Risk of adverse coronavirus disease 2019 outcomes for people living with HIV

Objective: To assess whether people living with HIV (PLWH) are at increased risk of coronavirus disease 2019 (COVID-19) mortality or adverse outcomes, and whether antiretroviral therapy (ART) influences this risk. Design: Rapid review with meta-analysis and narrative synthesis. Methods: We searched databases including Embase, Medline, medRxiv and Google Scholar up to 26 August 2020 for studies describing COVID-19 outcomes in PLWH and conducted a meta-analysis of higher quality studies. Results: We identified 1908 studies and included 19 in the review. In a meta-analysis of five studies, PLWH had a higher risk of COVID-19 mortality [hazard ratio 1.95, 95% confidence interval (CI): 1.62–2.34] compared with people without HIV. Risk of death remained elevated for PLWH in a subgroup analysis of hospitalized cohorts (hazard ratio 1.60, 95% CI: 1.12–2.27) and studies of PLWH across all settings (hazard ratio 2.08, 95% CI: 1.69–2.56). Eight other studies assessed the association between HIV and COVID-19 outcomes, but provided inconclusive, lower quality evidence due to potential confounding and selection bias. There were insufficient data on the effect of CD4+ T-cell count and HIV viral load on COVID-19 outcomes. Eleven studies reported COVID-19 outcomes by ART-regimen. In the two largest studies, tenofovir disoproxil fumarate-based regimens were associated with a lower risk of adverse COVID-19 outcomes, although these analyses are susceptible to confounding by co-morbidities. Conclusion: Emerging evidence suggests a moderately increased risk of COVID-19 mortality among PLWH. Further investigation into the relationship between COVID-19 outcomes and CD4+ T-cell count, HIV viral load, ART and the use of tenofovir disoproxil fumarate is warranted.


Introduction
By September 2020, over 30 million people worldwide had been diagnosed with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) [1]. Although SARS-CoV-2 infection may be asymptomatic or cause only mild symptoms, a proportion of people develop severe coronavirus disease 2019 , leading to hospitalization, acute respiratory distress syndrome or death. Established risk factors for severe COVID-19 among the general population include older age [2], chronic kidney disease and obesity [3].
People living with HIV (PLWH), who constitute approximately 0.5% of the global population [4], may have an increased risk of adverse outcomes from COVID-19 as a result of HIV-associated immune dysfunction [5]. There may also be a higher prevalence of co-morbidities among PLWH that predispose to unfavourable COVID-19 outcomes [6]. Conversely, PLWH may have more favourable outcomes due to increased health awareness or close medical follow-up. Some antiretroviral agents are under consideration as potential treatments for COVID-19 [7], but the influence of antiretroviral therapy (ART) on COVID-19 outcomes is not known. In this rapid review, we aim to evaluate the evidence regarding the risk of adverse COVID-19 outcomes in PLWH, and the extent to which this risk is modified by other factors including ART.

Methods
We used rapid review methods, a simplified version of a systematic review to allow for timely publication [8], to identify studies between 1 January 2020 and 26 August 2020 that described COVID-19 outcomes in PLWH and compared outcomes with HIV-negative people or the general population, or that compared outcomes by risk factors among PLWH. We searched Embase, Medline, medRxiv, LitCovid, Trip, Google and Google Scholar without language restrictions. Search terms are available in Table S1, http://links.lww.com/QAD/C9. One author with extensive literature search expertise performed the initial screen to exclude duplicates and studies not related to HIV. For remaining articles, one author performed title and abstract screening, with subsequent full text review by two authors using a standardized data extraction form. In case of disagreement, inclusion decisions were made by a third author. We included preprints to capture emerging evidence. Studies with 15 or less participants were excluded as they were unlikely to be powered to detect meaningful associations. We critically appraised the quality of studies using checklists for Case Series and for Cohort Studies from the Joanna Briggs Institute [9].
Cohort studies reporting COVID-19-related death in people with and without HIV that adjusted for age, sex and co-morbidities were included in a meta-analysis. Cohort-specific relative risks (RRs) and hazard ratios were combined with random effects model to account for variability of the true effect between studies. Hazard ratios and RRs numerically approximate each other with shorter follow-up, rarer endpoints and risks closer to 1 [10]. Subgroup analyses were conducted by study setting and method of confounder adjustment. Meta-analysis was performed in R (version 3.6.0; R Foundation for Statistical Computing, Vienna, Austria) using the meta package [11].
There were 10 studies that included a total of more than 1000 individuals, but among these the median number of PLWH with COVID-19 co-infection was only 55 (interquartile range . Seventeen of the studies were performed in high-income countries and most included a majority of patients on ART with well controlled HIV ( Table 1).
Quality of evidence and risk of bias assessment There were common limitations among the included studies (Tables S2 and S3, http://links.lww.com/QAD/ C9). Most were retrospective analyses of routinely collected clinical data, meaning identification of COVID-19 cases was not systematic and depended on the local approach to screening and diagnosis. This has varied over time and between settings, and may also differ between PLWH and the general population. Only five studies directly compared COVID-19 outcomes among PLWH and HIV-negative people in the same cohort, and accounted adequately for potential confounding by comorbidities associated with adverse COVID-19 outcomes. Other studies used inadequately matched HIVnegative controls, or general populations in various settings, which is susceptible to bias as the exposed and control groups were selected differently [2]. Across all studies, the numbers of PLWH and COVID-19 infection were relatively low.

