This retrospective cohort study was conducted at the outpatient rheumatology clinics of the First Medical Center of the Chinese People’s Liberation Army (PLA) General Hospital, a tertiary referral center in Beijing, China. Patients attending the clinics were invited to participate in the study and complete questionnaires that included demographic data, AS disease characteristics, and COVID-19 infection details. Besides, evaluation of AS disease activity (Bath Ankylosing Spondylitis Disease Activity Index, BASDAI) (Garrett et al., 1994), function level (Bath Ankylosing Spondylitis Functional Index, BASFI) (Calin, et al., 1994) and physical mobility (Bath Ankylosing Spondylitis Metrology Index, BASMI) (Jenkinson et al., 1994) were conducted by a fellowship-trained physician.

COVID-19 infection details included the results of SARS-CoV-2 PCR or antigen tests, vaccination status, COVID-19 symptoms, and the time taken for patients to recover. Patients aged 18 years or older were enrolled in the study from 20 December, 2022, to 31 March, 2023, if they met the 1984 modified New York criteria for AS (van der Linden et al., 1984) and had mild-moderate COVID-19. Patients were identified as having COVID-19 if they had a positive SARS-CoV-2 PCR or antigen test, and the day of the positive test was considered as the index day. The severity of COVID-19 was determined based on the Chinese Diagnosis and Treatment Protocol for COVID-19 (Trial Version ten) (General Office of the National Health Commission, 2023). Mild COVID-19 was defined as patients having upper respiratory infection symptoms as their predominant manifestation, such as fever, cough, or sore throat. Moderate COVID-19 was defined as patients having persistent fever, cough, or dyspnea, but without any of the following signs: respiratory rate ≥30 times per minute, oxygen saturation ≤93% when breathing ambient air, PaO2/FiO2 ≤ 300 mmHg, or lung infiltrates >50% area on images. Patients who denied having COVID-19 or whose AS medications were something other than non-steroidal anti-inflammatory drugs (NSAIDs), TNFi, and csDMARDs were excluded from the study. We retrospectively reviewed exposure variables before the index day and prospectively explored their influences on outcomes.

This study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines and complied with the Declaration of Helsinki.

NSAIDs, TNFi, and csDMARDs exposure was defined as patients being prescribed NSAIDs, TNFi, and csDMARDs within 12 months before the index date, respectively. TNFi included etanercept and its biosimilars, adalimumab, and infliximab. The range of csDMARDs in this study included sulfasalazine, methotrexate, leflunomide, thalidomide, and iguratimod. Given the clinical practice and the real-world background of the study, patients taking AS medications were grouped as NSAIDs monotherapy, NSAIDs + csDMARDs, csDMARDs monotherapy, and TNFi (with or without NSAIDs/csDMARDs).

Covariates such as age, sex, comorbidity, body mass index (BMI), smoking, alcohol consumption, and vaccination status might be associated with different outcomes in COVID-19. Therefore, they were considered confounding factors and adjusted for in further multivariable analysis. Comorbidities included diabetes, cardiovascular disease (CVD, including hypertension), and chronic obstructive lung disease (COPD). Vaccination status was classified as unvaccinated, partially vaccinated (one dose of inactive vaccine), fully vaccinated (two doses of inactive vaccine or one dose of adenovirus vaccine), and booster vaccinated (three doses of inactive vaccine or two doses of adenovirus vaccine).

The two key outcomes in this study were symptom burden and disease course. Fifteen symptoms related to COVID-19 were collected, including fever (peak temperature >37.3°C), chill, sore throat, hoarse voice, cough, nasal congestion, runny nose, headache, dizziness, dyspnea, myalgia, otologic symptoms, palpitation, abdominal pain, and diarrhea. Following a previous study (Hopkinson et al., 2021), we considered the number of self-reported COVID-19 symptoms as a proxy for disease burden and did not attempt to weigh different symptoms. In this study, the presence of more than 5 symptoms or 10 symptoms was arbitrarily categorized as increased symptom burden at two levels.

Defining long COVID in AS is challenging due to the lack of a globally accepted definition and the overlap of symptoms between long COVID and AS, such as fatigue and arthralgia (Baimukhamedov, 2023). In this study, we calculated the period between the index day and the day when patients reported returning to their “usual health.” Patients with a period longer than 10 days (LC10) or 28 days (LC28) were defined as having long COVID, to different extents.

