Abstract
Study question
What is the predictive performance of iDAScore® compared with KIDScore™D5 and conventional morphology grading for implantation after frozen blastocyst transfer?
Summary answer
iDAScore®, KIDScore™D5 models and conventional morphology grading are all significantly associated with implantation after frozen blastocyst transfer. However, their predictive performance remains fair.
What is known already
Embryo selection algorithms (ESA) have been used in conjunction with conventional morphology grading (CMG) to rank embryo quality in an attempt to optimise embryo selection prior to transfer. Traditionally, ESA, such as the KIDScore™D5 prediction model, have been based on predetermined morphokinetic parameters which vary according to the specific ESA used. Recently, algorithms based on artificial intelligence (AI) deep learning such as the iDAScore®, have been developed using raw time lapse image data of embryos with known implantation outcome. Embryo scores generated by both traditional and AI based ESA have been positively correlated with implantation potential in retrospective studies.
Study design, size, duration
This retrospective single centre study included data from 157 frozen embryo transfers carried out between January 2020 and January 2021 with embryos cultured in EmbryoScope+ time-lapse incubator. Embryos were selected for transfer using CMG. iDAScore® and KIDScore™D5 were generated retrospectively for each embryo transferred. Sensitivity and specificity for implantation in addition to concordance was determined for all three embryo scoring systems and compared.
Participants/materials, setting, methods
Statistical analysis was performed with SPSS software. ROC curve analysis was performed to investigate the predictive performance of the three embryo selection methods for implantation. Chi-Square test was used to determine if there was a significant association between iDAScore®, KIDScore™D5 and CMG with implantation. Spearman’s correlation tested for correlation between iDAScore®, KIDScore™D5 and CMG.
Main results and the role of chance
A statistically significant but limited predictive power was observed between iDAScore®, KIDScore™D5 and CMG methods of embryo selection and implantation rate (AUC=0.675, 95% CI 0.59, 0.76; AUC=0.683, 95% CI 0.60, 0.77 and AUC=0.638, 95% CI 0.55, 0.73) respectively. Unsurprisingly, higher values for each of iDAScore®, KIDScore™D5 and CMG were significantly associated with higher implantation (p = 0.002, p < 0.001, p = 0.003) respectively. Interestingly, there was a significantly moderate correlation between iDAScore® and CMG rs (155)=0.581, p < 0.001, while a significantly strong correlation between KIDScore™D5 and CMG rs (155)=0.775, p < 0.001 as well as between iDAScore® and KIDScore™D5 rs (155)=0.799, p < 0.001.
Limitations, reasons for caution
This is a retrospective single centre study with limited sample size, hence the results may be interpreted with caution. Further analysis involving a larger patient cohort must be carried out prior to implementing iDAScore® or KIDScore™D5 as primary embryo selection methods.
Wider implications of the findings: The findings of this study support the potential application of iDAScore® and KIDScore™D5 prediction models in automatic embryo selection, thereby minimising operator variability and time taken for embryo grading.
Trial registration number
Not applicable