Spatiotemporal trends in pancreatic cancer incidence from 2013 to 2023 in Gansu Province, Northwest China: A regional study

Abstract

BACKGROUND

Pancreatic cancer (PC) is a malignant neoplasm with a rising incidence, yet long-term provincial trends in China remain unclear.

AIM

To comprehensively analyze the spatiotemporal distribution of, and trends in, PC incidence in Gansu Province from 2013 to 2023.

METHODS

Data on PC cases were gathered from 262 hospitals in Gansu Province, focusing on age-standardized incidence rates (ASIRs) stratified by county, gender, and climate. Moran’s I was employed for spatial autocorrelation analysis to detect clustering patterns in PC incidence. SatScan was utilized to detect spatiotemporal clusters of PC, using a Poisson model as a basis. Joinpoint regression was used to evaluate temporal trends and determine the average annual percent change. Socioeconomic and healthcare indicators were examined using spatial regression analysis.

RESULTS

In 2013-2023, PC incidence in Gansu Province rose from 590 to 1017 cases, with the ASIR increasing from 2.39 per 100000 people to 3.23 per 100000 people. The ASIR was higher in males compared to females, and higher in the elderly than in the young people. Global spatial clustering highlighted hotspots in the Hexi region, and a high-risk region and two low-risk regions were identified. The provincial incidence exhibited an increasing trend. The ASIR of PC in Tianshui, Zhangye, Qingyang, Dingxi, and Linxia showed a continuous upward trend. The incidence of PC in the Monsoon climate of medium latitudes (Dwb climate zone) was high and rising. Area-level analyses suggested that socioeconomic development may partially contribute to spatial heterogeneity.

CONCLUSION

PC incidence increased in Gansu Province from 2013 to 2023, especially in the Hexi region and arid-cold areas, highlighting priority counties for targeted prevention, earlier diagnosis, and resource allocation.

Key Words: Pancreatic cancer; Spatiotemporal analysis; Incidence; Trend; China; Climate

Core Tip: Using province-level multi-center hospital surveillance data from 262 hospitals in Gansu Province, China from 2013 to 2023, we quantified pancreatic cancer incidence trends and mapped county-level spatiotemporal clusters. It was found that age-standarded incidence increased from 2.39 per 100000 people to 3.23 per 100000 population, with consistently higher rates in men and older adults. Spatial autocorrelation and SaTScan revealed persistent high-risk hotspots concentrated in the Hexi region, and an increasing burden in the Dwb climate zone. These results provide an evidence for targeted prevention, earlier diagnosis, and resource allocation in high-risk counties.

INTRODUCTION

Pancreatic cancer (PC) is a highly complex, difficult malignancy. According to the latest global cancer statistics, it is the 12^th most prevalent cancer, accounting for 2.6% of all cancer cases, but its mortality rate (4.8%) ranks 6^th[1]. PC is often diagnosed at an advanced stage due to the lack of effective early diagnostic methods, leading to a poor prognosis with a five-year survival rate of less than 10%[2]. Despite advances in surgery, chemotherapy and immunotherapy, treatment outcomes remain limited as most patients present with metastatic disease at diagnosis[3,4]. With the advancement of genetic research, there is a growing recognition that environmental exposures and lifestyle factors are also crucial contributors to the development and progression of PC[5]. These factors can vary significantly across different regions, contributing to spatial and temporal differences in cancer incidence. For instance, smoking, diet, chronic pancreatitis, and exposure to carcinogens, such as aflatoxins, are significant risk factors for PC and may explain some of the observed regional disparities in incidence. Consequently, it is essential to understand the spatiotemporal distribution of PC and identify high-risk populations for early intervention and targeted prevention.

In China, the incidence and mortality rates of PC are increasing[6], highlighting a growing public health concern. However, due to regional differences, the burden of PC varies significantly among provinces. These regional differences emphasize the need for localized healthcare strategies and a better understanding of their impact on cancer incidence. Moreover, analyzing the spatiotemporal distribution of PC will allow for more targeted interventions and resource allocation.

