Amy Hammers et al recently published a critical metaanalysis of indomethacin as a tocolytic that demonstrated evidence of harmful effects to preterm infants. The investigators have used rigorous methods for metaanalysis of observational studies. They acknowledge that the selection biases are difficult to control using metaregression analysis because many important confounders are not identified. The investigators did not study randomized clinical trials (RCT) because there are only 3 reported trials with a rather limited number of patients. Bronchopulmonary dysplasia (BPD) was not a significant risk factor in the study of Hammers et al, whereas indomethacin increased the risk of BPD in metaanalysis of RCTs.

The discordance in the results between cohort studies and RCT effects may be due to a selection bias. In the metaanalyses published on the topic, there have been fewer cohorts listing BPD as an outcome, compared with neonatal outcomes, such as intraventricular hemorrhage. Of the 7 cohort studies used by Hammers et al to estimate the risk of BPD, 3 listed preterm pregnancies with or without requirement of tocolysis as control subjects. As a result, the exposed pregnancies may have under-representation of intrauterine growth restriction and preeclampsia and over-representation of spontaneous birth cases compared with control subjects; some of the studies actually showed that these differences between the groups were statistically significant. This is likely to result in an overestimate of the risk of intraventricular hemorrhage and an underestimate of the risk of BPD in the indomethacin group. We reanalyzed the risk of BPD using the references listed by Hammer et al and included only the 4 studies that compared 2 populations that required tocolytic. We found a significant increase in the risk of BPD (odds ratio, 1.7; 95% confidence interval, 1.2–2.7). This estimate is consistent with the result of the randomized trial.

The harmful consequences of indomethacin tocolytic treatment shortly before the birth of a preterm infant may overshadow the benefits of prolonging the short duration of pregnancy. For instance a follow up of a randomized trial shows evidence of adverse long-term effect in small toddlers, despite a significant increase in the duration of pregnancy in the indomethacin group. Many investigators, including those who are involved in formulating consensus guidelines, prefer to evaluate RCTs rather than retrospective observational studies.