HIV infection increases the risk of acquiring Plasmodium vivax malaria: a 4-year cohort study in the Brazilian Amazon HIV and risk of vivax malaria

Globally, malaria and human immunodeficiency virus (HIV) are both independently associated with a massive burden of disease and death. While their co-infection has been well studied for Plasmodium falciparum, scarce data exist regarding the association of P. vivax and HIV. In this cohort study, we assessed the effect of HIV on the risk of vivax malaria infection and recurrence during a 4-year follow-up period in an endemic area of the Brazilian Amazon. For the purpose of this study, we obtained clinical information from January 2012 to December 2016 from two databases. HIV screening data were acquired from the clinical information system at the tropical hospital Fundação de Medicina Tropical Dr. Heitor Vieira Dourado (FMT-HVD). The National Malaria Surveillance database (SIVEP malaria) was utilized to identify malaria infections during a 4-year follow-up period after diagnosis of HIV. Both datasets were combined via data linkage. Between 2012 and 2016, a total of 42,121 people were screened for HIV, with 1569 testing positive (3.7%). Out of all the patients diagnosed with HIV, 198 had at least one episode of P. vivax malaria in the follow-up. In the HIV-negative group, 711 participants had at least one P. vivax malaria episode. When comparing both groups, HIV patients had a 6.48 [(5.37–7.83); P < 0.0001] (adjusted relative risk) greater chance of acquiring P. vivax malaria. Moreover, being of the male gender [ARR = 1.41 (1.17–1.71); P < 0.0001], Amerindian ethnicity [ARR = 2.77 (1.46–5.28); P < 0.0001], and a resident in a municipality of the Metropolitan region of Manaus [ARR = 1.48 (1.02–2.15); P = 0.038] were independent risk factors associated with an increased risk of clinical malaria. Education ≥ 8 years [ARR = 0.41 (0.26–0.64); P < 0.0001] and living in the urban area [ARR = 0.44 (0.24–0.80); P = 0.007] were associated to a lower risk of P. vivax malaria. A total of 28 (14.1%) and 180 (25.3%) recurrences (at least a second clinical malaria episode) were reported in the HIV-positive and HIV-negative groups, respectively. After adjusting for sex and education, HIV-positive status was associated with a tendency towards protection from P. vivax malaria recurrences [ARR = 0.55 (0.27–1.10); P = 0.090]. HIV status was not associated with hospitalizations due to P. vivax malaria. CD4 + counts and viral load were not associated with recurrences of P. vivax malaria. No significant differences were found in the distribution of parasitemia between HIV-negative and HIV-positive P. vivax malaria patients. Our results suggest that HIV-positive status is a risk factor for vivax malaria infection, which represents an additional challenge that should be addressed during elimination efforts.

HIV-status and malaria episode database matches. A total of 15,561,260 malaria tests were performed in the state of Amazonas between January 2012 and December 2020, according to the SIVEP-malaria database. The linkage process between the databases for malaria and HIV resulted in 1,954 and 15,342 matches between HIV-positive and HIV-negative groups, respectively. After a detailed analysis and further exclusion of false matches and malaria episodes reported before HIV testing, a total of 936 participants were identified as having had at least one episode of P. vivax malaria during follow-up, 198 in the HIV-positive group and 711 in the HIV-negative group (Fig. 1). Table 2   Impact of CD4+ counts and viral load on malaria incidence and recurrence of P. vivax in HIV-positive patients. Among individuals with HIV-positive status, no association was found between CD4+ counts at baseline and the rates of total malaria and P. vivax malaria. Individuals carrying a viral load   (Table 4). CD4 + counts and viral load were not associated with recurrences in P. vivax malaria ( Table 4).

