Prognostic value of a nomogram model for pancreatic cancer incorporating the systemic immune-inflammation and prognostic nutritional indices

Abstract

BACKGROUND

A nomogram model predictive of pancreatic ductal adenocarcinoma (PDAC) outcomes based on the systemic immune inflammation index (SII) and prognostic nutritional index (PNI) does not exist.

AIM

To determine the prognostic value of a nomogram model combining the SII and PNI in patients with PDAC after radical surgery.

METHODS

One hundred sixty-four patients who underwent radical surgery for PDAC between 2017 and 2023 were retrospectively enrolled. A nomogram prediction model for the survival rate after PDAC surgery was established based on the multivariable Cox regression results of the training cohort (n = 129). The model was validated in the external validation cohort (n = 35). The time-dependent receiver operating characteristic (ROC) curve, C-index, calibration curve, and decision curve were used to evaluate model efficacy, and an online prediction tool was developed.

RESULTS

An SII > 335.82, a PNI ≤ 44.45, poorly differentiated PDAC, no postoperative adjuvant chemotherapy, and a carbohydrate antigen 19-9 level > 37 U/mL were independent risk factors for poor prognosis of patients with PDAC after radical surgery. These five factors were included in the nomogram model. The internal validation C-index of the nomogram model was 0.691 (95% confidence interval: 0.626-0.755). The area under the ROC curve values for the 1-year, 2-year, and 3-year postoperative survival rates in the training cohort were 0.684, 0.762, and 0.822, respectively; the area under the ROC curve values for the external validation cohort were 0.772, 0.755, and 0.796, respectively.

CONCLUSION

The nomogram model based on the SII and PNI accurately predicted the survival risk and prognosis in patients with PDAC after surgery, providing a quantitative tool for individualized treatment decisions.

Key Words: Nomogram; Prognostic nutritional index; Pancreatic ductal adenocarcinoma; Prognosis; Systemic immune inflammation index

Core Tip: A nomogram prediction model that was constructed based on the systemic immune inflammation index and prognostic nutrition index accurately assessed the comprehensive preoperative inflammation-immune-nutrition status of patients with pancreatic cancer. The convenience and simplicity of the model, especially when combined with the online calculator, enhanced the predictive potential for the prognosis of patients with pancreatic cancer after surgery. This model is particularly suitable for primary hospitals with limited resources. However, the results of this study need to be verified by prospective samples from multiple centers with large samples.

INTRODUCTION

Pancreatic cancer is a highly aggressive malignant tumor of the digestive tract with the following clinical features: Insidious onset; challenging early-stage diagnosis; and poor prognosis[1]. The 5-year relative survival rate for patients with pancreatic cancer is < 8% in China[2]. Moreover, patients with the same pathologic type and clinical stage have heterogeneity in postoperative survival outcomes due to significant differences in immune status, nutritional level, and systemic inflammatory response[3-5].

Inflammation promotes the occurrence, progression, and metastasis of tumors by regulating the tumor microenvironment[6]. The systemic immune inflammation index (SII) is an inflammation-related index that integrates three hematologic indicators (neutrophils, lymphocytes, and platelets). Existing studies have confirmed that higher SII levels are significantly associated with a poor prognosis in patients with pancreatic cancer[5,7,8]. In addition, low prognostic nutrition index (PNI) levels have been confirmed to be significantly associated with a poor prognosis in patients with pancreatic cancer[7,9,10]. At present, nomogram models based on the SII[11] and PNI[10] have been verified to predict pancreatic cancer survival. However, a prediction nomogram model combining the SII and PNI has not been developed for pancreatic ductal adenocarcinoma (PDAC). Moreover, the mechanism underlying PDAC is unclear.

Therefore, this study aimed to develop a nomogram model based on the SII and PNI for predicting 1-year, 2-year, and 3-year survival rates in patients with PDAC who have undergone radical surgery. The efficacy and clinical significance of the nomogram model were also determined.

