# Obstetric and perinatal outcomes from the follow-up of a multicentre randomized controlled trial investigating time-lapse embryo monitoring

## TL;DR

A study found that using time-lapse embryo monitoring with or without machine learning does not increase pregnancy risks compared to traditional methods.

## Contribution

This study provides empirical evidence on the safety of time-lapse embryo monitoring in assisted reproduction.

## Key findings

- No significant differences in serious pregnancy complications across groups using time-lapse monitoring versus traditional methods.
- Birth weights and gestational ages were similar among all groups, indicating comparable perinatal outcomes.

## Abstract

Does uninterrupted culture in a time-lapse incubator with or without a commercially available machine learning embryo selection algorithm result in comparable obstetric and perinatal outcomes as interrupted culture and morphological embryo selection?

The application of uninterrupted culture in a time-lapse incubator with and without the use of an embryo selection algorithm is comparable to interrupted embryo culture and morphological embryo selection in terms of obstetric and perinatal results.

There is very limited evidence regarding the safety of time-lapse monitoring (TLM) from prospective randomized controlled trials (RCT). Recent RCTs have demonstrated that the application of TLM does not increase (cumulative) live birth rates or shorten the time to pregnancy within 1 year. Although most studies only report pregnancy rates, the safety of this commonly used method is also relevant for decision-making.

The obstetric and perinatal outcomes of patients scheduled for Day 3 single embryo transfer who participated in a multicentre RCT on TLM were studied (SelecTIMO trial). Three groups were compared: (i) TLE: embryo selection based on a commercially available Day 3 TLM algorithm, used adjunctively with morphology, and uninterrupted culture. (ii) TLR: routine morphological embryo selection and uninterrupted culture. (iii) CON: routine morphological embryo selection and interrupted culture.

In total, 1731 IVF/ICSI patients undergoing their first, second, or third oocyte retrieval cycle were randomized. Obstetric and perinatal data were registered for all pregnancies occurring after fresh and frozen embryo transfers associated with the initial oocyte retrieval cycle as well as natural conceptions within 1 year. Serious pregnancy complications and birth weight were considered main safety outcomes. Mean differences (MD) and age-adjusted relative risks (RRadj) and mean differences with 95% CI were calculated for TLE and TLR versus CON.

A total of 827 women gave birth to a singleton during the follow-up period (TLE = 275, TLR = 278, CON = 274; P = 0.99). Of the 827 women who gave birth to a singleton, 497 deliveries originated from a fresh embryo transfer (60%), 294 from a frozen embryo transfer (36%), and 36 women conceived naturally (4%), with similar proportions in each study group. The proportion of women with serious pregnancy complications was comparable across the three groups (TLE vs CON: RRadj 0.95, 95% CI 0.65–1.40 and TLR vs CON: RRadj 1.03, 95% CI 0.70–1.50; P = 0.89). Mean (SD) gestational age at birth was 39.4 (1.9) weeks, 39.5 (1.5) weeks, and 39.3 (1.9) weeks, respectively. We found no evidence of differences in preterm and very preterm births between groups. Mean (SD) weight at birth was 3413 (588) g, 3412 (588) g, and 3377 (578) g, respectively (TLE vs CON: MD 34, 95% CI −62 to 129 and TLR vs CON: MD 32, 95% CI −635 to 120; P = 0.70). We did not observe substantial differences in babies with low and very low birth weight. Health problems immediately after delivery were reported for eight babies in the TLE group, 12 in the TLR group, and 11 in the CON group. Major congenital malformations occurred in four children in the TLE group, four in the TLR group, and seven in the CON group. Minor congenital malformations occurred in five children in the TLE group, three in the TLR group, and five in the CON group.

This study reports safety outcomes for one type of time-lapse incubator, however, more systems are currently available.

Our results suggest that uninterrupted time-lapse culture with and without embryo selection based on machine learning can be regarded as safe compared to interrupted embryo culture and routine morphological selection in terms of obstetric and perinatal risks.

The authors received a grant from the Netherlands Organisation for Health Research and Development (ZonMw) for the execution of the SelecTIMO study (Health Care Efficiency Research programme grant 843001602). Merck (Germany and The Netherlands) supplied the six time-lapse incubators, funded the laboratory adjustments, and provided technical support and training to laboratory personnel before and during the study. D.C.K. received the Fertility Society of Australia exchange award. The following declarations of interest were made outside of the submitted work: F.B. reports additional financial support for the LUMO trial from Besins Healthcare Monaco, fellowship grants for ongoing basic research from Merck, consulting fees and payment or honoraria from Merck, Besins, and Ferring, and is member of the DSMB of the POISE study UK. J.M.J.S. has received grants or contracts from Ferring BV and Merck (payments to ETZ in both cases); consulting fees for an advisory board from Ferring BV; speakers fee from Merck BV; and support for conference attendance from Ferring BV, Merck, and Goodlife. M.v.W. is Senior Editor of Cochrane and Editor-in-Chief of Human Reproduction Update. C.B.L. reports a speakers honorarium from Organon (The Netherlands) and was Editor In Chief for Human Reproduction at the time of submitting this manuscript.

NTR5423: ICTRP Search Portal (who.int).