Development and validation of a nomogram for overall survival in pancreatic solid pseudopapillary neoplasm: a population-based study

Abstract

Solid pseudopapillary neoplasm (SPN) of the pancreas is an uncommon tumor, leading to the lack of a prognostic prediction model. This study aimed to develop and validate a nomogram for predicting overall survival (OS) in SPN patients. Data from patients diagnosed with SPN between 2000 and 2018 were obtained from the Surveillance, Epidemiology, and End Results (SEER) database and randomly split into training and validation cohorts (7:3). Independent prognostic factors for OS were identified by Cox regression analysis. A nomogram was constructed and internally validated using the concordance index (C-index), area under the curve (AUC), calibration curves, and decision curve analysis (DCA). External validation was performed using our institutional data from SPN patients between January 2008 and January 2026. The results showed that the total positive lymph nodes, age, surgery type, and SEER summary stage were all independent prognostic factors for OS chosen to develop a nomogram for SPN. C-indices of 0.927, 0.93, and 0.965 were found in the training, internal validation, and external verification cohorts respectively. The nomogram’s discriminative ability was proved by the AUC value (> 0.7), and the calibration curve showed alignment between the nomogram’s prediction and actual survival. Finally, the outcomes of DCA further demonstrated the nomogram’s clinical effectiveness. In conclusion, A nomogram was built and validated to help doctors to determine the prognosis of patients with SPN and tailor treatment options.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-47722-0.

Introduction

Solid pseudopapillary neoplasm (SPN) of the pancreas is an uncommon low-grade malignant tumor with uncertain tissue origin, comprising approximately 0.9%–2.7% of all exocrine pancreatic neoplasms and 5% of cystic pancreatic tumors^1,2. SPN mainly affects young women and accounts for about 1%–2% of all pancreatic tumors^3,4. Large series have shown that over 80%–90% of SPN patients are female, most commonly in the second to fourth decades of life, suggesting a role for sex hormones in SPN tumorigenesis^5–7. Frequent expression of progesterone receptors in SPN tumor cells provides a possible biological explanation for this sex predilection⁸. Some research has demonstrated high expression of estrogen receptors (ER) in SPN tumor tissues and the proliferative effect of estrogen in vitro^7,9, providing a potential therapeutic strategy for SPN through selective ER modulators. Additionally, multiple studies have reported that 79% to 100% of SPN patients have progesterone receptors^8,10,11. The frequent expression of progesterone receptors in SPN tumor cells also provides a possible biological explanation for this gender bias.

