Sample size, model stability, and the interpretation of adjusted effect estimates: correspondence on Moreira et al

We comment on the recent retrospective cohort by Moreira and colleagues evaluating intracranial-compliance-guided mean arterial pressure management. The adjusted multivariable logistic regression in the published analysis includes ten estimated coefficients against 22 ICU deaths, producing an adjusted odds ratio reported with a confidence interval spanning more than two orders of magnitude. We outline why these features are consistent with model instability rather than a precise causal estimate, identify a corroborating signal within the same model, and raise two secondary questions about control selection and the headline cost figure. We support the authors' call for prospective evaluation while suggesting that the inferential claims of the present analysis warrant more cautious presentation.