Preoperative computed tomography-based risk stratification model validation for postoperative pancreatic ductal adenocarcinoma recurrence

Abstract

BACKGROUND

The computed tomography (CT)-based preoperative risk score was developed to predict recurrence after upfront surgery in patients with resectable pancreatic ductal adenocarcinoma (PDAC) in South Korea. However, whether it performs well in other countries remains unknown.

AIM

To externally validate the CT-based preoperative risk score for PDAC in a country outside South Korea.

METHODS

Consecutive patients with PDAC who underwent upfront surgery from January 2016 to December 2019 at our institute in a country outside South Korea were retrospectively included. The study utilized the CT-based risk scoring system, which incorporates tumor size, portal venous phase density, tumor necrosis, peripancreatic infiltration, and suspicious metastatic lymph nodes. Patients were categorized into prognosis groups based on their risk score, as good (risk score < 2), moderate (risk score 2-4), and poor (risk score ≥ 5).

RESULTS

A total of 283 patients were evaluated, comprising 170 males and 113 females, with an average age of 63.52 ± 8.71 years. Follow-up was conducted until May 2023, and 76% of patients experienced tumor recurrence with median recurrence-free survival (RFS) of 29.1 ± 1.9 months. According to the evaluation results of Reader 1, the recurrence rates were 39.0% in the good prognosis group, 82.1% in the moderate group, and 84.5% in the poor group. In comparison, Reader 2 reported recurrence rates of 50.0%, 79.5%, and 88.9%, respectively, across the same prognostic categories. The study validated the effectiveness of the risk scoring system, demonstrating better RFS in the good prognosis group.

CONCLUSION

This research validated that the CT-based preoperative risk scoring system can effectively predict RFS in patients with PDAC, suggesting that it may be valuable in diverse populations.

Key Words: Pancreatic ductal adenocarcinoma; Postoperative recurrence; Risk assessment system; Computed tomography; Model validation

Core Tip: This multinational study externally validated a preoperative computed tomography-based risk model, originally developed in South Korea, for predicting recurrence-free survival (RFS) in patients with pancreatic ductal adenocarcinoma (PDAC), after surgery. Analyzing 283 patients from a distinct population, the model-incorporating tumor size, portal venous phase density, necrosis, peripancreatic infiltration, and metastatic lymph nodes-stratified patients into prognostic groups. The good prognosis group exhibited significantly longer RFS despite a 76% overall recurrence rate. The findings confirm its cross-population applicability, supporting its use as a preoperative tool to identify patients at high risk of PDAC and guide clinical decision-making.

INTRODUCTION

Pancreatic ductal adenocarcinoma (PDAC) is one of the most aggressive malignancies, with a poor 5-year survival rate of less than 10%[1,2]. Although surgical resection is currently the only potential cure, only 15% to 20% of patients present with localized disease amenable to surgery. Regrettably, a staggering 70% of these patients face tumor recurrence within 3 years post-resection. Accumulating evidence has highlighted the modest benefits of neoadjuvant chemotherapy in reducing the recurrence rate and improving survival outcomes[3]. However, the suboptimal integration of neoadjuvant therapy into routine clinical practice is largely attributable to the challenges in accurately assessing the disease preoperatively. Risk stratification holds significant clinical implications, as it helps tailor medical care based on a patient’s predicted prognosis. For example, patients in the high-risk group may benefit from preoperative neoadjuvant therapy or more intensive postoperative surveillance, whereas those in the low-risk group might be more appropriately treated with direct surgical intervention[4]. Consequently, there is a critical need for robust preoperative evaluation methodologies to facilitate risk stratification in patients with PDAC, thereby guiding optimal therapeutic management strategies.

As the cornerstone of imaging assessment in patients with PDAC, pancreatic computed tomography (CT) plays an indispensable role in clinical management. Notably, a cohort of researchers from South Korea previously developed a preoperative risk scoring system, integrating clinical parameters and CT-derived variables[3]. This innovative scoring system has demonstrated a significant correlation with postoperative tumor recurrence in patients with PDAC. We verified the effectiveness and universality of the risk scoring system in PDAC patient populations from geographic regions outside South Korea. This study maintained consistency with the Korean cohort in terms of CT protocols and imaging evaluation methods. However, the majority of patients in this study did not receive preoperative neoadjuvant chemotherapy. Validation of this system in a diverse patient cohort could potentially pave the way for its broader adoption in clinical practice, ultimately enhancing the precision of preoperative risk assessment and informing tailored therapeutic interventions for individuals affected by this formidable disease.

