Risk of prostate cancer associated with long-term air pollution exposure: a nationwide cohort study in Taiwan

Abstract

Background and objective

Prostate cancer is a critical health issue, particularly in developed countries. Emerging evidence suggests that air pollution is associated with increased risk of prostate cancer. This study aimed to evaluate the impact of long-term exposure to seven specific air pollutants on prostate cancer risk in Taiwanese men and establish dose-response relationships to guide prevention strategies.

Methods

This retrospective cohort study analyzed the data of 425,916 men from Taiwan’s National Health Insurance Research Database (2000–2013). Seven pollutants (SO₂, CO, PM₁₀, PM_2.5, NO_x, NO, and NO₂) were assessed and Cox regression models were adjusted for confounders. Prostate cancer incidence was the primary outcome, with significance set at P < 0.05.

Key findings and limitations

Increased exposure to air pollutants was associated with a 39–106% higher risk of prostate cancer per 1 standard deviation increase in pollutant levels. However, potential misclassification of exposure is a key limitation.

Conclusions and clinical implications

In over 400,000 Taiwanese men, higher levels of several air pollutants were associated with a greater likelihood of developing prostate cancer. These findings suggest that air pollution may be an important environmental factor in prostate cancer etiology. Further research is needed to confirm these associations before informing clinical practice or public health policy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-25229-z.

Introduction

Prostate cancer is the second most frequently diagnosed cancer in men and the fifth leading cause of cancer-related deaths worldwide, with a higher prevalence in developed countries [1]. The incidence of prostate cancer increases significantly with age, particularly in men aged >65 years. In Asia, the age-standardized incidence rates of prostate cancer from 2008 to 2012 ranged from 6.1 to 59 per 100,000 people [2]. Key risk factors for prostate cancer include advanced age, family history, ethnicity, obesity, and lifestyle factors [3]. Men with a first-degree relative diagnosed with prostate cancer are two to three times more likely to develop the disease, suggesting a strong genetic component.

The pathophysiology of prostate cancer involves a complex interplay between genetic mutations, inflammatory processes, and molecular alterations. Chronic inflammation, often triggered by infections or hormonal changes, is a critical factor in the development and progression of prostate cancer. Inflammatory cells release cytokines and reactive oxygen species (ROS), leading to DNA damage and creating a tumorigenic environment [4, 5]. The androgen receptor (AR) pathway is also significant, with mutations, overexpression, and splice variants of AR contributing to cancer progression [6]. Additionally, the epithelial-to-mesenchymal transition (EMT), driven by signaling pathways such as TGF-β, Wnt, and PI3K/AKT, enables tumor cells to gain migratory and invasive properties, contributing to metastasis and therapeutic resistance. Key EMT regulators and transcription factors promote an androgen-independent, aggressive phenotype, especially under anti-androgen pressure [7]. Environmental factors, such as particles less than 2.5 micrometers in diameter (PM_2.5), may induce the generation of ROS, which in turn can initiate EMT, further linking external exposure to the progression of aggressive prostate cancer [8, 9]. These molecular mechanisms could link environmental factors to the EMT process and progression of aggressive prostate cancer.

Air pollution is a significant global health issue, with over 99% of the world’s population exposed to pollutant levels exceeding the World Health Organization’s guidelines [10]. Although air pollution is well documented for its role in various health problems, including respiratory, cardiovascular, and neurological diseases and infertility [11–13], some studies have also suggested that pollutants such as PM_2.5 and nitrogen dioxide (NO₂) may increase the risk of prostate cancer [14, 15]. A recent systematic review and meta-analysis revealed that exposure to PM₂.₅ and NO₂, is associated with a 5% and 4% increased risk of prostate cancer, respectively [16]. These findings suggest a potential epidemiological link between ambient air pollution and prostate cancer risk. However, the effects of other air pollutants have not yet been thoroughly examined. Therefore, we conducted this retrospective study to investigate the effects of long-term exposure to seven air pollutants on the risk of prostate cancer in Taiwanese men. Our study explored the impact of seven air pollutants—sulfur dioxide (SO₂), carbon monoxide (CO), particles less than 10 micrometers in diameter, (PM₁₀), PM_2.5, nitrogen oxides (NO_X), nitrogen monoxide (NO), and NO₂ — over a decade-long period among Taiwanese individuals. This study aimed to explore potential associations between air pollution and prostate cancer, which could guide the development of effective preventive strategies and treatment options.

