The article examines the objective, findings, and clinical context described in What EMERGE Revealed About Early Metformin Beyond Statistical Significance.
Approach:
Trial: EMERGE randomized 535 pregnant women to metformin or placebo and assessed a composite of insulin initiation or fasting glucose ≥5.1 mmol/L at 32 or 38 weeks’ gestation.
Reanalysis: A secondary generalized pairwise comparison analysis ranked the three endpoint components and used win ratios.
Bayesian analysis: Using uninformative priors, the authors estimated posterior probabilities of any benefit and of a reduction exceeding 5 percentage points.
Key Findings:
The prespecified primary analysis estimated a risk difference of −6.9% (95% CI, −15.1% to 1.4%; P=.13), and the trial authors concluded metformin was not superior to placebo for the primary composite.
The secondary analysis reported a win ratio of 1.46 (P=.004).
The primary-analysis confidence interval was compatible with effects ranging from substantial benefit to minimal benefit and indicated a low probability of harm on the composite outcome.
The Bayesian analysis estimated an approximately 95% probability of any benefit and an approximately 70% probability of a reduction greater than 5 percentage points.
Interpretation:
The article states that the analyses are more consistent with benefit than harm, while the size and clinical importance of any benefit remain uncertain. It characterizes a well-justified reanalysis as exploratory or suggestive and as a rationale for further research, not as a replacement for the prespecified primary analysis.
Limitations:
The primary result did not reach statistical significance, and its confidence interval includes no effect.
The reanalysis is secondary and should be interpreted as exploratory or suggestive.
The article states that greater certainty about the magnitude of benefit requires additional research.
Conclusion:
The authors argue that reanalysis may be warranted when primary findings and their uncertainty suggest a possible effect and the methods are structured, ethical, and scientifically sound; they caution that reanalysis should not be routine.