MATERIALS AND METHODS

Patients

This study was approved by the Institutional Ethical Committee of Rugao People’s Hospital, approval No. KY202211013 and proceeded in compliance with the Declaration of Helsinki. Each participant signed a written informed consent for treatment and the use and publication of their clinical data and information. The confidentiality agreement has been honored, and private patient information has been protected.

Patients with PDAC who were treated at the Affiliated Hospital of Nantong University from January 2017 to December 2020 were designated as the training cohort. The patients with PDAC who were treated at Rugao People’s Hospital between January 2019 and December 2023 were designated as the external validation cohort. The inclusion criteria were as follows: (1) PDAC diagnosis according to the 2021 Guidelines for the Diagnosis and Treatment of Pancreatic Cancer in China[12] and the postoperative pathologic diagnosis was PDAC; (2) Underwent radical surgery for PDAC; (3) Preoperative examination and intraoperative exploration confirmed no distant tumor metastases; and (4) Complete clinical and follow-up data. The exclusion criteria were as follows: (1) Preoperative radiotherapy or chemotherapy; (2) Coexisting blood, inflammatory, or immune diseases; (3) History of other malignant tumors; (4) Severe postoperative complications resulting in < 30 days survival; and (5) Lack of clinical and follow-up data.

Data collection

Data collection included general clinical information, routine serum biochemical test results, and intraoperative and postoperative details. The general clinical information included age, gender, contact number, and a history of chronic diseases (hypertension, diabetes, chronic bronchitis emphysema, and coronary atherosclerotic heart disease). The following routine serum biochemical test results within 1 week preoperatively were included: Platelet count; lymphocyte count; neutrophil count; hemoglobin concentration; total bilirubin level; aspartate aminotransferase level; alanine aminotransferase level; albumin level; carcinoembryonic antigen level; and carbohydrate antigen 19-9 (CA19-9) level. Intraoperative and postoperative details included the following: Surgical method; intraoperative tumor location; postoperative pathologic report (tumor size, lymph node metastases, degree of tumor degree, surgical margins, vascular tumor thrombus, and presence or absence of vascular and nerve invasion); tumor stage according to the 8^th edition of the American Joint Committee on Cancer Tumor-node-metastasis Staging Manual; postoperative adjuvant chemotherapy; recent postoperative complications; and other indicators. The postoperative short-term complications were defined as any of the following complications that occurred within 1 month after surgery: Gastrointestinal anastomotic fistula; pancreatic fistula; biliary fistula; delayed gastric emptying; postoperative massive hemorrhage of the pancreas; abdominal cavity infection; pulmonary infection; and surgical incision infection.

Calculation of the SII and PNI

The SII and PNI values were calculated based on the biochemical and serum routine data 1 week preoperatively as follows: SII = platelet count (10⁹/L) × neutrophil count (10⁹/L)/Lymphocyte count (10⁹/L)[8]; and PNI = serum albumin concentration (g/L) + lymphocyte count (10⁹/L) × 5[9].

Follow-up data

Follow-up data, including patient mortality, the time of death, and the postoperative adjuvant therapy, were recorded through telephone interviews and outpatient follow-up evaluation. The follow-up period ended on December 31, 2024. The end of the follow-up period or patient death was the endpoint event. Patient survival status during the follow-up period was recorded, and the survival time was recorded monthly.

Statistical analysis

Statistical analysis was performed using SPSS 25.0 and R 4.2.2 software. The mean ± SD was applied to continuous variables with a normal distribution. Continuous variables with a non-normal distribution are represented by the median [lower and upper quartiles (median (P25, P75)]. Counting variables are expressed in terms of frequency and 100% [n (%)]. The optimal SII and PNI cutoff values were determined using the surv-cutpoint function in the R language surv-miner package based on the maximum selection test method. The patients were divided into high and low groups according to the SII and PNI levels. A χ² or Fisher’s exact test was used to analyze the correlation of clinicopathologic characteristics between the training and external validation cohorts. The independent risk factors influencing overall survival (OS) in patients after radical resection of PDAC were determined using the Cox regression model; a P value < 0.05 identified significant factors.

