Hydroxychloroquine Dose and Hospitalizations for Active Lupus

Abstract

Objectives:

We sought to determine the impact of hydroxychloroquine (HCQ) dose on the risk of hospitalizations for systemic lupus erythematosus (SLE).

Methods:

We conducted a case-crossover study within an academic health system, including patients with SLE who used HCQ and had ≥ 1 hospitalization for active SLE between January 2011 and December 2021. Case periods ended in hospitalization for SLE, whereas control periods did not. The exposures were the average weight-based HCQ dose, categorized as ≤5 or >5 mg/kg/day, and non-weight-based HCQ dose, categorized as <400 or 400 mg/day, assessed during each 6-month case or control period. Odds ratios (OR) were calculated using conditional logistic regression and adjusted for prior disease activity, kidney function, glucocorticoid use and other immunosuppressant use.

Results:

Of 2,974 patients with SLE who used HCQ (mean age 36.5 years; 92% female), 584 had ≥1 hospitalization with primary discharge diagnosis of SLE. Of these, 122 had ≥1 hospitalization for active SLE while using HCQ and had ≥1 control period with HCQ use during the study period. Lower HCQ weight-based dose (≤5 vs. >5 mg/kg/day) and non-weight-based dose (<400 vs. 400 mg/day) were each associated with increased hospitalizations for active SLE (adjusted OR 4.20 [95% CI 1.45–12.19] and 3.39 [95% CI 1.31–8.81]).

Conclusions:

The use of lower doses of HCQ was associated with an increased risk of hospitalizations for active SLE. Although the long-term risk of HCQ retinopathy must be acknowledged, this must be balanced with the short-term and cumulative risks of increased SLE activity.

INTRODUCTION

Systemic lupus erythematosus (SLE) is a heterogeneous and multifaceted disease that has the potential to affect multiple organ systems and leads to frequent hospitalizations (1). Hydroxychloroquine (HCQ), an anti-malarial medication associated with wide-ranging benefits for patients with SLE, is the cornerstone therapy for the majority of patients with SLE. Its benefits include reduction of SLE flares, damage progression, cardiovascular events, and pregnancy complications, as well as improving overall survival (2–4). Several mechanisms are proposed for how HCQ produces these benefits, prime among them being by inhibiting the function of lysosomes and antagonizing toll-like receptors (3).

In general, HCQ is well-tolerated by patients, but one of the most concerning potential toxicities is the long-term risk of HCQ retinopathy, which can, although rarely, result in irreversible vision loss (3, 5, 6). This retinopathy has been found to be associated with both dose and duration of HCQ administration (5–7). The mechanism of HCQ retinal toxicity is similarly thought to be through inhibition of lysosomal function, with reduced autophagy leading to increased concentration of lipofuscin in retinal pigment epithelial cells, resulting in photoreceptor degradation (7).

Evolving evidence and increased concern for the risk of retinopathy among long-term HCQ users led the American Academy of Ophthalmology (AAO) to recommend limiting HCQ dosing to <6.5 mg/kg of ideal body weight in 2011, and this recommended dose-threshold was reduced further by the AAO to <5 mg/kg in 2016, now based on real body weight (8, 9). In accordance with these ophthalmology guidelines, the European League Against Rheumatism now recommends using ≤5 mg/kg of HCQ for patients with SLE (10), and this HCQ dosing threshold is a 2023 quality measure in the American College of Rheumatology’s RISE registry (11).

These weight-based HCQ dosing guidelines have resulted in the use of lower doses of HCQ for many patients than they would otherwise receive. Whereas 400 mg/day had been previously prescribed for most patients with SLE, 400 mg exceeds 5 mg/kg/day for the majority of patients with SLE (8, 9). However, low-dose HCQ, per the current guidelines, may not maintain the same benefits as higher dosing on important outcomes for patients with SLE. A recent case-crossover study found an increased risk of SLE flares associated with using ≤5 mg/kg dosing, in accordance with current guidelines (12), and a study from the Systemic Lupus Erythematosus International Collaborating Clinics (SLICC) multinational lupus cohort found that HCQ dose-reduction and discontinuation were associated with an increased risk of lupus flares and hospitalizations, but did not examine the association between specific dose and this outcome (13).

Our study aimed to determine whether the use of lower HCQ dosing was associated with an increased risk of hospitalizations for active SLE.

PATIENTS AND METHODS

Study Setting and Source Population

We identified patients with SLE from the observational Mass General Brigham lupus cohort, including two US academic medical centers and associated community hospitals, using a previously validated electronic health record (EHR)-based algorithm (14). Within this group, we identified all patients who had one or more visits for SLE (ICD-9 710.0 or ICD-10 M32.*, excluding M32.0 Drug-Induced SLE) during the study period of January 2011 through December 2021 and used HCQ during that period.

