Abstract
Background:
Data regarding malignancy risk amongst rheumatoid arthritis (RA) patients using biological and targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs) was conflicting, and real-world Taiwanese data was lacking.
Objectives:
To compare the malignancy risk and to investigate factors associated with malignancies amongst RA patients initiating b/tsDMARDs.
Design:
Nationwide, population-based, retrospective cohort study.
Methods:
We conducted a retrospective cohort study using the Taiwan National Health Insurance Research Database from 2001 to 2020. Adult patients with newly diagnosed, biologic-naïve RA were included if they initiated b/tsDMARDs, and those with prior malignancies were excluded. The primary outcome was the first diagnosis of malignancies after b/tsDMARD initiation. Time-dependent Cox proportional hazards models were performed to evaluate malignancy risk across different b/tsDMARDs and to identify independent risk factors.
Results:
A total of 8732 adult RA patients were included. Over both 5-year and extended 18-year follow-up periods, there was no statistically significant difference in malignancy risk amongst the four treatment groups (tumour necrosis factor inhibitors (TNFis), tocilizumab, abatacept and tsDMARDs). Independent associated factors included male gender, older age at b/tsDMARD initiation, chronic lung diseases including chronic obstructive pulmonary disease and higher-dose corticosteroid prescription. Use of conventional DMARDs and more frequent ambulatory follow-up were associated with a lower risk.
Conclusion:
In this Taiwanese nationwide RA cohort, b/tsDMARD class demonstrated no difference in malignancy risk within 5 years after their initiation. Clinicians should instead focus on individual associated factors, particularly male gender, older age at b/tsDMARD initiation, chronic lung diseases and higher-dose corticosteroid exposure, while optimising the protective role of conventional DMARDs and regular follow-up for better disease control.
Introduction
Rheumatoid arthritis (RA) is one of the most prevalent diseases worldwide, and its complications are diverse. Notably, patients with RA face a higher risk of malignancies compared with the general population, which underscores the importance of investigating these comorbidities and constitutes an essential part of holistic care of RA by rheumatologists.1,2
Over the past two decades, biological and targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs) have become key components in the treatment of active RA. Previous studies have explored the potential association between these therapies and malignancy risk. 3 For instance, tumour necrosis factor inhibitors (TNFis) have generally not been linked to an increased risk of malignancies, and tocilizumab has also not shown an elevated risk. 4 Nevertheless, several studies have indicated that abatacept may be associated with a higher risk of malignancies than TNFis.5,6 Furthermore, the risk of malignancies with tofacitinib remains controversial 7 ; while some studies have suggested a higher risk relative to TNFis, others have not found an increased risk when compared with methotrexate or placebo.8,9 As information on unmet need, there is little available real-world analysis addressing this issue in Taiwan, especially including tsDMARDs.
Given the ongoing debate and conflicting data, we aimed to investigate and compare the risk of malignancies associated with b/tsDMARDs initiation and to identify corresponding risk factors in Taiwanese RA patients.
Patients and methods
Study design and data sources
This nationwide, population-based, retrospective cohort study retrieved data from Taiwan National Health Insurance Research Database (NHIRD), which was subsequently linked with Taiwan Cancer Registry and Catastrophic Illness Registry, 10 with study period spanning from 1 January 2000 to 31 December 2020. Taiwan National Health Insurance (NHI) programme, implemented on 1 March 1995, is an obligatory, single-payer healthcare system covering over 99.9% of the Taiwanese population. 11 NHIRD, managed by Health and Welfare Data Science Center of Taiwan Ministry of Health and Welfare, is derived from NHI database for research purposes and includes data on registration, demographics, prescriptions categorised by Anatomical Therapeutic Chemical (ATC) codes, diagnoses and procedures classified by International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification (ICD-9/10-CM) codes, dates and types of medical services and medical expenses reimbursed by NHI. However, it lacks certain personal data including body mass index, alcohol or tobacco exposure, laboratory results, serostatus, disease activity or severity measurements, damage or function assessments and prescriptions not reimbursed by NHI. Nonetheless, NHIRD has facilitated numerous population-based longitudinal studies to date. 12
Study population
In real-world practice, Taiwanese rheumatologists diagnose patients with RA following worldwide classification criteria.13,14 This study identified newly diagnosed, biologic-naïve RA patients aged ⩾20 years by including those holding a catastrophic illness certificate for RA, as confirmed by associated ICD codes from 1 January 2001 to 31 December 2020 from Catastrophic Illness Registry (Supplemental Table 1). 15 The issuance of a catastrophic illness certificate for RA requires a diagnosis confirmed by at least two other qualified rheumatologists after a critical review following Taiwan NHI regulations. To ensure all selected patients were initially prescribed NHI-reimbursed b/tsDMARDs approved for RA treatment, we excluded patients with recorded diagnoses of other rheumatic diseases before RA, those without b/tsDMARD prescriptions, those whose first b/tsDMARD prescriptions either did not meet NHI copayment criteria for RA or occurred before initial RA diagnosis, and those with incomplete data regarding urbanisation or insured amount. Patients with documented malignancies before their RA diagnosis were also excluded. Notably, b/tsDMARD prescriptions under NHI reimbursement are only available after approval by qualified rheumatologists following a critical review to ensure appropriate indications.
