Survival Analysis and Prediction Model of ASCP Based on SEER Database.

Background: This study aims to compare the incidence and clinical and survival characteristics of adenosquamous carcinoma of the pancreas (ASCP) and adenomatous carcinoma of the pancreas (ACP), analyze the survival factors of ASCP and construct a prognostic model. Method: Patients diagnosed with pancreatic cancer from 2000 to 2018 are selected from the SEER database. ASCP and ACP are compared in terms of epidemiology, clinical characteristics and prognosis. Cases are matched in a 1:2 ratio, and survival analysis is performed. The Cox proportional hazard model is used to determine covariates related to overall survival (OS), and an ASCP prognosis nomogram is constructed and verified by consistency index (C-index), calibration chart and decision curve analysis (DCA). The accuracy of the model is compared with that of AJCC.Stage and SEER.Stage to obtain the area under the receiver operating characteristic (ROC) curve.
Results: the age-adjusted incidence of ACP increased significantly over time from 2000 to 2008 and from 2008 to 2018 (P < 0.05). APC was 2.01% (95% CI: 1.95-2.21) and 1.08% (95% CI: 0.93-1.25) respectively. The age-adjusted incidence of ASCP increased with time from 2000 to 2018 (P < 0.05) and APC was 3.64% (95% CI: 3.25-4.01).After propensity score matching (PSM), the OS and cancer-specific survival (CSS) of ACP are better than those of ASCP. The survival time of ASCP is significantly improved by the combined treatment of surgery + chemotherapy + radiotherapy, with a median OS of 31 months. Cox proportional hazard regression analysis shows that age, race, surgery, radiotherapy, chemotherapy and tumor size are independent factors affecting the prognosis. DCA and area under the curve (AUC) value shows that the model has good discrimination ability.
Conclusion: The OS prognosis of ASCP is worse than that of ACP, and the nomogram has high accuracy for the prognosis prediction of ASCP.

Introduction

Pancreatic cancer is a highly malignant tumor and the deadliest gastrointestinal (GI) cancer, with morbidity and mortality approaching 1 (1). With half a million new cases diagnosed each year worldwide, it is one of the few cancers of which the incidence is still rising in the United States (2, 3). Most pancreatic cancers are adenomatous carcinoma of the pancreas (ACP), and only 0.4–4% are adenosquamous carcinoma of the pancreas (ASCP) (4). Histologically, ASCP is defined as consisting of at least 30% malignant squamous cell carcinoma with coexisting ductal adenocarcinoma (5, 6). Due to its low incidence rate, ASCP has been reported in individual or small cases in most literature (5, 7–12). Despite current advances in surgical techniques and clinical drugs, the overall survival (OS) is less than 1 year (13), with a median OS(MOS) of 12 months for resectable disease and 4–5 months for metastatic disease (14). Compared with ACP, ASCP is more aggressive and potentially metastatic with a worse prognosis (15). However, according to several studies, the OS of ASCP and ACP are inconsistent (8, 16). ASCP is a solid cancer and its prognostic factors are difficult to predict. Currently, the TNM staging system of the American Joint Committee on Cancer (AJCC) is used to evaluate the survival and prognosis of pancreatic cancer patients. Based on tumor size and extent, number of lymph node metastases and number of distant metastases (17, 18), the system provides a simple but incomplete tool for assessing ASCP development and disease treatment and decision-making. According to the reported literature, the independent prognostic factors of ASCP include age, gender, race, radiotherapy, chemotherapy, surgery, anatomic site, etc. (12, 14, 16, 19). Compared with TNM staging, the above factors have a convincing ability to predict the prognosis of ASCP. Thus, further studies on ASCP are necessary.

Surveillance, Epidemiology and End Results (SEER) (https://seer.cancer.gov/) is the largest and most authoritative database of tumor-related information in the United States, which collects tumor incidence and survival data from population-based cancer registries covering approximately 34.6% of the U.S. population (20). Studies combined with the SEER database can be targeted at different regions and larger populations, and population analysis can produce more convincing results for the rare tumor ASCP. Big data can generate individual probabilities of clinical events, meeting our needs for integrated biological and clinical models (21, 22). Therefore, this study explores the epidemiology, clinical characteristics and prognosis of ASCP and ACP through the SEER database. In addition, we analyzed the prognostic risk factors for ASCP and constructed a nomogram to provide clinicians with a convenient tool for implementing individualized treatment.

