1.1 Who this is for and what it covers

This reference is written for clinical embryologists, laboratory directors and the staff who train them. Its question is a practical one: what does an embryologist need to understand about preimplantation genetic testing (PGT) to take a good sample, protect its identity, read the report correctly and take part in the decisions that follow? It covers testing for aneuploidy (PGT-A), for monogenic conditions (PGT-M), for structural rearrangements (PGT-SR), assessment of ploidy, noninvasive approaches, polygenic risk screening (PGT-P) and the quality systems that sit around all of them.

It does not cover the genetic counselling of patients, the clinical indications for treatment, or the design of a genetics laboratory's own assays, except where these change what happens at the bench. Polar body biopsy and cleavage-stage biopsy are described for completeness and for laboratories working under legal restrictions, but the reference assumes trophectoderm biopsy at the blastocyst stage as the reference method, because that is where current practice and most of the evidence sit [1, 2].

Abbreviations used throughout:

• PGT, preimplantation genetic testing: PGT-A for aneuploidy, PGT-M for monogenic conditions, PGT-SR for structural rearrangements, PGT-P for polygenic risk.
• ICSI, intracytoplasmic sperm injection; ART, assisted reproductive technology.
• NGS, next-generation sequencing; SNP, single-nucleotide polymorphism; array CGH, array comparative genomic hybridisation; FISH, fluorescence in situ hybridisation; qPCR, quantitative polymerase chain reaction; NIPT, noninvasive prenatal testing.
• CI, confidence interval.
• ESHRE, European Society of Human Reproduction and Embryology; ASRM, American Society for Reproductive Medicine; SART, Society for Assisted Reproductive Technology; HFEA, Human Fertilisation and Embryology Authority (UK); PGDIS, Preimplantation Genetic Diagnosis International Society.

1.2 Evidence conventions

Each statement is one of four types: established biology; a consensus or guideline position, with the document named; laboratory practice that varies between centres; and open questions on which the literature disagrees. Where the literature disagrees, both positions are stated with their evidence and the reference does not try to settle the dispute. Statements of common laboratory practice carry no citation. Practical recommendations for the laboratory that are not attributed to a source, such as those closing several subsections of Sections 5, 8, 12, 15 and 16 and the "In practice" notes ending Sections 6, 7, 9 and 10, are this reference's own.

Numbers are given with their denominators where the source reports them, and the unit of analysis is named: per woman randomised, per cycle started, per retrieval, per transfer or per embryo. Age is that of the woman whose oocytes were used, at retrieval, unless stated otherwise: for donor oocytes it is the donor's age, and for vitrified oocytes the age at retrieval, not at warming or transfer. These denominators are not interchangeable, and much of the published disagreement about PGT-A turns on which one a study chose. A difference that is not statistically significant is described as no detected difference, not as equivalence, unless the study was designed and powered to show equivalence or non-inferiority. Results in mouse or other models are labelled as such.

Two kinds of study recur and are worth defining at the outset. In a randomised controlled trial, patients are allocated to testing or no testing and the outcome compared; this answers whether offering the test changes what happens to patients. In a non-selection study, embryos are biopsied and analysed but the result is withheld and embryos are transferred as if untested; the results are unblinded only after the outcome is known. A non-selection study answers a different question, how well a given result predicts what the embryo will do, and it measures the predictive value of an abnormal result with less bias from knowledge of the result than other designs, without knowingly transferring embryos classed as abnormal; it still depends on patients consenting to transfer without the result being used. Transfers made knowing the result and retrospective re-analyses add complementary evidence, each with its own selection.

This is a narrative educational review, not a systematic review: the four types of statement above describe the kind of claim being made, not a formal grading of the evidence, and naming the databases searched does not make the coverage complete. The literature was searched in PubMed, Europe PMC and CrossRef. The published output of one research group, whose embryology and genetics papers anchor many sections, was retrieved by author name and ORCID. Independent work was sought by topic searches for each section, aimed at studies by other groups, the trials and critiques that reach different conclusions, and society guidance and regulation, and by screening the reference lists of the cited papers and highly co-cited papers for omissions; these searches were not designed to systematic-review standards and cannot show that all independent work was found. The search record (search strings, dates and inclusion choices; searches last run on 30 October 2026) is available from IVF Store on request and will be published as an appendix to this reference.

That group centres on Antonio Capalbo, a reproductive geneticist who has authored more than 120 indexed papers on PGT, embryology and reproductive genetics, many with the embryology group of Laura Rienzi, Filippo Ubaldi and Danilo Cimadomo in Rome. Several of its authors, including Capalbo, are or have been employed by commercial providers of PGT (Igenomix and Juno Genetics), as their papers' declarations and affiliations state. Where the reference summarises the group's positions, it says so, and it sets out the evidence of those who disagree alongside. Capalbo and his co-authors did not write or review this reference, and it does not represent their views. IVF Store's Chief Scientific Officer, Thomas Elliott, is a co-author of two of the cited studies from this group [3, 4]. Several of the reference's own recommendations, such as biopsying poor-quality, day 7 and atypically pronucleated embryos, would increase the number of biopsies a laboratory performs, and readers should weigh them knowing that IVF Store sells biopsy consumables, including biopsy pipettes, and lists PGT services from NOVA Genomics (see the disclaimer above).

1.3 Using this reference

Nothing in this reference replaces the validation data and report of the genetics laboratory performing a test or the genetic counselling a patient receives. Patients and families who find it should discuss any individual result with their clinic and genetic counsellor: the figures here describe groups of embryos and patients, not any one person's chances.

2.1 Terms

PGT is the analysis of DNA from oocytes (through their polar bodies) or embryos, before transfer, to detect genetic changes [5]. The older terms preimplantation genetic diagnosis (PGD) and preimplantation genetic screening (PGS) were replaced by a single term with a suffix for the purpose of the test in the 2017 International Glossary on Infertility and Fertility Care, a consensus set of 283 terms [6], and the current usage is:

• PGT-A: testing for aneuploidy, an abnormal number of chromosomes, usually in patients with no known inherited chromosomal risk.
• PGT-M: testing for a monogenic (single-gene) condition for which one or both partners are known to be at risk [7].
• PGT-SR: testing for unbalanced products of a structural rearrangement, such as a reciprocal or Robertsonian translocation or an inversion, carried by one partner [8].
• PGT-P: estimation of polygenic risk for common diseases from genome-wide genotype data, sometimes called polygenic embryo screening (Section 14) [9].

Two further terms matter to the laboratory. Ploidy is the number of complete chromosome sets (haploid, diploid, triploid); most PGT-A methods count chromosome copy number relative to the rest of the genome and cannot, on their own, see an extra or missing complete set (Section 12). Mosaicism is the presence of two or more cell lines with different chromosome complements in one embryo; in practice the word is often used for a biopsy result that falls between the normal and abnormal thresholds, which is not the same thing (Section 8).

2.2 Thirty years in brief

The first pregnancies after embryo biopsy were reported in 1990, from single blastomeres removed after zona drilling with acid Tyrode's solution and analysed to identify the sex of embryos from couples at risk of X-linked disease [2, 10]. The method was extended to monogenic conditions and then to chromosomal rearrangements and numerical abnormalities, first by FISH on a handful of chromosomes; an early protocol scored five (X, Y, 13, 18 and 21) simultaneously and gave a successful analysis in 93% of 157 blastomeres from 30 embryos [11]. Cleavage-stage FISH screening for aneuploidy did not improve live birth rates in randomised trials and, in women of advanced maternal age, lowered them [12, 13]. A later pair of randomised trials did report a benefit in women aged 41 to 44 (live birth per patient 30 of 93, 32.3%, after day 3 FISH screening against 14 of 90, 15.5%, with day 5 transfer in both arms) and, in repeated implantation failure, a 20-point difference that did not reach significance in 91 women (23 of 48, 47.9%, against 12 of 43, 27.9%) [14]. The field moved, through the 2010s, to comprehensive analysis of all chromosomes on several trophectoderm cells taken from the blastocyst [2, 15, 16]. Trophectoderm biopsy itself had entered clinical use in the early 2000s: in one Australian clinic's 231 PGD cycles from 2002 to 2004, 974 of 1,050 biopsied blastocysts (93%) gave an unambiguous result [17].

That move changed the embryology laboratory as much as the genetics. Blastocyst biopsy brought extended culture, vitrification of every biopsied embryo while the result is awaited, and single frozen embryo transfer into routine practice. In one centre's pre- and post-intervention study of women aged 36 to 44, introducing an elective single embryo transfer policy with selection enhanced by blastocyst culture and PGT-A was followed by a fall in the multiple delivery rate from 21.0% to 6.8%, while the cumulative delivery rate per retrieval did not differ significantly (21.0% and 20.4%; 1,609 and 574 retrievals) [18]. In one Italian centre the proportion of retrievals followed by PGT rose from 0.2% in 2015 to 72.9% in 2024, with the number of embryologists trained to biopsy rising from one to four over the same period [19]. In the US SART registry, PGT-A was used in 11.5% of first autologous IVF cycles in 2014 and in 49.0% in 2020 (438,583 cycles analysed; donor oocyte, gestational carrier and fresh-transfer cycles were excluded, so the denominator is not all US IVF cycles) [20].

Two features of the history explain the present debate. First, the early, cleavage-stage version of PGT-A was introduced on a biological premise before its benefit had been tested, and in most trials it failed when tested. Second, the methods now in use are very different from those that failed, so evidence about one cannot simply be transferred to the other [21]. Section 10 sets out what the current evidence shows.

3.1 The oocyte and the maternal age effect

Aneuploidy is the most commonly identified chromosome abnormality in humans and the leading genetic cause of pregnancy loss; a review estimated that it affects at least 5% of clinically recognised pregnancies [22]. The single largest source of aneuploidy in human embryos is the oocyte. A review that searched the primary literature systematically estimated that about 20% of oocytes are aneuploid at baseline, that the proportion rises exponentially from 30 to 35 years, and that it reaches on average about 80% of oocytes by 42 years; the authors concluded that female reproductive senescence is predominantly determined by the declining genetic quality of the oocyte [23]. Trophectoderm biopsies show the same pattern. In a retrospective review of 15,169 consecutive biopsies analysed by comprehensive chromosome screening, aneuploidy rose predictably with the female partner's age after 26 years, with a smaller rise at younger ages to more than 40% of biopsies in women aged 23 and under. No euploid embryo was found in 2% to 6% of cohorts between 26 and 37 years, 33% at 42 and 53% at 44, a per-cohort figure rather than the proportion of embryos aneuploid, and trisomies and monosomies were about equally frequent [24]. Aneuploidy of paternal origin is less frequent [22]; in a multicentre retrospective SNP-array study of 2,409 blastocysts, paternal-origin whole-chromosome aneuploidy showed no significant association with paternal age, whereas maternal-origin aneuploidy followed maternal age, although paternal-origin segmental deletions were 2.9 times as frequent as maternal ones in the 867 embryos analysed for segmental changes [25].

Following chromosome segregation in oocytes from females aged 9 to 43 years showed that aneuploidy follows a U-shaped curve with age, higher in the youngest and the oldest, and that different error types explain the two arms: whole-chromosome nondisjunction contributes more in young girls, while loss of cohesion between sister chromatids, at the centromere and more extensively, limits fertility as women age [26].

Whether aneuploidy accounts for all of the age effect is disputed. In a multicentre retrospective study using array CGH, implantation of euploid blastocysts ranged from 51% to 40% between under 35 and 42 years and did not differ significantly, which the authors read as evidence that the decline in IVF success with age is primarily caused by aneuploidy [27]. A later meta-analysis of 7 studies covering 11,335 euploid embryo transfers found a higher ongoing pregnancy or live birth rate in women younger than 35 than in those aged 35 or older (odds ratio 1.29, 95% CI 1.07 to 1.54), with a gradient across age groups, and concluded that maternal age lowers success independently of ploidy [28].

3.2 How meiosis goes wrong

Classical nondisjunction, in which both homologues travel to the same pole at meiosis I, is not the most frequent error in human oocytes; premature separation of sister chromatids and reverse segregation are common [29, 30]. An early cytogenetic study pointed this way: among oocytes remaining unfertilized after IVF, 4 of 38 metaphase II oocytes contained single chromatids rather than whole extra chromosomes, which the study attributed to premature division of the centromeres at meiosis I, or predivision [31]. Recovering all three products of single female meioses and genotyping more than four million informative SNPs mapped 2,032 maternal crossovers and the segregation of 529 chromosome pairs, and revealed a pattern called reverse segregation, in which both homologues separate their sister chromatids at meiosis I. The same data showed selection against embryos from meioses with low recombination, and preferential segregation of recombined chromatids to the oocyte at meiosis II [29].

At the extreme, failure of recombination across the whole genome can produce oocytes that are ovulated but give rise to triploid conceptions. In a retrospective cohort of 96,660 trophectoderm biopsies of 2PN blastocysts from 20,187 ICSI cycles, 1,063 of the 93,986 with an informative result (1.1% of informative blastocyst biopsies) carried an extra or missing chromosome set, 83% triploid and 17% haploid; the authors call this 1.1% of conceptions, but the cohort contains only embryos that reached biopsy. The odds rose with maternal age (odds ratio 1.046 per year, an observational association), about one third of maternal triploidies arose at meiosis I and two thirds at meiosis II, and seven of the triploid embryos analysed appeared to come from oocytes that had failed to initiate recombination at all [32].

The practical point for the laboratory is that many age-associated errors are set up long before retrieval: crossovers between homologues are established in fetal oogenesis, and the cohesion that holds them must last through decades of arrest. Final segregation comes later, at meiosis I during maturation and at meiosis II, which the oocyte, arrested at metaphase II, completes only on activation at fertilization; the data above include errors at both divisions. Culture and handling can affect embryo survival and development, which does not establish them as a principal cause of age-related aneuploidy; whether ovarian stimulation influences oocyte aneuploidy is still debated.

3.3 Mitotic errors after fertilization

After fertilization, chromosome segregation errors in the first cleavage divisions are common: genome-wide analysis of single blastomeres found mosaicism for whole-chromosome aneuploidies and uniparental disomies in most cleavage-stage embryos, together with frequent segmental deletions, duplications and amplifications that were reciprocal in sister blastomeres [33]. Unlike meiotic errors, they affect only the descendants of the cell in which they occurred and so usually produce mosaic embryos; an error at the first cleavage, whose reciprocal products can leave both daughter lineages abnormal, can give an embryo with no normal cells. Many embryos with such errors arrest; a 2025 review reports that about half of human embryos stop developing in vitro or shortly after transfer and that around 90% of these are aneuploid, most of them from errors in early mitotic divisions and associated with abnormal cleavage, spindle and cohesion defects, mitochondrial dysfunction and variants in maternal-effect genes [34]. In primary data from 909 embryos tested (297 arrested embryos and 612 blastocysts), 94% of the arrested embryos were aneuploid, mostly with mitotic aneuploidies affecting several chromosomes, and on time-lapse imaging (843 embryos) 51% of embryos with abnormal first or second cleavage divisions carried mitotic aneuploidies against 23% of normally dividing embryos [35]. Maternal genotype may contribute: common variants spanning PLK4, a gene with a well-characterised role in centriole duplication, were associated with aneuploidy of putative mitotic origin in day 3 embryos, and women carrying the high-risk genotypes contributed fewer embryos for testing at day 5 [36].

Sequential analysis of polar bodies, a blastomere and a trophectoderm biopsy from the same embryos found the polar bodies the least accurate predictor of the blastocyst's chromosome complement, mainly because of post-zygotic events (Section 4.3) [30]. By the blastocyst stage, aneuploidy of mitotic origin appears to be the minority: a method that combines SNP genotyping with copy number to assign the cell division of origin, applied to 2,277 blastocysts with parental DNA (mean maternal age 34.4), classified 71% as euploid, 27% as meiotic aneuploid and 2% as mitotic aneuploid [37]. That figure depends on the assay, its criteria for calling an error mitotic and the cohort sampled; it describes trophectoderm biopsies of embryos that survived to the blastocyst stage, not the prevalence of mosaicism across whole embryos. Section 8 discusses what this means for intermediate copy-number results.

3.4 What morphology can and cannot tell you

Blastocyst morphology and chromosome complement are related but only moderately. In 956 blastocysts from 213 patients, euploidy rates were 56.4%, 39.1%, 42.8% and 25.5% in blastocysts graded excellent, good, average and poor, yet among 215 euploid blastocysts transferred singly, morphology and developmental rate were not statistically significant predictors of implantation in this retrospective analysis [4]. Later pooled data did find lower live birth after transfer of euploid blastocysts of poorer morphology (Section 10.4). Time-lapse morphokinetics have not provided a reliable substitute: a longitudinal cohort of 455 blastocysts from 138 patients at increased risk of aneuploidy found no statistically significant correlation between any of 16 morphokinetic parameters and aneuploidy [38]. In a recent single-centre retrospective study of 279 PGT-A cycles that each produced at least three blastocysts, including at least one euploid and one aneuploid, the better of two versions of a commercial artificial-intelligence model (one author is an employee of Vitrolife, as the paper declares) ranked a euploid blastocyst first in 68% of cycles (62% for the other version), against 58% for Gardner grading and 44% for random choice (the abstract reports no confidence intervals or tests). The authors described this as matching or slightly improving on conventional assessment, and concluded that no ranking method based on morphology or morphokinetics can prevent an aneuploid blastocyst from being ranked first [39]. In a prospective multicentre non-selection cohort, in which the best-looking blastocyst was transferred and its chromosome status assigned only afterwards, 53 of 188 transfers (28%) were of aneuploid embryos, with live birth after 8% of aneuploid against 50% of euploid transfers [40]. The aneuploid group combined 44 uniform whole-chromosome and 9 segmental results, and a later review counted 2 live births among the 44 whole-chromosome transfers [41]; the two categories carry different prognoses (Section 9.1).

