Rheumatoid arthritis (RA) is a chronic autoimmune disease characterized by systemic inflammation, primarily affecting the joints. Unlike osteoarthritis, RA can also affect multiple organs, including the cardiovascular system and lungs, where it may manifest as interstitial lung disease, ranging from mild to severe. In addition to its inflammatory effects, RA is associated with high morbidity and can lead to long-term disability and increased mortality.
The use of conventional, biologic, and targeted synthetic disease-modifying antirheumatic drugs (DMARDs) aims to control inflammation and prevent further joint damage. However, DMARDs can have risks such as liver and kidney complications. After more than 50 years of research, RA remains a leading cause of disability worldwide. Although prevalence rates may vary, it is estimated to affect from 0.2% to 1% of the population. RA disproportionately affects women, older adults, and individuals in higher-income populations. Although previous epidemiological studies have evaluated national-level trends, the subnational aspects of the disease, its temporal longitudinal trajectory, and its responsiveness to health policies remain incompletely understood. Standard forecasting methodologies, such as ARIMA, often fail to capture the complex real-world connections between demographic variables and health interventions.
This study addresses these gaps by utilizing a deep-learning-based framework to elucidate the global-to-local RA landscape and project its future trajectory. Researchers identified, compiled, and analyzed the largest spatiotemporal dataset on RA ever assembled, which encompasses 953 locations worldwide from 1980 to 2021. They employed an iTransformer (Inverted Transformer), a recently developed deep learning architecture designed for multivariate time-series forecasting. In addition to enabling high-accuracy predictions through 2040, the iTransformer allowed researchers to simulate various policy and decision scenarios.
To evaluate disparities and epidemiological trends, the team assessed key indicators including prevalence, incidence, mortality, disability-adjusted life years (DALYs), years of life lost (YLLs), and years lived with disability (YLDs). They also examined the impact of socioeconomic development, population ageing, and health infrastructure, which shape the gap between actual and optimal disease burden. Additionally, the burden of disease was analysed in the context of smoking control policies.
In 2021, an estimated 17.9 million individuals worldwide were living with RA. Between 1990 and 2021, the global incidence rate increased by 13.2%, with cases occurring at younger ages and across a wider population. Although the age-standardised death rate has reduced by 32.7% since 1980, RA-related DALYs nearly doubled between 1990 and 2021.
Subnational analysis discovered significant disparities. Among 652 regions, the highest age-standardized incidence rate was recorded in West Berkshire, UK, at 35.1 (95% uncertainty interval [UI]: 30.8-39.8). The region with the highest age-standardized discovery DALY rate was Zacatecas, Mexico, at 112.6 (95% UI: 87.2-142.7). Regions with high sociodemographic index (SDI) values experienced a greater burden of RA, and disparities have generally increased over the study period. Notably, over 90% of areas did not reach optimal RA management levels across multiple indicators.
Japan emerged as a notable exception. Despite its high SDI, the country made significant progress in reducing RA burden. In Tokyo, for example, the age-standardised DALY rate declined by 22.4% between 1990 and 2021, while other affluent regions showed minimal improvement or even worsening trends. Scenario analyses suggested that implementing smoking restrictions in high-risk countries like China could result in a 16.8% reduction in RA-related deaths and a 20.6% decline in DALYs among male patients.
This study highlights the evolving burden and widening inequality of RA over four decades at the global, national, and sub-national levels. The use of deep learning models allowed for more accurate forecasting and scenario simulation, revealing that while high-SDI regions bear a disproportionate RA burden, low-SDI regions are quickly catching up due to demographic shifts and limited infrastructure. Targeted localized interventions, such as local tobacco control, could significantly reduce RA-related health impacts in high-risk areas.
References: Jin W, Wang Q, Jin C, et al. Spatiotemporal distributions and regional disparities of rheumatoid arthritis in 953 global to local locations, 1980-2040, with deep learning-empowered forecasts and evaluation of interventional policies’ benefits. Ann Rheum Dis. 2025. doi:10.1136/ard-2024-224567