Materials and Methods

Ethical Approval

Informed consent is not required for SEER data, and this study is consistent with the 1964 “Declaration of Helsinki” and its subsequent amendments or similar ethical standards.

Patients

The data for this study was obtained from the SEER database, covering up to 97% of cancer incidence and 28% of the U.S. population, and obtained by SEER*Stat v8.3.9 under the registration number 17070-Nov2020.

Inclusion criteria: 1. International Classification of Diseases for Oncology ICD-0-3/WHO 2008 = Pancreas; 2. patients diagnosed between 2000 and 2018; 3. histologically, ACP under ICD-0-3 His/behav = 8140, 8141, 8142, 8144, 8500, 8501, 8503, 8504, 8507 and 8521, and ASCP under ICD-0-3 His/behav = 8560. Exclusion criteria: 1. previous malignant tumors and multiple tumors; 2. non-pathologically/cytologically confirmed patients and autopsy results; 3. unknown surgery information; 4. tumor size = 0 cm; 5. T0 Stage; 6. unknown survival time. OS is defined as the interval between cancer diagnosis and the last follow-up of patients who have died by any cause or are still alive. Cancer-specific survival (CSS) is defined as the time interval between cancer diagnosis and death from pancreatic cancer.

Statistical Analysis

Demographic and clinical characteristics were extracted from the SEER database, including age, gender, race, grade, tumor site, SEER stage, AJCC stage, T stage, lymph node metastasis, distant metastasis, adjuvant therapy (chemotherapy and radiation therapy) surgery and vital conditions at follow-up. Age-adjusted incidence per 100,000 patients diagnosed per year was calculated based on the number of cases. Annual percentage change (APC) in incidence change was assessed using Joinpoint software, and random 1:2 nearest neighbor propensity score matching (PSM) was used to balance all baseline covariates between ASCP and ACP. ASCP cases in the SEER database were then randomly assigned to the training and validation groups in a ratio of 7:3. The classified data was expressed as frequency and percentage, and verified by the Chi-square test or Fisher’s exact test. Continuous data was expressed as median and standard deviation (SD), and compared by the Mann-Whitney U test. The survival curvess were plotted using the Kaplan-Meier method and compared by the log series test. The Cox proportional hazard regression model was used for univariate and multivariate analyses, and the hazard ratio (HR) and corresponding 95% confidence interval (CI) were calculated. P values greater than 0.05 on both sides were considered statistically significant. All statistical analyses were performed using R-version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria, http://www.r-project.org) and related software packages. In the training group, a probability nomogram of OS at 6, 12 and 24 months was constructed on the basis of independent prognostic risk factors screened by Cox proportional hazard regression, then compared using AJCC.Stage and SEER.Stage to calculate the nomogram score. The sensitivity of the nomogram was assessed by C-index. A calibration chart (1,000 bootstrap resamples) was used to evaluate the consistency between model prediction and actual prediction, and DCA was used to evaluate the clinical effects and clinical benefit capacity of the nomogram. Finally, the accuracy of the nomogram, AJCC.Stage and SEER.Stage was compared using the area under the ROC curve.

Results

Incidence of ACP and ASCP

In this study, patients with ASCP and ACP from 2000 to 2018 were selected from the SEER database, and the incidence trend was described using Joinpoint software. As shown in , the age-adjusted incidence of ACP increased significantly over time from 2000 to 2008 and from 2008 to 2018 (P < 0.05). APC was 2.01% (95% CI: 1.95–2.21) and 1.08% (95% CI: 0.93–1.25) respectively, but the incidence of pancreatic adenocarcinoma slowed slightly with time after 2008. The age-adjusted incidence of ASCP increased with time from 2000 to 2018 (P < 0.05) and APC was 3.64% (95% CI: 3.25–4.01). The detailed incidence data of ACP and ASCP is shown in (2000–2018). ( shows the age-adjusted incidence figures of ASCP and ACP from 1975 to 1999, and shows the detailed incidence data from 1975 to 1999).