The forestplot, rms, survival receiver operating characteristic (ROC), ggplot2, and other packages in R language were applied to establish the nomogram model of the 1-year, 2-year, and 3-year postoperative survival rates of patients with PDAC, including the independent risk factors with statistical differences in the multivariable Cox regression analysis (SII, PNI, degree of tumor differentiation, postoperative adjuvant chemotherapy, and CA19-9 level). Internal validation was performed using the bootstrap resampling method. The efficacy of the nomogram model was evaluated and verified through a time-dependent ROC curve, C-index, calibration curve, and decision curve analysis (DCA). A web calculator was developed using the “shiny: Web Application Framework for R package.” Results were considered significant if the P value was < 0.05.

RESULTS

The training cohort was initially comprised of 155 patients. However, 3 patients received preoperative radiotherapy or chemotherapy, 3 patients had blood, inflammatory, or immune diseases, 8 patients had other coexisting malignant tumors, 2 patients had a survival time < 30 days due to complications or other reasons, and 10 patients were lost to follow-up after surgery. Therefore, 129 patients were finally included in the training cohort. The external validation cohort initially included 41 patients. However, 1 patient had preoperative radiotherapy or chemotherapy, 1 patient had a blood, inflammatory, or immune disease, 2 patients had coexisting malignant tumors, and 2 patients had incomplete follow-up postoperatively. Therefore, 35 patients were finally included in the external validation cohort. The flowchart illustrating the inclusion and exclusion of patients with PDAC is shown in Figure 1.

Figure 1  Flowchart for inclusion and exclusion of pancreatic ductal adenocarcinoma patients.

Baseline data

The baseline data from patients with PDAC are shown in Table 1. With the exception of gender, surgical methods, and the SII, no significant differences were detected between the training and external validation cohorts.

Table 1 Baseline data.

Clinicopathologic factors	Total number (n = 164)	Training cohort (n = 129)	External validation cohort (n = 35)	P value
Gender				0.022a
Male	89 (54.3)	76 (58.9)	13 (37.1)
Female	75 (45.7)	53 (41.1)	22 (62.9)
Age (years)				0.624
≤ 65	81 (49.4)	65 (50.4)	19 (54.3)
> 65	83 (50.6)	64 (49.6)	16 (45.7)
Underlying diseases				0.341
No	82 (50)	67 (51.9)	15 (42.9)
Yes	82 (50)	62 (48.1)	20 (57.1)
Surgical method				< 0.001a
Endoscopic surgery	32 (19.5)	16 (12.4)	16 (45.7)
Open surgery	132 (80.5)	113 (87.6)	19 (54.3)
Tumor location				0.804
Pancreatic body/tail	58 (35.4)	45 (34.9)	13 (37.1)
Pancreatic head	106 (64.6)	84 (65.1)	22 (62.9)
Tumor maximum diameter (cm)				0.301
≤ 3	81 (49.6)	61 (47.3)	20 (57.1)
> 3	83 (50.6)	68 (52.7)	15 (42.9)
T staging				0.543
T1-T2	120 (73.2)	93 (72.1)	27 (77.1)
T3-T4	42 (26.9)	36 (27.9)	8 (22.9)
N grouping				0.192
N0	100 (61.0)	82 (63.6)	18 (51.4)
N1-N2	64 (39.0)	47 (36.4)	17 (48.6)
TNM staging				0.771
I-II	144 (87.8)	114 (88.4)	30 (85.7)
III	20 (12.2)	15 (11.6)	5 (14.3)
Degree of tumor differentiation				0.871
Moderately or well-differentiated	105 (64.0)	83 (64.3)	22 (62.9)
Poorly differentiated	59 (36.0)	46 (35.7)	13 (37.1)
Surgical margins				0.343
Negative	158 (96.3)	123 (95.3)	35 (100.0)
Positive	6 (3.7)	6 (4.65)	0 (0.0)
Vascular tumor thrombus				0.072
No	81 (49.4)	59 (45.7)	22 (62.9)
Yes	83 (50.6)	70 (54.3)	13 (37.1)
Vascular/nerve invasion				0.381
No	20 (12.2)	14 (10.9)	6 (17.1)
Yes	144 (87.8)	115 (89.1)	29 (82.9)
Lymph node metastasis				0.139
No	102 (62.2)	84 (65.1)	18 (51.4)
Yes	62 (37.8)	45 (34.9)	17 (48.6)
Postoperative adjuvant chemotherapy				0.847
No	110 (67.1)	42 (32.6)	12 (34.3)
Yes	54 (32.9)	87 (67.4)	23 (65.7)
Recent postoperative complications				0.960
No	126 (76.8)	99 (76.7)	27 (77.1)
Yes	38 (23.2)	30 (23.3)	8 (22.9)
CEA (ng/mL)				0.322
≤ 5	123 (75.0)	99 (76.7)	24 (68.6)
> 5	41 (25.0)	30 (23.3)	11 (31.4)
CA19-9 (U/mL)				0.274
≤ 37	40 (24.4)	29 (22.5)	11 (31.4)
> 37	124 (75.6)	100 (77.5)	24 (68.6)
SII				0.011a
≤ 335.82	58 (35.4)	52 (40.3)	6 (17.1)
> 335.82	106 (64.6)	77 (59.7)	29 (82.9)
PNI				0.331
≤ 44.45	68 (41.5)	56 (43.4)	12 (34.3)
> 44.45	96 (58.5)	73 (56.6)	23 (65.7)

