We read with great interest the article by Lee et al. 1, published in the Journal of Gastroenterology and Hepatology, which investigated the long-term surveillance outcomes of branch-duct intraductal papillary mucinous neoplasms (BD-IPMN). The authors reported a 10-year cumulative malignancy incidence of 5.5%—specifically 2.7% for IPMN-derived carcinoma and 2.9% for de novo pancreatic ductal adenocarcinoma (PDAC)—and identified cyst wall thickening as the sole independent risk factor for both entities. While this study provides valuable longitudinal data that challenges the conventional 5-year surveillance threshold, we believe several methodological and interpretative issues warrant further discussion to prevent potential misinterpretation of the findings and to refine future risk stratification models. First, the conclusion that cyst wall thickening is the only independent predictor, effectively dismissing the relevance of mural nodules, contradicts established international consensuses such as the Fukuoka and Kyoto guidelines 2, 3. In the univariate analysis, mural nodules showed a striking odds ratio (OR) of 43.95 (p < 0.001), yet this significance vanished in the multivariate model (p = 0.503). This discrepancy is likely attributable to statistical overfitting and multicollinearity among morphological features. The exceptionally wide 95% confidence interval observed for cyst wall thickening in the multivariate analysis (95% CI: 3.992–436.943) is a hallmark of model instability. Given the limited number of outcome events (n = 16), including multiple strongly correlated variables (e.g., mural nodules, cyst wall thickening, and main pancreatic duct dilation) into a regression model violates the “events per variable” (EPV) principle, which generally requires at least 10 events per predictor variable to ensure validity. Consequently, the statistical power was likely insufficient to disentangle the true independent effects, potentially masking the established malignant potential of mural nodules and overemphasizing cyst wall thickening. Second, the finding that localized cyst wall thickening predicts distant de novo PDAC raises questions regarding biological plausibility. The authors suggest “stromal fibrosis” as a potential shared mechanism; however, cyst wall thickening is inherently a local morphological change associated with specific cystic inflammation or neoplasia. Predicting a tumor in a remote pancreatic segment based on a local cystic feature implies a “field cancerization” hypothesis that requires rigorous validation. It is plausible that the observed association is driven by unadjusted confounders of a “high-risk pancreas,” such as the degree of background parenchymal fibrosis, chronic pancreatitis, or genetic susceptibility (e.g., germline mutations), which were not fully accounted for in the analysis. Without controlling these systemic factors, attributing predictive power solely to local cyst morphology may confuse a marker of co-existing pancreatic pathology with a direct causal predictor. Third, the diagnostic accuracy of the outcome variables relies heavily on radiological assessment without uniform pathological confirmation 4. Among the 16 patients suspected of malignant transformation, only 12 underwent surgical resection, leaving 25% (n = 4) diagnosed solely based on serial imaging. Distinguishing high-grade dysplasia or inflammation-associated wall thickening (e.g., focal pancreatitis) from invasive carcinoma via imaging alone is challenging and prone to misclassification bias. Furthermore, a potential “detection bias” cannot be ruled out: patients with visible cyst thickening may have undergone more intensive surveillance or high-resolution imaging (e.g., EUS) compared to those without, thereby increasing the likelihood of detecting small de novo cancers. If any of the unconfirmed cases were false positives, the calculated cumulative incidence would be overestimated, and the specificity of cyst wall thickening as a risk factor would be compromised. We commend the authors for their significant contribution to the literature on the natural history of BD-IPMN. However, considering the statistical instability demonstrated by wide confidence intervals, along with the biological and diagnostic uncertainties, we suggest that the prominence of cyst wall thickening as the “sole” predictor should be interpreted with caution. Future multi-center prospective studies with larger cohorts, strict pathological validation, and integration of molecular markers are essential to confirm these findings before altering current surveillance strategies. The author declares no conflicts of interest. Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
Hongyi Jiang (Fri,) studied this question.