JBRA Assisted Reproduction 2025;29(Suppl.2 SBRA 2025):9
Oral Presentation
29^th Annual Congress of the SBRA. São Paulo/SP - Brazil, 2025
doi: 10.5935/1518-0557.20263523

O-09. Not All That Glitters Is Safe: Morphology, Time-Lapse, and AI Systems Fail to Deselect High-Quality Embryos with Simple Aneuploidies

Sebastião Evangelista Torquato¹, Eduardo Gomes Sá¹, Ellayne Cavalcanti Queiroz¹, Andrea Mesquita Lima¹, Gleicyane Sousa Santos Alam¹, Renata Reis Pimentel Castro¹, Maria Clara Parente Torquato¹, Marcus Aurélio Paiva Bessa¹, Eduardo Paula Miranda¹, Hamilton de Martin¹

¹Bios Centro de Medicina Reprodutiva - Fortaleza - CE - Brasil

Objective: To evaluate the ability of conventional morphology (Gardner), KIDScore (AI-based), and Blastocyst Fragmentation Indicator (BFI) to identify blastocysts with simple and complex chromosomal abnormalities, and to assess the performance of these systems specifically for single whole-chromosome aneuploidies involving small classical chromosomes (13, 18, 21, 22).
Methods: This retrospective cohort included 1,182 blastocysts individually cultured in time-lapse incubators and biopsied for preimplantation genetic testing for aneuploidy (PGT-A) between September 2022 and December 2024. Embryos were classified as euploid, monosomic, trisomic, complex (simultaneous chromosomal gain and loss), or chaotic (three or more abnormalities). Gardner scores were grouped as A (AA, AB, BA), B (BB, BC, AC), and C (CA, CB, CC). KIDScore groups (v3.1) were defined as A (score ≥7), B (score 4–7), and C (score <4). BFI classes were assigned as NN (no fragmentation), FN (fragmentation of the inner cell mass only), and FF (fragmentation of both the inner cell mass and trophectoderm). Multinomial generalized linear models assessed the association between each scoring system and PGT-A outcome, adjusting for maternal age. Odds ratios and post hoc analyses quantified discriminatory power for each aneuploidy type, with particular focus on high-quality embryos.
Results: All three scoring systems were significantly associated with increased risk of complex and chaotic aneuploidies (p<0.001 for all comparisons). Embryos classified as C by Gardner, C by KIDScore, or FF by BFI had markedly higher odds of being chaotic or complex, with BFI FF versus NN yielding an odds ratio (OR) of 72.2 (95% CI: 43.0–233.3) for chaotic aneuploidy. For complex aneuploidies, the OR was 20.7 (13.8–63.3) for BFI FF versus NN, and 2.65 (1.85–4.26) for Gardner C versus A. In contrast, the highest quality categories (Gardner A, KIDScore A, BFI NN) were enriched for euploid embryos. However, a substantial fraction of high-quality embryos remained undetected for single-chromosome aneuploidies. Among all embryos with high morphological or morphokinetic quality, 22.9% of those graded as Gardner A, 22.0% of KIDScore A, and 20.5% of BFI NN were monosomic or trisomic. Odds ratios for monosomic or trisomic versus euploid status were only modestly elevated for B or FN categories (Gardner B vs A: OR=2.06, 95% CI: 1.22–4.46; KIDScore B vs A: OR=2.23, 1.26–4.74; BFI FN vs NN: OR=7.46, 4.29–16.64), with overlapping confidence intervals, indicating limited discriminatory capacity. When focusing on small classical chromosomes, these limitations became more pronounced. Of all embryos with single aneuploidies in chromosomes 13, 18, 21, or 22, 62% of monosomies and 66% of trisomies were classified as Gardner A. Similarly, 71% of monosomies and 58% of trisomies were BFI NN, and 54% of monosomies and 47% of trisomies received a KIDScore A. Only a minority were categorized as C or FF, highlighting a blind spot in all three systems.
Conclusion: Morphological, morphokinetic, and fragmentation-based scoring systems are highly effective in deselecting embryos with complex or chaotic aneuploidies, but all fail to reliably exclude high-quality blastocysts with simple whole-chromosome aneuploidies, particularly those involving chromosomes 13, 18, 21, or 22. This diagnostic gap persists across all three platforms (Gardner, KIDScore, and BFI), highlighting the indispensable role of PGT-A in clinical embryo assessment. PGT-A remains essential for detecting clinically significant single-chromosome aneuploidies, and patients should be counseled about these limitations. This study represents a focused segment of a larger, ongoing investigation.