Comment on “Survival benefit of radiotherapy in pancreatic cancer: a real-world analysis based on SEER and an independent hospital-based cohort” Authors: Kangkang Ji, Ning Zou, Jing Tang Published: 2026 Dear Editor, The recent investigation by Ou et al provides valuable insights into radiotherapy’s role in pancreatic cancer management through SEER data analysis with institutional validation[1]. Although the application of propensity score matching (PSM) strengthens causal inference, three substantive methodological limitations warrant critical examination to contextualize the divergent findings between the SEER cohort and the hospital cohort and to guide future research. All AI-related statements comply with the TITAN guideline[2]. Methodological aggregation in radiotherapy classification The binary classification of radiotherapy (“received” versus “not received”) fails to capture critical technical heterogeneity in treatment delivery. The SEER database aggregates conventional radiotherapy, stereotactic body radiotherapy (SBRT), brachytherapy, and radioisotope-based therapy within a single variable, despite established evidence of different efficacy profiles across these radiotherapy modalities. This aggregation may inflate the reported survival benefit (HR = 0.69) by masking the suboptimal outcomes associated with outdated techniques. Furthermore, SEER’s absence of dosimetric parameters, fractionation details, and treatment intent data substantially limits the clinical applicability of these findings for contemporary radiation oncology decision-making. Validation cohort design limitations The non-significant result in the hospital cohort (HR = 0.78, P = 0.35), which is attributed by the authors to insufficient power, in fact reflects more fundamental methodological constraints. Critical design discrepancies exist between the SEER cohort (2019–2021) and the hospital cohort (2018–2025). The SEER cohort included 51.9% stage IV patients versus only 25.8% metastatic cases in the hospital cohort, indicating significant selection bias. This stage imbalance fundamentally compromises cohort comparability since metastatic disease inherently constrains radiotherapy’s therapeutic window. Temporal discordance introduces additional era bias. Specifically, the hospital cohort extending into 2025 captures advances in systemic therapy, such as increased use of the FOLFIRINOX chemotherapy regimen, which may diminish radiotherapy’s relative contribution. Crucially, immortal time bias remains unaddressed; SEER lacks radiotherapy initiation timelines, meaning patients surviving beyond this threshold disproportionately receive treatment, thereby artificially inflating survival estimates. The hospital cohort’s limited sample size (n = 132, post-PSM) and retrospective nature do not fully resolve these registry-specific biases. Unmeasured confounders in subgroup analyses Despite statistically significant interactions, the subgroup analyses remain vulnerable to residual confounding from unmeasured variables. SEER lacks critical data on resection margins (R0/R1), performance status (ECOG), and molecular biomarkers. Consequently, the reported benefit in stage III–IV adenocarcinoma may reflect biological selection rather than genuine radiotherapy efficacy. Reliance on ICD-O-3 histology codes (8140/3 vs 8500/3) disregards molecular subtyping (e.g., basal-like vs classical), which modulates treatment response – a clinically consequential omission. This limitation is exemplified by the LAP07 randomized trial, which demonstrated no significant survival benefit of chemoradiotherapy over chemotherapy alone in locally advanced pancreatic cancer despite rigorous prospective design, underscoring the impact of molecular and clinical confounders absent in SEER analyses[3]. Randomized evidence further confirms differential survival benefits in molecularly defined subgroups, such as BRCA/PALB2-mutated patients receiving platinum-based therapies[4]. Conclusion and discussion Although Ou et al underscore radiotherapy’s potential in advanced pancreatic cancer, this study highlights inherent limitations of real-world evidence. SEER’s inability to differentiate radiotherapy modalities, coupled with unmeasured confounders and validation cohort mismatches, precludes definitive clinical conclusions. Future studies must integrate radiotherapy-specific registries with comprehensive molecular profiling to identify true beneficiaries. This approach is exemplified by adaptive trials such as the phase 1b/2 study employing standardized ablative SBRT (50–55 Gy in 5 fractions) combined with molecular-targeted agents, which directly addresses technical heterogeneity and molecular confounders in locally advanced disease[5]. Genomic profiling studies further indicate that classifiers identifying homologous recombination deficiency can identify an additional 7–10% of PDAC patients beyond germline BRCA/PALB2 carriers who may benefit from DNA-damaging agents[6]. Until such integrated data become available, this study’s subgroup analyses should not supersede prospective evidence supporting modern precision radiotherapy in molecularly stratified populations. These findings affirm the clinical relevance of radiotherapy acknowledged in NCCN Guidelines, which endorse its role in multimodal therapy for locally advanced disease when technical heterogeneity is minimized through standardized protocols like SBRT[7].