This study was approved by the Mass General Brigham Institutional Review Board.

Study Design

We conducted a retrospective case-crossover study in which each patient served as his or her own control. This design mitigates potential time-invariant confounders. The outcome of interest was hospitalization for active SLE. To identify these outcomes, we identified patients with one or more hospitalizations with a primary discharge diagnosis of SLE during the study period. We reviewed the medical records for each hospitalization identified to determine if it was for active SLE. Rheumatology consults notes were reviewed to determine whether the presentation warranted treatment with immunosuppression indicating increased SLE activity and excluded hospitalizations for other primary indications (e.g., infection, renal failure other than attributed to active lupus nephritis, other comorbid conditions). To address the study question of interest, we also excluded hospitalizations for which patients were not treated with HCQ in the preceding 6 months.

For all patients with a hospitalization for active SLE while using HCQ, we identified case periods and control periods. We defined a case period as the 6-months prior to a hospitalization for SLE. When there were multiple SLE hospitalization events within a 6-month period, only the first event generated a case period. For each patient, we included up to three non-overlapping case periods for separate hospitalization events. We defined control periods as 6-month periods with no hospitalizations for SLE (Supplemental Figure 1). A rheumatology visit was required within 3 months of the start of each case or control period. Similar to case periods, we excluded control periods where patients did not use HCQ. Up to three non-overlapping control periods per patient were selected. To avoid potential bias in selecting control periods, the first eligible control periods were selected in the following order of priority: 1) 12–18 months prior to SLE hospitalization, 2) 6–12 months prior to SLE hospitalization, 3) 18–24 months prior to SLE hospitalization, 4) 6–12 months after SLE hospitalization, 5) 0–6 months after SLE hospitalization. Patients with at least one eligible case period and one eligible control period were included in the case-crossover analysis.

Exposure

The exposure of interest was the mean HCQ dose, assessed in each 6-month case and control period. The dose was derived from prescription records and modified when applicable according to physician report of the dose a patient was taking at clinical encounters. We assessed the mean weight-based HCQ dose, categorized as ≤5 or >5 mg/kg/day, (9, 10) and the low dose group was further stratified into <4 and 4–5 mg/kg/day. We also assessed the mean non-weight based HCQ dose, categorized as <400 or 400 mg/day, and the low dose group was further stratified into <300 and 300 to <400 mg/day. Dose changes that occurred during a 6-month period were incorporated into the mean dose calculations for that period. Six-month case or control periods without HCQ use did not contribute to the case-crossover analysis.

Covariates

For the overall lupus cohort, we assessed baseline demographics, anthropometrics, lupus nephritis, chronic kidney disease (CKD) stage ≥3, prescribed medications, including glucocorticoids and immunosuppressants, and SLE serologic markers (e.g., dsDNA positivity and hypocomplementemia) at the first encounter for SLE during the study period. Race and ethnicity were obtained from electronic medical records. For patients included in the case-crossover analysis, we assessed these covariates at the first outpatient rheumatology clinic visit of the case-crossover period. We also assessed time-varying estimated glomerular filtration rate (eGFR), glucocorticoid use, SLE immunosuppressant use, and disease activity (e.g., clinical Systemic Lupus Erythematosus Disease Activity Index [SLEDAI] score assessed from rheumatologist notes) at the most recent rheumatology visit prior to each case or control period.

Statistical Analysis

Odds ratios (OR) were calculated using conditional logistic regression, matched within a subject, to examine the association of HCQ dose with the risk of hospitalization for active SLE. We adjusted for eGFR, glucocorticoid use, immunosuppressant use, and clinical SLEDAI score prior to each 6-month observation period. All analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).

RESULTS

We identified a cohort of 2,974 patients with SLE who had at least one clinical encounter for SLE and used HCQ between January 2011 and December 2021, including 584 patients with at least one hospitalization with a primary discharge diagnosis of SLE during this period (Figure 1). Of these 584 patients, 122 patients had at least one hospitalization for active SLE while using HCQ during the study period and were included in the case-crossover study. These patients contributed 137 case periods and 266 control periods. The mean age was 36.5 years (SD 14.3) at the first visit of the case-crossover study, and 92% were female (Table 1). The majority (64%) of the overall HCQ user SLE cohort were White, but only 42% of the patients included in the case-crossover study were White. Of these patients included in the case-crossover study, 31% were Black, 12% were Hispanic or Latino/a, and 9% were Asian. At cohort entry, 5% of patients in the overall cohort had CKD stage ≥3 and 5% had lupus nephritis. Among patients included in the case-crossover study, 17% had CKD and 34% had LN at the onset of the case-crossover study. In addition to HCQ, 63% were on glucocorticoids and 53% used a SLE immunosuppressant.

