Introduction

The SARS-CoV-2 infection was declared a global health emergency and categorized as a pandemic by the World Health Organization (WHO)¹. As we learn more about the infection, it has been established that severity and mortality rates are associated with underlying health conditions, such as hypertension, diabetes, and cardiovascular diseases². Given these circumstances, many health institutions worldwide have questioned; whether there is an increased risk of morbidity and complications in immunocompromised patients³.

Understanding susceptibility to SARS-CoV-2 infection among people living with HIV (PLWH) or propensity to develop a critical condition are crucial questions for health providers worldwide. The rapid and globally progressive spread of the COVID-19 pandemic occurs among the last great pandemic of our era, HIV. Evidence to date does not suggest a higher incidence of the infection among PLWH, but once exposed, they have a greater risk of severe COVID-19⁴. PLWH can be disproportionately impacted by social determinants of health that may increase the risk of developing the infection and the severity of the disease⁴. In countries like Colombia, with high rates of late HIV presentation and still significant mortality⁸, PLWH can be especially vulnerable to the effects of SARS-CoV-2. They also have increased rates of comorbidities, such as cardiovascular disease, pulmonary disease, cancer, obesity, and diabetes, which can raise the risk of severe COVID-19⁴. For this reason, it is a priority to maintain close surveillance of this population and to evaluate the impact of COVID-19 on them and, above all, the factors associated with its susceptibility, the different clinical stages, and the presence of possible protective factors that could be later evaluated in another type of population.

Our objective was to evaluate the factors associated with the propensity of developing the infection in PLWH and differentiated by the patient's clinical condition.

Materials and methods

Setting

This study was conducted in a cohort of patients from Colombian health institutions providing attention to PLWH. These institutions are part of the VIHCOL group. The VIHCOL group is a research group that began its activities on February 2016, intending to establish an inclusive group of health institutions that offers integral attention to PLWH and generates a reliable and adequate information system. All centers report their indicators to the high-cost account of the Colombian Ministry of Health under its regulations and following the definitions of the Colombian clinical practice guideline. The institutions that took part in this study are located in Bogotá, Cali, and Medellín.

Participants

The criteria for selecting the participants and reviewing their medical history were: people older than 13 years with a confirmed diagnosis of HIV infection based on the recommendations of the Colombian clinical practice guideline for HIV⁹. Data from people who had declined authorization for research participation before the start of this study or during was excluded. When the study began, we did not know the prevalence of COVID-19 in PLWH or the risk of developing this comorbidity in our country. For this reason, the sampling was incidental and convenient; in a 1:1 ratio of cases and controls. Due to the analysis being carried out through a multiple logistic regression model, it was proposed to have at least ten events per variable⁷.

Variables of study

Based on the knowledge of the researchers, the following variables were defined for analysis as possible associated factors: i) Biological sex of the person, male or female; ii) Age in years at the time of inclusion in the study; iii) Category of the residency area: urban, semi-urban or rural; iv) Clinical stage of the HIV infection at the time of inclusion in the study according to the classification revised by the Centers for Disease Control (CDC) in 2008¹¹; v) CD4+ T lymphocyte count, CD8 + and CD4/CD8 index; vi) Viral load count in copies/ml of HIV RNA at the time of the diagnosis of COVID-19 or inclusion in the study; vii) Description of the type of antiretroviral combination and drug classes included in the regime; viii) If the patient has had their first antiretroviral treatment modified during the clinical follow up, a description of the causes and the number of months since the modification until the study entry; ix) Comorbidities present by each participant (obesity and diabetes); x) Habits of the participants (active smoking); xi) Vaccination status against influenza, pneumococcus and hepatitis B and xii) Immigration status in the country.

Analysis of data

The data was collected from each health institution, and personnel from each center was responsible for completing the database based on the electronic medical record. A Microsoft Excel spreadsheet was provided to share the database with the principal investigator. Patients' data were coded to avoid sharing personal identities. The missing data was less than 5%, except for the CD8+ TL variable, which was greater than 20%, for which it was decided not to analyze and not to impute the missing values for other variables. The information was analyzed using the R programming language, version 4.1.0, and the RStudio integrated development environment. The normal distribution of the variables was evaluated using the Shapiro Wilk test, reporting the median and the range when these variables did not present a normal distribution. Using the Mann-Whitney U test, the difference in medians for quantitative variables was evaluated when these variables had two categories; when they had more than two categories, the Kruskal-Wallis test was used. About qualitative variables, the hypothesis of independence was evaluated using the chi-square test or Fisher's F test in case of having expected values less than 5 in more than 25% of the values. To analyze both factors associated with the acquisition of COVID-19 and disease severity, multivariate analyzes were performed using binary logistic regressions with all the variables that were considered associated with the outcome in the bivariate analysis for plausibility and statistical criteria (in this case, a value of conservative p of 0,2 was the criterion). The best model was chosen after evaluating its AIC value, R2, and the area under the curve. This final selected model was assessed for its adjustment with the Hosmer Lemeshow test, and the variables were considered significant if their p-value was less than 0.05.

Results

Data collected between May 1, 2020, and February 28, 2021, included 476 patients; 238 cases, and 238 controls, were analyzed. The characteristics of the participants are presented in Table 1. The median evolution time of the HIV infection was 4.52 years (range 0.04 to 23.10), and there was no difference between cases and controls. Among the patients with COVID-19, 31 (13%) were women, and 207 (87%) were men. Most patients lived in urban areas and were classified in stages 2 or 3 of HIV infection. No differences were found between the association of CD4 count and COVID-19 incidence, but instead were observed among patients under antiretroviral treatment (ART), receiving integrase and protease inhibitors, and for the variables smoking, migrant status, and diabetes (Table 1).