Impact of HIV status and other variables on malaria incidence.
Impact of HIV status on malaria parasitemia. No significant differences were found in the distribution of parasitemia between HIV-negative and HIV-positive P. vivax malaria patients.  www.nature.com/scientificreports/

Discussion
The high prevalence of HIV and P. vivax malaria in tropical and subtropical regions may increase the simultaneous co-infection by these two pathogens. In this study, HIV-positive status was associated to increased risk of P. vivax malaria, as previously observed in Africa with P. falciparum malaria 3,13,37 . To the best of our knowledge, there is no information on the effect of HIV infection on P. vivax susceptibility, but as occurs in P. falciparum malaria [38][39][40] , abnormalities in the immune response may increase susceptibility to P. vivax malaria by compromising naturally acquired immunity [41][42][43] . A mixed pattern of proinflammatory and regulatory biomarkers involving a robust IL-10 and IL-6 axis is produced in P. vivax malaria 43 and, in recurrent P. vivax episodes, an unbalanced CD4 + /CD8 + T-cell ratio was found and is associated with a significant increase in IL-10 levels 44 . Genetic polymorphisms in TLRs play a role in P. vivax susceptibility and parasitemia 42 , possibly by activation of IL-6, IFN-γ, IL-10 pathways 45 . From an integrated perspective, mechanisms behind recurrences and subclinical P. vivax infections are related to inhibitory receptors, T regulatory cells, and IL-10, with a poor induction of immunological memory cells and inefficient T effector cells 46 . Since HIV infects CD4 + T cells, a disruption of  www.nature.com/scientificreports/ the components of both the cellular and humoral aspects of the immune response is observed, thus predisposing individuals to many infections 43 . Evidence from modelling studies 47 and historical observations suggests that fevers from other systemic parasitic and bacterial infections can activate P. vivax hypnozoites 48 . If this is true, opportunistic infections in HIV/ AIDS patients could also trigger new episodes of P. vivax malaria through relapse. Additionally, reports also suggest that antimalarial treatment failure in P. falciparum malaria is more common in HIV-infected patients with low CD4-cell counts compared to those not infected with HIV 49 . In this study, HIV-positive status was associated with a tendency towards protection from recurrences of P. vivax malaria. Although some preliminary evidence demonstrated antiplasmodial 50 and hypnozoiticidal 51 activities of HIV protease inhibitors, our results point to another explanation for a lower frequency of recurrences in HIV patients, since, overall, P. vivax malaria was significantly less common in this group. Moreover, prophylaxis of opportunistic infections using trimethoprim/ sulfamethoxazole would possibly minimize the clinical manifestations of recurrences, as seen in P. falciparum malaria 52 . As the HIV patients in this study were followed in a tertiary institution for infectious diseases, including malaria, they should have received more information about the need to complete antimalarial treatment, in a more rigorous clinical follow-up to reach a radical cure. It should also be considered that participants with HIV/ AIDS may have a lower body weight than those in the HIV-negative group, which could lead to proportionally higher plasma levels of primaquine and its active metabolites, resulting in higher radical cure rates.
In this study, no significant differences were found between HIV-negative and HIV-positive P. vivax malaria patients in the distribution of parasitemia and severity. This differs from studies linking HIV to P. falciparum parasitemia, which demonstrate that HIV is associated with an increase in parasitemia and severity [53][54][55][56] . However, the number of severe cases was very small, which may have limited the statistical power of this comparison.
To the best of our knowledge, this is the only cohort that has been studied to assess the influence of HIV infection on P. vivax infections. Regardless of the large number of coinfections, some limitations were imposed by the study design, and these should be stated. HIV patients may be more likely than their HIV negative counterparts to seek treatment for a febrile illness than HIV-negative individuals. HIV patients are more likely to need ongoing outpatient monitoring and therefore will interact with the medical system more frequently.
In accordance with the case reporting procedures, only positive results for malaria are registered in SIVEPmalaria, which prevents to determine the universe of tested individuals, both in HIV positive and controls, potentially generating a substantial detection bias. In the Brazilian Amazon, however, fever is closely associated with malaria by the inhabitants and, in the case of the presence of this sign, testing for malaria is routine in all health units. As a retrospective cohort study, such confounders related to different access to the health system could not be collected. Additionally, as per the Brazilian Ministry of Health and local protocols, the treatment for both diseases is provided free-of-charge and under prescription, which means the diseases are necessarily conditioned to case reporting, and this is expected to minimize underreporting and related bias. Another weakness is regarding the absence of a unique identifier that allows for a deterministic linkage between distinct databases. In order to ensure high accuracy in this process, we used a high throughput algorithm that was previously described 35 and that has both high sensitivity and specificity when used for probabilistic linkage of public health datasets.
This study hypothesizes that HIV-infected patients are more likely to present P. vivax malaria. HIV counselling and testing programs, as well as approaching patients with febrile illnesses for the diagnosis of malaria should be strengthened to improve HIV/AIDS control strategies and the determination of malaria and HIV co-infection to help and articulate prevention and HIV/malaria control. With continuing interest in the elimination of malaria, it is necessary to have a better understanding of HIV-P. vivax co-infection, with the goal of reducing the burden of P. vivax malaria. Prospective studies are needed to confirm our hypothesis.

