Epidemiological and viral characteristics of undiagnosed HIV infections in Botswana

doi:10.21203/rs.3.rs-1449096/v1

Background: HIV-1 is endemic in Botswana. The country’s primary challenge is identifying persons living with HIV who are unaware of their status. We evaluated factors associated with undiagnosed HIV infection using HIV-1 phylogenetic, behavioural, and demographic data.

Methods: As part of the Botswana Combination Prevention Project, 20% of adults residing in 30 villages were tested for HIV and followed from 2013 to 2018. A total of 12,610 participants were enrolled, 3,596 tested HIV-positive at enrolment, and 147 participants acquired HIV during the trial. Extensive socio-demographic and behavioural data were collected from participants and next-generation sequences were generated for HIV-positive cases. We compared three groups of participants: 1) those previously known to be HIV-positive at enrolment (n=2,995); 2) those newly diagnosed at enrolment (n=601) and 3) those who tested HIV-negative at enrolment but tested HIV-positive during follow-up (n=147). We searched for differences in demographic and behavioural factors between known and newly diagnosed group using logistic regression. We also compared the topology of each group in HIV-1 phylogenies and used a genetic diversity-based algorithm to classify infections as recent (<1 year) or chronic (≥1 year).

Results: Being male (aOR=2.23), younger than 35 years old (aOR=8.08), educated and married was associated with undiagnosed HIV infection (p<0.001). Inconsistent condom use (aOR=1.76), multiple sexual partners (aOR=1.73) and repeated HIV testing (aOR=3.96) increased the odds of having undiagnosed HIV infection, while knowledge of one’s partner’s positive HIV status decreased those odds as compared to knowing one’s partner to be negative (aOR=0.077, p<0.001). The genetic diversity-based algorithm classified most incident infections as recent (75.0%), but almost none of known infections (2.0%). The estimated proportion of recent infections among new diagnoses was 37.0% (p<0.001).

Conclusion: Our results indicate that those with undiagnosed infections are likely to be younger, educated, married, men who do not consistently use condoms with their partners. Notably, these men have been tested for HIV previously, and a sizeable proportion had recent infections based on a genetic diversity-based classifier, suggesting that they are aware of their risk. In the era of “test and treat all”, these identified predictors may help better prioritize groups for prevention and testing.

HIV

undiagnosed infection

phylogenetics

recent HIV infection

Botswana was among the first countries to reach the Joint United Nations Programme on HIV/AIDS (UNAIDS) 90-90-90 targets, defined as 90% of people living with human immunodeficiency virus (HIV) aware of their status, 90% of those diagnosed on antiretroviral treatment and 90% of those on treatment, virally suppressed [1]. Botswana reached 93-93-98 in 2021[2]. However, the estimated prevalence of human immunodeficiency virus type 1 (HIV-1) ranges between 18.5% and 21% among adults in the general population and HIV-1 incidence ranges from 0.59–1.35% [3–5]

One of the main challenges in curbing transmission is identifying individuals with HIV who are unaware of their HIV status in order to link them to care. Early HIV symptoms are not sufficiently clear or specific to warrant testing, thus there is often a delay between HIV infection and diagnosis. This delayed diagnosis poses an obstacle to the elimination of HIV, as early HIV infection is associated with a disproportionate amount of onward transmission [6–9]. In Sub-Saharan Africa, several factors have been associated with poor HIV testing behaviours and undiagnosed HIV infection [2, 10–16]. These include age, gender, marital status, religion, education, employment status, sexual experience, condom usage [10–12, 17–19]. However, many prior studies have focused on deliberately delineated epidemiological groups such as adolescents, older adults (> 50 years old), or public sector clinic attendees.

To reduce the incidence of HIV-1 in Botswana, it would help to deploy targeted testing strategies for individuals at high risk of being infected and of transmitting the virus. It may be possible to identify these individuals based on their demographic or behavioural characteristics. The Botswana Combination Prevention Project (BCPP) evaluated whether a wide-ranging strategy would reduce HIV incidence over time[3, 4]. 20% of the population, across 30 communities, was systematically tested for HIV between 2013 to 2018. The study found 601 infections among participants who reported no prior positive test at the start of the study, and 147 individuals seroconverted during follow-up [3]. All study participants completed detailed questionnaires, and HIV-1 full genome sequences were generated from all participants with HIV [15, 16]. Using data from the BCPP study, we sought to evaluate factors associated with undiagnosed HIV-1 infection in Botswana, and to estimate time since infection among new diagnoses.