In Taiwan, b/tsDMARDs can be reimbursed by NHI if RA patients show high disease activity (Disease Activity Score 28 >5.1) on two occasions at least 4 weeks apart, despite having received either: (i) two or more conventional synthetic DMARDs (csDMARDs) for at least 6 months and each with full target doses for at least 2 months at the same institution, unless intolerance occurs; or (ii) concomitant use of two or more csDMARDs for at least 3 months and each with full target doses for at least 2 months, along with prednisolone at a dose of at least 15 mg/day for at least 3 months, unless intolerance occurs. Rituximab is reimbursed only for patients who inadequately respond to NHI-reimbursed TNFis.
The selected patients were subsequently classified into four distinct groups based on pharmacological mechanisms of their initial b/tsDMARDs. Rituximab was not included as a separate group in the primary analysis, as it is designated as a second-line bDMARD under current Taiwan NHI copayment guidelines. Details on b/tsDMARDs approved for persistently active RA in Taiwan from 2001 to 2020 are provided in Supplemental Table 2. Each selected patient was given an index date, defined as the date of initiation of first NHI-reimbursed b/tsDMARD, that helps censor the occurrence of covariates within 1 year before b/tsDMARD initiation. Details on establishing the study population for primary analysis are depicted in Figure 1.
Flow diagram of establishment of study population and categorisation into b/tsDMARD groups.
Study outcome
The primary outcome was the newly diagnosed malignancies within 5 years after first NHI-reimbursed b/tsDMARD initiation, as identified by approval dates of catastrophic illness certificates with corresponding ICD codes for malignancies recorded in Taiwan Cancer Registry and Catastrophic Illness Registry (Supplemental Table 1). Issuance of a catastrophic illness certificate for malignancies requires endorsement by ⩾2 qualified haematologists/oncologists after a critical review. This process ensures that conditions such as dysplasia or carcinoma in situ are excluded. Follow-up period for each patient was defined as interval from index date to the earliest occurrence of one of the following censoring events: 31 December 2020; occurrence of any outcome event; date of discontinuation as 90 days after the latest prescription of first b/tsDMARD; direct switch of first b/tsDMARD; withdrawal from NHI programme for any reason; death or a maximum duration of 5 years after first b/tsDMARD initiation.
Covariates
We selected the following covariates: gender, age at index date, disease duration by NHI-reimbursed b/tsDMARDs initiation, socioeconomic status (urbanisation levels and categorised monthly insured income), comorbidities, frequencies of ambulatory visits during follow-up periods (per half-year) and medications during follow-up periods. Urbanisation was stratified into three levels based on a composite of factors: population density, physician number per 100,000 residents and proportions of elderly individuals, agricultural workers and residents with a college education or higher in the patient’s residential district. For comorbidities, we defined their presence as having at least one inpatient visit or at least three ambulatory visits with relevant ICD codes within 1 year before the index date, to mitigate potential misclassification bias. Besides common diseases, several comorbidities associated with malignancies were incorporated.
Concomitant medications were identified based on recorded prescriptions with associated ATC codes during follow-up periods. These medications included immunosuppressants, non-steroidal anti-inflammatory drugs (NSAIDs) and systemic corticosteroids (quantified in prednisolone equivalent daily dosages). Details regarding these covariates and their corresponding administrative codes were summarised in Supplemental Table 1.