Figure 1

The age-adjusted incidence of ACP and ASCP patients between 2000 and 2018 from the SEER database(ACP, Pancreatic adenocarcinoma; ASCP, Pancreatic adenosquamous carcinoma; APC, annual percent change).

Demographic and Clinical Characteristics

From 2000 to 2018, a total of 136,336 patients in the SEER database (according to ICD-O-3) were diagnosed as having ACP and ASCP. According to the inclusion and exclusion criteria, 101,796 patients were finally included in this study, including 101,012 ACP cases and 784 ASCP cases ( ). The clinical characteristics of two different tumor subtypes of two different pancreatic cancers are summarized in . It can be seen that in terms of demographic characteristics, the average age at diagnosis in both groups was around 67 years old, and the incidence was slightly higher in men than women, accounting for 51.9% in ACP and 52.3% in ASCP. There was a significant increase in the proportion of white patients compared with that of other races at 79.2% and 81.0% respectively. In terms of clinical characteristics, the incidence of cancer of the head of the pancreas was higher than that of other parts, accounting for 52.4% in ACP and 44.6% in ASCP. Interestingly, distant metastasis was higher in ACP than in ASCP at 44.6% and 37.9% respectively. the proportion of patients undergoing surgery in ASCP is higher than that of ACP (39.4% vs. 21%), and the proportion of patients with ASCP who received chemotherapy and radiotherapy was similar to that of ACP patients. To eliminate differences in baseline characteristics between the two groups, PSM was used to balance all characteristics, including age, gender, race, grade, tumor site, SEER.Stage, AJCC.Stage, T stage, lymph node metastasis, distant metastasis, radiotherapy, chemotherapy and surgery. After PSM, there were no significant differences in demographic and clinical characteristics between the two groups of patients (P > 0.05).

Figure 2

Flowchart for Selecting Patients from SEER Database.

Table 1

Comparison of clinicopathological characteristics between ASCP and ACP.

