOsm when medium was added to the outer wells, and day 3 transfers from the dry configuration had ongoing pregnancy of 37.8% against 50.2% and miscarriage of 8.3% against 2.3%, with day 5 miscarriage of 25.0% against 2.2%, a within-incubator humidification the section had not described [393]. Pre-equilibration adds a further evaporation step: media pre-equilibrated in a dry incubator for 24 h had higher osmolality at 96 h than those pre-equilibrated humid, pre-humidifying the oil itself made no difference, and a retrospective comparison found more good blastocysts after humid pre-equilibration with no difference in clinical pregnancy [394]. The practical response is to measure evaporation and osmolality in the laboratory's own dishes, oil and incubators over the full culture interval [391]. Drop preparation itself matters: pipetting drops before overlay on a warm stage with airflow raised osmolality by up to 40 mOsm/kg within 5 min, and overlaying first and then replacing the drop limited the rise [98].

12.5 Volatile organic compounds and air

Oil is also a sink for what is in the air. Cohen and colleagues found that incubator air can be dirtier than unfiltered outside air because of volatile organic compounds from compressed CO₂, plastics, cleaning agents, anaesthetic gases and refrigerants, with isopropanol dominant and aromatic hydrocarbons accumulating specifically in incubators, and noted that water pans and culture media absorb them [395]. Hall and colleagues detected elevated aldehydes including acrolein in a laboratory adjacent to a resurfaced car park and showed that acrolein impaired mouse embryo development at low parts-per-million concentrations, with activated carbon and oxidising filtration reducing laboratory levels below outdoor air [396]. Because lipophilic volatiles partition into oil, the overlay both protects the drop in clean air and concentrates the exposure in dirty air; the same property makes stored oil vulnerable to whatever it is stored next to. The Cairo consensus on the laboratory environment and air quality, a separate document from the culture-conditions consensus cited elsewhere on this page, gives more than fifty points on site assessment, ventilation design, particulate, microbial and volatile control, cleaning practice and ongoing VOC management, framed as aspirational benchmarks for existing laboratories [397].

13.1 Principles

Every cryopreservation solution is an answer to a single problem: water inside the cell must not form ice. Mazur's two-factor hypothesis holds that cells cooled too slowly are damaged by prolonged exposure to concentrated solutes as extracellular ice forms, and cells cooled too fast retain water that freezes intracellularly; survival is highest at an intermediate rate that depends on the cell's water permeability [360, 398]. Slow freezing exploits this by dehydrating the embryo in a low concentration of permeating cryoprotectant with controlled seeding, which was how Whittingham, Leibo and Mazur first froze mouse embryos to −196 °C in 1972 [399], how Trounson and Mohr obtained the first human pregnancy from a frozen eight-cell embryo with DMSO in 1983 [400], and how Lassalle, Testart and Renard established 1,2-propanediol as the cryoprotectant for cleavage-stage human embryos in 1985 [401]. The water permeability and its activation energy that set the achievable cooling rate were measured directly for mouse oocytes and ova by Leibo and by Toner and colleagues [402, 403].

Vitrification takes the other route. Fahy and colleagues proposed in 1984 that a sufficiently concentrated cryoprotectant solution cooled fast enough passes into a glass without any ice [404], and Rall and Fahy demonstrated in 1985 that mouse embryos survive ice-free cryopreservation of both intracellular and extracellular solutions at −196 °C [405]. Rall then set out the principle that full permeation is unnecessary and can itself cause toxicity and osmotic injury, whereas partial permeation with osmotic shrinkage concentrates the cytoplasmic macromolecules and increases the likelihood of intracellular vitrification, the basis of every two-step protocol since [406]. The solution must be concentrated enough to vitrify at the cooling rate available, and the cryoprotectants must be tolerable at that concentration for the exposure time used. Toxicity is the constraint that shapes every formulation, and its molecular mechanisms remain incompletely understood [407]. Ali and Shelton tested ten cryoprotectants and found five that vitrified on plunging into liquid nitrogen (about 6.5 M for ethylene glycol, 5 M for DMSO and glycerol, 4 M for propylene glycol, 3 M for butylene glycol), whereas methanol and the polymers dextran, Ficoll, polyethylene glycol and polyvinylpyrrolidone did not vitrify at the concentrations tested; none of the five, at the highest concentration tested, remained vitreous during warming in a 25 °C water bath, and the authors ranked the order of increasing toxicity as ethylene glycol, methanol, DMSO, glycerol, propylene glycol and butylene glycol, and found that among two-component mixtures those containing ethylene glycol and glycerol were the least toxic at vitrifying concentration [408]. Those figures are cooling thresholds under that study's conditions, not intrinsic constants of the solutes: the concentration needed to form and keep a glass depends on the rest of the formulation and on the cooling and warming rates, which is why the warming rate comes next.

The rate that matters most is the warming rate. Seki and Mazur cooled mouse oocytes at 37 to 1827 °C/min and warmed at 139 to 2950 °C/min: at the highest warming rate survival exceeded 80% across cooling rates from 187 to 1827 °C/min, and at the lowest warming rate survival was near zero regardless of cooling rate, which they attributed to recrystallisation of small intracellular ice crystals during slow warming [409]. Extending the range to 69,250 °C/min cooling and 118,000 °C/min warming confirmed that survival was 70 to 85% at the highest warming rate whatever the cooling rate and low at the lowest warming rate whatever the cooling rate [410]. These are mouse oocyte data in one solution system, but they are the reason the warming solutions and the speed of the first warming step are specified as carefully as the vitrification solutions, and why a solution, carrier, exposure, cooling rate and warming rate are validated as a combination rather than separately. Chilling injury is a distinct hazard that minimum-volume carriers were designed to outrun: bovine metaphase II oocytes held at 0 °C for 5 s had reduced cleavage and after 80 s about a tenth reached the blastocyst, whereas under 1 µL of oocyte suspension on electron microscope grids plunged into nitrogen slush gave cleavage (30%) and blastocyst rates no lower than those of oocytes exposed to the ethylene glycol solution without cooling [411], and the open pulled straw brought cooling and warming rates above 20,000 °C/min with under 30 s of contact with concentrated cryoprotectant above −180 °C [412].

13.2 Permeating cryoprotectants

Ethylene glycol became the workhorse of embryo vitrification because of its low toxicity and rapid permeation. Kasai and colleagues exposed mouse morulae to 30 to 50% ethylene glycol, glycerol or propylene glycol for 20 min at 20 °C and found the highest development after ethylene glycol; their EFS solution was made by diluting ethylene glycol to 40% (v/v) with phosphate-buffered PB1 medium containing 30% Ficoll and 0.5 M sucrose, so that the final working solution contained about 18% Ficoll and 0.3 M sucrose; it was tolerated for a 5 min exposure, did not crystallise on cooling or warming, and gave 97 to 98% development in culture and 51% live young after transfer, with no appreciable loss of viability [413]. Vanderzwalmen and colleagues vitrified human morulae and blastocysts in the same final composition, 40% ethylene glycol with 18% Ficoll and 0.3 M sucrose, showed that survival fell as the blastocoel expanded, and raised survival from 20% to 71% and implantation from 1.4% to 12% by collapsing the blastocoel with a needle before vitrification [414]. Whether a given protocol collapses expanded blastocysts before vitrification, and by what method, is specified by that protocol.

