Abstract
Background/Aim: Completion of adjuvant chemotherapy (AC) is a key determinant of long-term survival after curative-intent resection for pancreatic cancer. However, predictors of AC completion remain unclear. This study aimed to clarify the prognostic significance of AC non-completion and identify host-related factors after pancreaticoduodenectomy (PD).
Patients and Methods: We retrospectively analyzed 170 patients who were scheduled to receive AC after PD for pancreatic cancer. Associations between AC completion status and survival outcomes were evaluated. Independent predictors of AC non-completion were identified using logistic regression, and model performance was assessed through receiver operating characteristic analysis, calibration plots, and decision curve analysis (DCA).
Results: Of the 170 patients, 116 (68.2%) completed standard AC (complete group), whereas 54 (31.8%) discontinued treatment (incomplete group). The 5-year survival rate was significantly lower in the incomplete group than in the complete group (16.7% vs. 45.3%, p<0.001). Low serum total cholesterol (T-chol) independently predicted poor AC tolerance. The T-chol model demonstrated moderate discrimination (area under the curve: 0.697) and good calibration. DCA indicated a favorable net clinical benefit for T-chol within a threshold probability range of 0.2-0.4.
Conclusion: Serum T-chol is a strong predictor of AC completion and may help identify patients with pancreatic cancer who could benefit from targeted nutritional or supportive interventions during the perioperative period.
- Pancreatic cancer
- pancreaticoduodenectomy
- adjuvant chemotherapy tolerance
- nutritional biomarker
- serum total cholesterol
Introduction
Pancreatic cancer is one of the most lethal malignancies worldwide, largely because of its intrinsic therapeutic resistance and strong metastatic potential, despite decades of intensive research and numerous therapeutic advances (1). In recent years, systemic chemotherapy – particularly adjuvant chemotherapy (AC) – has become a central component of multidisciplinary treatment for pancreatic cancer (2). AC, including S-1–based regimens, is well established as the standard postoperative treatment for resected pancreatic cancer, and its therapeutic efficacy depends strongly on treatment completion (3). However, approximately 20-40% of patients fail to complete the planned course because of treatment-related toxicities or postoperative complications (4-7). Several studies have reported the clinical relevance of nutritional and inflammatory factors that contribute to AC non-completion (6, 8, 9); however, robust preoperative predictors of treatment adherence remain unclear.
Another important reason for AC non-completion lies in the surgical procedure itself. Pancreaticoduodenectomy (PD), the standard operation for pancreatic head cancer, causes major alterations in gastrointestinal continuity and exocrine function, often leading to impaired digestion, nutrient malabsorption, and delayed recovery of oral intake (10, 11). These surgery-related nutritional challenges further reduce tolerance to postoperative systemic chemotherapy.
Therefore, identifying reliable preoperative factors that predict AC non-completion is crucial for optimizing perioperative management and improving outcomes. In this study, we aimed to clarify the impact of AC completion on long-term survival after PD for pancreatic cancer and to identify key host-related nutritional and metabolic predictors – particularly serum total cholesterol (T-chol), a nutritional marker recently highlighted for its associations with both long-term outcomes after pancreatic cancer surgery and short-term risk of pancreatic cancer (12, 13) – that are linked to AC non-completion. Our goal was to establish a simple and practical framework for risk stratification and supportive treatment.
Patients and Methods
This study was approved by the Local Ethics Committee for Clinical Investigation at Osaka Metropolitan University (approval number: 2020-241) and was conducted in accordance with the principles of the Declaration of Helsinki.
Patients. Between 2007 and 2021, 263 patients with localized pancreatic head cancer underwent PD at Osaka Metropolitan University. All protocols were conducted in accordance with approved institutional procedures, and written informed consent was obtained from all patients. Of these 263 patients, 259 with invasive pancreatic cancer were included in this study, whereas four patients with non-invasive pancreatic cancer were excluded (Figure 1). Patients were categorized into two groups: the AC group, which received the standard AC regimen, and the no-AC group, which did not receive the standard AC because of early recurrence or a pathological diagnosis of stage IV disease.
Flow diagram of the patients with pancreatic cancer. Tumor-node-metastasis staging was determined according to the 8th edition of the American Joint Committee on Cancer TNM Staging System (14). PD: Pancreaticoduodenectomy; AC: adjuvant chemotherapy.