The reverse is also true: a poor-looking blastocyst is not necessarily an incompetent one when it is euploid, although its chances are much lower (live birth 10.8% after 195 euploid poor-quality transfers against 44.6% after 1,697 euploid good-quality transfers, adjusted odds ratio 0.22). In PGT-A cycles, poor-quality blastocysts contributed live births that would otherwise not have occurred, an argument for biopsying them rather than discarding them on appearance (Section 5.7) [42].

4.1 The three options and what each sees

Three stages can be sampled, and each sees a different part of the embryo's genetic history [1, 43].

• Polar bodies (first and second, either sequentially, the first within about 4 hours of retrieval and the second at the fertilization check, or together 6 to 9 hours after insemination) can reveal maternal meiotic segregation only. The second polar body is extruded after oocyte activation, so it is available only once the oocyte has been inseminated or injected; neither polar body samples paternal chromosomes or postzygotic mitotic events, so paternal meiotic errors, paternal variants and mosaicism are invisible to it [1]. Polar body biopsy remains an option where embryo biopsy is prohibited by law, or where only a maternal variant, rearrangement or meiotic aneuploidy is being investigated [1].
• A cleavage-stage blastomere (usually one cell on day 3) sees both parental contributions but samples a single cell at the stage when mitotic errors are most frequent, so one cell may not represent the embryo, and the embryo loses one-sixth to one-eighth of its cells, since ESHRE recommends biopsy at the six-cell stage or later and discourages taking two cells [1].
• Trophectoderm (several cells from the blastocyst, day 5 to 7) samples the lineage that forms the placenta, leaves the inner cell mass untouched and provides more DNA, at the cost of extended culture and, in practice, vitrification while the result is awaited [1].

Polar body testing has been assessed in one randomised trial. In a multicentre pragmatic trial in women aged 36 to 40, PGT-A by array CGH of the first and second polar bodies gave a cumulative live birth rate within one year that did not differ significantly from ICSI without testing (50 of 205, 24%, against 45 of 191, 24%; difference 0.83 percentage points, 95% CI −7.60 to 9.18), with fewer women reaching a transfer (relative risk 0.81, 95% CI 0.74 to 0.89) and fewer miscarriages (relative risk 0.48, 95% CI 0.26 to 0.90) [44].

4.2 Does the biopsy harm the embryo?

The clearest answer comes from a paired randomised design in which two embryos were selected for transfer, one was randomised to biopsy and the other left as control, both were transferred, and the source of each baby was identified by comparing the neonate's DNA with the biopsy. Removing a blastomere on day 3 reduced sustained implantation (implantation going on to a live-born infant, counted per embryo within pairs) to 30% of biopsied embryos against 50% of their unbiopsied partners, a relative reduction of 39%. For trophectoderm biopsy, sustained implantation was 51% for biopsied and 54% for control blastocysts, which the authors interpreted as no measurable impact; the abstract does not give the number of pairs or the power calculation, and the trial registration (NCT01219504) posts no results, so the precision of that comparison cannot be judged from the sources read for this reference [45]. This trial is widely regarded as a main reason PGT moved to the blastocyst. Earlier, a prospective cohort of day 5 single embryo transfers had shown that the amount removed at cleavage stage matters: live birth was 37.4% (95% CI 29.0% to 47.4%) after removal of one blastomere (182 embryos transferred) and 22.4% (95% CI 17.0% to 28.9%) after removal of two (259 embryos), against 35.0% in the reference ICSI population [46].

A later meta-analysis of 13 observational studies and trials, covering 11,469 live births after trophectoderm biopsy and 20,438 after IVF or ICSI without testing, found a slightly higher pooled odds of preterm delivery after biopsy (odds ratio 1.12, 95% CI 1.03 to 1.21) that did not persist in sensitivity analysis (odds ratio 0.97, 95% CI 0.84 to 1.11), and no detected increase in low birthweight (odds ratio 1.01, 95% CI 0.85 to 1.20) or in other obstetric and neonatal outcomes. The authors noted that the evidence is mainly observational [47]. Obstetric and neonatal outcomes after PGT-A more broadly are reviewed in Section 10.5.

4.3 Why the blastocyst is the reference stage

Taken together, the argument for the blastocyst rests on three findings: the polar bodies are the least accurate predictor of the blastocyst's chromosome complement and a single blastomere is an imperfect one (sensitivity per chromosome 61.7% and 86.4%, against the trophectoderm as reference, in 21 embryos from 9 couples, all with the woman older than 40) [30], and a day 3 blastomere agrees less well than a trophectoderm biopsy with the inner cell mass (kappa 0.553 against 0.857) [48]; trophectoderm biopsy, unlike blastomere biopsy, has shown no measurable effect on implantation in a randomised paired design [45]; and multicellular samples give a more reliable genetic analysis [1]. The cost is that only embryos that reach the blastocyst are tested, more laboratory time is needed across days 5 to 7, and a freeze-all strategy (vitrifying every biopsied embryo and transferring in a later cycle) is usually required [1].

5.1 Before the biopsy: insemination and culture

Copy-number tests cannot distinguish embryonic DNA from DNA that should not be there, and genotyping methods that detect contamination do so only within their validated limits (Section 5.4), so contamination control starts at insemination. ESHRE considers ICSI preferable for PGT to reduce paternal contamination from sperm bound to the zona pellucida, and careful removal of cumulus cells, with rinsing of the oocytes before injection, to reduce maternal contamination [1]. The 2026 ASRM and SART committee opinion on ICSI for non-male factor indications is indication-specific: it does not recommend routine ICSI in the absence of male factor infertility or prior fertilization failure, concludes from the PGT-A studies it reviewed that ICSI without male factor does not improve live birth, euploidy rates or embryo numbers and so does not recommend it for PGT-A alone, and considers that ICSI may be appropriate for PGT-M to reduce contamination from cumulus cells and extraneous sperm that could cause misdiagnosis [49]. The two positions therefore differ for PGT-A and agree for PGT-M. In a randomised sibling-oocyte pilot study of 30 couples with non-male factor infertility, conventional IVF and ICSI did not differ significantly in the number of euploid blastocysts per cycle (2.0 ± 1.8 and 1.9 ± 1.7), and the authors cited evidence that sperm DNA fails to amplify under the conditions used for trophectoderm samples [50]. A retrospective single-centre review of 251 blastocysts from conventional IVF and 724 from ICSI found no significant difference in euploid rates (27.9% and 30%) and a non-significant difference in mosaic results (25.9% after conventional IVF against 20.9% after ICSI) [51]. Neither study measured contamination or diagnostic error: comparable euploid proportions show that the insemination method did not shift the distribution of results, not that no extraneous DNA was amplified. This reference proposes that the insemination method and contamination-control plan for each indication be agreed with the testing laboratory during work-up. Until biopsy, routine culture conditions apply, but embryos should be cultured individually, in multiwell dishes or separate drops, so that each biopsy can be traced to one embryo, and rinsed free of biopsy medium before they return to culture or are vitrified [1].

A practical point follows: for PGT cycles, the denudation is not only a preparation for injection but the first contamination control in the test. For ICSI this means checking that no corona cells remain attached after denudation; where conventional insemination is used for PGT-A, ESHRE asks that zygotes be fully denuded and rinsed after the fertilization check [1].

5.2 Trophectoderm biopsy

The ESHRE good practice recommendations for biopsy set out the main points [1]:

• Biopsy is performed on days 5 to 7, according to the rate of development, once the inner cell mass is clearly visible, so that cells can be taken away from it.
• Five to ten trophectoderm cells are recommended; the effect of removing more than ten remains under investigation.
• Calcium- and magnesium-free medium, used to loosen blastomeres for cleavage-stage biopsy, should not be used for blastocyst biopsy.
• The zona can be opened by laser or mechanically, although a non-contact laser is highly recommended for blastocysts; the blastocyst is held with the inner cell mass between 7 and 11 o'clock, away from the opening and the holding pipette, trophectoderm cells are aspirated with gentle suction into a biopsy pipette of about 25 to 30 µm inner diameter, and they are released with as few laser shots as possible, aimed at cell junctions, and a fully hatched blastocyst can still be biopsied by combining laser pulses with a flicking movement.
• The biopsy pipette should be changed for each blastocyst, or, if rinsed instead, the laboratory should verify that rinsing prevents carry-over.

Three approaches to opening the zona are in use: laser opening on day 3 or 4 so that cells herniate by the time of biopsy; opening on the day of biopsy followed by a wait for herniation; and simultaneous opening and retrieval. ESHRE describes the first as more time-consuming but easier, unless herniation starts at the inner cell mass, the second as also time-consuming, since the blastocyst must be checked repeatedly (ideally in a time-lapse incubator), but the easiest, and the third as the quickest but the hardest to learn; the choice is a matter of laboratory policy [1]. Opening the zona at the expanded blastocyst stage, rather than on day 3, is the approach that keeps culture uninterrupted, which matters to laboratories using single-step media in time-lapse incubators [15]. A systematic review of factors associated with live birth after euploid blastocyst transfer found better results with simultaneous zona opening and biopsy than with the day 3 hatching-based protocol (3 studies, odds ratio 1.41, 95% CI 1.18 to 1.69), within a body of evidence it graded as low or very low quality [52].

Cellular debris and excluded fragments in the subzonal space should not be taken. In one reported case, fragments sampled alongside the trophectoderm gave a different karyotype from the biopsy, and genotyping showed that the fragment DNA came from the second polar body, which had persisted to the blastocyst stage [53].

5.3 Is the biopsy operator a variable?

It can be, which is why training and competency matter. In 2,586 consecutive blastocyst biopsies (counted per biopsy; several came from each patient) by seven practitioners in three centres, all trained the same way and using similar equipment, 94.2% gave a conclusive diagnosis, 4.6% gave a non-concurrent result and 1.2% failed to amplify. After adjustment, no difference was detected between practitioners in diagnostic rate or in the quality of the genetic data, nor in outcomes of 494 single euploid frozen transfers. The mean estimated number of cells per biopsy was 7.32 (range 2 to 15), and higher cellularity was associated with better-quality data and a higher conclusive rate [54]. Larger is not necessarily better for the embryo: in a retrospective single-centre cohort of first single transfers after biopsy, the quartile of biopsies with the highest relative DNA content had a lower live birth rate than the other three quartiles (relative risk 0.84, 95% CI 0.75 to 0.95); the authors could not separate reduced diagnostic accuracy from a mechanical effect of the biopsy [55].

The finding of the multi-operator study is not that the operator does not matter; it is that consistent results across operators were achieved under uniform training and equipment. A laboratory should be able to show the same thing from its own data, by monitoring each operator's conclusive and amplification-failure rates and the cellularity reported by the genetics laboratory. ESHRE does set a consensus minimum: preclinical practice on at least 50 oocytes or embryos for each biopsy stage, then at least 20 supervised clinical biopsies for practitioners experienced in micromanipulation or 40 for those without, with post-biopsy survival, sample lysis and amplification outcomes compared against the laboratory's standards [1]. These numbers are expert consensus; this reference identified no externally validated, universal case-number threshold for independent practice, only centre-level reports of the kind above. As this reference's suggestion, a programme should state which outcomes it uses for supervised sign-off (for example post-biopsy survival and re-expansion, conclusive and amplification-failure rates, and reported cellularity) and the acceptance limits it applies to each. IVF Store's own Training and Sign-Off Manual includes a biopsy module that laboratories may adapt, but it is a template, not a validated competency standard.

5.4 Tubing, negative controls and contamination

After biopsy, the cells are washed and transferred to a small reaction tube, a step called tubing, either in buffer or directly in lysis buffer according to the genetics laboratory's protocol. Transfer may be done with or without microscopic visualisation of the cells entering the tube. ESHRE recommends at least one negative control per buffer for each collection session (a blank taken through the same handling), more where single cells are handled; biopsied cells washed at least twice in fresh buffer with a new transfer pipette for each embryo; tubing in a dedicated clean area or hood decontaminated with UV-C, close to the biopsy station; and full protective clothing with powder-free, well-fitting gloves changed frequently [1]. ESHRE's 2026 good practice recommendations for the IVF laboratory repeat these points in brief (tubing in a separate area of the laboratory or in a separate tubing laboratory; a clean gown, hair cover, mask over nose and mouth and preferably shoe covers; gloves worn throughout and changed frequently) and refer to the 2020 PGT recommendations for the detail [56].

Contamination is not hypothetical. In a retrospective analysis of 57,292 clinical PGT-A samples from 32 IVF clinics, using a method that genotypes more than 4,000 SNPs and can detect foreign DNA, contamination was found in 0.45% of trophectoderm biopsies (256 of 57,292), about one in 222. Rates differed significantly between clinics, below one in 1,000 at some and more than ten times higher at others. Where a second, uncontaminated sample was available, comparison showed that the chromosomal classification could have been wrong in 69 of 155 embryos (44.5%) had the contamination gone undetected. The authors, all affiliated with the genetics laboratory that performed the testing (Juno Genetics), proposed that the contamination rate be treated as a key performance indicator for clinics offering PGT-A [57].

The consequence for the bench is twofold. The laboratory should know whether its genetics provider's method detects contamination; and it should ask for, and track, its own contamination rate alongside amplification failure and inconclusive rates.

5.5 Witnessing and traceability

A PGT cycle adds a new identity problem: each biopsy must remain linked to the one embryo it came from, through biopsy, tubing, vitrification, shipping, analysis and reporting. ESHRE's PGT-specific biopsy recommendations ask for witnessing immediately after biopsy to confirm that embryo and sample numbers match, at tubing to confirm that the sample identity matches the tube, when the embryo is placed and labelled in the culture dish or cryopreservation device, and after the diagnostic results are issued, again at warming and at selection of the embryo for transfer, and state that, to ensure embryo-based traceability, a witness is mandatory even when an electronic witnessing system is in place, that is, a human check in addition to the electronic one [1].

Electronic witnessing systems, which reduce mismatch risk in standard IVF, may not, unless configured for it, follow individual embryos and their sample tubes through a biopsy session, so in this reference's reading manual double witnessing remains the default at these steps. ESHRE's general guidance is less prescriptive than its PGT-specific one: its 2026 good practice recommendations for the IVF laboratory strongly advise a double check by a second person and/or an electronic identification system at critical steps, ask that manual witnessing be independent verification by two qualified staff rather than passive confirmation, and ask that any witnessing discrepancy trigger a predefined escalation, with the procedure halted, documented and reviewed by a supervisor [56]. For biopsy and tubing, ESHRE's PGT-specific requirement for a human witness, set out above, is the more specific of the two. In the United States, the 2026 ASRM and SRBT committee opinion on witnessing and protocol deviations allows critical steps, embryo biopsy among them, to be witnessed by a second person and/or a certified electronic system; it requires a second manual witness even with electronic witnessing when gametes are mixed at insemination or ICSI and when embryo and patient identity are confirmed before the transfer catheter is loaded, keeps discards manual, and for biopsy asks for checks that trophectoderm cells go into the correct tubes [58]. As this reference's recommendation, a programme using electronic witnessing should be able to show which embryo-to-sample transitions (biopsy droplet, tube, rack, cryo-device) its system actually witnesses and which rest on a human check. A failure mode and effects analysis of a large blastocyst PGT programme identified 44 failure modes, six of them with a moderate risk priority number; corrective measures reduced the estimated risks when the analysis was repeated six months later [59]. The method is transferable: any laboratory introducing or changing a PGT workflow can map its own steps the same way.

5.6 Vitrification after biopsy, and re-biopsy

Biopsied blastocysts must be vitrified individually on a properly labelled device, with witnessing, and as soon as possible after biopsy, before the blastocoel re-expands, particularly when the blastocyst is fully hatched [1]. In 2,129 single vitrified-warmed blastocyst transfers at two centres, blastocyst morphology, the day of full development and artificial shrinkage (by laser or as a result of biopsy) were the pre-vitrification factors most strongly associated with post-warming behaviour and implantation, whereas trophectoderm biopsy itself was not; the study was not randomised and the PGT-A and non-PGT populations differed [60].

