Influence of seasonal changes on disease activity and distribution of affected joints in rheumatoid arthritis

Background Previous studies suggest that RA activity is sensitive to seasonal changes. This study explored the influence of season on RA activity, particularly the distribution of affected joints, using a nationwide database in Japan. Methods We investigated 12,839 patients whose RA activity was recorded in spring (n = 3250), summer (n = 916), fall (n = 1021), and winter (n = 7652). Disease activity score (DAS) 28-CRP, simplified disease activity index (SDAI), and clinical disease activity index (CDAI) were used as indices of disease activity. Disease activity was also assessed according to DAS28-CRP scores (remission, low, moderate, or high). The affected joint distribution was investigated using novel joint indices (x, y, z), where x and y are indices for the upper and lower joints, respectively, and z is the index for large joint predominance. Results Mean DAS28-CRP and median SDAI and CDAI scores were highest in spring and lowest in fall. There was a significant difference in the DAS28-CRP for fall versus spring and winter. Fall was associated with a higher remission rate, and spring and winter with high and moderate RA activity, respectively. Significant differences in x, y, SDAI, and CDAI scores were found for spring versus summer, fall, and winter, in addition to fall versus winter (except in y). There was no seasonal difference in the z index. Conclusions RA activity in the upper and lower extremities may be highest in spring, followed by winter. Seasonal changes should be considered in patients with RA to better understand their symptoms.

In many parts of the world, the year is divided into four 3-month seasons (winter, spring, summer, and fall) that are characterized by specific weather conditions (temperature, atmospheric pressure, humidity, and precipitation) and hours of daylight [26]. Previous studies have suggested that seasonal changes could influence the clinical manifestations of RA. For example, it has been reported that the onset of RA is more likely in winter than in summer in the northern hemisphere and that working in a cold environment increases the risk of developing RA [27]. Moreover, the onset of symptoms of arthritis during winter or spring was reported to be associated with more radiographic joint damage after 6 months in patients with early RA [28]; however, that report was followed by a conflicting one [29]. It is also widely believed that symptoms in patients with musculoskeletal pain are worse in winter [30]. Another study demonstrated that rheumatic symptoms were exacerbated by seasonal changes in approximately half of patients with rheumatic diseases, including RA [31]. In a study based on data from a large RA database in Japan, Iikuni et al. demonstrated that RA activity was higher in spring and lower in fall [32].
The hallmark of RA is symmetric polyarthritis affecting the small joints of the hands and feet [33], but the large joints (shoulders, elbows, knees, and ankles) and wrists may also be involved [34]. However, there have been no reports in the literature on the effects of changes in season on the distribution of affected joints in RA. Nishiyama's joint indices (JIs; x, y, z) are novel measures of RA that have 3 components: x and y are indices of RA disease activity in the upper and lower extremities, respectively, and z is an index of the predominance of large joint involvement [35,36]. In this study, we assessed RA activity and the distribution of affected joints using the x, y, and z indices and data from a nationwide RA database in Japan (National Database of Rheumatic Diseases by iR-net in Japan, NinJa) [37]. Based on the data obtained, we discuss the influence of seasonal changes on the activity of RA.

Data source
NinJa is a nationwide, multicenter, observational database that was established in 2002 and contains data on RA patients treated in Japan [37]. RA patients at each participating medical institution are enrolled in NinJa at random. The clinical data of RA patients registered in NinJa are collected once a year from each patient at any time point during the indicated year. In this study, we used data from NinJa2016, which contains data for 15,341 patients with RA collected between April 1, 2016, and March 31, 2017. To analyze the effects of season on RA activity, we used the data from 12,839 patients for whom data on tender joint count (TJC), swollen joint count (SJC), 10-cm visual analog scale (VAS) pain score, patient's global assessment of disease activity (PGA), physician's global assessment of disease activity (PhGA), C-reactive protein (CRP) level, Steinbrocker classification of radiographic stage, and functional class were available. The cutoff age for defining the patients as older or younger was 65 years. The seasons were divided into spring (March, April, May), summer (June, July, August), fall (September, October, November), and winter (December, January, February), while acknowledging that weather conditions in Japan can be different even during the same season depending on the location of the participating institutions.
The NinJa study protocol was reviewed and approved by the ethics committee at each participating institution.

Novel joint index
We used the methodology previously described by Nishiyama et al. in order to measure RA activity and affected joint distribution using the joint indices consisting of 3 components (x, y, z) [35]. The indices x and y are the joint indices for upper and lower extremities, respectively. The index z reflects the predominance of large joints over small joints.