A widely used published formulation is the Cryotop system of Kuwayama and colleagues, which uses an equilibration solution of 7.5% ethylene glycol and 7.5% DMSO followed by a vitrification solution of 15% ethylene glycol, 15% DMSO and 0.5 M sucrose, loaded in a volume below 0.1 µL on a polypropylene strip. In its first human application, 58 of 64 vitrified oocytes (91%) were morphologically normal after warming, 52 fertilised after ICSI, and 32 developed to the blastocyst stage, which is 50% of the 64 vitrified oocytes and about 62% of the 52 fertilised; the five blastocysts examined were diploid by the fluorescence in situ hybridisation panel used, and 29 transfers gave 12 initial pregnancies and seven healthy babies at the time of report [415, 416]. The two-step exposure is a toxicity strategy: the equilibration solution lets the cell dehydrate and take up cryoprotectant at a concentration tolerated for the exposure used, so that the brief exposure to the full vitrification concentration, under a minute in the published protocol, raises intracellular cryoprotectant to a vitrifiable level without the exposure time that would make it toxic [407, 416]. These are the concentrations of the published Cryotop method; current commercial kits from the same and other manufacturers are formulated differently, and one product line's 2021 instructions for use list trehalose and hydroxypropyl cellulose rather than sucrose and protein [417]. Glycerol remains the cryoprotectant for sperm (Section 10.4) and, with ethylene glycol, in some embryo solutions, and propanediol persists in slow-freezing protocols. The retention of DMSO in the ethylene glycol mixture has mouse support: vitrification of mouse oocytes without DMSO gave lower survival at warming and inferior blastocyst development after parthenogenetic activation than with it, with no difference in maternal imprinted gene expression [418].

The osmotic behaviour of the oocyte during exposure is as important as the chemistry. Paynter and colleagues measured mature human oocyte volume in propanediol with and without sucrose: in cryoprotectant alone the oocyte shrank and re-expanded to 75 to 84% of its volume, whereas with 0.2 or 0.3 M sucrose it shrank continuously to 67% or 55%, and small changes in exposure time produced drastic changes in hydration [419]. A two-parameter permeability model with volumetric observation of research-donated human oocytes and zygotes later showed that, for the ethylene glycol, DMSO and sucrose solutions modelled, the minimum of the shrink-swell curve is reached within about 60 s, the point at which the authors consider water ejection complete and intra- and extracellular solute concentrations similar, and argued from this that prolonging the usual 8 to 15 min exposure does not further improve the glass-forming tendency of the cytosol; a two-minute protocol of two one-minute exposures gave post-warming survival and resumed cytokinesis, which the authors themselves distinguished from developmental competence, calling for further studies before clinical use [420]. The result belongs to that model and that solution system: a minimum on a volume curve does not mean that cryoprotectant transport has stopped or that every intracellular solute has equilibrated, which is why exposure windows remain part of each validated system rather than a universal constant. The first human birth from a vitrified oocyte used a DMSO-free solution of 40% ethylene glycol with 0.6 M sucrose in 1 to 2 µL open pulled straws, from which 11 of 17 oocytes survived, five zygotes were obtained after ICSI, and one embryo, assessed by preimplantation diagnosis as normal for the eight chromosomes tested (X, 13, 14, 15, 16, 18, 21 and 22) and donated to a 47-year-old recipient, gave a healthy girl at 37 weeks [421], and a protein-free solution of 25% ethylene glycol with 35% dextran or Ficoll gave mouse embryo survival no different from controls in a straw-in-straw closed package, the first use of that package throughout a rapid-cooling procedure [422]. Adherence to the specified exposure times and temperatures is therefore part of the formulation, not merely a procedural nicety, and the windows themselves differ between validated systems.

13.3 Non-permeating solutes: sucrose, trehalose and macromolecules

A non-permeating sugar, sucrose in the published Cryotop and EFS solutions (0.5 M and 0.3 M final respectively) and trehalose in at least one commercial kit as documented in its 2021 instructions for use [413, 415, 417], draws water out of the cell osmotically, reducing the intracellular water that must be vitrified, and lowers the concentration of permeating cryoprotectant required for the solution as a whole to form a glass. Shaw and colleagues showed by differential scanning calorimetry that 59% (w/w) ethylene glycol in saline forms a stable glass, that adding sucrose at 1 m in the authors' notation allowed vitrification with 31% ethylene glycol, and that the total solute concentration required could be estimated from the sum of the components [423]. Trehalose serves the same osmotic purpose extracellularly; introduced inside cells at 0.2 M, which normally requires a pore-forming agent, it markedly improved survival of cryopreserved mammalian cells [424]. A macromolecule, Ficoll in EFS, albumin or serum substitute in the Cryotop base, and a synthetic polymer such as the hydroxypropyl cellulose listed in that 2021 document, is added to raise viscosity and reduce ice nucleation; the physical chemistry is polymer-dependent, since on a weight-for-weight substitution Ficoll and dextran vitrified less readily than the ethylene glycol they replaced whereas polyvinylpyrrolidone was comparable, and glass formation in a solution is a necessary condition for survival, not a guarantee of it, because membrane transport, toxicity and osmotic stress decide the rest [413, 423]. The base of most contemporary clinical vitrification kits is a zwitterion-buffered handling medium (Section 8) with albumin, a serum substitute or a synthetic macromolecule, because the entire procedure is performed on the bench; published systems have used other bases, the EFS solutions being made up in phosphate-buffered PB1 [413]. Antioxidants have been proposed as a further composition variable: acetyl-L-carnitine, N-acetylcysteine and alpha-lipoic acid in the vitrification or warming solutions increased mouse blastocyst developmental potential, outgrowth and fetal development after transfer [425].

13.4 Warming and dilution solutions

Warming reverses the loading. The cell is placed at high warming rate into a thawing solution containing sucrose at a concentration above that of the vitrification solution's non-permeating component (commonly 1.0 M), so that the cryoprotectant leaves the cell down its gradient while the sucrose holds cell volume and prevents osmotic swelling; the cell is then stepped through descending sucrose (for example 0.5 M) into cryoprotectant-free handling medium and finally culture medium [413, 416]. In Kasai's original protocol, straws were warmed in 20 °C water and the embryos diluted directly into a separate 0.5 M sucrose PB1 solution; the EFS working solution itself contains about 0.3 M sucrose (Section 13.2), which acts as the osmotic buffer during loading [413]. The warming solutions are as temperature-critical as the vitrification solutions, both because the first step must be fast enough to outrun recrystallisation (Section 13.1) and because cryoprotectant permeability and toxicity both depend on temperature [407, 410]; in a matched retrospective comparison of 366 pairs of frozen transfers with one commercial kit, performing every warming step at 37 °C with shortened wash times, instead of dropping to room temperature after the first step as the instructions specified, gave higher clinical pregnancy and implantation for blastocysts, a reminder that the temperature and duration of the later steps are themselves variables [426]. The descending sucrose steps are now omitted in some laboratories, and the evidence for doing so is observational and tied to each laboratory's solutions, device and temperatures: a retrospective analysis of 3167 blastocyst transfers found survival above 99% with either protocol and higher clinical pregnancy (72.1% against 67.2%), ongoing pregnancy and live birth (63.7% against 57.0%) after one-step warming, with re-expansion slower at 0 to 2 h but higher by 4 h [427]; a time-interrupted comparison of 752 frozen transfers using a one-minute immersion in the thaw solution against the manufacturer's 14-minute protocol found 100% survival in both, no adjusted difference in implantation, clinical pregnancy, miscarriage, live birth or birthweight, and a slightly higher odds of multiple gestation [428]; a three-phase validation of a single-step protocol (risk analysis, 246 preclinical blastocysts, then 1925 cycles against 1744 historical multi-step cycles) found higher survival and intactness, no significant difference in ongoing pregnancy in any subgroup, and some early-pregnancy endpoints, such as clinical pregnancy in non-PGT day 5 transfers (40.5% against 35.7%), that favoured the multi-step protocol [429]; and a series of 1266 infants born after one-step warming for 1 min in 1 M sucrose at 37 °C reported live birth, miscarriage, gestational age (37.6 weeks in both groups) and birthweight (3267 against 3252 g) no different from 1323 infants born after multi-step warming [430]. Operational simplicity, embryo survival, live birth and neonatal outcome are separate questions, and the one-step data so far answer them for the specific systems studied. What the cell needs after warming is also known from the hamster: vitrified and warmed two-cell embryos had reduced Na⁺/H⁺ antiporter and HCO₃⁻/Cl⁻ exchanger activity for about 4 h, with elevated intracellular pH and impaired oxidative metabolism, normal only by 6 h, which is why the post-warming medium and its buffer matter (Section 8) [431]. In oocyte slow freezing, raising the sucrose of the propanediol solution from 0.2 to 0.3 M raised post-thaw survival from 60% to 82%, after an earlier gain from doubling it from 0.1 M, the modification behind the Table 5 row [432].

