Original conceptual illustration, not study data, a micrograph or a laboratory protocol. Objects are not to scale; the depicted embryos do not imply pooling of samples. Source: Talwar et al., Scientific Reports, 2026.
Testing an embryo without removing cells sounds like an obvious advance. But reducing invasiveness and helping choose an embryo are separate questions. A new cohort compares outcomes after transfer; it does not establish which strategy gives patients a better cumulative chance of a baby from an entire started IVF cycle.
Two routes to genetic information
In the biopsy-based PGT-A approach studied here, cells were taken from the trophectoderm. With non-invasive PGT-A, or niPGT-A, DNA in spent embryo culture medium was analysed. These are different biological materials. Avoiding biopsy describes a feature of the procedure, not proof that its results support transfer decisions equally well. The methods are described by Talwar and colleagues.
It helps to separate “Can the laboratory obtain a result?” from “Does that result help the patient?” The first concerns technical performance. The second connects information to a decision and its consequences. A simpler way to collect material still needs that assessment. This is not an argument against innovation; it is a condition for comparing innovations fairly.
What did the study compare?
The retrospective analysis covered one network of 48 centres, with 6,215 ICSI cycles and 15,635 tested blastocysts. Clinical outcomes were evaluated for 3,179 frozen embryo transfers: 2,231 after niPGT-A and 948 after biopsy-based PGT-A. Table 2 reports live-birth percentages of 34.5% and 46.2%, respectively; how missing follow-up was handled in these descriptive percentages requires clarification, as discussed below. In a model adjusting for parental age and endometrial thickness, the odds ratio for PGT-A was 1.95 (95% confidence interval 1.65–2.31). That live-birth model included 2,930 cases.
Crucially, patients chose the method after counselling; they were not randomly assigned. The association favoured biopsy-based testing, but it cannot establish that biopsy itself caused the better outcome. The groups may also have differed in other characteristics affecting prognosis. This distinction belongs alongside the result, not in a footnote readers are unlikely to notice. See the methods and Table 2 in the Article in Press.
The most important word: denominator
When patients hear “the chance of a baby”, they often imagine the whole journey: starting treatment, obtaining eggs, developing embryos and using available opportunities for transfer. An outcome calculated per transfer describes a narrower part of that journey. It does not automatically include people who never reached transfer.
Imagine two strategies that select embryos differently. Comparing only completed transfers does not reveal the entire story of embryos not selected or patients without a transfer. Assessing cumulative effectiveness requires a clear starting point, a definition of which subsequent transfers count and a follow-up period. Without these, a per-transfer result cannot become a promise per started cycle.
This is more than a wording preference. A denominator determines whom an estimate describes. A number can be correctly reported yet misleading when applied to a broader group than the analysis included. Counselling should make that boundary visible before discussing an individual patient’s expectations.
An odds ratio is not a doubling of probability
An odds ratio is different from a risk or probability ratio. A value close to two does not justify saying “twice the chance of a baby” if that phrase is understood as a twofold increase in probability. The safer approach is to name the measure and the model rather than translate it into an attractive but imprecise promise.
Nor should the difference between the reported percentages be converted into a number needed to treat. Such a calculation would suggest a causal benefit of choosing one intervention. Here, the way groups were formed remains an important limitation. The statistical result must stay connected to its study design.
Statistical adjustment does not replace randomisation
Including selected factors in a model answers a conditional question: what does the association look like after accounting for those particular variables? It does not mean everything influencing method selection and outcome has been balanced. A factor cannot be adequately adjusted for simply because its potential importance is recognised; appropriate data and assumptions are needed.
In a clinic-network analysis, it is also reasonable to ask whether outcomes reflect the test, team experience, selection practices or several elements together. A large organisation can provide valuable evidence without representing every laboratory. Multiple centres within one network should not automatically be described as independent external validation across unrelated services.
What about missing follow-up?
The 34.5% and 46.2% figures are percentages reported by the authors in Table 2A. That row does not explicitly provide birth counts and separate denominators after missing outcomes are considered. We therefore do not reconstruct numbers of babies from rounded percentages. A detailed comparison requires clarification of how incomplete follow-up was handled in that descriptive row.
The authors recorded missing live-birth outcomes for 119 people after niPGT-A and 130 after PGT-A. The final model consequently has fewer cases than the transfer dataset. Subtracting those missing outcomes yields the group sizes reported for the model. Missing observations should not simply be counted as either failures or successes.
There is a further reporting issue: the missing-follow-up percentage stated for the PGT-A group does not agree with dividing its reported missing count by its transfer total. The available text does not explain a different denominator. We retain the absolute missing counts rather than repeat that disputed percentage or silently replace the authors’ figure. This is a request for clarification, not evidence that a particular part of the analysis must be wrong.
If missing outcomes are not distributed randomly, their absence can affect interpretation. Completeness of follow-up therefore deserves attention alongside statistical significance. A large initial sample does not make the missing-data question disappear. Relevant sections include data collection and Tables 2A and 2G.
This was not PGT-A versus no testing
A patient planning IVF may ask a different question: is PGT-A worth considering in my situation at all? A comparison of two testing approaches cannot answer that by simple extension. Showing an association favouring test A over test B does not automatically establish a benefit of test A over using neither.
Likewise, the findings should not be treated as an assessment of every preimplantation testing technology. Each method has a defined purpose. Counselling needs to identify that purpose before addressing its benefits, limitations and alternatives for a particular couple.
Questions to ask before choosing a method
- What clinical problem is the proposed test intended to address?
- Does the quoted outcome concern one transfer or the whole course of treatment?
- How has the laboratory validated the method in its own practice?
- What happens when the result is inconclusive or no result is obtained?
- Which alternatives have been discussed, and how will the final outcome be assessed?
These questions do not replace genetic and embryological counselling. They help distinguish an appealing description of technology from an answer to the patient’s actual needs. A good consultation need not end with choosing the newest test. It should end with understanding what the test can and cannot be expected to deliver.
The takeaway
“Without biopsy” describes how material is obtained. It is not a synonym for demonstrated clinical equivalence. This cohort supports careful scrutiny of niPGT-A, but it does not justify causal promises about biopsy or converting the result into an individual prognosis. Decisions require discussion of indication, validation and the right denominator.
Source and transparency
Talwar D et al. Comparative clinical and reproductive outcomes following non-invasive versus biopsy-based preimplantation genetic testing for aneuploidy: a multicentric cohort study. Scientific Reports, Article in Press, 2026. DOI: 10.1038/s41598-026-69506-2.
The authors declared no competing interests. This version contains no separate funding statement; that does not establish an absence of funding. Several authors are affiliated with the clinic network studied. The publisher’s version may undergo further editing.
Educational material, not an individual recommendation for embryo testing. The questions and illustrative examples are editorial commentary, not additional study findings.