Statistical analysis
The RA patients whose data were recorded in spring (n = 3250), summer (n = 916), fall (n = 1021), and winter (n = 7652) were all independent patients, so paired analyses were not required for the present study. For continuous variables with a normal distribution, one-way analysis of variance (ANOVA) was used to examine the differences in average values among the four seasons, followed by the post hoc Tukey-Kramer method for multiple comparisons. The Kruskal-Wallis test was used for continuous variables with a skewed distribution. Pearson's chi-square test was used for categorical variables. A post hoc test based on adjusted standardized residuals (ASR) was used for multiple comparisons, whereby absolute values of ASR that were more than 1.96 and 2.58, respectively, were considered to correspond to significance levels of 0.05 and 0.01. The statistical analysis was performed using IBM SPSS Statistics version 24 software (IBM Corp., Armonk, NY) and JMP version 12.0.1 (SAS Institute Inc., Cary, NC). All significance levels were set at p < 0.05 (two-sided).

Demographic and clinical data
In NinJa, clinical data are collected once per year from each participating patient at any time point during the indicated year. We used the clinical data for 2016 (NinJa2016), which were collected between April 1, 2016 and March 31, 2017. Data on TJC, SJC, VAS pain score, PtGA, PhGA, CRP, disease stage, and functional class were available for 12,839 patients with RA registered in NinJa2016. Of these 12,839 patients, 3250, 916, 1021, and 7652 were evaluated in spring, summer, fall, and winter, respectively. One-way ANOVA and Pearson's chi-square tests did not reveal a statistically significant seasonal difference in age or sex distribution or in the proportion of patients with stage III-IV disease (Table 1). However, there were significant differences in disease duration and the proportion of patients with class 3-4 functional status. Post hoc analysis revealed that the disease duration was significantly longer in RA patients evaluated in fall than in those evaluated in winter (p < 0.05) and that the proportion of patients with class 3-4 functional status was significantly lower in patients evaluated in fall (p < 0.01) and higher in those evaluated in winter (p < 0.05).

Seasonal differences in ACR core set of disease activity variables
One-way ANOVA of the 7 parameters included in the ACR core set of measures used in clinical trials of RA (TJC, SJC, pain VAS score, PGA, PhGA, mHAQ, and CRP) revealed significant differences in the distributions of TJC, SJC, and PhGA (Table 1). A post hoc analysis using the Tukey-Kramer method revealed significant differences in the distribution of TJC in summer versus spring (p = 0.001) and winter (p < 0.001) as well as in fall versus winter (p = 0.013); the difference between fall and spring did not reach statistical significance (p = 0.054). There were statistically significant differences in the SJC for fall versus spring (p < 0.001), summer (p = 0.019), and winter (p = 0.015), in addition to spring versus summer (p < 0.001) and winter (p < 0.001). There were also statistically significant differences in PhGA for fall versus spring (p < 0.001) and winter (p < 0.001) in addition to spring versus summer (p = 0.003).

Seasonal differences in composite RA activity indices
DAS28-CRP, SDAI, and CDAI scores were highest in spring and lowest in fall. A post hoc analysis using the Tukey-Kramer method revealed statistically significant differences in the distribution of DAS28-CRP scores in fall versus spring (p < 0.001) and winter (p = 0.004; Table 2). There were also significant differences in the SDAI scores for spring versus summer (p = 0.021), fall (p < 0.001), and winter (p = 0.011) in addition to fall versus winter (p = 0.001). Similarly, there were significant differences in the distribution of CDAI scores for spring versus summer (p = 0.009), fall (p < 0.001), and winter (p = 0.007) as well as fall versus winter (p < 0.001).
We also analyzed the influence of seasonal changes on RA activity by categorizing the disease activity as remission, low, moderate, or high based on DAS28-CRP scores. There was a significantly higher proportion of patients in remission (p < 0.01) and a significantly lower proportion of patients with moderate disease activity (p < 0.01) in fall ( Fig. 1 a). In contrast, spring was associated with high disease activity (p < 0.01) and winter with moderate disease activity (p < 0.01). We divided the RA patients into an older group (≥65 years, n = 7584) and a younger group (< 65 years; n = 5255) to examine the influence of age on seasonal variations in RA disease activity. We found a significantly higher proportion of patients in remission in fall (p < 0.01) and summer (p < 0.01) and a significantly lower proportion of patients with moderate disease activity in fall (p < 0.01; Fig. 1b).

Seasonal differences in joint distribution
The x and y indices reflect the RA activity in the upper and lower extremities, respectively. The z value is an The average x, y and z values are plotted in Fig. 2. There were significant differences in these values for spring versus summer (p = 0.001), fall (p < 0.001), and winter (p = 0.003) as well as for fall versus winter (p = 0.006). There were significant differences in y for spring versus summer (p = 0.004) and fall (p = 0.002) as well as for fall versus winter (p = 0.037); the difference between spring and winter did not reach statistical significance (p = 0.25). There was no significant seasonal difference in the z value (0.06 ± 0.30, 0.07 ± 0.28, 0.07 ± 0.25, and 0.07 ± 0.28 in the order of spring to winter).