When the first analysis is inconclusive, the embryo can be warmed, biopsied again and re-vitrified. In 8,990 trophectoderm biopsies at six centres analysed by one genetics laboratory, 2.5% gave an inconclusive result (228 of 8,990: 176 amplification failures and 52 non-concurrent results), associated with the IVF centre and the day of biopsy, and among amplified samples with the number of cells taken. Of 213 such blastocysts warmed for re-biopsy, 206 survived (96.7%), 107 of 206 were euploid (51.9%) and were re-vitrified, all 49 warmed for transfer survived, and 19 of 49 single transfers resulted in a live birth (38.8%), with no reported obstetric or perinatal complication. The authors noted that the sample was too small to assess obstetric and perinatal outcomes with confidence [61]: with 19 births, an absence of reported complications is compatible with a true rate of up to about 18% (upper limit of a two-sided exact 95% interval for 0 of 19, calculated for this reference). ESHRE advises that re-biopsy be considered only after a failed, incomplete or inconclusive diagnosis, through the original zona opening, after allowing time for the warmed blastocyst to re-expand, with immediate re-vitrification afterwards, and in accordance with local law [1]. A later systematic review and meta-analysis of 10 studies, with evidence of low to moderate certainty, found lower live birth per transfer both after a second vitrification-warming with a single biopsy (odds ratio 0.56, 95% CI 0.38 to 0.81; six studies) and after a second biopsy with a second vitrification (odds ratio 0.51, 95% CI 0.34 to 0.77; six studies). Miscarriage was higher in both groups, while cryosurvival and perinatal outcomes were within expected norms, and the authors concluded that the risk should be weighed against the benefit of a diagnosis for each couple [62].

5.7 Which embryos to biopsy

ESHRE asks laboratories to set written biopsy criteria before clinical use and apply them consistently [1]; in principle, biopsy is offered to every blastocyst that meets them and could be transferred. Two groups are often excluded by habit rather than evidence. Poor-quality blastocysts (graded below BB in the Gardner and Schoolcraft system) have a lower chance of euploidy and implantation than their siblings. In an observational cohort of 2,757 PGT-A retrievals at one centre (mean maternal age 39.6 years), including them increased the number of cycles with at least one biopsied blastocyst by 12.4% and the number with at least one live birth by 2.6%; these are relative increases in counts of cycles, not percentage-point gains in live birth per retrieval. Among 195 transferred euploid poor-quality blastocysts, live birth was 10.8%, against 44.6% for 1,697 better-graded euploid blastocysts, and the 18 women whose only live birth came from a poor-quality blastocyst were mostly older than 42 or had few other blastocysts [42]. The study shows that such embryos can add births; it does not show that every low-grade blastocyst should be biopsied regardless of technical suitability. Blastocysts reaching full development only on day 7 are a separate, overlapping group (29.4% of the poor-quality blastocysts in that cohort developed on day 7) and are biopsied in many programmes; ESHRE's biopsy recommendations note that biopsy slots should be planned from day 5 to day 7 [1], and a 2023 commentary summarised the case for their clinical use [63]. A 2026 meta-analysis of 17 retrospective studies found that, among PGT-A euploid blastocysts (six studies), day 7 transfers had a lower live birth rate than day 5 transfers (odds ratio 0.21, 95% CI 0.17 to 0.28; 364 against 5,924 transfers pooled from retrospective studies, with the transfer rather than the patient as the unit) and day 6 transfers (odds ratio 0.36, 95% CI 0.28 to 0.46; 364 against 3,782), and a higher miscarriage rate than day 5. Day 7 blastocysts were also less often euploid, and the authors concluded that day 7 blastocysts may still add to cumulative live birth, notably for patients with a poorer prognosis [64]. Both bodies of evidence are retrospective and come from selected populations.

5.8 Workload and staffing

PGT adds work at every stage: extended culture, biopsy across three days, tubing with witnessing, individual vitrification, shipping and result reconciliation. In the single-centre experience cited in Section 2.2, the growth of PGT to most retrievals was matched by training a second, third and fourth biopsy practitioner, and staffing was planned with a published ART staffing model [19]. Laboratories planning PGT should budget for training and for the time to maintain competency in more than one operator.

6.1 How each method works

A trophectoderm biopsy contains about 6 to 7 picograms of DNA per diploid cell, some tens of picograms in a typical five-to-ten-cell sample, too little to measure without first amplifying it, either across the whole genome or, for targeted methods such as qPCR (below), at selected loci only [65]. The genetics laboratory first lyses the cells to release the DNA and then, for most methods, copies the whole genome many times over (whole-genome amplification, WGA) before measuring it [8]. Three families of WGA chemistry are in use: PCR-based methods such as degenerate oligonucleotide-primed PCR; multiple displacement amplification (MDA), an isothermal reaction driven by a strand-displacing polymerase; and hybrid methods, including multiple annealing and looping-based amplification cycles (MALBAC) and the methods of widely used commercial kits [65, 66]. Every chemistry distorts the template: one allele can be over-amplified or lost altogether (allele drop-out), coverage of the genome is uneven and partly dependent on its GC content, and strand-displacing polymerases create chimeric molecules that join unrelated sequences [67]. In a 2025 comparison of six methods on 230 single cells from two cancer cell lines and one fibroblast cell line (a cell-line model, not embryos), MDA gave higher yields and broader coverage, non-MDA methods amplified more uniformly and reproducibly, and the most accurate copy-number detection came from a non-MDA method; no method was best on every measure [67]. Amplification also becomes unreliable as the starting amount falls: in a bacterial-DNA model, three chemistries performed well from 15 pg or more but were unreliable from 1.5 pg or less, and the two chemistries with PCR cycles (MALBAC and a PCR-based kit, both classed there as PCR-based) were particularly prone to failure through contamination [68]. ESHRE notes that a trophectoderm sample of 5 to 10 cells in theory allows a more reliable diagnosis [8]. The alternative to WGA is to amplify only selected targets by PCR and quantify the products by qPCR or by sequencing, which gives more depth at the targets at the expense of breadth of coverage [65].

Low-pass NGS. The amplified DNA is converted into a barcoded sequencing library, so that many biopsies share one run, and sequenced into millions of short reads that are mapped to the human reference genome [8]. A depth of about 0.1× or less is enough for copy number, hence “low-pass” sequencing; the mapped reads are counted in windows (bins) along each chromosome [65]. Because WGA distorts the number of reads per chromosome reproducibly, presumably through differences in base composition and chromatin structure, the counts are normalised against chromosomally normal reference samples; in one early validation a disomic chromosome gave a test-to-reference ratio of 0.7 to 1.2, and ratios above 1.2 or below 0.7 were read as gain or loss [69]. In a disomic background, a trisomy present in every biopsied cell gives an expected ratio of 1.5 (three copies against two) and a uniform monosomy 0.5, so the profile sits at copy number 3 or 1 instead of 2 (measured ratios scatter around these expectations, which is why the calling thresholds above sit well inside them); a value of 2.5 is what a 50:50 mixture of disomic and trisomic cells would give, and in 2020 many laboratories treated the 20% to 80% span between whole numbers as consistent with mosaicism [65]. An intermediate value is not proof of mosaicism. Amplification or sequencing noise and poor biopsy technique can produce one [65]; so can contaminating DNA [57]. A segmental change moves only the bins it covers, usually to a whole-number level; averaged over the whole chromosome, or when close to the method's resolution, it can also look intermediate (an explanatory point, not a cited finding). Laboratories turn these values into reporting categories such as “mosaic” or “intermediate copy number”, each with its own thresholds (Sections 6.3, 7 and 8), and treating intermediate copy number alone as evidence of mosaicism has been argued to misestimate its prevalence [37].

Array CGH. Amplified sample DNA and reference DNA are labelled with different fluorochromes, for example green and red, and hybridised competitively to a slide carrying probes for defined chromosomal regions; the fluorescence ratio at each probe, expressed as a log ratio, indicates loss or gain in the sample [8].

SNP arrays and SNP-based sequencing. These read the genotype at many polymorphic sites. SNP arrays record two measures at each: total signal intensity, as a log R ratio, which tracks copy number, and the B-allele frequency, the share of signal from one of the two alleles; ESHRE recommends that SNP-array diagnoses of imbalance be supported by both [8]. Sequencing-based methods use the counterparts, normalised read depth and allelic fractions, in platform-specific implementations. In principle, heterozygous sites in a disomic region sit near a B-allele frequency of 0.5 and a trisomy moves them towards one-third and two-thirds; the same split on every chromosome with an unchanged log R ratio is the expected signature of triploidy (Section 12.2). ESHRE notes that SNP arrays and NGS-based haplotyping can identify polyploidy and haploidy [7]; one targeted method that analyses allele frequencies was validated for ploidy on 164 embryonic samples of known status, 9 of them triploid and 8 haploid [70]. Genotyping can also reveal contaminating DNA [57], and with parental genotypes the inherited haplotypes can be traced, showing which parent contributed an extra chromosome and whether the error arose in meiosis or mitosis [37, 71]. WGA distorts allele ratios in single cells, which is why these methods combine many sites and, often, phased parental genotypes [72]. The same haplotyping underlies PGT-M (Section 11.1).

qPCR. Selected loci on every chromosome are amplified by PCR without WGA and quantified in real time relative to a reference [65, 73].

FISH. Biopsied cells are fixed on a slide, chromosome-specific fluorescent probes are hybridised to the nuclei, and the signals are counted under a microscope [8]. ESHRE does not recommend FISH for PGT-A because only a subset of chromosomes can be examined and comprehensive methods for all 24 are available [8]; it was the method of the cleavage-stage trials that did not improve live birth (Section 10) [12, 13].

Table 1 compares the methods. What any one laboratory's version detects is set by its own validation (Section 6.4).

6.2 How the platforms compare

The methods of Section 6.1 have been validated to different depths. In its prospective, randomised and blinded validation, qPCR assigned the same 24-chromosome diagnosis as SNP array analysis to 70 of 71 discarded blastocysts (98.6%) [73]. Array CGH, the first widely used comprehensive method, has largely been replaced; WGA followed by low-pass NGS is now the workhorse of the field, and some targeted methods sequence selected amplicons or polymorphisms instead [65].

Methods differ in resolution, in their sensitivity to amplification artefacts and in what they can detect beyond whole-chromosome copy number. Methods that avoid whole-genome amplification were argued to offer higher reliability at lower cost and turnaround time [15]. When aneuploid blastocysts diagnosed by array CGH were re-biopsied and blinded for analysis by qPCR, 18.3% (22 of 120) gave a discordant result for at least one chromosome; resolving the discordances by SNP array showed that array CGH produced more of the errors, mostly false positives, with a discordant aneuploidy call rate per chromosome of 5.7% against 0.6% for qPCR [74]. The authors concluded that blastocyst-stage diagnosis was highly reliable with both methods but that array CGH can be expected to give a higher aneuploidy rate than other contemporary methods; the practical lesson is that platforms have measurable and different error profiles that a laboratory should know before choosing.

6.3 What the platform cannot see

Each method has blind spots that the embryologist should know when reading a report. What a genotype-based method adds depends on what it was validated to detect, which the laboratory's validation summary should state:

• Ploidy. Copy-number methods measure each chromosome relative to the others. A triploid or haploid embryo, in which every chromosome is equally over- or under-represented, can look normal; an unbalanced sex-chromosome ratio, as in 69,XXY and 69,XYY, can reveal those forms of triploidy [8], but this is a suggestive pattern, not a validated ploidy classification. A validated ploidy classification needs a method that reads allele ratios at polymorphic sites and was validated on embryos of known ploidy (Section 12) [70]; saying which parent contributed the extra set is a further claim, needing parental genotypes and its own validation. A uniparental diploid embryo, with both sets from one parent, also has a normal copy-number profile; genotyping shows it as genome-wide loss of heterozygosity when the two sets are identical, and needs parental genotypes otherwise (Section 12.2).
• Small deletions and duplications. In 2017 the resolution of methods applied to single cells was described as usually above 10 megabases, which excluded most recurrent microdeletion and microduplication syndromes [75]; resolution today is platform-specific and set by each laboratory's own validation [8]. Targeted approaches can now detect some common microdeletions (Section 9).
• Uniparental disomy and balanced rearrangements. Uniparental disomy, in which both copies of a chromosome come from one parent, and balanced rearrangements are invisible to copy-number counting [8]. Some genotype-based methods can assign the parental origin of chromosomes [37] and, with family references, distinguish an embryo carrying the balanced form of a familial rearrangement from a non-carrier [8], but only where validated for that purpose.
• Low-level mosaicism. An abnormal cell line present in only a small fraction of the biopsy produces a signal that cannot be distinguished from technical noise; in ESHRE's survey, centres most often placed the lower cut-off at an estimated 20% or 30% mosaicism, below which a biopsy is reported as euploid [76].

6.4 Validation: analytical and clinical

Two kinds of validation are needed and are often confused. Analytical validation shows that the test measures what it claims, for example by testing cell lines of known karyotype, mixtures of cells, or several biopsies of the same embryo. Such models can expose the cost of a threshold before it reaches patients: in a randomised, blinded cell-line mixture model of mosaic six-cell biopsies, qPCR and NGS at their default settings gave no false positives at any level of mosaicism, but published custom NGS criteria raised sensitivity at the price of a 33% false-positive prediction of aneuploidy [77]. Clinical validation shows that a given result predicts what happens after transfer and requires outcome data; for abnormal results, prospective blinded non-selection studies give the least biased estimate, while intended transfers and retrospective re-analyses add complementary evidence from selected embryos and patients [78]. Concordance with the rest of the embryo or the fetus, prediction of reproductive outcome and benefit to patients are separate questions, each needing its own evidence. A recent review of reporting practice argued that each reported category should be clinically validated against a defined reproductive outcome before it is used to decide an embryo's fate [79].

Accuracy is also not fixed once a test is validated. Performance can drift after validation; continued surveillance, described as a phase 4 of validation, is discussed in Section 15.2 [80].

7.1 The categories

Reports vary between laboratories, but most place each embryo in one of the categories below. The categories are not equally well supported, and the strength of the evidence behind each should govern how much weight it carries in decisions.

7.2 How reliable is an aneuploid result?

The question that matters most for embryo disposition is how often an embryo reported as aneuploid would in fact have produced a healthy baby. It is answered with least bias by transferring embryos without knowing their result; transfers made knowing the result, such as the cohort described below, are informative but selected. The first prospective, blinded non-selection data, from 255 embryos (113 cleavage-stage and 142 blastocyst biopsies) selected for transfer without influence from the result, found that 96% of embryos predicted aneuploid failed to sustain implantation, a clinical error rate of 4% for an aneuploid designation [84]. In a later, larger prospective, blinded, multicentre non-selection study, all usable blastocysts were biopsied, the single best-looking blastocyst was transferred before analysis, and results were unblinded after the outcome was known; across 484 single frozen transfers in 402 patients, the clinical error rate of an aneuploid diagnosis, defined by delivery, was 0%: no embryo called aneuploid delivered [83]. Failure to deliver does not by itself verify a chromosome diagnosis. A later review gives the denominator as 102 transferred embryos with a uniform aneuploidy, none of which delivered [41]; with 102 embryos, the data are compatible with a true delivery rate of up to 3.6% (exact 95% CI 0% to 3.6%, calculated for this reference). Sustained implantation showed no detected difference from a non-randomised, age-matched group whose embryos were not biopsied (47.9% against 45.8%) [83].

Reviewing the non-selection evidence, Capalbo and colleagues concluded that embryos with a uniform whole-chromosome aneuploidy very rarely result in the live birth of a healthy baby, and that their transfer exposes women to substantial risks of miscarriage and of chromosomally abnormal pregnancy, whereas embryos with low-range whole-chromosome mosaicism have shown reproductive capabilities "somewhat equivalent" to uniformly euploid embryos, in the review's words. The practical conclusion they drew is that careful differentiation between uniform and mosaic aneuploidy is critical, both when counselling patients and when reporting research [41]. The same group argued the point directly in a letter responding to the cohort study described next [85], and its authors replied [86].

A different view comes from a centre that accepted embryos classified as abnormal and refused transfer elsewhere. Among 69 couples who moved 444 such embryos, 50 patients had 141 embryos transferred in 57 cycles (141 in the abstract's design section and in its breakdown of 102, 30 and 9 embryos; 144 in its results section), with 8 live births (8 of 57 transfer cycles, 14.0%), 11 miscarriages and no voluntary terminations. Most transferred embryos had one or two abnormalities, and several had been called mosaic or had no result. Because most transfers contained several embryos, a birth cannot be attributed to a particular embryo or abnormality. One child was born with a segmental duplication and needed surgery for coarctation of the aorta. The authors argued that abnormal results lead clinics to refuse transfers that could succeed. They acknowledged that the patients were self-selected and may not represent PGT-A patients in general [87]. The study did not separate uniform from mosaic results in its headline figures, which is the distinction the non-selection literature rests on. Re-analysing the cohort, Capalbo and colleagues counted one live birth among 106 transferred embryos with at least one uniform whole-chromosome aneuploidy (the review counts embryos, not the 57 transfer procedures) [41], and a 2026 cohort from the International Registry of Mosaic Embryo Transfers (IRMET), which kept the categories apart, found no live birth after 168 whole-chromosome transfers against 17.1% after 82 segmental transfers, in a selected cohort with limited newborn follow-up (Section 9.1) [78]. A pooled live-birth figure for abnormal embryos mixes groups with very different prognoses.