Characteristic	Before PSM		P	After PSM		P
	ACPn=101012 (%)	ASCPn= 784 (%)		ACPn=1568 (%)	ASCPn=784(%)
Age	67.4(11.4)	67.9(10.7)	0.171	67.5 (10.6)	67.9 (10.7)	0.335
Race:			0.312			0.561
White	80051 (79.2%)	635 (81.0%)		157(10.0%)	84 (10.7%)
Black	12646 (12.5%)	84 (10.7%)		114 (7.27%)	65 (8.29%)
Other	8315 (8.23%)	65 (8.29%)		1297 (82.7%)	635 (81.0%)
Sex:			0.869			0.693
Male	52462 (51.9%)	410 (52.3%)		835 (53.3%)	410 (52.3%)
Female	48550 (48.1%)	374 (47.7%)		733 (46.7%)	374 (47.7%)
Marital.status:			0.023			0.886
Married	57136 (56.6%)	489 (62.4%)		971 (61.9%)	489 (62.4%)
Divorced/Separated	11020 (10.9%)	77 (9.82%)		149 (9.50%)	77 (9.82%)
UnMarried	13672 (13.5%)	95 (12.1%)		178 (11.4%)	95 (12.1%)
Widowed	15362 (15.2%)	101 (12.9%)		225 (14.3%)	101 (12.9%)
Unknown	3822 (3.78%)	22 (2.81%)		45 (2.87%)	22 (2.81%)
Grade			<0.001.			0.889
I	3965 (3.93%)	3 (0.38%)		7 (0.45%)	3 (0.38%)
II	16685 (16.5%)	100 (12.8%)		177 (11.3%)	100 (12.8%)
III	15773 (15.6%)	299 (38.1%)		605 (38.6%)	299 (38.1%)
IV	574 (0.57%)	14 (1.79%)		28 (1.79%)	14 (1.79%)
Unknown	64015 (63.4%)	368 (46.9%)		751 (47.9%)	368 (46.9%)
Primary.Site:			<0.001			0.94
Body	12721 (12.6%)	112 (14.3%)		217 (13.8%)	112 (14.3%)
Head	52949 (52.4%)	350 (44.6%)		712 (45.4%)	350 (44.6%)
Tail	13122 (13.0%)	176 (22.4%)		360 (23.0%)	176 (22.4%)
Other	22220 (22.0%)	146 (18.6%)		279 (17.8%)	146 (18.6%)
Seer.stage:			0.003			0.866
Localized	7714 (7.64%)	64 (8.16%)		131 (8.35%)	64 (8.16%)
Regional	36100 (35.7%)	323 (41.2%)		628 (40.1%)	323 (41.2%)
Distan	57198 (56.6%)	397 (50.6%)		809 (51.6%)	397 (50.6%)
AJCC.stage:			<0.001			0.675
I	5869 (5.81%)	50 (6.38%)		109 (6.95%)	50 (6.38%)
II	22602 (22.4%)	247 (31.5%)		474 (30.2%)	247 (31.5%)
III	8931 (8.84%)	59 (7.53%)		102 (6.51%)	59 (7.53%)
IV	45102 (44.7%)	297 (37.9%)		633 (40.4%)	297 (37.9%)
Unknown	18508 (18.3%)	131 (16.7%)		250 (15.9%)	131 (16.7%)
T.stage			<0.001			0.904
T1	2841 (2.81%)	5 (0.64%)		9 (0.57%)	5 (0.64%)
T2	18140 (18.0%)	138 (17.6%)		302 (19.3%)	138 (17.6%)
T3	33363 (33.0%)	350 (44.6%)		681 (43.4%)	350 (44.6%)
T4	17124 (17.0%)	106 (13.5%)		210 (13.4%)	106 (13.5%)
Unknown	29544 (29.2%)	185 (23.6%)		366 (23.3%)	185 (23.6%)
LN metastasis			<0.001			0.411
Yes	42022 (41.6%)	311 (39.7%)		599 (38.2%)	311 (39.7%)
No	29767 (29.5%)	282 (36.0%)		608 (38.8%)	282 (36.0%)
Unknown	29223 (28.9%)	191 (24.4%)		361 (23.0%)	191 (24.4%)
Distant metastasis			<0.001			0.508
No	39017 (38.6%)	365 (46.6%)		702 (44.8%)	365 (46.6%)
Yes	45073 (44.6%)	297 (37.9%)		633 (40.4%)	297 (37.9%)
Unknow	16922 (16.8%)	122 (15.6%)		233 (14.9%)	122 (15.6%)
Surgery:			<0.001			0.846
No	79832 (79.0%)	475 (60.6%)		958 (61.1%)	475 (60.6%)
Yes	21180 (21.0%)	309 (39.4%)		610 (38.9%)	309 (39.4%)
Radiation:			0.835			0.074
No	82878 (82.0%)	646 (82.4%)		1338 (85.3%)	646 (82.4%)
Yes	18134 (18.0%)	138 (17.6%)		230 (14.7%)	138 (17.6%)
Chemotherapy:			0.212			0.988
No	43921 (43.5%)	323 (41.2%)		644 (41.1%)	323 (41.2%)
Yes	57091 (56.5%)	461 (58.8%)		924 (58.9%)	461 (58.8%)