Methods
Ethics statement. This study was conducted in accordance with the principles of the Declaration of Helsinki and the guidelines of Good Clinical Practice of the International Harmonization Conference. The study was approved by the Ethics Review Board (ERB) of the Fundação de Medicina Tropical Dr. Heitor Vieira Dourado (CAAE: 09,714,619.2.0000.0005). The ERB gave a waiver for informed consent. After database linkages, the final dataset was anonymized before statistical analysis.

Study site.
Out of all malaria cases in Brazil, more than 99% occur in the Brazilian Amazon region, of which about 89% are attributable to P. vivax infections 28 . The national HIV prevalence is estimated at 0.4%, with a noticeably higher HIV incidence rate of 34.8/100,000 population in the Brazilian Amazon, compared to the overall national rate of 17.8/100,000 29 .
FMT-HVD is a tertiary, referral healthcare center for infectious diseases in Manaus, state of Amazonas, western Brazilian Amazon and, in its outpatient clinic and emergency department, it receives all patients seeking care for infectious and parasitic diseases or those referred from other health units in the city of Manaus and the surrounding areas. HIV testing is provided after careful health screening for possible epidemiological risk and exposure. Diagnosis and treatment are free of charge. For both diseases, any positive HIV or malaria diagnoses are compulsorily recorded in structured forms available on-line as part of the Brazilian Ministry of Health's notification system. FMT-FVD is responsible for the follow-up and management of ~ 80% of the HIV/ AIDS cases in the state of Amazonas.
Study design and exposure. This is a concurrent, cohort study, based on surveillance data from patients tested for HIV at FMT-HVD and malaria diagnosis data and case recurrences recorded in the National Malaria Epidemiological Surveillance Information System (SIVEP Malaria) between January 2012 and December 2016. www.nature.com/scientificreports/ In this study, exposure was defined as a confirmed HIV-positive status, according to the national HIV infection diagnosis guideline 30 , as reported on the FMT-HVD Information System (iDoctor). In short, a rapid diagnostic test (RDT) (Bio-Manguinhos, Fiocruz) with 100% sensitivity and 99.8% specificity 31 was used in mobile testing campaigns and at the FMT-HVD outpatient clinics. For each positive test, blood samples were collected at the FMT-HVD laboratory, and the western blot technique was used to confirm the diagnosis, and the patient was then referred to the HIV ward for consultations, psychological support and ART therapy. All the remaining individuals with HIV-negative test results were included as non-exposed controls. HIV-positive patients were stratified by CD4 cell counts (/mm 3 ) and viral load (copies/mL), via data gathered from the FMT-HVD SISCEL (Laboratory Test Control System) database. The predictor variables sex, age, education (in years of schooling), self-reported ethnicity, place of residence (rural/urban), and municipality of residence were also used. All HIV-positive patients were treated according to the Brazilian Ministry of Health Guidelines 32 .
Outcomes. The present study was designed to determine the risk of (i) malaria infection, (ii) P. vivax malaria, (iii) P. falciparum infection, (iv) P. vivax malaria recurrences up to 90, 180 days and 4-years, (v) malaria-associated hospitalizations, (vi) malaria-associated deaths, in a 4-year follow-up, and (vii) parasitemia. Recurrence was defined as the occurrence of at least one additional malaria infection during the follow-up period. Furthermore, the study aimed to determine any association between age, sex or parasitemia as a potential additional risk factor for malaria reinfection. In Brazil, thick blood smears (TBS) are routinely performed for the diagnosis of malaria, and are prepared according to the Walker technique 33 and then evaluated by a local