Methods

Study population

This study utilized data from Taiwan’s National Health Insurance Research Database (NHIRD), a comprehensive repository established in 1995 that contains the medical records of nearly all Taiwanese residents. The NHIRD encompasses a wide range of patient information, including gender, birth date, employment status, diagnosis times, medication usage, and treatment duration, using ICD-9-CM diagnosis codes [17]. For our analysis, we extracted patients’ baseline information from the Longitudinal Health Insurance Database 2000 (LHID2000), a subset of the NHIRD containing data on one million randomly selected patients. The Institutional Review Board (IRB) Committee of Changhua Christian Hospital in Taiwan granted ethical approval for this study (IRB number: 240310) and waived the requirement for informed consent due to the use of anonymized data.

Our study cohort comprised of all male patients between January 1, 2000, and December 31, 2013. To ensure data integrity, specific exclusion criteria were applied. We excluded subjects with unreasonable age entries (n = 0), those diagnosed with prostate cancer or metastatic prostate cancer before the follow-up period (n = 351), individuals whose follow-up start and end dates were identical (n = 573), those with less than ten years of air pollution exposure records (n = 11237), and those with a follow-up period of less than one year (n = 7007). Each participant was required to have at least ten consecutive years of valid exposure data prior to cohort entry (index date), followed by a minimum of one year of follow-up. The follow-up period began immediately after the exposure window and continued until prostate cancer diagnosis, death, withdrawal from the NHI system, or the end of the study (December 31, 2013). Thus, exposure and follow-up periods were consecutive but non-overlapping.

Exposure assessment

To investigate the potential link between long-term exposure to air pollutants and prostate cancer incidence, we conducted a detailed analysis over the past decade. We measured the concentrations of seven air pollutants: SO₂, CO, PM₁₀, PM_2.5, NO_X, NO, and NO₂. These pollutants were continuously monitored by stations operated by Taiwan’s Environmental Protection Administration (TEPA) from July 1, 1993, to December 31, 2013. Exposure periods were defined as the ten years immediately preceding each participant’s follow-up endpoint, which served as the index date [18]. For each pollutant, we computed the cumulative daily average concentration over the entire ten-year exposure-up period. To facilitate our analysis, we categorized pollutant exposure levels into tertiles, with cut-off points at the 33.3th and 66.6th percentiles. We meticulously linked the pollutant exposure data to the participants’ residential postal codes obtained from their insurance registration records, accounting for any changes in residence during the study period, to ensure accurate exposure assessment.

Outcome

The primary outcome of our study was the incidence of prostate cancer. Each participant was monitored until they were diagnosed with prostate cancer according to the ICD-9-CM. Participants who died, withdrew from the NHI system, or reached the end of the study period (December 31, 2013) without a diagnosis were censored. Survival duration was recorded in months.

Covariates

To account for potential covariates, we extracted and defined various variables from the NHIRD. These included age, level of urbanization (classified based on participants’ residential location at the beginning of the follow-up period), and insurance amount (averaged over the air pollution exposure assessment period) [19]. These covariates were selected based on prior epidemiological evidence and biological plausibility, as they are recognized as established risk or prognostic factors for prostate cancer and have also been linked to long-term exposure to air pollution in previous studies. We also incorporated ambient temperature, and the lag0-1 factor (2-day moving average of air pollutant levels prior to the primary endpoint) into our analysis to isolate the long-term impacts of air pollution exposure, which was the primary focus of our investigation. This approach was supported by the largest short-term effect estimates reported in the literature [20, 21].

Furthermore, we adjusted for several comorbidities based on ICD-9-CM [22]. Adjustments were made for a variety of demographic characteristics, and comorbidities, including alcohol-related disease, and metabolic disorders. All comorbidities were assessed prior to the exposure period, and the corresponding ICD-9-CM codes are provided in the Supplementary Methods. We defined medication usage as the prescription of one or more drugs for 90 days or more, aligning with the practices of Taiwan National Health Insurance system.