According to earlier studies, the majority of SPN patients show no specific clinical symptoms¹². Some patients are asymptomatic or only develop abdominal pain¹³. SPN tumors are soft and rarely obstruct the bile or pancreatic ducts¹⁴. Jaundice is rare even when the tumor is located in the head of the pancreas^15,16. Moreover, SPN presents with unique histological changes, such as homogeneous epithelioid cells with poor adhesion, characterized by varied proportions of solid and pseudopapillary structures with hemorrhage and necrosis, and cystic features^17,18. The histological morphology of SPN overlaps with that of pancreatic neuroendocrine tumor and pancreatic acinar cell carcinoma, making it easy to be misdiagnosed clinically^19–21.Thus, abdominal ultrasound, computed tomography (CT), magnetic resonance imaging (MRI), and endoscopic ultrasound play vital roles in diagnosing the disease^22,23. SPN is usually found during imaging examination and is typically large at the time of diagnosis²⁴. SPN lack alterations in common driver genes such as KRAS, TP53, and SMAD4, which are frequently mutated in ductal adenocarcinoma, and exhibit a relatively low incidence of chromosomal abnormalities^25–27. In contrast, point mutations in exon 3 of the CTNNB1 gene—a critical component of the Wnt/β-catenin signaling pathway—are present in over 90% of SPN cases²⁸. Subsequent studies have confirmed that CTNNB1 mutation acts as the principal oncogenic driver in SPN tumorigenesis^29,30. More recently, activation of genes associated with the Hedgehog and androgen receptor (AR) signaling pathways, as well as epithelial-mesenchymal transition (EMT)-related genes, has also been reported in SPN. To date, 17 microRNAs closely linked to the upregulation of Wnt/β-catenin, Hedgehog, and AR pathways, along with EMT genes, have been identified³¹. Whole-exome sequencing of metastatic SPN further revealed that, in addition to the activating CTNNB1 mutation, inactivating mutations in epigenetic regulators such as KDM6A, TET1, and BAP1 are consistently present in both primary and metastatic lesions³², suggesting their potential role in SPN dissemination. Collectively, these findings underscore the complex genetic landscape of SPN. As a result of the advancement and widespread use of CT as well as other imaging technologies, the detection rate of asymptomatic SPN patients has significantly increased³³. Surgery is the mainstay of treatment for SPN, with the surgical approach tailored to tumor size and location. Options range from parenchyma-preserving resections (PPR), including enucleation (EN), central pancreatectomy (CP), and duodenum-preserving cephalic pancreatectomy (DPCP), to more extensive procedures such as spleen-preserving distal pancreatectomy (DP), distal pancreatosplenectomy (DPS), and pancreatoduodenectomy (PD)^34,35. Given its low-grade malignant potential, PPR is considered an oncologically safe option³⁶. Recurrence occurs in 2–10% of cases and is associated with male sex, positive resection margins, lymph node involvement, perineural or vascular invasion, adjacent structure infiltration, high proliferative index, and undifferentiated carcinoma^37–39. Moreover, the mortality rate was low at 0.2%⁴⁰. However, most studies on SPN to date have been case-based reports, with minimal data available related to diagnosis, malignancy potential, and optimal surgical strategy of SPN^41–43. Generally, early diagnosis is difficult due to the absence of specific clinical signs and the lack of understanding of SPN⁴⁴. Based on this dilemma in the current clinical study of SPN, the aim of the present research is to retrospectively examine the clinicopathological data of SPN patients in the SEER database in order to find independent prognostic factors. Furthermore, although SPN is generally associated with a favorable prognosis, its clinical behavior is heterogeneous. A subset of patients experienced aggressive disease, including invasion, metastasis, or recurrence after resection^45,46, making it difficult to summarize their prognosis with a single survival rate. In clinical practice, a nomogram is an intuitive tool that integrates multiple prognostic factors into an individualized survival prediction⁴⁷. Clinicians can integrate relevant key clinical pathological variables of the patients and read their scores in the corresponding nomogram for each variable. Through this specific scoring method tailored for each individual, doctors can obtain the patient-specific overall survival probability, identify high-risk patients with potential aggressive behavior, thereby supporting risk stratification, surgical strategy formulation, and postoperative follow-up management^48,49. SPN is a rare tumor and currently lacks mature and unified prognostic assessment tools⁴⁹. Therefore, in this article, we further construct a quantitative individualized prognostic prediction model - a nomogram based on integrating multiple clinical pathological factors of SPN patients, in order to provide important basis for precise prediction of survival outcomes and formulation of individualized decisions. On this basis, to evaluate the clinical utility of our nomogram beyond conventional discrimination metrics, we performed Decision Curve Analysis (DCA). DCA quantifies the net benefit of using a prediction model across a range of threshold probabilities—the minimum predicted risk at which a patient or clinician would opt for intervention⁵⁰. Unlike the area under the ROC curve (AUC), which measures only rank-order discrimination, DCA incorporates the relative harms of false positives and false negatives, thereby addressing whether using the model leads to better clinical decisions. In addition, we externally validated our findings in a cohort of SPN patients treated in Affiliated Hospital of Southwest Medical University. We believe that the nomogram will aid clinicians in assessing the prognosis and identifying appropriate therapy options for patients with SPN.

Results

Characteristics of the patients and disease

In accordance with the inclusion and exclusion criteria, 341 patients were included in the study (Fig. 1). Then they were classified into two cohorts: a training cohort (n = 238) and a validation cohort (n = 103). The median survival time of the entire population was 36 (interquartile range [IQR]: 12.5–80) months, while the training cohort’s and the internal validation cohort’s median survival times were 35 (IQR: 11.75–77) months and 41 (IQR: 15–98) months, respectively. More than half of the patients were white (69.2%), female (86.8%), and single/unmarried (52.8%). For clinical treatment, 153 (44.9%) of all SPN patients underwent local excision of the tumor or partial pancreatectomy. The pancreatic tail (40.5%) was the most common site for SPN tumor, followed by the head (26.7%) and body of the pancreas (16.4%). Moreover, based on the SEER summary stage, over half of the patients (64.3%) with SPN were diagnosed at the localized stage. The demographic and clinical features of the patients with SPN are compiled in Table 1. There was no significant difference between the training and validation cohorts (P > 0.05) according to the results of the Chi-square test.

