Discontinuation of hydroxychloroquine in older patients with systemic lupus erythematosus: a multicenter retrospective study

SLE cases were identified from three SLE databases at New York University School of Medicine, Hospital for Special Surgery, and Columbia University College of Physicians and Surgeons. Data were obtained by retrospective chart review. This study was approved by the Institutional Review Board of New York University School of Medicine.

Twenty-six patients met the following inclusion criteria: (1) ≥ 4 American College of Rheumatology (ACR) criteria for SLE [35] or Systemic Lupus International Collaborating Clinics (SLICC) classification criteria [36], (2) disease duration ≥ 5 years, (3) HCQ use of 200–400 mg/day ≥ 5 years, (4) discontinuation of HCQ at age ≥ 55 years, (5) prednisone ≤ 7.5 mg/day, and (6) a clinical Safety of Estrogens in Lupus Erythematosus: National Assessment version of the Systemic Lupus Erythematosus Disease Activity Index (SELENA-SELEDAI) [37] instrument score of ≤ 4. The comparator group comprised 32 age- (within 3 years), gender-, and race/ethnicity-matched subjects with clinical data available during the time period matching the patients who discontinued HCQ.

This was a retrospective chart review study. The primary outcome was the occurrence of a lupus flare within 1 year of HCQ withdrawal or matched time of continuation, by the revised version of the SELENA-SLEDAI Flare composite index (rSFI) that separates mild from moderate flares, evaluates each organ system separately, and incorporates increases in corticosteroid dose and/or addition of immunosuppressive agents [37, 38]. Secondary outcomes included incidence of mild, moderate, or severe flares, classified by the rSFI. The adjudication of flares and their severity was performed by two of the investigators (PI and JB), blinded to the patients’ group. Additional clinical secondary outcomes included death from any cause, venous thrombosis, and cardiovascular events within 1 year of HCQ discontinuation or matched time of continuation.

Categorical variables were summarized by computing counts and proportions of patients (%). Continuous variables are expressed as mean ± standard deviation (SD) or median with interquartile range (IQR), as appropriate. Baseline characteristics were compared between the HCQ withdrawal and continuation groups in analyses unadjusted for matching and confounders using the chi-square or Fisher’s exact test for categorical variables and the two-sample T test or Mann-Whitney U test for continuous variables. To assess the association between HCQ status and the occurrence of flare during the 12-month period of interest, which was initially considered a binary outcome, generalized linear mixed models (GLMM) using the logit link were fit to the data to account for the matched design and potential confounders. Because the sample size and number of flares in the study limited the number of confounders that could be included as independent covariates in the model, a propensity score analysis was also conducted. Specifically, for each patient, a propensity score was estimated from a logistic regression model that was fit with HCQ withdrawal status as the outcome and years since diagnosis of SLE, years of HCQ use, low C3 or C4, SLEDAI, number of ACR criteria for SLE, history of lupus nephritis, immunosuppressive use, and presence of anti-double stranded DNA antibodies as predictors. Given that the patients were already matched by age, race, and gender, the covariate adjustment method was used in the propensity analysis, where the propensity score was included as a covariate, along with HCQ withdrawal status, in the GLMM model. Missing data rates ranged from 0 to 13.7% across study variables and were addressed in the GLMM analysis using multiple imputation with chained equations. The distribution of time to flare was estimated by the Kaplan-Meier method and compared between groups using the log-rank test. Two-sided p values < 0.05 were considered significant for all statistical analyses. All analyses were performed using SAS version 9.4 and SPSS version 26.