7.3 How accurate is the test overall?

A 2025 meta-analysis of 109 studies compared biopsy results with reference standards. Against the whole embryo or the inner cell mass (40 studies), the positive predictive value of an aneuploid result was 89.2% (95% CI 83.1 to 94.0) and the negative predictive value of a euploid result was 94.2% (95% CI 91.1 to 96.7; I² 42%, where I² is the percentage of variation between studies attributable to heterogeneity rather than chance). Against pregnancy outcome (43 studies), the misdiagnosis rate after euploid transfer was 0.2% (95% CI 0.0 to 0.7; I² 65%; prediction interval 0% to 3.4%), which the authors calculated as false-negative results divided by all euploid embryos transferred, because a per-pregnancy denominator could not be established. Ascertainment was incomplete: of 10,641 pregnancies from 20,196 euploid embryos transferred, 1,367 had their status verified by the methods the review accepted (prenatal or postnatal genetic testing, products of conception or physical examination, the last with different detection limits), and 22 microdeletions below the detection limit found in products of conception were counted as incidental findings rather than misdiagnoses. Embryos that never implanted and pregnancies never tested count in the denominator as correctly diagnosed, so 0.2% is a pooled observed discordance per embryo transferred; the probability that a pregnancy after a euploid result is affected cannot be estimated reliably from these data. The estimate is an observed misdiagnosis rate for what PGT-A is designed to detect, not a sensitivity for all fetal genetic abnormalities, and it does not replace the offer of prenatal screening or diagnosis (Section 11.4). For mosaic results the picture was very different: the positive predictive value of a confirmatory mosaic or aneuploid finding was 52.8% (95% CI 37.9 to 67.5), and the rate for mosaic transfer with a confirmatory euploid pregnancy outcome was 21.7% (95% CI 9.6 to 36.9), with very high heterogeneity [81].

Two studies that re-sampled the embryo itself give a broadly consistent picture for the trophectoderm, though not for day 3 biopsy. Re-analysis of the inner cell mass and three trophectoderm sections of 70 blastocysts by FISH for nine chromosomes gave a sensitivity of 98.0% and specificity of 100% per embryo for the original array CGH trophectoderm diagnosis (the embryos were chosen by their original result, 50 aneuploid and 20 diploid, so the specificity rests on about 20 embryos; as an illustration, 20 of 20 would have an exact 95% CI of 83% to 100%, calculated for this reference). Mosaic chromosome errors of any kind were seen in 11 of 70 embryos (15.7%), diploid-aneuploid mosaicism in only 2 of 70 (2.9%) [3]. In 84 embryos biopsied on day 3 and again at the trophectoderm and inner cell mass, agreement between trophectoderm and inner cell mass was "almost perfect" on the conventional kappa scale (kappa 0.857), although the trophectoderm still gave a false-positive aneuploid call against the inner cell mass in 7.5% of comparisons, and agreement between a day 3 blastomere and the inner cell mass was only "moderate" (kappa 0.553). Only 50 of 84 embryos (59.5%) were concordant across all three biopsies, and the authors concluded that the discordance underlines the limitations of PGT-A. Embryos selected for transfer were not re-biopsied, a possible source of bias [48]. Re-analysis with NGS separates whole-chromosome from other findings. In 100 donated blastocysts called uniformly aneuploid at one centre, the inner cell mass was aneuploid in 90 of 93 with a whole-chromosome aneuploidy (96.8%) but in only 3 of 7 with segmental aneuploidy alone (42.9%) [88]. In a blinded prospective cohort of 300 donated blastocysts re-biopsied on a targeted NGS platform, euploid and whole-chromosome aneuploid results were reconfirmed in more than 99% of re-biopsies, against 15% to 18% for whole-chromosome or segmental mosaic results and 56.6% for segmental aneuploidy [89].

8.1 Biology and measurement are different things

True mosaicism, two or more chromosomally distinct cell populations derived from one zygote, exists and arises from mitotic errors after fertilization. A "mosaic" result is something else: an intermediate copy-number value in a biopsy of a few cells. That value can reflect a mixture of normal and abnormal cells in the biopsy, but it can also reflect amplification noise, a biopsy that included a damaged cell, or a uniform aneuploidy measured imprecisely. Genotype studies disagree on how common the last explanation is, and a meiotic origin can also mean a meiotic error partly corrected after fertilization. In a retrospective single-centre cohort, SNP genotyping of the same amplified DNA found an error of meiotic origin in 46 of 141 embryos reported as mosaic by NGS (32.6%) [90], whereas in a larger single-centre cohort tested on a SNP array a meiotic origin was found in only 15 of 384 blastocysts with a mosaic result (3.9%), more often with high-level than low-level values (7.2% against 2.0%) and in women of advanced age [91]; the proportion depends on platform, thresholds and cohort. The cell-division assay described in Section 3.3 classified only 2% of 2,277 blastocysts as mitotic aneuploid, an assay-defined figure for one cohort [37]. Reported rates of mosaicism in preimplantation embryos have ranged from 4% to 90%, while true fetal mosaicism is found in fewer than 0.5% of clinical pregnancies, a contrast that led Capalbo and colleagues to argue that technical variation may be one cause contributing to overestimation of the incidence in embryos [92]; developmental selection between the blastocyst and the pregnancy, and the different tissues sampled, may also contribute to the contrast. Capalbo's listed affiliation was then a PGT laboratory, and co-authors of later papers cited in this section list PGT providers as their employers or affiliations. In their view a diagnosis of certainty for mosaicism from a single trophectoderm biopsy is conceptually impracticable, because technical and biological sources of error cannot be separated in one sample [93]. Whole-embryo single-cell data point the other way on the biology: aneuploidy inferred from gene expression and allelic imbalance in published single-cell RNA sequencing of 74 human embryos found at least one putative aneuploid cell in 59 (80%) and mitotic aneuploidies in 55 (74%), and the authors concluded that low-level mosaicism is a common feature of early human development [94].

8.2 What a mosaic biopsy predicts about the rest of the embryo

The most direct test is to take several biopsies from the same embryo. In 73 unselected embryos sampled at four trophectoderm sites and the inner cell mass (365 biopsies), when mosaicism affected fewer than 50% of the cells in one trophectoderm biopsy, only 1% of the aneuploidies were found in any other part of the embryo, which the authors interpreted as low- and medium-level mosaicism mostly arising after the trophectoderm and inner cell mass have separated and staying confined to the trophectoderm [95]; removal of abnormal cells in the biopsy itself or their depletion during development would give a similar pattern, and the data cannot show which operated. A mouse model suggests one possible mechanism: in chimeric mouse embryos, aneuploid cells were eliminated by apoptosis in the fetal lineage and showed severe proliferative defects in the placental lineage, and mosaic mouse embryos had full developmental potential provided they contained sufficient euploid cells [96].

Thresholds matter. In 117 donated blastocysts each divided into the inner cell mass and four trophectoderm pieces (585 samples), applying a 30% to 70% threshold to the original biopsy identified a true mosaic configuration in only 8.4% of abnormal patterns, and 87.3% of predicted mosaic patterns were confined to the single original specimen. Widening the range to 20% to 80% doubled the mosaicism rate per biopsy (from 20.2% to 40.2%) without finding any additional true mosaic patterns, and increased the proportion of false-positive mosaic calls from 57.8% to 79.5%. A binary classification with a single 50% cut-off gave the highest sensitivity and specificity. Uniform whole-chromosome aneuploidies were confirmed in every further biopsy in 46 of 47 cases (97.9%) [97]. Testing the re-biopsies on a different platform gives a similar picture. In a blinded prospective cohort, 100 donated blastocysts were each resampled four times (three trophectoderm re-biopsies and the remaining embryo) and tested on a SNP array platform that combines copy number with allele ratios: NGS whole-chromosome aneuploidy was reconfirmed in 95% of the 40 such embryos, whereas 35% of the 20 whole-chromosome mosaic and 65% of the 20 segmental mosaic embryos were called euploid on all conclusive re-biopsies [98].

8.3 What happens after transfer

Two kinds of data are available, and they give different answers.

Blinded non-selection data. In a double-blinded prospective non-selection trial, embryos were transferred without the mosaic classification being known to clinicians or patients. The trial was designed to test non-inferiority of live birth, with a margin of 7.5 percentage points, not equivalence. Live birth was 43.4% (210 of 484) for euploid, 42.9% (121 of 282) for low-grade and 42.0% (55 of 131) for medium-grade mosaic embryos. With the two mosaic groups combined, non-inferiority was met (difference, euploid minus mosaic, 95% CI −5.7 to 7.3 percentage points, upper limit inside the 7.5-point margin); neither mosaic group was large enough to show it on its own (for the medium-grade group the interval extends to about 11 points, calculated for this reference). No difference in miscarriage was detected (adjusted odds ratio 0.89, 95% CI 0.50 to 1.55), but the interval is too wide to show non-inferiority against the 2-point margin the authors set for miscarriage. The unit of analysis was the transferred embryo, and only the best embryo by morphology was transferred. The authors described the outcomes as equivalent, and no significant differences were detected in obstetric and neonatal outcomes. Genetic follow-up covered only a subset: 4 of 52 miscarriages were karyotyped, all euploid; 26 of 388 sustained pregnancies (6.7%) had prenatal diagnosis, 9 of them after mosaic transfers, all with a euploid fetal result; and 38 newborns from mosaic transfers (9.8% of all newborns in the study) were genotyped with SNP arrays, with no mosaicism or uniparental disomy found. One child from a medium-grade mosaic transfer was born with Beckwith-Wiedemann syndrome caused by a methylation (imprinting) defect, which the authors noted is not a target of PGT-A and not due to chromosomal mosaicism. Numbers this small cannot exclude rare fetal mosaicism or uniparental disomy. The trial included only whole-chromosome results in the 20% to 50% range, and the authors cautioned that its findings cannot be extended to segmental mosaicism [95].

A larger multisite double-blinded study of 9,828 single-embryo transfers in the United States, with independent validation on 5,487 euploid transfers in Europe, found that 84.7% of embryos had no intermediate copy number, 8.8% had segmental and 5.6% whole-chromosome intermediate copy number. Live birth was modestly lower with intermediate copy number (53.2% against 60.0%; adjusted odds ratio 0.79, 95% CI 0.70 to 0.89), driven mainly by the small group with high-level values (odds ratio 0.61, 95% CI 0.49 to 0.75). Adding intermediate copy number to a model of clinical and embryological factors did not, however, improve prediction (area under the curve 0.552 against 0.555, where 0.5 is no better than chance); live birth after a euploid or mosaic transfer was poorly predictable with or without it. Miscarriage, obstetric and neonatal outcomes were reported as comparable. The authors concluded that, in their setting, reporting putative mosaicism provides no clinical benefit [99]. The study measured added discrimination in a prediction model; it did not randomise reporting or transfer policies, and a lack of added discrimination across all transfers does not by itself show that the information has no use in any individual decision. A 2026 review from the same group argued that the residual signal in high-level categories is increasingly attributable to misclassified uniform aneuploidy rather than to mosaicism itself [100].

Retrospective re-analyses. Two retrospective re-analyses of embryos transferred as euploid, a design that also avoids selection by indication, found lower outcomes. Re-reading array CGH profiles identified 102 mosaic embryos, 101 of them estimated at no more than 50% mosaicism, and 268 matched euploid control embryos; live birth per transfer cycle was 46.6% against 59.1% (odds ratio 0.60, 95% CI 0.38 to 0.95) [101]. Among 377 single euploid transfers, the 20 embryos found afterwards to carry segmental mosaicism had live birth of 30.0% against 53.8%, with an odds ratio of 0.34 (95% CI 0.13 to 0.92) after adjustment for age and day of biopsy [102].

Data from embryos transferred knowing they were mosaic. The first report of healthy babies after transfer of embryos classified as mosaic was published in 2015 [103]. A compiled analysis of 1,000 mosaic and 5,561 euploid transfers found lower implantation (46.5% against 57.2%) and ongoing pregnancy or birth (37.0% against 52.3%), and more miscarriage (20.4% against 8.6%), for mosaic embryos, with worse outcomes for whole-chromosome mosaicism, levels of 50% or more and more chromosomes affected [104]. In a prospective study of 77 women with no euploid embryo, mosaic embryos below 50% had a live birth rate of 42.2%, not significantly different from 46.6% for 251 euploid control blastocysts, whereas those at 50% or more had a live birth rate of 15.2% [105]. A later multicentre analysis by an overlapping group of authors found miscarriage in 22.2% of implanted mosaic embryos against 8.9% of euploid, similar birthweights and gestational lengths among 488 infants from mosaic transfers, and prenatal results reflecting the embryonic mosaicism in 3 of 250 pregnancies with a prenatal test (1.2%) [106]. A 2026 meta-analysis of 1,904 mosaic and 9,332 euploid transfers found lower live birth or ongoing pregnancy after mosaic transfer (risk ratio 0.74, 95% CI 0.64 to 0.85; I² 78%, 95% prediction interval 0.41 to 1.34), with segmental mosaicism close to euploid outcomes (risk ratio 0.90, 95% CI 0.81 to 1.00) and whole-chromosome mosaicism clearly worse (risk ratio 0.60, 95% CI 0.53 to 0.68). It noted that the one double-blind trial had shown non-inferiority (risk ratio 0.98) [107].

The disagreement is real and one likely source is identifiable. In unblinded studies, mosaic embryos are usually transferred only when no euploid embryo is available, so they may come from patients with poorer prognosis, and lower outcomes may reflect the patient's prognosis rather than the embryo; the 2026 meta-analysis attributed much of the difference to this confounding by indication [107]. Blinded designs remove that bias but answer the question for one laboratory's platform and thresholds. Both kinds of study agree that birthweight, gestational length and neonatal outcomes among live-born infants are broadly reassuring and that a small fraction of pregnancies carry the mosaicism forward, which is why prenatal diagnosis is offered after a mosaic transfer; unblinded registry data also show more miscarriage after a mosaic embryo implants [76, 106].

8.4 What the guidelines say

An ESHRE working group surveyed 239 centres and found that 80% of centres taking three or more cells reported mosaicism, although of the 133 centres answering the validation question only 89 (66.9%) had validated their technology (59 of 67 testing in-house, 30 of 66 outsourcing) and only 55 of those 89 (61.8%) had validated it specifically for mosaicism; designation criteria and transfer policies varied widely. It formulated thirty good practice recommendations covering validation, risk assessment, designation and reporting, transfer policy, prenatal testing and follow-up [76]. The variation reaches patients: in a multicentre cohort of 36,395 blastocyst biopsies tested by NGS at eight providers, the seven providers that reported mosaicism did so in 3.1% to 25.0% of biopsies, and adjusted euploidy rates were highest in the providers reporting the least mosaicism [108]. The 2021 PGDIS position statement stated that recent data suggest mosaic embryos can be transferred without added risk of abnormal birth outcomes, a reassurance resting on reported births and the few pregnancies and newborns genetically tested (Section 8.3), not on systematic offspring follow-up; it added that they may be associated with increased implantation failure and miscarriage, with higher levels of mosaicism appearing less favourable [109]. ASRM, a US body, revised its committee opinion on the clinical management of mosaic results in 2023, with a corrigendum in 2026 [110, 111]. Its practical advice is that each IVF programme have a policy on the reporting, transfer, storage and disposition of mosaic embryos, known to staff and shared with patients before testing; that clinicians understand how their reference laboratory reports mosaicism; that patients considering a mosaic transfer be offered consultation with a genetic counsellor specialising in PGT; and that outcomes after mosaic transfer be tracked and published. It states that no evidence-based method yet ranks mosaic embryos or predicts an adverse outcome, and puts fetal aneuploidy related to the mosaic result at likely below 1% on current data, citing a 2021 review in which fewer than 1% of 2,759 mosaic embryos transferred led to an ongoing aneuploid pregnancy related to the original result [110]. That denominator is transferred embryos, most never tested in pregnancy; the registry's 3 of 250 (Section 8.3) is pregnancies with a prenatal test. There is no common denominator from which to quote a personalised fetal risk.

For the laboratory, several things follow. A mosaic result should not be treated as equivalent to a uniform aneuploidy in laboratory disposal or research policies, and patients deciding what happens to a mosaic embryo should be told the difference. Whether to report mosaicism at all, and with what thresholds, is a decision for the genetics laboratory that should rest on its own validation; the IVF centre should know those thresholds and agree in advance how each category will be handled (Section 15.1). For low-range results ESHRE's good practice recommendations go further: discarding low-range mosaic embryos is not acceptable; a new stimulation cycle is not recommended while such embryos are available for transfer; when choosing among euploid and low-range mosaic embryos, the biopsy result should be weighed together with morphology rather than on its own; and re-biopsy of a mosaic embryo to confirm its status is not recommended, because there is no evidence that it has diagnostic value and it may harm further development. ESHRE could make no specific recommendation for high-range mosaic embryos and regards thorough genetic counselling as a prerequisite for their transfer; after PGT-M or PGT-SR, it accepts a new cycle to obtain embryos with an accurate diagnosis [76]. ASRM's emphasis differs: it considers it reasonable, depending on the patient's preference, to transfer euploid embryos before mosaic ones to reduce the burden of uncertainty, while noting that outcome data do not necessarily support this as a universal rule and that morphology may also be considered [110]. When a mosaic embryo is considered for transfer, three steps stay distinct: the laboratory documents technical suitability and the scope of testing; the clinical and genetics teams explain which options applicable law and clinic policy allow and what is known about outcomes; and the patient decides among those options after genetic counselling. ESHRE recommends that counselling on prenatal diagnosis be offered after any pregnancy following PGT, but states that a low-range mosaic result in a trophectoderm biopsy is not in itself an indication for invasive prenatal diagnosis [76]. The choice of prenatal test matters: chorionic villi and the cell-free DNA measured by standard noninvasive prenatal testing (NIPT) come largely from the placenta, which, like the trophectoderm, can carry a mosaicism absent from the fetus (confined placental mosaicism); ESHRE, ASRM and the Italian Society of Human Genetics differ on which test to use after a mosaic transfer, and Table 4 (Section 11.4) sets their positions side by side [76, 82, 110].

