Pharmacoepidemiological aspects of the COVID-19 pandemic

The emergence of the severe acute respiratory syndrome corona virus 2 (SARS-CoV-2) that causes corona virus disease 2019 (COVID-19) has been the most urgent threat to public health in recent times, with more than 461 million confirmed cases worldwide and 6 million related deaths. The novelty of the disease, and a resulting lack of knowledge on interactions between COVID-19 and pharmacological treatments, led to a surge of new hypotheses being formed regarding beneficial and detrimental drug effects. In this PhD-project, we aimed to investigate such hypotheses using population-based cohort studies.
The first study was prompted by reports of several young individuals being admitted to the intensive care unit of the university hospital of Toulouse, France for COVID-19. Common to all of them was exposure to ibuprofen during the early phase of a SARS-CoV-2 infection. Therefore, we conducted a population-based cohort study among all Danish residents with documented SARS-CoV-2 infection between 27 February 2020 and 29 April 2020. We compared the occurrence of hospitalisation, intensive care unit admission and death following SARS-CoV-2 infection among 248 users and 8988 non-users of NSAIDs and found no increased risk of death, hospitalisation, intensive care unit admission or receiving mechanical ventilation 30 days after testing positive for SARS-CoV-2.
Thyroid hormone levels influence the tissue concentration of the angiotensin converting enzyme 2 receptor, which SARS-CoV-2 uses to enter lung-epithelial cells. Therefore, in- or decreased levels of thyroid hormones could influence the risk of SARS-CoV-2 infection and subsequent clinical outcomes. Using a test-negative case-control design, we found that individuals exposed to levothyroxine or antithyroid drugs were not overrepresented among 28,078 individuals infected with SARS-CoV-2 during the period 27 February and 30 September 2020 compared to individuals not treated for thyroid disorders. We subsequently followed all infected individuals for 30 days and found no increased risk of adverse outcomes, except for hospitalisation (risk ratio 1.19, 95% confidence interval 1.02 to 1.40), when comparing users of levothyroxine or antithyroid therapy to untreated individuals.

Increased media attention on the occurrence of complications to SARS-CoV-2 infection weeks to months after the primary infection prompted us to investigate the occurrence of such events. We included all individuals who tested positive for SARS-CoV-2 on or prior to 31 May 2020 and did not require hospitalisation during the following two weeks. We compared the occurrence of a range of delayed acute complications, chronic disease and persisting symptoms among SARS-CoV-2 positive individuals to a test-negative reference cohort six months after testing. We found no increased risk of acute delayed complications, except for a slightly increased risk of venous thromboembolism (risk difference +0.1%, 0.0 to 0.2). For non-acute outcome, we found an increased risk of initiating short-acting beta-2 agonists (+0.4%, 0.1 to 0.7), triptans (+0.1%, 0.0 to 0.3) and a hospital diagnosis of dyspnoea (+0.6%, 0.4 to 0.8). Further, SARS-CoV-2 positive individuals had higher rates of healthcare utilisation compared to individuals who tested negative (adjusted rate ratio 1.18, 1.15 to 1.22). Still, the absolute risk of post-acute complications following SARS-CoV-2 infection not requiring hospital admission was low.

Study IV was prompted by reports of a high risk of venous thromboembolism among patients hospitalised for COVID-19, and subsequent publication of guidelines recommending prophylactic anticoagulation with low-molecular weight heparin for all adults hospitalised for COVID-19. Using electronic health records from the Capital region of Denmark and Karolinska Institutet in Sweden, we followed all individuals hospitalised for COVID-19 and compared the occurrence of death, venous thromboembolism, and bleeding episodes after 30 days between individuals who received prophylactic anticoagulation within 48 hours of admission to those who did not. We found no beneficial effect of prophylactic anticoagulation on mortality, and no increased risk of bleeding. The risk ratio for venous thromboembolism was 0.80 (0.43 to 1.47).

The fifth and final study evaluated methods to account for calendar time in observational studies regarding COVID-19 vaccine effectiveness and safety. We simulated key parameters of the COVID-19 pandemic (vaccine uptake, SARS-CoV-2 incidence) and evaluated an effectiveness outcome (SARS-CoV-2 infection) and a safety outcome with seasonal variation using five different estimators: unadjusted and calendar-time adjusted Poisson regression, unadjusted and adjusted Cox regression with time under observation as the underlying time scale, and Cox regression with a calendar time scale. We found that adjusted Poisson regression and Cox regression with a calendar time scale yielded unbiased estimates, while all other estimators were biased due to temporal trends in the exposure or outcome. When analysing empirical data from the year of 2021 we obtained estimates compatible with the simulation and prior evidence.

Overall, we found null results in studies I through IV. We found no increased risk of adverse outcomes of SARS-CoV-2 infection with specific pharmacological exposures (study I and II). The occurrence of post-acute complication following SARS-CoV-2 infection not requiring hospital admission was lower than expected from media reports. Likewise, we found no beneficial effect of prophylactic anticoagulation on COVID-19 related mortality. Finally, we found that properly accounting for calendar time in observational analyses of COVID-19 vaccine effectiveness and safety is crucial to obtaining unbiased results. The COVID-19 pandemic has generated additional hypotheses, which also were dismissed when evaluated with scientific rigor. Such results reinforce, that patients should not initiate or discontinue treatments based on weak evidence. Further, the studies in this thesis demonstrate the utility of frequently updated population-based registries for public health decision making.