ABSTRACT
Introduction
Body composition abnormalities, including sarcopenia and adipose tissue alterations, have emerged as potential prognostic biomarkers in pancreatic ductal adenocarcinoma (PDAC). The present study investigated the relationship between preoperative computed tomography (CT)-derived adipose and muscle parameters and survival outcomes in patients undergoing pancreaticoduodenectomy for PDAC.
Methods
This retrospective analysis enrolled 69 patients with histologically confirmed PDAC who underwent pancreaticoduodenectomy between 2018 and 2024. Preoperative contrast-enhanced CT images acquired within 30 days before surgery were assessed at the level of the third lumbar vertebra (L3). Measurements included the area and attenuation of visceral and subcutaneous adipose tissue, as well as the area and attenuation of the psoas muscle. Overall survival (OS) and disease-free survival (DFS) were evaluated using Kaplan-Meier survival curves and Cox proportional hazards models.
Results
Median OS and DFS were 10.0 and 7.0 months, respectively. Kaplan-Meier analyses showed no statistically significant differences in survival across body composition parameters (all p>0.05), although several parameters exhibited consistent numerical trends in survival. In multivariable analysis, higher T stage was independently associated with worse OS [hazard ratio (HR): 1.83, 95% confidence interval (CI): 1.12-3.00, p=0.015], whereas adjuvant chemotherapy remained independently associated with improved OS (HR: 0.16, 95% CI: 0.07-0.34, p<0.001). Because of the limited number of DFS events, multivariable analysis was not performed.
Conclusion
CT-derived body composition analysis may provide complementary prognostic information in patients undergoing pancreaticoduodenectomy for PDAC. Although independent associations were not consistently demonstrated, the observed survival patterns support further investigation of CT-derived muscle and adipose tissue parameters as potential imaging biomarkers. Larger prospective multicenter studies are warranted to clarify their prognostic value.
Introduction
Pancreatic ductal adenocarcinoma (PDAC) is among the most aggressive and difficult-to-treat malignancies owing to its biological characteristics. According to GLOBOCAN 2020, pancreatic cancer accounted for near 500,000 new cases and over 450,000 deaths worldwide, reflecting its disproportionately high mortality burden (1). Surgical resection remains the primary curative treatment; for tumors located in the pancreatic head, pancreaticoduodenectomy (Whipple procedure) is the only surgical option. Yet, even in patients who undergo resection, outcomes remain disappointing, with a substantial proportion experiencing early recurrence and limited long-term survival (2, 3).
In this context, improving preoperative risk stratification has become increasingly important. Conventional prognostic factors such as tumor stage, lymph node involvement, and margin status are typically determined postoperatively, limiting their utility in preoperative decision-making. Therefore, there is growing interest in identifying non-invasive, imaging-based biomarkers that can provide prognostic information before surgery. Computed tomography (CT), routinely performed in the preoperative setting, offers an opportunity to extract such information beyond standard anatomical assessment.
Among CT-derived parameters, body composition analysis at the third lumbar vertebra (L3) level provides a reliable representation of whole-body tissue distribution. A single cross-sectional CT image at this level can be used to quantify both adipose and lean tissue compartments with high accuracy (4). This has enabled the use of routine imaging data for quantitative assessment of patient-specific physiological characteristics.
Adipose tissue is increasingly recognized as an active metabolic organ rather than a passive energy reservoir. Both the quantity and quality of adipose tissue—reflected by cross-sectional area and attenuation values measured in Hounsfield units (HUs)—have been associated with systemic inflammation, cachexia, and tumor progression (5). Similarly, skeletal muscle parameters such as psoas muscle area and attenuation may reflect sarcopenia and myosteatosis, both of which have been linked to impaired physiological reserve and adverse oncological outcomes (6). Importantly, these measurements can be easily derived from standard preoperative CT scans without incurring extra costs or imposing an additional burden on patients.
Evidence from gastrointestinal malignancies suggests that both adipose tissue and skeletal muscle characteristics may carry prognostic significance (7, 8). Increased visceral adiposity has been associated with adverse oncologic outcomes in colorectal cancer (9), while higher visceral fat attenuation has been linked to worse survival in gastric cancer (10). Likewise, reduced muscle attenuation and myosteatosis have been associated with inferior postoperative and long-term outcomes in several gastrointestinal cancers. Nevertheless, the prognostic significance of these CT-derived body composition parameters in PDAC remains incompletely defined. In particular, the relative contributions of adipose tissue quantity and quality and of muscle-related parameters to survival outcomes remain uncertain in surgically treated PDAC patients.
