Dear Editor, We read a recent study by Leem et al, titled “Comparison of second-line chemotherapy regimens in advanced biliary tract cancer: a systematic review, meta-analysis, and population-based cohort study” which integrated a single-center retrospective cohort with a systematic review and meta-analysis to evaluate FOLFIRINOX as a second-line treatment for advanced biliary tract cancer (BTC) and compared it with other regimens1. The authors’ work in a field lacking a standard second-line protocol is commendable, yet several limitations warrant discussion to contextualize the findings within the existing literature. This retrospective, single-center cohort with 54 patients was inherently susceptible to selection, information, and confounding biases. The choice of FOLFIRINOX was likely influenced by patient performance status, organ function, and physician preference, resulting in a nonrepresentative sample. Furthermore, excluding approximately 30% of patients from the efficacy analysis due to early clinical deterioration, while methodologically rigorous, may introduce survivorship bias and overestimate progression-free survival (PFS) and overall survival (OS) compared with an intent-to-treat analysis. This discrepancy underscores how patient selection in retrospective studies limits the generalizability of the reported outcomes (PFS 2.7 months, OS 8.9 months) to the broader second-line BTC population. Substantial heterogeneity complicates the meta-analysis of different regimens. Although FOLFIRINOX showed the highest pooled median PFS (4.60 months), its wide confidence interval (2.60–8.32) reflects limited data or high variability. The I2 statistic for the FOLFOX group was 93%, indicating extreme inconsistency among studies. The cohort’s efficacy outcomes were also lower than those of a prior phase II trial, a difference that the authors attributed to broader patient inclusion and more treatment modifications. This contrast highlights a critical point; the apparent advantage of FOLFIRINOX may be confined to a strictly selected patient population that can tolerate full-dose therapy, mirroring its use in pancreatic cancer2. Current evidence suggests that intensive chemotherapy does not benefit all patients undergoing second-line treatment, leading to the identification of an optimal patient population for the FOLFIRINOX regimen. However, the article did not define specific patient characteristics that warrant intensive chemotherapy. Consequently, this study primarily supplements real-world efficacy data for FOLFIRINOX and highlights its conditional effectiveness rather than challenging established treatment paradigms. Robust quantitative synthesis of severe toxicities across regimens was not possible because of the inconsistent reporting in the included studies. Consequently, the conclusion of “manageable toxicity” of FOLFIRINOX relies heavily on the low discontinuation rate (3.7%) observed in the authors’ cohort, a rate not replicated in other studies3. The study did not specify whether patients underwent genetic testing to determine the possibility of targeted therapy4,5. Given the molecular heterogeneity of BTC, proposing FOLFIRINOX as a preferred option without considering actionable mutations may deny patients access to more effective, less toxic targeted therapies. For instance, patients with IDH1 mutations achieved a PFS of 6 months with ivosidenib in the ClarIDHy trial, suggesting that molecular profiling should precede the selection of intensive chemotherapy4. The lack of this perspective represents a gap in the current shift toward individualized treatment. The primary contribution of this study is the systematic positioning of FOLFIRINOX within the second-line BTC treatment landscape. This finding should be interpreted as a compelling rationale for prospective phase III trials that directly compare FOLFIRINOX with standard regimens. Future trials must incorporate biomarker exploration to predict efficacy and toxicity, moving toward truly optimized and individualized second-line therapy for advanced BTC.
Liu et al. (2026) studied this question.