Our study produced several key findings. Firstly, we observed that the values of DAS28-ESR, sDAI, pGA, ESR, and the population of patients using bDAMRDs were significantly higher in RF-positive RA patients. This suggests that the severity of RA in our registry was different depending on the presence of auto-antibodies. The presence of RF allowed for the identification of different subgroups in RA patients, including those with different genetic backgrounds, clinical courses, and treatment responses (11).
Secondly, our study indicated that disease activities of RA were significantly higher in winter and lower in summer, particularly in patients with positive RF. Statistical significance was observed in RF-positive patients when comparing summer with winter for ESR and global health assessment but not for CRP (Table 2). However, we found that these effects were limited to bDMARDs users or those with positive RF and no statistical significance was found in other groups. Since bDMARDs users had a higher RF-positive rate (82.4% vs. 71.5%, p = 0.016, Table 1), further research is needed to determine the main factors associated with weather changes. In Table 3, a positive Beta value indicates a positive correlation with weather factors. We found that the DAS28-ESR, sDAI, and pGA in RF-positive patients using bDMARDs were associated with weather factors, while those with RF-negative did not show correlations with weather factors. Among csDMARDs users, only pGA and DAS28-ESR in those with positive RF were associated with weather factors. To summarize, bDMARDs users or patients with RF were more likely to have disease activity fluctuations due to weather changes, and RF may be the major determinant. Different weather factors (pressure, temperature, rain amount, relative humidity, and sunshine hours) were found to have distinctive correlation coefficients with RA disease activities (Table 3). If it is necessary to select one weather factor and one disease activity for further study, relative humidity and pGA would be the recommended candidates. Interestingly, we found that ESR was more sensitive than CRP when weather factors were considered.
Our study is consistent with previous research indicating that patients' pain is associated with humidity (7) and lower temperatures are associated with RA flares (5). In Savage's study, lower humidity and more sunshine hours were associated with low DAS28 in Northern Ireland (12). Although there was no statistical significance in Savage's data, higher temperatures were associated with a reduction in DAS28, suggesting that lower humidity and more sunshine hours (typical climate characters during summer in Northern Ireland) were associated with a decrease in DAS28 (13). These findings imply that disease activities of RA patients in Savage's study also had seasonal fluctuations. In Iikuni's registry with 1665 patients, DAS28 and HAQ showed seasonal changes in spring and fall, with higher disease activity observed in spring than in fall (6). Although studies worldwide have indicated that RA patients seem to have seasonal changes in disease activities, it is still too early to conclusively determine the association between weather changes and RA disease. Research is still needed to examine whether disease activity changes before, during, or after the changes of specific weather parameters. The evidence we have can only support the association between weather changes and RA disease activities.
There are several hypotheses regarding the relationship between weather changes and joint pain. Patberg's hypothesis is that the microenvironment for skin contact, rather than natural weather variables recorded from weather stations, is the major weather variable affecting people who spend most of their time indoors and are largely impacted by air conditioning (14). However, our study showed that only RF-positive RA patients had a significant positive correlation with relative humidity. Iikuni's study also reported lower DAS28 in fall, which is a relatively wet season in Japan. Another contributing factor to joint pain is pressure change, which can cause tendon and associated tissue swelling or contraction, leading to sensitizing neurological pain (3). Nevertheless, in our data, barometric pressure does not seem to be significantly associated with DAS28-ESR even in RF-positive patients. However, if we focus on the pGA instead of DAS28-ESR, patients with positive RF may feel more discomfort when the atmospheric pressure rises. Rentschler's study (2) confirmed that joint comfort levels had an inverse relationship with barometric pressure, albeit in patients who did not match the current classification criteria for RA. It is assumed that a minimal change in the joints and their peripheral tissues due to a pressure change may cause pain or discomfort rather than changes in ESR or DAS28. Additionally, patients with positive RF may be more sensitive to joint pain than those without RF. Although the pathophysiology remains unclear, as mentioned above, patients with auto-antibodies (RF-positive) in RA may have different genetic backgrounds, clinical courses, and treatment responses from RF-negative patients. It is also possible that the distribution of baroreceptors in joints may vary based on auto-antibodies such as RF, but this has not been confirmed.
The third key finding of our study revealed that a real-world registry under the national health policy imposes limitations on clinical works, not only in a timely manner but also in terms of drug dispensing. The gender distribution of our registry was comparable to that of Taiwan National Health Insurance (NHI) Research Database(15), which indicates no significant demographic bias among the selected RA patients in our registry. However, the National Health Insurance Administration of Taiwan restricts clinician's decision-making ability by reducing NHI payment for medications and laboratory examinations. Therefore, we found missing data in our registry, such as 74.3% of anti-CCP antibody tests and 62.6% of IGRA tests. The missing data, in turn, reflects excessive clinical workload due to the lack of sufficient time for registry recording, regardless of whether TRACER, our web-based registry tool, was used or not, and may generate selection bias of disease activities in RA patients, especially in the peak month of February (as seen in Figs. 1 and 2). Chinese New Year, which usually falls in February in Taiwan, leads to an extended vacation for physicians and patients. If a patient has to visit the hospital during the Chinese New Year vacation, it would be inconvenient. Therefore, physicians often provide longer-term prescriptions for the Chinese New Year vacation if the patient is relatively stable. Consequently, the number of patients seeking medical assistance in February is often lower, and only those with more severe conditions (higher disease activities) are likely to seek medical aid during this period. This phenomenon may not be observed in other countries or clinical trials.
The fourth major finding of our study was that the NHI research database eliminates personal data to maintain patient privacy before releasing datasets, which limits opportunities for linking with outbound data. However, our registry allows us to merge patient data first and then delink their private data, allowing for greater flexibility in further analysis. This feature of our registry provides a unique advantage over the NHI research database and allows researchers to conduct more comprehensive studies on specific patient groups, such as those with positive RF or those with different disease activity levels. Our registry data also provides opportunities for future investigations, such as linking to environmental data or social determinants of health, to better understand the disease burden of RA in Taiwan and ultimately improve patient care.
Our study had several limitations that must be acknowledged. Firstly, there was a significant amount of missing data due to NHI policies in Taiwan, including payment cuts resulting in data loss in the NHI Research Database and inadequate time and resources within real-world clinical practice. Secondly, the weather data collected near the hospital may not accurately represent individual patient residence data, potentially introducing bias in our results. Additionally, since the weather changes in Taiwan are relatively minor compared to more extreme climate zones, it is possible that all weather factors may contribute to RA disease activity, rather than a single factor. Furthermore, short-term weather changes, such as typhoons or the Plum Rains in Taiwan, were not accounted for in our study, and these changes may also impact disease activity.
Despite these limitations, our study's major contribution was demonstrating that TRACER is a brand new approach/registry in rheumatology that differs significantly from national registries and modularized studies. In future studies, we intend to merge more data and provide a more comprehensive understanding of RA disease in Taiwan's real world.