Authors' reply to the comment from Benavides-Zora et al

Even if our analysis would violate the proportional hazards assumption, this would only lead to an underestimation of pooled effect size, further strengthening the robustness of our findings [2]. [...]when repeating the analysis using random-effects model and trim-and-fill approach, results remained consistent with the main analysis (Additional file 1: The ICU subgroup also had > 10% relative mortality increase (15% vs. 13%) with Bayesian approach indicating 75.7% probability of harm. Since the outcome is death, it is maybe cavalier to dismiss such probability as “no difference,” especially given a pediatric RCT suggested harm leading to a FDA warning [5] and the manufacturer’s promise for a second RCT which was never conducted.