13.5 Clinical performance and the open-versus-closed question

Vitrification has displaced slow freezing on the strength of survival. The systematic review commissioned for World Health Organization guidance found that vitrification gave higher ongoing clinical pregnancy per cycle than slow freezing for oocytes (RR 2.81 in the one RCT), higher clinical pregnancy per cycle for embryos (RR 1.89, borderline), and, across seven RCTs and 3615 embryos, a cryosurvival rate 1.59 times that of slow freezing, and it recommended that laboratories still slow-freezing consider transitioning [433]. Cobo and colleagues' randomised trial of 600 donation recipients established the non-inferiority of vitrified to fresh donor oocytes, with ongoing pregnancy of 43.7% versus 41.7% [434]. Edgar and Gook's 2012 critical appraisal, written while the transition was under way, concluded that vitrification was already the method of choice for metaphase II oocytes, that early cleavage-stage embryos could then be cryopreserved with equal success by optimised slow cooling or vitrification, and that blastocyst cryopreservation might be more consistently achieved with vitrification although optimal slow cooling could produce similar results [435]; the current position is the 2026 ESHRE recommendation that for oocytes, pronuclear and cleavage-stage embryos and blastocysts vitrification is the more effective method and is recommended, with slow freezing remaining the method of choice for ejaculated sperm [89]; their earlier review documents the sucrose concentration and sodium-depletion modifications of oocyte slow freezing that preceded vitrification's adoption [436]. Because the highest cooling rates come from direct contact between the solution and liquid nitrogen, the choice between open and closed carriers is a trade between rate and a theoretical contamination risk. The one documented transmission through liquid nitrogen involved storage bags, not gamete carriers: six patients developed acute hepatitis B after transplantation of bone marrow or stem cells stored in one tank, with HBV DNA matching the infected patients' sequences recovered from the liquid nitrogen and leaking bags from the index patient implicated [437]. On the rate side, the series that compared vitrification with slow cooling in more than 13,000 embryos included a separate comparison in human blastocysts of closed CryoTip against open Cryotop, with survival of 93% against 97%, pregnancy 51% against 59% and delivery 48% against 51%, none significantly different [438], and a randomised trial of 737 patients and 4029 vitrified oocytes found closed high-security straws against open Cryotop survival of 70.3% against 73.3%, fertilization 70.8% against 74.9%, pregnancy per transfer 32.0% against 31.8% and miscarriage 22.1% against 21.5%, none significantly different, so that for at least one closed device the trade has shown no measurable clinical cost [439]; Bielanski and Vajta reviewed the evidence on contamination through liquid nitrogen and cryobanking [440], and Vajta, Rienzi and Ubaldi argued that closed systems are not entirely closed, that transmission through open systems has not been documented, which is not the same as a demonstrated absence of risk, and that rational measures can lower the theoretical risk without sacrificing efficiency [441]. The solution chemistry can be the same in either carrier; what changes is the cooling and warming rate available, and therefore the margin within which the validated cryoprotectant concentrations and exposure times were established.

14.1 Ammonium

Sources. Ammonium enters the drop from two directions. Free glutamine breaks down spontaneously at 37 °C, cyclising to pyrrolidone carboxylic acid with the loss of its side-chain amide nitrogen as ammonia, a deamidation with first-order kinetics that accelerate with pH [170, 171]; glutamine is by far the least stable of the amino acids in solution, so its concentration and form, free or dipeptide, set most of the ammonium a medium generates without an embryo, and the embryo itself releases ammonium as it metabolises amino acids [172]. Lane and Gardner found that ammonium accumulates linearly with time when amino-acid-containing media are simply incubated at 37 °C [7]. In sheep, blastocysts produced significantly more ammonium than mouse blastocysts, so the embryonic contribution is species- and stage-dependent [69]. The measurements on clinical products came later: all fifteen protein-supplemented ready-to-use media tested accumulated ammonium over six weeks of refrigerated storage, and ready-to-use media and the protein supplements themselves accumulated it over four days at 37 °C, which makes the protein supplement a third source and the refrigerator, not only the incubator, a place where ammonium forms [239]. Gamete co-incubation is a fourth: spent medium after 16 to 18 h of conventional co-incubation contained more ammonium (about 27 µM) than after 2 h of brief co-incubation (about 21 µM) or medium incubated without gametes (about 19 µM), in a small series measured with an assay whose range was 7 to 286 µM (Section 9.2) [335].

Effects. The 1993, 1994 and 1996 papers established that ammonium reduces mouse blastocyst development and cell number, retards fetal growth and induces exencephaly in a time- and concentration-dependent manner [50, 51, 155]; an independent replication with 0.3 to 0.6 mM ammonium chloride found no consistent effect on fetal or placental weight but confirmed the malformations at low frequency (5 abnormal fetuses of 217 against 0 of 363 controls, three exencephalic), and concluded that the risk may be lower than first suggested [442]. Ammonium generated from amino acid breakdown significantly reduced blastocyst development of CF1 zygotes [155]. The 2003 study then showed that ammonium reduces blastocyst cell number and inner cell mass, increases apoptosis, perturbs metabolism, impairs the embryo's ability to regulate intracellular pH, and alters expression of the imprinted gene H19, while blastocyst rate and morphology remain normal, so that the damage is invisible to morphological grading; after transfer, pregnancy rates were reduced and fetuses exposed to 300 µM ammonium were developmentally retarded by 1.5 days at day 15 [7]. Zander, Thompson and Lane located the sensitive window: exposure to 300 µM ammonium before compaction, at either the zygote to two-cell or the two-cell to eight-cell stage, reduced blastocyst and inner cell mass cell numbers, increased apoptosis, perturbed glucose transporter expression and uptake, and reduced fetal number and maturity after transfer, whereas the post-compaction embryo was largely resistant [443]. In human embryos, a gradient of ammonium that mimicked the spontaneous deamination of Eagle's amino acids with 1 mM glutamine impaired development before compaction, inhibited pyruvate consumption after only 24 and 48 h, and altered the blastocyst transcriptome in both human and mouse, with over-representation of genes for metabolism, cell growth and maintenance, transcription, cell communication, transport and development [172]. Clinically, Virant-Klun and colleagues measured ammonium in spent sequential media from 100 cycles: it was increased in 62% of cycles and was negatively correlated with blastocyst development after conventional IVF, though not after ICSI [444]; in 254 cleavage-stage embryos cultured individually in 30 µL of a current blastocyst medium, embryo-derived ammonium above a cut-off of 16.1 µmol/L separated blastocyst formation of 59.0% from 80.5% and good-quality blastocyst rates of 3.4% from 21.0%, with higher Bax and lower Oct4 expression in the high-ammonium group, so that ammonium produced by the embryo may mark poor development as much as cause it [445]. The embryo's own defence has been identified: in the mouse blastocyst, glutamine synthetase sequesters ammonium into glutamine, inhibiting the enzyme raised ammonium production, and the sequestration worked at 5% oxygen but reversed at 20%, where glutamine was consumed instead, so atmospheric oxygen disables the embryo's handling of ammonium [446]. The chemistry behind the pHi effect is that ammonium (NH₄⁺) exists in equilibrium with ammonia (NH₃), the uncharged weak base, which crosses membranes and is protonated again inside the cell; with the embryo's limited capacity to regulate pHi before compaction, this is one proposed route for the disturbance Lane and Gardner measured, although the study established the effect rather than the mechanism [7, 106].

