Background
Pancreatic cancer is a highly fatal disease with poor prognosis. The 5-year survival rate is only 9%, and the incidence rate is still rising steadily [
1]. Surgical resection is considered to be the only treatment that can be cured. However, only a few patients with pancreatic cancer are suitable for initial resection. Since pancreatic cancer is usually asymptomatic in the early stage, and most patients are diagnosed as advanced stage [
2‐
4]. Some patients can find the disease during physical examination and undergo early resection, but most patients still relapse and die. Therefore, it is very important to find out the risk factors of postoperative patients with pancreatic cancer and to evaluate the survival prognosis.
In recent years, nomogram has been widely used in tumor prediction, so that clinicians can use it to predict the prognosis of patients [
5‐
7]. A recent investigation has elucidated that a comprehensive analysis encompassing variables such as age, race, histological grade, surgical interventions, and chemotherapy among patients afflicted with bone metastases from pancreatic cancer yields a proficient prediction of survival prognosis. The nomogram's C-index, indicative of model performance, exhibited commendable accuracy [
8]. In the study conducted by Wu Mengwei and colleagues, the identification of nine distinctive gene characteristics facilitated the establishment of a prognostic nomogram for the overall survival period in pancreatic cancer. Remarkably, the predictive efficiency surpassed that of the AJCC staging system [
9]. Furthermore, the utility of the nomogram has transcended disciplinary boundaries, proving its superior predictive performance over traditional tumor staging methodologies in diverse domains [
10‐
12]. This superiority can be attributed to the nomogram's holistic consideration of a broader spectrum of influential factors.
Nevertheless, investigations concerning postoperative patients with pancreatic cancer remain scarce. Consequently, there exists a critical need for a personalized prediction model tailored specifically to postoperative patients with pancreatic cancer. This imperative underscores our commitment to constructing models aimed at assessing the prognosis and survival rates of individuals post pancreatic cancer surgery.
Materials and methods
Patient selection
Patients diagnosed with pancreatic cancer between 2004 and 2015 were initially identified from the SEER database, utilizing SEER * Sta 8.4.0.1 (Surveillance, Epidemiology, and End Results Program at cancer.gov). The external validation cohorts, diagnosed with pancreatic cancer between January 2018 and January 2023, were obtained from the Affiliated Hospital of Hunan Academy of Traditional Chinese Medicine. Inclusion criteria were as follows:Patients with pancreatic cancer who underwent surgery. Availability of clear information on survival status and survival time. Exclusion criteria encompassed:Lack of information on age, sex, marital status, AJCC TNM stage, tumor size, radiotherapy, chemotherapy, and liver metastasis. Patients who died within 1 month or were followed up for less than 1 month after the initial diagnosis. Other causes of death or cases where the cause of death was unknown.
Cohort definition and variable recode
The entire cohort was randomly divided into training and internal validation cohorts at a ratio of 7:3. The training cohort was employed for risk factor screening and model establishment, while both the internal and external validation cohorts were utilized to validate the results. From the SEER database, 12 variables were screened, encompassing age (at diagnosis), sex, pathological grade, AJCC TNM stage, radiotherapy and chemotherapy status, presence of liver metastasis, tumor size, marital status, and primary site. These variables were crucial in assessing and understanding the factors influencing postoperative survival in patients with pancreatic cancer.
Statistical analysis
The optimal cut-off values for tumor size and age were determined using X-tile [
13]. Univariate and multivariate Cox regression analyses were applied to calculate the corresponding hazard ratios (HR) and 95% confidence intervals (CI) for the training cohort. Independent risk factors identified through these analyses were then incorporated into the nomogram. To assess the nomogram's discriminative ability, the area under the time-dependent curve (AUC value) was calculated. The effectiveness and calibration of the nomogram were evaluated using a calibration curve. The clinical benefit and utility of the nomogram were assessed through decision curve analysis (DCA) [
14]. X-tile software was utilized to stratify the risk of the nomogram based on total scores. The Kaplan–Meier method compared the risk stratification of the nomogram with the AJCC stage. Statistical significance was set at
P < 0.05. All data analyses were conducted using R software in accordance with relevant guidelines and regulations.