Table 1.

Baseline characteristics of SPN patients in the training and internal validation cohorts.

Variables	Whole population [cases (%)]	Training cohort [cases (%)]	Validation cohort [cases (%)]	P value
Total	341	238	103
Age				0.147
< 40	213 (62.5)	153 (64.3)	60 (58.2)
40–59	103 (30.2)	65 (27.3)	38 (36.9)
≥ 60	25 (7.3)	20 (8.4)	5 (4.9)
Sex				0.887
Male	45 (13.2)	31 (13.0)	14 (13.6)
Female	296 (86.8)	207 (87.0)	89 (86.4)
Year of diagnosis				0.587
2000–2006	45 (13.2)	32 (13.4)	13 (12.6)
2007–2012	71 (20.8)	46 (19.3)	25 (24.3)
2013–2018	225 (66.0)	160 (67.3)	65 (63.1)
Race				0.634
White	236 (69.2)	167 (70.2)	69 (67.0)
Black	61 (17.9)	43 (18.1)	18 (17.5)
Asian/Alaska Indian	44 (12.9)	28 (11.7)	16 (15.5)
Marital status				0.429
Single/Unmarried	180 (52.8)	130 (54.6)	50 (48.6)
Married	128 (37.5)	84 (35.3)	44 (42.7)
Divorced/Separated/Widowed	33 (9.7)	24 (10.1)	9 (8.7)
Tumor location				0.579
Head of pancreas	91 (26.7)	62 (26.1)	29 (28.2)
Body of pancreas	56 (16.4)	37 (15.5)	19 (18.4)
Tail of pancreas	138 (40.5)	102 (42.9)	36 (35.0)
Others*	56 (16.4)	37 (15.5)	19 (18.4)
Tumor size (mm)				0.412
≤ 60	207 (60.7)	139 (58.4)	68 (66.0)
> 60	116 (34.0)	86 (36.1)	30 (29.1)
Not specified	18 (5.3)	13 (5.5)	5 (4.9)
Surgery type				0.313
0	39 (11.4)	27 (11.3)	12 (11.7)
1	153 (44.9)	104 (43.7)	49 (47.6)
2	98 (28.7)	75 (31.5)	23 (22.3)
3	51 (15.0)	32 (13.5)	19 (18.4)
Lymph nodes positive				0.679
0 LN+	238 (69.8)	165 (69.3)	73 (70.9)
1–4 LN+	15 (4.4)	12 (5.1)	3 (2.9)
Not specified	88 (25.8)	61 (25.6)	27 (26.2)
Regional LN dissected				0.414
1–3 Reg	44 (12.9)	27 (11.3)	17 (16.5)
4 or more Reg	200 (58.7)	143 (60.1)	57 (55.3)
None	97 (28.4)	68 (28.6)	29 (28.2)
SEER summary stage				0.876
Localized	220 (64.5)	153 (64.3)	67 (65.0)
Regional	80 (23.5)	58 (24.4)	22 (21.4)
Distant	28 (8.2)	18 (7.5)	10 (9.7)
Unstaged	13 (3.8)	9 (3.8)	4 (3.9)

*Others include Islets of Langerhans, other specified parts of pancreas and overlapping lesion of pancreas.

Surgery type: 0: No surgery of primary site; 1: Local excision of tumor; Partial pancreatectomy; 2: Local or partial pancreatectomy and duodenectomy with or without distal/partial gastrectomy; 3: Other surgical approaches.

Identification of prognostic factors of patients with SPN in the training cohort

Eleven variables were included in the univariate Cox analysis for the identification of OS-related variables. A total of seven variables (age, marital status, diagnosis year, tumor size, surgery type, number of positive regional lymph nodes, SEER summary stage) were remarkably linked to OS in the univariate regression analysis. Following that, the multivariate Cox analysis took into account all seven variables that conformed to the analysis (Tables 2 and 3). Age, number of positive regional lymph nodes, surgery type, and SEER summary stage were all identified as independent prognostic factors with P-values < 0.05 .

Table 2.

Univariate and multivariate Cox regression analysis of SPN patients based on clinicopathological characteristics in the training cohort.