Statistical analysis
Continuous variables were presented as mean ± standard deviation, while categorical variables were shown as counts (percentages). Comparisons amongst the b/tsDMARD groups were performed using independent t tests for continuous variables and Chi-square or Fisher’s exact tests for categorical variables. Incidence rate ratios (IRRs) for study outcomes, with corresponding log-rank test p-values, were calculated, and Kaplan–Meier survival curves illustrating cumulative incidence of events free of outcomes were generated.
After confirming the proportional hazards assumption with Schoenfeld residuals, risk of outcomes following b/tsDMARD initiation was assessed using Cox proportional hazards models, with adjusted hazard ratios (aHRs) and 95% confidence intervals (CIs) reported. A two-tailed p-value of <0.05 was considered statistically significant. All statistical analyses were performed using SAS software (version 9.3; SAS Institute Inc., Cary, NC, USA).
Results
Baseline characteristics of the study population
We enrolled 8732 newly diagnosed adult RA patients without any record of malignancy who initiated NHI-reimbursed b/tsDMARDs between 2001 and 2020. Amongst these patients, 6384 received TNFis, 667 received tocilizumab, 656 received abatacept and 1025 received tsDMARDs. Baseline characteristics are detailed in Table 1. About 75.3% of the patients were female, and the mean age at index date was 55.1 years, with TNFi initiators being the youngest and abatacept initiators the oldest. Hypertension was the most common comorbidity, present in approximately one-quarter of patients. For the abatacept group, a higher prevalence of patients was recorded with hypertension (31.1%), renal diseases (4.4%), mental illnesses (14.8%) and interstitial lung diseases (1.1%) before initiation. Hyperlipidaemia was found to be the highest in the tsDMARD group. Methotrexate, NSAIDs and corticosteroids were prescribed in over 80.0% of patients. Most of the medications were most prevalently prescribed in the TNFi group, while the tsDMARD group was prescribed the lowest average steroid dose at 2.5 mg prednisolone equivalent per day.
Baseline characteristics of the study population.
Age at initiation of first b/tsDMARDs.
Insured monthly income lower than the median income (⩽22,800 New Taiwan dollars/month).
The composite result of chronic obstructive pulmonary disease, pneumoconiosis or asbestosis.
The composite result of liver cirrhosis, non-alcoholic fatty liver disease, non-alcoholic steatohepatitis, alcoholic hepatitis or chronic cholecystitis.
A p-value <0.05 is considered bold statistically significant.
bDMARDs, biological disease-modifying antirheumatic drugs; H. pylori, Helicobacter pylori; NA, not applicable; SD, standard deviation; TNFis, tumour necrosis factor inhibitors; tsDMARDs, targeted synthetic disease-modifying antirheumatic drugs.
Crude incidence rates of malignancies
Over a 5-year follow-up period, the overall event rate in the entire study population was 2.16%. When stratified by treatments, event rates were 2.16% for TNFi, 1.95% for tocilizumab, 2.74% for abatacept and 1.95% for tsDMARDs. The corresponding crude incidence rates were 769 in total, and 760, 774, 1001 and 948 per 100,000 person-years for TNFi, tocilizumab, abatacept and tsDMARDs, respectively (Table 2). Kaplan–Meier curve was depicted in Figure 2. Though abatacept (IRR = 1.32) and tsDMARDs (IRR = 1.25) had slightly higher IRRs compared with TNFi, no significant difference amongst the b/tsDMARDs was found, and none of the b/tsDMARDs differed significantly from TNFis by IRR.
Crude incidence rates of malignancies within two follow-up periods after b/tsDMARDs initiation.
A p-value <0.05 is considered statistically significant.
bDMARDs, biological disease-modifying antirheumatic drugs; CI, confidence interval; IRR, incidence rate ratio; TNFis, tumour necrosis factor inhibitors; tsDMARDs, targeted synthetic disease-modifying antirheumatic drugs.
Kaplan–Meier curves for cumulative incidence of free of malignancies.