Therefore, this study was designed to investigate the association between preoperative CT-derived measures at the L3 vertebral level (subcutaneous and visceral adipose tissue area, adipose tissue attenuation values, and psoas muscle area and attenuation) and overall survival (OS) in patients undergoing pancreaticoduodenectomy for PDAC. Secondary objectives included evaluating the relationship between these parameters and disease-free survival (DFS) and 30-day postoperative mortality.
Methods
Study Design and Patient Selection
This retrospective cohort study included patients who underwent pancreaticoduodenectomy (Whipple procedure) at University of Health Sciences Türkiye, Ümraniye Training and Research Hospital between January 2018 and December 2024. Eligible patients were identified using the institutional surgical registry. The study was performed and reported in compliance with the Strengthening the Reporting of Observational Studies in Epidemiology recommendations.
Patients were included if they had histopathologically confirmed PDAC, were aged 18 years or older, had a preoperative abdominal CT scan obtained within 30 days before surgery, and had available OS data.
Patients were excluded if the final histopathology was not consistent with PDAC; if preoperative CT imaging was unavailable or technically inadequate for adipose tissue analysis; if OS data were unknown; if they were younger than 18 years; or if pancreaticoduodenectomy had been performed emergently for hemorrhage. PDAC patients located in the tail of the pancreas were excluded due to differences in the surgical approach and surgical burden.
Data Collection
Demographic, clinical, operative, and pathological data were retrieved retrospectively from electronic medical records. Variables of interest, where available, included age, sex, tumor-related pathological findings, nodal status, resection margin status, and adjuvant treatment status.
Follow-up data were obtained from hospital records and other accessible institutional follow-up sources. OS status, recurrence status, and 30-day postoperative mortality were recorded. Because of the retrospective nature of the study, DFS data were not expected to be available for all patients.
CT Acquisition and Image Analysis
Preoperative contrast-enhanced abdominal CT examinations obtained within 30 days of surgery were retrospectively reviewed. Body composition analysis was performed at the level of the L3 vertebra, which served as the predefined anatomical landmark. Portal venous phase CT images were used for all body composition measurements.
On a single representative axial CT image at the L3 level, subcutaneous adipose tissue, visceral adipose tissue, and bilateral psoas muscles were evaluated. Subcutaneous and visceral adipose tissue areas were segmented using a threshold-based semi-automatic method with a predefined adipose tissue attenuation range of -190 to -30 HUs, followed by manual correction when necessary. During segmentation, bowel contents, vessels, and bone structures were carefully excluded.
For each patient, subcutaneous and visceral adipose tissue areas and their mean attenuations were recorded. All area measurements were expressed in mm2, whereas attenuation values were expressed in HU. Representative examples of visceral adipose tissue, subcutaneous adipose tissue, and psoas muscle measurements are shown in Figure 1.
Bilateral psoas muscles were segmented on the same axial slice using a semi-automatic method followed by manual correction. The right and left psoas muscle areas were summed to obtain the total psoas muscle area. Psoas muscle attenuation was measured as the mean attenuation of the segmented bilateral psoas muscles.
All CT measurements were performed by two radiologists in consensus: an abdominal radiologist with 12 years of experience and a radiology resident with 4 years of experience. Both readers were blinded to clinical, laboratory, and survival outcome data. For each patient, a single final consensus measurement was recorded for each CT-derived body composition parameter. Total psoas muscle area was used as a pragmatic surrogate marker of skeletal muscle mass because height-normalized skeletal muscle index (SMI) measurements were not consistently available owing to the retrospective study design.
Outcomes
The primary outcome was OS, measured from the date of surgery until death from any cause or the last available follow-up.
Secondary outcomes included DFS, defined as the interval from surgery to recurrence, death, or last follow-up among patients with documented recurrence status; and early postoperative mortality, defined as mortality occurring within 30 days of surgery.
Ethical Approval
The study adhered to the principles of the Declaration of Helsinki and received approval from the University of Health Sciences Türkiye, Ümraniye Training and Research Hospital (decision number: 146, date: 22.04.2026). Because of the retrospective design and the exclusive use of anonymized patient data, the Institutional Ethics Committee waived the requirement for written informed consent.