Study population

The BCPP study enrolled participants from 2013 to 2018 [3]. Approximately 20% of randomly selected individuals aged 16 or older in 30 communities across Botswana completed questionnaires and had blood drawn [3]. We collected questionnaire data in two broad categories: 1) Socio-demographics and community environment: participant and community variables, residency and mobility, education, employment, 2) HIV risk behaviour: HIV testing history and sexual behaviour (see [20] for protocols, questionnaires and data). Negative participants were followed up with yearly HIV tests.

HIV-1 full genome sequencing

HIV samples from all positively diagnosed participants were sequenced, regardless of antiretroviral therapy (ART) status and viral load. The majority of persons with known HIV were already on ART with viral suppression defined as viral load ≤ 400 copies/mL; in these instances, integrated virus was sequenced from viral RNA and proviral DNA templates. Next-generation sequencing (NGS) was performed by the BioPolimers Facility at Harvard Medical School [21] and through collaboration with the PANGEA HIV consortium [22–24] using Illumina platforms MiSeq and HiSeq, as previously described [25–27]. In brief, the nucleic acids were reverse-transcribed and PCR amplified. Amplicons were pooled in equimolar amounts for Illumina library preparation. Sequence assembly was performed de novo using SPAdes version 2.4.0. The HIV-1 strain, HXB2 (NC_001802), was used as the reference strain for sequence alignment and a consensus sequence was generated using abacas version 1.3.1 and MUMmer version 3.2. Next, sequence reads were mapped against the consensus sequence using SMALT version 0.5.0 [27]. For each participant, nucleotide frequency data at each HIV-1 site were generated.

HIV consensus sequences were subtyped using COMET [28], and only HIV-1 subtype C sequences were included in our analysis (accounting for > 99% of BCPP sequences). Based on the NGS reads, we were provided with nucleotide frequency files for each patient, detailing the relative frequency of each nucleotide at each site in the alignment. HIV sequences and basic demographic and clinical data are available upon request to the PANGEA consortium [24].

Statistical analysis

We stratified participants with HIV-1 into three groups as follows: 2,995 with previously diagnosed HIV (henceforth referred to as “known cases”), 601 persons with newly diagnosed HIV at enrolment (“new cases”), and 147 persons who seroconverted to HIV-positive during follow-up (“incident cases”). As study participants who tested negative at enrolment were then tested yearly, incident cases were known to have been infected for ≤ 1 year when they were diagnosed. Baseline characteristics and descriptive statistics for all participants have been previously described [3].

We compared responses to the questionnaire across our three groups. We included variables known to be associated with HIV infection and variables known to be associated with undiagnosed infection. We employed logistic regression to compare the characteristics of new cases versus known cases to identify predictors associated with undiagnosed infection. For each variable of interest, we performed univariate analyses comparing the two groups (new cases versus known cases). A single multivariable analysis was fit including demographic and behavioural predictors that were significant (p < 0.05) in the univariate analyses. We did not adjust for missing data in the multivariable analysis, and the proportion of complete cases in the data was 53.9%. Demographic predictors evaluated included gender, age, marital status, number of children, religious affiliation, education, and employment status. The behavioural predictors assessed were previous number of HIV tests, sexual activity (yes/no), number of partners, partner concurrency, condom use, condom use frequency in the past year, number of nights spent away from home and partner’s HIV status. Logistic regression was adjusted by community (n = 30) with a random effect using a model with robust standard errors in R. All variables were analysed as categorical variables except for age, which was analysed as a continuous predictor in the primary analysis. A sensitivity analysis was run with age as a categorical variable with five age ranges: 16–24, 25–34, 35–44, 45–54 and 55–64 years old.

We sought to use viral genetic data to determine whether the time from infection to diagnosis varied significantly between the three groups (new, incident, and known cases). First, we compared the assigned stage of infection based on within-host genetic diversity (< or ≥ 1 year) for each group, using a χ² test. We then compared the recency probability distributions between the three groups using a Kruskal-Wallis test. These comparisons were conducted for participants for whom deep sequencing nucleotide frequency data were available (n = 1,867). Finally, we compared the length of terminal branch lengths across the groups, and across each HIV-1 gene (gag, pol, env), in a pairwise manner, using Kolmogorov–Smirnov (KS) tests. These comparisons were conducted for participants for whom consensus genetic sequences were available for at least one gene (n = 2,872). This dataset included 2,339 known cases, 399 new cases and 134 incident cases. To account for multiple non-independent comparisons (across different genes), we used Bonferroni’s correction to assess statistical significance where appropriate. All statistical analyses were conducted in R (version 3.6.0) [29].