Figure 1. Flowchart of Patients Included in the Case-Crossover Study with Inclusion and Exclusion Criteria.

During the study period, patients with SLE were identified who had at least one outpatient visit with a diagnosis of SLE, and were prescribed HCQ during the study period of January, 2011-December, 2021. Of this larger cohort, patients were identified who were hospitalized with a primary discharge diagnosis for SLE (ICD-9 710.0 or ICD-10 M32.* excluding M32.0) during the study period. Patients were excluded from the case-crossover study if they were not hospitalized for active SLE, but for alternative indications such as infection, and they were excluded if they did not use HCQ during the 6-month case period prior to the hospitalization for SLE. Patients were also excluded if they did not have an eligible 6-month control period, with HCQ use and no hospitalizations for SLE, during the study period. The remaining patients were included in a case-crossover study.

Table 1.

Baseline Characteristics

Baseline Characteristics*	Overall (n= 2,974)	Hospitalized for Active SLE, Included in Case-Crossover Study (n=122)
Age (years), mean (SD)	44.6 (15.9)	36.5 (14.3)
Female, n (%)	2707 (91.1)	112 (91.8)
Race, n (%)
White	1909 (64.2)	51 (41.8)
Black	486 (16.4)	38 (31.1)
Asian	182 (6.1)	11 (9.0)
Other/unknown	397 (13.3)	22 (18.0)
Ethnicity
Not Hispanic or Latino/a	2,841 (95.5)	107 (87.7)
Hispanic or Latino/a	133 (4.5)	15 (12.3)
Weight (kg), mean (SD)	72.4 (19.5)	76.8 (26.5)
Weight < 80 kg, n (%)	1996 (72.0)	73 (59.8)
Chronic kidney disease stage ≥3, n (%)	138 (4.6)	21 (17.2)
Lupus Nephritis, n (%)	135 (4.5)	42 (34.4)
SLE Serologies, n (%)
dsDNA antibody	653 (45.7)	62 (73.8)
Low complement levels	659 (52.0)	70 (78.7)
SLE Medications, n (%)
Glucocorticoids	462 (15.5)	77 (63.1)
Immunosuppressant	243 (8.2)	64 (52.5)

^*Assessed at the first clinical encounter for SLE during the study period for the overall cohort and assessed at the first encounter during the case-crossover study for patients hospitalized for active SLE and included in the case-crossover study.

In the case-crossover study, when comparing the risk of hospitalizations for SLE according to weight-based HCQ dose in the preceding 6-month period, there was an elevated unadjusted OR of 4.68 (95% CI 1.66–13.20) associated with using ≤ 5mg/kg/day vs >5 mg/kg/day (Table 2). The adjusted OR was 4.20 (95% CI 1.45–12.19). An increased odds of SLE hospitalizations was also observed for HCQ dosing <4 mg/kg/day compared with >5 mg/kg/day dosing (adjusted OR 4.62 [95% CI 1.50–14.20]). There was a trend towards a higher OR for SLE hospitalization associated with HCQ dosing in the range of 4–5 mg/kg/day compared with >5 mg/kg/day. When comparing the risk by non-weight-based HCQ dose, there was an elevated unadjusted OR of 3.78 (95% CI 1.51–9.51) associated with using <400 mg/day vs 400 mg/day, and the adjusted OR was 3.39 (95% CI 1.31–8.81). Within the low-dose non-weight based group, there was not a significant difference between dosing 300 to <400 mg/day and 400 mg/day dosing, but there was an increased OR for hospitalizations for SLE associated with less than 300 mg/day dosing (adjusted OR 4.68 [95% CI 1.66–13.17]).

Table 2.

Association Between Hydroxychloroquine Dose and Hospitalizations for Active SLE Among 122 Patients in a Case-Crossover Study

Hydroxychloroquine Dose	Hospitalizations/Case Periods (n)	Control Periods (n)	Unadjusted OR (95% CI)	Adjusted OR* (95% CI)
Overall	137	266	--	--
Weight-Based Dose
High Dose, > 5mg/kg/day	39	91	1.0	1.0
Low Dose, ≤ 5mg/kg/day	98	175	4.68 (1.66–13.20)	4.20 (1.45–12.19)
4–5mg/kg/day	44	92	3.43 (0.94–12.42)	3.41 (0.93–12.58)
< 4 mg/kg/day	54	83	5.27 (1.79–15.56)	4.62 (1.50–14.20)
Non-Weight-Based Dose
High Dose, 400mg/day	64	155	1.0	1.0
Low Dose, < 400mg/day	73	111	3.78 (1.51–9.51)	3.39 (1.31–8.81)
300 to < 400mg/day	26	61	1.81 (0.57–5.76)	1.74 (0.53–5.69)
< 300 mg/day	47	50	5.28 (1.94–14.36)	4.68 (1.66–13.17)