Definition of completion status of standard AC. The standard AC protocol was defined as either gemcitabine monotherapy (1,000 mg/m2 on days 1, 8, and 15 every four weeks for six cycles) or S-1 therapy (80-120 mg/day orally for four weeks followed by two weeks of rest, repeated for six cycles), according to national guidelines (2, 3) during the study period. As there is no universally accepted definition of AC completion, we defined completion status as administration of AC for a total duration of six months after surgery, based on previous reports (6, 9, 15).
Preoperative management and follow-up. The pretreatment protocol varied according to resectability status based on the National Comprehensive Cancer Network Pancreatic Cancer guidelines, version 2.2023 (16). Patients with resectable pancreatic head cancer received neoadjuvant chemotherapy (NAC) with gemcitabine plus S-1 (GS) followed by surgical resection. Patients with borderline resectable pancreatic head cancer received NAC with gemcitabine plus nanoparticle albumin-bound paclitaxel (GnP) or FOLFIRINOX (fluorouracil, leucovorin, irinotecan, and oxaliplatin), with or without concurrent chemoradiation. For unresectable-locally advanced pancreatic cancer, we considered PD following chemotherapy with GnP combined with radiation therapy when the tumor abutted the superior mesenteric artery (SMA) (17), or PD combined with celiac artery resection for pancreatic neck cancer involving both the celiac and gastroduodenal arteries (without SMA involvement) after chemotherapy with GS and radiation therapy (18).
Assessment of skeletal muscle volume and nutritional-inflammatory markers. Skeletal muscle area at the third lumbar vertebral level was assessed on preoperative computed tomography scans using SYNAPSE VINCENT software (Fujifilm, Tokyo, Japan). To standardize muscle mass for body size, both the skeletal muscle index and psoas muscle index (PMI) were calculated by dividing each muscle area (cm2) by the square of the patient’s height (m2) (9, 19, 20). We evaluated preoperative nutritional and inflammatory status using several validated indices, including nutritional indicators such as Onodera’s prognostic nutritional index (PNI), the geriatric nutritional risk index, and controlling nutritional status (8, 21-23). We also assessed inflammatory or combined inflammatory-nutritional markers such as the modified Glasgow prognostic score (mGPS), neutrophil–lymphocyte ratio, platelet–lymphocyte ratio, and C-reactive protein/albumin ratio (24-26).
Evaluation of predictive performance. We evaluated the predictive performance of preoperative nutritional indicators using receiver operating characteristic (ROC) analysis, calibration plots, and decision curve analysis (DCA). For DCA, we constructed three logistic regression models incorporating age and each nutritional parameter (T-chol, albumin, and PNI) to estimate the net clinical benefit of predicting AC non-completion. We plotted the net clinical benefit across threshold probabilities following the method of Vickers and Elkin (Med Decis Making 2006), using the “rmda” package (version 1.6) in R (version 4.3.1; R Foundation for Statistical Computing, Vienna, Austria). To improve visual interpretability, DCA curves were smoothed using a locally estimated scatterplot smoothing method.
Statistical analysis. Group comparisons for continuous variables were conducted using the Mann–Whitney U-test or the Wilcoxon signed-rank test. Correlations between AC completion status and clinicopathological factors were evaluated using the Chi-squared test. ROC analysis was used to determine optimal cutoff values for continuous variables, including serum albumin and serum T-chol levels. Survival rates were estimated using the Kaplan–Meier method and compared using the log-rank test. Univariate and multivariate analyses were performed to estimate hazard ratios (HRs) and odds ratios (ORs) using the Cox proportional hazards model and logistic regression, respectively. Statistical significance was set at p<0.05. All statistical analyses were performed using GraphPad Prism 10, SPSS (version 29.0; IBM Corp., Armonk, NY, USA), and R (version 4.3.1; R Foundation for Statistical Computing).
Results
Background and long-term outcomes. Of the 259 patients, 170 were scheduled to receive standard AC after PD for pancreatic cancer (AC group), whereas 89 patients did not receive standard AC (no-AC group). The 5-year survival rate was significantly higher in the AC group than in the no-AC group (37.0% vs. 10.8%, p<0.001) (Figure 2A).
