Instrumental variable meta-analysis of randomised trials of epidural analgesia in labour to adjust for non-compliance

Abstract

Objective: Intention-to-treat analysis of randomised controlled trials may cause bias towards the null where non-compliance with the allocated intervention occurs. Instrumental variable analysis allows estimation of the causal effect adjusted for non-compliance. The aim of this study is to compare intention-to-treat and instrumental variable meta-analysis of the association between epidural analgesia in labour and caesarean section. Study design and Setting: The study was restricted to 27 trials in a recent Cochrane Systematic Review. For trials with data on compliance, the association between epidural analgesia in labour and caesarean section was calculated using intention-to-treat analysis and instrumental variable analysis. Fixed-effects meta-analysis was used to calculate pooled risk ratios. Results: In 18 trials with data on compliance, 23% of women allocated to epidural analgesia did not comply and 27% of women allocated to the control received epidural analgesia. Data on outcomes in non-compliant groups were available for 10 trials. The pooled risk ratio for caesarean section following epidural analgesia in labour was 1.37 (95% CI 1.00-1.89, p=0.049) using instrumental variable analysis compared to 1.19 (95% CI 0.93-1.51, p=0.16) using intention-to-treat analysis. Conclusion: Intention-to-treat meta-analysis underestimates the effect of receiving epidural analgesia in labour on caesarean section compared to instrumental variable meta-analysis.