Survival Analysis

The Kaplan-Meier method was used primarily to evaluate OS and CSS before and after PSM in all patients of both types. Before matching ( ), the OS and CSS of ASCP patients were lower than those of ACP patients, but there was no significant difference between the two groups (P = 0.96, P = 0.85). After matching ( ), the OS and CSS of ASCP patients were still lower than those of ACP patients, with statistical significance (P = 0.016 and P = 0.02 respectively),ASCP and ACP patients had a median OS of 6 months and 7 months respectively. In addition, the matched data was used to compare the OS of ASCP and ACP patients without any treatment and after surgery, radiotherapy and chemotherapy. In the two groups of patients who underwent surgery, considering that they were at different stages of tumor and their prognoses were also different, we plotted the survival curves of three stages according to SEER.Stage. At all stages ( ), ACP or ASCP patients who underwent surgery had a better prognosis than those who did not. Notably, the post-operative MOS of ASCP patients were higher than those of ACP patients in both the Localized and Distant stages, but were lower than those of ACP patients in the Regional stage (6 months vs 9 months) (P < 0.001). A total of 368 subjects were treated with adjuvant radiotherapy in this study, of which 249 were treated in the Regional stage, often in combination with adjuvant chemotherapy. In patients with ACP ( ), the MOS of patients receiving surgery+chemotherapy were higher than those receiving surgery+adjuvant chemotherapy+adjuvant radiotherapy (18 months vs. 17 months) (P < 0.001). Surprisingly, ( ) the MOS were significantly higher in ASCP patients receiving concurrent surgery+adjuvant radiotherapy+adjuvant chemotherapy than in those receiving other treatment modalities, reaching a staggering 31 months (P < 0.001). In addition, as shown in , among the patients receiving surgery+chemotherapy+radiotherapy, the benefits for ASCP patients were significantly higher than those for ACP patients (31 months vs. 17 months) (P < 0.001). It can be seen that ACP patients did not benefit from radiotherapy. Surgery+adjuvant chemotherapy+adjuvant radiotherapy may be the optimal treatment for ASCP patients in the Regional stage. Patients in the Distant stage accounted for 43.62% of all patients. According to the 2011 literature, pancreatic cancer is treated by a multi-drug combination regimen. Due to monotherapy, taking 2011 as the boundary, we divided the patients into two groups to assess the role of chemotherapy at this stage ( ). From 2000 to 2011, among ASCP and ACP patients, MOS were 4 months and 5 months respectively. From 2012 to 2018, the two showed improved MOS of 6 and 7 months respectively, and this difference was statistically significant in ACP patients (P = 0.0062) ( ), whereas it was not significant in ASCP patients (P = 0.28) ( ).

Figure 3

OS and CSS of ACP and ASCP before PSM (A, B) and after PSM (C, D) (PSM, propensity score matching).

Figure 4

Survival Curves of ACP and ASCP: (A) Surgical Treatment in Localized Stage; (B) Surgical Treatment in Regional Stage; (C) Surgical Treatment in Distant Stage; (D) Comparison of Treatment Methods in Regional Stage (S, surgery; C, chemotherapy; R, radiotherapy).

Figure 5

Role of Radiotherapy in ACP and ASCP: (A) Role of Radiotherapy in ACP Regional Stage; (B) Role of Radiotherapy in ASCP Regional Stage (S, surgery; C, chemotherapy; R, radiotherapy).

Figure 6

Role of Chemotherapy in ACP and ASCP (2000–2011 and 2012–2018): (A, C) Comparison of Treatment Methods between ACP and ASCP in Distant Stage; (B, D) Comparison of Chemotherapy Efficacy between ACP and ASCP in Different Time Periods in Distant Stage (S, surgery; C, chemotherapy).

Cox Regression Analysis

According to the inclusion and exclusion criteria, 784 patients were included in the ASCP group, randomly divided into the training group (N = 552) and test group (N = 232) in a ratio of 7:3. lists the basic demographic and clinical characteristics, with no significant differences between the two groups. As shown in , Cox univariate analysis suggested that age, marital status, race, SEER.Stage, AJCC.Stage, lymph node metastasis, distant metastasis, surgery (yes/no), adjuvant radiotherapy (yes/no), adjuvant chemotherapy (yes/no) and tumor size were independent prognostic factors that ultimately affected patient survival; while Cox multivariate analysis suggested that age, race, SEER.Stage, AJCC.Stage, surgery (yes/no), radiotherapy (yes/no), chemotherapy (yes/no) and tumor size were independent prognostic factors that ultimately affected patient survival. Cox multivariate analysis excluded lymph node metastasis and M stage, which are considered to have a strong correlation with AJCC.Stage.

Table 2

Univariate and multivariate cox regression analyses of the prognoses of the ASCP.