Survival outcomes according to adjuvant chemotherapy status. Kaplan–Meier survival curves illustrate overall survival based on AC status. (A) The MST was 15.4 months in the no-AC group and 41.4 months in the AC group (p<0.001). (B) Within the AC group, the MST was 17.2 months in the incomplete group and 55.9 months in the complete group (p<0.001). AC: Adjuvant chemotherapy; MST: median survival time.
Among the 170 patients in the AC group, 116 completed the standard AC protocol over six months (complete group), whereas 54 did not complete the planned regimen (incomplete group) (Figure 1). In this study, we focused on these 170 patients to assess the clinical impact of AC completion.
Table I summarizes the baseline patient demographics in the incomplete and complete groups. Patients in the incomplete group were significantly older than those in the complete group (75 vs. 69 years, p<0.001). No significant difference was observed in pathological margin-positive resection rates between the incomplete and complete groups (20.4% vs. 9.6%, p=0.052). S-1 was the initial adjuvant regimen in approximately 90% of patients in both groups. We compared preoperative nutritional and inflammatory parameters between the incomplete and complete groups (Table II). Patients in the incomplete group had poorer nutritional status, including lower PMI (p=0.046), serum albumin levels (p=0.048), and T-chol levels (p<0.001), as well as a higher incidence of mGPS ≥1 (p=0.041) than those in the complete group. The reasons for AC non-completion, summarized in Supplementary Table I, were predominantly treatment-related toxicity or inability to initiate therapy due to infection or poor performance status.
Baseline characteristics.
Patient characteristics of skeletal muscle volume and nutritional-inflammatory markers.
Kaplan–Meier curves of postoperative survival by AC status showed that the 5-year survival rate was significantly lower in the incomplete group than in the complete group (16.7% vs. 45.3%; p<0.001) (Figure 2B). In multivariate analysis of postoperative survival among patients who underwent resection, AC non-completion was an independent prognostic factor for poorer survival [hazard ratio (HR)=3.088, p<0.001] (Table III).
Multivariate analysis of risk factors for overall survival.
Supplementary Table II summarizes the location of initial recurrence (local vs. distant). No significant difference in initial recurrence site was observed between the two groups; however, Supplementary Figure 1 shows that recurrence-free survival (RFS) was significantly shorter in the incomplete group than that in the complete group (median RFS: 10.3 vs. 24.2 months; p<0.001).
Predictive factors for non-completion of AC. To identify significant factors associated with AC non-completion, univariate and multivariate logistic regression analyses were performed (Table IV). In univariate analysis, advanced age, borderline resectable or unresectable-locally advanced disease, low PMI, low T-chol, and high mGPS were significantly associated with AC non-completion. Multivariate analysis revealed that advanced age [odds ratio (OR)=4.986, p<0.001] and low T-chol level (OR=3.103, p=0.007) were independent predictors of AC non-completion.
Univariate and multivariate analysis of preoperative predictors associated with incomplete adjuvant chemotherapy.
Given that T-chol was the strongest independent predictor, we compared its predictive performance with other nutritional indicators. Figure 3A shows ROC curves for predicting AC non-completion using T-chol, serum albumin, and PNI. The areas under the curve for T-chol, albumin, and PNI were 0.697, 0.594, and 0.590, respectively, indicating that T-chol had the highest predictive accuracy. The calibration curve of the T-chol–based model showed close agreement between predicted and observed probabilities of AC non-completion (Figure 3B). DCA demonstrated that the multivariable model incorporating age and T-chol provided a greater net clinical benefit than models based on albumin or PNI, particularly within the threshold probability range of 0.2-0.4 (Figure 3C, Supplementary Table III).
Predictive performance of the total cholesterol-based model for adjuvant chemotherapy non-completion. (A) Receiver operating characteristic curves of T-chol, albumin, and PNI for predicting AC non-completion. (B) Calibration plot of the T-chol–based model for predicting AC non-completion. The gray shaded area indicates the 95% confidence interval of the calibration curve. (C) Decision curve analysis of multivariable models incorporating age and each nutritional parameter (T-chol, albumin, or PNI). T-chol: Total cholesterol; PNI: prognostic nutritional index; AC: adjuvant chemotherapy; AUC: area under the curve; FPR: false-positive rate.