There are marked regional differences in the epidemiological characteristics of PC. The southeastern provinces of China[7,8] exhibit a significantly higher incidence compared to the northwestern provinces[9,10]. This disparity may be associated with variations in economic development, environmental pollutants, and the climatic and geographical characteristics of these regions. Gansu Province, situated in northwest China, features a variety of climate zones and landforms. Its unique geographical and climatic features could influence the risk of PC through multiple pathways. In addition, the dietary and living habits, such as smoking[11] and drinking[12], of residents of Gansu Province, and specific environmental exposures in some areas[13] (aflatoxin, etc.), may elevate the risk of PC. However, there is a paucity of research on the geographical distribution of PC within Gansu Province, and comprehensive epidemiological data and analysis are lacking.

Consequently, comprehensive analysis of the incidence trends and regional distribution characteristics of PC in Gansu Province is of great importance. Such analysis would significantly contribute to the formulation of targeted prevention strategies, the optimization of early screening protocols, and the advancement of clinical diagnosis and treatment methodologies.

Utilizing cancer registry data from Gansu Province, we analyzed spatiotemporal trends in PC incidence and identified high-risk regions from 2013 to 2023. Stratification was conducted according to age, gender, and climate zone to analyze changes in the incidence of PC. Our research might facilitate early detection of PC in Gansu Province.

MATERIALS AND METHODS

Study population and data processing

In this study, cancer registration data and hospital-based data from Gansu Province were collected from the Gansu Cancer Registration System between 2013 and 2023. The data were collected from a multi-source surveillance system involving 262 hospitals across all 87 counties and districts of the province, with an initial identification of 22099 records. The system integrates various data sources, including inpatient records, outpatient records, pathology reports, and discharge summaries. Patients diagnosed with PC as their first primary malignancy were included. PC cases were defined using the International Classification of Diseases for Oncology, 3^rd edition codes for C25.0-C25.9.

Data quality was assessed according to the Guidelines for Cancer Registration in China. Data cleaning and case selection followed a rigorous multi-step procedure, as detailed in Supplementary Figure 1. This procedure removed the majority of duplicate entries (n = 10054), resulting in 12045 unique, confirmed cases. These 12045 cases underwent manual review to verify the completeness of demographic information. Each record should contain the information including sex, year of diagnosis, and a registered residential address specified at the county or district level to ensure feasibility for subsequent spatial analyses. The final analytical cohort consisted of 12045 unique cases with complete information, representing 54.5% of the initial records. Of these cases, 59.02% were male and 40.98% were female. County-level age- and sex-specific population denominators were obtained from the Seventh National Population Census, available by county administrative code, sex, and age group. Because denominators were available as a 2020 census snapshot, annual rates for each calendar year were calculated using year-specific case counts with fixed 2020 denominators. Age was categorized into 18 subgroups (0-4, 5-9, 10-14, ..., 80-84, and ≥ 85 years). There are seven distinct climate types in Gansu Province based on the Köppen-Geiger climate classification (Supplementary Table 1).

Crude incidence rates and age-standardized incidence rates (ASIRs) were calculated using direct age standardization and expressed per 100000. For each stratum (county/district × sex × year), age-specific incidence rates were computed using the corresponding case counts and population denominators by age group, where population denominators were obtained from the Seventh National Population Census (2020) snapshot. ASIRs were then derived by applying age-group weights from the China 2010 standard population (based on the Sixth National Population Census). Detailed formulas and computational steps are provided in the Supplementary material. Rate smoothing was not applied for county/district estimates. Relative risk (RR) for each cluster and county was computed by SatScan (Version 10.3.3, Kulldorff M; http://www.satscan.org/).

Spatial-temporal aggregation analysis

Global spatial autocorrelation was assessed using Global Moran’s I. Spatial weights were defined at the county level using first-order queen contiguity and were row-standardized (style = “W”; zero. Policy = TRUE). Statistical significance was evaluated using a two-sided permutation test with 999 random permutations, and P < 0.05 was considered statistically significant. Local Moran’s I was used to identify local spatial clusters based on the same spatial weights, with significance assessed by 999 permutations. Significant counties were classified into high-high, low-low, high-low, and low-high categories. No multiple-testing correction (e.g., false discovery rate adjustment) was applied; therefore, local Moran’s I results were interpreted as exploratory and considered hypothesis-generating rather than confirmatory. The analysis was conducted, and the spatial data processed, using the ‘sf’ and ‘spdep’ R (version 4.5.1) packages, encompassing province boundaries and centroids.