Variables	Univariate analysis				Multivariate analysis
	HR	95%CI	P		HR	95%CI	P
Age(years)	1.019	1.01~1.029	<0.001		1.019	1.008~1.03	0.001
Marital status
Married	Reference
Divorced/Separated	1.018	0.754~1.374	0.908		1.007	0.738~1.375	0.963
Unmarried	1.455	1.109~1.909	0.007		1.323	0.999~1.752	0.051
Widowed	1.474	1.13~1.922	0.004		1.151	0.757~1.343	0.955
Unknow	1.498	0.891~2.519	0.127		1.008	0.664`1.994	0.617
Race
White	Reference
Black	1.344	1.022~1.768	0.035		1.58	1.191~2.096	0.002
Other	0.935	0.665~1.316	0.70		1.047	0.739~1.485	0.794
Sex
Male	Reference
Female	1.082	0.907~1.292	0.383
Primary site	Reference
Body
Head	0.842	0.643~1.102	0.21
Tail	0.907	0.669~1.229	0.527
Other	1.175	0.863~1.599	0.306
Grade
I	Reference
II	1.098	0.268~4.502	0.896
III	1.333	0.33~5.38	0.686
IV	1.131	0.228~5.612	0.88
Unknow	2.457	0.61~9.901	0.206
Seer.stage
Localized	Reference
Regional	0.389	0.27~0.561	<0.001		0.82	0.422~1.594	0.559
Distant	0.39	0.321~0.475	<0.001		0.699	0.502~0.974	0.035
AJCC.stage
I	Reference
II	1.246	0.788	1.97	0.347	2.469	1.111~5.487	0.026
III	1.844	1.072	3.172	0.027	2.223	0.965~5.122	0.061
IV	3.181	2.021	5.006	<0.001	2.034	0.774~5.345	0.15
Unknown	1.963	1.219	3.162	0.006	1.443	0.46~4.53	0.53
T stage
T1	Reference
T2	1.8	0.568~5.704	0.318
T3	1.638	0.524~5.124	0.396
T4	2.491	0.785~7.907	0.121
Unknow	2.347	0.746~7.389	0.145
Lymph node metastasis
No	Reference
Yes vs	0.988	0.805~1.212	0.907		1.007	0.806~1.258	0.954
Unknow vs No	1.279	1.015~1.61	0.037		1.047	0.752~1.459	0.785
Distant metastasis
No	Reference
Yes	2.38	1.944~2.914	<0.001		1.008	0.599~1.696	0.977
Unknow	1.427	1.108~1.837	0.006		1.361	0.45~4.117	0.586
Surgery
No	Reference
Yes	0.297	0.242~0.364	<0.001		0.354	0.267~0.469	<0.001
Radiotherapy
No	Reference
Yes	0.452	0.352~0.58	<0.001		0.617	0.469~0.813	0.001
Chemotherapy					0.479	0.39~0.588	<0.001
No	Reference
Yes	0.461	0.384~0.553	<0.001		1.437	1.15~1.796	0.001
Tumor size (cm)
<4.6	Reference
4.7~7.0	1.62	1.309~2.004	<0.001		1.437	1.15~1.796	0.001
>7.0 vs <4.6	2.475	1.876~3.266	<0.001		1.984	1.477~2.664	<0.001
Unknow vs <4.6	2.544	1.925~3.362	<0.001		1.368	1.009~1.855	0.043

Construction and Validation of Nomogram

Based on Cox analysis, 8 independent prognostic factors were used to construct the nomogram ( ), and a specific score for each independent prognostic factor was obtained ( ). According to the total score, the survival of ASCP patients at 6, 12 and 24 months can be confidently predicted. The stability of the nomogram was verified by the correction curves in the training and verification groups. shows that there is good consistency between the predicted survival results and actual survival results. demonstrates the superiority of this model over AJCC.Stage and SEER.Stage in predicting the net clinical benefits of 6, 12 and 24-month survival.

Figure 7

Nomogram of ASCP. *P<0.05 for independent prognostic factors screened in the multivariate cox regression analysis of ASCP prognosis. ***P<0.01 of independent prognostic factors screened in the multivariate cox regression analysis of ASCP prognosis. They can reflect the influence of the independent prognostic factor on the prognosis of patients, and "***" is more strongly correlated with the prognosis of patients than "*".

Figure 8

Calibration curves of nomogram (A, C, F) in training group at 6, 12 and 24 months; calibration curves of nomogram (B, D, E) in validation group at 6, 12 and 24 months.

