A Simple Method to Establish Sufficiency and Stability in Meta‐Analyses: With Application to Fine Particulate Matter Air Pollution and Preterm Birth

Abstract

Fine particulate matter air pollution (PM2.5) is a potential cause of preterm birth. Incon-sistent findings from observational studies have motivated researchers to conduct more studies, but some degree of study heterogeneity is inevitable. The consequence of this feedback is a burgeoning research effort that results in marginal gains. The aim of this study was to develop and apply a method to establish the sufficiency and stability of estimates of associations as they have been published over time. Cohort studies identified in a recent systematic review and meta‐analysis on the association between preterm birth and whole‐pregnancy exposure to PM2.5 were selected. The estimates of the cohort studies were pooled with cumulative meta‐analysis, whereby a new meta‐anal-ysis was run for each new study published over time. The relative risks (RR) and 95% confidence interval (CI) limits needed for a new study to move the cumulative RR to 1.00 were calculated. Findings indicate that the cumulative relative risks (cRR) for PM2.5 (cRR 1.07, 95% CI 1.03, 1.12) con-verged in 2015 (RR 1.07, 95% CI 1.01, 1.14). To change conclusions to a null association, a new study would need to observe a protective RR of 0.93 (95% CI limit 1.02) with precision equivalent to that achieved by all past 24 cohort studies combined. Preterm birth is associated with elevated PM2.5, and it is highly unlikely that any new observational study will alter this conclusion. Consequently, establishing whether an observational association exists is now less relevant an objective for future studies than characterising risk (magnitude, impact, pathways, populations and potential bias) and interventions. Sufficiency and stability can be effectively applied in meta‐analyses and have the potential to reduce research waste.