Pancreatic adenosquamous carcinoma: A population level analysis of epidemiological trends and prognosis

The incidence and mortality of pancreatic adenosquamous carcinoma (PASC) have received little attention. The goal of our study was to explore the overall epidemiological trend of PASC at the population level.


| Date source
The SEER database is a reliable source of follow-up information about cancer patients maintained by the National Cancer Institute (NCI). 22 Based on the SEER-18 database, which includes cancer registries representing 28% of the US population, we studied the incidence and mortality of PASC patient from 2000 to 2017, as well as the characteristics of the patients who suffer from this condition.

| Study population
To identify patients with PASC from 2000 to 2017, we used the International Classification of Diseases for Oncology 3 (ICD-O-3) codes (8560,8570) and SEER site codes (C25). In this study, data from patients who died within 1 month of their diagnosis were excluded to avoid cases where the survival time was zero because the smallest unit of survival was the month, rather than the day. In addition, we excluded patients whose diagnosis was not microscopically confirmed by imaging or clinical diagnosis. Overall survival (OS) was analyzed using the case data screened above ( Figure 1). Patients who died of non-pancreatic adenosquamous carcinoma or whose cause of death was unknown were excluded. Disease-specific survival (DSS) was analyzed using data from the screened patients ( Figure 1). We used the same screening method to obtain data from patients with PDAC and pancreatic squamous cell carcinoma (PSCC) for comparison. We included 90,343 patients with PDAC and 306 with PSCC in OS analysis. Additionally, 85,133 patients with PDAC and 286 patients with PSCC were included in DSS analysis. As the American Joint Committee on Cancer Staging Classification has been updated several times during our study period, we used the SEER stage classification, which provides a more consistent classification standard. According to the SEER staging system, tumors are classified as a localized stage that is limited to the primary site, a regional stage that has spread to regional lymph nodes, and a distant stage that has spread to distant tissues and organs.

| Statistical analyses
In this study, the SEER*Stat software (version 8.40) was used for collecting data on incidence, IB mortality, and patient characteristics. Since death certificates do not contain histological information about tumors, we studied IB mortality rather than standard mortality. 23 IB mortality rates were calculated by combining morbidity-and its characteristics at the time of onset-with death certificates. Patients who are categorized as dying from PASC must have been previously diagnosed, not just based on information on the death certificate. In addition, the incidence and IB mortality rates were adjusted for age in the U.S. standard population in 2000.
The joinpoint regression of NCI (version 4.5.01) was used to calculate the annual percent change (APC) and 95% confidence intervals (CIs). The APC method analyzes the trends in incidence and mortality over time, describing the slope, gradient, and direction of each segment. In this method, a logarithmic linear segmentation model is applied, and the incidence and mortality of tumors are assumed to change in a stable percentage compared with the previous year. 24 We classified the population according to the location of the lesions and used descriptive statistics to summarize the demographic and clinical characteristics of each group. Chi-square tests were used to compare categorical variables. We used the Kaplan-Meier method to construct the survival curve and log-rank test to compare survival differences. The Cox proportional hazards model was used to examine factors associated with mortality. Based on the results of Cox regression analysis, nomograms were developed to predict 1-and 2-year survival rate. Moreover, we assessed the prediction discrimination of the nomogram model using the Harrell consistency index (C-index), which can determine the difference between predicted and actual survival. Calibration curves were calculated to test whether the outcome prediction matched the actual outcome. All p values were two-sided, and values of p < 0.05 were considered statistically significant. All statistical analyses were performed using R (version 4.1.3) and GraphPad Prism (version 8.0.2).