Adjusted analyses of HIV and risk of death in coronavirus disease 2019
In a meta-analysis of five cohort studies which accounted for confounding, the risk of death from COVID-19 for PLWH was almost double that of HIV-negative people [hazard ratio ¼ 1.95, 95% confidence interval (CI) 1.62-2.34] (Fig. 2) [12,18,[20][21][22]. Three of these studies used large routine databases to identify PLWH across community and hospital settings, in South Africa [20], the United Kingdom [22] and United States [12], and two studies were limited to hospitalized PLWH and COVID-19 in the United Kingdom [21] and United States [18]. In a subgroup analysis there was no significant difference between study settings (P ¼ 0.20), although a weaker hazard ratio was seen in hospitalized patients (Fig. S1, http://links.lww.com/QAD/C9). Among the three studies which used multi-variable adjustment to account for confounding [20][21][22], the crude risk of COVID-19 death was similar between people with and without HIV, but after adjustment for age, the adjusted risk among    PLWH was higher. Subsequent adjustment for comorbidities did not drastically alter hazard ratios (Table  1). In subgroup analysis by method of accounting for confounders, a weaker hazard ratio was seen in the smaller two studies which used propensity score matching (Fig.  S2, http://links.lww.com/QAD/C9) [12,18].  [18].

Unadjusted analyses of coronavirus disease 2019-related outcomes in people living with HIV
Eight studies provided lower quality evidence regarding COVID-19 outcomes in PLWH as they did not compare HIV-positive and negative people in the same cohort, or did not adequately account for confounders [13][14][15][16][17]23,29,30] (Table 1, Tables S2 and S3, http:// links.lww.com/QAD/C9). The largest of these was a Spanish multi-centre study of 77 590 PLWH, of whom 236 were diagnosed with COVID-19 and 20 died. In keeping with our meta-analysis result, age-standardized and sex-standardized mortality from COVID-19 were found to be higher in PLWH (3.7 per 10 000) compared with the general population (2.1 per 10 000) [13]. The other seven studies [14][15][16][17]23,29,30] were limited due to being at single sites, having small sample sizes (median 64, range n ¼ 31-161 PLWH diagnosed with COVID- 19) and not accounting for potential confounding. These studies report conflicting results with one suggesting a higher rate of hospitalization and mortality among PLWH compared with the general COVID-19-positive population [14], two studies suggesting lower COVID-19 mortality in PLWH [16,29] and four studies reporting no significant difference in the risk of adverse outcomes from COVID-19 between PLWH and HIV-negative cohorts [15,23] or the general population [17,30].

Risk of death and hospitalization in relation to CD4 R T-cell count and HIV viral load
Several of the large cohort studies did not include data on CD4 þ T-cell count or HIV viral loads [12,21,22]. In the South African study, lower CD4 þ T-cell counts (measured during the COVID-19 episode) were associated with mortality, but this could be a result of, rather than causing, severe disease (Table 2). There was no difference in outcomes by HIV viral load, although viral load data were incomplete and numbers with unsuppressed viral loads were small [20]. A London HIV clinic found that 18 PLWH who were hospitalized with COVID-19 had a lower median CD4 þ T-cell count (395 vs. 573 cells/ml, P ¼ 0.03) compared with their 2699 PLWH outpatients (    [20]. While this analysis was adjusted for certain co-morbidities, the observed association may be confounded due to patients receiving TDF having less complex healthcare needs. In the Spanish multi-centre study, PLWH receiving TDF and emtricitabine had the lowest risk for COVID-19 diagnosis (16.9 per 10 000) and hospitalization (10.5 per 10 000) compared with all other ART regimens investigated, but without adjusting for co-morbidities [13]. A US hospital study found that PLWH and COVID-19 who survived were more likely to have been treated with nucleotide reverse transcriptase inhibitors (NRTIs) than those PLWH who died (99 vs. 89%, P ¼ 0.04) in univariate analysis [15]. Seven smaller studies (n ¼ 18-93 PLWH with  reported no significant association between ART-regimen and COVID-19 severity among PLWH [16,17,19,[24][25][26][27][28].  [19,22]. Other smaller analyses suggested that among PLWH, factors such as older age [27], metabolic disorders [27], obesity [16], African ethnicity [27] and organ transplantation [15] were associated with COVID-19 infection or severity.

