Human embryos with segmental aneuploidies display delayed early development: a multi-centre morphokinetic analysis.

Objectives: To assess if segmental aneuploid embryos display unique morphokinetic patterns DESIGN: Retrospective multicentre study including a total of 7,027 embryos cultured between 2016 and 2021 in 3 European IVF centres. Analysis was performed on aggregated multicentre data and separately for data from each centre. Embryos with no more than 4 chromosomal alterations were considered in the analysis, resulting in 3,040 euploids, 2,818 whole-chromosome and 697 segmental aneuploids. Overall, the dataset contained 3,742 distinct euploid-segmental sibling pairs.

Subjects: Standard morphokinetic features were annotated using various time-lapse systems. Blastocysts were subjected to comprehensive chromosomal screening via PGT-A.

Exposure: Morphokinetic patterns were compared among euploid, whole-chromosome aneuploid, and segmental aneuploid embryos.

Main outcomes measures: Morphokinetic timings across groups were compared using statistical analysis and associations with cleavage features were assessed. Multi-centre and centre-specific multivariate logistic regression models were calibrated, and their predictive performance was evaluated on independent test set data using Area-Under-ROC curve metrics.

Results: Segmental aneuploid embryos cleaved significantly slower than their euploid siblings across the first three cell cycles, with a delay reaching the blastocyst stage of development. Specifically during these early cell cycles, segmental aneuploid embryos were also shown to be significantly slower than their aneuploid siblings . A logistic model based on morphokinetic data from the multicentre dataset and regressed against type of aneuploidy displayed modest predictive performance on an independent test set (train-AUROC=0.58; test-AUROC=0.57). Predictive performance improved based on data from a single centre displaying adequate predictive performance on an independent test set from the same centre (train-AUROC=0.74; test-AUROC=0.64). However, the predictive value diminished when tested on data from other centres (AUROC=0.52-0.55). Finally, the presence of multinucleation and blastomere exclusion at cleavage stage, were associated with segmental aneuploidies. The combination of morphokinetic features and these discrete embryo morphological features into the logistic regression model (train AUROC=0.71) provided an improved prediction of segmental aneuploidy, supporting future investigations using more comprehensive annotation systems.

Conclusion: The developed predictive framework might help improve decision-making in PGT-A cycles, helping in the evaluation of embryos showing segmental aneuploidy and distinguishing which embryos are more likely to not have lethal uniform aneuploidies for transfer.