Can dissenting findings regarding the comparative effectiveness of ICSI and IVF be explained by a learning curve?

Abstract

To determine whether between-trial heterogeneity in relative risk of fertilisation for intracytoplasmic sperm injection (ICSI) compared to in vitro fertilisation (IVF) can be explained by learning or by between-trial variation in patient characteristics. Systematic review and meta-analysis of trials comparing fertilisation outcomes for ICSI and IVF (without surgical sperm retrieval). Meta-regressions to identify associations between treatment effect and trial characteristics. Coefficients on individually significant covariates from the meta-regressions confirm that the ICSI versus IVF treatment effect is increased when patients are "unsuited for IVF" but reduced as semen quality improves and when IVF insemination concentrations are increased. However, the relative risk of fertilisation varies inversely with publication date; contrary to the hypothesised learning effect. While it is recognised that publication date might proxy for unobserved covariates, the possibility of a learning effect in favour of ICSI is not supported by the meta-regression.