Space-time cluster analysis

SaTScan™ (v10.3) was used to evaluate spatiotemporal clustering of PC in Gansu Province using a retrospective space-time scan statistic of the discrete Poisson model, with county-level annual case counts and corresponding populations at risk as inputs[14]. For space-time analyses, a cylindrical scanning window with a circular geographic base centered on county centroids was applied, with both the spatial radius and temporal duration allowed to vary up to prespecified maxima. The maximum spatial cluster size was set to 50% of the population at risk, and the maximum temporal cluster size was set to 50% of the study period. Statistical significance was assessed using 999 Monte Carlo replications, and P < 0.05 was considered statistically significant.

To examine robustness of the spatial window size, sensitivity analyses were performed using alternative maximum spatial windows of 10%, 20%, and 30% of the population at risk, respectively. Sex differences were examined using sex-stratified analyses, with separate Poisson space-time scans conducted for males and females using sex-specific annual case counts and corresponding sex-specific population denominators. In the overall analysis, age adjustment was implemented by providing age-stratified case and population counts (18 age groups) and specifying age group as a categorical covariate in the discrete Poisson model.

Joinpoint regression model

Joinpoint regression analysis, proposed by Kim in 2000, is widely used for analyzing temporal trends in diseases[15]. We used Joinpoint regression to assess temporal trends in the ASIR of PC in Gansu Province from 2013 to 2023, with the average annual percent change (AAPC) calculated using Joinpoint software (version 4.9). Statistical significance over periods was assessed using the Monte Carlo substitution method. The AAPC was utilized to analyze the temporal trend in the ASIR of PC. We assessed the statistical significance of the changing trend across different periods by comparing the AAPC to zero using its 95% confidence interval (CI). An increasing trend was defined when AAPC > 0 and the 95%CI did not include 0; a decreasing trend was defined when AAPC < 0 and the 95%CI did not include 0.

Socioeconomic and healthcare indicators and ecological regression analysis

Socioeconomic and healthcare data of Gansu Province’s 87 counties were compiled for 2003-2013 from the China County Statistical Yearbook. Variables included GDP per capita (CNY/capita/year), primary/secondary/tertiary industry output values (10000 CNY), and hospital beds per 10000 population. To represent long-term contextual conditions and reduce year-to-year fluctuations, these indicators were summarized as 2003-2013 averages and linked to county-level ASIR for 2013-2023. Industrial structure was expressed as the tertiary-industry share (tertiary output divided by the sum of primary, secondary, and tertiary outputs). Descriptive statistics are summarized in Supplementary Table 2. No missing values were observed for these county-level indicators.

To assess whether socioeconomic development and healthcare capacity were associated with PC incidence, we performed an ecological multivariable regression with log-transformed ASIR (2013-2023) as the dependent variable. GDP per capita and hospital beds were log-transformed, and the tertiary-industry share was included to represent industrial structure. Only this share variable was retained to reduce multicollinearity, as assessed by variance inflation factor.

After fitting the ordinary least squares (OLS) model, residual spatial autocorrelation was evaluated using a permutation-based Moran’s I test (moran.mc, 999 permutations) with a first-order queen contiguity, row-standardized spatial weights matrix (style = “W”, zero. Policy = TRUE). When residual spatial dependence was detected, we fitted spatial regression models (SEM: Errorsarlm; SLM: Lagsarlm) using the same weights and selected the preferred specification by Akaike information criteria. Analyses were conducted in R 4.5.1 using the sf, spdep, and spatialreg packages, with two-sided P < 0.05 considered statistically significant.