| Patient and tumor characteristics
Between 2000 and 2017, 914 patients were diagnosed with PASC. Of these, 815 patients were included in our OS and descriptive analyses, whereas 773 patients were included in our DSS analysis ( Figure 1). We divided the study population into four subgroups based on tumor site (head of pancreas, body of pancreas, tail of pancreas, and others) ( Table 1). In the study population, the median age at diagnosis was 69 year old (interquartile range, [IQR]: 62-76 year old). There was a similar proportion of males (n = 417, 51.2%) and females (n = 398, 48.8%) in the overall cohort. Caucasians comprised the vast majority of the patients (n = 666, 81.7%). Based on SEER stage, most patients had regional (n = 355, 43.6%) and distant PASC (n = 375, 46.0%). Localized tumors occurred in a relatively small number of patients (n = 68, 8.3%). Based on the degree of differentiation, most patients had poorly differentiated PASC (n = 311, 38.2%), followed by moderately differentiated disease (n = 115, 14.1%).
As shown in Table 1, the median age of diagnosis in patients with PASC of the head was 70 years old (IQR: 61.5-76 years old). The median age of patients with PASC in the body and tail of the pancreas was same at the time of diagnosis, both 69 years old (IQR, body: 60-74 years old; tail: 62-77 years old). Similar to the general study population, the proportions of males and females with tumors at different sites were approximately equal. Interestingly, the rate of regional disease was significantly higher in patients with PASC of the head than in those with PASC of the body (p = 0.009) and tail (p < 0.001) of the pancreas. The rate of distant disease was significantly lower in patients with PASC of the head than in those with PASC of the body (p = 0.006) and tail (p < 0.001) of the pancreas. However, the proportion of moderately differentiated disease was significantly higher in patients with PASC of the head than in patients with PASC of the tail (p = 0.004).
Patients who underwent surgery had a higher proportion of regional PASC than those who did not (77.8% vs. 22.5%, p < 0.001) ( Figure S1A). In contrast, the proportion of distant PASC in patients who received surgical treatment was lower than that in patients without surgical treatment (17.5% vs. 70.0%, p < 0.001). In addition, we studied the relationship between tumor size and SEER stage (Figure S1B). Regional PASC was more common in patients with tumors ≤3.5 cm (58.9% vs. 43.9%, p = 0.003), and distant PASC was more common in patients with tumors >3.5 cm (49.4% vs. 27.0%, p < 0.001).

| Incidence and IB mortality trends by stage
We then stratified the study cohort according to the SEER stage. Overall, the incidence of distant and regional PASC was considerably higher than that of localized PASC ( Figure S3A). Moreover, the increasing trends in the incidence of PASC were similar at both distant and regional stages, with APCs of 4.1% (95% CI: 1.0%-7.4%, p < 0.05) in distant PASC and 4.3% (95% CI: 1.4%-7.3%, p < 0.05) in regional PASC. As with incidence, the IB mortality of localized PASC remained the lowest of the three subgroups ( Figure S3B). Furthermore, the increasing trends in IB mortality of PASC were also similar at both distant and regional stages, with APCs of 5.0% (95% CI: 1.4%-8.7%, p < 0.05) and 5.8% (95% CI: 3.1%-8.7%, p < 0.05) in distant and regional PASC, respectively.

| Long-term survival outcomes
The overall median survival of patients with PASC was 6 months (95% CI: 6-7 months), and the disease-specific median survival of patients with PASC was 6 months (95% CI: 5-7 months). The 1-and 2-year OS were 26.6% and 13.6%, respectively, and the corresponding DSS were 25.8% and 12.7%, respectively. Compared to patients with PASC, there was no statistical difference in the OS and DSS of patients with PDAC, but the OS (median survival 4 vs. 6 months, p < 0.001) and DSS (median survival 4 vs. 6 months, p < 0.001) of patients with PSCC were significantly worse ( Figure 3A,B). Patients older than 60 years had significantly worse OS (p < 0.001) and DSS (p < 0.001) [ Figure 3C,D]. In contrast, patients with PASC who underwent surgery had significantly longer OS (median survival: 12 vs. 4 months, p < 0.001) and DSS (median survival: 12 vs. 4 months, p < 0.001) than those who did not ( Figure 3E,F). Moreover, positive regional lymph nodes and tumors larger than 3.5 cm led to shorter OS (p < 0.001) and DSS (p < 0.001) ( Figure 3G-J). According to SEER stage, the OS (median survival: 10 vs. 5 months, p < 0.001) and DSS (median survival 10 vs. 4 months, p < 0.001) of patients with localized PASC were better than those of patients with distant PASC (Figure 3K,L). However, there was no difference in OS and DSS between regional and localized PASC patients. There was also no significant difference in OS and DSS among patients with different sexes and different grades of PASC. (Figure S5A-D). Interestingly, OS (median survival 6 vs. 5 months, p = 0.017) and DSS (median survival 6 vs. 5 months, p = 0.031) were longer in White patients than in Black patients ( Figure S5E,F). Furthermore, the location of the tumor along the pancreas did not affect OS and DSS ( Figure S5G,H).
We first used COX regression to conduct univariate analysis and found that age at diagnosis, SEER stage, treatment, regional lymph node involvement, and tumor size were significantly correlated with OS (p < 0. 1) and DSS (p < 0. 1) (Table S1). These factors were then incorporated into the multivariate COX regression analysis (Table 2). Here, age at diagnosis, treatment, regional lymph node involvement, and tumor size were identified as independent prognosticators of OS and DSS.