^aP < 0.05.

Data are presented as n (%). TNM: Tumor-node-metastasis; CEA: Carcinoembryonic antigen; CA19-9: Carbohydrate antigen 19-9; SII: Systemic immune inflammation index; PNI: Prognostic nutritional index.

Optimal SII and PNI cutoff values

The SII and PNI of the patients in the training cohort were 412.9 (278.0, 686.7) and 45.23 ± 5.06, respectively. The surv-cutpoint function in the R language surv-miner package was used to determine the partitioning threshold and the optimal SII and PNI cutoff values (Figure 2). The optimal SII cutoff value was 335.82. The patients were then divided into high (SII > 335.82) and low SII groups (SII ≤ 335.82). The optimal PNI cutoff value was 44.45. The patients were then divided into high (PNI > 44.45) and low PNI groups (PNI ≤ 44.45).

Figure 2 Calculation for the optimal systemic immune inflammation index and prognostic nutritional index cutoff values. The basic principle of the surv-cutpoint function was based on patient overall survival, and the Kaplan-Meier curve was used to determine the best cutoff value. The survival differences between the two groups were tested by a log-rank test. The survival rates on both sides of a specific point had the greatest difference when the χ² statistic of the log-rank test was the largest. The measured values corresponding to the maximum χ² statistic were the optimal systemic immune inflammation index and prognostic nutritional index cutoff values. SII: Systemic immune inflammation index; PNI: Prognostic nutritional index; grps: Groups.

Univariable Cox regression analyses for the prognosis of patients with PDAC

Univariable Cox regression analyses showed that an SII > 335.82 [hazard ratio (HR) = 1.844, 95% confidence interval (CI): 1.194-2.849; P = 0.006], a PNI ≤ 44.45 (HR = 1.990, 95%CI: 1.305-3.034; P = 0.001), a maximum tumor diameter > 3 cm (HR = 1.816, 95%CI: 1.194-2.762; P = 0.005), T3-T4 staging (HR = 1.568, 95%CI: 1.013-2.427; P = 0.043), a poorly differentiated tumor (HR = 1.607, 95%CI: 1.057-2.442; P = 0.026), no postoperative adjuvant chemotherapy (HR = 0.575, 95%CI: 0.373-0.885; P = 0.012), and a CA19-9 level > 37 U/mL (HR = 2.021, 95%CI: 1.142-3.576; P = 0.016) were significantly related to poor prognosis in patients with PDAC after radical surgery (Table 2).

