This is a systematic review and meta-analysis of published observational studies. Articles were selected if they fulfilled the following entry criteria: compliance with the PICO framework, namely P = confirmed positive COVID-19 patients, I = patients living with HIV, C = HIV-uninfected persons and O = mortality in COVID-19-confirmed patients not attributable to unrelated conditions such as trauma. The studies included were randomised clinical trials, cohort, case-cohort and cross-over design, and the full-text paper had to be available and to have been published. Excluded studies included non-original research such as review articles, letters or commentaries; case reports; studies in a language other than English; studies of children and youths <18 years of age and pregnant women.

A systematic search of PubMed and Europe PMC provided many papers. Additional articles were located by analysing the papers cited by the authors of the identified studies. The search terms included ‘HIV’ or ‘human immunodeficiency virus’ or ‘immunocompromised’ or ‘immune-deficient’ or ‘AIDS’ or ‘acquired immunodeficiency syndrome’ and ‘SARS-CoV-2’ or ‘coronavirus disease 2019’ or ‘COVID-19’ or ‘novel coronavirus’ or ‘nCoV’. The selected time-range included 01 December 2019 to 19 January 2021. Only English-language articles were evaluated. Details of the search strategy are listed in Table 1. Studies of HIV and SARS-CoV-2 co-infection with a valid definition of ‘mortality’ were included. The search strategy is presented in the preferred reporting items for systematic reviews and meta-analyses (PRISMA) diagram.

The initial investigation located 10 733 studies. After the removal of duplicates, 8653 records remained. A further 8585 studies were excluded after screening of the titles and abstracts failed to match with the inclusion and exclusion criteria. Of the 68 full-text articles evaluated for eligibility, 22 that lacked control or comparator groups were excluded, and 15 more were excluded because they lacked outcomes pertinent to our study. Three articles that were not in the English language were rejected. The final meta-analysis included 28 observational studies^{21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44} that reported on 18 255 040 COVID-19-infected persons, of whom 48 703 were co-infected with both HIV and SARS-CoV-2 (see Figure 1). Of the included articles, 25 were retrospective and 3 were prospective (see Table 2).

The study’s outcome of interest was mortality from COVID-19. This was defined as the number of patients with COVID-19 whose death could not be attributed to a cause other than COVID-19. Two authors performed the data extraction. Relevant demographic, laboratory and clinical information was recorded on a dataform: age, gender, ethnicity, the number of PLWH, the number of patients with a CD4 cell count of <200 cells/μL, the use of antiretroviral therapy (ART) and the mortality outcomes of both HIV-infected and HIV-uninfected participants. Two authors independently assessed the quality of each study using the Newcastle–Ottawa Scale.⁴⁵ The selection, comparability and outcome of each study were assigned a score from zero to nine. Studies with scores of ≥7 were considered to be of good quality (see Table 3). All included studies were rated ‘good’. In summary, all studies were deemed fit to be included in the meta-analysis.

Review Manager version 5.4 (Cochrane Collaboration) and the Comprehensive Meta-Analysis version 3 software were used in the meta-analysis, and Mantel-Haenszel’s formula gave odds ratios (ORs) and 95% confidence intervals (CIs). The heterogeneity was assessed using the I² statistic with values of <25%, 26% – 50% and >50% providing low, moderate and high degrees of heterogeneity, respectively. Significance was obtained if the two-tailed P-value was ≤0.05. The qualitative risk of publication bias was assessed using Begg’s funnel plot analysis.

Our pooled analysis indicated that HIV was associated with mortality from COVID-19 [OR = 1.19 (95% CI 1.01–1.39), p = 0.03; I² = 72%, random-effect modelling] (see Figure 2).

However, meta-regression showed that the association between HIV and mortality from COVID-19 was unaffected by age (p = 0.208), gender (p = 0.608) (see Figure 3a), Black ethnicity (p = 0.389), CD4 cell count of <200 cells/μL (p = 0.353) (see Figure 3b) or ART (p = 0.647) (see Figure 3c).

The subgroup analysis revealed that the association between HIV and mortality from COVID-19 was only statistically significant for studies from African regions [OR = 1.13 (95% CI = 1.04–1.23), p = 0.004; I² = 0%, random-effect modelling] and the United States of America (USA) [OR = 1.30 (95% CI = 1.08–1.59), p = 0.006; I² = 61%] but not for studies from Asia [OR = 2.41 (95% CI = 0.16–36.57), p = 0.53; I² = 76%], or Europe [OR = 0.90 (95% CI = 0.70–1.15), p = 0.40; I² = 5%].

The funnel plot analysis revealed a qualitatively symmetrically inverted funnel plot for the association between HIV and a mortality outcome, suggesting no publication bias. This is demonstrated in Figure 4.

Firstly, only a limited number of our included studies reported on CD4 cell counts, viral loads and ART – a fact that is likely to have impacted the precision of the meta-regression analysis of this study. Indeed, most studies focussed on the characteristics of COVID-19 patients rather than its effects on PLWH. Secondly, the studies utilised in this review and meta-analysis were primarily observational and thus, may reflect occult confounders or biases unique to the particular study. Finally, we included some preprint studies to minimise the risk of publication bias; however, we made exhaustive efforts to ensure that only sound studies were included that we expect will eventually be published. We hope that this study can give further insight into the management of COVID-19 patients.