Discussion
In this study, using a nationwide RA database (NinJa2016) in Japan, we investigated the influence of season on clinical parameters of RA, including TJC, SJC, VAS pain score, PGA, PhGA, CRP, stage, and functional class data, for 12,839 patients with RA. Our results show a significant difference in RA activity among the seasons. Furthermore, we have demonstrated for the first time that arthritic symptoms are affected by seasonal change in both the upper and lower extremities, as assessed by a newly developed joint index.
Iikuni et al. previously demonstrated that RA activity, evaluated both subjectively and objectively, is higher in spring and lower in fall (n = 1665) using the large IORRA (Institute of Rheumatology, Rheumatoid Arthritis) cohort database, from which they retrieved clinical information for registered patients with RA in spring and fall for 5 consecutive years [32]. Sadamoto et al. also investigated the relationship between RA activity and seasonal change in 280 patients with RA in Japan, and found that 63% of patients recognized seasonal deterioration, which was frequent in spring and winter [40]. Consistent with their findings, we found that fall was associated with a high proportion of patients in remission as well as a low proportion of patients with moderate disease activity, whereas spring and winter were significantly associated with moderate and high disease activity, respectively. Furthermore, DAS28-CRP scores were numerically highest in spring, followed by winter, summer, and fall. Although a significant seasonal difference in DAS28-CRP scores was observed only in comparisons between fall and spring and between fall and winter, the SDAI and CDAI scores were significantly higher in spring than in summer, fall, or winter. The  SDAI and CDAI scores were also significantly higher in winter than in fall. Therefore, it was considered that spring was the season associated with the highest RA activity.
In this study, we analyzed the influence of seasonal changes on the distribution of affected joints in patients with RA using the newly developed Nishiyama's joint index [35]. To our knowledge, this is the first study to do so. We demonstrated that the x index was numerically elevated in the order of spring, winter, summer, and fall. A statistically significant difference in the x index was found between spring and the other seasons and between winter and fall. A similar statistically significant seasonal pattern was observed for the y index, except between spring and winter. Given that the x and y indices reflect RA activity in the upper and lower extremities, respectively, it was considered that the RA activity in both the upper and lower extremities was greater in spring, presumably followed by winter. However, there was no significant change in the z index, suggesting that large and small joints were similarly affected, irrespective of season.
It is unknown why RA activity was higher in spring and winter in our study, as in earlier reports. Winter is generally associated with an increase in musculoskeletal symptoms [41], partly because cold weather is likely to cause an increase in muscle spasm and a decrease in the blood circulation. In a longitudinal observational study of 133 patients with RA, Savage et al. demonstrated a significant decrease in disease activity in conditions of more sunshine and low humidity [21]. With regard to spring, it should be noted that individuals with chronic pain may be vulnerable to changes in the local climate [42]. Japan has a unique climate pattern in early spring known as san-kan-shi-on (3 days of cold weather followed by 4 days of warm weather) because of alternating high and low atmospheric air pressure systems over a period of approximately 7 days [43]. Of interest, Terao et al. have recently investigated the relationship between atmospheric air pressure and RA activity using data for 2131 patients with RA in the KURAMA (Kyoto University Rheumatoid Arthritis Management Alliance) database and demonstrated an inverse association of air pressure with activity of rheumatoid synovitis. This finding raises the possibility that rapid and repeated changes in local atmospheric pressure could lead to unstable RA and subsequent deterioration of disease activity during the san-kan-shi-on period in spring. Alternatively, it is possible that this high disease activity is triggered by stressful life events that coincide with spring. The financial year starts in April in Japan, so spring is the season in which patients with RA may feel the most financial stress. Other events likely to occur in spring include changes in lifestyle, sitting entrance examinations, which are anxiety-provoking for both parents and children, and changes in personnel in the workplace. RA activity is known to be affected by psychological distress [44] and could be elevated in spring because of these stressful events.
Previous studies have demonstrated that elderly individuals are generally more likely to experience frequent and prolonged pain than their younger counterparts [45,46]. However, in the present study, there was a significant seasonal variation in RA disease activity in younger RA patients as well, and significant associations of fall with a higher proportion of remission and a lower proportion of moderate disease activity. Therefore, we need to recognize that seasonal changes in RA activity may occur in any age group.
The biological mechanism for seasonal variation in RA activity remains to be elucidated. Vitamin D metabolism might be a contributing factor [47][48][49], given that vitamin D has effects on innate and acquired immunity [47] and exposure to sunlight is required for ultraviolet B-induced synthesis of vitamin D in the skin. Ishikawa et al. demonstrated that serum vitamin D levels follow a lagged pattern relative to the astronomical seasons, peaking in late summer (August) and troughing in late winter (February) [47], possibly as a result of fewer daylight hours and less outdoor activity because of cool or cold weather. A meta-analysis by Lee et al. demonstrated that vitamin D deficiency is prevalent in patients with RA and found a significant inverse correlation between low serum vitamin D levels and RA activity in these patients [50], suggesting that hypovitaminosis D could play an important role in the seasonality of RA activity. Furthermore, in a comprehensive whole-transcriptome analysis of peripheral blood mononuclear cells in vitro, Dopico et al. demonstrated significant and widespread seasonal changes in the immune system, including a predominance of pro-inflammatory gene expression in peripheral blood mononuclear cells in winter as compared with summer [51]. Their findings suggest that seasonality of RA activity might be explained at least in part by seasonal changes in transcriptional signatures within the immune system.
This study has several limitations that need to be addressed. First, the patients with RA in NinJa were evaluated once a year at any time point at the discretion of their physicians, which enabled us to study the seasonal influence on RA activity. However, this was not a longitudinal study that examined changes in disease activity in a cohort of RA patients over a defined period of time. Therefore, one limitation of this study is its cross-sectional unpaired design, which could potentially have introduced bias. Second, although the age at RA onset has increased significantly over the last 10 years in Japan [52], NinJa does not collect data on concomitant diseases, such as osteoarthritis. Timmermans et al. demonstrated that elderly patients with osteoarthritis often perceived the weather as affecting their pain [53]. In the present study, younger patients were also vulnerable to seasonal changes in RA activity, suggesting that symptoms of osteoarthritis do not play a major role in seasonal variation in RA activity. Nevertheless, we cannot exclude the possibility that unidentified comorbidities, such as osteoarthritis, might be confounding factors contributing to seasonal variation in RA activity. Third, the study did not include a control group for comparison, such as patients with osteoarthritis or low back pain. Therefore, it is unclear whether or not the observed seasonal influence is specific for RA or includes musculoskeletal diseases other than RA. Fourth, there is a regional difference in terms of the periods of the four seasons in Japan, which is located at a latitude of 24°-46°N and a longitude of 123°-146°E. If we had defined the period of each individual season corresponding to each region based on local weather conditions, a more significant seasonal influence on RA activity might have been found. It should also be noted that there were significant differences in disease duration and the proportion of patients with class 3-4 functional status according to season, which could have affected our observations. With respect to functional impairment, we consider it possible that the modest RA activity in fall improved the functional status of our study population, resulting in a decreased proportion in functional class 3-4. The difference in disease duration among the seasons (ranging from 13.2 years in winter to 14.2 years in fall) appears to be clinically insignificant, given that functional impairment was reported to develop most rapidly during the first year after disease onset, with a slow and almost linear rate of increase thereafter [54]. However, this study has some strengths, including the relatively large patient population studied and use of a novel joint index that enabled us to demonstrate a significant influence of season on the distribution of affected joints in patients with RA.

