Abstract
Objective
Hydroxychloroquine (HCQ) has been associated with improved survival among SLE patients from tertiary referral centers. We aimed to determine the potential impact of HCQ use on the risk of mortality among SLE patients in the general population.
Methods
We conducted a nested case-control study within an incident SLE cohort from the entire population of British Columbia, Canada. Deceased cases were matched with up to three living controls by age, sex, and SLE disease duration. HCQ exposure was categorized by the time between the last HCQ prescription date covered (i.e., end of supply) and the index date (i.e., death date) as current (<30 days), recent (30–365 days), remote (>365 days), or never used. We used conditional logistic regression to assess the risk of all-cause mortality associated with current or recent HCQ exposure compared with remote HCQ users.
Results
Among 6,241 patients with incident SLE, we identified 290 deceased cases and 502 matched SLE controls. Adjusted odd ratios (ORs) for all-cause mortality were 0.50 (95% CI: 0.30–0.82) for current users and 2.47 (95% CI: 1.21–5.05) for recent users compared with remote users. Associations were similar in subgroups according to SLE duration (≤ 5 years vs > 5 years).
Conclusion
Our general population data support a substantial survival benefit associated with current HCQ use. Increased mortality among patients who had discontinued HCQ recently could be due to a sick stopper effect or the loss of actual HCQ benefits.
Keywords: systemic lupus erythematosus, mortality, hydroxychloroquine
INTRODUCTION
Hydroxychloroquine (HCQ) is near-universally recommended for patients with SLE. Its use has been associated with multiple benefits including reduced disease activity and damage, a lower risk of lupus nephritis, and lower risks of several comorbidities including hyperglycemia, hyperlipidemia, venous thromboembolism, and pregnancy complications.(1) HCQ has also been associated with a substantial overall survival benefit among SLE patients. Three prior cohort studies from tertiary referral/lupus expert centers compared HCQ users and non-users among individuals with SLE and demonstrated a 38–85% reduction in overall mortality associated with HCQ use.(2–4) However, this benefit has not been previously demonstrated in a general population context. Furthermore, patients who are never prescribed HCQ may have systematic differences from active users of this medication. We aimed to investigate the potential survival impact of HCQ use at the general population level in a cohort of incident SLE patients, using remote users as the comparison group to reduce potential confounding by indication.(5, 6)
METHODS
Data Source, Study Population, and Study Design
We conducted a nested case-control study within an incident SLE cohort. The source population was identified using linked administrative health databases from Population Data BC, which cover the entire population of the province of British Columbia, Canada.(7) These databases capture demographics, vital statistics, and healthcare utilization data since 1990 including all provincially-funded outpatient medical visits and hospital admissions and discharges. They also capture medications through the comprehensive prescription database PharmaNet, which includes all outpatient dispensed medications for all residents of British Columbia since 1996. Each PharmaNet record contains information on the medication and dose dispensed (via the Drug Information Number (DIN)), dispensing date, and quantity and days’ supply dispensed. These databases have been used previously to conduct population-based assessments of mortality in other inflammatory rheumatic conditions.(8)
An inception cohort of SLE patients has been previously reported.(9) This cohort includes 6,241 patients at least 18 years of age with SLE diagnosed between 1997–2015. Subjects were classified as having SLE if they met the following criteria: ≥1 International Classification of Diseases, Ninth Revision (ICD-9) or Tenth Revision, Clinical Modification (ICD-10-CM) code for SLE by a rheumatologist or from a hospital encounter (710.0 or ICD-10 M32.1, M32.8, and M32.9) or at least two ICD-9 codes for SLE (710.0) at least two months apart within two years by a non-rheumatologist physician. We excluded individuals with diagnoses of other inflammatory rheumatic diseases (i.e., rheumatoid arthritis, psoriatic arthritis, and ankylosing spondylitis) occurring in at least two physician visits at least two months apart after the first SLE diagnosis. To ensure incident cases, all individuals were required to have no SLE diagnosis recorded for at least seven years prior to the index date (i.e., from January 1990, the earliest data available).(9)
From this incident SLE cohort, we conducted a nested case-control study by identifying patients who died (cases) and up to three living controls selected from risk set samples by age, sex, and SLE disease duration at the index date (e.g., death date).