Discussion Summary
Emerging evidence suggests an increased risk of COVID-19-related death in PLWH. Whether this increased risk is associated with to HIV viral load, CD4 þ T-cell counts or ART use was not clear as data in the included studies was insufficient. Regarding differences in effects of specific ART regimens, we found some evidence that TDF-based regimens may be associated with lower frequency of SARS-CoV-2 infection and milder courses of COVID-19 compared with other ART regimens, although this was not consistent between studies and was susceptible to confounding. Risk factors for severe COVID-19 among PLWH include older age, obesity and black ethnicity, and appear similar to the general population.
Risk of coronavirus disease 2019-related mortality among people living with HIV In our review, the two population-based studies from South Africa and the United Kingdom both suggested almost double the risk of COVID-19-related death among PLWH, despite having very different demographic profiles [20,22]. In contrast, studies restricted to cohorts of PLWH diagnosed with COVID-19 [12], and hospitalized patients with COVID-19 [18,21] found a weaker or null effect. These studies are more at risk of selection bias, as PLWH with milder symptoms may be more likely to test for SARS-CoV-2 or be hospitalized by clinicians (due to a higher perceived risk), compared with people without HIV who may only be tested or hospitalized once more severely unwell. This would lead to the cohort of PLWH being less unwell at baseline compared with the HIV-negative cohort, leading to underestimation of any association between HIV status and COVID-19-related mortality. Furthermore, studies restricted to hospitalized patients cannot account for the effect of HIV (or any other potential risk factor) on SARS-CoV-2 infection and COVID-19 severity which result in hospitalization, and therefore may underestimate the effect of risk factors on COVID-19 death, compared with studies in the general population [31].

Influence of antiretroviral therapy
We found no evidence to determine whether ART use reduces COVID-19 severity through immune reconstitution, as most studies only included PLWH on ART. Regarding specific antiretrovirals, the potential therapeutic value of TDF for COVID-19 is supported by results from molecular docking studies [32]. However, TDF is relatively contra-indicated in renal impairment [33], meaning patients receiving TDF-based ART are likely to have less co-morbidities, which may explain the observed better COVID-19 outcomes. Randomized trials of TDF prophylaxis for SARS-CoV-2 are underway [34].
Comparisons with existing literature PLWH are known to be at higher risk of respiratory bacterial infections, but the evidence regarding acute viral infections is less clear [35]. A review from the H1N1 influenza pandemic in 2009/2010 found some evidence of a higher risk of adverse H1N1 outcomes among PLWH who were severely immunocompromised [36].
However, the quality of the evidence was weak with a lack of rigorously designed prospective cohort studies, reflecting the challenges of in-pandemic research [36].

Limitations
Our meta-analysis of five studies is potentially limited by the small numbers of PLWH with COVID-19 who died. This presented challenges when accounting for confounding; studies that used multi-variable analyses to adjust for confounding were susceptible to overfitting of models and potential overadjustment by factors which could be on the causal pathway between HIV and death (e.g. malignancy or tuberculosis). Studies that used matching were potentially underpowered, which may explain why they tended to report no independent association between HIV and COVID-19 death. Concomitant treatment with corticosteroids, which reduce COVID-19 mortality [44] and may have been used differently by HIV status, was only reported by one study [18]. In our narrative synthesis, the majority of the studies were small case series or cohort studies that did not adequately account for confounders such as age. Most were performed in high-income countries, and the majority of participants had well controlled HIVon ART. This may limit the applicability to populations of PLWH in other settings. Only 68% of adults and 53% of children living with HIV globally are receiving ART [4], highlighting a crucial need to examine the risk of COVID-19 complications in these populations.

Conclusion
We present evidence which suggests a moderately increased risk of COVID-19 death among PLWH. Measures to mitigate COVID-19 risk among PLWH should be included in HIV programs. Further research into the role of ART, immunosuppression and viral suppression is needed to quantify and address risks for PLWH in diverse settings.