Variable	Univariate analysis
	HR	95% CI	P
Age
< 40	Reference
40–59	4.50	1.38–14.68	0.013
≥ 60	7.69	1.91–31	0.004
Sex
Male	Reference
Female	0.41	0.13–1.28	0.124
Year of diagnosis
2000–2006	Reference
2007–2012	0.59	0.18–1.86	0.364
2013–2018	0.20	0.05–0.88	0.033
Race
White	Reference
Black	0.99	0.28–3.52	0.988
Asian/Alaska Indian	1.04	0.23–4.7	0.962
Marital status
Single/Unmarried	Reference
Married	1.62	0.49–5.33	0.424
Divorced/Separated/Widowed	4.69	1.41–15.57	0.012
Tumor location
Head of pancreas	Reference
Body of pancreas	0.28	0.03–2.36	0.244
Tail of pancreas	0.42	0.13–1.32	0.136
Others*	1.07	0.3–3.8	0.917
Tumor size (mm)
≤ 60	Reference
> 60	1.5	0.5–4.49	0.472
Not specified	3.87	1.11–13.57	0.034
Surgery type
0	Reference
1	0.09	0.02–0.34	< 0.001
2	0.19	0.06–0.65	0.008
3	0.16	0.03–0.76	0.021
Lymph nodes positive
0 LN+	Reference
1–4 LN+	11.57	2.56–52.3	0.001
Not specified	5.42	1.68–17.5	0.005
Regional LN dissected
1–3 Reg	Reference
4 or more Reg	0.63	0.12–3.3	0.588
None	2.08	0.46–9.5	0.345
SEER summary stage
Localized	Reference
Regional	1.8	0.4–8.05	0.442
Distant	18.1	5.55–59.05	< 0.001
Unstaged	2.26	0.24–20.96	0.471

*Others include Islets of Langerhans, other specified parts of pancreas and overlapping lesion of pancreas.

Operation category: 0: No surgery of primary site; 1: Local excision of tumor; Partial pancreatectomy; 2: Local or partial pancreatectomy and duodenectomy with or without distal/partial gastrectomy; 3: Other surgical approaches.

Table 3.

Multivariate Cox regression analysis of SPN patients based on clinicopathological characteristics in the training cohort.

Variable	Multivariate analysis
	HR	95% CI	P
Age
< 40	Reference
40–59	5.45	0.99–29.98	0.051
≥ 60	21.21	1.85-243.66	0.014
Sex
Male	-	-	-
Female	-	-	-
Year of diagnosis
2000–2006	Reference
2007–2012	0.34	0.06–1.86	0.211
2013–2018	0.17	0.02–1.33	0.092
Race
White	-	-	-
Black	-	-	-
Asian/Alaska Indian	-	-	-
Marital status
Single/Unmarried	Reference
Married	0.98	0.25–3.94	0.980
Divorced/Separated/Widowed	1.20	0.20–7.03	0.840
Tumor location
Head of pancreas	-	-	-
Body of pancreas	-	-	-
Tail of pancreas	-	-	-
Others*	-	-	-
Tumor size (mm)
≤ 60	Reference
> 60	3.81	0.82–17.64	0.088
Not specified	0.91	0.11–7.70	0.930
Surgery type
0	Reference
1	0.01	0–0.25	0.005
2	0.01	0–0.18	0.002
3	0.01	0–0.47	0.013
Lymph nodes positive
0 LN+	Reference
1–4 LN+	40.16	2.31-696.93	0.011
Not specified	0.62	0.07–5.46	0.665
Regional LN dissected
1–3 Reg	-	-	-
4 or more Reg	-	-	-
None	-	-	-
SEER summary stage
Localized	Reference
Regional	0.43	0.04–4.41	0.481
Distant	8.31	1.80-38.34	0.007
Unstaged	0.39	0.02–1.33	0.533

*Others include Islets of Langerhans, other specified parts of pancreas and overlapping lesion of pancreas.

Operation category: 0: No surgery of primary site; 1: Local excision of tumor; Partial pancreatectomy; 2: Local or partial pancreatectomy and duodenectomy with or without distal/partial gastrectomy; 3: Other surgical approaches.

Nomogram construction

With the aid of the evaluated variables, we then constructed a nomogram for SPN. The nomogram was found to be accurate in predicting the 5-, 6-, and 7-year OS (Fig. 2). To gain a score for each item, the individual sub-variables of each independent prognostic factor were mapped to a score scale on the rote. The higher total score led to a worse prognosis for patients with SPN. The nomogram can be used to personalize OS prediction and clarify the outcomes of the prediction model based on information provided by the patients. The created nomogram makes patient evaluation easier while also enhancing the prediction accuracy and efficiency. Based on the linear predictor derived from the nomogram, we used surv_cutpoint analysis to identify an optimal cut‑off value of -0.138, stratifying patients into low‑risk (n = 140) and high‑risk (n = 98) groups. The two groups showed a highly significant difference in overall survival (log‑rank p = 0.00085 Supplementary Fig. 1), with 5‑year overall survival rates of 98.1% and 87.4%, respectively. High‑risk status was associated with a significantly increased risk of death (HR = 8.34, 95% CI 1.89–36.77, p = 0.005).