Mitigation. Four measures follow from the data. Supply glutamine as a stable dipeptide (Section 5.4) [173, 174]. Keep the time any amino-acid-containing medium spends at 37 °C within what the product was validated for: in the mouse work, blastocyst formation and cell number rose when embryos were moved to fresh medium after 48 to 72 h [50, 51], Gardner's 2013 paper recommends renewal at intervals for this reason [172], and in clinical sequential systems the change to the second medium at the manufacturer's specified point, day 3 in most protocols, is where that renewal happens. The interval is product-specific; a universal 48 h rule is not what the evidence supports, and modern formulations differ in glutamine source and validated exposure duration. Equilibrate dishes within the window the manufacturer specifies, which for many current products includes overnight preparation, since ammonium begins to accumulate in the incubator before the embryo arrives and glutamine-containing media generate more of it than dipeptide media over the same hours [7]. And recognise that single-step media formulated with dipeptides can end culture with lower ammonium than a sequential blastocyst medium, as Hardarson and colleagues measured, which is one of the legitimate arguments for uninterrupted culture [175]. Renewal is not the only way to remove ammonium, and the alternative shows what renewal costs: an in situ glutamate dehydrogenase system that converted ammonium to glutamate in the drop removed 0.30 mM within 3 h and increased mouse blastocyst cell number, implantation, fetal development and fetal weight, whereas renewing the medium increased cell number but not post-implantation development, which is direct evidence that renewal takes the conditioning with the ammonium (Section 14.4) [447]. The trade-off in single-step culture is that whatever ammonium does accumulate is present throughout the sensitive pre-compaction window [443].

14.2 pH drift: CO₂ loss, equilibration and gas set points

Outside the incubator a bicarbonate-buffered medium loses CO₂ and its pH rises within minutes; the rate depends on surface area, volume, the oil overlay and the temperature, and it is the reason handling media exist (Section 8) [8]. Inside the incubator the reverse process, equilibration, takes hours, and a dish used before it has equilibrated exposes the embryo to alkaline medium at the stage when it can least correct it [105, 107]. Because manufacturers formulate bicarbonate to reach their target pH at a stated CO₂ percentage, two media in the same incubator can sit at different pH, and a laboratory that changes medium, CO₂ set point, altitude or incubator must re-measure pH in the medium under its own conditions rather than assume it [88, 101, 102]. The embryo's own metabolism is a further variable: the blastocyst releases lactate through aerobic glycolysis [151], although whether embryo-derived lactate measurably shifts the pH of a bicarbonate-buffered drop has not been shown directly, and the arguments for adequate drop volume and appropriate embryo density rest on the conditioning and osmolality data rather than on pH [268]. Alkalosis and acidosis both disrupt mitochondrial and cytoskeletal organisation in the early embryo [109], and the tolerance is narrow: a fall in mouse intracellular pH of under 0.2 units produced by a weak acid reduced blastocyst cell number and inner cell mass and increased apoptosis with morphology unchanged, exposure during the first cleavage alone reduced fetal weight and crown-rump length, and continuous exposure also reduced implantation [448]; and the pre-compaction human embryo's capacity to correct an acid load is at best limited: Dale and colleagues found no recovery from acidosis before the blastocyst stage, whereas Phillips and colleagues observed Na⁺-dependent recovery in two- to eight-cell embryos, a disagreement that has not been resolved [106, 107].

14.3 Osmolality drift: evaporation and concentration

Water leaves the drop through the oil and into the incubator atmosphere, and every solute concentrates as it goes. Over five days in non-humidified benchtop incubators the osmolality of drops under oil rose linearly by up to 20 mOsm/kg, more in smaller drops, and did not rise at all in a humidified incubator [9]. Humidity, oil volume, medium type and dish design set the rate, protein does not, and media that start near the top of the osmolality range are the ones that reach hypertonic levels [391]. Preparation adds to the load before the embryo arrives, by up to 40 mOsm/kg within 5 min under adverse conditions [98]. The biological threshold is not far away: mouse development fell above 310 mOsm/kg [98], raised osmolality is itself a cause of the two-cell block in mouse embryos [90, 96], and in the human randomised trial dry culture significantly reduced clinical and ongoing pregnancy relative to humidified culture (OR 0.57 and 0.54) [392], with a second clinical dataset finding higher miscarriage from 20 µL drops that reached 308 mOsm in a dry incubator than from drops held at 285 to 290 mOsm by medium in the outer wells (Section 12.4) [393]; against that, six days of osmolality rise under light oil in a dry incubator did not change human blastocyst formation, quality or cell count in one study [389], so the human threshold is not settled. Because the early embryo regulates its volume with glycine and other organic osmolytes drawn from the medium [91, 92], the amino acid content of a medium is part of its defence against its own evaporation, and a medium without those osmolytes is more vulnerable to the same loss of water. Tarahomi and colleagues found that sham culture for three days significantly changed the osmolality of the fifteen media they tested, in addition to changing 13 of 37 measured components [136].

14.4 Substrate depletion, metabolite accumulation and conditioning

The embryo is a small consumer and a small producer, but the drop is small too. A human embryo consumes on the order of 30 to 40 pmol of pyruvate per hour before compaction and 8 to 24 pmol of glucose per hour, rising at the blastocyst stage [148], and blastocysts release lactate [151] and alanine [168] while depleting leucine and other amino acids [168, 169]. An illustrative calculation shows the scale. A 20 µL drop of HTF-type medium at 0.33 mM pyruvate holds about 6.6 nmol; a single embryo taking up about 35 pmol per hour (within Hardy's 28 to 40 pmol range) would remove roughly 0.8 nmol per day, about a quarter of the initial amount over 48 h, so that with five embryos in one drop the demand at a constant rate over two days, about 8.4 nmol, would exceed the 6.6 nmol present, which shows only that the constant-rate assumption cannot hold [33, 148]. Uptake falls as the concentration falls and varies with stage and formulation, so this is an illustration of scale rather than a validated depletion forecast, an upper bound or a density prescription, and measured spent-medium data are needed for any real system; the practical reading is that pyruvate supply is worth checking in small-volume group culture. Glucose in a 0.5 mM cleavage medium (10 nmol in 20 µL) is drawn down more slowly at the pre-compaction uptake rates measured by Hardy, by a few percent per day per embryo, and the 3.15 mM of a blastocyst-stage medium by a smaller fraction still even at the blastocyst uptake of 24 pmol per hour [55, 148]. The point stands in a weaker form: substrate supply is a real consideration in the choice of drop volume, embryo density and renewal interval for pyruvate in small-volume group culture, and not only a question of products accumulating. What accumulates is not only waste. Group culture in small volumes can improve development in particular systems, because embryos condition the medium with autocrine and paracrine factors [198, 268], which in the mouse act as survival signals as early as the two-cell stage [199], and in two human series embryos cultured in groups compacted, blastulated and scored better than embryos cultured alone [269, 270]; the human evidence is mixed, however, with one randomised comparison favouring individual culture for blastocyst yield, and the effect depends on density, volume, dish, medium, oxygen and the endpoint chosen (Section 7.3). Renewal therefore removes both the ammonium and the conditioning, and the interval and volume are a compromise between the two, which is why the same laboratory may reasonably renew a sequential system at the product's specified change point and leave a single-step medium undisturbed under time-lapse [172, 263]. The composition of spent medium is itself informative: pyruvate uptake [148], glucose uptake [156] and amino acid turnover [168, 169] have each been associated with developmental competence in research datasets, and amino acid turnover tracked DNA damage in the blastocyst [162], although none is a validated clinical selection test (Section 5.3).