In the descriptive analysis, we assessed the differences in proportions and medians of variables between the exposure group and the unexposed group using chi-squared tests for categorical variables and t-tests or Mann-Whitney U tests for continuous variables. Missing data were addressed using multiple imputations with five iterations, assuming that the data were missing at random.

The main analysis compared baseline medication exposure versus no medication use on COVID-19 symptom burden and disease course. First, to evaluate the association between COVID-19 symptom burden and baseline medication exposure, we used modified Poisson regression with a robust (sandwich) estimation of variance (which allows for binary variables) (Zou, 2004) to calculate the relative risk (RR) and 95% confidence interval (CI). We performed multivariable analysis to adjust for confounding factors, including age, gender, comorbidity, overweight (BMI >25) (Carnethon et al., 2012), smoking, alcohol consumption, and vaccination status. Similar analyses were conducted for the outcome of long COVID. Subsequently, we subgrouped patients treated with TNFi into TNFi monotherapy or combination with NSAIDs or csDMARDs to further explore the influence of TNFi on COVID-19 outcomes using the same approach. Forest plots were used to visualize the results using the R statistical program (Ver 4.0.3) with the forest plot package. Other statistical analyses were performed using SPSS Statistics (version 22; IBM Corp.). A p-value <0.05 was considered statistically significant for all analyses.

Questionnaires were collected from 658 AS patients, out of which 112 patients reported negative results for COVID-19 infection. Additionally, 25 patients were taking other AS medications, including 11 patients using Secukinumab, 6 patients using Tofacitinib, and 8 patients using traditional Chinese medicines. These patients were excluded from the analysis, leaving a total of 521 patients (Figure 1). Among these patients, the median age was 34.8 (interquartile range, IQR: 27.2–46.7). Of the 521 patients, 420 (80.6%) were male, 52 (10.0%) had at least one comorbidity, 442 (84.8%) had HLA-B27 positivity, and 443 (85%) had been vaccinated (Table 1).

Regarding AS treatments, 71 patients (13.6%) underwent NSAIDs monotherapy, 95 (18.2%) received NSAIDs in combination with csDMARDs, 38 (7.3%) had csDMARDs monotherapy, 232 (44.5%) were treated with TNFi, and 85 (16.3%) did not receive any of the mentioned medications. Compared to patients who reported no AS medication usage, patients who received TNFi were more likely to be men (196 [84.5%] vs 66 [77.6%]) and less likely to be older than 55 years (6 [2.6%] vs 5 [5.9%]). Patients with NSAIDs monotherapy were more likely to be overweight (38 [53.5%] vs 31 [36.5%]) and have COPD (4 [5.6%] vs 1 [1.2%]), but less likely to be older than 55 years (1 [1.4%] vs 5 [5.9%]). On the other hand, patients with csDMARDs monotherapy were more likely to be older than 55 years (5 [13.2%] vs 5 [5.9%]), but less likely to be male (26 [68.4%] vs 66 [77.6%]). Besides, patients treated with csDMARDs monotherapy, NSAIDs & csDMARDs or TNFi had lower BASDAI (median and IQR was 2.2, (1.2, 3.8), 2.1 (1.0, 3.5) and 2.2 (1.2, 4.0), respectively) and BASMI (3.0 (0, 6.3), 1.0 (0, 5.0) and 3.0 (0, 5.0), respectively) than patients without previous medication use (3.0 (1.6, 4.4) for BASDAI and 4.0 (0, 5.0) for BASMI), whereas patients with NSAIDs monotherapy had worse BASFI (1.8 (0.2, 4.8)) and comparable BASDAI (3.0 (1.4, 4.2)), BASMI (4.0 (0, 5.0)) than patients without previous medication use ((3.0 (1.6, 4.4) for BASDAI, 4.0 (0, 5.0) for BASMI and 1.2 (0.1, 4.4) for BASFI) (Table 1).