MATERIALS AND METHODS

Patients

A retrospective and consecutive cohort of patients with resectable PDAC, who underwent primary surgery at our institute between January 2016 and December 2019, was assessed. After excluding patients with concurrent extrapancreatic malignancies or intraoperative metastases (n = 18), those who underwent incomplete resections (n = 15), and those who did not have a preoperative CT examination within 14 days (n = 34), a total of 283 patients were eligible for analysis.

CT examination and variables

Examinations were conducted using one of two multidetector CT scanners (SOMATOM Definition AS + 128; Siemens, München, Germany or Erlangen Brilliance 128; Philips Healthcare, Best, the Netherlands). All scans were performed at settings of 120 kVp and 250 mAs, including non-enhanced imaging and imaging during the arterial phase and portal venous phase after the administration of intravenous contrast medium. The slice thickness was 5 mm.

CT variables were based on the radiology reports template recommended by the Abdominal Radiological Society and the American Pancreatic Association. These variables included tumor location and size, density of the tumor's solid portion in the arterial and venous phases (hypo, iso, or hyperdense), tumor infiltration of the pancreas, tumor contact with the portal vein or superior mesenteric vein, suspicious lymph nodes, indistinct attenuation increase with major vessel contact, and invasion of adjacent organs. Assessment of these variables was conducted based on the definitions provided in the report template. Additional CT variables included tumor conspicuity (good: Clear tumor margins; moderate: Visible but less distinct tumor; or poor: Indistinct or poorly visualized tumor) and tumor necrosis defined as non-enhancement of the tumor tissue in the arterial and portal venous phases.

All variables were independently evaluated by two radiologists with 10 years of experience in interpreting abdominal imaging who were blinded to the postoperative clinical data. One radiologist assigned a score of 1, while the other assigned a score of 2.

Kim et al[3] have identified five independent predictive factors for postoperative recurrence or mortality: Tumor size, low-density tumors in the portal venous phase, tumor necrosis, peripancreatic tumor infiltration, and suspicious metastatic lymph nodes.

Outcome measures

The primary endpoint was recurrence-free survival (RFS), defined as the interval from surgery to recurrence. Patients were followed up every 3 months for the first postoperative year and every 3-6 months thereafter until May 2023.

Pathologic findings

According to the International Study Group on Pancreatic Surgery, this study documented the types of pancreatic surgery (standard and extended) and the pathological results of surgical specimens, including the pathological margin status as defined by the Royal College of Pathologists guidelines. Pathological tumor staging, tumor differentiation, neural invasion, and lymphovascular invasion were determined in accordance with the American Joint Committee on Cancer 8^th Edition Staging System.

Statistical analysis

All data were statistically analyzed using SPSS 22.0 (IBM Corp., Armonk, NY, United States). Cohen's Kappa statistic was used to evaluate the consistency between readers. The model’s discriminatory power was assessed using hazard ratios (HRs) and 95%CI derived from a Cox model. Kaplan-Meier survival curves were used to visualize RFS across the three risk groups, and differences in survival were assessed using the log-rank test. P < 0.05 was considered statistically significant.

RESULTS

Patient characteristics

The clinical and imaging characteristics of the research samples are shown in Table 1. The median follow-up time was 24 months (range: 0.8-63.0 months). The median RFS was 29.1 months (range: 1.0-85.0 months). There were 113 cases of liver metastasis, 84 cases of local recurrence, 20 cases of lung metastasis, 27 cases of peritoneal seeding metastasis, and 4 cases of metastasis to other sites (adrenal glands, brain, vertebrae), with 37 cases having multiple site recurrences.

Table 1 Clinical and imaging characteristics of the study sample, n (%).

