Chapter 4: Examples of COVID-19 Real-World Evidence Studies

Cohort Study

Study citation: Rentsch CT, DeVito NJ, MacKenna B, Morton CE, Bhaskaran K, Brown JP, Schultze A, Hulme WJ, Croker R, Walker AJ, Williamson EJ, Bates C, Bacon S, Mehrkar A, Curtis HJ, Evans D, Wing K, Inglesby P, Mathur R, Drysdale H, Wong AYS, McDonald HI, Cockburn J, Forbes H, Parry J, Hester F, Harper S, Smeeth L, Douglas IJ, Dixon WG, Evans SJW, Tomlinson L, Goldacre B. Effect of pre-exposure use of hydroxychloroquine on COVID-19 mortality: a population-based cohort study in patients with rheumatoid arthritis or systemic lupus erythematosus using the OpenSAFELY platform. Lancet Rheumatol. 2021 Jan;3(1):e19-e27. doi: 10.1016/S2665-9913(20)30378-7. Epub 2020 Nov 5. PMID: 33349815; PMCID: PMC7745258.2

Study objective: This study sought to understand the potential effectiveness of hydroxychloroquine in preventing COVID-19 mortality in the general population. Early in the pandemic, there was interest in repurposing pharmaceuticals for the prevention of SARS-CoV-2 infections. Hydroxychloroquine was quickly thrown into a global spotlight after it garnered simultaneous praise and criticism from a number of high-profile sources. At the time of this study, most randomized clinical trials and observational studies showed no evidence of a benefit of hydroxychloroquine as a treatment for patients admitted to hospital who already have COVID-19. However, there remained uncertainty whether routine ongoing use of hydroxychloroquine in people without SARS-CoV-2 infection protected against new infections or severe COVID-19 outcomes.

PICO: This was an observational, population-based cohort study analyzing electronic health records for individuals over the age of 18 years who were established with a primary care provider in England, United Kingdom (UK). Patients who carried a diagnosis for which hydroxychloroquine was indicated (systemic lupus erythematosus or rheumatoid arthritis) and were on hydroxychloroquine therapy for at least 6 months were selected to be compared with those with systemic lupus erythematosus or rheumatoid arthritis but not treated with this medication. Death due to COVID-19 was studied as the primary endpoint, with non-COVID-19 mortality used as a negative control outcome.

Data source: The OpenSAFELY platform, a new, secure, transparent, open-source software platform for analysis of electronic health record data and linked SARS-CoV-2 testing, hospitalization, and death records for approximately 40% of the population in England.

Study period: March 1 and July 13, 2020

Key sources of error and how they were handled: Because the OpenSAFELY platform collects longitudinal data from routine clinical practice, it enabled an examination of previously approved indications and real-world use of hydroxychloroquine in COVID-19 with adjustment for confounding by indication. The authors developed a directed acyclic graph to inform the selection of potential confounders to include in multivariable regression. The data source had substantial inclusion of patient medication data but was limited by a lack of information on medications prescribed by specialists, such as biologic treatments for rheumatological disease. Therefore, the authors calculated bias-adjusted effect estimates using quantitative bias analysis to assess how adjustment for biologic treatments (as a proxy for disease severity) might have produced different results. Sensitivity analyses were conducted to assess the potential of exposure misclassification, including shortening the exposure ascertainment window from 6 to 3 months prior to baseline to be more sure patients were prescribed hydroxychloroquine closer to baseline. Multiple imputation was used to address missing ethnicity data for 23% of individuals. The authors pre-specified and published an open-source protocol prior to analyzing the data, which reduced the potential for subjectivity in decision-making during analysis. All code for data management and analyses was archived with version control on GitHub for review and re-use to promote reproducibility of research findings.