Fig. 2.

Nomogram to predict 5-, 6-, and 7-year OS in patients developing SPN. Surgery type: 0: no surgery of the primary site, 1: local excision ofthe tumor or partial pancreatectomy, 2: local or partial pancreatectomy and duodenectomy with or without distal/partial gastrectomy; and 3. othersurgical methods.

Nomogram validation

Internal validation

The training and validation cohorts both verified the model. The training cohort had higher C-index (0.927; 95% confidence interval [CI] = 0.841–1.013) than the SEER summary stage (0.812; 95% CI = 0.675–0.949). The C-index value of the nomogram in the validation cohort was 0.93 (95% CI = 0.877–0.983), which was greater than that of the SEER summary stage (0.692; 95% CI = 0.472–0.912). Furthermore, the nomogram’s AUC values were greater than those of the SEER summary stage in the training cohort (5-year AUC: 0.839 vs. 0.669, 6-year AUC: 0.863 vs. 0.690, 7-year AUC: 0.885 vs. 0.727) (Fig. 3A). The same trend was observed in the validation cohort. The nomogram’s AUC values were higher than those of the SEER summary stage (5-year AUC: 0.948 vs. 0731, 6-year AUC: 0.957 vs. 0.598, 7-year AUC: 0.853 vs. 0.539) (Fig. 3B). Moreover, the calibration curves of the nomogram showed a greater consistency between the predicted and observed survival probabilities in both the training cohort (Fig. 4A) and the validation cohort (Fig. 4B). Finally, the findings of the DCA analysis revealed that the nomogram could serve as an effective clinical tool in clinical settings (Fig. 5). This indicates that within this risk spectrum, using the nomogram to guide surgery type would improve clinical outcomes by increasing appropriate interventions while reducing unnecessary procedures.

Fig. 3.

The area under the curve (AUC) of the nomogram and the SEER summary stage in predicting prognosis in the training and validation cohorts. (A) AUC curves of the nomogram and the SEER summary stage in predicting prognosis at 5, 6, and 7 years in the training cohort. (B) AUC curves of thenomogram and the SEER summary stage in predicting prognosis at 5, 6, and 7 years in the validation cohort.

Fig. 4.

(A) Calibration curve of the nomogram over 5, 6, and 7 years in the training cohort. (B) Calibration curve of the nomogram at 5, 6, and 7years in the validation cohort.

Fig. 5.

(A) Decision curve analysis (DCA) curves of the nomogram at 5, 6, and 7 years in the training cohorts. (B) DCA curves of thenomogram at 5, 6, and 7 years in the validation cohorts.

External validation

External validation was performed on the constructed nomogram. Twenty-six patients with SPN in Affiliated Hospital of Southwest Medical University were collected from 2008 to 2026. The study’s inclusion and exclusion criteria were met by all 26 patients. The SEER summary stage was divided based on the SEER database and patients’ basic clinical information (https://seer.cancer.gov/seerstat/variables/seer/lrd-stage/). The mean age of all patients was 33.8 ± 15.36 years (14–64 years), and the median age was 33 years. Only one SPN patient in this cohort was male. The nomogram’s C-index was greater (0.965; 95% CI = 0.898–1.032) than that of the SEER summary stage (0.688; 95% CI = 0.486–0.890). Besides, the AUC values at 5, 6, and 7 years were 0.951, 0.937, and 0.934 in the external validation cohort, respectively (Fig. 6). The nomogram showed excellent performance according to internal and external validation results.

Fig. 6.

Area under the curve (AUC) obtained from the nomogram in the external validation cohorts at (A) 5-year, (B) 6-year, and (C) 7-yearpoints.

Discussion

SPN is a sporadic tumor, leading to the lack of clinical evidence to predict patient prognosis¹⁷. Therefore, in order to produce reliable prognostic determinations for SPN, deeper exploration is required⁵¹. However, medical staff currently have a paucity of reliable and extensive sample data to carry out retrospective analysis in clinical practice⁵². Hence, we used the SEER database, which collected data on cancer diagnosis, treatment, and survival for around 30% of the US population, to thoroughly analyze the clinical and pathological characteristics of SPN⁵³.