14.5 Protein, lipids and peroxides

Albumin is a component whose character can change during use. It binds the peroxides, aldehydes and alkenals that peroxidised oil releases and, in Otsuki's tracer experiments, increased the passage of lipophilic agents into the zona pellucida [232, 387]; it binds the transition metals that some supplements carry and that catalyse further oxidation [223]; and it carries the octanoate stabiliser whose effects on development and fetal weight were shown in mice [227]. Under atmospheric oxygen the pro-oxidant load of the supplement interacts with the oxygen concentration to reduce blastocyst development, an interaction that reduced oxygen mitigates [126, 223]. Group culture masks oil-peroxide toxicity in the mouse assay; protein supplementation with HSA supported blastocyst development in the extended assay but had minimal effect on its sensitivity [278]. The albumin content of a medium is itself among the components that change on storage and during sham culture in the analysis by Tarahomi and colleagues [136].

14.6 Light and photoproducts

Media are photochemically active. The photosensitiser triad was identified in 1974 with near-ultraviolet light: irradiated tissue culture medium killed cells added afterwards, riboflavin, tryptophan and tyrosine were the photoactive components, omitting riboflavin almost abolished the killing, and riboflavin with either amino acid, but none singly, reproduced it [449]. Riboflavin with tryptophan or tyrosine generates hydrogen peroxide under daylight fluorescent light, and about 40% of the toxicity of irradiated culture medium to human cells was attributable to that peroxide [181, 286]; HEPES amplifies the reaction, and hydrogen peroxide is again the principal cytotoxic product [182]. Zygotes are directly affected: cool white fluorescent light, rich in short wavelengths, generated more reactive oxygen species in mouse and hamster zygotes than warm white light, and blastocysts from zygotes shielded from light developed best to term, followed by those exposed to warm white and then cool white light [450]. Ottosen and colleagues measured the spectral composition and intensity of light during IVF procedures in Danish clinics of that time and calculated that normal handling delivered stressing doses unless filters were used, that 95% of the radiation reaching the embryo came from microscopes, and that ambient room lighting contributed little, from which they concluded that darkened laboratories were not justified but microscope filtering was; the proportions depend on the light sources, optics and exposure times of the laboratory concerned [287]. The one clinical dataset is a non-randomised comparison of light-protected against conventional handling, with red filters on laboratory lamps and ultraviolet or infrared filters on microscopes, in which protection was associated with higher ICSI fertilization, blastocyst rates higher by 20.9 percentage points after IVF and 38.6 after ICSI, and more clinical pregnancies per ICSI day 5 transfer; the whole-laboratory design cannot separate microscope from ambient filtering and the study was not randomised [451]. Media formulated without riboflavin and stored in the dark remove that photosensitiser from the medium, though not the direct effects of light on the zygote shown by Takenaka and colleagues; handling media exposed under the microscope, and any medium left on a lit bench, are where photoproducts form.

14.7 Storage instability: pyruvate, glutamine, temperature excursions and expiry

Media change before they are opened. Wales and Whittingham reported in 1971 that sodium pyruvate decomposes in culture media stored at 5 °C, with effects on mouse embryo development [224]; free glutamine decomposes with a half-life measured in days to weeks at 37 °C and more slowly, but measurably, at refrigerator temperature [170, 171]. Tarahomi and colleagues tested fifteen complete media on arrival, after three days of sham culture, just before expiry, and after sham culture at expiry, and found that storage produced significant changes in 17 of 37 components, including magnesium, chloride, phosphate, albumin, total protein and twelve amino acids among them glutamine, asparagine and tryptophan; they also found that two media contained 50% D-lactate, which the embryo cannot use, and in a secondary analysis detected human liver enzymes in more than half of the complete media, a finding the authors called surprising and did not explain, and for which the plasma proteins that accompany plasma-derived albumin (Section 6.2) are the most plausible source [136]. Dyrlund and colleagues found that the non-declared protein profile of albumin-supplemented media varies from batch to batch [226]. The Cairo consensus devotes a section to cold chain and storage [88], and the chemistry above is the reason its recommendations matter: receipt at the manufacturer's specified temperature, storage at 2 to 8 °C protected from light, use within the shelf life and the stated post-opening interval, and avoidance of repeated warming and cooling, each cycle of which advances the temperature-dependent decomposition of pyruvate and glutamine and, for oil, peroxidation [171, 224, 383]. Whether the storage chemistry reaches the patient is a separate question, and the largest clinical dataset says that within shelf life it did not: in 9680 IVF and ICSI cycles from eight clinics using one single-step medium in time-lapse incubators, medium age at use, from 38 to 365 days within a one-year labelled shelf life, was not associated with blastocyst or usable blastocyst rates, cumulative clinical or ongoing pregnancy or live birth, or with birthweight or preterm birth among 1070 singletons, with the caveats that a single formulation was studied and post-opening handling was not captured [452]. A laboratory that logs receipt temperature, opening date and in-use time for every bottle is not being bureaucratic; it is tracking a chemical reaction, and the tracking is what allows a finding like that one to be trusted for the products it covers.

14.8 Leachables, volatiles and microbial contamination

The medium also acquires what its container and surroundings give it. Gatimel and colleagues found no bisphenol A in two single-step media, two sequential media or three protein supplements, and none leaching from 33 plastic consumables at levels above those in normal serum and follicular fluid, but the plastic of the three oocyte denudation strippers tested did contain bisphenol A, two being made of polycarbonate [453]. Glassware belongs on the list: zinc eluted from glassware delayed mouse development, disrupted chromosome segregation and zygotic gene activation, reduced blastocyst formation and raised birthweight by 18% after transfer, affected bovine and human embryos as well, and was prevented by chelating agents [454]. Pyrogens can arrive on consumables too: a human cell-based pyrogen test (a monocyte activation test reading interleukin-6, which detects pyrogens generally rather than endotoxin specifically and expresses its result in endotoxin-equivalent units) detected pyrogenic activity in 7 of 20 ART consumables, the cells being cultured in medium in direct contact with each product rather than in a water extract; some product types were tested as several pieces immersed together in one flask, and for those the per-product figures, up to about 1.1 endotoxin-equivalent units, are upper estimates that assume the whole signal sat on a single piece, with the same signal spread evenly across the pieces giving values several-fold lower, whereas the products tested piece by piece gave directly measured values of about 0.01 to 0.02 units. In the same manufacturing study low-temperature ozone and hydrogen peroxide gas sterilisation reduced the activity below detection, with the residual peroxide reduced by drying to a level that did not affect sperm viability or embryo development in the authors' tests; that is a demonstration of what a manufacturer can achieve, not an instruction to reprocess clinical consumables [455]. Nijs and colleagues screened 36 products of 72 brands with the human sperm survival test over four years and found 13 reprotoxic, including eight brands of surgical glove, an oocyte retrieval needle, an embryo transfer catheter, tubing, a Pasteur pipette and petri dish, and a specimen container cover [456]; Ackerman and colleagues had found the same pattern in 1985 for instruments sterilised with glutaraldehyde or ethylene oxide, gloves and catheter liners [457]. Volatile organic compounds from the incubator atmosphere partition into oil and medium (Section 12.5) [395, 396], and oils differ in the volatiles they carry and transfer [382]. Microbial contamination is rare but not random: in the one series reviewed here that reports it in detail, it occurred only after conventional IVF and never after ICSI, and most isolates were resistant to the penicillin and streptomycin in the medium [185]. Each of these is a reason that the FDA definition of reproductive media extends beyond the culture medium to its water, rinsing solutions, supplements and oil [85], that other contact devices carry their own classifications, and that the laboratory's qualification of contact materials is part of its media management (Section 16).