Statistical Analysis
Statistical analyses were performed using IBM SPSS Statistics (version 26.0; IBM Corp., Armonk, NY, USA); data visualization and forest-plot generation were performed using RStudio (version 2025.05.1). Continuous variables were summarized as mean ± standard deviation or median with interquartile range (IQR) according to data distribution, whereas categorical variables were expressed as frequencies and percentages.
Radiological body composition parameters were primarily analyzed as continuous variables in Cox proportional hazards regression models to minimize information loss associated with dichotomization. For illustrative Kaplan–Meier analyses, body composition parameters were dichotomized according to median values. Survival curves were estimated using the Kaplan-Meier method and compared using the log-rank test. Additional Kaplan-Meier analysis was performed, stratified by adjuvant chemotherapy status, to provide an unadjusted comparison of OS.
Associations between clinicopathological variables, radiological body composition parameters, and survival outcomes were further evaluated using Cox proportional hazards regression analysis. Imaging parameters were analyzed as continuous variables in Cox models. Univariable Cox regression analyses were initially performed for both OS and DFS. Variables demonstrating statistical or borderline significance in univariable analyses were considered for multivariable Cox regression. T stage was also retained in the final model because of its established prognostic importance in PDAC and its borderline association with OS in the univariable analysis. To minimize the risk of model overfitting, the final multivariable model was restricted to a limited number of covariates. Because of the limited number of DFS events and concerns regarding model stability, multivariable Cox regression analysis was not performed.
The proportional hazards assumption was assessed using both graphical log-minus-log survival plots and formal testing based on Schoenfeld residuals.
Thirty-day postoperative mortality and complications were analyzed descriptively. Because of the limited number of early mortality events, regression analysis was not performed for the 30-day mortality endpoint.
A two-sided p value of less than 0.05 was considered statistically significant.
Results
A total of 231 patients who underwent pancreaticoduodenectomy (Whipple procedure) were screened for eligibility. Among them, 73 patients had histopathologically confirmed PDAC. Three patients were excluded because preoperative CT images were unavailable, and one patient was excluded due to loss to follow-up. Consequently, 69 patients were included in the OS analysis. Eight additional patients were excluded from DFS analysis because postoperative imaging required for recurrence assessment was unavailable, resulting in 61 patients being included in the DFS cohort (Figure 2).
The mean age of the patients was 64.0±7.8 years. Most patients had advanced pathological disease, with T2 and T3 tumors accounting for 49.3% and 36.2% of cases, respectively, while N2 nodal involvement was observed in 46.4% of patients. The mean positive lymph node count was 5.8±4.4, and the majority of tumors were grade 2 (81.2%). Margin positivity was identified in 31.9% of patients, whereas lymphovascular invasion and perineural invasion were present in 72.5% and 98.6% of cases, respectively. Adjuvant therapy was administered to 78.3% of patients, whereas only two patients (2.9%) received neoadjuvant treatment. Additional demographic, clinicopathological, biochemical, and radiological body composition characteristics are summarized in Table 1.
Postoperative pancreatic fistula developed in 21 (30.4%) patients, based on the criteria established by the International Study Group on Pancreatic Surgery, irrespective of fistula grade. Grade B/C fistulas occurred in 12 (17.3%) patients. Thirty-day mortality occurred in 5 patients (7.2%). Two deaths were attributed to sepsis secondary to pancreaticojejunostomy (PJ) leakage, while two additional patients died during intensive care follow-up after hemorrhagic complications associated with PJ leakage. One patient died from myocardial infarction on postoperative day 9, unrelated to any surgical morbidity.
Median OS was 10.0 months (IQR: 11.0), while median DFS was 7.0 months (IQR: 5.0). The median follow-up duration was 12.0 months (IQR: 10.0 months). Overall, mortality was observed in 53 patients (76.8%), and recurrence occurred in 35 patients (57.3%). The OS analysis was based on all-cause mortality.
For illustrative Kaplan-Meier analyses, body composition parameters were dichotomized according to median values. Kaplan-Meier analysis demonstrated no significant differences in OS by visceral fat area, visceral fat attenuation, subcutaneous fat area, subcutaneous fat attenuation, psoas muscle area, or psoas muscle attenuation (all p>0.05). Nevertheless, patients with higher visceral fat area demonstrated a non-significant trend toward longer OS than those with lower visceral fat area (median: 8.0 vs. 14.0, p=0.161). Similarly, a higher psoas muscle area was associated with longer OS (median: 9.0 vs. 13.0 months, p=0.135). Detailed OS analyses are summarized in Table 2 and Figure 3.
