In surgically treated patients diagnosed with appendiceal epithelial malignancies, an observational treatment proxy designed to capture hyperthermic intraperitoneal chemotherapy delivered minimal incremental prognostic discrimination in internal validation and only modest gains at longer follow-up in an external cohort. The study evaluated whether incorporating an administrative proxy for the regional treatment into survival models could improve predictions of cancer-specific survival beyond standard clinicopathologic characteristics. While the proxy was associated with cancer-specific outcomes in statistical models, investigators observed that its inclusion did not produce statistically significant improvements in discriminative performance within the primary internal validation sample.
Appendiceal epithelial malignancies represent rare and heterogeneous gastrointestinal tumors with diverse clinical trajectories. In routine population-based cancer surveillance, evaluating the prognostic influence of specialized multimodal surgical interventions has been constrained by database recording practices. Specifically, confirmed administration of hyperthermic intraperitoneal chemotherapy is not directly recorded within the Surveillance, Epidemiology, and End Results program database. To address this structural data limitation, the research team constructed a derived proxy variable intended to identify patients who likely underwent the regional procedure, defining proxy positivity by the presence of recorded intraoperative systemic therapy alongside chemotherapy in the registry records.
The primary investigation drew upon a nationwide registry cohort comprising 5,520 adult patients diagnosed with appendiceal epithelial malignancies between 2007 and 2017 within the Surveillance, Epidemiology, and End Results database. To rigorously develop and test the predictive models, investigators partitioned this overall surgical cohort into two distinct subsets: a training cohort consisting of 3,863 patients and an internal validation cohort containing 1,657 patients. In addition to these registry-derived groups, the researchers gathered an independent external validation cohort of 209 institutional patients to evaluate the generalizability of the models across clinical settings without refitting the model parameters.
To assess the prognostic associations, the investigators conducted cause-specific Cox proportional hazards regression analyses on cancer-specific survival outcomes. They formulated a baseline statistical model founded exclusively on conventional clinicopathologic variables and compared its performance against an expanded model that incorporated the newly established treatment proxy. Prognostic performance was systematically characterized across three distinct follow-up intervals at one, three, and five years. The authors quantified model behavior through multiple complementary statistical dimensions, evaluating discrimination using the area under the receiver operating characteristic curve, tracking calibration curves, and calculating the Brier score alongside the integrated discrimination improvement metric.
Within the internal validation cohort of 1,657 patients, both statistical models demonstrated useful baseline prognostic discrimination, but the addition of the treatment proxy offered negligible predictive improvement. The expanded model incorporating the proxy achieved area under the curve values of 0.815 at one year, 0.810 at three years, and 0.782 at five years. By comparison, the baseline model relying solely on conventional clinicopathologic features demonstrated corresponding area under the curve values of 0.812 at one year, 0.808 at three years, and 0.780 at five years. Statistical comparisons of these paired receiver operating characteristic differences confirmed that the modest numerical increments provided by the proxy were not statistically significant in internal validation.
When evaluated externally in the 209 institutional patients without model refitting, the proxy model recorded area under the curve values of 0.775 at one year, 0.797 at three years, and 0.783 at five years, compared with baseline model values of 0.753 at one year, 0.759 at three years, and 0.751 at five years. Paired differences in the external cohort indicated modest prognostic improvements that achieved statistical support at the three-year and five-year marks, whereas the one-year estimate remained imprecise. External calibration was generally acceptable across evaluations, though the researchers noted some overprediction of survival probabilities at longer follow-up horizons. In supplementary sensitivity analyses designed to address proportional hazards violations, the derived proxy estimates remained stable and consistent.
Despite the observed statistical associations, the study authors explicitly outlined critical limitations regarding how the administrative proxy must be understood and applied. Because the proxy relies on secondary registry indicators rather than verified clinical documentation, it should not be interpreted as confirmed exposure to hyperthermic intraperitoneal chemotherapy or as evidence of direct therapeutic efficacy. Furthermore, the researchers emphasized that the proxy cannot serve as a reliable basis for individualized postoperative prognosis or clinical treatment selection in surgical oncology. Instead, the authors concluded that these findings provide rationale for further exploring proxy-based treatment patterns within specialized healthcare datasets that maintain more detailed, granular treatment characterizations.
Subject of Research: Prognostic evaluation of a registry-based hyperthermic intraperitoneal chemotherapy proxy in appendiceal epithelial malignancies
Article Title: Prognostic association of a SEER-derived HIPEC-related treatment proxy with cancer-specific survival in surgically treated appendiceal epithelial malignancies: a SEER-based study
Article References: Wang, Y., Li, K., Chen, J., Yang, C., Song, D., & Zheng, Y. (2026). Prognostic association of a SEER-derived HIPEC-related treatment proxy with cancer-specific survival in surgically treated appendiceal epithelial malignancies: a SEER-based study. BMC Cancer. https://doi.org/10.1186/s12885-026-17258-7
Image Credits: AI Generated
DOI: 10.1186/s12885-026-17258-7
Keywords: appendiceal epithelial malignancies, hyperthermic intraperitoneal chemotherapy, HIPEC proxy, SEER database, cancer-specific survival, prognostic modeling, surgical oncology, internal validation, external cohort evaluation, cause-specific Cox analysis, area under the curve, discrimination performance, model calibration, Brier score, integrated discrimination improvement, intraoperative systemic therapy, proportional hazards, clinicopathologic variables
