Key points are not available for this paper at this time.
Abstract Study question How do iDAScore® and KIDScoreTM perform in predicting live birth outcomes after single frozen Day 5 blastocyst transfers? Summary answer In a euploid major cohort, both KIDScoreTM and iDAScore® demonstrated comparable predictive ability for live birth. However, iDAScore® offers the advantage of being fully automated. What is known already In the field of in vitro fertilization (IVF), the escalating use of artificial intelligence (AI) for predicting clinical outcomes is a burgeoning trend. One key contributor to this trend is time-lapse incubation, which captures images of developing images, generating a wealth of data. This data has facilitated the creation of various predictive models for implantation. KIDScoreTM is one such model that uses morphokinetic data to predict the probability of implantation after a Day 5 embryo transfer. Similarly, iDAScore® is an AI-driven automated scoring system specifically designed to automatically identify embryos with the greatest likelihood of implantation. Study design, size, duration This retrospective study included 779 patients (maternal age range: 19-44, mean age: 33.38) who underwent single frozen day 5 blastocyst transfer cycle from January 2021 to January 2023. Of these, 621 patients transferred euploid blastocysts, 48 transferred mosaic blastocysts, and 110 transferred non-PGT-A tested blastocysts. This study comprised of 12.1% (94/779) fair-to-good and fair grade (below AB and BA), 51.5% (401/779) good grade (AB and BA), and 36.4% (284/779) excellent grade (AA) blastocysts. Participants/materials, setting, methods After PIEZO-ICSI, embryos were cultured in EmbryoScopeTM Plus (Vitrolife, Denmark) until they reached the blastocyst stage. Blastocysts graded at least as fair according to Gardner (1999), were either biopsied for PGT-A and vitrified (Cryotec, Japan), or vitrified without biopsy. Clinical pregnancy and the number of gestational sacs were determined using ultrasound and successful pregnancies were tracked until delivery. KIDScoreTM (KIDScoreTM Day 5 v3.2) and iDAScore® (iDAScore® v2.0) were retrospectively generated for each transferred blastocyst. Main results and the role of chance Box plot showed that both KIDScoreTM and iDAScore® have positive correlation with morphology (Excellent vs. Good vs. Fair-to-Good and Fair; KIDScoreTM median: 8.8 vs. 6.8 vs. 5.7; iDAScore® median: 8.57 vs. 7.68 vs. 6.615), ICM (A vs. B; KIDScoreTM median: 7.5 vs. 6.5; iDAScore® median: 8.14 vs. 7.26) and TE (A vs. B vs. C; KIDScoreTM median: 8.7 vs. 6.5 vs. 4.65; iDAScore® median: 8.41 vs. 7.515 vs. 5.2). In this cohort, the overall implantation rate was 67.91% (529/779), and the live birth rate was 59.31% (462/779). Further statistical analysis revealed interesting correlations. A linear regression analysis indicated a statistically significant positive correlation between KIDScoreTM and iDAScore® (F(1,777)=297.6, p 0.001, adjusted R2=0.276). Wilcoxon rank sum tests showed that KIDScoreTM has significant higher mean scores in blastocysts resulting in live birth compared to those that did not (7.44 vs. 7.22, w = 66577, p 0.05). iDAScore® also had higher mean scores in blastocysts resulting in live birth but without statistically significant difference (7.56 vs. 7.48, w = 71328, p 0.05). ROC curve analysis showed that AUC for KIDScoreTM was 54.4 and AUC for iDAScore® was 51.9, indicating no significant difference in their predictive ability for live birth (p = 0.2455). Limitations, reasons for caution The study’s cohort predominantly consisted of blastocysts with good morphology and were euploid, potentially skewing the results. Therefore, the findings may not be representative of all IVF patients and should be interpreted with caution. Additionally, further study should investigate the capability of iDAScore® 2.0 to predict euploidy. Wider implications of the findings While KIDScoreTM had a higher AUC than iDAScore®, there was no statistically significant difference. The major advantage of iDAScore® lies in its automation, eliminating the need for manual embryo evaluation. This can greatly enhance consistency in embryo grading and selection. Trial registration number NOT APPLICABLE
Chan et al. (Mon,) studied this question.