In contrast, over an 18-year follow-up period, the overall event rate and crude incidence rate increased to 2.89% and 834 per 100,000 person-years. By treatment groups, event rates were 3.09% for TNFi, 1.95% for tocilizumab, 3.35% for abatacept and 1.95% for tsDMARDs, with crude incidence rates of 811, 728, 1091 and 942 per 100,000 person-years, respectively (Table 2). Kaplan–Meier curve was depicted in Supplemental Figure 1. Although abatacept (IRR = 1.35) and tsDMARD (IRR = 1.16) showed numerically higher IRRs compared with TNFi, no significant difference amongst the b/tsDMARDs was found, and neither of b/tsDMARDs differed significantly from TNFis.
Independent factors associated with malignancies
Table 3 presents the results of time-dependent Cox regression models. In univariate analysis, male gender (aHR = 2.09, p < 0.001), older age at b/tsDMARD initiation (aHR = 1.05 per year, p < 0.001), coronary heart disease (aHR = 1.97, p = 0.012), chronic lung diseases (composite result of chronic obstructive pulmonary disease (COPD), pneumoconiosis or asbestosis; aHR = 2.31, p < 0.001) and tuberculous infection (aHR = 4.93, p = 0.006) were positively associated with risk of malignancies. Higher frequency of ambulatory visits during follow-up periods was also associated with a reduced hazard (aHR = 0.93, p < 0.001). Amongst medication-related factors, higher steroid doses (>5 mg/day) showed a higher risk (aHR = 1.79, p = 0.002), whereas leflunomide (aHR = 0.58, p = 0.023), methotrexate (aHR = 0.54, p < 0.001), sulfasalazine (aHR = 0.70, p = 0.028) and NSAIDs (aHR = 0.56, p = 0.001) each demonstrated a lower risk.
Risk factors associated with malignancies within 5 years after b/tsDMARDs initiation.
Age at initiation of first b/tsDMARDs.
Insured monthly income lower than the median income (⩽22,800 New Taiwan dollars/month).
The composite result of chronic obstructive pulmonary disease, pneumoconiosis or asbestosis.
The composite result of liver cirrhosis, non-alcoholic fatty liver disease, non-alcoholic steatohepatitis, alcoholic hepatitis or chronic cholecystitis.
Time-dependent multivariable model 1 adjusted for gender and age at index date.
Time-dependent multivariable model 2 adjusted for all adjustable covariates in this study.
A p-value <0.05 is considered bold statistically significant.
aHR, adjusted hazard ratio; AIC, Akaike information criterion; bDMARDs, biological disease-modifying antirheumatic drugs; CI, confidence interval; H. pylori, Helicobacter pylori; HR, hazard ratio; NHI, national health insurance; TNFis, tumour necrosis factor inhibitors; tsDMARDs, targeted synthetic disease-modifying antirheumatic drugs.
In multivariable regression model 1, adjusting for gender and age at index date, no b/tsDMARD group differed significantly from TNFis, while male gender (aHR = 1.96, p < 0.001) and older age at b/tsDMARD initiation (aHR = 1.05, p < 0.001) were associated with a higher risk. In multivariable regression model 2, adjusting for all included covariates, neither group differed significantly from TNFis. Male gender (aHR = 1.80, p < 0.001) and older age at b/tsDMARD initiation (aHR = 1.05, p < 0.001) remained independent risk factors. Notably, chronic lung diseases continued to be associated with a higher risk (aHR = 2.03, p = 0.003), while more frequent ambulatory visits remained protective (aHR = 0.90, p < 0.001). Leflunomide (aHR = 0.44, p = 0.001), methotrexate (aHR = 0.55, p < 0.001) and sulfasalazine (aHR = 0.70, p = 0.030) all showed persistent protective effects. Furthermore, steroid doses exceeding 5 mg/day remained a significant risk factor (aHR = 2.62, p < 0.001).
Discussion
In this nationwide, population-based study of 8732 RA patients initiating b/tsDMARDs, we observed an overall malignancy event rate of 2.16% over a 5-year follow-up period, which increased to 2.89% when extended to 18 years. Crude incidence rates of malignancies were comparable across b/tsDMARD groups over both 5 and 18 years. Within 5 years after first NHI-reimbursed b/tsDMARD initiation, neither non-TNF b/tsDMARDs differed significantly from TNFis in malignancy incidence in multivariable regression models. Independent factors associated with malignancies included male gender, older age at initiation, chronic lung diseases and higher-dose corticosteroid prescription, while leflunomide, methotrexate and sulfasalazine prescription and higher ambulatory visit frequencies were associated with a protective effect.