9.1 Segmental findings

A segmental aneuploidy is a gain or loss of part of a chromosome. A review found them in about 8.15% of preimplantation embryos, compared with 3.55% of prenatal diagnosis samples, and a mean confirmation rate on re-biopsy only slightly above 30% [112]. Multifocal analysis of 78 donated blastocysts (four trophectoderm pieces and the inner cell mass each, 390 biopsies) led its authors to attribute about 70% of segmental changes to mitotic errors, unlike whole-chromosome aneuploidies, which are predominantly meiotic; accordingly, the positive predictive value of a trophectoderm finding for the inner cell mass was 70.8% for segmental against 97.18% for whole-chromosome aneuploidies, and a risk model based on confirmation by a second trophectoderm biopsy and on segment length improved prediction of inner cell mass status in over 86% of clinical cases studied [113]. A later multisample study using haplotyping reached a different conclusion on origin: of 105 donated blastocysts with a non-mosaic segmental call in the first biopsy, 81% (85 of 105) gave discordant results across the three samples (the first and a second trophectoderm biopsy and the inner cell mass), consistent with mosaicism, yet the authors inferred a meiotic origin in more than half, often paternal [114]. A separate research re-analysis of 89 embryos, each given a second trophectoderm biopsy with the rest of the embryo then tested, suggests that a second biopsy may help interpret segmental findings: when a segmental change in the first trophectoderm biopsy appeared again in a second, it was almost always present in the rest of the embryo (17 of 18), and when it did not, the rest of the embryo was unaffected in 12 of 13 cases [115]. The authors proposed offering a second biopsy for segmental results, but these studies analysed embryos not transferred; none has shown that a clinical second biopsy improves live birth or safety, and it adds manipulation and a further vitrification-warming cycle, both associated with lower live birth (Section 5.6). A second biopsy that does not repeat a segmental finding is a discordant result, not a euploid diagnosis. In this reference's view, a second-biopsy strategy for segmental findings needs further evaluation before routine use; it is a separate question from ESHRE's advice against re-biopsy to confirm a mosaic result (Section 8.4).

Segmental aneuploid embryos cleave more slowly than their euploid siblings over the first three cell cycles, but morphokinetic models to predict them performed modestly and did not transfer between centres [116]. Transfer data on embryos with non-mosaic segmental findings remain sparse; a 2023 review judged them too few to estimate reproductive potential or gestational and newborn risks [112]. The 2026 IRMET cohort, from eight of the registry's 26 centres, has since reported 250 transfers of embryos with non-mosaic aneuploid results and complete outcome documentation, made either knowing the result or, in non-selection settings, without knowledge of it. None of 168 whole-chromosome aneuploid transfers, 111 of them made without knowledge of the result, ended in a live birth (exact 95% CI 0% to 2.2%, calculated for this reference), whereas 17.1% of 82 segmental transfers did. Neonatal data were available for only 10 segmental births: gestational age and birthweight showed no detected difference from euploid births at the same centres, and prenatal or neonatal genetic tests were normal in all tested cases, numbers too small to establish safety [78]. Which embryos were transferred knowingly was not random, so the segmental figure cannot supply a transfer rule for a particular segmental finding, and it may not generalise to other platforms; it does support keeping segmental and whole-chromosome results apart in counselling and in any pooled figure for abnormal embryos. A different pattern has a different meaning: when several embryos of one couple show segmental imbalances of the same chromosome regions, typically a gain of one segment with a loss of another, an unrecognised balanced rearrangement in a parent is possible, and the appropriate response is genetic counselling and parental karyotyping, as in a reported case [117], rather than a decision about one embryo.

9.2 Microdeletions and inherited copy-number variants

In 2017 the resolution of methods on single cells was described as too coarse to detect the main recurrent microdeletion and microduplication syndromes, there were no reference databases for the pathogenicity of copy-number changes in embryos, and the recommended course was not to report subchromosomal changes in PGT-A [75]. Since then, targeted sequencing with genotype analysis has been validated for ploidy and for several common microdeletion syndromes, with two regions excluded for poor reliability; applied clinically it identified a 22q11.21 deletion, retrospectively, in a transferred embryo whose pregnancy had ended in miscarriage, and estimated an incidence of ploidy abnormalities or microdeletions of 1.5% (3 of 202; 95% CI 0.5 to 4.5% as reported; an exact binomial interval is 0.3% to 4.3%) among embryos otherwise considered euploid. Because about half of those embryos had been chosen because their transfer ended in pregnancy loss, the figure is not an estimate for unselected embryos [70].

Small inherited copy-number variants can also appear incidentally. On one targeted platform, a small variant seen in at least two embryos from the same cycle was reported in 75 of 12,157 patients (0.62%); in all 65 couples with follow-up microarray testing, one partner carried it, and 40% of the variants were benign or likely benign, 46.2% of uncertain significance and 13.8% pathogenic or likely pathogenic [118]. Such findings are about the parents as much as the embryos, and the consent and counselling for PGT-A should anticipate them; ESHRE recommends that patients be told the centre's policy on incidental findings before treatment [5]. Where donor gametes were used, the finding concerns the donor as well. A pathogenic variant may matter for a parent's own health and for relatives, so it needs parental confirmation and genetic counselling. The converse does not hold: the authors noted that a reliable negative predictive value for small copy-number variants is unknown, so the absence of such a finding is not reassurance [118].

10.1 First, what question is being asked?

PGT-A does not make an embryo more competent. It sorts the embryos a patient already has and removes some of them from use. Whether that helps depends on which outcome is counted and over what denominator, and most of the published disagreement follows from this.

• Per embryo transferred (implantation): how often an embryo implants or goes on to delivery. When two embryos are transferred together, each counts separately.
• Per transfer procedure: how often a transfer results in a pregnancy or birth; a delivery after a double transfer counts once. In one trial, for example, 134 and 163 embryos were transferred in 72 and 83 transfers, so its implantation and delivery rates have different denominators [119]. Women whose embryos are all aneuploid have no transfer in a PGT-A arm and so are not counted, while their counterparts in an untested arm do transfer, usually without success. This denominator therefore tends to favour PGT-A, although not every trial found a significant per-transfer difference (Table 3), and once women are excluded from the denominator the comparison is no longer protected by randomisation [120].
• Per participant randomised, cumulatively: the chance of a baby for every woman or couple randomised, including those with no transfer, within a stated number of transfers and follow-up (in the Yan and Lin trials, up to three transfers within 12 months of randomisation, which is not the same as using every embryo from a retrieval), counted from the point of randomisation, which in most trials came after blastocysts were available rather than at the start of stimulation. For the same embryo cohort, if every untested embryo would eventually be transferred and none were lost to treatment, testing cannot raise this ceiling. Within a limited number of transfers or a fixed follow-up, however, moving a viable embryo earlier in the transfer order can change the chance of observing a first live birth, and treatment dropout, embryo losses, delays and exclusion of potentially viable embryos also affect that outcome. As an illustration of the logic only, not an estimate of any treatment effect: if the one embryo able to give a live birth is ranked fourth by morphology and first after testing, a count limited to three transfers records a birth with testing and none without, even with complete follow-up and no miscarriage. Testing can also lower the cumulative rate through several routes whose relative contributions have not been measured: exclusion of embryos with mosaic or false-positive aneuploid results, embryo loss at biopsy or warming, embryos left without a result, and delay or dropout before all embryos are used. A first live birth is also not the same outcome as completed family size, which depends on the embryos left afterwards.
• Miscarriage, number of transfers and time to pregnancy: outcomes that measure the burden of reaching a birth rather than whether it happens.

Higher success per transfer with a similar cumulative rate per woman is therefore not a contradiction. Which result matters depends on whether the question is the efficiency and burden of treatment, or the overall chance of a baby from a started cycle.

10.2 The randomised trials

The first generation of PGT-A, FISH on a single cleavage-stage blastomere, was tested in randomised trials and failed. In women aged 35 to 41, live birth per woman randomised was 24% (49 of 206) with screening against 35% (71 of 202) without, a rate ratio of 0.68 (95% CI 0.50 to 0.92) [12], and a meta-analysis of nine trials confirmed lower live birth with the method in women of advanced maternal age (risk difference −8 percentage points, 95% CI −13 to −3) [13]. The trials of the current method, comprehensive testing of trophectoderm, are summarised in Table 3.

Three points from the table need more space than a cell. Yan's interval has its upper limit at about zero, so the data do not support a gain with PGT-A in this population; the trial was not designed to test superiority of conventional IVF, and its results do not extend to older or low-response patients [124]. In Lin, couples could enter with cleavage-stage embryos only, and some never had a blastocyst: at 12 months, 34 of 225 couples in the PGT-A arm and 13 of 225 without testing had had no transfer, including 7 per arm in whom no blastocyst was obtained and 22 in the PGT-A arm with no blastocyst available for transfer after testing; the odds ratio for live birth after the first transfer was 1.09 (95% CI 0.76 to 1.58) [125]. In Peregrino, the difference in live birth was 4.9 points (95% CI −8.4 to 18.1, calculated for this reference by normal approximation) [126].

Three further trials tested older or different methods. In women aged 38 to 41, day 3 biopsy with array CGH improved delivery at the first transfer and reduced miscarriage (2.7% against 39.0%) and time to pregnancy, but cumulative delivery per patient did not differ significantly, and the analysis included only the 205 of 326 recruited women who completed the study [127]. Polar body testing in women aged 36 to 40 (Section 4.1) gave cumulative live birth of 24% in both arms per woman randomised [44]. A single-centre pilot of 100 women aged 35 to 42, designed to test feasibility, reported cumulative live birth of 72% against 52%, significant before adjustment (P = 0.04 by an uncorrected chi-square test; a two-sided Fisher exact test on the same counts gives P = 0.06, calculated for this reference) but not after (P = 0.08); the abstract describes the difference as not significant, and the comparison also differed by fresh against frozen transfer [128]. A further trial asked a different question: with randomisation once at least two blastocysts could be biopsied, in women under 43, ongoing pregnancy per patient randomised after the first transfer was 60.7% with one euploid blastocyst (89 patients) against 65.1% with two untested blastocysts (86 patients), relative risk 0.9 (95% CI 0.7 to 1.2); the non-inferiority design excluded an absolute difference of more than 20 percentage points in favour of two untested blastocysts (a wide margin; the relative risk interval, down to 0.7, does not exclude a relative loss of up to about 30%), and multiple gestation fell from 53.4% to 0%. Because the arms also differed in the number of embryos transferred, the trial does not isolate the effect of testing [129].

The 2020 Cochrane review, which predates the Yan and Lin trials, found insufficient good-quality evidence on cumulative live birth for comprehensive methods (one polar body trial, low-certainty evidence, and no cumulative data from the one blastocyst trial), and judged that FISH-based testing probably reduces live birth after the first transfer [130]. No later Cochrane update had been published when this reference was written. A 2026 systematic review (literature searched to November 2024) restricted to seven randomised trials of all-chromosome PGT-A, with GRADE (Grading of Recommendations, Assessment, Development and Evaluation) assessment of the certainty of evidence, found that live birth per woman randomised did not differ significantly with and without testing, either after the first planned transfer (odds ratio 1.16, 95% CI 0.94 to 1.45) or cumulatively (the review's "per oocyte retrieval", although in most included trials women entered only once blastocysts were available; odds ratio 0.87, 95% CI 0.64 to 1.19), including in women over 35, with low certainty of evidence. Live birth per transfer was higher with PGT-A, and the effect on miscarriage could not be assessed because certainty was very low. Because they found no clear patient benefit and estimated that, in a Swedish setting, PGT-A almost doubles the cost of IVF, its authors concluded that PGT-A is currently not considered ethically justifiable [131]. Because its search ended in November 2024, the Lin trial (published December 2025), the Peregrino trial (February 2026) and the 100-woman pilot (July 2025) fall outside it; this reference does not add them to its pooled estimates.

10.3 Why reasonable experts read the trials differently

No adequately powered randomised trial of current PGT-A has shown a higher cumulative live birth rate per woman randomised, within follow-ups of up to three transfers or 12 months; the only nominally significant difference, in a 100-woman feasibility pilot, did not survive adjustment. Beyond that point, readings diverge, and the reasons can be named.

Population. The trials enrolled women with a good prognosis: under 38 in five of them, and with two or more biopsiable or good-quality blastocysts (STAR, Ozgur, Peregrino) or three or more good-quality blastocysts (Yan) at randomisation. These are the patients in whom selection has least to do. No trial of trophectoderm PGT-A has yet reported a cumulative live birth endpoint powered for women over 38 or for those with few embryos [132]. In the STAR trial, a subgroup of women aged 35 to 40 had higher ongoing pregnancy per transfer (51% against 37%), but this analysis was post hoc and the difference was not significant per woman randomised [123]. A prospective multicentre Danish cohort of 260 PGT-A patients who were aged 36 or over, had recurrent implantation failure or had recurrent pregnancy loss (or more than one of these), compared with a historical reference group of 3,060 patients having conventional single frozen transfer, reported higher odds of ongoing pregnancy per transfer in each group (for age, odds ratio 2.20, 95% CI 1.52 to 3.18) and lower odds of pregnancy loss, while cumulative pregnancy rate did not differ significantly (risk ratio 0.94, 95% CI 0.80 to 1.10); the outcome is pregnancy rather than live birth, and the design is observational with a non-concurrent comparator [133].

The fixed embryo pool. Critics argue that in a cohort of a fixed size, testing can only reorder or remove embryos, so a cumulative gain should not be expected and the absence of one is not a failure; supporters of PGT-A argue the same point to defend it, as a test whose value lies in fewer failed transfers and miscarriages rather than more births. One commentary argued that the Yan design fixed the embryos for transfer before randomisation and excluded mosaic embryos, which may have cost deliveries in the PGT-A arm, and that the trial is best read as a safety study [134]. From the opposite direction, another argued that, analysed properly, the trials show a deficit with PGT-A [135].

Endpoints. An appraisal of the six trials published by 2019 found that only two reported cumulative live birth per started cycle, both without significant difference, and argued that a focus on the first transfer misrepresents the test as superior [136]. A 2025 review concluded that the trials share shortcomings (limited power, good-prognosis patients, few cumulative outcomes, exclusion of mosaic transfers) and that the evidence does not justify widespread routine use [132]. A meta-analysis that pooled trials and cohort studies per transfer reported a risk ratio of 1.09 (95% CI 1.02 to 1.16) for live birth or ongoing pregnancy in six trials, using a fixed-effect model; its trial pool included one trial of single against double embryo transfer, and the result answers the per-transfer question only [137]. A 2021 network meta-analysis of 11 trials of comprehensive testing at day 3 or day 5, which predates the Yan and Lin trials, found no significant improvement in live birth per patient overall (risk ratio 1.11, 95% CI 0.87 to 1.42; 1,513 patients; I² 75%). It did, however, find higher cumulative live birth per patient (risk ratio 1.36, 95% CI 1.13 to 1.64; 580 patients) and, in an age subgroup analysis, higher live birth in women over 35 (risk ratio 1.29, 95% CI 1.05 to 1.60; 692 patients) [138].

Registry data. Large analyses of the US SART registry, the SART Clinic Outcome Reporting System (SART CORS), which follow patients from retrieval, point in both directions. Among 133,494 first autologous cycles from 2014 and 2015, PGT-A was associated with lower cumulative live birth at all ages except over 40, most markedly under 35, among patients who had blastocysts available for transfer or testing, counting those left with no transferable embryo [139]. In an overlapping cohort of 31,900 women aged 37 or under, followed from their first retrieval, cumulative live birth per patient was 70.6% with against 71.1% without PGT-A under 35 (adjusted odds ratio 0.82, 95% CI 0.72 to 0.93) and 66.6% against 62.5% at 35 to 37 (adjusted odds ratio 1.14, 95% CI 0.99 to 1.31) [140]. That last odds ratio is from the paper's Table 2; the abstract gives 0.92 (95% CI 0.83 to 1.01), which the table reports for the whole sample. The models adjusted for the number of embryos transferred, which PGT-A itself changes, so these adjusted odds ratios do not estimate the total effect of choosing PGT-A. In a selected cohort of 56,469 first retrievals from 2016 to 2019 in women aged 21 to 40 with at least five fertilized oocytes and subsequent single blastocyst transfer (how many retrievals ended with no transfer is not reported in the sources checked, so these estimates should not be read as outcomes for all started cycles), cumulative live birth per retrieval was slightly lower with PGT-A under 35 (risk ratio 0.96, 95% CI 0.93 to 0.99) and higher at 38 to 40 (risk ratio 1.14, 95% CI 1.07 to 1.20), with a borderline estimate at 35 to 37 (risk ratio 1.04, 95% CI 1.00 to 1.08) [141].

These are observational comparisons: patients who choose PGT-A differ from those who do not, and adjustment cannot remove all of that difference. A 2026 systematic review of seven large real-world studies found that all those reporting live birth per transfer favoured PGT-A, whereas those reporting per started cycle or per retrieval favoured the untested group [142].

Platform and reporting policy. The trials used FISH, array CGH, qPCR and NGS on different cells, with different mosaic thresholds and transfer policies, so their results are not simply interchangeable. Excluding embryos with mosaic results, as some trials did, is one of the routes listed in Section 10.1 by which testing can reduce cumulative live birth, and the trials do not allow their contributions to be ranked. Mosaic exclusion is a property of the reporting policy rather than of testing as such (Section 8).