Table 2 Univariable and multivariable Cox regression analyses for the prognosis of pancreatic ductal adenocarcinoma patients.

Clinicopathologic factors	Univariable analysis		P value	Multivariable analysis		P value
	HR	95%CI		HR	95%CI
Gender
Female	1 (reference)		-	-	-	-
Male	0.955	0.634-1.438	0.824	-	-	-
Age (years)
≤ 65	1 (reference)		-	-	-	-
> 65	1.107	0.736-1.665	0.626	-	-	-
Underlying diseases
No	1 (reference)		-	-	-	-
Yes	0.992	0.661-1.491	0.970	-	-	-
Surgical methods
Endoscopic surgery	1 (reference)		-	-	-	-
Open surgery	1.101	0.570-2.128	0.774	-	-	-
Tumor location
Pancreatic body/tail	1 (reference)		-	-	-	-
Pancreatic head	0.754	0.495-1.147	0.187	-	-	-
Tumor maximum diameter (cm)
≤ 3	1 (reference)		-	-	-	-
> 3	1.816	1.194-2.762	0.005a	1.592	0.962-2.633	0.070
T staging
T1-T2	1 (reference)	-	-	-	-	-
T3-T4	1.568	1.013-2.427	0.043a	1.061	0.624-1.802	0.827
N grouping
N0	1 (reference)	-	-	-	-	-
N1-N2	1.325	0.877-2.002	0.181	-		-
TNM staging
I-II	1 (reference)	-	-	-	-	-
III	1.129	0.601-2.120	0.707	-	-	-
Degree of tumor differentiation
Moderately or well-differentiated	1 (reference)	-	-	-	-	-
Poorly differentiated	1.607	1.057-2.442	0.026a	1.614	1.036-2.515	0.034a
Surgical margins
Negative	1 (reference)	-	-	-	-	-
Positive	1.033	0.451-2.371	0.938	-	-	-
Vascular tumor thrombus
No	1 (reference)	-	-	-	-	-
Yes	1.058	0.700-1.599	0.790	-	-	-
Vascular/nerve invasion
No	1 (reference)	-	-	-	-	-
Yes	1.715	0.859-3.426	0.126	-	-	-
Lymph node metastasis
No	1 (reference)	-	-	-	-	-
Yes	1.251	0.825-1.897	0.292	-	-	-
Postoperative adjuvant chemotherapy
No	1 (reference)	-	-	-	-	-
Yes	0.575	0.373-0.885	0.012a	0.561	0.360-0.874	0.011a
Recent postoperative complications
No	1 (reference)	-	-	-	-	-
Yes	1.218	0.763-1.945	0.408	-	-	-
CEA (ng/mL)
≤ 5	1 (reference)	-	-	-	-	-
> 5	0.917	0.562-1.495	0.728	-	-	-
CA19-9 (U/mL)
≤ 37	1 (reference)	-	-	-	-	-
> 37	2.021	1.142-3.576	0.016a	1.844	1.021-3.330	0.043a
SII
≤ 335.82	1 (reference)	-	-	-	-	-
> 335.82	1.844	1.194-2.849	0.006a	1.891	1.217-2.938	0.005a
PNI
≤ 44.45	1 (reference)	-	-	-	-	-
> 44.45	1.990	1.305-3.034	0.001a	2.149	1.381-3.343	0.001a

^aP < 0.05.

HR: Hazard ratio; CI: Confidence interval; TNM: Tumor-node-metastasis; SII: Systemic immune inflammation index; PNI: Prognostic nutritional index; CEA: Carcinoembryonic antigen; CA19-9: Carbohydrate antigen 19-9.