Fifty-eight patients were included in the study. Twenty-six patients discontinued HCQ, and 32 patients on HCQ were matched at the time of discontinuation. Baseline characteristics are summarized in Table 1. There were no significant differences between the two groups with regard to age, gender, race/ethnicity, C3 and C4 levels, clinical SLEDAI score, proportion of patients with positive anti-dsDNA antibodies, or history of lupus nephritis. The duration of SLE was longer in the HCQ withdrawal group than in the HCQ continuation group (24.3 ± 10.6 years vs. 17.8 ± 11.8 years, p = 0.03). Patients who discontinued HCQ had a slightly lower number of accumulated ACR classification criteria than the comparator group (4.6 ± 0.9 vs. 5.4 ± 1.5, respectively, p = 0.04). The proportion of patients on prednisone (7.7% vs. 15.6%) and other immunosuppressants (15.4% vs. 25%) were lower in the HCQ withdrawal group. C3 and C4 levels were marginally higher in the HCQ withdrawal group (108.1 ± 16.4 vs. 100.5 ± 27.6 for C3; 26.0 ± 10.6 vs. 21.3 ± 10.3 for C4) as compared to the comparator group, with the proportion of patients with low C3 or C4 being higher in the HCQ continuation than in the HCQ withdrawal group (14/32 [43.8%] vs. 3/25 [12%], respectively; p = 0.02). The proportion of patients with type 2 diabetes mellitus (T2DM) at baseline was similar between the groups (7.7% of patients in the HCQ withdrawal group vs. 6.3% in the comparator group). A higher proportion of patients in the HCQ withdrawal group was taking statins at baseline compared to the HCQ continuation group (5/26 [19.2%] vs. 3/32 [9.4%]).

Five out of 26 patients in the HCQ withdrawal group (19.2%) and 5 out of 32 patients in the HCQ continuation group (15.6%) experienced a flare of any severity, corresponding to an estimated odds ratio of 1.28 (95% CI 0.31, 5.30; p = 0.73). After adjusting for years since diagnosis of SLE, low C3 or C4, number of ACR criteria, and SELENA-SLEDAI score (i.e., the baseline characteristics that differed between the HCQ withdrawal and continuation groups at the p < 0.10 level in Table 1), results were similar (OR = 1.31; 95% CI 0.18, 9.49; p = 0.78). The estimated odds ratio for flare from the propensity score analysis was lower (OR = 1.18; 95% CI 0.23, 6.16; p = 0.84). The Kaplan-Meier plots of time to flare for both the HCQ continuation and withdrawal groups are shown in Fig. 1 (log-rank = 0.67).

There were no severe flares during the 12 months following HCQ discontinuation or continuation. The rate of moderate flares was lower in the HCQ withdrawal group than in the HCQ continuation group (7.7% vs. 15.6%), corresponding to an unadjusted odds ratio of 0.45 (95% CI 0.10, 2.72; p = 0.37). After adjusting for propensity score, the estimated odds ratio for moderate flare was 0.50 (95% CI 0.07, 3.82; p = 0.50).

The clinical manifestations and treatment of flares are described in Table 2. Four patients in the HCQ withdrawal group developed cutaneous manifestations, including alopecia and pernio, discoid, or malar rash. One of these patients also developed arthritis, and another patient had polyarthritis requiring methotrexate. In the HCQ continuation group, there were three patients with cutaneous flares (discoid rash and alopecia), two of which also had polyarthritis. One additional patient who continued HCQ had polyarthritis as the only clinical manifestation, and another patient had pericarditis in this group. No patient in either group developed new manifestations of lupus during the study period.

The most common reason for HCQ discontinuation was retinal toxicity (11/26, 42.3%), followed by patient’s preference (9/26, 34.6%), other confirmed or suspected adverse effects (4/26, 15.4%), ophthalmologist recommendation for macular degeneration (1/26, 3.8%), and rheumatologist recommendation for quiescent SLE (1/26, 3.8%). One patient discontinued HCQ for biopsy-proven cardiac toxicity. No lupus flares occurred in the patients who discontinued HCQ due to maculopathy or cardiotoxicity. Two patients in the HCQ continuation group developed maculopathy after the 12-month post-matching period, requiring discontinuation of the drug. As maculopathy was diagnosed after matching and chart reviewing process, and due to a lack of follow-up after HCQ discontinuation, these patients were maintained in the HCQ continuation group.

No patients in the HCQ withdrawal group (excluding patients with T2DM at baseline) and only one out of 30 patients in the comparator group developed T2DM during the 12 months following HCQ discontinuation or time of matching. There were no acute coronary syndromes, strokes, or other events associated with arterial thrombosis in either group during the 12-month study period. In the HCQ withdrawal group, one patient developed a postsurgical episode of unilateral lower extremity deep venous thrombosis and pulmonary embolism that was managed as a provoked venous thromboembolic event with 3 months of anticoagulation. One patient in the HCQ discontinuation group was started on a statin during the study period. There were no deaths during the 12-month post-matching period in any of the groups.