Resection Rates and Predictors of Resectability of Pancreatic Tumors at Mulago Hospital: A Retrospective Cross-Sectional Study

Background

Pancreatic tumors are among the most lethal malignancies globally, ranking 12th in incidence and 7th in mortality, and accounting for 4.6% of all cancer-related deaths.^1,2 In Uganda, the Global Cancer Observatory (GLOBOCAN 2023) ranks pancreatic cancer 22nd in incidence and 17th in cancer-related deaths, with 290 new cases and 271 deaths reported. Records at the Upper GIT Unit of Mulago National Referral Hospital (MNRH) show a rising burden, with an average of six new cases admitted monthly.

The overall 5-year survival for pancreatic tumors remains poor, ranging from 2% to 9% globally, regardless of income level.³ Surgical resection is the only potentially curative intervention and can raise the 5-year survival rate to approximately 17.5%.⁴ However, only 20% of patients are eligible for resection at diagnosis, due to advanced local or metastatic disease.^5,6 The complexity of pancreatic surgery, coupled with its high perioperative risks,⁷ necessitates thorough preoperative evaluation to select appropriate surgical candidates.

Traditionally, resectability is assessed radiologically, primarily using contrast-enhanced CT with a positive predictive value of 75–86%.⁸ This remains the most accessible tool in low-income countries such as Uganda. Internationally recognized frameworks such as the NCCN guidelines⁹ and the Resectability Scoring System¹⁰ guide imaging-based evaluation. However, clinical and biochemical markers, including hemoglobin, albumin, CA 19–9, and CEA have also been shown to contribute meaningfully to resectability prediction.¹¹

Despite advances in diagnostics, resectability prediction remains imprecise. Studies show that nearly 50% of tumors deemed resectable preoperatively are found unresectable at surgery.¹² A pilot study at MNRH (unpublished) showed a resection rate below 10%, significantly lower than the global benchmark of 20%, suggesting a major gap in surgical outcomes.

This study, therefore, aims to determine the resection rate and identify clinical, biochemical, and radiological predictors of resectability of pancreatic tumors at Mulago National Referral Hospital. The findings will help improve preoperative decision-making and patient outcomes in low-resource settings.

Methods

Study Design and Setting

This was a retrospective cross-sectional study conducted at Mulago National Referral Hospital (MNRH) from January 2021 to December 2024. MNRH is Uganda’s largest tertiary and teaching hospital, affiliated with Makerere University College of Health Sciences. It has a bed capacity of about 1500 and provides specialized surgical and medical services.

Patients with pancreatic tumors are managed in the Upper Gastrointestinal and Pancreato-hepato-biliary Unit of the Directorate of Surgery. This ward has a capacity of 35 beds and admits an average of four pancreatic tumor cases monthly. Patients are referred through the surgical outpatient clinic, accident and emergency unit, or from the Uganda Cancer Institute. Diagnosis is based on triphasic abdominal CT scans performed at MNRH or peripheral facilities, with reports confirmed by MNRH radiologists. Elective pancreatic surgeries such as Whipple’s procedures and distal pancreatectomies are performed twice weekly, usually following tumor board review, with postoperative intensive care provided for a minimum of 72 hours.

Study Population and Sampling

The study population comprised patients with pancreatic tumors admitted to MNRH between January 2021 and December 2024. Eligible participants were adult patients admitted to Mulago National Referral Hospital with radiologically suspected pancreatic tumors confirmed on contrast-enhanced CT scan because in our setting, histopathological confirmation was often not feasible for all patients due to limited access to endoscopic ultrasound-guided biopsy and the advanced stage at presentation, where many patients are not surgical candidates. Only records with adequate clinical, biochemical, radiological, and operative documentation were included in the analysis.

Exclusion criteria were: Refusal of surgery despite eligibility, concurrent malignancies, previous pancreatic surgery, patients with incomplete records or imaging findings inconsistent with pancreatic neoplasms were excluded.

Sample size estimation using Cohen’s method yielded 215 participants; however, after adjusting for the finite population of 144 admissions in four years and allowing for missing records, the final required sample was 95 patient files. A consecutive sampling method was employed to obtain all eligible records until the target was reached.

Study Procedure and Data Collection

Names and in-patient numbers of patients with pancreatic tumors were retrieved from the Upper GI unit admission register. Patient files were then obtained from the records office and screened against eligibility criteria. Data were extracted using a structured tool capturing demographic characteristics, clinical symptoms (jaundice, abdominal pain, weight loss), biochemical parameters (hemoglobin, CA 19–9, CEA), radiological features (tumor size, TNM stage, vascular involvement, metastases), operative findings, and resectability status at surgery. Two trained research assistants collected the data under supervision of the principal investigator, with completeness and accuracy checked through double entry.

Assessment of Resectability

Tumor resectability was assessed preoperatively based on contrast-enhanced CT findings. Criteria for resectability included absence of distant metastasis and lack of major arterial encasement involving the celiac axis, superior mesenteric artery, or common hepatic artery. Limited involvement of the portal vein or superior mesenteric vein was considered potentially resectable.