microscopist. The results of peripheral parasitemia are given using the following semi-quantitative system: < 1/2 + (< 200 parasites/mm 3 ); 1/2 + (200-300 parasites/mm 3 ); 1 + (301-500 parasites/mm 3 ); 2 + (501-10,000 parasites/mm 3 ); 3 + (10,001-100,000 parasites/mm 3 ); and 4 + (> 100,000 parasites/mm 3 ). All positive slides and 10% of negative slides are routinely reviewed at a referral unit with experienced microscopists. In case of divergence, the reviewed result is updated on the on-line system. Only participants with a positive TBS for Plasmodium spp. are treated according to the Brazilian Anti-Malarial Treatment Guidelines 34 . The malaria patients' data are entered into the SIVEP Malaria (National Malaria Surveillance) database. Data on malaria-associated hospitalizations and deaths were extracted from HIS (Hospitalization Information System) and MIS (Mortality Information System) databases, respectively.
Data processing and record linkage strategy. Duplicated HIV tests were excluded from the iDoctor database (HIV-status database; 2012-2016 period) and, in the case of individuals that had had more than one HIV test in the period, the first positive result was considered. Patients exposed to disease (HIV-positive) and the controls (HIV-negative) were included in the study according to the day of confirmation of HIV-status. Patients entered and left the study at different points in time. The HIV-status database and the SIVEP-malaria database were both deduplicated and further the record linkage was performed (Malaria database; 2012-2020) of Tucuxi-BLAST software, which uses the DNA-encoded approach 35 and was installed on a Linux Workstation, Intel Core i7-8700, with 32 GB of RAM and 12 physical processing cores. In the absence of a unique identifier, Tucuxi-BLAST used the patient's name, date of birth and mother's name. Malaria episodes reported before the participants' entries were then excluded from the generated database. All remaining matches were double-checked and false matches were also excluded. This inspection included visual confirmation of homonyms, possible siblings (twins), and duplicity. Finally, the list of malaria cases obtained was linked to the HIS (Hospitalization Information System) and MIS (Mortality Information System) databases to investigate these outcomes. We obtained a final selection of pairs identified as likely to be from the same patients by automatic verification, applying a probability threshold (probability > 0.7) for all linkages. A final identification (ID) was created for all included participants.
Statistical analysis. Descriptive statistics were used for demographic variables. Student's t test was used to compare means while Fisher's exact or the Chi-squared (X 2 ) test were used to compare proportions, as appropriate. Crude relative risk (RR) with its respective 95% confidence interval (95% CI) was determined in a univariate analysis. Logistic regression was used for the multivariate analyses and the adjusted RR (ARR) with 95% CI were also estimated. A log binomial multivariate generalized linear regression was performed using an automated backward and forward stepwise estimation. All variables that were associated with dependent variables at a significance level of P < 0.200 in the univariate analysis were included in the multivariate analysis. Statistical significance was considered when P < 0.05 in the Hosmer-Lemeshow goodness-of-fit test. Kaplan-Meier survival estimates were used to compare recurrences over follow-up time. Additionally, a survival analysis for recurrent events was performed 36 . A two-tailed P < 0.05 was considered significant. The statistical analyses were carried out using R software (version 4.1.0), R Studio (version 1.4.17), and Stata v.13.0 (Stata Corp. LP, College Station, TX).

Data availability
All data generated or analysed during this study are included in this published article [and its supplementary information files].