Figure 9

DCA of nomogram (A, C, F) in training group at 6, 12 and 24 months; DCA of nomogram (B, D, E) in validation group at 6, 12 and 24 months (DCA, decision curve analysis).

Comparison of OS Prediction Accuracy Between Nomogram, AJCC.Stage and SEER.Stage

The accuracy of the prediction model was verified by the area under the ROC curve. As shown in , in the training group, the AUCs of the nomogram, AJCC.Stage and SEER.Stage were 0.770, 0.567 and 0.626 respectively; in the training group, the AUCs of the nomogram, AJCC.Stage and SEER.Stage were 0.789, 0.579 and 0.581 respectively. The results show that this model has higher accuracy in predicting the OS of ASCP patients, making it more suitable for predicting the survival prognosis of ASCP.

Figure 10

Accuracy comparison of nomogram, AJCC.Stage, SEER.Stage in training group (A) and validation group (B), AUC, area under the curve; ROC, receiver operating characteristic.

Discussion

In this population-based retrospective analysis, the incidence of ASCP is extremely low, and the overall incidence of pancreatic cancer is about 0.7%, which is similar to previous studies (4, 23). It is worth noting that the incidence of ASCP and ACP is increasing year by year from 2000 to 2018, which may be related to pancreatic inflammation and obesity (24). This study is the first to use PSM to compare demographic characteristics between ASCP and ACP. The results show that ASCP and ACP are very similar in age of onset, gender and race; the average age of onset is 67, and the incidence is highest in white males, this has been reported in previous studies (14, 16, 25). In terms of clinicopathological characteristics, it is worth noting that Grade II in pathology has a lower OS than Grade IV, which may be related to data sources and caused by bias; more prospective studies may be needed for validation. Interestingly, consistent with previous research (15, 25, 26), the histological grading of ASCP indicates a higher degree of malignancy, with Grades III and IV accounting for about 40%, compared with only about 16% for ACP, this may be related to the squamous cell component (27). In addition, the two have similar clinical manifestations, such as abdominal pain, weight loss, and jaundice, making it difficult to distinguish ASCP from ACP.Therefore, endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) is extremely important for the preoperative diagnosis of solid pancreatic lesions and is considered the gold standard (28–30).

Many studies (15, 31) have shown that ASCP is always more aggressive and malignant histologically than ACP, but previous studies on ASCP (10, 14, 16) have confirmed that there is no difference in OS between ASCP and ACP in the whole population, which is different from the conclusions reported in the literature (8, 12, 25). In order to arrive at a reliable conclusion and minimize the influence of various biases, we used PSM to balance the baseline characteristics. It was found that before PSM, there was no difference in OS and CSS between ASCP and ACP, but after PSM, ACP had a good prognosis with a median OS of 7 months, while the median survival time for ASCP was 6 months (P < 0.05). This conclusion is based on the fact that the data sample is by far the largest, so the conclusion is highly reliable. Subsequently, we compared the prognosis improvement of ASCP and ACP by different treatment methods.Surgery remains a vital treatment modality for pancreatic cancer and may significantly prolong the OS of patients (5, 13, 32–34). According to the results of this study, the surgical rate of ASCP patients is significantly higher than that of ACP patients (up to nearly 40%), and both ASCP and ACP patients can benefit from surgery. Pancreatic cancer can be divided into three categories: resectable, borderline and unresectable (35). However, since this is not documented in the SEER database, we demonstrate from the three stages of SEER.Stage that patients can benefit from surgery when it is available. Studies (36, 37) have shown that the FOLFIRINOX regimen and the gemcitabine plus albumin-bound paclitaxel regimen have significantly improved the prognosis of metastatic pancreatic cancer. Besides, in the updated NCCN guidelines for pancreatic cancer, the multi-drug combination therapy model is adopted as the acceptable regimen for advanced pancreatic cancer (38). In this study, it is also demonstrated that the prognosis of ACP patients receiving chemotherapy after 2012 in the Distant stage was significantly higher than before (7 months vs. 5 months) (P < 0.05), and the survival time after ASCP chemotherapy was increased as compared with before 2011 (6 months vs. 4 months), but not statistically significant. Our results suggest that in the Regional stage, the survival time of ASCP and ACP patients undergoing chemotherapy after surgery is significantly higher than that of patients undergoing surgery alone, which is consistent with the literature (39, 40). Unfortunately, chemotherapy regimens are not recorded in the SEER database, which is one of the limitations of this study. Besides, radiotherapy plays a vital role in the treatment of pancreatic adenosquamous carcinoma. Our results suggest that in the Regional stage, the MOS of ASCP reaches 31 months after surgery combined with adjuvant radiotherapy and chemotherapy, which is significantly higher than that of ACP (17 months). This indicates that radiotherapy can significantly improve the overall survival time of ASCP patients (41–43), and surgery+adjuvant radiotherapy+adjuvant chemotherapy fails to significantly improve the prognosis of ACP patients, the reason might be that squamous carcinoma in ASCP is more sensitive to radiotherapy (43), Compared with ACP, the role of non-surgical treatment in patients with ASCP is unclear (14), and in practice, due to limited data on ASCP, standard chemotherapy regimens for ACP such as FOLFIRINOX, gemcitabine and capecitabine may be used (44, 45), so we need to further explore the standard treatment of ASCP. It is noteworthy that recent studies (46) have shown that the expression of PD-L1 is limited to squamous cell components, which may be the key to subsequent targeted therapies. The role of squamous cells in ASC also requires further investigation. In addition, neoadjuvant therapy may play an important role in the future to improve the R0 resection rate in patients with locally advanced or distant metastases (47–49).

