Effect of labor epidural analgesia on the probability of Cesarean section: an updated analysis
Abstract Number: F1B-4
Abstract Type: Original Research
Numerous study designs, including randomized controlled trials (RCT’s), before-and-after studies, and observational studies with concurrent controls (analyzed by propensity scores), have asked whether labor epidural analgesia (LEA) influences the probability of Cesarean section (C/S). One limitation of the RCT’s is the high rate of crossovers in many of these studies. In contrast to our previous analysis, a recent RCT meta-analysis using instrumental variables to adjust for crossovers  concluded that LEA increased the probability of C/S. To further investigate this topic, we updated a previous analysis based on the paired availability design (PAD), a meta-analysis of before-and-after studies adjusted for different availabilities of treatment . The revised PAD (with 1 additional study and modified data extractions) yielded similar results as before, namely no effect of epidural analgesia on the probability of C/S (Figure 1). We also updated our previous meta-analysis of randomized trials which used instrumental variables to adjust for crossovers . For the 8 studies in which we obtained additional outcome data for testing the assumptions needed to adjust for crossovers, we found the assumptions did not hold, likely because crossovers did not occur immediately after randomization. Because crossovers generally occur well after randomization, results from randomized trials (which showed no effect of epidural analgesia on the probability of C/S) are inconclusive. A recent study using propensity scores  showed an effect of epidural analgesia on the probability of Caesarean section. However, we question the interpretation of these results because, as with previous studies using a propensity score analysis , pain levels were not included as a covariate, which could lead to substantial bias. We conclude the PAD is a reasonable way to evaluate the influence of LEA on C/S, given the limitations of RCT’s and propensity score analyses, and that there is no compelling evidence that LEA increases the probability of C/S. We recommend more before-and-after studies, particularly if there is a large change in the fraction receiving LEA.
1. J Clin Epidemiol. 2015; 68(5):525-33.
2. Matern Child Health J. 2010 Sep;14(5):705-712.
3. J Causal Inference 2013; 1:51-82