Multivariable Cox regression analyses for the prognosis of patients with PDAC

The factors with statistical differences (P < 0.05) based on univariable Cox analysis were further analyzed by multivariable Cox analysis. The multivariable Cox regression analysis showed that an SII >335.82 (HR = 1.891, 95%CI: 1.217-2.938; P = 0.005), a PNI ≤ 44.45 (HR = 2.149, 95%CI: 1.381-3.34;, P = 0.001), a poorly differentiated tumor (HR = 1.614, 95%CI: 1.036-2.515; P = 0.034), no postoperative adjuvant chemotherapy (HR = 0.561, 95%CI: 0.360-0.874; P < 0.011), and a CA19-9 level > 37 U/mL (HR = 1.844, 95%CI: 1.021-3.330; P = 0.043) were independent risk factors for the poor prognosis of patients with PDAC (Table 2).

Establishment of a nomogram model for predicting the 1-year, 2-year, and 3-year postoperative survival rates among patients with PDAC

The predictive nomogram model was established using the rms package in R language software, including the independent risk factors with statistical differences in the multivariable Cox regression analysis (SII, PNI, degree of tumor differentiation, postoperative adjuvant chemotherapy, and CA19-9). The nomogram model for predicting the 1-year, 2-year, and 3-year postoperative survival rates among patients with PDAC is shown in Figure 3. The scores of each variable were summed to calculate the total score of the nomogram model. A vertical line was drawn downward from the total points axis, which intersected the 1-year, 2-year, and 3-year postoperative survival rate axes. This intersection point corresponded to the predicted probability of the postoperative survival rate.

Figure 3 Nomogram model for predicting the 1-year, 2-year, and 3-year postoperative survival rates among patients with pancreatic ductal adenocarcinoma. The assigned scores of relevant indicators were calculated through R language based on the regression coefficients of each variable. Usually the indicator with the largest absolute value of the regression coefficient was set at 100 points, and the scores of other indicators were converted based on the relative relationship between the regression coefficients and the benchmark indicator. Prognostic nutritional index ≤ 44.45 was 100 points; systemic immune inflammation index > 335.82 was 85 points; poorly differentiated tumor was 74 points; no postoperative adjuvant chemotherapy was 81 points; and the carbohydrate antigen 19-9 level > 37 U/mL was 92 points. The calculated total score was matched with the 1-year and 3-year survival rates. The higher the total score, the lower the postoperative survival rate. CA19-9: Carbohydrate antigen 19-9; SII: Systemic immune inflammation index; PNI: Prognostic nutritional index; OS: Overall survival.

Discrimination evaluation and internal model validation

The bootstrap resampling method verified the model in the internal dataset using the survival ROC and ggplot2 packages in the R language software. The calculated C-index was 0.691 (95%CI: 0.626-0.755), indicating that the prediction model had good discrimination. The ROC curves of the nomogram prediction model in the training and external validation cohorts 1 year, 2 years, and 3 years after surgery were plotted and the area under the ROC curve (AUC) performance were analyzed. The results suggested that AUC values of the 1-year, 2-year, and 3-year postoperative survival rates in the training cohort were 0.684, 0.762, and 0.822, respectively; the 1-year, 2-year, and 3-year postoperative survival rates in the external validation cohort were 0.772, 0.755, and 0.796, respectively (Figure 4A and B). Time-dependent C-index analysis revealed that the nomogram model had better prognostic accuracy in predicting the clinical outcomes of OS compared with any other single prognostic marker (Figure 4C and D). The nomogram had a higher time-dependent C-index for the OS of patients with pancreatic cancer compared with the degree of tumor differentiation, chemotherapy, CA19-9 level, SII, and PNI.

Figure 4 Discrimination evaluation and internal model validation. A and B: Receiver operating characteristic curves of the nomogram prediction model in the training and external validation cohorts 1 year, 2 years, and 3 years after surgery; C and D: Time-dependent C-index analysis. AUC: Area under the receiver operating characteristic curve; CA19-9: Carbohydrate antigen 19-9; SII: Systemic immune inflammation index; PNI: Prognostic nutritional index.