Discussion
Completion of AC emerged as a strong and independent determinant of long-term outcomes after PD for pancreatic cancer, with overall survival significantly higher among completers than non-completers, consistent with previous studies (5, 6, 15, 29). We found that advanced age and low T-chol – both patient-related factors – independently predicted AC non-completion, whereas tumor-related variables, such as resectability category and tumor markers, were not associated. Furthermore, receipt of NAC did not affect AC completion. These findings suggest that host status at the time of AC initiation is the primary determinant of tolerance to postoperative systemic therapy.
Mechanistically, low T-chol is unlikely to directly cause AC non-completion; rather, it likely serves as a surrogate marker of nutritional reserve. Patients with lower T-chol may have reduced metabolic capacity and greater vulnerability to treatment-related toxicities or postoperative complications, which were the leading causes of AC non-completion in our study (Supplementary Table I). Notably, T-chol reflects intrinsic aspects of lipid metabolism and, therefore, provides a more stable predictive measure, whereas serum albumin and PNI are strongly influenced by inflammation and treatment-related fluctuations (Supplementary Figure 2). This stability likely explains why the T-chol–based model, but not albumin or PNI, demonstrated favorable net clinical benefit within the threshold range of 0.2-0.4 (Figure 3C, Supplementary Table III). Reported AC non-completion rates typically range from 20% to 40% in prior studies (3, 6, 8, 9, 29), making this threshold clinically relevant and suggesting that the predictive utility of T-chol may generalize across patient populations. To our knowledge, this is the first study to demonstrate a strong association between the serum T-chol levels and AC completion in patients with pancreatic cancer. Although our exploratory paired-sample analysis was limited by small numbers, patients whose T-chol declined during NAC tended to have poorer survival. While not detailed here, this observation highlights the potential prognostic value of dynamic T-chol monitoring and warrants further investigation.
Clinically, these results support the use of T-chol as a practical, robust, and inexpensive biomarker for perioperative risk stratification. In particular, T-chol may help identify high-risk patients who could benefit from targeted supportive strategies such as nutritional counseling, pancreatic enzyme replacement, glycemic control when indicated, early mobilization, and prehabilitation. Applying these measures selectively to high-risk individuals may improve resource efficiency and enhance AC adherence without over-treating lower-risk patients. Consistent with our findings, a previous study reported that postoperative cholesterol levels measured one month after surgery predicted long-term outcomes in patients with pancreatic cancer, with lower cholesterol associated with significantly worse survival (12). Moreover, low T-chol has been linked to an increased short-term risk of developing pancreatic cancer, supporting its potential value as an early diagnostic indicator (13). Thus, serum T-chol may have practical utility across multiple phases of pancreatic cancer management and may help guide individualized treatment strategies.
Study limitations. First, it included a relatively small number of patients from a single institution, and the prognostic analyses were retrospective. Larger prospective studies are needed to further evaluate the clinical relevance of serum T-chol monitoring and its role in predicting chemotherapy tolerance. In addition, although we focused primarily on serum T-chol levels, other lipid parameters, such as low-density lipoprotein, high-density lipoprotein, and very-low-density lipoprotein, were not assessed in detail.
Conclusion
Completion of AC is a key determinant of survival after PD. Low serum T-chol, together with advanced age, independently predicted AC non-completion, establishing T-chol as a simple and robust biomarker for risk stratification and supportive care planning. Prospective validation and dynamic monitoring of T-chol may help optimize individualized strategies to improve treatment adherence and outcomes.
Acknowledgements
The Authors would like to thank Editage (www.editage.com) for English language editing.
Footnotes
Authors’ Contributions
Conception and design: K.H., S.N. and R.T.; Collection and assembly of data: K.H., S.N., R.T., H.S., K.N., M.K., J.T. and S.K.; Data analysis and interpretation: K.H., S.N. and R.T.; Manuscript preparation: K.H.; Revision of the manuscript: T.I.; Final approval of manuscript: All Authors.
Supplementary Material
Supplementary Figures and Tables are available at the Zenodo repository: https://doi.org/10.5281/zenodo.18878053
Conflicts of Interest
The Authors declare no conflicts of interest in relation to this study.
Funding
Not applicable.
Artificial Intelligence (AI) Disclosure
No artificial intelligence (AI) tools, including large language models or machine learning software, were used in the preparation, analysis, or presentation of this manuscript.
- Received March 10, 2026.
- Revision received March 28, 2026.
- Accepted April 3, 2026.
- Copyright © 2026 The Author(s). Published by the International Institute of Anticancer Research.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.