In order to further explore the independent prognostic risk factors for ASCP, we performed univariate and multivariate Cox analysis, which revealed that the prognosis of advanced-stage white elderly patients was poor, while surgery, radiotherapy and chemotherapy were protective factors (50–53)that were beneficial for prognosis. This is consistent with the previous survival analysis. According to the multivariate Cox analysis, we established a nomogram with C-indexes of 0.780 and 0.781 respectively, indicating good consistency between predicted survival rate and actual survival rate. We then validated the prediction effects of the nomogram at 6, 12 and 24 months, which revealed that the nomogram has good performance. DCA indicated that the net clinical benefit of this model is higher than that of AJCC.Stage and SEER.Stage. Finally, the area under the ROC curve indicated that the prediction of this model is more comprehensive and accurate. This model may be used for individualized prognostic assessment, and may become an effective diagnostic tool for making treatment-related decisions (21).

However, our analysis had several limitations. First, since it is a retrospective analysis, the selection bias was inevitable even if PSM was used to reduce the bias. Second, the absence of smoking and drinking history, pancreatitis history and CA19-9 and CA125 data in the SEER database may have affected the screening of independent prognostic factors. Third, since the SEER database fails to record the primary type of tumor before surgery such as resectable, borderline and unresectable for surgical treatment, this imposes certain limitations on the analysis of this study. Fourth, this study covers a period of 18 years during which the replacement of the chemotherapy regimen, monotherapy to multi-drug combination and the emergence of neoadjuvant therapy were included. However, in this study, we only included adjuvant therapy (radiotherapy and chemotherapy), but not neoadjuvant therapy; if neoadjuvant therapy had been included, there would have been a lot of missing data, imposing a limitation on this study. Fifth, the lack of external data does not support external validation, so this prediction model may not be highly applicable to populations in other regions. To this end, ASCP patients at our center will be collected for external validation at a later date.

Conclusion

In conclusion, the incidence of both ACP and ASCP is increasing year by year, which requires greater attention. in the Regional stage, surgery + chemotherapy + radiotherapy can significantly improve the prognosis of ASCP. Finally, the accuracy of our prediction model is higher than that of AJCC.Stage and SEER.Stage, and can help clinicians to better implement individual treatment.

Data Availability Statement

The original contributions presented in the study are included in the article/ . Further inquiries can be directed to the corresponding author.

Author Contributions

S-YL designed the study and wrote the manuscript. Z-QY and M-JL extracted the data and conducted statistical analysis. S-YL and C-NX made and improved the charts. Z-MW controlled the quality of the study and revised the paper. All authors contributed to the paper and approved the submission of the manuscript.

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.