The median (IQR) number of symptoms reported by patients with no medication usage, NSAIDs monotherapy, csDMARDs monotherapy, and TNFi were 6.0 (5.0, 8.0), 8.0 (6.0, 9.0), 7.5 (5.0, 9.0), and 7.0 (5.0, 9.0), respectively (Table 2). Detailed COVID-19 symptoms are presented in Supplementary Table S1. The univariate analysis revealed that patients treated with AS medications had similar odds of experiencing more than 5 symptoms compared to those without medication use, except for patients with NSAIDs monotherapy (RR 1.26, 95% CI: 1.05–1.51). After adjusting for age, gender, obesity, smoking, alcohol drinking, coexisting comorbidities, and vaccination status, the association of csDMARDs or TNFi with the risk of experiencing more than five symptoms remained insignificant (adjusted RR (RRa) 1.07, 95% CI: 0.83–1.39 or 1.10, 0.93–1.30, respectively). However, patients with NSAIDs monotherapy still had increased odds of experiencing more than 5 symptoms (RRa 1.22, 95% CI: 1.01–1.46). Additionally, patients with COPD had a significantly increased risk of experiencing more than 5 symptoms compared to those without baseline coexisting diseases (RRa 1.34, 95% CI: 1.16–1.56) (Figure 2).

A similar analysis was conducted to evaluate the risk of experiencing more than 10 symptoms in AS patients. Patients treated with NSAIDs, csDMARDs and TNFi had similar odds of experiencing more than 10 symptoms compared to those without medication use in both univariable analysis and after adjusting for confounding factors (RRa 0.65, 95% CI: 0.26–1.64; 0.95, 0.36–2.54; and 1.01, 0.53–1.91 for NSAIDs, csDMARDs and TNFi, respectively). Likewise, COPD was associated with greater odds of experiencing more than 10 symptoms (RRa 5.94, 95% CI: 2.54–13.90) (Figure 2).

The median (IQR) duration of COVID-19 symptoms in patients with no medication use, NSAIDs monotherapy, csDMARDs monotherapy, and TNFi were 8.0 (5.5, 13.5), 9.0 (6.0, 13.0), 9.5 (6.8, 14.5), and 9.0 (6.0, 16.0) days, respectively (Table 2). The univariate analysis revealed that patients with different AS medications had similar odds of long COVID (LC10) compared to those without medication use. Importantly, these associations were reproduced after adjusting for confounding factors (RRa 1.17, 95% CI: 0.82–1.66; 1.18, 0.78–1.77; and 1.22, 0.92–1.63 for NSAIDs, csDMARDs and TNFi, respectively) (Figure 3).

Similarly, when LC28 was considered as an increased disease course in COVID-19, patients with NSAIDs, csDMARDs, and TNFi had similar odds of LC28 compared to those without medication usage, both in univariate analysis and after adjustment (RRa 0.94, 95% CI: 0.31–2.81; 0.97, 0.25–3.74; and 1.05, 0.44–2.49 for NSAIDs csDMARDs and TNFi, respectively). Additionally, having cardiovascular disease (CVD) was found to be associated with greater odds of LC10 (RRa 1.52, 95% CI: 1.12–2.06), but not LC28 (RRa 1.53, 95% CI: 0.59–3.92) in the population (Figure 3).

Among the 232 patients treated with TNFi, 104 patients received TNFi monotherapy (Group 1), while 128 patients were prescribed NSAIDs or csDMARDs concomitantly (Group 2). The median (IQR) number of symptoms reported by patients in Groups 1 and 2 were 7.0 (5.0, 9.0) and 8.0 (6.0, 10.0), respectively (Table 2). After adjustment for potential confounding factors, patients with TNFi monotherapy or combination therapy had similar odds of experiencing more than five symptoms compared to those without medication use (RRa 1.02, 95% CI: 0.84–1.24 and 1.18, 0.99–1.40 for Group 1 and 2, respectively). It was also revealed that combining NSAIDs or csDMARDs did not affect the odds of experiencing more than 10 symptoms in TNFi treatment (RRa 0.65, 95% CI: 0.29–1.45 and 1.32, 0.68–2.56 for Group 1 and 2, respectively) (Figure 2).

The median (IQR) duration of COVID-19 symptoms in Groups 1 and 2 were 9.0 (6.0, 14.0) and 10.0 (6.0, 18.0) days, respectively (Table 2). In multivariable analysis, patients with TNFi monotherapy or combination therapy had similar odds of long COVID (LC10) compared to those without medication use (RRa 1.10, 95% CI: 0.79–1.54 and 1.32, 0.97–1.78 for Group 1 and 2, respectively). Similar results were obtained when LC28 was used as the outcome variable (RRa 1.33, 95% CI: 0.54–3.28 and 0.82, 0.30–2.23 for Groups 1 and 2, respectively) (Figure 3).