A commentary on “Assessing kidney outcomes in childhood-onset lupus nephritis: role of National Institutes of Health-modified histological indices”

Article information

Clin Exp Pediatr. 2026;69(4):362-363

Publication date (electronic) : 2026 March 6

doi : https://doi.org/10.3345/cep.2025.02712

Yanlan Ma, MBBS¹, Yunyun Zhu, MD^,2

¹Department of Nephrology, Pan'an Hospital of Traditional Chinese Medicine, Jinhua, China

²Department of Nephrology, Tongde Hospital of Zhejiang Province Affiliated to Zhejiang Chinese Medical University (Tongde hospital of Zhejiang Province), Hangzhou, China

Corresponding author: Yunyun Zhu. Department of Nephrology, Tongde Hospital of Zhejiang Province Affiliated to Zhejiang Chinese Medical University (Tongde hospital of Zhejiang Province), Hangzhou, Zhejiang, China Email: zhuyunyun-520@163.com

Received 2025 November 12; Revised 2025 December 24; Accepted 2025 December 26.

To the editor,

We read with great interest the study by Penboon and Rianthavorn [1], which assessed the utility of the National Institutes of Health-modified activity and chronicity indices for predicting kidney outcomes and identified histopathological features associated with kidney function impairment (KFI) in childhood-onset lupus nephritis (cLN). The authors' finding that tubular atrophy is an independent predictor of KFI is clinically relevant. However, we would like to raise several methodological concerns regarding the statistical analysis.

First, widely accepted guidelines for multivariable analysis recommend approximately 10 outcome events per predictor variable to ensure model stability [2-5]. Although this study explicitly framed the Cox model as exploratory/etiologic, not as a fully developed prediction tool, the author already highlighted as limitations in the original paper. The final multivariate model in this study (Table 4) included 6 variables, which would require approximately 60 KFI events. However, the analysis was based on only 18 events. This low event-to-variable ratio can lead to severe model overfitting, potentially compromising the reliability of the reported estimates. Therefore, we caution that these specific findings require validation in a larger, ideally multicenter, cohort. Second, established methodological best practices suggest that multivariate models should primarily include variables that show univariate association and/or are prespecified on clinical grounds [4]. The inclusion of nonsignificant variables, as seen in Table 4, can introduce noise and lead to inaccurate estimates of the true associations.

Additionally, we noted that the chronicity index (CI) was significantly different between the KFI and non-KFI groups in Table 2. It is therefore surprising that this variable was not included in the final multivariate model for predicting KFI. Given that the CI is a composite score encompassing chronic pathological changes, including tubular atrophy, it is plausible that the CI could be a stronger, or at least a contributory, predictor of KFI than tubular atrophy alone. Its omission may affect the interpretation of the final model. Furthermore, we suggest the authors explicitly recognize that including both the CI and its components (e.g., tubular atrophy) in the same model may induce collinearity, and that the choice between the composite score and individual components partly reflects an a priori analytic strategy. Clarifying this nuance would strengthen their argument without overstating the impact on the study’s conclusions.

Despite these methodological considerations, we commend the authors for their valuable contribution to the challenging field of cLN outcomes.