RESULTS

Increasing burden of PC in Gansu Province from 2013 to 2023

We analyzed the epidemiological characteristics of 12045 patients with PC in Gansu Province from 2013 to 2023. There were 7109 male patients (59.02%) and 4936 female patients (40.98%; male-to-female ratio of 1.44:1). The median age was 65.5 years, and the mean age of onset was 69.5 ± 13.3 years. Group 14 had the most patients (n = 2111), accounting for 17.52% of the total patients (Figure 1A). There was a rise in PC cases among individuals aged 40-60 and ≥ 60 years (Figure 1B). The mean and median age of PC onset varied between years in Gansu Province, with an upward trend from 2013 to 2023 (Figure 1C). The peak age of onset was at 50-80 years (Figure 1D). The ASIR exhibited an upward trend from 2013 to 2023, with the number of cases rising from 2.39 per 100000 people to 3.23 per 100000 people (Figure 1E). The same patterns were seen when the data were categorized by gender (Figure 1F). The ASIR increased for both genders, with a more significant rise observed in males.

Figure 1 The incidence of pancreatic cancer in Gansu Province from 2013 to 2023. A: Number of cases in age groups; B: Incidence trends across 3 age groups; C: The mean and median age in each year; D: Annual proportion of cases by age group; E: Trends in overall age-standardized incidence rate (ASIR) over time; F: Trends in male and female ASIR over time; G: Overall ASIR distribution; H: ASIR distribution in males; I: ASIR distribution in females. ASIR: Age-standardized incidence rate.

Distribution maps of overall ASIR showed elevated incidence rates in the Hexi region and lower rates in Longnan City (Figure 1G). Pingchuan District in Baiyin City had the highest ASIR for males and females, followed by Sunan Yugur Autonomous County of Zhangye City. The lowest ASIR in males was seen in Liangdang County in Longnan City (Figure 1H). The counties with the lowest ASIR for females were Liangdang County in Longnan City and Aksai Kazak Autonomous County in Jiuquan City (Figure 1I). From 2013 to 2023, there was a clear pattern of annual distribution of ASIR overall, and in males and females in counties of Gansu Province. Figure 2 illustrates that the ASIR of PC in Gansu Province was inclined to the Hexi region, with a notable upward trend extending toward Gannan Tibetan Autonomous Prefecture in the southeast and Jiuquan City in the northwest. The ASIR among males was consistent with the overall ASIR (Figure 3). However, among females, high ASIR was mainly seen in the Hexi region and northwest part of Gansu (Figure 4).

Figure 2 Distribution of annual age-standardized incidence rate of pancreatic cancer in Gansu Province from 2013 to 2023. A-K: Distribution map of age-standardized incidence rate in 2013 (A), 2014 (B), 2015 (C), 2016 (D), 2017 (E), 2018 (F), 2019 (G), 2020 (H), 2021 (I), 2022 (J), and 2023 (K).

Figure 3 Distribution map of male age-standardized incidence rate of pancreatic cancer in Gansu Province from 2013 to 2023. A-K: Distribution map of male age-standardized incidence rate in 2013 (A), 2014 (B), 2015 (C), 2016 (D), 2017 (E), 2018 (F), 2019 (G), 2020 (H), 2021 (I), 2022 (J) and 2023 (K).

Figure 4 Distribution map of female age-standardized incidence rate of pancreatic cancer in Gansu Province from 2013 to 2023. A-K: Distribution map of female age-standardized incidence rate in 2013 (A), 2014 (B), 2015 (C), 2016 (D), 2017 (E), 2018 (F), 2019 (G), 2020 (H), 2021 (I), 2022 (J), and 2023 (K).

Spatial shift in PC burden in Gansu Province

The global spatial autocorrelation of PC in Gansu Province was analyzed using Moran’s I (0.30, Z = 4.45, P < 0.05). Local hotspot analysis identified notable clustering within certain counties. As shown in Figure 5A, the central region of Gansu Province was a hotspot area with a relatively high incidence rate, while the southern region had a relatively low incidence rate and was the main cold-spot area. Subsequently, we analyzed male (global Moran’s I = 0.28, Z = 4.13, P < 0.05) and female (global Moran’s I = 0.21, Z = 3.21, P < 0.01) PC incidence. The spatial autocorrelation of global ASIR indicated a significant positive spatial correlation of ASIR between males and females in Gansu Province. According to the local Moran’s I statistics, Jinchang, Lanzhou, Baiyin, and Gannan Prefectures were significant high-high cluster areas for males (Figure 5B). Zhangye, Jinchang, Wuwei, and Baiyin were significant high-high clusters for females (Figure 5C). Longnan was a low-low cluster area for both male and female PC.

