HIV and COVID-19: Intersecting Epidemics With Many Unknowns

Catherine R. Lesko; Angela M. Bengtson


Am J Epidemiol. 2021;190(1):10-16. 

In This Article

HIV as a "Risk Factor"

Useful epidemiologic investigations into the impact of COVID-19 on PLWH will need to carefully consider the research question of interest and how results will be used. There is justified concern about labeling HIV as an "independent risk factor" for poor COVID-19 outcomes based on an arbitrary multivariable model given that it might then be inappropriately used to ration care or guide treatment decisions. Ambiguity about the meaning of the term "independent risk factor" makes it highly likely that results will be misinterpreted and misapplied.[20,21] If interest is in identifying groups that should be monitored more closely for SARS-CoV-2 infection, this is a descriptive epidemiology question, and crude analyses (or perhaps age- and sex-adjusted analyses) might be sufficient.[22] While the most appropriate adjustment set for descriptive epidemiology is an unresolved question, associations from a multivariable model are interpretable as hypothetical assumptions that would exist if we could "hold constant" the other covariates in the model and thus are necessarily not descriptive of the world as it exists.[22] If interest is in estimating the etiological influence of the HIV virus and associated immune activation/immune system suppression on the progression of COVID-19, other analyses are warranted.

In particular, for etiological questions, we need to be thoughtful about the role of comorbidities in our analyses. The prevalence of some comorbidities, such as alcohol and other drug use disorders, is higher among PLWH because they are causal factors for HIV acquisition and are therefore confounders in any etiological analysis. The prevalence of other comorbidities, such as depleted bone and kidney health, might be higher among PLWH because of the effects of the virus and antiretroviral therapy, and are mediators of any effect of HIV on poor COVID-19 outcomes; adjusting for these comorbidities would be inappropriate. The role of other comorbidities, such as cardiovascular disease and diabetes, is more complex. For example, tobacco use increases the risk of cardiovascular disease that might precede HIV infection (implying cardiovascular disease is a confounder), but uncontrolled viremia and some antiretroviral medications themselves increase the risk of cardiovascular disease (implying cardiovascular disease is also a mediator).

Existing studies examining the relationship between HIV and COVID-19 outcomes have not always been clear about their research question. In a retrospective matched cohort of PLWH and people without HIV hospitalized for COVID-19 in New York, outcomes were similar. Matching factors included some confounders of the effect of HIV, such as admission date, age, gender, and tobacco history, but also included variables that might be considered mediators, such as body mass index and history of chronic kidney disease, hypertension, asthma, chronic obstructive pulmonary disease, and heart failure.[23] In another matched cohort in New York, outcomes of people with and without HIV hospitalized for COVID-19 were similar even without adjusting for higher prevalence of chronic obstructive pulmonary disease, prior cancer, cirrhosis, and current smoking among PLWH.[24]

Markers of HIV disease that might be expected to be the strongest mediators of a direct effect of HIV infection on COVID-19 outcomes, such as HIV viral load and CD4 cell count, have not been strongly associated with COVID-19 morbidity among PLWH.[15,17,24] However, these data should be interpreted cautiously because, as yet, only SARS-CoV-2 infections that resulted in symptomatic disease have been studied; HIV might influence whether SARS-CoV-2 infections are detected, either because PLWH might have more or less access to screening or because HIV might increase the proportion of infections that are symptomatic. Furthermore, HIV-infected patients without updated clinical data (presumably because they are out of care) have generally been excluded for having missing data; the prevalence of HIV viral suppression in identified COVID-19 cases has been high, which might indicate that PLWH who are not well-linked to HIV care are less likely to be diagnosed with COVID-19 or identified as HIV-infected in the data.