| Construction of the Nomogram
The prediction models for OS and DSS contained all of the independent prognostic factors, which were visually represented as nomograms ( Figure 4A, B). By adding the related scores for each parameter and projecting the overall score to the bottom level, the probability of 1-and 2-year OS and DSS may be calculated. The C-index values of OS and DSS were 0.724 (95% CI: 0.685-0.763) and 0.695 (95% CI: 0.636-0.753), reflecting the good prediction discrimination of the nomogram models. The calibration curves of internal validation revealed good agreement between predicted and actual OS and DSS at 1, and 2 years ( Figure 4C, D, E, F). Hue et al. also studied the treatment strategy of PASC through a large cancer database. However, the incidence and mortality of PASC are still unclear, including the changing trend of incidence and mortality. 25,26 Therefore, a thorough grasp of PASC epidemiological patterns is critical to assist clinicians in making more effective clinical decisions.

| DISCUSSION
In our study, the incidence of PASC increased steadily at a rate of approximately 3.93% per year over the study period, although PASC was relatively rare. This rate of increase is of great concern because it is even faster than The OS (p < 0.001) and DSS (p < 0.001) of patients with distant disease were shorter than those with localized and regional disease; There were no statistically significant differences in OS (p = 0.593) and DSS (p = 0.815) among patients with localized and regional disease. that of PDAC. Here, the continuous rise in PASC incidence might imply that prevention has not improved in recent years. Another reason may be that advances in histopathology have led to an increase in the number of diagnoses. Compared with PASC, the overall incidence of PDAC also showed an increasing trend, but the rate of increase has gradually slowed down in recent years. This difference in incidence trends may suggest that precautionary measures for PDAC are more sophisticated and useful than those for PASC, ameliorating the continued rise in incidence. The large number of patients with PDAC has raised concern, leading to the development of preventive measures. The incidence of PASC is increasing much faster than expected; therefore, it should be given sufficient attention. The overall incidence rate of male is significantly higher than that of female, which may suggest that more emphasis should be placed on men when formulating prevention and surveillance strategies.
During the study period, IB mortality rate of PASC also showed a continuous upward trend, with an APC of 5.0%. This might imply that PASC therapy and intervention programs have also not improved significantly in recent years. Interestingly, the overall incidence of PDAC is rising, but the increasing trend has slowed down in recent years. This difference in IB mortality trends may indicate that the treatment methods and strategies for PDAC have made progress and improvement in recent years, as opposed to PASC. As a result, greater resources should be allocated to PASC as well as attempts to establish standardized and focused treatment procedures for these patients. Epidemiological findings could be used to guide both primary prevention through lifestyle modification and secondary prevention through the identification and targeted screening of high-risk groups with the goal of cancer prevention and early detection. It is important to stress that identification of populations at risk, linkageto-care, improvement in diagnostic tests and implementation of rigorous screening and surveillance strategies may facilitate early-stage diagnosis of PASC and improve response to currently available treatment options.
Depending on the number of patients with PASC included in the research, previous studies revealed varying survival outcomes, with the overall median survival ranging from 4 to 6 months. 21,26,27 Consistent with the results of Hester et al. our study found that PASC patients had an overall median survival of 6 months. 26 Moreover, we found no significant differences in OS and DSS between patients with PASC and PDAC. However, patients with PSCC have shorter OS and DSS than those with PASC and PDAC, despite the poor prognosis for all three diseases.
We discovered that localized and regional PASC significantly delayed OS and DSS compared with distant disease, which is consistent with earlier studies. 27 However, in our study cohort, patients with distant PASC accounted for the T A B L E 2 Multivariate Cox's proportional hazards model assessing factors associated with mortality after diagnosis of pancreatic adenosquamous carcinoma. majority. Furthermore, the incidence of distant diseases was significantly higher than that of local and localized diseases, indicating that early diagnosis of PASC was not optimistic. Moreover, according to our results, patients with PASC who underwent surgery had considerably longer OS and DSS than those who did not. However, many patients have an advanced stage when the disease is diagnosed and have lost the opportunity for surgical treatment due to complications or distant metastases. 10,14,27 The proportion of distal PASC in patients who underwent surgical therapy and survived was significantly lower than that in patients who did not receive surgical treatment in our study (17.5% vs. 70.0%, p < 0.001). Therefore, a breakthrough in the early diagnosis of PASC is the main direction of current efforts so that more patients can survive. Exploring comprehensive treatments for patients with advanced disease is also indispensable. At present, there is no exceptional clinical treatment for PASC. Local control and aggressive resection are the first choice treatments for patients with PASC. 28 In addition, chemotherapy and radiotherapy, as well as combination therapy, are increasingly being used for PASC.

