Study population.
During the evaluation period, 2091 patients were surveyed. The rate of non-responders was 1.96% and, among them, the great majority (35 out of 41, 85.3%) were confirmed to be alive and without signs of infections by a relative, although they could not be reached directly for a telephone interview. These patients were not included in the study population, which finally encompassed 2050 subjects. Demographic and clinical characteristics of the study population are summarized in Table 1. Mean (±SD) age was 58 (±15) years, and the majority of the patients were female (66%). Age and gender distribution within different diagnostic subgroups were as expected according to the specific type of arthritis. Nearly all (95.4%) had residency in a Province of Northern Italy with COVID-19 incidence ≥0.5%, with approximatively 10% residing in high-risk areas (incidence ≥1%) (Figure 1). Patients had established arthritis of long duration (median 10 years), and 62.3% were on treatment with biologic (b) or targeted synthetic (ts) disease modifying anti-rheumatic drugs (DMARDs) (a tumor necrosis factor (TNF) antagonist in the majority of the cases), alone or in combination with conventional synthetic (cs) DMARDs. A small proportion of our cohort (18.6%) was receiving hydroxychloroquine (HCQ). Detailed treatment disposition is shown in supplementary Table S1. Approximatively one third of the patients was on concurrent chronic treatment with GC. No significant difference in demographic, clinical and treatment characteristics were observed between the two recruiting centers (supplementary Table S2).
Frequency and characteristics of SARS-CoV-2 infection
During the observation period, 23 patients (1.1%) were diagnosed with COVID-19 according to nasopharingeal-swab (confirmed COVID-19). Twenty-nine additional patients (1.4%) reported acute respiratory illness having been in close contact with a confirmed COVID-19 case in the last 14 days prior to onset of symptoms (highly suspicious COVID-19). Two-hundred and sixty-one (12.7%) instead described respiratory symptoms without known contact with a positive COVID-19 case (unlikely COVID-19).
Table 2 summarizes the demographic and clinical characteristics of SARS-CoV-2 infection according to the different definitions. Of the 10 patients developing COVID-19 in course of b/tsDMARD treatment, 6 were receiving a TNF antagonist, 2 a JAK inhibitor, 1 abatacept and 1 secukinumab. The b/tsDMARD was taken in combination with a csDMARD in 5 cases (3 methotrexate (MTX), 1 sulfasalazine (SSZ), 1 leflunomide (LFN)). Four patients were on concurrent HCQ treatment. Of the cases developing COVID-19 in the absence of b/tsDMARD therapy, 8 were on MTX, 1 on SSZ, 1 on cyclosporin, and 1 on HCQ; none was receiving more than 1 csDMARD in combination. The two patients receiving neither a csDMARD nor HCQ were taking GC. Eleven patients (47.8%) required hospitalization with low-flow oxygen supplementation, and none was managed in intensive care unit; all recovered and were readmitted home. Of the 29 cases of highly probable SARS-CoV-2 infections, 11 (37.9%) were on b/tsDMARD treatment (45.4% on a TNF antagonist, 27.3% on a JAK inhibitor, 9.1% on secukinumab, 9.1% on tocilizumab, 9.1% on others), which was taken in combination with a csDMARD in 27.3% of the cases (MTX in all). One patient was receiving HCQ. Of the 18 highly suspicious cases not being treated with b/tsDMARDs, 6 were on MTX, 2 on MTX + HCQ, 2 on SSZ, and 2 on HCQ. Finally, treatment disposition of the 261 patients with acute respiratory symptoms without known contacts with COVID-19 cases was similar to that of non-COVID-19 cases.
Predictors of SARS-CoV-2 infection
Confirmed infection
Results from univariable analysis are presented in Table 3. None of the demographic characteristics significantly predicted laboratory-confirmed SARS-CoV-2 infection in our patients. Presence of hypertension conferred an OR (95% CI) of 2.09 (0.87 to 5.06). Precautions taken to prevent contagion were not significantly associated with outcome, whilst, as expected, SARS-CoV-2-positive patients more often reported close contacts with COVID-19 cases. Relevantly, arthritis treatment significantly impacted on the probability of SARS-CoV-2 infection. Use of GC indeed increased risk (OR [95% CI] 2.89 [1.26 to 6.62]), with the highest risk associated with daily prednisone doses ≥2.5 mg. Similarly, use of csDMARDs was associated with a trend towards higher odds of COVID-19, whilst use of b/tsDMARDs tended to reduce risk. No clear relationships between HCQ and confirmed COVID-19 cases were observed.
In multiple bivariavable models in which steroids, csDMARDs and b/tsDMARDs were entered separately (Table 4), use of GC was confirmed to independently predict increased risk of SARS-CoV-2 infection irrespective of comorbidities, precautions taken to prevent contagion and contacts with COVID-19 cases. In contrast, the trends for higher odds of infection in csDMARDs-treated and lower odds in b/tsDMARDs-treated patients were apparently not independent from other variables. GC maintained independent association also in bivariavable models including the other classes of anti-rheumatic drugs.
Highly suspicious infection
When the analysis was extended to include also highly suspicious SARS-CoV-2 infection, none of the demographic variables appeared consistently associated with the outcome. Again, the risk of SARS-CoV-2 infection appeared differently conditioned by the different classes of anti-rheumatic drugs. Patients on treatment with b/tsDMARDs were indeed significantly less affected, whilst use of GC, especially at doses ≥2.5 mg/d tended to increase the risk of infection (supplementary Table S3). No effects were seen in association with csDMARDs.
In multivariable analyses, treatment variables were confirmed as independent predictors of SARS-CoV-2 infection. In particular, use of GC emerged as a risk factor also in this setting including milder cases, whilst b/tsDMARD-treated patients less frequently developed symptoms of COVID-19. Even when forced in the model, csDMARDs did not modify the odds of infection (Table 5).
Unlikely infection
We then tested the associations of patients’ and disease characteristics with a more permissive definition of SARS-CoV-2 infection also including patients with acute respiratory symptoms in the absence of known contacts with COVID-19 cases. As shown in Supplementary Table S4, no significant predictors emerged among demographic and clinical characteristics apart from higher odds in younger patients. In contrast to confirmed and highly suspicious SARS-CoV-2 infection, none of the classes of anti-rheumatic drugs conferred increased or reduced risk. The neutral impact of GC and b/tsDMARDs was confirmed even when treatments were forced in multivariable analyses.