Aim
This study aimed to investigate the relationship between low-dose GC (≤7.5 mg PSL or equivalent) and infection occurrence in SLE patients, using a database of the multicenter registry of SLE patients in Japan.
Study design
This study was a prospective cohort study using data from the Lupus Registry of Nationwide Institutions (LUNA), a multicenter database of patients with SLE in Japan.
Setting
The participating institutions included the rheumatology departments of eight university hospitals and one general hospital. Because this study included data from multiple institutions from various regions of Japan, we were able to minimize selection bias and achieve geographical representativeness.
We collected data about GC use during the study period, and data regarding other patient characteristics were collected once a year at each hospital. Patient data were collected until the patient’s death, transfer out of the hospital, or patient opt-out from the registry.
Participants
Patients were recruited from February 2016 to September 2019. We included SLE patients over 20 years who fulfilled the American College of Rheumatology criteria for SLE classification,[14] including patients with SLE complicated with other connective tissue diseases (such as Sjogren’s syndrome). The included patients were followed up for more than 1 year. We excluded patients who received daily PSL >15 mg or equivalent during the study period because this dose is already known to be an infection risk, and our focus was on the relationship between low-dose PSL and infection. Patients with missing data regarding GC, infection, or follow-up were also excluded.
Variables
We recorded the following from the database: age, sex, comorbid diabetes mellitus (defined by a serum HbA1c >6.5%), biopsy-proven lupus nephritis (Class Ⅲ, Ⅳ, and Ⅴ according to the classification of the International Society of Nephrology and the Renal Pathology Society 2003 [15]), the SLE Disease Activity Index 2000,[16] blood serum data (C3, C4, CH50, anti-ds-DNA IgG antibody, and IgG), white blood cell count, concurrent use of multiple immunosuppressants use with GC (tacrolimus, cyclosporine, mycophenolate mofetil, azathioprine, and mizoribine), past use of intravenous immunosuppressant (methylprednisolone pulse therapy, intravenous cyclophosphamide, and rituximab), and hydroxychloroquine (HCQ) use. We recorded variables at the time of registration, and time-varying variables were also collected once a year when the patients were followed up in the registry.
Primary exposure
The primary exposure was systemic GC administration. The GC dose was averaged to the daily dose and converted to the PSL dose. The PSL dose was categorized as PSL 0‒2.5, 2.6‒5.0, 5.1‒7.5, and 7.6‒15.0 mg, as previously described.[17] Moreover, we considered this categorization to be clinically relevant as it is unknown whether PSL doses <7.5 mg could pose an infection risk, and the LLDAS aims to use a PSL dose <7.5 mg.
Outcome measure
The primary outcome was infection requiring hospitalization. This outcome was selected as it is clinically relevant for patients; further, infections requiring hospitalization are usually severe and may require changes in the therapy plan. Data on hospitalizations for infection were identified from the records of the hospitals participating in this study. Hospital records were also reviewed to verify whether patients were admitted to any other hospitals not participating in LUNA. Infections were classified as respiratory infection, urinary tract infection, abdominal infection, soft tissue and bone infection, neurological infection, and others. In patients with multiple infections, each infection was considered and evaluated.
Statistical analysis
Age, SLE Disease Activity Index, blood serum data, and white blood cell count were used as continuous variables. Sex, comorbid diabetes mellitus, lupus nephritis, PSL dose group, concurrent use of multiple immunosuppressants, past use of intravenous immunosuppressive therapy, immunosuppressive therapy, HCQ use, and infection occurrence were the categorical variables analyzed. Summary statistics are presented as mean values with standard deviations and as numbers with proportions.
We conducted a multivariable analysis using time-dependent Cox regression analysis to assess the hazard ratio of infection occurrence with PSL dose, referenced to PSL or equivalent dose of 0‒2.5 mg. The covariates selected were age, sex, and concurrent immunosuppressant use, because these variables are clinically important for infection occurrence in SLE patients, as previously reported.[18–20] The covariate data were collected at registration and every other year.
We conducted two sensitivity analyses. First, we conducted a time-dependent Cox regression analysis among the patients who did not receive HCQ, because HCQ has a protective effect against infection in SLE.[21] Second, we conducted a Cox regression analysis for infection risk with total PSL dose in 1 year in each group among the patients whose PSL dose did not change over the course of 1 year. This was because the PSL dose was changed every hospital visit and therefore we were unable to accurately determine the effect of GC after every dose change. We conducted the same analysis as the primary analysis in the sensitivity analyses.
We assumed that data were missing at random, and thus missing data were handled by using multiple imputations, a method for handling missing data in epidemiological and clinical research, for multivariate analysis. Statistical significance was defined as a two-sided p-value <0.05. All statistical analyses were conducted using STATA 15.1 (Stata Corp LP, College Station, TX)
Patient and public involvement
Patients and the public were not involved in the design, conduct, and reporting of this research.