A. Jorge, R. Melles, M. Marmor, C. Conell, B. Zhou, J. Niu, Y. Zhang, H. Choi
2022 Annals of the Rheumatic Diseases  
BackgroundHydroxychloroquine (HCQ) is a key treatment for patients with lupus, but the major long-term toxicity is HCQ retinopathy. A large cross-sectional study found a prevalence of HCQ retinopathy of 7.5% overall and 5 times higher odds associated with HCQ dose >5 mg/kg/day, which led to ophthalmology guidelines recommending the avoidance of HCQ dosing >5 mg/kg.1 However, whether this dosing recommendation is applicable to the future risk of HCQ retinopathy is unknown, as is the incidence of
more » ... severe bullseye retinopathy vs. the mild, pre-symptomatic stage. Furthermore, recent studies have indicated that the use of lower doses of HCQ may increase the risk of lupus flares and hospitalizations.2ObjectivesTo determine the incidence of HCQ retinopathy associated with long-term HCQ use and compare them according to HCQ dose.MethodsWe emulated a hypothetical target trial using observational data3 from the US integrated health network Kaiser Permanente Northern California to compare two HCQ weight-based dosing strategies, >5 vs ≤5 mg/kg/day, based on dispensed tablets per year. A secondary analysis evaluated >80% of prescription days covered by dispensed tablets per year with HCQ dose >5 vs ≤5 mg/kg. We included 4,677 patients who initiated and continued HCQ for at least 5 years between 1997-2020. We emulated randomization of treatment strategy by cloning each subject and assigning a replicate to each treatment group.3 We censored replicates if and when they deviated from the assigned treatment group, assessed as the average dose in the first 5 years and annually thereafter. We used inverse probability weighting to account for censoring. The primary outcome was HCQ retinopathy, assessed by expert adjudication of spectral domain-optical coherence tomography (SD-OCT) and graded by severity. All SD-OCTs were prospectively reviewed by an expert ophthalmologist (RM), and a second expert ophthalmologist (MM) reviewed all abnormal scans and a random subset of normal scans. We assessed intra-rater reliability of SD-OCT findings. We used pooled logistic regression to estimate the cumulative incidence of HCQ retinopathy for each HCQ dose strategy from initiation. The odds ratios approximated hazard ratios (HRs) because the outcome at each one-year time block is <5%.ResultsAmong 4,677 patients, the mean age at HCQ initiation was 52 years; 83% were female. The racial/ethnic composition included 51% non-Hispanic White, 19% Hispanic, 14% Asian, and 11% Black patients. 756 (16.2%) and 3,921 (83.8%) patients initiated HCQ with the primary definition of the treatment strategies >5 and ≤5 mg/kg/day, respectively. The weighted kappa was 0.80 for SD-OCT reliability. 164 patients developed HCQ retinopathy (100 mild, 38 moderate, and 26 severe cases). The cumulative incidence of retinopathy over 18 years was 37.6% for >5 and 5.7% for ≤5mg/kg of HCQ in our primary analysis. The corresponding risk was 26.5% for >5 and 3.2% for ≤5 mg/kg in our secondary analysis using >80% of prescription days. Compared with ≤5 mg/kg of HCQ, the HRs of retinopathy were 9.65 (95% CI 5.73-16.65) and 10.79 (95% CI 6.40-20.07) for >5 mg/kg using the primary and secondary definitions of HCQ dose categories, respectively (Figure 1).ConclusionThe risk of HCQ retinopathy associated with long-term adherence to >5 mg/kg dosing was high, approximately 10 times that of ≤5 mg/kg dosing. However, most cases identified during the study were mild and pre-symptomatic, supporting the value of regular screening. These data should be incorporated into individualized decisions about long-term use of HCQ.References[1]Melles RB, Marmor MF. The risk of toxic retinopathy in patients on long-term hydroxychloroquine therapy. JAMA Ophthalmol 2014;132(12):1453-60[2]Almeida-Brasil CC et al. Flares after hydroxychloroquine reduction or discontinuation: results from the SLICC inception cohort. Ann Rheum Dis 2021 Dec 15, epub ahead of print[3]Hernán MA, Robins JM. Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available. Am J Epidemiol 2016;183(8):758-64Disclosure of InterestsApril Jorge: None declared, Ronald Melles: None declared, Michael Marmor: None declared, Carol Conell: None declared, Baijun Zhou: None declared, Jingbo Niu: None declared, Yuqing Zhang: None declared, Hyon Choi Consultant of: Ironwood, Selecta, Horizon, Takeda, Kowa, and Vaxart., Grant/research support from: Ironwood and Horizon
doi:10.1136/annrheumdis-2022-eular.1671 fatcat:xwnt42sysnhflllrlxpiqpumfa