Our results suggest that the risk of malignancies in RA patients initiating b/tsDMARDs is not significantly affected by specific classes of b/tsDMARDs, and that patient demographics, comorbidities and certain concomitant medications play a more pivotal role in determining malignancy risk. Notably, cumulative real-world reports on overall malignancy risk with abatacept compared with other cs/b/tsDMARDs have demonstrated conflicting results, probably owing to differences in design, follow-up period, ethnicity, selected RA population and focused types of malignancy.5,16–18 Our study demonstrates a numerically higher risk associated with abatacept initiation, which may reflect an older mean age of initiation than other b/tsDMARD groups, and the risk didn’t significantly rise after adjustment for all included covariates. No direct association between abatacept initiation and malignancy onset aligns with the results of some studies.16,19–21 As ORAL Surveillance study reported a higher risk of cancers in cardiovascular risk-enriched patients using tofacitinib than TNFis, safety concerns have been raised regarding tsDMARDs prescription. 8 Emerging real-world analyses have demonstrated no association between tsDMARDs and overall malignancy risk amongst older RA patients. 7 Our study demonstrates a numerically slightly higher malignancy risk but a non-significant difference in hazard between tsDMARDs and TNFis. However, we present analysis results from the pooled adult patients instead of patients with a higher malignancy risk. Though evidence between malignancy risk and use of b/tsDMARDs remains controversial, our findings suggest that their use should not be categorically avoided and patients receiving them may benefit from closer monitoring.
Male gender and older age at initiation are previously reported as risk factors for malignancies in RA patients, which are compatible with our results.22,23 Prior studies have reported that male RA patients exhibit higher IRRs for various malignancies, including mesothelioma, Kaposi sarcoma, cancers of lip, larynx, hypopharynx, urinary bladder, oesophagus, tonsil, oropharynx and other urinary organs, compared with their female counterparts. 24 Moreover, older age is associated with an increased malignancy risk in general population, which is also reflected in RA patients. These findings underscore the need for heightened vigilance and careful monitoring for malignancies during treatment, especially in male and older RA patients. Further studies focusing on specific malignancy subtypes are warranted to better elucidate the differential risks and inform tailored monitoring strategies in Taiwanese RA patients. Previous studies have demonstrated that COPD is an independent risk factor for lung cancer, with smoking playing a central role in the development of both diseases. 25 RA has been associated with an increased risk of COPD, 26 with smoking serving as a common risk factor for both. Moreover, recent studies suggest that periodontal pathogens may contribute to the pathogenesis of RA, highlighting a potential link between oral health and systemic autoimmunity. 27 In summary, smoking cessation and maintaining good oral hygiene should be included in the management of RA and are crucial for rheumatologists to actively promote to reduce the risk of complications and improve overall patient outcomes.
Higher doses of corticosteroid exposure are associated with the development of malignancies, whereas certain csDMARDs appear protective. This suggests that the risk of malignancies stemming from suboptimal control of disease activity outweighs that from csDMARDs under indicated doses, and that better control of disease activity under a steroid-sparing strategy may reduce malignancy risk in RA patients. Several studies support the association between RA disease activity, use of csDMARDs and malignancy risk.28–30 Therefore, optimising RA management to achieve better disease control and reduce corticosteroid dosage should be a critical strategy in lowering the risk of malignancies.
There are several strengths in this study. First, the study utilises data from Taiwan NHIRD, which covers over 99.9% of the Taiwanese population. This comprehensive coverage enhances the representativeness of our findings and minimises selection bias, ensuring that our results are reflective of real-world clinical practice. The recording of prescriptions and procedures also minimises recall bias. Second, the requirement for catastrophic illness certificates for both RA and malignancy diagnoses, each confirmed by at least two qualified specialists, ensures high diagnostic accuracy. This rigorous verification process strengthens the reliability of our outcome assessments and reduces the likelihood of misclassification. Third, indications of b/tsDMARDs for persistently active RA were also qualified by specialists, which confirmed their adequate prescription. Fourth, by examining outcomes over both 5-year and extended 18-year periods, the study provides insights into both short-term and long-term malignancy risks. This dual time-frame analysis not only captures immediate treatment effects but also reflects the impact of long-term disease management, aligning closely with actual clinical practice. Fifth, incorporating time-dependent Cox models allows us to account for changes in ambulatory visit frequencies and medication use over time. This dynamic approach offers a more realistic evaluation of risk factors compared with static baseline analyses, as it adjusts for temporal variations in patient management and disease evolution. Adjustment for ambulatory visit frequencies also minimises detection bias. Overall, these strengths highlight the robustness of our study design and enhance the clinical relevance of our findings.