15.1 Birthweight and early growth

The clearest human evidence that medium composition reaches beyond the culture period comes from the Maastricht group. Dumoulin and colleagues allocated 826 first IVF cycles alternately, a pseudo-randomised design, between two commercial sequential systems and found that singletons from the Vitrolife group weighed 3453 g against 3208 g in the Cook group, with the difference persisting after adjustment for gestational age, sex and other covariates [10]. The cohort was defined in the group's 2012 paper, which extended the effect beyond first cycles and found a lower adjusted birthweight for fresh-transfer singletons from the Cook medium (difference 112 g) with more low-birthweight infants, alongside frozen-transfer and twin analyses [458]. In the extended series of 1432 alternately allocated cycles, 265 of the 294 eligible singletons were followed (29 lost to follow-up), and those from the Vitrolife group remained heavier than those from the Cook group throughout the first two years of life [459]; at 9 years, 136 of the 294 children were examined and the difference attributable to medium was 1.58 kg in weight and 0.84 kg/m² in BMI, with higher waist circumference and truncal adiposity in the Vitrolife group and no difference in blood pressure, fasting metabolic markers or endothelial function [460]. Growth diverges earlier than birth: in a prospective cohort with serial three-dimensional ultrasound at 7, 9 and 11 weeks, embryos cultured in one single-step medium grew faster than those cultured in another sequential product and than naturally conceived embryos, most markedly in males [461]. In the multicentre double-blind randomised trial that followed, 836 couples were allocated to HTF or G5 for all cycles within a year: cumulative live birth over all cycles within the year was 44.1% with G5 and 37.9% with HTF (RR 1.2, 95% CI 0.99 to 1.37, not significant in the trial's own analysis; the 2026 Cochrane review grades this as moderate-certainty evidence of a likely increase, Section 7.2), but utilisable embryos, implantation and clinical pregnancy were higher with G5, and birthweight was 158 g lower in the G5 group, with more preterm singletons [11]. Zandstra and colleagues reviewed eleven human studies, five of which found significant birthweight differences between media, and noted that because nearly every study compared a different pair of products no conventional meta-analysis was possible [462]. Not every randomised comparison of the same product pair found a difference: 449 cycles randomised between the Cook and Vitrolife media gave birthweight z-scores of −0.19 against 0.08, not significant, and a retrospective series of 2518 cycles across three media likewise found none [463], and a cohort of 6352 first cycles found live birth, pregnancy and miscarriage comparable between the same two brands, with the weight excess this time on the Cook side (macrosomia 7.51% against 5.39%) and more placenta previa with Vitrolife, so that the direction of a product-pair difference is not stable across populations [464]. A UK national analysis merging a survey of media, oxygen and incubator type with outcome data for 2011 to 2013 found live birth to differ between eight media systems but no culture factor associated with singleton birthweight after adjustment, with a strong clinic-site effect that may have masked culture conditions [465]. The protein-source comparison by Zhu and colleagues (Section 6.6) adds a within-manufacturer contrast [253]. A small cohort of 73 singletons from an earlier randomised comparison that had been stopped early, followed to five years across many developmental variables, found no effect of medium on birthweight, malformation, growth or medical concerns, but a difference in developmental problems between the two media, an exploratory finding the authors presented cautiously [466]. The newer syntheses do not settle the question. A 2024 systematic review with a network meta-analysis of 18 media studies found no significant effect of culture medium on birthweight or preterm birth, and no effect of oxygen tension on birthweight or gestational age in the three studies available [13], whereas a retrospective cohort of 23,403 fresh transfer cycles in one Chinese centre, with media used in different periods, found higher birthweight z-scores and more large-for-gestational-age infants after G5 and HTF than after Cook, with medium ranking as a factor of moderate importance among those examined [467]. The human evidence is therefore mixed: some randomised comparisons of particular product pairs have found differences and others have not, pooled analyses have not, and neither a difference in mean birthweight nor its absence is by itself evidence of harm or safety.

15.2 Imprinting and the animal models

The mechanism most often proposed is epigenetic. Doherty and colleagues showed that culture of mouse embryos in Whitten's medium caused aberrant expression of the normally silent paternal H19 allele with loss of methylation at its imprinting control region, whereas KSOM with amino acids largely preserved imprinted expression and methylation, and that Snrpn imprinting was maintained in both, so that culture affects imprinted genes selectively and medium-dependently [468]. Mann and colleagues found that loss of H19 imprinting after culture persisted to mid-gestation, with activation of the silent alleles of H19, Ascl2, Snrpn, Peg3 and Xist in placental tissues while the embryo proper largely preserved imprinting, suggesting that trophectoderm-derived tissues cannot restore lost imprints [469]. In a side-by-side comparison of five commercial media systems (KSOM^AA, Global, HTF, P-1/Multiblast and G1/G2 v5 PLUS) against in vivo embryos and Whitten's medium, all commercial systems showed loss of imprinted methylation relative to in vivo embryos, some preserved it better than others, all lost imprinted H19 expression, and superovulation added to the perturbation, leading those authors to recommend, from mouse data, that time in culture and the number of procedures be minimised [12]. Lane and Gardner's ammonium study identified altered H19 expression as one consequence of ammonium exposure [7], connecting a specific in-use change in the medium to the imprinting findings. In ruminants, serum-containing culture produced the large offspring syndrome [73, 74], and in the mouse the link between serum, imprinting and fetal growth was made in a single experiment: only a third of blastocysts cultured with fetal calf serum gave viable day 14 fetuses, and those were lighter, with reduced H19 and Igf2 expression, gain of methylation at the H19 imprinting control region and increased Grb10 [470]; halving insulin and the branched-chain amino acids in a defined medium is the composition-specific counterpart, with a postnatal cardiometabolic phenotype and no signal at the blastocyst (Section 5.10) [209]. The human epigenetic tests of the birthweight cohorts have been reassuring. Placentas from the HTF against G5 trial (43 and 54) showed no difference in mean methylation of imprinted-gene regions between media [471]; cord blood from neonates of the same trial showed no systematic CpG, regional, imprinted or birthweight-associated differences and no epigenetic gestational-age acceleration [472]; saliva from 106 nine-year-olds of the Maastricht comparison showed no significantly differentially methylated site, gene, promoter or island across 659,708 CpG sites between the two media [473]; and buccal cells from children of a randomised comparison of two single-step media showed differences between ART and naturally conceived children in imprinted genes and transposable elements but none between the two media [474], which separates an ART-versus-natural signal from any medium signal.