Assessment of Exposure
The exposure of interest was HCQ use status relative to the index date (death date). Using the PharmaNet dispensing database, we determined the final prescription date covered by the last HCQ prescription to determine the end of the medication supply and categorized HCQ use as current, recent, remote, or never. For current users, their HCQ supply ended within 30 days prior to the index date or they had an active supply of HCQ spanning the index date. For recent users, their HCQ supply ended between 30–365 days prior to the index date, whereas for remote users, their HCQ supply ended greater than 365 days prior to the index date. Never users had no HCQ prescriptions dispensed during the study period. This classification was designed to minimize exposure misclassification due to delayed prescription refills.
Assessment of Covariates
Covariates were assessed during the year prior to the underlying inception cohort entry and included chronic kidney disease (CKD), the Charlson comorbidity index,(10) glucocorticoid use, cardiovascular medication use (including statins, anti-hypertensives, cardiac glycosides, diuretics, anti-arrhythmics, nitrates, and anti-coagulants), other immunosuppressive drug use (including azathioprine, methotrexate, mycophenolate, leflunomide, cyclosporine, cyclophosphamide), financial assistance for public health insurance as a surrogate for socioeconomic status, and healthcare utilization (including hospitalizations and outpatient visits).
Statistical Analysis
We used conditional logistic regression to calculate the odds ratios (ORs) for all-cause mortality associated with current or recent HCQ use compared with remote HCQ users. Remote users were chosen as the reference group over non-users to minimize confounding by indication, as in previous studies.(5, 6) To evaluate the potential impact of SLE duration and other patient characteristics on our findings, we conducted a subgroup analysis by the disease duration (≤ 5 years vs > 5 years), age (> 60 years or ≤ 60 years), female sex, and CKD. Statistical analyses were performed using SAS (version 9.3, SAS Institute Inc); all p-values were two-sided (α=0.05).
All procedures were conducted in compliance with British Columbia’s Freedom of Information and Privacy Protection Act. Ethics approval was obtained from the University of British Columbia’s Behavioral Research Ethics Board.
RESULTS
Among 6,241 patients with incident SLE between 1997 and 2015, we identified 290 deceased cases and 502 matched living controls. Both cases and controls were predominately female (88% and 91% of cases and controls, respectively), with a mean age of 60 and 59 years, respectively. The mean SLE disease duration on the index date (matching time point) was 5.3 years. As expected, the cases had a higher mean Charlson comorbidity index score and higher rates of cardiovascular medication usage than controls at cohort entry. They also had higher rates of immunosuppressive drug and glucocorticoid usage. Cases had more frequent hospitalizations and outpatient visits than the controls and were more likely to have financial assistance for public health insurance (Table 1).
Table 1.
Characteristics† of Deceased Cases and Living Controls with Systemic Lupus Erythematosus
| Variable | Cases (n=290) | Controls (n=502) | P-Value |
|---|---|---|---|
| Age (mean, SD) | 60 (13) | 59 (13) | 0.31 |
| Sex (% female) | 88 | 91 | 0.11 |
| SLE Disease Duration (years, mean, SD)* | 5.3 (3.9) | 5.3 (3.9) | 0.85 |
| Chronic Kidney Disease (%) | 37 | 21 | <.001 |
| Charlson comorbidity index, mean (SD) | 2.12 (2.03) | 1.32 (1.06) | <.001 |
| Medications (%) | |||
| Glucocorticoids | 60 | 41 | <.001 |
| Cardiovascular medications | 57 | 41 | <.001 |
| Immunosuppressives | 22 | 16 | 0.038 |
| Financial Assistance (%) | 47.2 | 30.7 | <.001 |
| Healthcare Utilization | |||
| Number of hospitalizations, mean (SD) | 3.0 (2.5) | 0.5 (1.1) | <.001 |
| Number of outpatient visits, mean (SD) | 63.2 (39.8) | 27.4 (23.2) | <.001 |
Immunosuppressives, including azathioprine, methotrexate, mycophenolate, leflunomide, cyclosporine, cyclophosphamide.
Cardiovascular medications include anti-hypertensives, cardiac glycosides, diuretics, anti-arrhythmics, nitrates, and anticoagulants.
At the cohort entry date unless otherwise specified.
At the index date.
Current HCQ use was associated with an unadjusted odds ratio (OR) of 0.55 (95% CI: 0.36–0.85) for all-cause mortality compared with remote users (Table 2). The fully-adjusted OR was 0.50 (95% CI: 0.30–0.82) for current users. Recent HCQ use was associated with an unadjusted OR of 2.00 (95% CI: 1.08–3.69) and a fully-adjusted OR of 2.32 (95% CI: 1.13–4.77) compared with remote users. HCQ non-users had similar risk of death as remote users (fully-adjusted OR 0.80, 95% CI: 0.53–1.20]).