OS
In clinical practice, adjuvant chemoradiotherapy is more common than adjuvant radiotherapy or adjuvant chemotherapy. Adjuvant chemoradiotherapy significantly improved the overall survival time of PASC patients compared with adjuvant chemotherapy or adjuvant radiotherapy. 29 It is important to note that PD-L1 immune checkpoint is only expressed in squamous carcinoma, and immune checkpoint therapy may be effective in patients with PASC with this characteristic. 14,30 However, Lee et al. reported that there was no significant difference in the survival of PASC patients between pd-l1 positive group and pd-l1 negative group, which may be related to the insensitivity of pancreatic cancer to immunotherapy alone. 14 Therefore, surgery is the preferred treatment for PASC patients, but adjuvant chemoradiotherapy and comprehensive targeted immunotherapy should also be provided. According to our analysis, patients with a tumor size >3.5 cm had a worse prognosis. A previous study has also shown that patients with tumors larger than 4 cm tend to have a shorter survival. 26 We discovered that patients with a tumor size >3.5 cm had a higher frequency of distant PASC than those with a tumor size ≤3.5 cm. In contrast, the proportion of localized disease was smaller in patients with tumors size >3.5 cm than in patients with tumors size ≤3.5 cm. These data imply that tumor size may have an impact on the biological activity of PASCs.
The nomogram has long been employed as an important prediction model for estimating individual survival. 31 According to our multivariate COX regression analysis, age at diagnosis, SEER stage, treatment, regional lymph node and tumor size were independent prognostic factors affecting OS and DSS of PASC patients. Based on these factors, we developed nomogram 1-year and 2-year OS and DSS prediction techniques. The higher C-index values reflected the good prediction discrimination of the nomogram models. In addition, the calibration curve between the anticipated and actual survival rates for the whole cohort is quite consistent. We may use this model to forecast the transition's survival probability and utilize it as a guide when making clinical decisions. This is the first research to use a large multi-center database to study the epidemiological trends of PASC. Despite a detailed analysis of trends in PASC incidence and IB mortality, and subgroup trends by age, grade, and tumor site, there are still some shortcomings in this study. A common problem with large multi-center databases is selection bias and missing data entry. 32 In addition, the SEER database does not provide information on serological tests and complications for a more comprehensive analysis.

| CONCLUSIONS
The incidence and IB mortality of PASC have been steadily increasing in recent years, manifesting that the preventive and treatment measures of PASC are not ideal. Therefore, we must identify the significance of this condition as soon as possible and commit greater attention and resources to it. This can help us further understand PASC and develop more targeted and standardized preventive and therapeutic measures. In addition, a nomogram model was developed to forecast the probability of survival of an outcome, which could guide clinical decision making. Further studies on the molecular characteristics of PASC are needed to better determine the biological characteristics and prognosis of the tumor.