15.3 Where this leaves the laboratory

The ESHRE working group's position, set out by Sunde and colleagues, is that the environment of the early embryo can reprogramme growth trajectory, birthweight, childhood growth and long-term disease risk, that the mechanism is likely epigenetic, that commercial formulations have proliferated without full documentation or disclosure, and that there is a strong case for full transparency of composition [475]. Tarahomi and colleagues reached the same conclusion from their compositional survey [136]. For the laboratory director the practical implications are three. This page's recommendations are that the choice of medium be documented and justified like any other validated process, since it is a decision with possible consequences beyond the culture period; that a change of medium or protein product be treated as a change of exposure and validated in proportion to its scale, with outcome endpoints where the change is a new formulation (Section 16.3); and that the in-use changes of Section 14, ammonium above all, be recognised as the part of that exposure the laboratory itself controls. What the mouse imprinting studies contribute is mechanism and plausibility; they do not by themselves establish medium-specific imprinting damage in human offspring.

16.1 The mouse embryo assay and its limits

The mouse embryo assay has been the release test for media since the early days of clinical IVF. Ackerman and colleagues used it in 1985 to identify toxic instruments, gloves and catheter liners and to show a wide range of medium quality across nine programmes [457]. Its sensitivity depends entirely on how it is run. The one-cell assay detected changes in osmolality, pH and trace glutaraldehyde that the two-cell assay missed [476]; the one-cell assay was twice as sensitive as the two-cell assay and twenty times as sensitive as the sperm motility assay to peroxides in oil, and individual culture doubled its sensitivity again [277]; extending the assay to 144 h or counting cells detected toxicity invisible at the 96 h blastocyst rate, with group culture masking it while protein supplementation and oxygen concentration had minimal effect on assay sensitivity [278]; and time-lapse morphokinetics detected sub-lethal concentrations of cumene hydroperoxide and Triton X-100, and two lots of toxic oil, that did not change blastocyst rate [385]. Gardner and colleagues argued that assessing embryos at multiple time points and counting blastocyst cells lets the assay identify impaired as well as lethal conditions, and that several assays in concert are better than one [477]. Punt-van der Zalm and colleagues formalised a stratified, randomised, powered design requiring 69 embryos per group and identified five toxic samples among 52 tested [478]. Quinn's own 1984 caution remains true: the mouse assay did not predict human fertilization and cleavage in his hands, and the mouse is a good model for substrate utilisation but not for every aspect of human embryo physiology [29, 38].

Three further determinants of sensitivity are documented. Strain: outbred CF1 one-cell embryos detected cumene hydroperoxide and Triton X-100 at less than half the concentration that affected inbred, F1 or F2 embryos [479], and in four commercial single-step media the ranking of media changed with the mouse strain, so that no single strain can rank media for human use [480]. Endpoint: across 13 human ART culture protocols tested in one laboratory, blastocyst rates of 49.9 to 91.9% correlated only weakly with fetal rates of 15.7 to 62.0%, the two extremes differed in lineage composition, gene expression and litter size, and the authors called for post-implantation endpoints in an assay whose release criterion is 80% blastocysts [481]. And configuration in general: the foundational statement that the bioassay cannot distinguish media suitable for human embryos, because development is strain-, stage- and medium-dependent, but that pronuclear embryos in protein-free medium detect water, pH, temperature and contaminant faults, dates from 1993 [482]. How the manufacturers themselves run the assays was surveyed by questionnaire in 2018: of 21 companies approached, 11 responded, and their mouse embryo and sperm survival assay methodologies and thresholds, in strain, embryo number, culture conditions, extraction protocol and subcontracting, differed significantly with no regulation governing them, which is the primary evidence behind asking for the assay configuration (Section 16.2) [483]. Cumulative exposure is the newest question: 4 of 53 lots of disposables failed a sperm motility assay individually, but 17 of 48 combinations of three to five items that had passed individually failed, with risk rising with item number [484], whereas a mouse assay of ten combinations of 7 to 25 disposables covering a whole cycle found 2 of 10 combinations toxic and traced each to specific sperm straws and transfer catheters that were toxic alone despite certification, concluding that there is no inherent cumulative toxicity and attributing the failure of manufacturer certification to catch those devices to variability in assay protocols [485]. The bovine embryo assay is the most recent challenger. In a comparative study in which sperm selection, fertilization and culture media from two human IVF suppliers were used to produce bovine blastocysts from slaughterhouse oocytes (4118 oocytes, at least four independent replicates and at least 50 oocytes per group), the bovine assay detected differences in blastocyst rate, developmental kinetics, post-warming re-expansion and hatching, cell number and lineage allocation between products that the standard mouse embryo assay did not detect at all and that a mouse IVF-based assay detected only as a day 4 blastocyst difference; the authors concluded that the bovine assay reveals functional and quality differences invisible to the mouse assays and could strengthen safety assessment while reducing animal sacrifice [486]. A companion article sets out the case for standardising and validating the bovine assay as a replacement for the mouse assay on ethical, scientific, practical and economic grounds, noting the FDA's stated aim of phasing out animal toxicity testing, and proposes that regulators incorporate it [487]. The bovine assay uses oocytes from abattoir ovaries rather than purpose-bred animals and runs from fertilization rather than from a retrieved zygote. Two cautions apply. In the comparative study the bovine control medium outperformed both human products, so part of the difference between products may reflect how well a human formulation suits a bovine embryo rather than its quality, and a difference in a bovine assay is not a difference in human clinical outcome. And the FDA special controls for reproductive media specify the mouse embryo assay and do not name the bovine assay [85], so a laboratory reading a certificate of analysis should treat a bovine result as additional evidence rather than as a substitute for the mouse assay result those controls require.

The human sperm survival or motility assay complements the mouse assay for contact materials. Its sensitivity is greatest without albumin and at 4 to 8 h and 48 h, and items toxic within 8 h concord closely with mouse embryo results [488]; it identified 13 reprotoxic products among 36 in a four-year survey [456]. Neither assay detects endotoxin reliably, which is why endotoxin is measured directly by the Limulus test [86, 87].

16.2 Reading the certificate of analysis

What a certificate of analysis contains is set by the manufacturer within the framework of the special controls and the product's clearance; typically it reports pH under stated CO₂ conditions, osmolality, endotoxin, sterility and a mouse embryo assay result. The author's recommendation, beyond what is mandatory, is to ask for the assay configuration (strain, one-cell or two-cell, group or individual culture, endpoint and time), the peroxide value for oil, and the source, endotoxin and stabiliser content for a protein supplement, because those are the details on which the assay's sensitivity and the supplement's behaviour depend (Section 16.1). The compositional surveys make clear what a certificate does not tell the laboratory, and the most recent survey, 47 media and supplements from seven suppliers bought in 2019 and 2020 and analysed for 40 components, found glucose following a high-low-high pattern in every sequential system (2.5 to 3 mM in most fertilization media, 0.5 mM or below in all cleavage media, 2.5 to 3.3 mM in most blastocyst media, with continuous media resembling cleavage media) but lactate, glycine and potassium differing clearly between brands, no two media identical, brands under the same parent company differing from each other, and, in the authors' words, limited scientific backing for the specific concentrations used [489]. A certificate does not report the concentration of each amino acid and carbohydrate [134, 135, 489], the non-declared protein cargo [226], the metal content of the supplement [223], whether lactate is supplied as the L-isomer or the racemate [136], or how the composition has changed since the lot was released [136]. The receiving inspection should therefore confirm the certificate values against the laboratory's own acceptance limits, measure pH and osmolality on receipt where the laboratory has the instruments, and record the shipping temperature and the opening date [88, 89].