Phylogenetic analysis

For participants for whom viral genetic data were available (n = 2,872), we constructed phylogenies separately for each HIV gene region: gag, pol and env. Sequences were available for 2,339 known cases, 399 new cases and 134 incident cases. Maximum likelihood phylogenies were reconstructed using RaxML [30] under a GTR model with four gamma rates. Phylogenies included sequences from an additional 3,277 patients from clinics, to serve as local controls. Final phylogeny sizes were: gag (n = 5,631), pol (n = 6,084) and env (n = 5,840). We time-resolved the phylogenies using the treedater package, available in R [31], using sample times as tip dates. For each tip, we extracted terminal branch lengths (measured in time) using the “Analyses of Phylogenetics and Evolution” (APE) [32] and phytools [33] R packages. We compared the distributions of terminal branch lengths across our three groups (known cases, new cases, incident cases) using Kolmogorov-Smirnov tests.

Inference of stage of HIV infection

For each study participant for whom deep sequencing nucleotide frequency data were available (n = 1,867), we calculated the probability (0–1) of their infection being recent (< 1 year) based on within-host genetic diversity, demographic (age, sex) and clinical (treatment status, viral load) predictors using an xgboost gradient boosting [34] machine learning algorithm. The machine learning classifier is trained on a dataset of known recent (< 1 year) and chronic (≥ 1 year) infections to classify stage of HIV infection for individuals in each of our three groups. This analysis was conducted on all individuals for whom NGS coverage was sufficient to derive site-specific nucleotide frequency distributions (n = 1,867). We selected the threshold that optimized for accuracy (the highest number of overall correct classifications). The algorithm has been previously developed for and trained on the BCPP dataset on participants with known duration of infection [35]. For each participant in the present study, we used the algorithm to predict the probability of recency. We carefully excluded individuals comprised in the present study from the dataset used to train the classifier.

Participant demographics

In total, 12,610 people were enrolled in BCPP. Of these, 3,596 (29%) participants tested positive for HIV-1 at enrolment and 147 participants seroconverted during the study.

Predictors associated with undiagnosed HIV infection at baseline

Table 1

Demographic factors associated with newly diagnosed HIV-1 infections (n = 601) compared to known HIV-cases (n = 2,995) in Botswana
Variable	Category	Known cases (%)	New cases (%)	OR (95% CI)^a	aOR (95% CI)^b
Sex	Female	2252 (75.2%)	383 (63.7%)	Ref
	Male	743 (24.8%)	218 (36.3%)	1.85^***(1.48–2.3)	2.23^*** (1.84–2.7)
Age (years)	Median (IQR)	40.8 (34.5–49.1)	34.7(28.4–43.6)	0.95^*** (0.94–0.96)	0.94^*** (0.93– 0.95)
Marital status	Married	478 (16%)	62 (10.3%)	Ref
	Single/Never married	2288 (76.4%)	511 (85%)	0.99(0.89–1.11)	0.46^*** (0.38–0.58)
	Divorced/Widowed	227 (7.6%)	27 (4.5%)	0.5^***(0.41–0.6)	0.49^*** (0.36–0.67)
	Missing	2 (0.1%)	1 (0.2%)
Number of children^c	None	109 (3.6%)	38 (6.3%)	Ref
	1–5 children	1552 (51.8%)	269 (44.8%)	0.53^**(0.33–0.87)	–
	≥ 5 children	525 (17.5%)	63 (10.5%)	0.36^***(0.21–0.64)	–
	Missing	809 (27%)	231 (38.4%)
Religious affiliation	Affiliated with religion	2166 (72.3%)	381 (63.4%)	Ref
	No religious affiliation	828 (27.6%)	220 (36.6%)	1.57^***(1.26–1.96)	1.14(0.95–1.35)
	Missing	1 (0%)	0 (0%)
Education level	Tertiary	189 (6.3%)	57 (9.5%)	Ref
	Senior secondary	225 (7.5%)	86 (14.3%)	0.99(0.81–1.22)	0.83(0.61–1.16)
	Junior secondary	1169 (39%)	243 (40.4%)	0.34^***(0.29–0.4)	0.43^***(0.32–0.58)
	Primary	927 (31%)	145 (24.1%)	0.19^***(0.16–022)	0.55^***(0.38–0.79)
	Non-formal	462 (15.4%)	68 (11.3%)	0.22^***(0.18–0.26)	0.69 (0.48–1.0)
	Missing	23 (0.8%)	2 (0.3%)
^* - p < 0.05, ^**- p < 0.01, ^**- p < 0.001. n – number of participants, Und – undiagnosed, D –diagnosed, OR – odds ratio, aOR – adjusted odds ratio for clustering by community, CI – confidence interval, Ref – reference group, NA – not applicable (excluded men, only women were asked about number of children, women who refused to answer the question), ^a – univariate logistic regression, ^b – multivariable logistic regression, ^c – “number of children” was excluded from multivariable analysis due to interactions with sex. New cases are those that were newly diagnosed during BCPP.*