Nomograms can correctly predict the survival prognosis of patients based on disease characteristics and make clinical decisions for patients with a variety of malignancies^47,54. Therefore, we used a number of clinicopathological variables from the SEER database for the purpose of constructing a nomogram for the OS prediction of SPN patients. Finally, we successfully developed the first SPN-specific overall survival nomogram that integrates age, lymph node status, surgical procedure, and SEER stage to generate individualized survival probabilities. The nomogram’s validation revealed that it had excellent discrimination ability and calibration accuracy, and we externally validated the model in an independent institutional cohort, supporting its generalizability and clinical applicability. Compared to traditional staging alone, the nomogram showed superior predictive performance, as reflected in its high C-index values (0.927, 0.930, and 0.965 across the training, internal, and external validation cohorts, respectively) and robust AUCs (> 0.7). The DCA findings provide critical clinical translation of our statistical model. DCA bridges the gap between predictive analytics and patient-centered care—an essential step toward the responsible implementation of risk models in real-world practice. Additionally, we established a risk stratification system based on the nomogram to identify high-risk SPN patients who may require closer surveillance and tailored surgical strategies. Age, surgery type, positive lymph nodes, and SEER summary stage were found as four independent predictive variables for OS in the present research.

Previous studies on SPN have proposed factors such as age, surgical interventions, sex, and tumor location as potentially affecting the OS of patients with SPN, all of which were fully considered in this study^55–57. In this study, some of our results have similar or opposite conclusions to those of previous studies, which may require continuous attention and supplement to explore the truth. For example, Wu J et al.⁵⁸ reported that the most common site for SPN tumors was the pancreatic head and the tumor location was not a significant prognostic factor for OS in patients with SPN. However, in agreement with the results from Meng et al.⁵⁶ and Uğuz et al.⁵⁹, our study found that the most common SPN tumor location was the pancreatic tail rather than the head.

SPN is currently recognized as a type of cystic, solid, and localized tumor with a good prognosis⁶⁰. Regardless of whether metastasis or vascular invasion is present, complete resection is the primary treatment for SPN⁶¹. The surgical approach should be selected based on tumor location, tumor size, and degree of invasion^62,63. Small tumors with intact capsules are typically treated with local tumor resection or partial pancreatectomy, whereas pancreatic head tumors are treated with pancreaticoduodenectomy (Whipple surgery). In addition, for non-invasive SPN, distal pancreatectomy and limited resection may be the best treatment option for tumors of the pancreatic body and/or tail⁶⁴. Our study identified that local excision of the tumor and partial pancreatectomy were found to be independent prognostic factors for OS. Meanwhile, local excision of the tumor and partial pancreatectomy were also the most common surgical procedures in our data (44.9%) owing to indolent behavior of SPN and a relatively low probability of metastasis or invasion. These findings were in line with the outcomes reported by Hanada K and his colleagues⁵⁷.

Several independent prognostic factors identified in our multivariable analysis are strongly supported by established histopathological and biological evidence. Patients aged ≥ 60 years exhibited a markedly elevated risk of death (HR = 21.21, 95% CI 1.85–243.66, P = 0.014; Table 3), consistent with the observation that SPN in older individuals more frequently displays adverse histological features, including solid growth patterns, nuclear atypia, and stromal degeneration⁶⁵. The presence of 1–4 positive lymph nodes conferred an exceptionally high risk of mortality (HR = 40.16, 95% CI 2.31–696.93, P = 0.011; Table 3), reflecting the pathological fact that lymphatic invasion is a direct morphological correlate of tumor dissemination and represents an established gateway to systemic progression⁶⁶. Distant stage at diagnosis (HR = 8.31, 95% CI 1.80–38.34, P = 0.007; Table 3) pathologically corresponds to hematogenous metastasis, most commonly to the liver, and is accompanied by loss of encapsulation, vascular infiltration, and often more aggressive cytological features⁶⁷. In contrast, surgical resection (all types) emerged as the strongest protective factor (HR = 0.01 for surgery types 1–3, all P < 0.05; Table 3), which aligns with the principle that complete tumor excision eliminates the primary pathological substrate and offers the only curative opportunity for SPN⁴⁰. Collectively, these clinicopathological correlations validate the robustness of our model and reinforce the central role of pathological risk stratification in guiding clinical decision-making for SPN. SPN is thought to predominantly affect young women, according to some infrequent case reports or cohort studies. Wu et al.⁶⁸. performed a retrospective analysis of 54 SPN patients, and 45 of them were female and only 9 were male, with an average age of 32.7 years. Similarly, Hao EIU and his colleagues³ conducted a meta-analysis on aggressive SPNs incorporating 59 patients with aggressive SPN, 52 of which were females, with an average age of 37.44 ± 2.21 years. A study has revealed that the prognosis of SPN patients differs depending on gender⁵⁸.From the SEER database, we obtained 341 SPN patients, and 296 of them were females. There were 238 SPN patients in the randomized training cohort, 207 of whom were females, and a majority of them were under the age of 40. Our statistical analysis, however, took into account all possible prognostic parameters and failed to find such a difference, which was consistent with the findings of Uğuz et al.⁵⁹. Further research may be needed to probe into whether the sex-related difference is an independent prognostic factor for SPN.