Figure 5 Spatial autocorrelation of age-standardized incidence rate for pancreatic cancer in the Gansu Province. A: Overall age-standardized incidence rate (ASIR); B: Male ASIR; C: Female ASIR. LISA: Local Moran’s I.

Using SaTScan software, we identified significant spatial clusters of PC in Gansu Province. In the baseline retrospective space-time analysis, the most likely high-risk cluster occurred during 2017-2021, comprising 39 counties (RR = 1.63, 95%CI: 1.57-1.69; log-likelihood ratio = 292.20, P ≤ 0.001) (Figure 6A). We conducted an additional analysis of the spatial distribution of PC in Gansu Province (Figure 6B). High-risk populations of PC, including male (Figure 6C) and female cases (Figure 6D), were predominantly located in Wuwei, Jinchang, Zhangye, and Baiyin, within the Hexi area. Sensitivity analyses using alternative maximum spatial windows (10%, 20%, and 30%) yielded largely consistent high-risk cluster locations compared with the 50% setting, with differences mainly in cluster extent (Supplementary Table 3). In the age-adjusted overall analysis, the most likely high-risk cluster remained in 2017-2021 and showed substantial overlap with the baseline cluster (36/39 counties, 92.3% overlap), indicating that the primary clustering pattern was robust in age adjustment (RR = 1.56, 95%CI: 1.50-1.62; log-likelihood ratio = 236.51, P ≤ 0.001) (Supplementary Table 4).

Figure 6 Spatial clusters of pancreatic cancer in Gansu Province. A: High and low-risk clusters; B-D: Distribution of overall (B), male (C), female (D) relative risk of pancreatic cancer in Gansu Province from 2013 to 2023.

Joinpoint regression analysis of temporal trends in the ASIR for PC

We calculated the AAPC of PC ASIR in all counties and districts of Gansu Province from 2013 to 2023 (Supplementary Table 5). Qingshui County (AAPC: 14.14, 95%CI: 3.74-26.51, P = 0.025) and Gangu County (AAPC: 8.94, 95%CI: 0.58-18.61, P = 0.035) in Tianshui City; Ganzhou District (AAPC: 16.66, 95%CI: 4.25-35.86, P = 0.010), Minle County (AAPC: 21.90, 95%CI: 1.49-45.36, P = 0.035), and Linze County in Zhangye City (AAPC: 13.30, 95%CI: 7.46-20.85, P = 0); Ning County (AAPC: 29.04, 95%CI: 19.35-41.11, P = 0) in Qingyang City; Longxi County (AAPC: 10.54, 95%CI: 3.62-18.07, P = 0.004) and Lintao County (AAPC: 13.92, 95%CI: 6.48-21.22, P = 0.001) in Dingxi City; and Linxia County (AAPC: 15.28, 95%CI: 8.94-22.52, P = 0) and Yongjing County (AAPC: 15.68, 95%CI: 5.87- 26.61, P = 0.001) in Linxia Prefecture showed an upward trend. Jinchuan District (AAPC: -7.7, 95%CI: -11.95 to -2.95, P = 0.001) in Jinchang City, and Baiyin (AAPC: -12.01, 95%CI: -20.97 to -1.61, P = 0.025) and Pingchuan District (AAPC: -22.91, 95%CI: -34.72 to -8.66, P = 0.001) in Baiyin City, showed a downward trend.

