Clinical outcomes and the impact of treatment modalities in children with carbapenem-resistant Enterobacteriaceae bloodstream infections: a retrospective cohort study from a Tertiary University Hospital—right to reply—author’s response
Makaleler > Tam Makale · 2025
Yazarlar: Gülhadiye AVCU
Journal of Antimicrobial Chemotherapy, Uluslararası, Wos, 2025, cilt 80, ss 2332-2333
DOI: 10.1093/jac/dkaf201
Özet
We read carefully the statistical concern regarding the potential overfitting in our multivariable logistic regression analysis by Qinmei Huang et al. Their concern is based on the widely recognized ‘10 events per variable’ (EPV) rule, a heuristic commonly used to minimize overfitting and ensure stability in prediction modelling. Although the ‘10 EPV’ rule was originally proposed to prevent overfitting, subsequent studies have demonstrated that it may be overly conservative. The literature has shown that valid and reliable results can be achieved with fewer events per variable, particularly when the model is clinically justified and parsimonious.1–3 In our original manuscript, we openly specified the limitations posed by our sample size and the retrospective nature of the study. Variables included in the multivariable model were selected based on both statistical significance in univariate analysis and clinical relevance, in line with previous literature on paediatric and adult CRE-BSI cases. While we agree that the limited number of deaths (n = 16) restricts the generalizability of our findings, we believe our study offers meaningful insights—particularly from a low-resource paediatric setting where data on CRE-BSI management remain limited.