Statistical Analysis

Data were analyzed using statistical software. Continuous variables were summarized using means and standard deviations, while categorical variables were presented as frequencies and percentages.

Associations between predictor variables and tumor resectability were assessed using independent-sample t-tests and crude odds ratios. Variables with clinical relevance were further evaluated using multivariate logistic regression to determine independent predictors of resectability.

A p-value of less than 0.05 was considered statistically significant.

Ethical Considerations

Ethical approval for this study was obtained from the Makerere University School of Medicine Research and Ethics Committee (Mak-SOMREC-2024-1077) and administrative clearance from Mulago National Referral Hospital (MHREC 2892).

Because this was a retrospective study involving review of existing medical records, the requirement for individual patient consent was waived by the ethics committee. All patient data were anonymized prior to analysis to ensure confidentiality. The study was conducted in accordance with the principles of the Declaration of Helsinki.

Results

Description of Study Participants and Their Clinical Characteristics

Participants were middle-aged, with a mean age of 57.1 ± 14.7 years. Symptom duration before presentation was prolonged: jaundice persisted for an average of 10.5 ± 8.9 weeks, abdominal pain for 16.1 ± 16.1 weeks, and weight loss for 15.1 ± 14.5 weeks. Laboratory findings showed marked hyperbilirubinemia (mean total bilirubin 257.7 µmol/L, direct bilirubin 188.7 µmol/L), reflecting obstructive jaundice in most patients. The mean hemoglobin level was 10.4 ± 2.2 g/dL, indicating mild-to-moderate anemia. Tumor markers were frequently elevated, with CA 19–9 averaging 448.2 U/mL and CEA averaging 12.0 ng/mL, though both demonstrated wide variability (Table 1).

Table 1 Baseline Clinical and Biochemical Characteristics of Study Participants

In the categorical demographic and radiologic characteristics. The sex distribution was balanced (48% male, 52% female). Tumors were predominantly located in the pancreatic head (92%), with few in the body (7%) or tail (1%). Lymph node involvement was observed in 28.3%, and distant metastases in 20% of patients. Using Manchester stratification, 49% had early-stage disease (TNM I–II), while 51% presented with late-stage disease (TNM III–IV). Vascular involvement was frequent: SMA (23%), celiac axis (15%), SMV (11%), portal vein (8%), and CHA (3%). These findings illustrate a predominance of locally advanced or metastatic tumors among patients evaluated at MNRH (Table 2).

Table 2 Participant Characteristics

Resection Rates of Pancreatic Tumors at MNRH

Out of 100 patients with pancreatic tumors, only 21 underwent surgical resection, yielding an overall resection rate of 21%. Preoperatively, 55 patients (55%) had been classified as resectable on CT imaging, but intraoperative findings revealed that 34 of these (61.8%) were in fact unresectable, mostly due to vascular invasion and previously undetected metastases.

Resection rates varied across clinical and radiological variables. Patients with tumors ≤4 cm were more likely to undergo resection (62.5%) compared to those with tumors >4 cm, where resection was rare (8%). Similarly, all patients with distant metastases were unresectable, giving a 0% resection rate in this subgroup, while patients without metastases had a resection rate of 26.3%. Among those with lymph node involvement, only 7% were resected compared to 25% in those without nodal spread. Tumor location also showed differences: resection was slightly more common in tumors of the pancreatic head (19.5%) compared to body/tail lesions (37.5%), although numbers were small. Patients with hemoglobin >10 g/dL had higher resection rates (24.6%) than those below this threshold (15.4%), suggesting better surgical candidacy in less anemic patients.

Univariate Analysis of Variables

Differences in mean values between resected and non-resected patients were compared using the independent samples t-test and among the variables analyzed, only tumor size differed significantly between resected and non-resected patients. Resected patients had a mean tumor size of 3.6 cm compared to 6.0 cm in the unresected group (p < 0.001). Other parameters including age, duration of symptoms, bilirubin levels, hemoglobin, and tumor markers (CEA, CA 19–9) showed no significant differences between the two groups (Table 3).

Table 3 Differences in Mean Between Resected and Non-Resected Patients

The association between major vascular involvement and resectability was also explored. None of the patients with significant arterial involvement (superior mesenteric artery, celiac axis, or common hepatic artery) underwent surgical resection. Similarly, resection rates were markedly reduced among patients with venous involvement (superior mesenteric vein and portal vein). Due to the low frequency of resectable cases within these subgroups, statistical estimation of odds ratios was limited.

Further analysis assessed associations between patient and tumor characteristics and the likelihood of tumor resectability. Only three variables had p-values < 0.2 and were considered to have potential significant associations with resectability: tumor size, presence of metastases, and lymph node involvement. Notably, none of the patients with metastases underwent surgical resection, resulting in an odds ratio of zero for this group (Table 4).