Statistical analyses

Chi-square tests and one-way ANOVA were used to compare prostate cancer incidence across tertiles of pollutant exposure. Post-hoc tests were conducted following significant ANOVA results. Cumulative incidence curves were plotted for prostate cancer across these groups, and log-rank tests were used to assess the differences. To quantify the association between pollutant exposure and prostate cancer risk, hazard ratios (HRs) were calculated for each standard deviation (SD) increase in exposure over a decade using Cox regression models adjusted for confounding variables. Age-stratified analyses were conducted to assess whether the association between air pollution and prostate cancer risk differed by age group, particularly among individuals aged ≥ 60, given the higher incidence in this population in Asia [23]. The proportional hazards assumption was assessed using a log-log Kaplan-Meier plot, and it was found that no pollutants contradicted this assumption. Pearson correlation coefficients among the seven pollutants were calculated to assess collinearity, and are presented in Supplementary Table 8. In multivariate models, we included only those co-pollutants with low intercorrelation (r < 0.3) to avoid multicollinearity, as demonstrated in Supplementary Tables 9–15. All statistical tests were two-sided with significance set at P < 0.05. We also reported the direction, magnitude, and precision of the effect estimates using hazard ratios and their corresponding 95% confidence intervals. Analyses were performed using the MetaTrial Research Platform developed by the Biomedica Corporation in Taipei, Taiwan. This thorough analytical approach enabled us to robustly assess the potential link between long-term air pollution exposure and prostate cancer risk in our study population.

Results

Study population and sample selection

Our study initially included 882,391 patients with complete data, from which we identified 439,846 male individuals. After applying the exclusion criteria, we excluded 13,930 subjects, resulting in a final cohort of 425,916 patients for analysis. Figure 1 illustrates the selection process.

Fig. 1.

Patient selection flowchart

Demographic characteristics and descriptive statistics

The demographic characteristics of the study cohort are summarized in Table 1. Among the 425,916 individuals, we observed 2,275 (0.53%) with prostate cancer. The mean age of the cohort was 45.03 (± 19.05) years. Most participants resided in highly urbanized areas and fell into the category with the lowest insurance amount. The age distribution showed that 42.72% (181,964) were younger than 40 years, 33.67% (143,403) were between 40 and 59 years, and 23.61% (100,549) were aged 60 years or older. We stratified the cohort into tertiles based on pollutant exposure levels, with detailed information about the distribution of population characteristics by exposure tertiles provided in Supplementary Tables 1 to 7. Air pollutant exposure levels varied significantly across population characteristics. Higher exposure groups tended to have younger age, higher urbanization level and lower income.

Table 1.

Cohort baseline characteristics

Characteristics	Number (%)
Prostate cancer	2,275 (0.53)
Age (years)
< 40	181,964 (42.72)
40–59	143,403 (33.67)
≥ 60	100,549 (23.61)
Mean ± SDa	45.03 ± 19.50
Urbanization level
1 (highest)	234,528 (55.06)
2	144,338 (33.89)
3	30,666 (7.20)
4 (lowest)	4,086 (0.96)
Unknown	12,298 (2.89)
Insurance amount (New Taiwan Dollar)
Financially dependent	4,117 (0.97)
1 to 19,999	180,289 (42.33)
20,000 to 39,999	112,997 (26.53)
≥ 40,000	76,469 (17.95)
Unknown	52,044 (12.22)

^aSD Standard deviation

Table 2 presents the mean values, standard deviations, and tertile distributions of each pollutant over the 10-year exposure period. The average concentrations were as follows: SO₂ (4.16 ppb), CO (0.53 ppm), PM₁₀ (54.62 µg/m³), PM_2.5 (32.68 µg/m³), NO_x (25.42 ppb), NO (7.39 ppb), and NO₂ (18.03 ppb).

Table 2.

Mean concentrations and distribution of air pollutants during the exposure period

	SO2(ppbc)	CO (ppmd)	PM10(µg/m3)	PM2.5 (µg/m3)	NOX (ppbc)	NO (ppbc)	NO2(ppbc)
Mean	4.16	0.53	54.62	32.68	25.42	7.39	18.03
SD	1.16	0.11	9.03	6.17	6.98	3.78	3.58
T1 cutoffa	3.59	0.47	48.57	27.99	21.03	4.82	16.17
T2 cutoffb	4.04	0.58	56.78	35.12	28.72	7.92	20.17

^aT1: 33.3rd percentile cutoff

^bT2: 66.6th percentile cutoff

^cppb: parts per billion

^dppm: parts per million

Association between air pollutant exposure and prostate cancer risk: Dose-Response analysis

Table 3 presents prostate cancer incidence across tertiles of average pollutant exposure, expressed as cases per 100,000 person-years. Incidence rates increased across tertiles for all seven pollutants. Post-hoc comparisons (A, B, C) were statistically significant (P < 0.001) for each pollutant. Figure 2 graphically portrays the cumulative incidence curves, revealing significant variations in the relationship between each air pollutant and the incidence of prostate cancer, elucidating both directionality and strength of association.

