Non-invasive assessment of embryo development and ploidy: morphokinetics predict blastulation, and exploratory...

To identify more reliable tools for embryo selection, this study aimed to investigate whether embryonic kinetic parameters and metabolomic profiling of spent culture medium on D3 can predict subsequent blastocyst development and chromosomal euploidy.

This prospective diagnostic study analyzed 1,260 embryos from 111 PGT-M or PGT-A patients (April 2024-July 2025) cultured in time-lapse incubators. For the morphokinetic analysis, generalized linear mixed models with patient ID as a random effect were used to identify kinetic parameters associated with blastocyst formation, and the final model was internally validated using a 70/30 hold-out split. Additionally, metabolomic profiling of D3 spent culture medium from 33 patients was performed using LC-MS, with features prioritized by VIP >1.0 and FDR controlled via the Benjamini-Hochberg procedure for multiple comparisons.

Our results delineated specific abnormal cleavage patterns (Direct cleavage, Uneven blastomere, Blastomere fragmentation) and most kinetic parameters (all except t3 and t5) were significantly associated with usable blastocyst formation potential (P < 0.05) but showed no association with ploidy status (P > 0.05). Optimal developmental range was identified for each parameter, and embryos developing within these windows demonstrated significantly higher usable blastocyst formation rates (P < 0.001). Our model demonstrated good performance in the training set (AUC = 0.826) and maintained acceptable discrimination in the validation set (AUC = 0.724), comparable to the KID score (AUC = 0.738; DeLong test P = 0.682). In contrast, metabolomic screening identified 86 VIP-ranked features; stringent FDR correction rendered none significant (all adjusted P-value > 0.05). Based on endogenous retention among the VIP-ranked features, 11 metabolites were retained as exploratory candidates; six of which-sphingosine, phytosphingosine, N,N-dimethylsphingosine, histidylproline, L-valine, and indoleacetaldehyde-displayed AUC >0.7 trends (pooled n = 17 euploid, 10 aneuploid) requiring independent validation.

Our results demonstrate that early kinetic parameters robustly predict blastocyst formation, whereas D3 SCM metabolomics provides only tentative, effect-size-dependent clues for ploidy status that require cautious interpretation. These complementary findings lay the groundwork for a future integrated noninvasive assessment strategy on D3.