Likewise, DFS analyses showed no statistically significant differences in survival by visceral fat area, visceral fat attenuation, subcutaneous fat area, subcutaneous fat attenuation, psoas muscle area, or psoas muscle attenuation (all p>0.05). However, patients with lower psoas muscle attenuation had a numerically longer DFS (median: 9.0 vs. 7.0) than those with higher attenuation values; this difference was not statistically significant (p=0.418). Detailed DFS analyses are presented in Table 2 and Figure 4.
Kaplan-Meier analysis stratified by adjuvant chemotherapy status demonstrated significantly longer OS for patients who received adjuvant chemotherapy than for those who did not (log-rank p<0.001; Supplementary Figure 1).
Univariable Cox regression analysis for OS demonstrated borderline associations with T stage [hazard ratio (HR): 1.344, 95% confidence interval (CI): 0.855-2.113, p=0.200] and psoas muscle area (HR: 0.946, 95% CI: 0.894-1.000, p=0.052), whereas adjuvant therapy emerged as a significant protective factor (HR: 0.198, 95% CI: 0.099-0.396, p<0.001). In multivariable analysis, higher T stage was associated with worse OS (HR: 1.83, 95% CI: 1.12-3.00, p=0.015), whereas adjuvant chemotherapy was associated with improved OS (HR: 0.16, 95% CI: 0.07-0.34, p<0.001). Although higher psoas muscle area showed a protective effect (HR: 0.96 per 100 mm2 increase), this association was not statistically significant (95% CI: 0.90-1.02, p=0.200).
Graphical assessment using log-minus-log survival plots did not demonstrate major violations of the proportional hazards assumption. Formal Schoenfeld residual testing was performed for the revised OS multivariable model. A statistically significant violation was identified for adjuvant chemotherapy (p=0.003), suggesting that the effect of adjuvant treatment on OS may not be constant over the entire follow-up period. Therefore, the HR for adjuvant chemotherapy should be regarded as an average effect throughout the follow-up period rather than a fixed treatment effect.
For DFS, univariate Cox regression analysis demonstrated a significant association for neoadjuvant therapy (HR: 11.326, 95% CI: 1.178-108.903, p=0.036), although this finding should be interpreted cautiously because only two patients received neoadjuvant therapy. Positive lymph node count (p=0.134), psoas muscle attenuation (p=0.137), and CA19-9 level (p=0.092) demonstrated borderline significance. Because of the limited number of DFS events and concerns regarding model stability, multivariable Cox regression analysis was not performed for DFS. Detailed Cox regression analyses are summarized in Table 3, and the OS multivariable model is visualized in Figure 5.
Discussion
In this retrospective analysis of patients undergoing pancreaticoduodenectomy for PDAC, CT-derived body composition parameters demonstrated heterogeneous associations with long-term oncological outcomes. Although most Kaplan–Meier comparisons did not reach statistical significance, several radiological markers showed consistent numerical separation of survival curves, suggesting a potential prognostic signal that may become more apparent in larger cohorts. Among the evaluated body composition parameters, psoas muscle area showed a borderline association with OS in univariable analysis and a consistently protective effect in multivariable analysis; however, this association did not remain statistically significant after adjustment. Higher T stage remained independently associated with worse OS, while adjuvant therapy consistently demonstrated a strong protective effect on OS.
Body composition analysis has attracted increasing interest in PDAC because conventional anthropometric measures often fail to capture cancer-related metabolic and muscular alterations. In particular, sarcopenia and myosteatosis have been linked to impaired functional reserve, systemic inflammation, treatment intolerance, and poorer oncological outcomes in several gastrointestinal malignancies (11, 12). Recent studies have suggested that muscle quality, reflected by muscle attenuation on CT, may carry greater prognostic relevance than muscle quantity alone (13-16). A recent meta-analysis has also suggested that myosteatosis may outperform traditional sarcopenia measurements as a prognostic biomarker in pancreatic malignancies (17).