Of the 238 patients who acquired COVID-19, 196 (82,35%) received outpatient treatment, and 42 (17,64%) were hospitalized. Differences were observed between these outcomes regarding age, CD4 count, clinical stage of HIV infection, ART, vaccination status, causes of ART modification, obesity, and diabetes (Table 2).

Multivariate analysis: Regarding the diagnosis of COVID-19, the use of integrase inhibitors or protease inhibitors within the antiretroviral regime remained as factors that decreased the risk of acquiring COVID-19 by 56% and 46%, respectively (Table 3). Concerning hospitalization or outpatient treatment, age remains a factor that increases the possibility of being hospitalized as the person gets older; however, we are aware that PLWH and older than 50 years have a higher risk of cardiovascular comorbidity and that is the reason why this variable was categorized above this limit (50 years), with this analysis, the variable was significant between outpatient and hospitalized patients (p 0,0001). Patients under antiretroviral treatment were 83% less likely to be hospitalized, similar to those vaccinated against influenza had 86% fewer possibilities for the same outcome. Finally, patients with obesity were associated with a 389% higher risk of hospitalization (Table 4).

Discussion

This study was conducted in a cohort of PLWH from 3 health institutes located in the three main cities of Colombia. We described the factors associated with the COVID-19 acquisition and severity; based on a series of variables considered of interest by knowledge and prior experience of clinical experts. Regarding the acquisition of COVID-19, integrase or protease inhibitors used within the antiretroviral regime were associated with decreased risk, adjusted to active smoking, immigration status, and TLCD4+ count. Patients co-infected with SARS-CoV-2 and HIV, who are 50 years or older, and obese were more likely to be hospitalized. In contrast, receiving antiretroviral treatment and vaccination against influenza were protective factors for this outcome.

As the pandemic continues, the susceptibility of PLWH to SARS-CoV-2 infection remains a concern; due to a higher burden of some comorbidities and a weakened adaptive immune response⁸. A cohort study from 60 health institutions that care for 77.590 PLWHs in Madrid (Spain); estimated a higher risk of acquiring COVID-19 and requiring hospitalization among PLWHs aged 70 years or older. On the contrary, receiving Tenofovir Disoproxil Fumarate/ Emtricitabine (TDF/ FTC) within the antiretroviral treatment; resulted in a lower risk for both outcomes when compared with other drugs⁹. On the other hand, a multicenter cohort of 286 PLWH in the United States; observed that having three or more comorbidities was associated with both hospitalization and severe outcomes¹⁰. The authors of a systematic review reported that age and comorbidities appear to be the strongest predictors of severity and mortality in PLWH. Most patients that developed symptomatic COVID-19; had at least one comorbidity, more commonly hypertension, dyslipidemia, or type 2 diabetes¹¹. However, these same authors concluded that, although prior case series and cohort studies did not find an increased risk of SARS-CoV-2 infection or severe COVID-19 outcomes among PLWH, recent studies have pointed out an increased risk of severity even in the context of virologically controlled patients. Although, it is unknown whether this is due to a higher prevalence of comorbidities and other social determinants of health among PLWH¹¹.

In contrast to other studies, our results suggest a protective effect of protease inhibitors. This outcome can be related to the findings of studies that have analyzed the effect of inhibiting the protease on SARS CoV2 viral isolates, realizing that drugs such as ritonavir have in vitro activity against the virus¹². However, several studies on this medication, including randomized control trials, systematic reviews, and meta-analyses, have failed to demonstrate benefits on mortality, hospital admission, or mechanical ventilation^8,13. Distinctively, this effect has not been reported for integrase inhibitors, whereby our results regarding these drugs have to be deeply analyzed; before dropping out conclusions on this association.

Our study has various limitations. Despite being a case-control study nested in a cohort, the sample was selected from a cohort of patients located in the three main cities of the country; for this reason, patients may not represent the majority of sociodemographic contexts along Colombia. Additionally, the study is exploratory; since it was not developed from a previous hypothesis; other than the reports of variables that decreased the risk of severe COVID-19 (such as the use of TDF, for example). The variables analyzed were established by experts on the attention of PLWH and infectious diseases specialists, avoiding considering the analysis only based on statistical significance. Furthermore, bias surveillance may exist due to the possibility of undiagnosed cases within controls, for example, patients with an asymptomatic course of COVID-19 disease. This bias can overestimate some associations. The acquisition of COVID-19 was the principal outcome, and the sample was calculated for a logistics regression model for this result. Rates of hospitalization were evaluated to recognize variables that could be controlled to avoid this outcome. However, the small sample size could still be insufficient to find differences between variables and could affect the precision of the inferences.

On the other hand, the period in which the study was conducted had some specific epidemiological characteristics, including SARS-CoV2 variants and no access to effective vaccines; these factors added to a short followup limits the generalizability of the results.

Despite its limitations, this study is the first report providing a glance at the COVID-19 pandemic among PLWH in Colombia. The analysis and conclusions should be taken as an initial approach to this coinfection, not as confirmatory of associations and much less as causality. Our results would be helpful to generate hypotheses around variables associated with COVID-19, both for its acquisition and its severity, helping to propose new research designs that will help in their solving. In conclusion, although, there are still many questions regarding whether there is a differential risk of acquiring COVID-19 among PLWH. Antiretroviral treatment with integrase or protease inhibitors was associated with a lower probability of developing the co-infection. Some factors related to co-morbidities, such as older age and obesity, draw attention to possible risk factors for hospitalization in this population. Discussing the feasibility of new studies with proposed causal hypotheses that allow directing the research designs toward a more precise answer to these questions is essential.