Table 2.
All-Cause Mortality According to Hydroxychloroquine Exposure
| All-Cause Mortality | Cases, N | Controls, N | Crude Odds Ratio (95% CI) | Adjusted* Odds Ratio (95% CI) |
|---|---|---|---|---|
| All Patients | ||||
| Remote HCQ Users | 72 | 106 | 1.0 (reference) | 1.0 (reference) |
| Recent HCQ users | 37 | 27 | 2.00 (1.08–3.69) | 2.47 (1.21–5.05) |
| Current HCQ users | 60 | 163 | 0.55 (0.36–0.85) | 0.50 (0.30–0.82) |
| HCQ non-users | 121 | 206 | 0.80 (0.53–1.20) | 0.80 (0.50–1.27) |
| Age > 60 Years | ||||
| Remote HCQ Users | 45 | 68 | 1.0 (reference) | 1.0 (reference) |
| Recent HCQ users | 26 | 19 | 2.33 (1.11–4.92) | 2.56 (1.08–6.05) |
| Current HCQ users | 32 | 97 | 0.51 (0.29–0.88) | 0.45 (0.24–0.85) |
| HCQ non-users | 92 | 141 | 0.94 (0.58–1.53) | 0.86 (0.51–1.47) |
| Age ≤ 60 Years | ||||
| Remote HCQ Users | 27 | 38 | 1.0 (reference) | 1.0 (reference) |
| Recent HCQ users | 11 | 8 | 1.80 (0.65–4.95) | 3.10 (0.69–13.94) |
| Current HCQ users | 28 | 66 | 0.62 (0.32–1.19) | 0.52 (0.18–1.48) |
| HCQ non-users | 29 | 65 | 0.62 (0.30–1.24) | 0.61 (0.21–1.78) |
| Female | ||||
| Remote HCQ Users | 65 | 102 | 1.0 (reference) | 1.0 (reference) |
| Recent HCQ users | 32 | 23 | 2.36 (1.25–4.46) | 2.75 (1.30–5.84) |
| Current HCQ users | 56 | 150 | 0.59 (0.38–0.91) | 0.53 (0.32–0.90) |
| HCQ non-users | 102 | 184 | 0.87 (0.57–1.31) | 0.86 (0.53–1.39) |
| SLE Duration ≤ 5 yrs | ||||
| Remote HCQ Users | 31 | 42 | 1.0 (reference) | 1.0 (reference) |
| Recent HCQ users | 22 | 19 | 1.68 (0.76–3.73) | 2.05 (0.76–5.52) |
| Current HCQ users | 36 | 96 | 0.51 (0.28–0.92) | 0.45 (0.21–0.97) |
| HCQ non-users | 64 | 105 | 0.80 (0.43–1.47) | 0.69 (0.32–1.51) |
| SLE Duration > 5 yrs | ||||
| Remote HCQ Users | 41 | 64 | 1.0 (reference) | 1.0 (reference) |
| Recent HCQ users | 15 | 8 | 2.91 (1.10–7.72) | 2.60 (0.83–8.15) |
| Current HCQ users | 24 | 67 | 0.57 (0.30–1.09) | 0.45 (0.20–1.01) |
| HCQ non-users | 57 | 101 | 0.89 (0.52–1.52) | 0.83 (0.44–1.54) |
| CKD | ||||
| Remote HCQ Users | 23 | 24 | 1.0 (reference) | 1.0 (reference) |
| Recent HCQ users | 12 | <6 | N/A** | N/A** |
| Current HCQ users | 24 | 39 | 0.71 (0.21–2.45) | 0.88 (0.10–8.12) |
| HCQ non-users | 48 | 39 | 1.10 (0.29–4.14) | 0.61 (0.07–5.07) |
| No CKD | ||||
| Remote HCQ Users | 49 | 82 | 1.0 (reference) | 1.0 (reference) |
| Recent HCQ users | 25 | 26 | 2.24 (1.04–4.83) | 2.23 (0.87–5.69) |
| Current HCQ users | 36 | 124 | 0.66 (0.38–1.16) | 0.52 (0.26–1.07) |
| HCQ non-users | 73 | 167 | 0.78 (0.46–1.34) | 0.85 (0.46–1.58) |
CKD, chronic kidney disease
Adjusted for Charlson comorbidity index, glucocorticoid use, disease-modifying anti-rheumatic drug use, cardiovascular medication use, healthcare utilization, and financial assistance.