The supporters' case. Capalbo and colleagues, whose research network has produced much of the non-selection evidence and whose authors include employees of genetics laboratories that provide PGT, frame PGT-A as a deselection test: it removes embryos with no realistic prospect of a healthy birth, so that failed transfers, miscarriages and abnormal pregnancies are avoided without a meaningful loss of cumulative live birth [41]. Their observational data are consistent with that view. In older women at one centre (308 retrievals with PGT-A, 2,905 without), live birth per transferred embryo was 40.3% with PGT-A against 11.0% without, while cumulative delivery per retrieval did not differ significantly (26.3%, 95% CI 21.5 to 31.6, against 24.0%, 95% CI 22.5 to 25.6) and pregnancy loss was lower (3.6% against 22.6%); a further 106 retrievals in which PGT-A was consented to but abandoned after a poor embryological outcome were analysed as a separate drop-out group, which removes poor-prognosis cycles from the PGT-A arm and favours it [143]. In a propensity-matched comparison of 242 patients per group with only one or two blastocysts, matched on age, oocytes inseminated and blastocyst number only, cumulative live birth was 27.3% (95% CI 21.9 to 33.4) with and 23.6% (95% CI 18.5 to 29.5) without PGT-A, with about 74 against 131 days from retrieval to the end of treatment [144]. These are not randomised comparisons. Capalbo, Rubio and Simon disputed a modelling argument that PGT-A loses potential implantations [145] in a letter [146], and the model's author replied rejecting their critique [147].

The critics' case. Others argue that PGT-A has never been clinically validated in its ability to classify embryos as normal, mosaic or aneuploid, that its false-positive rate reduces the chance of a live birth for many patients, and that it should be offered only within research [148]. A 2026 PGDIS task group report, whose authors include investigators of the earlier trials, took the opposite view and concluded that properly implemented PGT-A benefits embryo selection [149]. Readers should note that authors on both sides have professional interests in the answer, and that several authors of the task group report list affiliations with PGT or genomics companies.

10.4 Why euploid embryos still fail

A euploid result removes one cause of failure; it is not a promise of a baby. A systematic review of 372 studies, most of them retrospective, noted at the outset that the positive predictive value of a euploid result for implantation is no higher than about 50% to 60%, and identified factors associated with failure of euploid transfers, including repeated implantation failure, higher body mass index, poorer inner cell mass grade and, more weakly, maternal age of 38 or more (odds ratio 0.87, 95% CI 0.75 to 1.00); the quality of the evidence was low or very low [52]. In a retrospective propensity-matched study of women aged 39 to 46 (278 per group), live birth after a first single euploid frozen transfer did not differ significantly between autologous and donor oocytes (42.45% against 41.01% per transfer; 95% CI for the difference about −7 to 10 points, calculated for this reference by normal approximation, so a clinically relevant difference is not excluded), which the authors took to support the view that the age-related decline in IVF outcome is driven largely by aneuploidy [150]. Morphology still matters among euploid blastocysts, as do the endometrium and the transfer [52]; Section 14.2 describes research on markers beyond chromosome copy number.

10.5 Miscarriage, time, safety and cost

Miscarriage and pregnancy loss. Four trials reported significantly less pregnancy loss with PGT-A, but they measured different things. In the day 3 biopsy trial, miscarriage was 2.7% against 39.0% (type of loss and denominator not stated in the abstract) [127]. In the Yan trial, cumulative clinical pregnancy loss was 8.7% against 12.6% (difference −3.9 points, 95% CI −7.5 to −0.2) [124]; a systematic review reproducing the trial's results gives these as 46 of 526 against 72 of 571 women, denominators larger than the numbers with a clinical pregnancy (505 and 556) and so apparently women with a positive pregnancy test [151]; per woman randomised, the same counts would be 7.6% against 11.9% (46 and 72 of 606, calculated for this reference). In the Lin trial, total pregnancy loss within 12 months, including biochemical losses, was 11.1% (25 of 225) against 22.7% (51 of 225) per couple randomised, and clinical pregnancy loss alone 7.1% (16 of 225) against 13.8% (31 of 225); both were secondary outcomes, and the trial's stated multiplicity procedure concerned its two primary outcomes [125]. With polar body testing, fewer women randomised had a miscarriage (risk ratio 0.48, 95% CI 0.26 to 0.90) [44]. In the STAR trial the difference was not significant (odds ratio 0.89, 95% CI 0.52 to 1.54, in the Cochrane analysis) [130], and in the 100-woman pilot clinical miscarriage after the first transfer was 12% against 8% of women randomised (6 and 4 of 50) [128]. Fewer losses do not by themselves mean more births: in the Lin trial, cumulative live birth did not differ (Table 3). In 2024, before the Lin trial was published, ASRM and SART described the miscarriage benefit as unclear [152], and ESHRE noted in 2023 that claims of reduced miscarriage and time to pregnancy in specific groups rested largely on post hoc analyses [153]. Loss per clinical pregnancy depends on a post-randomisation event; loss per woman randomised is the cleaner measure.

Time to pregnancy. Time is the natural measure of the efficiency argument, but few trials have used it as a primary outcome. The day 3 biopsy trial reported 7.7 against 14.9 weeks to pregnancy and 1.8 against 3.7 transfers per live birth [127]; the Lin trial detected no difference in the time from randomisation to confirmation of an ongoing pregnancy at 20 weeks or later that went on to a live birth, within 12 months (a secondary outcome; hazard ratio 1.00, 95% CI 0.79 to 1.27) [125]. That clock starts at randomisation and stops at ongoing pregnancy, not delivery, so it is not directly comparable with the registry measure below, which starts at stimulation. In the SART registry analysis of women aged 37 or under described in Section 10.3, the average time from the start of stimulation to a pregnancy resulting in live birth, among women who had a live birth, was 4.58 months for PGT-A transfers against 2.37 months for untested transfers [140].

Obstetric and neonatal safety. A 2026 meta-analysis of 12 mainly retrospective studies, covering 56,113 live births of which 17,254 followed PGT-A, detected no significant differences in the obstetric and perinatal outcomes examined [154]; this is an absence of detected difference in observational data, not proof of equivalence, and rare outcomes need larger studies. The trophectoderm biopsy evidence is discussed in Section 4.2.

Other indications. Meta-analyses in recurrent pregnancy loss [155] and recurrent implantation failure [156] report favourable odds ratios, but they rest largely or entirely on observational studies in selected patients, and their authors call for randomised trials or further research before routine use.

Cost. A model using Italian public costs found that the cost-effectiveness of PGT-A improved with age and number of blastocysts and that it could be cost-effective in specific subgroups [157], and its arithmetic was disputed [158]. The matched cohort described above estimated an incremental cost of about €499 for each month of treatment time saved by PGT-A (incremental cost-effectiveness ratio; €170 to €2,065 by blastocyst number and age) [144]. The results depend on inputs the trials leave uncertain, above all whether testing causes any loss of cumulative live birth. A decision model built on data from the Yan (CESE-PGS) trial with Chinese healthcare costs estimated the average cost per live birth with PGT-A at about 16.8% higher than with conventional treatment and judged it not suitable for routine use in China [159]. The 2026 systematic review's Swedish analysis (Section 10.2) found that PGT-A almost doubles the cost of IVF without any increase in live births [131].

10.6 What the professional bodies say

ASRM and SART concluded in 2024 that the value of PGT-A as a routine screening test for all patients undergoing IVF has not been demonstrated [152]. ESHRE's 2023 good practice recommendations on add-ons (optional treatments offered in addition to standard IVF, often at extra cost) state that PGT-A is currently not recommended for routine clinical use, a category ESHRE defines as one that may change when more evidence on efficacy and safety becomes available, as distinct from its 'not recommended' category [153]. In England, NICE guideline NG257 (March 2026) recommends: "Do not offer pre-implantation genetic testing for aneuploidy (PGT-A) as part of fertility treatment to improve live birth rates" (recommendation 1.48.1), a recommendation framed around the intended benefit for live birth [160]. The UK regulator, the HFEA, rates the chance of having a baby and the chance of miscarriage separately: red for increasing the chance of a baby and green for reducing miscarriage for most patients, and, in separate ratings for older women, grey for both outcomes because of insufficient moderate or high quality evidence [161]. In Canada, Ontario Health's health technology assessment (October 2026; six randomised trials) found that PGT-A may not improve cumulative live birth (very low certainty), may make little or no difference to live birth after the first planned transfer per person (low certainty) and may reduce cumulative clinical pregnancy loss (low certainty); it did not attempt a cost-utility analysis because of the uncertain benefit [162]. An assessment is not a funding decision: Ontario Health makes its funding recommendation separately, on the advice of the Ontario Health Technology Advisory Committee, and its draft recommendation, against public funding of PGT-A, was released for consultation in June 2026 and was still listed as in development in November 2026 [163]. The PGDIS 2026 task group took a more favourable view [149]. These statements concern whether PGT-A should be offered. The laboratory's own question is narrower: given that a patient has chosen PGT-A, how to make the result as reliable as possible. ESHRE addresses it directly in its PGT Consortium good practice recommendations on biopsy, sample handling, witnessing and the organisation of PGT [1, 5], and it is the subject of Sections 5 and 15.

11.1 Testing for a known monogenic condition

PGT-M answers a narrower question than PGT-A, whether an embryo has inherited a specific familial pathogenic variant, but it is technically harder, because the test must detect a change at one locus in a few cells. The two main risks are amplification failure and allele drop-out, in which one allele of a heterozygous sample amplifies and the other does not, so that an affected embryo can appear unaffected. Cell number matters here: in a prospective blinded validation of a qPCR method for concurrent PGT-M and aneuploidy testing, allele drop-out frequency was 1.64% with single cells from a fibroblast line but 0.02% when two or more cells were tested (reported as 18/1,096 and 1/4,426; the denominators count assay results, not cells, since the same abstract reports 800 single-cell samples) [164], which favours a multicellular trophectoderm biopsy for PGT-M. Because drop-out cannot be excluded, most methods rely on haplotyping: polymorphic markers close to the gene are genotyped in the couple and in relatives of known status during a preclinical work-up, and the embryo's inheritance of the high-risk haplotype is inferred alongside, or instead of, direct detection of the variant. ESHRE's minimum marker numbers depend on the strategy and on validated performance. Where the variant itself is tested together with linked markers and amplification failure and drop-out stay below 5% per locus in validation, it recommends at least one informative short tandem repeat (STR) or three SNPs on each side of the variant, preferably within 1 Mb; a marker-only (indirect) test needs at least two STRs or six SNPs on each side; and higher failure or drop-out rates call for further optimisation or more markers. These are design requirements for the genetics laboratory, not choices the embryology laboratory makes. Linkage-based testing has its own residual risk, from recombination between a marker and the variant or from partially informative markers, and ESHRE asks that the report state the test's limitations and residual risk of misdiagnosis, including a figure [7]. Genome-wide versions of haplotyping, first described as karyomapping, use the SNP genotypes of the couple and a relative to map the inheritance of all four parental haplotypes across every chromosome, so that one test can be applied to many familial monogenic conditions, after case-specific feasibility assessment, without disease-specific development, and also detects trisomy and monosomy [71]; a later method, haplarithmisis, combines genome-wide haplotyping with copy-number profiling of single cells and can distinguish meiotic from mitotic errors [72].

Two consequences matter at the bench. First, contamination is a more serious threat in PGT-M than in PGT-A, because a single contaminating cumulus cell, or a sperm attached to the zona after conventional insemination, can supply the parental allele that has dropped out; this is the reason for ICSI and thorough denudation in PGT cycles (Section 5.1) [1]. Second, the preclinical work-up takes time and requires DNA from family members, so the IVF cycle should not start until the genetics laboratory confirms that the test is ready. Combined testing of the monogenic condition and aneuploidy from the same trophectoderm biopsy is technically established and increasingly offered [165, 166]. Technical feasibility does not establish benefit: whether to add PGT-A is a separate clinical and consent decision, made before the cycle in light of age, the expected number of embryos and the laboratory's reporting policy, given that the value of PGT-A as a routine test has not been demonstrated [152]. An embryo reported unaffected for the target condition has not been tested for most other genetic disease. Haplotyping also leaves the clinical preconditions in place. ESHRE excludes variants of uncertain significance, and cases with an uncertain diagnosis or mode of inheritance, from PGT, and expects counselling on the severity and variability of the condition; a de novo variant or missing relatives can mean the high-risk haplotype has to be established during the clinical cycle itself; repeat expansions such as those in FMR1 are often tested indirectly by linked markers; and mitochondrial DNA variants require measurement of variant load, with PGT not indicated for homoplasmy [5, 7]. ASRM's committee opinion on PGT-M sets out the indications and the clinical and technical complexities for the United States. Unlike ESHRE, it leaves requests to test for a variant of uncertain significance to laboratory policy and counselling, and it notes that such a test can only identify whether the embryo carries the variant, so that PGT-M would not reduce the risk if the condition has another cause, and that most such variants are eventually reclassified as benign [167]. Finding the variant in an embryo therefore does not by itself establish that the embryo is at increased risk of the familial disease.

11.2 Testing for a structural rearrangement

Carriers of balanced reciprocal or Robertsonian translocations and inversions produce a high proportion of embryos with unbalanced segregation products, and familial translocations are the most common indication for PGT-SR. ESHRE recommends that the risk assessment for each couple state the residual risk of a viable unbalanced pregnancy where the test cannot detect every possible unbalanced product, for example when one of the translocated segments is below the method's resolution. Most copy-number methods cannot distinguish an embryo with a normal karyotype from one that carries the same balanced rearrangement as the parent, and neither FISH nor copy-number methods such as array CGH and NGS detect uniparental disomy [8]. Haplotype-based methods can make the normal-versus-balanced distinction in some cases; the laboratory's report should say which applies. Where it cannot, an embryo reported as normal or balanced may carry the parental rearrangement; most carriers of the familial balanced rearrangement are clinically unaffected but may face the same reproductive risks as adults. Rearrangements involving chromosomes 6, 7, 11, 14, 15 or 20, including Robertsonian translocations of chromosome 14 or 15, carry an increased risk of uniparental disomy, which PGT-SR by copy number does not detect; ESHRE considers prenatal testing for it acceptable after individual assessment [8], and the Italian Society of Human Genetics puts the residual risk for a specific chromosome usually at no more than 1% [82].

11.3 Carrier screening as the route into PGT-M

Most couples for whom PGT-M could be relevant do not know it, because recessive conditions usually appear without family history. In exome data from 14,125 people, 5,845 gamete donors and 8,280 patients with infertility with no known family history, analysis of 114 selected gene-condition pairs identified a reproductive risk of a highly penetrant childhood-onset disorder in about 1 in 337 conceptions, with rare conditions (carrier rate below 1 in 100) and X-linked conditions contributing 16.7% and 41.2% of the total fetal risk. In a subgroup analysis, 37 of 776 IVF couples (4.8%, 95% CI 3.4 to 6.5) were at increased risk [168]. In a focused panel built on criteria of the American College of Obstetricians and Gynecologists (ACOG), 20 of 766 couples tested (2.6%) were at increased risk [169].

Practice varies widely. In a 2022 survey of 299 professionals from 40 countries, 42.5% worked at centres offering expanded carrier screening, with wide variation in panel design and in whom it was offered to [170]. Professional societies differ in what they recommend [171, 172]; a comparison of Italian, American and Australasian recommendations found a high level of agreement on core principles but not on every point [173]. In the United States, the American College of Medical Genetics and Genomics recommends that all pregnant patients and those planning a pregnancy be offered "Tier 3" screening, for conditions with a carrier frequency of at least 1 in 200 in any ethnic group with reasonable representation in the United States, together with selected X-linked conditions, and notes that a negative result reduces but never removes the chance of being a carrier. It advises considering wider "Tier 4" screening, with no lower carrier-frequency limit, for consanguineous couples (second cousins or closer) or where family or medical history warrants it [174]. The National Society of Genetic Counselors' evidence-based guideline recommends offering expanded screening to everyone pregnant or considering pregnancy, rather than selecting tests by ancestry, with the decision reached by shared decision-making [175]. Scope elsewhere follows national policy. A screening result is not a PGT-M indication in itself: an autosomal recessive carrier finding first needs testing of the partner or gamete donor, whereas for a woman heterozygous for an X-linked variant each pregnancy generally has a 1 in 2 chance of inheriting the variant whatever her partner's result, with consequences that depend on the condition, the fetal sex and penetrance (daughters can also be affected), and a couple found to be at risk needs clinical genetics review, confirming the variant classification and the severity of the condition, before a PGT-M work-up is considered.

11.4 Confirming the result in pregnancy

No PGT result is a prenatal diagnosis, and PGT does not replace routine prenatal care. PGT-M and PGT-SR answer a targeted question: an embryo reported unaffected can still carry a de novo variant or a genetic condition that was not tested, and the report should name the conditions tested for and the limitations of the method [5, 176]. ESHRE also asks that counselling be non-directive and cover all reproductive options open to the couple [5], which can include natural conception with prenatal diagnosis, donor gametes, not testing, or other routes to parenthood. ESHRE advises that counselling include the possibility of misdiagnosis, with error rates based on the laboratory's own work-up and follow-up, and the option of prenatal diagnosis to confirm the result [5]. The societies agree that prenatal testing should be offered and discussed after PGT and that accepting or declining it is the patient's choice; they differ on which tests suit which situation (Table 4). Each option has limits: invasive procedures carry a small procedure-related risk of miscarriage, chorionic villi share the placental origin of the trophectoderm, and amniocentesis, although it samples cells closest to the fetus, can still miss low-level mosaicism [110].

