Does the use of corticosteroids and immunosuppressants increase the risk of COVID-19 infection among people with systemic lupus erythematosus?

Introduction

Patients with SLE are at higher risk for serious infections than the general population1 2 and infections are an important cause of morbidity and mortality among patients with SLE.3–5 While these facts may be due to immunological abnormalities, numerous studies suggest that the use of corticosteroids and immunosuppressants also increase the risk of infection.4 6 7

This issue is an especially important concern given the high incidence and relative virulence of COVID-19 infections, especially among patients with SLE.8 Concern about the increased susceptibility to infection has led some patients to discontinue immunosuppressant treatments leading to the potential increased risk of SLE flare.9 Several studies of patients without SLE have addressed the question of whether corticosteroids and immunosuppressants increase the risk of infection with COVID-19,10–13 but to our knowledge, no sizeable study has addressed this question among patients with SLE.

To address this question, we determined the rates of incident COVID-19 infection in our large single-centre cohort of patients with SLE and examined whether they were associated with recent use of corticosteroids, immunosuppressants or biologics. We also explored whether other variables, such as lupus disease activity and use of hydroxychloroquine, were associated with an increased or decreased incidence of incident COVID-19 infection.

Materials and methods

The Hopkins Lupus Cohort is a large clinical cohort of patients with SLE seen in Baltimore, Maryland. Participants are seen in the clinic on a quarterly basis at which time an extensive standardised set of measurements are made. These include information on treatments received, lupus disease activity, immunological biomarkers and numerous other clinical measures.

The cohort dataset was augmented by information on COVID-19 infection (based on a positive PCR or antigen test) and vaccination. This additional information was obtained from the electronic medical records and from patient self-report.

The analysis was based on the cohort data from March 2020 through August 2022. Each patient’s data were formatted into a dataset with one record per person month of follow-up. Each person-month (PM) record contained information on the treatments that the patient received that month, the most recent measure of disease activity, the time since vaccination and whether they experienced an episode of COVID-19 in that month. Some patients missed quarterly visits. If the time between two clinic visits exceeded 1 year, the clinical information for time-varying predictors was coded as missing after 1 year and the subsequent follow-up was excluded from the analysis. In addition, some predictor variables (eg, income) had missing values. Those with missing values were excluded from the analysis of those variables. Generally, this occurred <5% of the time.

To assess the association between exposures (eg, corticosteroid use) and risk of incident COVID-19 infection, we compared subgroups of follow-up months with respect to the proportion with an incident infection. Some quantitative predictors were categorised in multiple subgroups to provide information on whether the relationship between the predictor and rates observed a monotonic or threshold association. These categories were chosen prior to examining the outcomes.

The statistical significance of observed differences was assessed using logistic regression models fit by generalised estimating equations to account for the fact that some participants were infected with COVID-19 more than once. Multivariable models were fit to examine associations controlling for potential confounding.

Results

The analysis was based on 24 629 person-months of follow-up from 1052 different patients. The number of follow-up months per patient ranged from 1 to 33 months, with a median of 28 months. During the follow-up there were 339 different episodes of incident COVID-19 infection from 319 different patients. The rate of infection was 1.38% per person-month.

Table 1 shows the rate of infection by demographics and by vaccination phase. In our cohort, the rate was significantly lower among those over 45 years of age. There was no strong difference observed by other demographic variables. The rate of infection was highest in the most recent pandemic phase. There was a significant protective effect of the vaccination, except for during the most recent pandemic phase.

Table 2 shows the relationship between prospectively measured clinical variables and rate of COVID-19 infection. We did not observe a statistically significant difference in rate of infection in subgroups defined by the most recently measured SLE-related clinical variables such as total SLE disease activity, renal or haematological disease activity or serological markers.

Tables 3 and 4 show the rate of infection in subgroups of months defined by current treatments. There was no evidence of an association between either corticosteroid use or hydroxychloroquine and COVID-19 incidence. There was a surprisingly significant lower risk of COVID-19 infection among those on diuretics. We observed a significantly increased risk of COVID-19 infection among those taking either belimumab or tacrolimus.