Conclusions
We have demonstrated an effect of seasonality on disease activity and affected joint distribution in patients with RA using a nationwide RA database in Japan. Health care providers should be aware of seasonal fluctuations in the manifestations of RA, particularly the functional deterioration that can occur in winter and spring, and help patients with the disease to adjust to seasonal changes in terms of their activities of daily living (e.g., by suggesting they wear warmer clothes in the cold weather and temporarily increasing their doses of analgesic medication if necessary). A large-scale, longitudinal study of the relationship between daily weather conditions and RA activity, including patient-reported outcomes, is needed to clarify how seasonal changes and other meteorological factors affect rheumatoid synovitis and to explore new anti-rheumatic interventions.

Funding
This study was supported by the Grants of the Practical Research Project for Allergic Diseases and Immunology from Japan Agency for Medical Research and Development (16ek0410014h0003) for the design of the study and collection, analysis, and interpretation of data. The writing of the manuscript was funded by Tokyo Medical University Hospital.

Availability of data and materials
The datasets used in this study can be made available by the corresponding author on reasonable request.
Authors' contributions TM and ST are in charge of the management of the NinJa database. HM, TS, KS, and SN were the lead investigators in the study. TS, KS, TM, and ST contributed to the conception and design of the study. The joint index used in the study was devised by SN. HM, TS, KT, HH, EK, and MT actively contributed to the statistical analysis. All authors participated in interpretation of the data, critically reviewed the manuscript, and approved the final version of the article for publication.
Ethics approval and consent to participate NinJa database is managed, operated, and organized by National Hospital Organization (NHO) Sagamihara hospital. The NinJa study protocol was reviewed and approved by the Ethics Committee of NHO Sagamihara hospital (2014031816)