After PGT-M or PGT-SR, confirmation is directed at the original target, the familial variant or rearrangement, tested on a fetal or placental sample or, for some monogenic conditions, by noninvasive prenatal diagnosis designed for that variant [5, 82]. Standard cell-free DNA screening for common aneuploidies does not examine the familial variant, so a normal result does not confirm a PGT-M result; after PGT-SR it answers a screening question about common trisomies, not the diagnostic question of whether the fetus carries an unbalanced product of the familial rearrangement. The bodies also differ on cell-free DNA screening after a euploid PGT-A result: the Italian society considers it unsuitable because it analyses DNA of the same trophoblast origin as the biopsy, whereas the US general guidance, which does not address PGT, recommends cell-free DNA as the most sensitive and specific screening test for common trisomies in any patient population [82, 178]. The choice of test is for the clinician and genetic counsellor with the patient, not the laboratory.

12.1 What pronuclear observations can tell us about fertilization and ploidy

IVF laboratories score pronuclei after insemination or ICSI, at 16 to 17 hours in the 2025 update of the ESHRE/Alpha Istanbul consensus [179], and, by convention, most have not used embryos from zygotes with one (1PN) or three or more pronuclei; policies for zygotes with a second polar body but no visible pronuclei vary more. These are often recorded as 0PN, a term the consensus update advises against after static observation, because time-lapse data show that a significant proportion of normally fertilized zygotes lose their pronuclei before the check; it proposes "PN not observed" instead. Capalbo and colleagues have pointed out that this is in effect a morphological test for ploidy, carried out because abnormal ploidy causes implantation failure, miscarriage and molar pregnancy. They also argue that it is not a very specific one: by their review of the literature it misclassifies more than half of the blastocysts that develop from atypically pronucleated zygotes, leading to the routine disposal of embryos that could have been viable [180].

12.2 What genotyping shows

Genotype-based methods can measure ploidy directly, from the ratio of alleles at heterozygous polymorphisms in the same biopsy used for PGT-A: one to one for a diploid embryo, two to one for a triploid, and genome-wide loss of heterozygosity for a haploid embryo or for a uniparental isodiploid one, with two identical chromosome sets from one parent, which copy number cannot tell apart from a normal embryo; one large cohort found maternal isodiploidy in 0.05% of biopsies (41 of 74,009) [32]. In 719 PGT-A cycles in 556 women, 5.2% of 5,026 injected oocytes showed one pronucleus and 0.7% showed two pronuclei with a smaller third (2.1PN); 9 of 13 blastocysts from 1PN zygotes and 12 of 14 from 2.1PN zygotes were diploid (small numbers: 95% CIs 39% to 91% and 57% to 98%, calculated for this reference). Rescue added eight balanced-diploid transferable embryos, in three cycles the only such embryo came from an atypical zygote, and three live births resulted [181]. A prospective cohort study from Leuven reached a similar conclusion in PGT-M cycles, using genome-wide haplotyping of a single blastomere biopsied on day 3: of 216 embryos analysed in 42 cycles, 49 came from 0PN (no visible pronuclei) and 15 from 1PN zygotes. Of these, 37 of 49 0PN-derived embryos (75.5%) and 6 of the 14 1PN-derived embryos analysed (42.9%) were diploid and biparental, a precondition for diagnosis rather than a finding of euploidy or of suitability for transfer (27 of 49 and 5 of 14 were euploid). Thirteen of the 42 cycles (31%) reached or could repeat transfer with an unaffected 0PN or 1PN embryo, and 7 of the 42 cycles gave a live birth, against the 4 that came from 2PN-derived embryos alone (P > 0.05); this is the contribution observed in one series, not a controlled comparison of two policies [182]. A healthy birth has also been reported after transfer of a diploid blastocyst from a four-pronucleate zygote [183]. A clinic series from Sydney adds a caution: among 24 female 1PN-derived blastocysts tested for biparental inheritance, one ICSI-derived embryo had only a paternal contribution and so carried a risk of complete hydatidiform mole, showing that biparental inheritance, not only ploidy, needs to be confirmed; single transfers of 26 1PN-derived blastocysts with no detectable abnormality gave nine clinical pregnancies [184].

A 2026 systematic review of seven studies of 2.1PN embryos reported descriptively lower blastulation, euploidy and diploidy than for 2PN embryos, without a meta-analysis, and estimated, by multiplying crude pooled rates, that a single 2.1PN zygote has about a 9.3% chance of becoming a euploid diploid blastocyst; four studies reported 7 live births from 23 transfers (95% CI 13% to 53%, calculated for this reference), and the authors supported clinical use when diploidy is confirmed by a validated genotyping method [185]. A model-based cost-effectiveness analysis from one Italian centre, using 8,334 ICSI cycles with PGT-A, estimated that including PGT-A-confirmed euploid blastocysts from atypical zygotes would raise the probability of at least one live birth by 1.42 percentage points while slightly reducing mean cost per couple [186].

The gain per cycle is small and much of the evidence comes from one research network, whose authors include employees of a PGT provider, although groups outside that network, an academic group in Leuven and a Sydney IVF and genetics company, have reported similar findings [182, 184]. Current guidance treats the pronuclear categories separately: the 2025 Istanbul consensus update (its zygote table and text) and ESHRE's 2026 good practice recommendations for the IVF laboratory (Table 1), which adopt its categories [56, 179].

• PN not observed, second polar body present: clinical use accepted after culture to blastocyst, to separate these zygotes from rare unfertilized oocytes that cleave abortively; ESHRE does not require genetic testing.
• 1PN: cautious clinical use, with blastocyst culture and, where available, genetic testing able to assess biparental diploidy.
• 2.1PN: use may be considered, in the consensus update "especially" with such testing; ESHRE's table lists blastocyst culture and such testing together.
• 3PN: the consensus update does not recommend clinical use and encourages pre-clinical studies; ESHRE's table lists it as "not recommended as a routine practice, but admissible in pilot clinical studies", with blastocyst culture and testing. Neither supports routine clinical use.
• 4PN: neither document gives a category; the single birth cited above is a case report and cannot support a policy.

Use of 1PN and 2.1PN zygotes should be discussed with the clinical team and the patients, governed by an internally approved policy and fully traceable. This reference's recommendation, within that frame, is that a laboratory with access to validated testing of ploidy and parental origin should replace a policy of discarding every atypical zygote on day 1 with a written policy for each category; that embryos cultured under it be identified to the genetics laboratory by pronuclear category; and that consent to biopsy and testing, and to what will be done with each possible result, be obtained before biopsy rather than after it. Three questions stay separate: the laboratory documents technical suitability, including whether the assay established diploidy, biparental inheritance or both; the clinical and genetics teams explain which options applicable law and clinic policy allow, with the small number of reported births and the residual genetic risks (above and Section 12.3), which are uncertainties of genetic risk, not of morphology; and the patient chooses among those options. A diploid, biparental result can make a 1PN- or 2.1PN-derived embryo technically suitable for consideration under such a policy; it does not make a 3PN- or 4PN-derived embryo eligible for transfer outside a study.

12.3 Triploidy among normally fertilized embryos

The converse also holds: an embryo from a 2PN zygote can be triploid or haploid. In the large cohort described in Section 3.2, restricted to blastocysts from ICSI zygotes scored 2PN, 1.1% of informative trophectoderm biopsies (1,063 of 93,986) showed an extra or missing chromosome set, 181 of them haploid; this is a frequency among biopsied blastocysts, not among all fertilized oocytes [32], and in the validation of an integrated ploidy test one triploid embryo was found among 100 2PN-derived blastocysts previously reported as euploid [70]. Copy-number PGT-A alone does not detect most of these embryos (Section 6.3). The clinical stakes are real. Full triploidy usually ends in pregnancy loss, and the cited authors attribute about 10% of pregnancy losses to triploid and haploid conceptions together [32]. Rare triploid infants are born alive and survival is generally brief: a 2022 report of a girl with complete triploidy who died on day 36 counted 12 earlier reported cases of survival beyond 30 days [187]. A triploidy with two paternal sets can present as a partial hydatidiform mole, which needs follow-up for persistent trophoblastic disease. Haploidy, a single parental genome, is a separate problem, which the cited authors regard as not viable.

13.1 The idea and the sources of DNA

The main randomised result first: in women aged 35 to 42 with at least two blastocysts, ongoing pregnancy after the first single frozen transfer was 38.7% (221 of 571) when embryos were ranked by spent-medium testing and 36.8% (214 of 581) when they were ranked by morphology, and no embryo was excluded on the basis of the test (Section 13.3) [188].

Embryos release DNA into the culture medium and the blastocoel fluid. Testing spent medium instead of a biopsy would avoid micromanipulation and the need for biopsy skills, and might sample both the trophectoderm and the inner cell mass; aspirating blastocoel fluid still requires micromanipulation and is better described as minimally invasive. The difficulty is that the DNA is scarce and degraded, and the medium can also contain DNA from cumulus cells, polar bodies, sperm and the operator [189, 190]. Studies of the medium have confirmed the problem: in 227 blastocysts from 70 patients, DNA in spent medium was of mixed origin, with contamination from the culture medium and from maternal cumulus cells [191]. In a study of 113 embryos, of which 56 spent-medium samples were sequenced for aneuploidy, the spent-medium result differed from the trophectoderm biopsy in most cases (67%), mainly because of maternal DNA, with a median embryonic fraction estimated at 8% [192].

13.2 How well it agrees with biopsy

The results depend heavily on laboratory protocol. In a prospective blinded study in standard clinical conditions, blastocoel fluid and spent medium were compared with trophectoderm biopsies from 26 patients. For monogenic testing, only 2.9% of blastocoel fluid and 20.8% of spent medium samples gave a diagnosis concordant with the biopsy, and the maternal pathogenic variant was found in the medium of embryos that were homozygous normal, evidence of persisting maternal DNA; for aneuploidy, 65.2% of blastocoel fluid samples failed to amplify. The authors concluded that neither source was reliable enough for clinical use and that spent medium should not be used for PGT-M until maternal contamination could be prevented [189].

With protocols designed for the purpose, agreement improves. In an early validation of sequencing of spent blastocyst medium, results from 42 blastocysts were compared with the corresponding whole donated embryos, with sensitivity of 0.882 and specificity of 0.840 (estimates from 42 embryos, so imprecise), and five live births were reported among seven couples treated with the method [193]. In a multicentre prospective study of 1,301 day 6 and 7 blastocysts, 1,108 gave informative results from both spent medium and biopsy. Of these pairs, 866 (78.2%) were concordant; concordance ranged from 72.5% to 86.3% across centres, with no significant difference between them, and exceeded 86% when all defined culture-laboratory steps were controlled to minimise maternal and operator contamination; 13 agreed on ploidy but not on the sex chromosomes; and 92 (8.3%) and 137 (12.4%) fell into the authors' false-negative and false-positive groups. These last two are fractions of all informative pairs, not 1 − sensitivity or 1 − specificity, which have different denominators [190]. These figures measure agreement with the trophectoderm biopsy, itself an imperfect reference, not accuracy against the embryo. A smaller study of 52 donated blastocysts, using whole-embryo sequencing as the reference, reported that spent-medium testing outperformed trophectoderm biopsy on some measures after complete removal of cumulus cells, although the threshold for calling mosaicism was chosen on the same 52 embryos, so the comparison is likely to be optimistic [194]; critics argued that the comparison could not rescue a test whose clinical premise they disputed [195]. A 2023 meta-analysis of 20 diagnostic accuracy studies reported pooled sensitivity of 0.84 (95% CI 0.72 to 0.91) and specificity of 0.85 (95% CI 0.74 to 0.92), with higher estimates in studies of fewer than 100 cases than in larger ones and no gain in accuracy from adding blastocoel fluid to spent medium [196].

13.3 Randomised evidence

A large double-blind randomised trial in 13 Chinese centres randomised 1,189 women aged 35 to 42 with at least two blastocysts to embryo selection by noninvasive testing or by morphology for a single frozen transfer. After 37 post-randomisation exclusions, 24 of them withdrawals to request a double transfer, the modified intention-to-treat population comprised 1,152 of the 1,189. Ongoing pregnancy after the first transfer (Section 13.1) differed by 1.9 points (95% CI −3.7 to 7.5; adjusted P = 0.49), an interval that excludes the 10-point gain the trial was powered to detect. Because the test only changed the order of transfer, the trial tested a ranking, not a diagnosis on which embryos could be excluded; it did not report cumulative outcomes beyond the first transfer. Clinical pregnancy, miscarriage (adjusted risk ratio 0.75, 95% CI 0.56 to 1.01) and live birth did not differ significantly; early miscarriage, a secondary outcome measured per clinical pregnancy, was lower with noninvasive testing (18.0% against 25.2%, adjusted P = 0.03), with no adjustment for multiple comparisons. The authors concluded that the trial does not support routine use in this good-prognosis population [188]. A further open-label trial in three Chinese centres, with cumulative ongoing pregnancy within 12 months as its primary endpoint, is under way: the protocol reports that recruitment and randomisation began at the lead site on 14 February 2025 (ChiCTR2400088283) [197].

For the laboratory, noninvasive testing is not simply biopsy-free PGT-A. It changes culture practice (individual culture, timed medium changes, complete denudation, contamination controls) and carries its own validation burden. ESHRE's 2023 add-ons recommendations consider noninvasive PGT to be in the research phase [153]. On present evidence it remains a research and validation question rather than a replacement for trophectoderm biopsy.

14.1 Polygenic embryo screening (PGT-P)

Genome-wide genotyping of embryos makes it possible to calculate polygenic risk scores for common diseases such as type 2 diabetes, coronary disease or breast cancer, and the service is already sold to IVF patients in some countries. The first clinical case was reported in 2019 by authors employed by a commercial provider, who modelled a 45% to 72% reduction in type 1 diabetes risk from selecting between adult siblings [198]; authors from a second company later reported genotype accuracy of 99.0% to 99.4% at sites relevant to polygenic scoring in day 5 biopsies, in a study of 110 embryos that combined rare and common variants [199]. An ESHRE-supported review led by Capalbo set out the arguments. Models suggest large relative risk reductions under ideal assumptions, but the realised benefit may be smaller because few embryos are available, many do not result in a live birth, and the accuracy of the scores is uncertain and lower in people of non-European ancestry. The harms identified included unnecessary IVF, lower IVF success, confusing counselling, discarded embryos and wider social effects, and the review concluded that polygenic screening should be offered only within research until the balance of benefit and harm is clearer [9].

A simulation in real IVF data from the same research network, reported in a preprint that has not been peer reviewed, illustrates the gap between theory and practice. Randomly generated polygenic scores were assigned to the patients and embryos of 6,944 stimulation cycles in 4,452 patients treated for infertility, and disease risk was computed with a liability-threshold model: transferring embryos in order of lowest simulated score reduced the child's risk, on a relative scale, by only about 1% to 3% in cycles in which every embryo was transferred and at least one child was born, by under 0.5% across all completed cycles, by about 5% to 10% when all of a patient's cycles were pooled, and by about 20% in egg-donor cycles. These are model-based projections, not observed reductions in disease, and the absolute change depends on the prevalence of the disease [200]. A peer-reviewed retrospective study from one Chinese centre makes the complementary point about embryo numbers: in 310 PGT-M cycles (1,722 embryos), 55% had at least three transferable embryos, and in simulation, ranking by a breast-cancer polygenic score lowered the chosen embryo's mean score percentile by about 12 points compared with morphology. The score discriminated only modestly within families, the outcome is a modelled surrogate, and the authors state that it shows no reduction in disease [201]. Neither study shows prevention of disease in children. Earlier modelling projected average gains of about 2.5 cm for height and 2.5 IQ points for cognitive ability, assuming five viable embryos and the scores then available, with wide prediction intervals; these are projections, not outcomes observed in children [202], and a review in a general medical journal set out further reasons why predictive power falls when scores are used to choose between siblings [203]. The selection strategy also matters: in liability-threshold modelling checked against simulated couples built from real genomes, excluding only embryos with a very high score gave minimal risk reduction, whereas selecting the embryo with the lowest score could give substantial relative reductions when enough viable embryos were available [204].

The positions of professional bodies and regulators rest on clinical utility, ethics and law rather than on any one model's estimate, and they are cautious. The European Society of Human Genetics called the practice unproven and unethical and called for a societal debate on acceptable trait selection before any further implementation [205]; in the United Kingdom the HFEA has stated that PGT-P falls outside every purpose for which the Human Fertilisation and Embryology Act permits embryo testing and is therefore unlawful, and that UK clinicians may not select embryos using PGT-P results obtained abroad [206]; the American College of Medical Genetics and Genomics published a points-to-consider statement on the clinical utility of polygenic scores for embryo selection in 2024 which, as summarised in the College's announcement, found insufficient evidence of clinical utility and did not recommend PGT-P as a clinical service [207]; and the ASRM Ethics and Practice Committees, in an opinion published in 2026, concluded that PGT-P should not be offered clinically at this time and that any use should be confined to research under institutional review board oversight, and excluded nonmedical trait selection from the scope of reproductive medicine [208].