Table 3.

Prostate cancer incidence by exposure levels

Pollutants	Tertiles of average daily exposure, number of cases/person-years (incidence rate per 100,000 Person-Years)			P values	Significance in post-hoca	Total, n (%)
	T1 (lowest)	T2	T3 (highest)
SO2	431/1,838,815(23.4)	763/1,696,562 (45)	1081/1,893,197 (57.1)	< 0.001	A, B, C	2275/425,916 (0.53)
CO	343/1,853,986 (18.5)	560/1,829,809 (30.6)	1372/1,744,779 (78.6)	< 0.001	A, B, C	2275/425,916 (0.53)
PM10	610/1,837,666 (33.2)	771/1,815,441 (42.5)	894/1,775,468 (50.4)	< 0.001	A, B, C	2275/425,916 (0.53)
PM2.5	483/1,835,720 (26.3)	642/1,784,245 (36)	1063/1,801,785 (59)	< 0.001	A, B, C	2188/424,165 (0.52)
NOX	413/1,844,240 (22.4)	783/1,805,352 (43.4)	1079/1,778,982 (60.7)	< 0.001	A, B, C	2275/425,916 (0.53)
NO	387/1,841,879 (21)	755/1,809,132 (41.7)	1133/1,777,563 (63.7)	< 0.001	A, B, C	2275/425,916 (0.53)
NO2	421/1,843,890 (22.8)	713/1,814,346 (39.3)	1141/1,770,338 (64.5)	< 0.001	A, B, C	2275/425,916 (0.53)

^aIndicating the significant groups shown in post-hoc analysis, A: T1 vs. T2; B: T1 vs. T3; C: T2 vs. T3

Fig. 2.

Cumulative incidence curves of prostate cancer by pollutant exposure tertiles. A: SO₂. B: CO. C: PM₁₀. D: PM_2.5. E: NO_x. F: NO. G: NO₂

Table 4 presents the HRs of prostate cancer associated with a 1 SD increase in average exposure to each pollutant. Across all pollutants, we observed a consistent positive association with prostate cancer risk. Specifically, a 1 SD increase in exposure to SO₂, CO, PM₁₀, PM_2.5, NO_x, NO, and NO₂ was associated with a 62%, 106%, 39%, 77%, 105%, 80%, and 101% higher hazard of prostate cancer, respectively.

Table 4.

Risk of prostate cancer linked to long-term exposure to air pollutants: hazard ratios corresponding to a 1 standard deviation increase over a 10-year period

Pollutants	Age groups	Adjusteda HRb (95% CIc)	P values	SD
SO2	Total	1.62 (1.54,1.70)	< 0.001	1.16 ppbd
	Age ≥ 60	1.58 (1.49,1.67)	< 0.001
CO	Total	2.06 (1.99,2.15)	< 0.001	0.11 ppme
	Age ≥ 60	1.97 (1.88,2.06)	< 0.001	0.11 ppme
PM10	Total	1.39 (1.30,1.48)	< 0.001	9.03 µg/m3
	Age ≥ 60	1.35 (1.26,1.46)	< 0.001
PM2.5	Total	1.77 (1.66,1.88)	< 0.001	6.17 µg/m3
	Age ≥ 60	1.59 (1.49,1.70)	< 0.001
NOX	Total	2.05 (1.96,2.15)	< 0.001	6.98 ppbd
	Age ≥ 60	1.96 (1.85,2.07)	< 0.001
NO	Total	1.80 (1.72,1.87)	< 0.001	3.78 ppbd
	Age ≥ 60	1.73 (1.64,1.81)	< 0.001
NO2	Total	2.01 (1.91,2.12)	< 0.001	3.58 ppbd
	Age ≥ 60	1.96 (1.83,2.09)	< 0.001

^aCox regression models were adjusted for age, sex, urbanization level, insurance amount, comorbidities, ambient temperature, and lag0-1

^bHR: hazard ratio

^cCI: confidence interval

^dppb: parts per billion

^eppm: parts per million

Co-adjusted analyses showed changes in HRs among pollutants in Supplement Tables 9, 10, 11, 12, 13, 14 and 15. For SO₂ (single-pollutant HR: 1.62), the HR decreased to 1.29 after adjusting for CO, 1.31 after adjusting for NO_x, and 1.50 after adjusting for NO.