Our findings further support the growing interest in skeletal muscle characteristics as potential imaging biomarkers in PDAC. Although higher psoas muscle area showed a numerical association with longer OS in Kaplan-Meier analysis and a protective direction of effect in multivariable analysis, this association did not reach statistical significance after adjustment. Likewise, psoas muscle attenuation demonstrated a borderline association with DFS in univariable analysis but did not reach statistical significance. Overall, these findings indicate that both muscle quantity and quality may warrant further investigation as potential imaging biomarkers. Given the exploratory design and relatively small sample size, these findings should be interpreted with caution.
The relationship between adiposity and pancreatic cancer outcomes remains complex. While visceral obesity has traditionally been linked to adverse oncological biology and postoperative morbidity (18), several recent reports have suggested that severe depletion of adipose tissue may also represent an advanced catabolic state associated with poor survival (19). In our cohort, higher visceral fat area was numerically associated with longer OS and DFS, although these differences did not reach statistical significance. Rather than contradicting prior literature, these findings further support the concept that body composition in PDAC likely reflects a dynamic interaction between nutritional reserve, systemic inflammation, and tumor-related metabolic stress, and warrant further investigation in larger studies specifically designed to evaluate adipose tissue biology and distribution patterns.
Another notable finding of the present study was an independent association between adjuvant therapy and improved survival outcomes. Adjuvant chemotherapy remained independently associated with improved OS in multivariable analysis, underscoring the continued importance of systemic therapy even in surgically treated cohorts. This observation aligns closely with modern PDAC management paradigms and contemporary randomized trials demonstrating substantial survival benefits with adjuvant chemotherapy following resection (2, 20). However, formal assessment of the proportional hazards assumption using Schoenfeld residual testing revealed a potential violation for adjuvant therapy in the OS model (p=0.003), suggesting that the effect of adjuvant treatment on survival may not be constant over time. Accordingly, the estimated hazard ratio for adjuvant therapy should be interpreted as an average effect over the follow-up period rather than a constant treatment effect.
The present study also provides clinically relevant perioperative context. Postoperative pancreatic fistula occurred in approximately one-third of patients, and early mortality was predominantly related to septic or hemorrhagic complications associated with PJ leakage. These findings reflect the continued technical and physiological complexity of pancreatic surgery, despite ongoing improvements in perioperative care. Although postoperative complications showed no significant association with OS in univariable analysis, larger studies specifically designed to evaluate postoperative morbidities are needed to better define their long-term oncological impact.
This study has several strengths. First, all body composition measurements were derived from routinely obtained preoperative CT imaging, which supports potential real-world applicability without additional cost or invasive assessment. Second, both adipose tissue and muscle-related parameters were evaluated simultaneously, allowing a more integrated assessment of body composition. Third, survival analyses incorporated both Kaplan–Meier and Cox regression approaches, enabling complementary evaluation of categorical survival patterns and continuous imaging parameters.
Study Limitations
This study has several limitations. First, its retrospective single-center design may have introduced selection bias. The relatively limited sample size and event burden may have reduced statistical power, particularly for subgroup analyses and multivariable modeling. Consequently, multivariable analysis was restricted to OS, whereas DFS was evaluated using univariable Cox regression because of concerns regarding model stability. In addition, body composition was measured at a single preoperative time point and therefore could not capture longitudinal changes in muscle or adipose tissue during treatment. Because multiple exploratory statistical comparisons were performed, the possibility of a type I error should be considered when interpreting isolated statistically significant findings.
Furthermore, total psoas muscle area was used as a surrogate for skeletal muscle mass rather than the height-normalized SMI, given the retrospective nature of the study. Although height data are typically available in medical records, SMI-based measurements were not consistently performed; future studies should adopt standardized SMI to allow cross-study comparisons. Finally, although several radiological parameters demonstrated consistent numerical survival trends, most associations did not reach statistical significance after adjustment. Accordingly, these results should be regarded as preliminary and confirmed in larger, prospective, multicenter studies before being translated into clinical practice.
Conclusion
Overall, the present study suggests that CT-derived body composition parameters may warrant further investigation as potential imaging biomarkers in patients undergoing pancreaticoduodenectomy for PDAC. While most survival differences did not achieve statistical significance, muscle-related CT parameters showed consistent numerical associations with survival outcomes across multiple analyses, supporting their potential biological and clinical relevance. Future multicenter studies with larger cohorts and standardized radiological assessment protocols may help clarify the precise prognostic role of adipose- and muscle-related imaging biomarkers in PDAC.
