Letter to the editor regarding the article “Development of a new predictive clinico-biological score for acute appendicitis in the pediatric population”

Abstract

The recently published pediatric appendicitis score demonstrates promising discriminative performance. However, key methodological aspects—including predictor handling, possible predictor redundancy or multicollinearity, absence of calibration analysis, internal validation, and post hoc cut-off selection—limit interpretability and clinical transportability. In this comment, these methodological aspects are critically and constructively examined to facilitate clearer interpretation and enhance methodological transparency.

Main Text

The recently published study by Youssef et al. [1], which describes the development of a clinico-biological score for pediatric acute appendicitis, was read with great interest. The effort to prospectively develop a diagnostic tool for pediatric emergencies is acknowledged. Nevertheless, several methodological aspects merit closer examination.

First, several clinical examination variables included in the model—such as tenderness, guarding, Rovsing’s sign, Blumberg’s sign, and the single-leg hop test—represent closely related manifestations of peritoneal irritation. In such contexts, the presence of collinearity may influence coefficient estimates and their interpretation. Providing information on how relationships between predictors were assessed (for example, using correlation matrices or variance inflation factors) would improve interpretability and allow readers to better understand the contribution of individual variables within the model.

Second, several continuous predictors, including neutrophil percentage, leukocyte count, CRP, and age, were dichotomized. Although this approach may increase bedside simplicity, information loss and distortion of risk estimation may result [2], particularly when cut-off values are not clearly pre-specified. Modeling predictors on their continuous scale, potentially incorporating non-linear effects, would preserve diagnostic information and could improve calibration.

Third, the strategy used for model construction was only partially described. Variables with p < 0.2 or deemed clinically relevant were included in the multivariable model; however, the exact variable-selection pathway, the handling of sparse predictors, and the reasons for retaining or excluding specific candidate variables were not fully reported. It should be acknowledged that several widely used appendicitis scores were originally derived using conventional logistic regression approaches without formal penalisation or internal validation, reflecting the methodological standards available at the time of their development. Furthermore, no form of penalization or shrinkage appears to have been applied [3]. As performance metrics were calculated on the same dataset used for model derivation, optimism bias cannot be excluded. These considerations are not intended to diminish the value of derivation studies, but rather to contextualise the results within current methodological standards and to support their appropriate interpretation.

Fourth, model performance assessment focused on discrimination and threshold-based classification metrics, as reflected by the reported AUC, sensitivity, specificity, predictive values, and likelihood ratios. Calibration, which is essential for evaluating agreement between predicted and observed risks, was not reported. A calibration slope, calibration intercept, or graphical calibration assessment would enable a more comprehensive evaluation of the model [4].

Finally, cut-off values and the proposed management strata were derived and evaluated within the same cohort. Post hoc threshold optimization may lead to overestimation of clinical performance [5]. In addition, as a clinical decision algorithm is proposed, formal evaluation of clinical utility—such as through decision curve analysis—would provide stronger evidence for practical implementation [6]. It should also be acknowledged that the original study appropriately recognises its own limitations and clearly states the need for further validation prior to clinical implementation.

Conclusions

Although promising discriminative performance was observed, further methodological clarification and validation are required before confident clinical adoption can be supported.

Abbreviations

AUC

Area under the curve

CRP

C-reactive protein

Author’s contributions

Javier Arredondo Montero: Conceptualization, Methodology, Formal analysis, Investigation, Writing – original draft, Writing – review & editing.

Funding

No funding was received.

Data availability

Not applicable.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The author declares no competing interests.