^*Analyses were completed using conditional logistic regression matched within a patient and adjusted for clinical Systemic Lupus Erythematosus Disease Activity Index (SLEDAI), glucocorticoid use, SLE immunosuppressant use, and estimated glomerular filtration rate. These variables were assessed at the most recent rheumatology visit prior to each case or control period.

DISCUSSION

In this case-crossover study within a SLE cohort spanning two academic medical centers and multiple community hospitals, we found that the use of lower doses of HCQ was associated with an increased risk of hospitalization for SLE compared with higher dosing. This finding was observed with lower HCQ dosing according to both weight-based and non-weight-based dosing. Our study spanned a recent 10-year period corresponding with ophthalmology recommendations to limit weight-based dosing of HCQ to reduce the risk of retinopathy. These results are concerning and are applicable to a large portion of the SLE population, who may receive lower HCQ dosing in accordance with guidelines aimed at preventing retinopathy. Both the 2011 AAO guidelines based on ideal body weight-based dosing and the 2016 guidelines based on the lower 5 mg/kg/day cutoff for actual body weight-based dosing effectively lead to a recommendation of HCQ dosing under 400 mg/day for the majority of patients with SLE (8, 9). These guidelines, while clearly of good intent of minimizing retinal toxicity, need to be balanced with the numerous benefits that HCQ can provide patients with SLE, including preventing SLE flares and hospitalizations.

The need for data on the efficacy of hydroxychloroquine dosing regimens to inform appropriate use in lupus care has been widely recognized (15). A recent study found lower HCQ dosing ≤5 mg/kg/day, in accordance with current guidelines, is associated with a higher risk of SLE flares (12). Another study found an increased risk of SLE hospitalizations and flares associated with HCQ dose-reduction, irrespective of dose (13). Ours is the first study to examine the impact of HCQ dosing in accordance with the current recommended ≤5 mg/kg/day cutoff on hospitalizations for active SLE and can therefore inform risk/benefit evaluations. An important consideration is the timing of HCQ retinal toxicity, which must be balanced with acute lupus manifestations. The current low dose HCQ guidelines prioritize the possibility of retinal toxicity in 10 to 20 years, whereas severe lupus complications, such as lupus nephritis, can occur early in the disease course. Therefore, the risk/benefit trade-offs regarding HCQ dose may vary over time and according to specific patient characteristics. Larger studies would be needed to evaluate the impact of HCQ dose on the risk of worsening of certain domains of SLE, such as nephritis or cutaneous disease. Long-term studies could also evaluate whether lower HCQ dosing leads to greater use of immunosuppressive agents, such as glucocorticoids, mycophenolate, or biologic immunosuppressants, increasing the risk of infectious complications.

Our study does have limitations. As this was an observational study, there is a possibility of misclassification of medication exposures and outcomes. We may have missed hospitalizations that were not captured in our health system. Although the case-crossover design has many benefits, including the ability to compare patients to themselves, bias can be introduced in the analysis of chronic exposures if there are temporal trends of the exposure and the timing of case periods. However, HCQ dose can increase or decrease over time, and we captured control periods at multiple time points before and after the case periods, so such temporal associations are unlikely to explain our findings. While lower dose HCQ may have been used during the study period due to the change in ophthalmology guidelines, it is also possible that other reasons influenced HCQ dose, including minimizing side effects such as gastrointestinal intolerance. Additionally, we assessed the association with the HCQ dose the patient was reported to be taking during a particular period. We did not assess medication adherence, as information on medication possession ratios and HCQ blood levels, which could serve as a measure of both adherence and absorption, were not available. These exposures could serve as the basis of future studies. Our study included a relatively small sample size, as our experimental design excluded hospitalizations where patients had not been previously taking HCQ and hospitalizations for etiologies other than SLE activity, such as infection.

In summary, this study suggests that lower HCQ dosing is associated with a higher risk of hospitalizations for active SLE than with higher dosing. These findings are juxtaposed with the established risk of retinal toxicity with the potential for vision loss that is associated with long-term HCQ use and which has been found to increase with higher weight-based dosing. (5–7) Therefore, these findings highlight the need to balance the risk of retinal toxicity in the long term with the prevention of SLE complications in the short term when choosing the optimal HCQ dose.