The prognostic factors identified in our nomogram—particularly age, lymph node metastasis, distant stage, and surgical resection—are not only statistically robust but also biologically anchored in the current understanding of SPN pathogenesis. SPN is characterized by near-universal activating mutations in CTNNB1 (exon 3)⁶⁹, leading to constitutive Wnt/β‑catenin signaling, which serves as the primary oncogenic driver⁷. This pathway activation promotes epithelial–mesenchymal transition (EMT) and cell proliferation, histopathologically manifesting as solid pseudopapillary architecture and, in aggressive cases, invasive growth and metastasis⁷⁰. The strong independent prognostic weight of lymph node positivity and distant metastasis in our model is therefore consistent with the biological capacity of Wnt‑driven SPN clones to disseminate. Conversely, the protective effect of complete surgical resectionreflects the removal of the entire CTNNB1‑mutated clone, which is currently the only curative intervention⁴⁰. Beyond CTNNB1, recent whole‑exome sequencing has identified inactivating mutations in epigenetic regulators (KDM6A, TET1, BAP1) in metastatic SPN, suggesting that epigenetic dysregulation cooperates with Wnt activation to promote tumor progression³². Although our SEER‑based analysis could not incorporate these molecular markers, they provide a biological rationale for refining risk stratification: patients classified as high‑risk by our nomogram may be enriched for such additional genetic alterations, and future integration of these biomarkers could further enhance predictive accuracy. From a therapeutic perspective, the nomogram may inform personalized management strategies. For patients predicted to have excellent prognosis (e.g., young age, node‑negative, localized disease), surgery alone remains curative, and aggressive adjuvant therapy is unwarranted. In contrast, high‑risk patients—particularly those with nodal involvement or distant metastasis—may be candidates for more intensive surveillance or clinical trials evaluating novel agents. Although no targeted therapies are currently approved for SPN, preclinical evidence suggests that inhibitors of Wnt/β‑catenin signaling or epigenetic therapies could be explored in the metastatic setting⁷. Our nomogram could help identify patients who might derive the greatest benefit from such investigational treatments once they become available.

The present research has a few limitations. Firstly, this was a retrospective study based on a publicly available database. Thus, the results might carry selection bias⁴⁴. Secondly, due to the limitations of the SEER database, we could not obtain more useful prognostic information(e.g., reproductive histories, tumor calcification, solid component, immunohistochemical markers, and preoperative examinations) and biomarker data(e.g., expression levels of specific proteins like Ki-67, p53, or gene mutations)⁷¹. In addition, the SEER database does not contain histopathological images. Although our classification relies on standardized codes derived from the original pathology reports, it is not possible to directly visually confirm the morphological features in this dataset. Finally, data size from the external validation cohort was relatively small acquiring only from one tertiary hospital. Moreover, all of the cases in the external validation cohort were Asian, which could lead to selection bias.

In conclusion, the SEER database was used for the construction of a nomogram that reliably assessed 5-, 6-, and 7-year OS for patients suffering from SPN. The nomogram’s performance was verified and it was found to have excellent performance. SPN prognosis prediction nomograms may aid clinicians in more accurately predicting the prognosis outcomes and providing tailored treatment. As far as we know, this nomogram is the first model that could be used in clinical practice for the prediction of OS in patients suffering from SPN.