The AAPC in the ASIR analysis of male PC (Supplementary Table 6) showed that Ganzhou District (AAPC: 21.38, 95%CI: 10.21-33.68, P = 0.001) and Shandan (AAPC: 19.61, 95%CI: 2.21-39.98, P = 0.030) in Zhangye City; Tianzhu Tibetan Autonomous County (AAPC: 16.28, 95%CI: 6.83-26.57, P = 0.003) in Wuwei City; Zhang County (AAPC: 13.16, 95%CI: 5.92-20.88, P = 0.002), Longxi County (AAPC: 9.63, 95%CI: 2.86-16.85, P = 0.010), and Lintao County (AAPC: 10.32, 95%CI: 0.01-21.69, P = 0.049) in Dingxi City; Linxia County (AAPC: 13.08, 95%CI: 5.91-20.74, P = 0.002) and Yongjing County (AAPC: 18.04, 95%CI: 4.89-32.83, P = 0.011) in Linxia City; Zhenyuan County (AAPC: 12.89, 95%CI: 3.14-23.57, P = 0.014) in Qingyang City; and Zangjiachuan Hui Autonomous County (AAPC: 20.42, 95%CI: 0.34-44.52, P = 0.046) in Tianshui City experienced an upward trend. Minxian County (AAPC: -4.15, 95%CI: -7.24 to 0.96, P = 0.011) in Dingxi City, Anning District (AAPC: -7.93, 95%CI: -15.12 to 0.14, P = 0.047) in Lanzhou City, Kongtong District (AAPC: -12.03, 95%CI: -18.47 to -5.09, P = 0.004) in Pingliang City, and Pingchuan District (AAPC: -21.78, 95%CI: -32.59 to -9.25, P = 0.005) in Baiyin City showed a decreasing trend. An increasing temporal trend of ASIR in women (Supplementary Table 7) was observed in Ganzhou District, Zhangye City (AAPC: 23.14, 95%CI: 9.77-38.15, P = 0.003) and Longxi County, Dingxi City (AAPC: 13.14, 95%CI: 4.38-22.63, P = 0.007) and Linxia (AAPC: 16.13, 95%CI: 3.45-30.37, P = 0.011) and Yongjing County (AAPC: 15.39, 95%CI: 2.55-29.83, P = 0.022) of Linxia City. There were no counties with downward trend.

Spatiotemporal analysis of PC in Gansu Province based on climate zones

We analyzed the ASIR of PC overall, and in males and females, in Gansu Province from 2013 to 2023, in different climate zones. Gansu Province’s lowest overall ASIR was 0.018 per 100000, in the ET climate zone, while the highest ASIR was 1.424 per 100000, in the Dwb climate zone (Figure 7A). The distributions for males (Figure 7B) and females (Figure 7C) were consistent with the overall pattern. In summary, PC incidence, overall and in males and females, exhibited an upward trend within the Dwb climate zone, whereas the incidence remained relatively stable across other climate zones (Figure 7D-F). High-incidence areas for PC were mainly located in the BSk climate zone in Baiyin City and the BWk climate zone in Jinchang City, Wuwei City, and Jiuquan City (Figure 7G-I).

Figure 7 Incidence trend and distribution of pancreatic cancer by climate zone in Gansu Province. A-C: Overall (A), male (B), female (C) age-standardized incidence rate (ASIR) in different climate zones; D-F: Annual ASIR of overall (D), male (E) and female (F) in climate zones; G-I: Distribution map of overall (G), male (H) and female (I) ASIR. ASIR: Age-standardized incidence rate.

We further used Joinpoint to assess the AAPC of PC ASIR in each climate zone of Gansu Province from 2013 to 2023. The results indicated an increasing ASIR of PC within the Cwa (AAPC: 9.87, 95%CI: 4.90-16.61, P = 0.0004) and Dwb (AAPC: 5.02, 95%CI: 0.33-8.83, P = 0.039) climate zones. Specifically, there was an increasing trend in the ASIR of male PC in the Dwb (AAPC: 3.72, 95%CI: 0.32-6.34, P = 0.030) climate zone. The ASIR of PC among females within the Dwc climate zone showed an increasing trend (AAPC: 17.85, 95%CI: 5.22-30.28, P = 0.011).

Socioeconomic and healthcare correlates with county-level ASIR

Using county/district-level socioeconomic and healthcare indicators (2003-2013 averages), we fitted an area-level multivariable regression with log-transformed ASIR (2013-2023) as the outcome. In the OLS model including log(GDP per capita), log(hospital beds per 10000 population), and tertiary-industry share, GDP per capita was positively associated with ASIR (β = 0.157, 95%CI: 0.006-0.308; P = 0.042), whereas hospital beds and tertiary-industry share were not statistically significant (Table 1).

