key: cord-0694810-iw5yry8t
authors: Kim, Alfred H J; Eisen, Seth A
title: The contribution of the observational research design to COVID-19 research
date: 2020-09-21
journal: Lancet Rheumatol
DOI: 10.1016/s2665-9913(20)30336-2
sha: 12f33801a0e8894b7ddc46f1d3593003d0a78152
doc_id: 694810
cord_uid: iw5yry8t

nan

The contribution of the observational research design to COVID-19 research

As the COVID-19 pandemic continues to influence global health, the search for effective therapies has been vigorous. An analysis published early during the pandemic suggested that hydroxychloroquine, with or without azithromycin, might improve nasopharyngeal viral clearance of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of COVID-19. 1 Despite the low quality of this study due to poor handling of confounders and participants lost to follow-up who had poor outcomes, a surge of prescriptions for the therapeutic and prophylactic use of hydroxychloroquine created shortages for patients with systemic lupus erythematosus and other rheumatic diseases who rely on this medication to treat their disease. Subsequently, an increased inci dence of cardiac arrhythmias was observed in patients with COVID-19 treated with hydroxychloroquine. Thus, there is a need to determine whether the benefits of hydroxychloroquine for COVID-19 outweigh the risk of harms. Randomised controlled trials (RCTs) are the gold standard by which the efficacy of an intervention is evaluated. Several RCTs have been published that largely demonstrated no benefit when hydroxy chloroquine was used for COVID-19. [2] [3] [4] [5] [6] An alternative research design, the retrospective observational study, can usually be carried out relatively quickly and at comparatively low cost, but is limited by numerous potential sources of bias. The appropriate scientific contribution of observational studies, it is argued, is to provide estimates of disease outcomes in real-world patient populations and to generate hypoth eses to support further research. Improving the validity of the results of observational studies requires construction of comprehensive models that identify all potential confounding and modulating variables that could link the intervention to the outcome. So far, validated measures of important variables are not routinely available.

In The Lancet Rheumatology, Chris Gentry and col leagues 7 leveraged a large dataset derived from the elec tronic health record of the US Veterans Health Administration (VHA) to carry out a retrospective, observational, propensity-matched analysis comparing the rate of laboratory-confirmed COVID-19 in adults with rheumatic diseases prescribed hydroxychloroquine with patients treated with other rheumatic disease medications. From 70 270 patients with an International Classification of Diseases (10th revision; ICD-10) diagnosis of rheumatic disease, the authors compared 10 703 patients treated with hydroxychloroquine who demonstrated satisfactory hydroxychloroquine adherence to 21 406 patients not treated with hydroxychloroquine. Commendably, the two groups were propensity matched for several dozen variables, including zip code of residence, comorbidities, laboratory test results, emergency or urgent care encounters, influenza titres, SARS-CoV-2 titres, date of death, medications, and hospital admissions.

Non-statistically significant differences in the proportion of SARS-CoV-2 infection were observed between those on hydroxy chloroquine compared with those not on hydroxy chloroquine (31 [0·3%] of 10 703 vs 78 [0·4%] of 21 406; odds ratio [OR] 0·79, 95% CI 0·52-1·20, p=0·27), and between those prescribed hydroxy chloroquine at more than 400 mg per day versus those prescribed 400 mg per day or less (13 [0·4%] of 2928 vs 18 [0·2%] of 7775, p=0·081). In multivariate logistic regression analyses, hydroxychloroquine use was not independently associated with SARS-CoV-2 infection risk. The results from this male-predominant population of veterans with rheumatic disease are similar to those reported by the COVID-19 Global Rheumatology Alliance 6, 7 and in the general population within the setting of an RCT. 5 Although the groups exposed and not exposed to hydroxychloroquine were extensively matched, concerns remain about inadequately measured or unmeasured confound ers that are not controlled by randomisation and might threaten study validity. One example is the potential for false positive rheumatological diagnoses. The authors' method for selecting patients with rheumatological dis ease was based exclusively on ICD-10 codes, and this might have resulted in inclusion of a substantial propor tion of patients with non-rheumatological dis orders (ie, false positives), the numbers of which might differ between hydroxychloroquine and control cohorts. There is also the possibility of bias by treatment indication. Patients with rheumatological diseases treated with hydroxychloroquine might differ from patients not treated with hydroxychloroquine in ways that affect SARS-CoV-2 infection risk. There might also be unmeasurable health behaviour bias. Notably, the hydroxy chloroquine cohort (but not the control cohort) was selected for good medication adherence, which might be associated with better general health behaviour. Finally, social determinants of health might differ between the groups. In addition to well recognised sociodemo graphic characteristics, there is increasing appreciation of the potential impact of variation in health service access; individual social norms, attitudes, and culture; residential segregation; food insecurity; attention to mass media; and social support on intervention outcomes. 8 In an effort to improve research using observational data, the US Veterans Affairs (VA) Office of Research and Development recently initiated a VA Phenomics Library (VAPheLib) with the objective of providing an encyclopedia of electronic health records phenomic definitions for use by VA investigators and clinical operations staff. 9 Once established, the phenomic library will greatly facilitate observational research using validated measures. To facilitate COVID-19-related research, the VA has implemented a standardised and validated definition for COVID-19 and is developing a COVID-19-specific Social Risks Screening questionnaire for inclusion in the electronic health record. 10 Over the next several years, the number of standardised and validated phenomic definitions will be expanded, patient reported outcome questionnaires will be routinely administered, and social determinants of disease data will be increasingly incorporated into the electronic health record. As a consequence, validity threats associated with the retrospective observational design might be reduced. These incremental improvements will permit investigators to carry out higher-quality observational studies that supplement (and in some cases replace) evidence from RCTs. AHJK reports personal fees from Exagen Diagnostics, GlaxoSmithKline, Aurinia Pharmaceuticals, and Annexon Biosciences, and is supported by the US National Institute of Arthritis and Musculoskeletal and Skin Disease (National Institutes of Health) and Rheumatology Research Foundation. SAE is supported by an investigator-initiated research grant from Bristol Myers Squibb.

seth.eisen@wustl.edu

Hydroxychloroquine and azithromycin as a treatment of COVID-19: results of an open-label non-randomized clinical trial

A randomized trial of hydroxychloroquine as postexposure prophylaxis for Covid-19

Observational study of hydroxychloroquine in hospitalized patients with Covid-19

Rheumatic disease and COVID-19: initial data from the COVID-19 Global Rheumatology Alliance provider registries

Baseline use of hydroxychloroquine in systemic lupus erythematosus does not preclude SARS-CoV-2 infection and severe COVID-19

Association of treatment with hydroxychloroquine or azithromycin with in-hospital mortality in patients with COVID-19 in New York State

Long-term hydroxychloroquine use in patients with rheumatic conditions and development of SARS-CoV-2 infection: a retrospective cohort study

US Office of Disease Prevention and Health Promotion

Introduction to VA Phenomics Library

HSR&D Collaborates to Develop VA COVID-19 Social Risks Screening Questions