When positive studies of novel therapies are subsequently nullified: cumulative meta-analyses in preeclampsia.

PURPOSE: The purpose of this study was to examine changes over time in the pooled effect size of randomized, double-blinded, placebo-controlled trials (RCTs) published on the protective effects of antioxidants and low dose aspirin against preeclampsia, and to identify determinants that may affect such changes.

METHODS

Two recently published meta-analyses of RCTs examining the effects of antioxidant treatment or low dose aspirin on the rates of preeclampsia and its adverse effects were used. Chronological, cumulative meta-analyses were conducted to investigate the possibility of a time-dependent effect. The journal's impact factor, citation numbers of each paper, and their sample size, were correlated with the risk ratio (RR) of the study.

RESULTS

The median sample size of positive antioxidant trials (i.e., showing protective effect) was tenfold smaller (median 267) than that of the negative trials (median 2120) (P = 0.017). A similar trend was seen for low dose aspirin studies. There was a significant correlation between study size and RR for the effects of antioxidants and low dose aspirin on intrauterine growth restriction (IUGR). There was no correlation between RR and citation number, or between RR and the journal's impact factor for the two therapeutic modalities. For both modalities, the journal's impact factor correlated significantly with the number of citations per year. Cumulative meta-analyses revealed that during the first few years and studies, there was a seeming significant protective effect of antioxidant or aspirin against preeclampsia. For both treatment, the initial protective effects gradually disappeared and nullified by larger, later studies.

CONCLUSIONS

Initial studies, often published in high impact factor journals, are cited significantly more times but do not exhibit a higher likelihood of predicting a correct long term answer. Studies with smaller sample sizes are more likely to be biased against the null hypothesis. As such, cumulative meta-analysis is an effective tool in predicting potential bias against the null hypothesis and the need for additional studies.