There are several limitations in this study. First, limitations regarding the administrative database exist. Key variables such as body mass index, smoking and alcohol history, family history, stressful events, laboratory results and measures of RA disease activity or severity were not recorded in the database and therefore could not be included in our analysis. We attempt to mitigate this limitation by incorporating recordable complications, such as chronic lung diseases including COPD, interstitial lung disease, mental illnesses and chronic hepatobiliary diseases including alcoholic hepatitis as surrogates, but the absence of these clinical parameters may still cause residual confounding. Prescription dosage, prescription compliance and actual date of the first b/tsDMARD discontinuation or switch are not definitely determined. However, both 90 days after the latest prescription and direct switch of first b/tsDMARD are set for follow-up termination to minimise confounding by first b/tsDMARD discontinuation or switch. Higher disease activity can be partially reflected by higher dosages of corticosteroid prescription and ambulatory visit frequencies in real-world practice; it can also be partially adjusted through time-dependent Cox models. Prescriptions or interventions not covered by NHI or the use of substances and alternative medications were not fully captured in our study. This may result in some misclassification of medication exposure; however, as these instances likely represent a small portion of overall treatment, their impact on our results over both 5-year and 18-year follow-up periods is expected to be minimal. Second, we claim associations and associated factors merely because a retrospective database study provides a limited causal relationship. Third, some malignancy subtypes, specific comorbidities or less commonly used therapies (e.g. certain tsDMARDs) are represented by relatively small patient numbers, limiting the statistical power to perform subgroup analysis for outcomes or detect significant differences or interactions for these variables. Fourth, the time scale within 1 year before the index date is set to examine continuous exposure to covariates, while the influence of covariates occurring earlier than the time scale can not be determined. Fifth, our results can not be extended to RA patients without the use of b/tsDMARDs, and the global generalisability of results of our regional research is also limited.
Our study can provide reference data for worldwide epidemiological research, future research on molecular biopharmacological mechanisms and Taiwanese health administrations for pharmacovigilance of b/tsDMARDs. Future studies endorsing more robust causal relationships, accounting for smoking and drinking status and RA disease activity and allowing censoring events within longer follow-up periods are warranted.
Conclusion
In this nationwide cohort of adult Taiwanese RA patients initiating b/tsDMARDs, no specific b/tsDMARD was associated with a higher overall malignancy risk over a 5-year follow-up period. Associated factors included male gender, older age at b/tsDMARD initiation, chronic lung diseases and exposure to higher corticosteroid doses, while certain csDMARDs and more frequent ambulatory follow-up appeared protective. These findings highlight the importance of individualised risk assessment and underscore the need for optimised disease activity control under a steroid-sparing strategy and preventive measures and instructions in RA management.
Supplemental Material
sj-docx-1-tab-10.1177_1759720X261423830 – Supplemental material for Malignancy risk in patients with rheumatoid arthritis initiating biological or targeted synthetic disease-modifying antirheumatic drugs: a 5-year population-based cohort study
Supplemental material, sj-docx-1-tab-10.1177_1759720X261423830 for Malignancy risk in patients with rheumatoid arthritis initiating biological or targeted synthetic disease-modifying antirheumatic drugs: a 5-year population-based cohort study by Chung-Mao Kao, Po-Cheng Shih, Der-Yuan Chen and Hsin-Hua Chen in Therapeutic Advances in Musculoskeletal Disease
Footnotes
Acknowledgements
We kindly thank Ms Chia-Hui Yu and the Biostatistics Task Force of Taichung Veterans General Hospital for their assistance in performing and instructing the statistical analysis. This study utilised data from Taiwan NHIRD, provided by the Bureau of National Health Insurance and the Health and Welfare Data Science Center of Ministry of Health and Welfare. Interpretations and conclusions in this study do not represent those of these administrative units.
Declarations
Availability of data and materials
The datasets used and/or analysed in the current study are available from the corresponding author upon reasonable request.
Supplemental material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