Consistency analysis

The calibration curves of the 1-year, 2-year, and 3-year survival rates of patients with pancreatic cancer in the training (Figure 5A) and external validation cohorts (Figure 5B) were plotted using the rms package of the R language software. The horizontal axis represents the predicted probability of the model, and the vertical axis represents the actual probability. The calibration curve shows a good degree of agreement between the predicted probability of the nomogram and the actual probability. The predicted curve had a good fit with the ideal curve, suggesting that the prediction model had good consistency.

Figure 5 Calibration curves of the nomogram prediction model in the training and external validation cohorts for predicting overall survival 1 year, 2 years, and 3 years after surgery. The horizontal axis represents the predicted probability of the model, and the vertical axis represents the actual probability. The higher the overlap between the predicted curve and the ideal curve, the better the consistency of the predicted model. A: The calibration curves of the nomogram prediction model in the training cohorts; B: The calibration curves of the nomogram prediction model in the external validation cohorts. OS: Overall survival.

DCA curve

The DCA curves of the training cohort (Figure 6A-C) and the external validation cohort (Figure 6D-F) showed that the nomogram provided a moderate additional net benefit in terms of survival probabilities 1 year, 2 years, and 3 years after surgery. The DCA curve indicated that the model has potential clinical practicability.

Figure 6 Decision curves of the nomogram prediction model in the training and external validation cohorts for predicting overall survival 1 year, 2 years, and 3 years after surgery. The blue line represents the net benefit of the intervention strategy for all patients. The black line indicates the net benefit without any intervention strategy. The vertical axis represents the overall net benefit. A: The decision curves of the nomogram prediction model in the training cohorts for predicting overall survival 1 year after surgery; B: The decision curves of the nomogram prediction model in the training cohorts for predicting overall survival 2 years after surgery; C: The decision curves of the nomogram prediction model in the training cohorts for predicting overall survival 3 years after surgery; D: The decision curves of the nomogram prediction model in the external validation cohorts for predicting overall survival 1 year after surgery; E: The decision curves of the nomogram prediction model in the external validation cohorts for predicting overall survival 2 years after surgery; F: The decision curves of the nomogram prediction model in the external validation cohorts for predicting overall survival 3 years after surgery.

Web calculator

A web calculator was used to further enhance the practicality of the nomogram prediction model. Users can visit https://tj190019.shinyapps.io/DynNomapp/ directly to implement the dynamic model of the nomogram in Figure 7 and then input the clinical pathologic characteristics. The network server will automatically generate the survival probability curve and 95%CI. The website easily predicts patient OS.

Figure 7  An example of using the online prediction website.

DISCUSSION

A multidimensional integrated nomogram model, including inflammation, immunity, and nutrition, was developed. The model, which consisted of SII and PNI indicators, was confirmed to have good predictive value for the postoperative survival of patients with PDAC.

Most studies have used the ROC curve method to determine the best SII and PNI cutoff values for postoperative survival of patients with PDAC, but the ROC curve method has limitations. Specifically, the ROC curve method only considers disease outcomes and ignores survival time and missing data. The current study determined the cutoff value using the maximum selection test method with the Kaplan-Meier curve and the R language surv-miner package. This method fully incorporates survival time and is more in line with the characteristics of cancer prognosis research than the ROC curve method.

Patients with an elevated SII before surgery usually have an increased platelet count, increased neutrophil count, or decreased lymphocyte count, suggesting the coexistence of inflammation and weakened immunity[8]. Platelets can disguise tumor cells[13-15], and low lymphocyte counts can impair the immune response[16]. Inflammation and malnutrition can exacerbate each other, creating a two-way vicious cycle[11,17-20]. Chronic inflammatory conditions can induce metabolic disorders and increase the risk of malnutrition[11,17]. In contrast, malnutrition can weaken immune function and further exacerbate the inflammatory process[18]. Albumin helps prevent platelet function and thrombosis[19] while inflammatory factors, like interleukin-1 and interleukin-6, promote tumor growth[20]. Therefore, the PNI integrates serum albumin levels and peripheral blood lymphocyte counts, effectively identifying patients at high risk of tumor-associated cachexia[7,9,10]. The serum CA19-9 level is the most frequently used tumor marker to assess PDAC[21]. The degree of tumor differentiation is also an important indicator for evaluating prognosis in patients with PDAC[22]. Adjuvant chemotherapy has also effectively improved the survival of patients with pancreatic cancer[23]. Therefore, these factors warrant inclusion in the predictive model.

