Study population and setting
This cohort study included 17,282,905 adults (18 years or older) followed through the OpenSAFELY database (primary care data for 40% of English population) with at least one year of follow-up as of February 1, 2020 and no missing age, sex, or deprivation data (those missing n=142,785). The authors collected participants’ HIV status, age, sex, deprivation, ethnicity, smoking status, BMI, and other comorbidity data through OpenSAFELY. Records were linked to a national death registry and COVID-19 related deaths through June 22, 2020 were identified via ICD-10 codes (participants were censored at their time of death if they died from any cause other than COVID-19). They used Cox proportional hazards models to estimate the relationship between HIV and COVID-19 related deaths, adjusting for age, sex, an index of deprivation, ethnicity, smoking, and obesity. They also assessed the direct effect of HIV on COVID-19 death by adjusting for medical comorbidities that might mediate or confound the relationship and assessed for effect modification via stratification by age ( 59 years), sex, ethnicity (Black versus other), comorbidities (any versus none), and epidemic period (0-59, 60-89, or 90+ days to represent different national social distancing policies).
Summary of Main Findings
Of the 17,282,905 participants, 27,480 (0.2%) were living with HIV. During the study period, 25 people living with HIV (in 10,680 person-years) and 14,857 people without HIV (in 6,700,000 person-years) died from COVID-19 for a crude hazard ratio (HR) of 1.03 (95% CI: 0.70, 1.52). After adjustment for suspected confounders, people living with HIV were at a 2.59 times increased risk of death from COVID-19 as compared to those without HIV (95% CI: 1.74, 3.84), which was similar after further adjusting for potential comorbidity-induced mediation (HR 2.30, 95% CI: 1.55, 3.41). Among participants who self-identified as Black, living with HIV was associated with an even greater risk of death from COVID-19 (HR 4.31, 95% CI: 2.42, 7.65) than among participants who self-identified as not Black (HR 1.84, 95% CI: 1.03, 3.26) compared to Black and non-Black participants without HIV (interaction p-value: 0.044).
This study assessed the association between HIV and COVID-19 related death in a large study population and used a directed acyclic graph to select covariates that could have confounded or mediated the relationship.
The risk of COVID-19 death has two distinct components, each with a different potential causal mechanism linked to HIV infection: risk of SARS-CoV-2 infection, and risk of progression to death conditional on infection. This analysis, which only considered COVID-19 death, does not present sufficient data (such as on infection risk by HIV status) to allow for the disaggregation of these components, which limits interpretability. Underlying causal relationships with covariates may differ for each component (e.g., occupation might correlate with HIV status and alter risk of SARS-CoV-2 infection but not alter risk of mortality given infection). The authors do not provide sufficient detail to assess which 40% of the English population were included in the OpenSAFELY database. If the database systematically excluded regions with higher or lower HIV prevalence or COVID-19 related deaths, the analysis could either underestimate or overestimate the relationship between the two. The authors did not have access to viral load or CD4 count data among participants living with HIV, which could also mediate the relationship between HIV and COVID-19 related deaths. Finally, despite the large number of recorded person-years, there were a small number of deaths among people living with HIV, which makes it very difficult to assess how these findings may apply to people living with HIV elsewhere the United Kingdom or around the world.
This study suggests that people living with HIV may be at increased risk for COVID-19 related death.
This review was posted on: 19 February 2021