14.2 Beyond copy number: transcriptomics and other signals

Euploidy does not guarantee implantation (Section 10.4), and research is looking for other markers of competence. MicroRNAs in spent blastocyst medium were shown to come mainly from the trophectoderm, with two, miR-20a and miR-30c, at higher levels in the media of implanted than non-implanted euploid blastocysts in a small study [209]. A pilot study has tested transcriptomic profiling of the trophectoderm biopsy (sometimes called PGT-T) [210], and trophectoderm transcriptomics in an in vitro model have been used to look for signatures of euploid embryos that fail to implant [211]. None of these is a clinical test.

14.3 Artificial intelligence

Image-based artificial intelligence is increasingly used to rank embryos. In the 279-cycle study described in Section 3.4, a commercial AI model matched or slightly improved on Gardner grading in ranking a euploid blastocyst first, but no morphology-based ranking method prevented aneuploid embryos from being prioritised [39]. In an earlier retrospective external validation of a deep-learning score, whose authors included employees and shareholders of Vitrolife, on 3,604 blastocysts from 1,232 PGT-A cycles, the area under the curve was 0.60 for euploidy and 0.66 for live birth, comparable to embryologists' performance [212]. This is consistent with the older observation that morphology and morphokinetics are only weakly related to chromosome complement (Section 3.4). Ranking has also been tested in a double-blind randomised trial in 14 clinics in Australia and Europe: 1,066 women under 42 with at least two early blastocysts on day 5 were randomised to selection by a deep-learning ranking algorithm or by morphology with a predefined prioritisation scheme. Clinical pregnancy was 46.5% (248 of 533) against 48.2% (257 of 533), a difference of −1.7 percentage points (95% CI −7.7 to 4.3), and noninferiority within the 5-point margin was not shown. The trial tested ranking for clinical pregnancy, not detection of aneuploidy; its authors included employees and shareholders of the algorithm's manufacturer [213]. On present evidence, AI ranking and PGT-A answer different questions.

14.4 Where is the limit?

Whole-genome sequencing of embryos, and in the longer term genome editing, raise questions of privacy, consent and proportionality that go well beyond the laboratory. A 2025 review argued that variability between methods calls for greater standardisation and transparency, and that new technologies need careful ethical consideration, public engagement and thorough clinical research before they are used [21]. The embryologist's role here is modest but real: a new test reaches patients through the laboratory's biopsy, and the laboratory is entitled to ask for the validation evidence before it becomes routine.

15.1 Two laboratories, one result

Every PGT result is produced jointly by an IVF laboratory and a genetics laboratory, often in different organisations and sometimes in different countries. ESHRE's recommendations on the organisation of PGT state that the examination and reporting must comply with local guidelines or law, or with ISO 15189 (the international standard for quality and competence in medical laboratories), and recommend that each step of the examination process be documented, that the embryo transfer policy be agreed between the IVF centre, the genetics centre, genetic counsellors, clinicians and patients, and that pregnancies and children be followed up; they also describe accreditation together with internal and external quality assessment as the means of maintaining quality [5]. Where biopsy and analysis take place in different centres, ESHRE recommends a service-level agreement covering legal, insurance and accountability issues, validated tubing and shipment protocols, a clear assignment of who is responsible for each stage of treatment, written lines of communication, agreed reporting, and one or more preclinical runs, with negative controls, before clinical samples are sent [5]. In PGT-M and PGT-SR, ESHRE does not recommend transfer of embryos with no result or an inconclusive result; for PGT-A it allows an exception, but only with the patient's fully informed consent [5].

For a laboratory choosing or reviewing a genetics provider, the following questions are this reference's suggestion, drawn from the ESHRE recommendations and the studies cited in Sections 5 to 9 rather than from a published checklist: which platform is used, and whether it detects contamination, ploidy and uniparental disomy (Sections 5.4, 6.3 and 12); what thresholds and categories are reported for mosaic and segmental findings, and what analytical and clinical validation supports each (Sections 6.4, 8 and 9); what its amplification-failure, inconclusive and contamination rates are across client clinics, and whether it will report the laboratory's own rates by operator, with cellularity (Section 15.2); what tubing protocol, buffer, negative controls, shipping conditions and turnaround time it requires and offers; how it handles re-biopsy samples and the exceptions in Table 6 (Section 15.2); and which accreditations it holds (Section 15.3).

15.2 Surveillance after validation

A validated test can drift. Capalbo and colleagues have argued that, once a test is in clinical use, both its analytical stability and its clinical outcomes must be monitored continuously by the genetics laboratory and by the IVF clinics it serves, with large-scale data collection, quality control and follow-up of abnormal pregnancy results, and that this is part of the legal and ethical obligation of genetics and IVF laboratory directors and clinicians [80]. The contamination study described in Section 5.4 shows why the IVF side matters: contamination rates differed significantly between the 32 clinics using the same genetics laboratory and method, which makes the IVF side a natural place to look. The difference is a signal for investigation rather than proof of where the problem lies: case mix, day of biopsy, which embryos were biopsied, transport and assay batch can also differ between clinics, and rates at clinics sending few samples are imprecise [57]. Pooled data give an external benchmark: the ESHRE PGT Consortium's voluntary data collection for 2019 to 2021, covering 14,731 analyses from 45 centres, reported diagnostic efficiency of 87% for PGT-M, 92% for PGT-SR, 98% for PGT-A and 90% for combined testing, and better outcomes after day 5 biopsy than after day 3 or day 6 biopsy, while noting that the data do not reflect global trends [214].

The indicators in Table 5 are this reference's suggestion for what an IVF laboratory offering PGT should track; they are drawn from the sources cited in Sections 5 and 15 but are not a published consensus set. ESHRE's 2026 good practice recommendations ask, in general terms, that performance indicators be defined for a specific patient reference group, monitored with appropriate statistics, and that operator performance be assessed regularly by direct observation of procedural skills or by individual performance indicators [56]. Each indicator needs its denominator (biopsies, embryos, cycles or transfers) and stratification by maternal age band, day of biopsy, operator, transport and assay version, and the set should include no-result and no-transfer rates and the completeness of pregnancy and genetic follow-up, not only outcomes among transferred euploid embryos. Outcome-based monitoring of individual operators is possible but needs care: one centre compared actual with AI-predicted implantation in groups of 20 consecutive single blastocyst transfers per provider and identified one embryologist whose transfers implanted less often than predicted (30% against 60%), but found no significant difference for trophectoderm biopsy (120 biopsies) [215]; with 20 transfers per group, several providers compared and no adjustment for multiple comparisons reported, such a signal is a reason to review practice, and its absence is weak reassurance.

Table 6 sets out what a laboratory might do when a PGT result can no longer be taken at face value. It is this reference's own proposal, not a published operational standard. It builds on ESHRE's general requirements for a documented system for non-conformities with corrective and preventive action, for procedures to withdraw or restrict the release of reproductive material, and for a predefined escalation protocol after any witnessing discrepancy [56], and on the PGT-specific recommendations that a report identify any commercial kit with its version and that discordant re-analyses count towards a centre's misdiagnosis rate [5]. For each trigger, the written agreement with the genetics laboratory (Section 15.1) should name the responsible IVF and genetics leads and the clinician or genetic counsellor who tells the patient (Section 16.4); each resolution should be documented with its audit trail and the name of whoever authorised release, and the report version, assay version and reporting thresholds recorded against each embryo. If an implicated result has already been communicated or used, disclosure follows the clinic's written policy. ASRM's 2026 witnessing opinion asks that significant protocol deviations be escalated to the laboratory and medical directors and that affected patients be alerted as soon as feasible [58]; its Ethics Committee holds that there is an ethical duty to disclose clinically significant errors involving gametes and embryos, and recommends, without making it mandatory, disclosure of near misses [216].

15.3 Regulation in outline

The regulatory layer differs by country and changes over time; the following is an outline, with national law last checked on 30 October 2026, and each laboratory must check the current rules for its jurisdiction. It covers only the United States, the United Kingdom and Europe; laboratories elsewhere, including in Asia-Pacific and Latin America, work under national rules not summarised here.

• United States. PGT is performed by clinical genetics laboratories subject to the Clinical Laboratory Improvement Amendments (CLIA), administered by the Centers for Medicare & Medicaid Services [217], commonly with College of American Pathologists accreditation, and some states impose their own licensing; New York's permit programme, for example, covers laboratories in the state and laboratories elsewhere that test specimens originating in New York, subject to exceptions set out in its programme guide [218]. The embryology laboratory that takes the biopsy is commonly accredited separately, for example under the College of American Pathologists Reproductive Laboratory Accreditation Program, so the two halves of one result sit under different inspection regimes, linked by the written agreement (Section 15.1). ASRM's 2026 committee opinion on witnessing, which is practice guidance rather than regulation, states that critical procedures, including biopsy, be witnessed by a second person and/or a certified electronic witnessing system, that some steps, such as mixing of gametes at insemination or ICSI and identification of embryos and patient before loading the transfer catheter, be validated by a second manual witness, and that an electronic system be validated with continued auditing of the processes it witnesses [58]; in this reference's suggestion, a programme should be able to show which embryo-to-sample transitions its electronic system actually witnesses. FDA's page on laboratory-developed tests records that its May 2024 rule regulating them as medical devices was vacated by a federal district court on 31 March 2025, and that on 19 September 2025 FDA reverted its regulation to the text in force before that rule [219]. ASRM and SART issued a committee opinion on the use of PGT-A in 2024 (Section 10.6) [152].
• United Kingdom. The HFEA licenses PGT-M and PGT-SR for conditions it has approved (a rearrangement known to raise the risk of miscarriage may not need specific approval) [220] and treats PGT-A as a treatment add-on, with separate ratings, summarised in Section 10.6, for the chance of having a baby and for miscarriage [161]. NICE's 2026 fertility guideline recommends not offering PGT-A as part of fertility treatment to improve live birth rates (recommendation 1.48.1) [160], a recommendation about an intended treatment benefit rather than a rating of each outcome. Regulation and funding shape use: in a cross-sectional analysis of national registry data for 2014 to 2016, PGT accounted for just under 2% of ART cycles in the UK (3,385 PGT cycles) but rose from 13% to 27% in the USA (94,935 PGT cycles), differences the authors suggested may be related to funding and to the lack of HFEA support for PGT-A [221].
• Europe. ESHRE's good practice recommendations on add-ons, published in 2023, state that PGT-A and noninvasive PGT are currently not recommended for routine clinical use [153], and the ESHRE PGT Consortium recommendations set out laboratory and organisational standards [1, 5, 7, 8]; ESHRE's 2026 general good practice recommendations for the IVF laboratory add a section on biopsy for PGT but state that they do not establish a standard of care [56]. National law varies widely, from permissive regimes to restrictions on embryo testing or on embryo disposition: in Germany, for example, PGT is lawful only for a high risk of a serious hereditary disease or of serious embryo damage likely to cause stillbirth or miscarriage, after a positive assessment by an interdisciplinary ethics committee and at an approved centre [222], and in Italy the law prohibits the suppression of embryos and allows research on an embryo only for therapeutic and diagnostic purposes directed at its own health and development (Section 16.3) [223]. Regulation (EU) 2024/1938 on substances of human origin, which treats embryos as reproductive substances of human origin, replaces the EU tissues and cells legislation and applies generally from 7 August 2027, with exceptions in Article 87, including provisions that apply from 7 August 2028; it sets standards of quality and safety and does not affect national law on other aspects (Article 2(3)) [224].

16.1 What the embryologist decides and what the patient decides

Most ethical decisions in PGT belong to the patient, informed by genetic counselling and the clinician. The laboratory's responsibilities are narrower but real: to produce a result that can be trusted, to keep every embryo's identity intact, to report accurately what was and was not tested, and not to let laboratory convenience decide an embryo's fate. Three examples from earlier sections show where laboratory policy has ethical weight: discarding blastocysts because they look poor (Section 5.7), discarding every atypically pronucleated zygote on day 1 without a written policy for each pronuclear category (Section 12), and treating a mosaic result as if it were a uniform aneuploidy (Section 8). Where such policies limit which embryos are cultured, biopsied, stored or offered for transfer, patients should know before treatment: ESHRE recommends that decisions about which embryos are acceptable for transfer or vitrification be discussed with patients before a treatment cycle begins [5].

16.2 Nonmedical selection

A PGT-A analysis can generate information on the sex chromosomes whether or not the report discloses it; whether it does, and whether that information is validated, reported or may lawfully be used, depends on the assay's coverage and validation, the reporting policy and the legal setting. ESHRE states that selection of embryos by sex for social reasons is not acceptable [8]. The ASRM Ethics Committee takes a different position: it describes nonmedical sex selection as ethically controversial, holds that practitioners are under no ethical obligation either to provide or to refuse it, and encourages clinics to develop and make available their policies [225]; a separate ASRM opinion states that clinics may decline to select embryos by sex when sex is revealed incidentally by PGT [226]. The law in many countries prohibits nonmedical sex selection, while in others, including the United States, it is permitted. In the UK the Human Fertilisation and Embryology Act bars any licensed practice designed to secure a child of one sex rather than the other, and allows the sex of an embryo to be determined only where there is a particular risk of a serious gender-related condition [227]; the Council of Europe's Oviedo Convention prohibits sex selection, except to avoid serious hereditary sex-related disease, in the states that have ratified it [228]. Laboratories should know the legal position where they work and have a written policy on how sex-chromosome information is reported and used. The ASRM ethics opinion on polygenic testing excludes nonmedical trait selection from the scope of reproductive medicine [208]. Public opinion draws a similar line: in a nationally representative US survey of 1,427 adults, 77.7% approved of polygenic embryo selection for physical health conditions but only 30.3% for physical traits, and 780 of 1,423 (54.8%) found the risk of promoting eugenic practices very to extremely concerning [229].

16.3 What happens to embryos that are not transferred

PGT leaves some embryos awaiting a disposition decision: those diagnosed as affected or aneuploid, those with mosaic or segmental results, and those with no result. Where the law allows a choice, patients may keep them in storage, discard them or donate them to research. In a survey of patients with affected or aneuploid embryos at an Italian centre, where the law prohibits both discarding embryos and donating them to research, so that they remain in storage, 126 of 149 respondents (84%) said they would donate them to research if allowed, although only 18% of the 832 patients approached responded, so the 126 represent 15% of those approached [230]. Italian law still prohibits the suppression of embryos (Article 14) and permits research on an embryo only for therapeutic and diagnostic purposes directed at its own health and development where no alternative exists (Article 13); in judgment 84/2016 the Constitutional Court declared inadmissible a challenge to that restriction as applied to surplus embryos that are affected or not suitable for transfer, leaving any change to the legislature (law as checked on 30 October 2026) [223]. Two Italian scientific societies have published criteria for defining embryos that are developmentally incompetent, taking genetic results into account, for jurisdictions where only non-viable embryos may be discarded [231].

For the laboratory the point is practical as well as ethical: the number of stored embryos awaiting a disposition decision grows, every stored embryo must remain correctly identified for as long as it is kept, and the consent forms should address disposition before the cycle begins. In this reference's suggestion, that consent names each result category (affected or aneuploid, mosaic, segmental, no result) and the options for each, says whether re-biopsy is accepted, and states the storage term and who pays for it. The cryostorage inventory should keep separate fields for the laboratory result as reported, with its report version; the clinical and counselling review; the patient's documented decision; any legal constraint; and the current transfer authorisation, so that an embryo not currently authorised for transfer cannot be warmed for transfer by mistake. A result category is not a fixed property of the embryo: mosaic, segmental and no-result embryos are not inherently untransferable, a PGT-M result for a condition with incomplete penetrance or variable expressivity may not settle eligibility, an amended report or a change in clinical policy can alter what a result means, and ESHRE notes that decisions about which embryos are acceptable for transfer may need to be revisited [5]. The ASRM Ethics Committee considers that in most cases it is ethically permissible for a physician to assist or to decline to assist a request to transfer an embryo affected by a monogenic condition, but that transfer is highly discouraged where a child is highly likely to be born with a life-threatening condition causing severe and early debility [232]. In this reference's recommendation, any change to an embryo's transfer authorisation, and any use of an embryo for staff training, validation or research, follows a documented review against the current report, the patient's consent and the law. Such use is one the patients must have specifically consented to; ASRM advises that, without specific written instructions, unclaimed embryos may not be used in research [233].

16.4 Access, cost and honest communication

PGT adds cost, and in many countries the patient pays it. Whether it is good value depends on the patient group and on what outcome is valued (Section 10.5), and the professional bodies differ in their assessments (Section 10.6). Cost also shapes access: PGT use differs widely between countries, which the authors of a registry comparison suggested may relate to funding and regulation [221], and polygenic scores are less accurate in people of non-European ancestry (Section 14.1). Laboratory staff who talk to patients, at a consultation, a tour or a transfer, should be able to say plainly what the test does, what it does not do and where the evidence is still disputed. They should not advise whether a particular patient should have PGT, interpret an individual embryo's result, quote residual risks, or advise on transferring a mosaic or segmental embryo or on which prenatal test to have; this reference recommends that every laboratory have a clear route for referring such questions to the clinician and genetic counsellor.