Listing pH and osmolality as measurements is not the same as measuring them comparably. A pH value is only interpretable with the CO₂ percentage, the temperature and the equilibration time at which it was taken: the same medium reads differently at 5% and 6% CO₂ because its pH is set by the bicarbonate and the CO₂ together [101, 102], and, as a matter of the chemistry of a bicarbonate buffer, differently at 37 °C and at room temperature and before and after equilibration, while a sample carried across the laboratory to an electrode loses CO₂ on the way, so a manufacturer's certificate value and a laboratory reading taken under different conditions are not expected to agree; the validation of a portable blood gas analyser measured under oil in the culture dish itself, with within-run and between-day coefficients of variation below 0.35%, an overestimate of about 0.03 pH units relative to the reference electrode in that study's hands, and a reading after 24 to 48 h of incubation as the most representative of what the embryo sees, shows the level of definition needed before a pH limit can be enforced [490]. Osmolality has its own conditions: the dish preparation and the minutes a drop spends uncovered raise it before any measurement [98], and the practical rule is that a freezing-point osmometer calibrated with certified standards, run on a sample taken from the unopened bottle and on a drop prepared the laboratory's usual way, may give different numbers, both of them legitimate for what they describe. Acceptance limits therefore belong to a product, an instrument and a stated measurement condition. Repeated measurements of a standard estimate precision, which is only one component of measurement uncertainty; an uncertainty assessment also has to consider the calibration and reference-material uncertainty, bias against a reference method, between-day variation, and the sampling and matrix effects of measuring a protein-containing medium under oil rather than a standard in a cup, all under the laboratory's stated measurement conditions. A laboratory that has written that assessment down knows how far a reading can drift before it means anything.

16.3 Qualifying a change of medium, protein, oil or lot

No product change in this category is a drop-in, but the scale of qualification should match the scale of the change. ESHRE's laboratory recommendations and the Cairo consensus expect documented validation with predefined acceptance criteria [88, 89, 491]. For an incoming lot of a product already in use, verification is the appropriate level: confirm the certificate, endotoxin and assay result against the laboratory's acceptance limits, measure pH at the laboratory's CO₂ set point and osmolality on receipt where instruments allow, since pH depends on the product's bicarbonate and the local CO₂ condition [101, 102] (a portable blood gas analyser validated under oil in two dish types gave within-run and between-day coefficients of variation below 0.35%, and a reading after 24 to 48 h of incubation best represents the pH the embryo sees [490]), and bridge the new lot alongside the old with routine key performance indicators watched for a shift, the verification level described in the consensus documents [88, 89]. For a new formulation, protein product, oil or culture system, the change is one of exposure and a prospective comparison is warranted: for oil or dishes, a sham culture under the laboratory's own incubator conditions to measure osmolality drift [9, 391]; for a genuinely new formulation, a parallel or sibling-oocyte comparison with fertilization, day 3 quality, blastulation and utilisation as endpoints, on the pattern of the paired designs of Werner and colleagues and Hardarson and colleagues, remembering that sibling embryos share patient-level factors and are not independent patients [175, 262], and on the worked example of a sibling-oocyte study of 1274 oocytes powered to detect a 7 percentage point difference in blastulation per cohort, which found 42.8% against 43.1% for a newly commercialised medium alongside slower morphokinetics and more poor-quality blastocysts, with euploidy and implantation after euploid transfer as secondary endpoints [492]; the extent of qualification for a change of protein product should follow a documented risk assessment of what actually differs, because two plasma-derived albumins at the same nominal concentration can differ in stabiliser, lipid cargo, impurities and biological effect (Sections 6.2 and 6.6), so a same-concentration substitution is not pre-classified here as like-for-like; routine verification is for a new lot of an established product, and few laboratories can power a pregnancy endpoint for a single component. Singleton birthweight belongs in the laboratory's ongoing surveillance, because randomised and pseudo-randomised human comparisons have found it to differ between media [10, 11], but it is a long-term indicator and cannot serve as a release endpoint. ESHRE and the Cairo consensus expect the change to be documented [88, 89]; this page's further recommendation, which goes beyond either document, is to retain a bridge stock of the outgoing product, update the SOP, training records and inventory, and file the qualification package with the quality records.

When outcomes fall and a medium, protein supplement or oil is suspected, the same records make the investigation possible. The exposed cycles should be linked to the lots of medium, supplement and oil they used and to the dates those bottles were opened, and the outcome trend examined stage by stage, since a fertilization failure, a cleavage-stage arrest and a loss of blastulation point to different exposures (Sections 9, 5 and 14); every other change made in the same period, of incubator, gas supply, dish, oil depth, operator or protocol, should be listed before the medium is blamed; unopened material from the suspect lot should be retained rather than discarded, so that pH, osmolality, a sperm survival or mouse embryo assay and, where a laboratory has access to one, a compositional analysis can be run alongside a comparator lot; and the manufacturer should be told, with lot numbers, because a signal seen in one laboratory may be one of several, and a device incident is reportable under the frameworks of Section 16.4. Three kinds of limit should be kept distinct throughout: the manufacturer's release specification on the certificate, the laboratory's own acceptance limits, and any investigational threshold set for the enquiry. None of this is evidence that a product caused an observed outcome; it is the quality-system context in which that question can be asked [88, 89].

16.4 The regulatory frame

In the United States media, supplements, sperm separation media and oil are Class II medical devices under 21 CFR 884.6180, cleared through premarket notification on the basis of special controls that include the mouse embryo assay, endotoxin testing, sterilisation validation, biocompatibility and clinical data, with washing PBS, overlay oil and water exempt from premarket notification within stated limits; a product's clearance status should be confirmed in the FDA database rather than assumed [85]. Research-use-only or unlisted products in clinical embryo contact are a compliance problem before they are a quality one. In the European Union the classification is stricter. Rule 3 of Annex VIII to the Medical Device Regulation places substances or mixtures of substances used in vitro in direct contact with human cells, tissues or organs, or with human embryos before implantation, in Class III [493], and the Commission's classification guidance gives IVF and ART products without a principal pharmacological or metabolic action as the example, lists a density-gradient medium for sperm separation, as a separation by physical means, as Class IIa, and gives long-term storage containers for sperm and embryos as Class IIa; media containing human albumin fall under the separate rule for devices incorporating a substance derived from human blood, which is also Class III and adds consultation of a medicines authority [494]. Class III certification under the Regulation means notified-body assessment at the highest device class, but the transitional provisions allow products certified under the earlier directive to remain on the market until the end of 2027 or 2028 depending on class, so a CE mark on a bottle today does not by itself show which regime it was assessed under [495]. A medium cleared in the United States and one CE-marked in Europe have passed different regulatory tests, and a laboratory using imported media should establish which one applies to the product on its shelf. Professional guidance sits in a different layer from law. ESHRE's recommendations on good practice in the IVF laboratory, revised in 2015 and updated in 2026 with separate sections on culture and on transfer, an expanded andrology section and a new biopsy section, set expectations for consumables, media handling, equilibration, culture and cryopreservation, and for the emergency planning around them [89, 491]; ESHRE's 2023 add-on recommendations address GM-CSF media and hyaluronan transfer media specifically [2]; and the Cairo consensus provides more than fifty specific guideline points on temperature, humidity, CO₂ and pH, oxygen, buffering, composition and protein supplementation, sequential and single-step systems, cold chain and storage, and test equipment [88]. Semen analysis and sperm preparation follow the WHO laboratory manual, sixth edition [353]. These are recommendations, not regulation or accreditation requirements in themselves, though accreditation bodies draw on them. None substitutes for the manufacturer's instructions for use, and none of them, or this page, substitutes for the laboratory's own validated procedures.