RESULTS
A total of 2,544 individuals were examined for eligibility and 2,494 fulfilled inclusion/exclusion criteria and were finally included in the analysis, 902 (36.2%) men and 1,592 (63.8%) women.Table 1 shows the description of the studied population. The mean age (SD) was 58.7 (15.7) and the most prevalent underlying pathologies were spondyloarthritis (32.6%), rheumatoid arthritis (21.6%) and osteoarthritis (25.1%). Almost half of individuals had at least one of the following comorbidities: hypertension (34%), diabetes (11.5%), pulmonary disease (14%), cardiovascular disease (11%), chronic kidney disease (5%), active cancer or treatment (3%) and post-transplant (0.3%). In terms of treatments, 45% of individuals were taking biologic DMARDs (59% in men and 36% in women), primarily anti-TNFα (30% in total; 42% in men and 24% in women) and anti-IL6/12/17/23 compounds (11%, 2% of which corresponded to anti-IL6, 3% to anti-IL17 and 6% to anti-IL12/23). A third of the population were exposed to synthetic DMARDs, being methotrexate, leflunomide and chloroquine/hydroxychloroquine the most prevalent ones (22%, 5% and 5%, respectively). Glucocorticoid consumption in women was twice that in men (26% vs 13%) but, in both cases, doses of glucocorticoids higher than 10 mg/day were unusual (<4%). NSAIDs and anti-hypertensive drugs were taken by the 20% and 27% of individuals, respectively. A 15.8% of the population (18.4% in women and 11.2% in men) did not take any of the registered treatments (Supplementary Table 3 ).
The incidence of confirmed or hsCOVID-19 symptoms during March 2020 was 6.3% (N=156), being slightly higher in women than in men (6.8% vs. 5.3%, respectively). Among them, a 22% were confirmed cases of COVID-19 and the remaining 78% had not been tested for SARS-CoV-2. There were 26 individuals (8 men and 18 women) hospitalized and there were 4 deaths due to COVID-19.
As shown in Table 2 , those presenting hsCOVID-19 symptoms had less spondylarthritis, rheumatoid arthritis or dermatological diseases, and higher osteoarthritis. The proportion of diabetics in the group of individuals with hsCOVID-19 symptoms was 20.5%, while in the group without symptoms was 11.5%. In the case of pulmonary disease, these percentages were 22.4% and 14.1%, respectively. The proportion of both bDMARDs and sDMARDs was also lower in the group with hsCOVID-19 symptoms.
Adjusted associations between different exposure variables (clinical characteristics and treatments) and hsCOVID-19 symptoms are shown inTable 3 . Diabetes and pulmonary disease were associated with an increased RR of hsCOVID-19 symptoms, with overall RRm1of 1.64 (95%CI 1.09, 2.47) and 1.47 (95%CI 1.02, 2.13). Regarding treatments, all bDMARDs presented a RR of 0.46 (95%CI 0.31, 0.67) and all sDMARDs presented a RR of 0.62 (95%CI 0.43, 0.91). Specifically, TNF-α antagonists presented RR of 0.50 (95%CI 0.33, 0.75) in the whole population. This protective effect was even higher in women (RR = 0.33; 95%CI 0.17, 0.64), while in men the effect size of the association was smaller and not statistically significant (RR = 0.76; 95%CI 0.41, 1.43). All types of TNF-α antagonists (adalimumab, certolizumab, etanercept, golimumab and infliximab) showed RR estimates < 1, although were only statistically significant for adalimumab (RR=0.53, CI95% 0.31, 0.93) and etanercept (RR=0.37, CI95% 0.16, 0.88). The RR of anti-IL6/12/17/23 compounds were 0.47 (95%CI 0.24, 0.92); with potentially higher effects of anti-IL17 (RR=0.20; CI95% 0.03-1.38) than anti-IL23(12) (RR=0.80; CI95% 0.39, 1.65). Methotrexate and chloroquine/hydroxychloroquine presented an RR of 0.71 (95%CI 0.46, 1.08) and 0.76 (95%CI 0.36, 1.62), respectively. The RR of leflunomide was 0.66 (95%CI 0.28, 1.58) in the whole population, with higher risk reduction in men (RR = 0.36; 95%CI 0.07, 1.75) than in women (RR = 0.81; 95%CI 0.29, 2.87). Glucocorticoids at doses of ≤ 10 mg/day also showed a risk reduction of about 30% in women (RRm1 = 0.72, 95%CI 0.42, 1.22). Figure 1 represents the adjusted RR for presenting hsCOVID-19 symptoms according to the exposure to different treatments in men and women. The interactions between most prevalent combinations of treatments were included in Model 4(Supplementary Table 4 ) and were not statistically significant, except for the interaction between biologic and synthetic DMARDs (RR = 4.3; CI95% 2.00, 9.25).
A sub-analysis in individuals aged 60 or older (N=1228), compared to individuals under 60 (N=1266) was performed (Supplementary Table 5 ). The association between anti-TNF-α and hsCOVID-19 symptoms was maintained in both groups (RR<60 = 0.55; CI95% 0.33, 0.91 vs RR = 0.43; CI95% 0.20, 0.92). In the case of anti-IL6/12/17/23 compounds their effect was statistically significant in individuals under 60 years (RR<60 = 0.36; CI95% 0.15, 0.85). The effect of all biologic DMARDs was the same in both age groups, but this was not in the case of synthetic DMARDs, with RR<60 = 0.77 (CI95% 0.46, 1.29) and RR≥60 = 0.48 (CI95% 0.27, 0.86). The effect of glucocorticoids (≤ 10 mg/day) was higher in younger women (RR<60 = 0.50; CI95% 0.20, 1.24) than older women (RR = 0.91; CI95% 0.46, 1.79).
Finally, the adjusted RR estimates using propensity score matching for the exposure to each treatment are shown in Supplementary Table 6 . The RR of anti-TNFα was almost equivalent to the observed in the unmatched database (RR=0.65; CI95% 0.37-1.16) while for anti-IL6/12/17/23 compounds, glucocorticoids and chloroquine/hydroxychloroquine, the associations presented greater differences.