To assess whether the observed associations (or lack of associations) seen in the univariate analyses could have been due to confounding by related variables, we fit two multivariate models (table 5). The first model focused on the association between corticosteroids and rate of infection controlling for age, pandemic phase, vaccination, SLE disease activity and taking other immunosuppressants. This model provided no evidence of an increased risk of COVID-19 infection with either low or higher dose of corticosteroids. The second model was fit to tease out the independent effect of some of the immunosuppressants that are sometimes used in combination, adjusting for age, pandemic phase, vaccination and disease activity. After adjusting for these variables and other immunosuppressants, the association between COVID-19 and exposure to either belimumab or tacrolimus were no longer statistically significant.

Discussion

In our SLE cohort, we did not observe strong evidence of an increased risk of COVID-19 infection among patients receiving corticosteroids. However, the CI for the rate ratio among those with corticosteroid doses exceeding 7 mg/day ranged from 0.67 to 1.82, indicating that we cannot rule out the possibility of a moderate increase in risk in this group. These findings regarding corticosteroids are reassuring, given previous studies that found a strong association between corticosteroid use and major infections in SLE.4 6 7 These findings are consistent with the findings of Adir et al,11 who found no evidence of an increased rate of COVID-19 positivity among patients with asthma taking corticosteroids. Our findings are also consistent with Patil et al,14 who observed similar rates of corticosteroid use among patients with and without a positive COVID-19 test in a large prospective cohort of patients with autoimmune rheumatic diseases in India. However, both Adir et al and Patil et al and several other studies11 14–16 observed an association between corticosteroid use and greater COVID-19 severity or mortality among those with COVID-19 infection. We did not have data on severity and death.

Regarding immunosuppressants, we observed an increased risk of COVID-19 among those taking tacrolimus or belimumab. However, after adjustment for potential confounding variables, these exposures were associated with an increased risk that was not statistically significant. Previously, we reported that the mean antibody response to COVID-19 vaccines was attenuated among those taking these medications.17 The association between tacrolimus and risk of COVID-19 infection was observed in the study by Nørgård et al.18 It should be noted that patients in our cohort on immunosuppressants other than rituximab were not prescribed Evusheld.

We observed no evidence of an increased risk among those taking other immunosuppressants including mycophenolate, methotrexate, azathioprine and rituximab. However, many of our patients on rituximab were given Evusheld to reduce their risk of COVID-19. In contrast, Patil et al observed a positive association between use of rituximab and incidence of COVID-19.14 In addition, several studies found an association between previous rituximab use and COVID-19 outcomes among patients infected with COVID-19.15 16 19

We found no evidence that hydroxychloroquine increased or reduced the risk of COVID-19 infection. This is consistent with the findings of Walbi et al.20

In our cohort, we did not observe a significantly lower rate of incident infection in vaccinated patients during the more recent phases of the pandemic. This is consistent with the limited efficacy of the vaccine against the Omicron variant as reported by Andrews et al.21

Our study has some limitations. First, our analysis only included symptomatic COVID-19 infections. Some asymptomatic infections might have been missed, and it is possible that such infections were more common in certain groups or given certain treatments, biasing our estimates of association. Second, we did not have information on the severity of outcomes of the COVID-19 infections. Risk factors for infection may not be the same as risk factors for serious infections. Third, we did not have information on social distancing behaviours, hand washing or use of masks that may have influenced the risk. Finally, the numbers of observations for some medications or subgroups were small leading to imprecise evaluations of some potential risk factors.

In summary, our data do not suggest that patients with SLE taking corticosteroids were at higher risk of incident COVID-19 infection. There was an increased risk among those taking belimumab or tacrolimus only in univariate models. The results were not significant in the multivariable models. Hydroxychloroquine use was not observed to be protective.

Our previous report22 (Petri et al17) found that belimumab and tacrolimus reduced vaccine response. In contrast with our previous report that found that mycophenolate reduced vaccine response, we did not find a significant increase in incident COVID-19 infection. With greater use of both belimumab and calcineurin inhibitors in lupus nephritis, our results point to patients with lupus nephritis as being at higher risk for poorer vaccine response but not for higher incident infection.