Variables	Total, n = 283	Recurrence, n = 216
Age, mean ± SD	63.52 ± 8.71	63.75 ± 8.43
Sex
Female	113 (39.9)	85 (39.3)
Male	170 (60.1)	131 (60.7)
Dominant location
Head and neck	164 (57.9)	126 (58.3)
Body and tail	119 (42.1)	90 (41.7)
Type of pancreatic surgery
Standard	183 (64.7)	132 (61.1)
Extended	100 (35.3)	84 (38.9)
Primary tumor (T) stage
T1	53 (18.7)	26 (12.0)
T2	190 (67.1)	155 (71.8)
T3	26 (9.2)	23 (10.6)
T4	14 (5.0)	12 (5.6)
Regional lymph node (N) stage
N0	164 (58.0)	118 (54.6)
N1	100 (35.3)	82 (38.0)
N2	19 (6.7)	16 (7.4)
Tumor size in cm
≤ 2	66 (23.3)	33 (15.3)
2-4	193 (68.2)	161 (74.5)
≥ 4	24 (8.5)	22 (10.2)
Tumor density in portal venous phase
Isodense or hyperdense	29 (10.2)	20 (9.3)
Hypodense	254 (89.8)	196 (90.7)
Tumor necrosis	248 (87.6)	188 (87.0)
Peripancreatic tumor infiltration	140 (49.5)	114 (52.8)
Suspicious metastatic lymph nodes	89 (31.5)	81 (37.5)

Extended: Extended pancreaticoduodenectomy, extended distal pancreatectomy; Standard: Standard pancreaticoduodenectomy, standard distal pancreatectomy, standard total pancreatectomy.

Validation of prognostic stratification based on risk scoring

A simple risk score was created using significant predictors from a multivariable model (Table 2), which categorized patients into risk groups, as: Favorable (risk score < 2), intermediate (risk score 2-4), and poor prognosis (risk score ≥ 5). According to the evaluation results of Reader 1, the recurrence rates were 39.0% in the good prognosis group, 82.1% in the moderate group, and 84.5% in the poor group. In comparison, Reader 2 reported recurrence rates of 50.0%, 79.5%, and 88.9%, respectively, across the same prognostic categories. These findings demonstrated a consistent trend across both raters, with higher prognostic scores associated with an increased risk of recurrence. The risk score showed good discrimination capability with an HR of 1.80 (95%CI: 1.44, 2.25) for Reader 1 and 1.74 (95%CI: 1.34, 2.18) for Reader 2. The evaluation of imaging features by the two radiologists showed good consistency (Kappa value of 0.78).

Table 2 Risk score for factors associated with recurrence-free survival of patients with resectable pancreatic ductal adenocarcinoma.

Parameters	Score
Tumor size in cm
≤ 2	0
2-4	1
≥ 4	2
Tumor density in portal venous phase
Isodense/hyperdense	0
Hypodense	1
Tumor necrosis
No	0
Yes	2
Peripancreatic tumor infiltration
No	0
Yes	1
Suspicious metastatic lymph nodes
No	0
Yes	2

The Kaplan-Meier survival curves (Figure 1) illustrated the time-dependent changes in survival probabilities across the different risk groups. The low-risk group exhibited significantly higher survival probabilities compared to the moderate- and high-risk groups, with the high-risk group showing the poorest survival outcomes. The log-rank test yielded a P value less than 0.001, indicating a statistically significant difference in survival among the three risk groups. These findings suggest that the risk score is an effective predictor of patient prognosis and provides valuable guidance for clinical decision-making. Representative patients are presented in Figures 2, 3 and 4.

Figure 1 Kaplan Meier survival curves for patients stratified by risk group. The low-risk group showed significantly better survival outcomes than the moderate and high-risk groups, whereas the high-risk group exhibited the poorest prognosis. Differences in survival among the groups were statistically significant (log-rank test, P < 0.001), indicating the predictive value of the risk score for patient prognosis. A: Risk score 1 from Reader 1; B: Risk score 2 from Reader 2.

Figure 2 A 61-year-old male with resectable pancreatic ductal adenocarcinoma. Axial contrast-enhanced computed tomography images revealed a 1.7 cm mass in the body of the pancreas (indicated by the arrow), which was hypodense in both. The image scale was approximately 0.55 mm per pixel. A: Arterial phase; B: Portal venous phase. There was no evidence of tumor necrosis, peritumoral invasion, or suspicious metastatic lymph nodes. The patient was given a risk score of 1, placing him in the good prognosis group. Following a standard pancreaticoduodenectomy, the patient survived for 65 months without any tumor recurrence.