Odds ratio not available due to small numbers.
Among the subgroups according to SLE disease duration (≤ 5 years versus > 5 years), female sex, age, and patients without CKD, the ORs associated with current HCQ use remained very similar (Table 2). Current HCQ users with CKD also had numerically lower odds of death relative to remote users.
DISCUSSION
In this study nested in a general population-based incident SLE cohort, we found a substantial survival benefit associated with current HCQ use compared with past use. These findings were similar among patients with shorter disease duration and those with longstanding SLE as well as younger and older patients alike. These findings support the generalizability of previous study findings generated from tertiary or lupus specialist centers.(2–4) Furthermore, by using the remote HCQ users as the reference unlike previous studies,(2–4) our study population consisted of those who were started on the medication, helping to minimize confounding by indication.
We also found a two-fold increased risk of death associated with recent HCQ discontinuation compared with remote discontinuation. We speculate that patients may have been less likely to adhere to taking or be prescribed this chronic medication near the end of life, due to potentially unrelated illness.(11) Although numbers were too small to conduct a meaningful cause of death analysis, we did note a higher number of cancer deaths among recent users than the other HCQ exposure categories (8 cancer deaths among recent users, 6 among current users and 6 among remote users, with an unadjusted cancer-specific mortality OR of 5.04 [95% CI 0.91–27.78] compared with remote users and 0.53 [95% CI 0.15–1.88] for current users). An alternative possibility is that this observed increase in mortality could be partially explained by a true biologic protective effect of HCQ that is lost and even reversed in the short-term following discontinuation, such as its impact on endothelial function(12, 13) and/or platelet aggregation.(14) However, further studies are needed to understand the short-term impact of HCQ discontinuation on mortality, cardiovascular disease, and other important SLE-related outcomes.
This study has several strengths and limitations worth noting. The main limitations are those inherent with the use of administrative data. The SLE diagnoses were not clinically confirmed, but patients were identified using a strict case definition that employed ICD codes and additional exclusion criteria. This SLE definition has been previously validated with 98% sensitivity and 72% specificity.(15) We did not have data on SLE disease activity. However, we adjusted for several indicators of illness including comorbidities, medication use, and health resource use. A major strength of our study is the use of a comprehensive prescription drug database, which captures all dispensed outpatient medications and the timing of refills regardless of age or funding. Our ascertainment of HCQ exposure status by actual prescription refills was less susceptible to misclassification of non-adherent patients as HCQ users than could occur with reliance on prescribing data alone. Furthermore, our population-based data source adds to the generalizability of our findings. Additionally, as mentioned, our remote user design, which has been employed in prior pharmacoepidemiology studies,(5) reduced the potential for confounding by indication for HCQ use. Although we demonstrated a survival benefit among patients with over five years’ disease duration, future studies could further evaluate the impact of HCQ on late mortality in patients with SLE.
In conclusion, our findings confirm the survival benefit of ongoing HCQ use among patients with SLE in a general population context. Future studies should further explore the potential link between HCQ discontinuation and increased mortality as well as its impact on other SLE related outcomes.
SIGNIFICANCE AND INNOVATIONS.
We investigated the potential survival benefit associated with hydroxychloroquine use at the general population level.
We used remote users as the comparison group rather than patients who had never used hydroxychloroquine to reduce potential confounding by indication.
We found a substantial survival benefit associated with current hydroxychloroquine use and increased mortality associated with recent hydroxychloroquine discontinuation.
ACKNOWLEDGEMENTS AND AFFILIATIONS
All inferences, opinions, and conclusions drawn in this manuscript are those of the authors, and do not reflect the opinions or policies of the Data Steward(s).
Funding:
This work was supported by the Canadian Institutes of Health Research [grant number THC-135235]; the National Institute for Arthritis and Musculoskeletal Diseases at the National Institutes of Health [grant numbers T32-AR-007258, P50-AR-060772]; the Rheumatology Research Foundation Scientist Development Award; the BC Lupus Society; the Michael Smith Foundation; and Walter & Marilyn Booth Research Scholar.