Discussion

This study found that prolonged exposure to SO₂, CO, PM₁₀, PM_2.5, NO_X, NO, and NO₂ was associated with a higher risk of prostate cancer. The most pronounced association was observed for CO, where a one standard deviation increase in exposure corresponded to a 106% higher hazard of prostate cancer. These associations were consistent in the overall population and among men aged 60 years and older. Co-adjusted analyses revealed complex interrelations among pollutants, such as mutual HR increases between CO and PM_2.5 suggesting negative confounding, and mutual HR decreases between CO and SO² indicating collinearity. Divergent effects between PM_2.5 and NO₂ further highlight the need for multipollutant models to account for correlated exposures. While these results suggest a potential link between air pollution and prostate cancer risk, the observational nature of the study precludes any definitive conclusions about causality.

Probable mechanisms

While the exact underlying mechanisms remain elusive, several hypotheses have been proposed to explain the observed association between exposure to air pollutants and increased prostate cancer risk.

Exposure to air pollution, particularly in urban and industrial areas where pollutant concentrations are generally higher due to dense traffic, industrial activities, and energy production, has significant biological effects. It alters the expression of genes involved in DNA damage, repair, inflammation, and oxidative stress response [24, 25]. For instance, particulate matter generates ROS through various mechanisms, including direct surface reactions and activation of inflammatory cells. Oxidative stress contributes to DNA damage and cancer development [26, 27]. In prostate cancer, elevated ROS levels are observed, driven by increased activity of the NADPH oxidase system. Studies have shown that inhibition of ROS production in prostate cancer cells reduces their proliferation and invasiveness [4]. This suggests that air pollution-induced ROS production promotes the initiation, progression, and dissemination of prostate cancer.

Emerging evidence suggests that components of air pollution, such as diesel exhaust particles (DEPs), exhibit anti-androgenic effects that may contribute to prostate cancer development [28, 29]. DEPs can inhibit AR activity by blocking its ligand-binding domains, preventing the proper transcriptional regulation necessary for normal prostate cell function. This disruption of AR signaling may lead to tumorigenesis, as prostate cancer progression is closely linked to AR activity [30–32].

In addition to DEPs, NO₂ and NO_x—common traffic-related pollutants—can induce oxidative stress and DNA damage. Chronic NO_x exposure has been linked to genetic instability and a pro-tumorigenic secretory phenotype in prostate cells. NO_x may also activate NADPH oxidase enzymes, increasing ROS production, which can enhance tumor invasiveness and progression [33]. SO₂, primarily from industrial emissions, may also play a role in PCa carcinogenesis. Animal studies have demonstrated that SO₂ exposure can disrupt testosterone biosynthesis by activating the ERK1/2 signaling pathway and impairing autophagy in Leydig cells, leading to hormonal imbalances that may influence prostate carcinogenesis [34]. While direct epidemiological evidence linking SO₂ exposure to increased prostate cancer risk remains limited, these findings suggest that SO₂-induced hormonal disruptions could potentially contribute to prostate cancer development. CO, a byproduct of incomplete combustion, may reduce oxygen delivery and contribute to tissue hypoxia—a condition known to promote tumor progression via hypoxia-inducible pathways [35]. Although direct evidence linking CO to prostate cancer remains limited, its role in hypoxia-related pathways warrants further investigation.

Recent studies have identified a connection between air pollution, specifically PM_2.5, and tumorigenesis mediated by EMT [8]. PM_2.5 exposure induces oxidative stress, leading to the activation of pathways such as TGF-β/SMAD, which are pivotal in EMT regulation and contribute to the initial stages of tumor development [8, 36]. In prostate cancer, these environmental pollutants may initiate tumorigenesis by altering the cellular microenvironment, promoting EMT, and fostering conditions favorable for malignant transformation [7, 37].