Table 1 The area-level multivariable regression analysis of socioeconomic and healthcare indicators.

Variable	β (95%CI)	SE	t	P value
Intercept	-0.682 (-2.959 to 1.594)	1.145	-0.596	0.553
log(GDP per capita)	0.157 (0.006-0.308)	0.076	2.070	0.042
log(hospital beds per 10000)	0.092 (-0.115 to 0.299)	0.104	0.882	0.381
Tertiary-industry share	-0.262 (-0.830 to 0.307)	0.286	-0.915	0.363

CI: Confidence interval.

Residual diagnostics indicated strong spatial dependence; OLS residuals exhibited significant spatial autocorrelation (permutation Moran’s I = 0.396, P < 0.001). Therefore, spatial regression models were fitted. The SEM provided better fit than the SLM and OLS (Akaike information criteria: 122.684 vs 128.906 and 149.590, respectively), and the spatial error parameter was significant (λ = 0.681, P < 0.001). In the SEM, log(GDP per capita) remained positively associated with ASIR (β = 0.207; P = 0.009), while hospital beds were not significant (P = 0.858); tertiary-industry share showed a marginal negative association (P = 0.098) (Supplementary Table 8). In the SLM, the spatial lag parameter was significant (ρ = 0.592, P < 0.001), but covariate associations were not statistically significant (Supplementary Table 9).

DISCUSSION

Using high-quality population-based registries in Gansu Province, this study revealed long-term patterns of PC incidence and provides a comprehensive overview of the current landscape. The findings revealed an increase in both incidence cases and ASIR in Gansu Province from 2013 to 2023. Furthermore, the ASIR of PC was higher in older age groups and males. These findings underscore the increasing impact of PC at the provincial level.

PC ranks among the most lethal cancers and contributes substantially to the global public health burden. The ASIR of PC has risen significantly worldwide in recent decades, paralleling rapid societal development[16-18]. The ASIR has increased worldwide in recent decades, with higher rates in very high Human Development Index countries than in less developed regions[1]. Comparative analyses across countries indicate substantial heterogeneity in PC burden. For example, Japan is projected to exceed an ASIR of 15 per 100000 by 2030, while the United States is around 10.3 per 100000, whereas South Korea has shown relatively stable trends in recent decades[19]. PC burden exhibited an upward trend in half of the included countries, with the highest mortality seen in China and Thailand, increasing by 2%-3% each year from 2013 to 2017[20-22]. This underscores the significant impact of PC in China. In 2021, China’s ASIR for PC was 5.64 per 100000 people[23]. The ASIR in Gansu Province was almost the same, at 5.38/100000. Although Gansu’s burden remains below that of highly developed countries, its alignment with China’s rising national trend underscores the need for strengthened prevention and early detection in the high-risk populations and regions identified in our study.

This study revealed notable regional differences in the incidence of PC across Gansu Province. High-ASIR areas were mainly concentrated in the Hexi region. Baiyin and the Hexi region (Jinchang and Wuwei) were also high-risk cluster areas. Longnan was a low-risk cluster area. PC incidence may be influenced by genetic or environmental factors. Saline-alkali soil is mainly distributed in the Hexi region, which contains diverse microbial resources[24]. Studies indicate a decrease in humic-like and high-molecular-weight fractions, with lower biodegradability, from northern to southern China, potentially linked to PC cases[25]. According to an epidemiological survey, drinking and smoking, which are risk factors for gastrointestinal tumors, are very common in the Hexi region[26,27]. The disparity in PC incidence between northern and southern regions may be attributed to variations in dietary patterns. Northern regions have a diet rich in fats, proteins, and calories. This dietary pattern is strongly associated with diabetes, a known risk factor for PC[28]. Vegetable intake may also be a factor for PC to some extent. A systematic review indicates an inverse relationship between the consumption of cruciferous vegetables and the risk of PC[29]. Moreover, there was a distinct difference between males and females in the present study. The ASIR among females was increased in the northwest region of Gansu Province, whereas high-ASIR areas for males were also distributed in the southern part of Gansu Province, including Gannan and Linxia Prefectures. To further contextualize the observed spatial heterogeneity, we incorporated county-level socioeconomic development and healthcare capacity into an area-level regression framework. Socioeconomic development showed a positive association with PC incidence, whereas healthcare capacity (as measured by hospital beds) did not explain the spatial variation. Residual spatial dependence remained evident after adjustment, and spatial regression models provided a better fit, suggesting that additional unmeasured, spatially structured determinants (e.g., environmental and behavioral factors) may contribute to the persistent geographic disparities.