Financial Disclosure Statement:
The authors have not received any financial support from commercial sources for the work reported on in this manuscript. The authors have no other financial conflicts of interest associated with the work reported on in this manuscript.
REFERENCES
- 1.Wallace DJ, Gudsoorkar VS, Weisman MH, Venuturupalli SR. New insights into mechanisms of therapeutic effects of antimalarial agents in SLE. Nat Rev Rheumatol. 2012;8(9):522–33. [DOI] [PubMed] [Google Scholar]
- 2.Alarcon GS, McGwin G, Bertoli AM, Fessler BJ, Calvo-Alen J, Bastian HM, et al. Effect of hydroxychloroquine on the survival of patients with systemic lupus erythematosus: data from LUMINA, a multiethnic US cohort (LUMINA L). Ann Rheum Dis. 2007;66(9):1168–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ruiz-Irastorza G, Egurbide MV, Pijoan JI, Garmendia M, Villar I, Martinez-Berriotxoa A, et al. Effect of antimalarials on thrombosis and survival in patients with systemic lupus erythematosus. Lupus. 2006;15(9):577–83. [DOI] [PubMed] [Google Scholar]
- 4.Shinjo SK, Bonfa E, Wojdyla D, Borba EF, Ramirez LA, Scherbarth HR, et al. Antimalarial treatment may have a time-dependent effect on lupus survival: data from a multinational Latin American inception cohort. Arthritis Rheum. 2010;62(3):855–62. [DOI] [PubMed] [Google Scholar]
- 5.Graham DJ, Campen D, Hui R, Spence M, Cheetham C, Levy G, et al. Risk of acute myocardial infarction and sudden cardiac death in patients treated with cyclo-oxygenase 2 selective and non-selective non-steroidal anti-inflammatory drugs: nested case-control study. Lancet. 2005;365(9458):475–81. [DOI] [PubMed] [Google Scholar]
- 6.Dubreuil M, Louie-Gao Q, Peloquin CE, Choi HK, Zhang Y, Neogi T. Risk of myocardial infarction with use of selected non-steroidal anti-inflammatory drugs in patients with spondyloarthritis and osteoarthritis. Ann Rheum Dis. 2018;77(8):1137–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Population Data BC [Available from: http://www.popdata.bc.ca/data.
- 8.Lacaille D, Avina-Zubieta JA, Sayre EC, Abrahamowicz M. Improvement in 5-year mortality in incident rheumatoid arthritis compared with the general population-closing the mortality gap. Ann Rheum Dis. 2017;76(6):1057–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Avina-Zubieta JA, To F, Vostretsova K, De Vera M, Sayre EC, Esdaile JM. Risk of Myocardial Infarction and Stroke in Newly Diagnosed Systemic Lupus Erythematosus: A General Population-Based Study. Arthritis Care Res (Hoboken). 2017;69(6):849–56. [DOI] [PubMed] [Google Scholar]
- 10.Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. Journal of clinical epidemiology. 1993;46(10):1075–9; discussion 81–90. [DOI] [PubMed] [Google Scholar]
- 11.Glynn RJ, Knight EL, Levin R, Avorn J. Paradoxical relations of drug treatment with mortality in older persons. Epidemiology. 2001;12(6):682–9. [DOI] [PubMed] [Google Scholar]
- 12.Miranda S, Billoir P, Damian L, Thiebaut PA, Schapman D, Le Besnerais M, et al. Hydroxychloroquine reverses the prothrombotic state in a mouse model of antiphospholipid syndrome: Role of reduced inflammation and endothelial dysfunction. PloS one. 2019;14(3):e0212614. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Gomez-Guzman M, Jimenez R, Romero M, Sanchez M, Zarzuelo MJ, Gomez-Morales M, et al. Chronic hydroxychloroquine improves endothelial dysfunction and protects kidney in a mouse model of systemic lupus erythematosus. Hypertension (Dallas, Tex : 1979). 2014;64(2):330–7. [DOI] [PubMed] [Google Scholar]
- 14.Jancinova V, Nosal R, Petrikova M. On the inhibitory effect of chloroquine on blood platelet aggregation. Thrombosis research. 1994;74(5):495–504. [DOI] [PubMed] [Google Scholar]
- 15.Bernatsky S, Linehan T, Hanly JG. The accuracy of administrative data diagnoses of systemic autoimmune rheumatic diseases. J Rheumatol. 2011;38(8):1612–6. [DOI] [PubMed] [Google Scholar]