Comparison to previous study

Previous studies have explored the link between ambient air pollution and prostate cancer, including a Montreal case–control study focusing on NO₂ with a small sample and no follow-up period [14], and Huang et al.’s analysis of PM₁₀ in Taiwan without other pollutants [38]. More recent U.S. and Canadian studies reported associations with long-term PM_2.5 and NO₂ exposure [15, 39]. The UK Biobank cohort analyzed multiple pollutants but was limited by its participant age range (40–69 years) [40]. Our findings are consistent with previous reports linking PM_2.5, PM₁₀, and NO₂ to prostate cancer risk. To address these limitations, our study is more comprehensive, analyzing a wide range of pollutants over a period exceeding ten years in a nationwide cohort that includes individuals of all ages. This larger sample size and extended duration provided a more representative and realistic reflection of the true situation.

Strengths

Our study has several strengths that contribute to the reliability and validity of our results. First, it covered a nationwide population with a large and varied sample of 425,916 participants. Second, this research spans a decade, allowing us to evaluate the long-term effects of exposure. Third, this study examined the impact of seven major air pollutants, providing a comprehensive understanding of their effects. Fourth, we stratified our analysis by age, demonstrating that our findings were consistent across both the entire population and those over 60 years of age. Finally, we adjusted for a wide range of important confounders, including temperature, socioeconomic factors (insurance amount and urbanization level), and comorbidities like metabolic disorders. Together, these strengths reinforce the reliability of our findings and support the robustness of our conclusions.

Limitations

In conducting this research, several limitations warrant consideration. First, using health insurance registration addresses to estimate air pollution exposure may lead to exposure misclassification if the registered address does not reflect actual residence. Although residential changes were accounted for when possible, this limitation remains. Additionally, exposure assessment did not capture time spent in other environments such as workplaces or commuting, but this misclassification is likely non-differential and would attenuate observed associations. We adjusted for key covariates; however, unmeasured confounders such as genetic predisposition and dietary factors, unavailable in our dataset, may bias results if related to both exposure and prostate cancer risk. While co-pollutants with low correlations were adjusted for in multipollutant models, high correlations among some pollutants limited assessment of independent effects. Finally, potential selection bias—particularly collider bias—may arise since both air pollution exposure and prostate cancer affect healthcare utilization. Conditioning on inclusion in the health database could induce spurious associations; although the magnitude of this bias is uncertain, we acknowledge it as a limitation in interpreting our findings.

Conclusions

In this nationwide retrospective cohort study, we observed that higher long-term exposure to air pollutants—including SO₂, CO, PM₁₀, PM_2.5, NO_x, NO, and NO₂—was statistically associated with an increased hazard of prostate cancer among Taiwanese males. These findings highlight the importance for clinicians to consider environmental risk factors when diagnosing and advising patients with prostate cancer. Moreover, our study provides strong evidence that can help governments instill policies aimed at reducing pollution and mitigating its impact on public health, thereby alleviating the burden on healthcare systems. Future research incorporating well-adjusted confounders should further investigate the intricate relationship between air pollution and prostate cancer along with the underlying pathophysiological mechanisms.

Authors’ contributions

conception and design: Y.-M.S. acquisition of data: Y.-P.C.analysis and interpretation of data: Y.-M.S., S.-H.C., Y.-P.C.drafting of the manuscript: Y.-M.S.critical revision of the manuscript for important intellectual content: S.-H.C., P.-H.C., Y.-P.C.statistical analysis: H.-W.Z., H.-C.P.obtaining funding: P.-H.C., Y.-P.C.administrative, technical, or material support: C.-H.C., Y.-J.K., S.-J.C., Y.-P.C., C.H. C., Y.-S. L. supervision: P.-H.C., Y.-P.C.

Funding

The authors are grateful to Wan Fang Hospital (Grant numbers 1111-wf-eva-26) for financially supporting this research.

Data availability

The data utilized in this research, including analyzed datasets, can be obtained by contacting the corresponding author. Requests will be considered and fulfilled if deemed reasonable and in line with data sharing policies.

Declarations

Ethics approval and consent to participate

This study received approval from the Institutional Review Board at Changhua Christian Hospital in Taiwan, with the approval number 240310. The requirement for informed consent was waived because the claims data were deidentified prior to analysis. The National Health Research Institute, which provided the database for research purposes, had previously encrypted all personal identifiers to ensure confidentiality.

Competing interests

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.