Methodological Concerns in “Prognostic Usefulness of Clinical Complete Response after PD-1 Inhibitor-Based Combination Therapy for Unresectable Hepatocellular Carcinoma”

Dear Editor,

We read with great interest the study by Ma et al. [1] (2025) investigating the prognostic value of clinical complete response (cCR) following PD-1 inhibitor-based combination therapy in unresectable hepatocellular carcinoma. While the authors present compelling findings on survival benefits associated with cCR, we wish to highlight methodological concerns that may affect the interpretation of the results.

First, the study combines PD-1 inhibitors with interventional therapies (e.g., transarterial chemoembolization) and antiangiogenic agents, yet it does not fully address potential confounding by concomitant medications. For instance, patients with hepatitis B/C received routine antiviral therapy, which may influence immune responses and outcomes such as cCR or overall survival. Ignoring these potential drug effects may introduce bias. We suggest that future analyses more explicitly account for concomitant medications, including antiviral treatment, as potential sources of residual confounding.

Second, regarding the multivariable analyses reported in Tables 2–4 of the original article, the inclusion of numerous predictors without clear justification warrants cautious interpretation. Although the traditional 10 events per variable rule has often been cited as a pragmatic rule of thumb, it should not be regarded as a strict threshold for model adequacy. Contemporary methodological research has shown that the required sample size for multivariable modelling depends on multiple factors, including expected model performance, total sample size, event rate, degree of shrinkage, model complexity, and the intended use of the model. In this context, our concern is not a rigid breach of an EPV cutoff per se, but the potential risk of overfitting, coefficient instability, and optimistic estimates when the effective amount of information may be limited relative to model complexity. More robust approaches, such as formal sample size considerations for prediction models, penalised regression, shrinkage methods, and internal validation using bootstrapping, would provide a stronger basis for assessing model stability [2–4].

Third, laboratory parameters such as α-fetoprotein may be influenced by treatment exposure if they are not consistently defined at baseline. We suggest that variables intended for prognostic modelling should ideally be measured at a clearly specified pretreatment time point to minimise temporal ambiguity.

Despite these concerns, we commend the authors for their large-scale, multicentre effort. Addressing these issues in future studies would help strengthen the methodological robustness and prognostic interpretation of cCR in hepatocellular carcinoma.

Statement of Ethics

Ethics approval and informed consent were not required for this correspondence.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This study was not supported by any sponsor or funder.

Author Contributions

S.S. and J.C. drafted the letter. B.Y. revised the manuscript and supervised the work. All authors reviewed and approved the final version of the manuscript and agreed to its submission.

Funding Statement

This study was not supported by any sponsor or funder.