Discussion
Pancreatic cancer stands out as one of the most invasive and fatally aggressive malignancies. Projections indicate that by the year 2030, it is poised to ascend to the position of the second leading cause of cancer-related fatalities. While radical surgery holds the potential for cancer cure [
15], the rates of postoperative recurrence and mortality continue to register high figures [
16,
17]. In light of these challenges, the predictive assessment of survival rates among postoperative cancer patients assumes paramount significance.
Several studies have consistently demonstrated that factors such as advanced age, elevated histological grade, and larger tumor size exhibit a negative correlation with long-term survival outcomes [
8,
18,
19]. In our investigation, the findings underscore a significant disparity in survival rates between patients who underwent surgical treatment and those who did not. Notably, patients with pancreatic cancer who actively pursued surgical resection exhibited markedly enhanced survival probabilities [
20,
21]. This observation aligns with the conclusions drawn by Hester et al., who based on an analysis of the National Cancer Database, established the beneficial impact of surgical resection on the overall survival of pancreatic cancer patients [
22]. Nevertheless, reliance on surgery alone is insufficient for achieving prolonged survival. The median survival time for the majority of patients typically hovers around 8 to 10 months, with frequent tumor relapses [
23,
24]. Our study incorporated patients receiving chemotherapy, encompassing both preoperative neoadjuvant chemotherapy and postoperative adjuvant chemotherapy. Cox regression analysis identified chemotherapy as an independent risk factor for postoperative pancreatic cancer patients, consistent with prior research [
7,
25,
26]. Notably, the median postoperative survival of patients undergoing adjuvant chemotherapy doubled compared to those who did not [
27]. Neoadjuvant chemotherapy emerged as an independent predictor and an enhancer of overall survival for postoperative pancreatic cancer patients [
28,
29], concurrently improving the R0 removal rate [
24,
30]. Consequently, it presents a favorable therapeutic option for both patients and healthcare practitioners. Additionally, age emerged as an independent risk factor for pancreatic cancer patients [
8]. Our study revealed a lower survival rate among patients aged 70 years and older. This age-related discrepancy in survival rates may be associated with compromised immunity and physical deterioration commonly observed in elderly patients [
18,
19,
31].
The validation of predictive models is crucial for determining generalization and avoiding overfitting [
32]. In our investigation, the nomogram exhibited a superior AUC value in comparison to the AJCC staging system, indicative of enhanced discriminative ability. The calibration chart further underscored the robust consistency between the predicted nomogram and the observed 1-year and 2-year cancer-specific survival (CSS), affirming the reliability and repeatability of the established nomogram. Decision Curve Analysis (DCA) analysis reinforced the nomogram's heightened clinical benefits over traditional AJCC staging models. Additionally, nomogram's risk stratification model proficiently categorizes patients into high-risk, medium-risk, and low-risk groups. To our knowledge, this study marks the inaugural utilization of a nomogram for survival prediction, leveraging the SEER database and undergoing external validation, specifically tailored for postoperative cancer patient prognostication. Insights gleaned from our research suggest that characteristics indicative of high-risk status among postoperative cancer patients encompass advanced age, male gender, lower histological grading, larger tumors, and absence of chemotherapy. Crucially, our nomogram surpasses the capabilities and value of the traditional TNM staging system. We contend that meticulously designed nomogram hold the potential to accurately predict the prognosis of each patient, thereby conferring substantial benefits to both clinical practitioners and patients.
This study holds significant clinical importance as nomogram can be employed to assess individualized prognoses in postoperative cancer patients. However, our research is not without limitations. Firstly, being a large-scale retrospective study based on the SEER database, inherent biases associated with retrospective designs cannot be entirely mitigated. Secondly, crucial information related to tumor markers, chemotherapy regimens, and comorbidities is absent from the database, factors known to influence the survival and prognosis of cancer. Lastly, external validation cohorts exclusively comprise the Asian population, with a relatively modest sample size. To validate our research findings, future endeavors should involve prospective clinical trials with expanded sample sizes and diverse ethnic groups. Despite these limitations, our nomogram, rooted in an extensive dataset from the SEER database, offers a robust opportunity to predict cancer-specific survival (CSS) in postoperative patients with pancreatic cancer. This provides valuable support for individualized treatment strategies and more precise clinical decision-making.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.