Table 4 Univariate Analysis of Predictors of Tumor Resectability

Multivariate Analysis for Predictors of Resectability

In the multivariate analysis, variables with clinical relevance in the univariate analysis were included in the model. These variables were tumor size (≤4 cm vs >4 cm) with p < 0.001, lymph node involvement (p = 0.141), and tumor location (head vs non-head) with p = 0.115. Additionally, hemoglobin levels (<10 g/dL vs ≥10 g/dL) were included despite a p-value of 0.285, due to their clinical relevance (Table 5).

Table 5 Multivariate Logistic Regression Analysis of Predictors of Tumor Resectability

In the multivariate model, tumor size emerged as the only independent predictor of resectability. Patients with tumors ≤4 cm were significantly more likely to undergo successful resection compared to those with tumors >4 cm (AOR = 0.054; p < 0.001). Although not statistically significant, higher hemoglobin levels (>10 g/dl) and absence of lymph node involvement were associated with increased likelihood of resection. Tumor location outside the pancreatic head showed a trend toward lower resectability. Overall, these findings highlight tumor size as the most reliable determinant of resectability in this cohort.

Discussion

In this study, the overall resection rate of pancreatic tumors at Mulago National Referral Hospital (MNRH) was 21% consistent with findings from other low-resource settings where late presentation significantly limits surgical eligibility. This finding lies within the global resection rate range of 15–25% reported in multiple international studies.^5,6 When compared with African data, our resection rate was notably higher than that reported in Kenya (7%) and Tanzania (6%),^13,14 but comparable to the 23% documented in Ethiopia.¹⁵

Despite over half (55%) of the patients being classified as resectable preoperatively, more than 60% of these were deemed unresectable intraoperatively due to vascular involvement or previously undetected metastases. This high discrepancy between preoperative and intraoperative findings is consistent with previous findings.¹²

The mean age of patients in this study was 57.1 years, and the sex distribution was nearly equal. These findings are consistent with studies conducted in Saudi Arabia, Ethiopia, and East Africa.¹¹

Clinically, patients most frequently presented with jaundice, abdominal pain, and weight loss. These findings align with previous observations.¹⁶

Biochemical evaluation revealed that most patients had anemia, consistent with previous reports.¹⁷ Tumor markers showed numerical differences but were not statistically significant predictors.¹⁸

Radiologically, lymph node involvement and metastases were seen in 28.3% and 20% of patients respectively. In line with previous studies, patients with nodal disease or metastases were less likely to undergo resection.^8,19

Major vascular involvement demonstrated a strong association with unresectability. Notably, none of the patients with arterial encasement involving the superior mesenteric artery (SMA), celiac axis (CA), or common hepatic artery (CHA) underwent successful resection. Similarly, venous involvement of the superior mesenteric vein (SMV) and portal vein (PV) was associated with markedly reduced resection rates. These findings are consistent with established oncologic principles, where arterial encasement is generally considered a contraindication to upfront surgical resection, while limited venous involvement may be amenable to resection in selected cases with vascular reconstruction.^20,21 Findings reinforce the critical role of vascular involvement in determining pancreatic tumor resectability, while also highlighting the challenges of quantitatively modeling this relationship in small cohorts within low-resource settings.

Tumor size greater than 4 cm emerged as the strongest predictor of unresectability. Larger tumors are more likely to invade adjacent vascular structures and metastasize, reducing the likelihood of successful surgical resection. Previous studies have similarly demonstrated that increasing tumor size is associated with poorer surgical outcomes.²²

At the molecular level, pancreatic ductal adenocarcinoma develops through accumulation of genetic alterations involving KRAS, TP53, CDKN2A, and SMAD4 genes, which drive tumor progression and metastatic potential. These molecular mechanisms do not demonstrate strong sex-specific predilection, which may explain the nearly equal sex distribution observed in this study.

The predominance of advanced disease at presentation likely reflects delayed health-seeking behavior, limited diagnostic capacity, and referral patterns typical of low-resource health systems.

Limitations

This study has several limitations. First, the relatively small sample size and the limited number of resected cases may reduce the statistical power of the multivariate model. Second, as a single-center study conducted at a national referral hospital, the patient population may include more complex or advanced cases, introducing potential referral bias. The retrospective design relied on existing medical records, which may contain incomplete data. Finally, diagnosis of pancreatic tumors was primarily based on contrast-enhanced CT without universal histopathological confirmation. While CT is central to assessment of pancreatic malignancy, it cannot definitively distinguish pancreatic ductal adenocarcinoma from other lesions, which may have influenced the observed resectability rates and predictors.

Despite these limitations, the study provides important insights into pancreatic tumor management in a low-resource setting.

Conclusion

The resection rate for pancreatic tumors at MNRH was 21%, comparable with global resection rates with tumor size <4 cm being the most robust predictor of resectability. Although variables such as CA 19–9, CEA, and hemoglobin levels showed numerical differences between groups, these were not statistically significant predictors of resectability in this study. Improved early detection and access to advanced imaging may help increase surgical eligibility and improve patient outcomes in low-resource settings.