Figure 3 Computed tomography images of a 61-year-old female with resectable pancreatic ductal adenocarcinoma. A: Arterial phase; B: Portal venous phase axial contrast-enhanced computed tomography images revealed a 3.8 cm hypodense mass in the tail of the pancreas (indicated by a white arrow) with invasion of the splenic artery (indicated by a orange arrow). The image scale was approximately 0.31 mm per pixel. There was no evidence of tumor necrosis or suspicious metastatic lymph nodes. With a risk score of 3, the patient was placed in the moderate prognosis group. A standard distal pancreatectomy was performed, and the tumor recurred 18 months postoperatively.

Figure 4 Computed tomography images of a 64-year-old male with resectable pancreatic ductal adenocarcinoma. A: Arterial phase; B and C: Portal venous phase computed tomography axial contrast enhancement showed a 3.4 cm hypodense mass in the head of the pancreas with adjacent vascular invasion (white arrow in A), and a suspicious metastatic lymph node adjacent to the pancreatic head with central necrosis (orange arrow in C). The image scale was approximately 0.41 mm per pixel. The patient’s risk score was 5, placing him in the poor prognosis group. A standard pancreaticoduodenectomy was performed, and pathological examination confirmed the diagnosis of poorly differentiated pancreatic ductal adenocarcinoma. The tumor recurred 9 months postoperatively.

DISCUSSION

Increasing evidence has indicated a significant correlation between the morphological characteristics observed on CT scan and the prognosis of PDAC[5-13]. Traditional prognostic models rely heavily on postoperative pathological findings, limiting their clinical utility. In recent years, radiomics-based prognostic assessments have garnered attention[14-16]. Despite the potential of radiomics in preoperative risk assessment, its clinical application faces several challenges including data heterogeneity, insufficient multimodal integration, subjectivity in manual segmentation, and difficulties in biological interpretation and clinical translation[17,18]. Compared to radiomics models, the risk scoring system proposed based on conventional CT imaging features is more clinically applicable and easier to implement, facilitating broader application[19].

The findings of this study support and extend prior research confirming that tumor size determined by preoperative CT imaging is a reliable independent predictor of tumor recurrence[5-13]. Furthermore, this study substantiates the well-established associations among lymph node metastasis, invasion of adjacent structures, and RFS[6,7,11-13,20,21].

Tumor necrosis, which is tissue death within a tumor[22], is intrinsically linked to heightened tumor aggressiveness. Empirical data suggest that diminished enhancement of a tumor during the portal venous phase is a common finding in malignant lesions[23]. Complementing these findings, studies have identified factors strongly linked to the outcome of pancreatic cancer including preoperative levels of carbohydrate antigen 19-9, maximum standardized uptake value, systemic inflammatory response indices, decrements in skeletal muscle mass, and the anatomical distribution of tumor growth[6-8,10,12,16,21,24-27]. The liver is a common and prominent site for tumor metastasis[28-30].

This study had some limitations. First, neoadjuvant therapy is playing an increasingly important role in improving resectability and overall survival[15,31], as it can improve both resection and survival rates, and provide early control of micrometastatic lesions[32,33]. However, the cohort in this study was predominantly composed of individuals who had not received preoperative neoadjuvant chemotherapy. Second, the majority of patients (76%) exhibited recurrence, with a mere 24% remaining recurrence-free, thus introducing an imbalance in the dataset. Finally, this study did not take into account the differences in population epidemiology in different regions. Thus, future studies could consider optimizing the model based on the epidemiological characteristics of local populations (e.g., genetic variation profiles, distribution of chronic diseases) to enhance the applicability of a risk scoring system across diverse patient groups[34].

CONCLUSION

In summary, this study validated the effectiveness of a risk scoring system in predicting RFS in patients with resectable PDAC after surgery. Parameters such as tumor size, tumor density during the portal venous phase, presence of necrosis, pancreatic infiltration, and suspicious metastatic lymph nodes are easily accessible and display a high degree of reproducibility. Thus, the risk scoring system may be introduced to diverse patient populations and serve as a preoperative tool to identify patients at high risk of PDAC and guide clinical decision-making.