Methods

Data collection and study cohort

Two cohorts were included in this population-based retrospective analysis. For one cohort, data were collected from the SEER database of the National Cancer Institute using SEER*Stat 8.3.9 software (https://seer.cancer.gov/seerstat/) (account ID:17004-Nov2020). Patients diagnosed with SPN between January 2000 and November 2018 were found in this cohort with the aid of the SEER database. According to the regulations of the National Cancer Institute of the United States, the SEER project does not need to undergo the approval of the Institutional Review Board (IRB) when conducting research. All data have been anonymized and are publicly accessible. The following were the criteria for patient inclusion: (i) International Classification of Diseases (ICD) code O-3 morphology code of 8452, (ii) Positive histology or/and cytology confirms the diagnosis, (iii) active follow-up to ensure the accuracy of the patient’s, (iv) SPN was the only or first primary tumor, and (v) patients were diagnosed > 1 month prior to death. Patients with unknown surgical types, race, or marital status were not included in the present research. Age, sex, diagnosis year, marital status, race, tumor site and size, types of surgery, number of positive regional lymph nodes (LN), regional LN dissection experience, and SEER summary stage were all retrieved from the available clinicopathological data. The entire approach for patients’ selection is presented in Fig. 1. All patients with SPN from the SEER database were classified into two groups in a random manner: the training cohort (70%) and the internal validation cohort (30%). The training cohort was used for the identification of independent prognostic factors for SPN patients and the construction of a prognostic nomogram, whereas the internal validation cohort was used for the validation of the nomogram. For the purpose of further validating our nomogram, a subset of patients pathologically diagnosed with SPN from January 2008 to January 2026 collected at our hospital (LuZhou, China) formed an external validation cohort. These patients were evaluated retrospectively, and all eligible cases were included. This retrospective study was conducted in accordance with the Declaration of Helsinki and relevant national guidelines. Given the use of de-identified clinical data, the study met the criteria for a waiver of informed consent. The study protocol was formally reviewed and approved by the Institutional Review Board of the Affiliated Hospital of Southwest Medical University (Approval No.: KY2026120).

Fig. 1.

Flow chart for selection of the study cohorts. Abbreviations: SEER: Surveillance, Epidemiology, and End Results, SPN: Solid pseudopapillary neoplasm of the pancreas.

Statistical analysis

All statistical analyses were carried out using SPSS version 26.0 (SPSS, Chicago, Illinois, USA) or R software (Version 4.5.1) (http://www.r-project.org/). All p-values that were reported were two-sided, and a P < 0.05 was set as the criterion of statistical significance. The X-tile software v3.6.1 (Yale University, New Haven, Connecticut, USA) was used to identify the appropriate cut-off points for tumor size⁷². In the training cohort, variables with P < 0.05 from univariate Cox analysis were included in the multivariate Cox analysis. Then, a nomogram was created on the basis of the independent variables with P-values < 0.05 in the multivariate Cox analysis. The nomogram calculated the 5, 6, and 7-year overall survival (OS) probabilities. To enhance clinical applicability, the continuous risk score derived from the nomogram was categorized into discrete risk groups. The risk score for each patient was defined as the linear predictor (LP) obtained from the multivariable Cox model. An optimal cut-off value was determined using the R software, which maximizes the log-rank statistic to separate patients into low- and high-risk groups. Patients with a risk score ≤ -0.138 were classified as low-risk (n = 140), and those with a score > -0.138 as high-risk (n = 98). The prognostic performance of this dichotomous classification was evaluated by Kaplan–Meier curves with log-rank tests.

The C-index and the corresponding time-dependent AUC were also employed with the aim of assessing the nomogram’s discrimination performance. In general, greater C-Index values were related to more accurate prediction, while AUC values near 1 indicated better model discrimination^73–75. The bootstrap approach helped to re-sample the data 1,000 times and construct calibration charts for evaluating the calibration capabilities. In addition, decision curve analysis (DCA) was used to assess the nomogram’s clinical utility.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

PZ, QT, and FH conceived and designed the study. PZ and QT contributed to the literature research and graphics. PZ drafted the manuscript. The manuscript was revised by PZ, QT and FH. All authors read and revised this manuscript.

Data availability

Publicly available datasets were analyzed in this study. The data can be found in the SEER database (https://seer.cancer.gov/).

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

This study was conducted in accordance with the Declaration of Helsinki. This study was approved by the Institutional Review Board (IRB) of The Affiliated Hospital of Southwest Medical University (Approval No. KY2026120).