Life’s Essential 8 is a cardiovascular health assessment tool comprising eight components: Physical activity, diet, nicotine exposure, sleep, body mass index, blood pressure, blood glucose, and non-high-density lipoprotein. A large prospective cohort study found an association between Life’s Essential 8 and PC risk[30]. This indicates that there might be a relationship between cardiovascular health and the occurrence of PC. A cross-sectional study conducted in Gansu Province between 2017 and 2022 revealed a higher prevalence rate in the Hexi region compared to other areas[31]. In Gansu Province, being male, a smoker, and overweight or obese are significant risk factors for cardiovascular disease. These data suggest that under the same geographical and environmental conditions, the risk factors for cardiovascular diseases are similar to those for PC. Thus, increased allocation of medical resources and implementation of preventive measures may be necessary in high-risk areas. Additionally, it is imperative to conduct screenings for high-risk populations, mitigate risk factors, and enhance cardiovascular health.

This study evaluated temporal trends in the incidence of PC via AAPC analysis. The findings indicated a rising incidence of PC in Gansu Province from 2013 to 2023, potentially attributable to a confluence of environmental, lifestyle, and genetic factors. The global prevalence of these factors is on the rise, contributing to an increased age-adjusted incidence of PC. Studies indicate that diabetes[32,33] and obesity[34] may contribute to higher rates of PC. Economically developed nations experience a higher burden of PC, primarily due to the higher prevalence of obesity, diabetes, smoking, and alcohol consumption in high-income countries. Furthermore, mortality rates for PC linked to high body mass index and elevated fasting plasma glucose have risen across various countries and age groups[35]. A healthy lifestyle and balanced diet may significantly influence the development of PC. In 2019, the ASIR of PC was 5.64 per 100000 in Jiangsu Province[36], while that in Gansu Province was 4.12 per 100000 people. Regional development may explain this difference. Prevention of PC remains a serious challenge.

Gansu Province, with its unique climate features, is located in the arid-to-semi-arid region of northwestern China[37]. Climate zones are differentially related to the cancer epidemic due to differences in temperature, relative humidity, wind speed, and sunshine[38]. For example, the prevalence of aflatoxin may be affected by climate change, which resulted in the difference in the incidence of liver disease[39]. Our study identified specific climatic patterns correlated with PC incidence. Notably, the incidence was relatively high in arid and cold climate zones. A pro-inflammatory diet could elevate the risk of PC[40]. Within cold climates, diets often include foods that are high in calories and fat, while the availability of fresh fruits may be restricted, thereby elevating the risk of cancer.

This study has several limitations. First, we were unable to include some established risk factors (e.g., environmental pollution, obesity, smoking, and dietary habits) because county-level data were not available. Second, temporal analyses were limited to 2013-2023 because earlier records were incomplete. A longer time series would better capture potential inflection points in incidence trends. Third, cases among Gansu residents diagnosed or treated outside the province and those managed in non-participating settings may not have been fully captured, potentially introducing under-ascertainment and selection bias and affecting the representativeness of incidence estimates. Fourth, while we compared our results with published national and international estimates, we lacked comparable raw datasets for unified reanalysis, and absolute ASIR values may vary across sources due to differences in ascertainment and standardization. Future multi-region studies with longer follow-up and integrated exposure data are warranted.

CONCLUSION

The rising provincial burden of PC is anticipated to continue, highlighting the necessity for ongoing research and public health interventions. Targeted prevention and treatment strategies are crucial to ensure healthier and safer outcomes for an expanding and aging population.