The current study fully referred to the results of previous studies and developed the nomogram model that consisted of the degree of tumor differentiation, the CA19-9 level, chemotherapy status, the SII, and the PNI. In the current study the AUC values of the 1-year, 2-year, and 3-year survival rates after surgery in the training cohort were 0.684, 0.762, and 0.822, respectively. The AUC values of the 1-year, 2-year, and 3-year survival rates after surgery in the external validation cohort were 0.772, 0.755, and 0.796, respectively. Furthermore, the calibration curve analysis confirmed that there was good consistency between the predicted probability of the model and the actual observed probability. These results indicate that the nomogram model has good discrimination value.

Xu et al[24] developed a nomogram model to predict survival of patients with PDAC using magnetic resonance imaging and clinical data with a predictive efficacy (C-index) of 0.78. Huang et al[25] developed a model using the microRNA-24 level with a predictive efficacy of 0.82. However, due to the high cost and professional skills required for these models involving magnetic resonance imaging radiomics[24] and RNA sequencing[25], the clinical applicability is limited. The model developed in the current study was based on routine hematologic indicators and can be directly obtained from routine blood and biochemical tests. The core advantage of this model lies in the fact that the hematologic indicators are noninvasive, easy to obtain, and low cost. The AUC results showed good predictive value of the nomogram model, and the DCA showed significant clinical benefits. Combined with the developed online prognostic risk calculator, the visualization and convenience of prognostic assessment have been achieved. Therefore, this model may have greater application potential in resource-limited environments, such as grassroots hospitals.

The previous PNI model had AUC values of 0.826, 0.798, and 0.846 for predicting 1-year, 3-year, and 5-year OS[10]. Sun et al[11] reported an AUC of 0.689 for SII. The current study combined the SII and PNI to predict OS in PDAC, unlike previous reports that used the PNI[10] or SII[11]. The AUC in the training cohort of the current study was lower than the AUC reported by Yang et al[10] but was higher than the AUC reported by Sun et al[11]. Moreover, the current study validated the prediction value in the external cohort and achieved moderately good AUC values (0.772, 0.755, and 0.796 for 1-year, 2-year, and 3-year, respectively, survival rates).

This study had some limitations that should be acknowledged. First, this study focused solely on the patients with PDAC after radical resection and did not explore the prognostic predictive value of the SII and PNI for patients with other types of pancreatic cancer, including unresectable pancreatic cancer. Second, there were regional and population biases, and the sample size of the external validation cohort in this study was relatively small. Third, existing studies show variability in the optimal SII and PNI cutoff values. This heterogeneity may stem from differences in study populations, detection methods, or statistical models, potentially limiting the clinical application and promotion of research outcomes. Therefore, further multicenter, large-sample, and prospective studies need to be conducted to determine the reference range for the optimal SII and PNI cutoff values, thereby providing unified and reliable guidance for clinical work. Finally, the postoperative adjuvant therapy details were not analyzed in the current study due to the retrospective design. Therefore, a stratified analysis of postoperative adjuvant therapy details are recommended to further refine the study results.

CONCLUSION

The nomogram prediction model presented herein was based on the SII and PNI and accurately assessed the comprehensive preoperative inflammation-immune-nutrition status of patients with pancreatic cancer, demonstrating significant postoperative prognostic significance for patients with PDAC. Due to the convenience and simplicity of this model, the nomogram is more likely to be applicable in primary hospitals with limited resources. The results of this study need to be verified by prospective samples from multiple centers with large samples given the population limitations of this study and the failure to standardize key parameters.