BACKGROUND A time lapse system (TLS) is utilized in some fertility clinics with the aim of predicting embryo viability and chance of live birth during IVF. It has been hypothesized that aneuploid embryos display altered morphokinetics as a consequence of their abnormal chromosome complement. Since aneuploidy is one of the fundamental reasons for IVF failure and miscarriage, attention has focused on utilizing morphokinetics to develop models to non-invasively risk stratify embryos for ploidy status. This could avoid or reduce the costs associated with pre-implantation genetic testing for aneuploidy (PGT-A). Furthermore, TLS have provided an understanding of the true prevalence of other dysmorphisms. Hypothetically, the incorporation of morphological features into a model could act synergistically, improving a model’s discriminative ability to predict ploidy status. OBJECTIVE AND RATIONALE The aim of this systematic review and meta-analysis was to investigate associations between ploidy status and morphokinetic or morphological features commonly denoted on a TLS. This will determine the feasibility of a prediction model for euploidy and summarize the most useful prognostic markers to be included in model development. SEARCH METHODS Five separate searches were conducted in Medline, Embase, PubMed and Cinahl from inception to 1 July 2021. Search terms and word variants included, among others, PGT-A, ploidy, morphokinetics and time lapse, and the latter were successively substituted for the following morphological parameters: fragmentation, multinucleation, abnormal cleavage and contraction. Studies were limited to human studies. OUTCOMES Overall, 58 studies were included incorporating over 40 000 embryos. All except one study had a moderate risk of bias in at least one domain when assessed by the quality in prognostic studies tool. Ten morphokinetic variables were significantly delayed in aneuploid embryos. When excluding studies using less reliable genetic technologies, the most notable variables were: time to eight cells (t8, 1.13 h, 95% CI: 0.21–2.05; three studies; n = 742; I2 = 0%), t9 (2.27 h, 95% CI: 0.5–4.03; two studies; n = 671; I2 = 33%), time to formation of a full blastocyst (tB, 1.99 h, 95% CI 0.15-3.81; four studies; n = 1640; I2 = 76%) and time to expanded blastocyst (tEB, 2.35 h, 95% CI: 0.06–4.63; four studies; n = 1640; I2 = 83%). There is potentially some prognostic potential in the degree of fragmentation, multinucleation persisting to the four-cell stage and frequency of embryo contractions. Reverse cleavage was associated with euploidy in this meta-analysis; however, this article argues that these are likely spurious results requiring further investigation. There was no association with direct unequal cleavage in an embryo that progressed to a blastocyst, or with multinucleation assessed on Day 2 or at the two-cell stage. However, owing to heterogeneous results and poor-quality evidence, associations between these morphological components needs to be investigated further before conclusions can be reliably drawn. WIDER IMPLICATIONS This first systematic review and meta-analysis of morphological and morphokinetic associations with ploidy status demonstrates the most useful morphokinetic variables, namely t8, t9 and tEB to be included in future model development. There is considerable variability within aneuploid and euploid embryos making definitively classifying them impossible; however, it is feasible that embryos could be prioritized for biopsy. Furthermore, these results support the mechanism by which algorithms for live birth may have predictive ability, suggesting aneuploidy causes delayed cytokinesis. We highlight significant heterogeneity in our results secondary to local conditions and diverse patient populations, therefore calling for future models to be robustly developed and tested in-house. If successful, such a model would constitute a meaningful breakthrough when accessing PGT-A is unsuitable for couples.

STUDY QUESTION Are machine learning methods superior to traditional statistics in predicting blastocyst ploidy status using morphokinetic and clinical biodata? SUMMARY ANSWER Mixed effects logistic regression performed better than all machine learning methods for ploidy prediction using our dataset of 8147 embryos. WHAT IS KNOWN ALREADY Morphokinetic timings have been demonstrated to be delayed in aneuploid embryos. Machine learning and statistical models are increasingly being built, however, until now they have been limited by data insufficiency. STUDY DESIGN, SIZE, DURATION This is a multicentre cohort study. Data were obtained from 8147 biopsied blastocysts from 1725 patients, treated from 2012 to 2020. PARTICIPANTS/MATERIALS, SETTING, METHODS All embryos were cultured in a time-lapse system at nine IVF clinics in the UK. A total of 3004 euploid embryos and 5023 aneuploid embryos were included in the final verified dataset. We developed a total of 12 models using four different approaches: mixed effects multivariable logistic regression, random forest classifiers, extreme gradient boosting, and deep learning. For each of the four algorithms, two models were created, the first consisting of 22 covariates using 8027 embryos (Dataset 1) and the second, a dataset of 2373 embryos and 26 covariates (Dataset 2). Four final models were created by switching the target outcome from euploid to aneuploid for each algorithm (Dataset 1). Models were validated using internal–external cross-validation and external validation. MAIN RESULTS AND THE ROLE OF CHANCE All morphokinetic variables were significantly delayed in aneuploid embryos. The likelihood of euploidy was significantly increased the more expanded the blastocyst (P < 0.001) and the better the trophectoderm grade (P < 0.01). Univariable analysis showed no association with ploidy status for morula or cleavage stage fragmentation, morula grade, fertilization method, sperm concentration, or progressive motility. Male age did not correlate with the percentage of euploid embryos when stratified for female age. Multinucleation at the two-cell or four-cell stage was not associated with ploidy status. The best-performing model was logistic regression built using the larger dataset with 22 predictors (F1 score 0.59 for predicting euploidy; F1 score 0.77 for predicting aneuploidy; AUC 0.71; 95% CI 0.67–0.73). The best-performing models using the algorithms from random forest, extreme gradient boosting, and deep learning achieved an AUC of 0.68, 0.63, and 0.63, respectively. When using only morphokinetic predictors the AUC was 0.61 for predicting ploidy status, whereas a model incorporating only embryo grading was unable to discriminate aneuploid embryos (AUC = 0.52). The ploidy prediction model’s performance improved with increasing age of the egg provider. LIMITATIONS, REASONS FOR CAUTION The models have not been validated in a prospective study design or yet been used to determine whether they improve clinical outcomes WIDER IMPLICATIONS OF THE FINDINGS This model may aid decision-making, particularly where pre-implantation genetic testing for aneuploidy is not permitted or for prioritizing embryos for biopsy. STUDY FUNDING/COMPETING INTEREST(S) No specific funding was sought for this study; university funds supported the first author. A.Ca. is a minor shareholder of participating centres. TRIAL REGISTRATION NUMBER N/A.

No abstract

No abstract