Our monthly Maternal-Fetal Medicine journal club recently discussed the recent article by Vigil-De Gracia et al with much interest. During our review, we noted that the sample size calculation was described in the following manner:

“The sample size was calculated, assuming a 15% rate of perinatal mortality (fetal death and neonatal death at ≤28 days) in the prompt delivery group, a 2-sided α error of 5%, and 80% power to detect a reduction of 30% in the rate of perinatal mortality with the expectant management. The high rate of perinatal mortality assumed in the study reflects the reality in Latin American countries. To detect that difference, 260 patients with 130 in each group needed to be recruited.”

Assuming that perinatal death was treated as a dichotomous outcome (either yes or no) and there was no intent to include time to event (survival time), a 2-sided test of proportions with PASS software results in 862 cases needed per group to obtain 80% power using the aforementioned 30% reduction from a baseline perinatal mortality rate of 15% (15% vs 0.7 × 15% = 10.5%). If 130 cases per group are used, the statistical power for this study (15% vs 10.5%) is only 0.23. Could the authors please clarify how they calculated their sample size?

Given that the authors report in favor of the null hypothesis (expectant management does not significantly affect perinatal death) and encourage prompt delivery for severe preeclampsia at <34 weeks’ gestation in Latin America in their discussion, we would like to fully understand the statistical methods used in this study.