Demographic predictors of undiagnosed HIV infection at baseline

As participants aged, they were less likely to have an undiagnosed HIV infection (aOR = 0.94 per year, p < 0.001; Table 1). When age was examined as a categorical variable, the 16–24-year-olds and the 25–34 years olds were 8.1 and 1.6 times more likely respectively (p < 0.01) than the 55–64-year-olds, to have undiagnosed HIV infections (data not shown). Being single (aOR = 0.46, p < 0.001) and divorced or widowed (aOR = 0.49, p < 0.001) significantly reduced the odds of having an undiagnosed infection as compared to being married. Next, not having children increased the odds of harbouring undiagnosed infection compared to having children (OR = 0.53, p < 0.001). However, due to interactions with gender, number of children was excluded from the multivariable analysis. Individuals who were not affiliated with any religion were more likely to have undiagnosed HIV infections in the univariate analysis (OR = 1.57, p < 0.001), but religious affiliation was not significant in the multivariable analysis. The odds of undiagnosed infection increased with attaining higher education beyond Senior Secondary level (p < 0.001).

Table 2

Behavioural predictors associated with undiagnosed HIV-1 infection (n = 601) compared to those with diagnosed HIV-1 infection (n = 2,995) in Botswana
Variable	Category	Known cases (%)	New cases (%)	OR (95% CI)^a	aOR (95% CI)^b
Number of HIV tests^c	1–2	2180 (72.8%)	268 (44.6%)	Ref
	3–4	474 (15.8%)	99 (16.5%)	3.22^***(2.87–3.61)	3.96^***(3.10–5.07)
	4–9	228 (7.6%)	61 (10.1%)	5.41^***(4.67–6.29)	8.13^***(5.80–11.50)
	³10	35 (1.2%)	14 (2.3%)	9.26^***(6.64–13.35)	14.4^***(6.60–31.20)
	Missing	78 (2.6%)	159 (26.5%)
Number partners in 1 year	1 partner	1704 (56.9%)	327 (54.4%)	Ref
	2 partners	515 (17.2%)	131 (21.8%)	1.31^***(1.18–1.47)	0.76(0.30– 1.93)
	3 partners	26 (0.9%)	16 (2.7%)	3.18^***(2.15–4.90)	1.73(0.53–5.68)
	4 partners	25 (0.8%)	17 (2.8%)	2.57^***(1.72–4.01)	0.81(0.27–2.45)
	None	476 (15.9%)	66 (11%)	0.74^***(0.66–0.84)	–
	Missing	249 (8.3%)	44 (7.3%)
Concurrency in past year	No concurrent partners	1720 (57.4%)	332 (55.2%)	Ref
	Concurrent partners	550 (18.4%)	159 (26.5%)	1.42^***(1.28–1.59)	1.49(0.58–3.82)
	Missing	725 (24.2%)	110 (18.3%)
Partner’s HIV status^d	HIV negative	576 (19.2%)	161 (26.8%)	Ref
	HIV positive	1221 (40.8%)	102 (17%)	0.06^***(0.05–0.06)	0.077^***(0.06–0.09)
	Missing	1198 (40%)	338 (56.2%)
Condom use in past year ^e	No	1261 (42.1%)	166 (27.6%)	Ref
	Yes	945 (31.6%)	310 (51.6%)	2.16^***(1.96–2.38)	–
	Missing	789 (26.3%)	125 (20.8%)
Condom frequency	Always	1281 (42.8%)	181 (30.1%)	Ref
	Sometimes	783 (26.1%)	221 (36.8%)	1.81^***(1.63–2.01)	1.76^***(1.49–2.10)
	Never	181 (6%)	72 (12%)	2.73^***(2.31–3.23)	4.47^*** (3.42–5.84)
	Missing	750 (25%)	127 (21.1%)
Nights away from home	Zero	1568 (52.4%)	270 (44.9%)	Ref
	< 1 week – < 1 month	603 (20.1%)	133 (22.1%)	1.21^***(1.09–1.35)	0.86(0.68–1.10)
	> 1 month	822 (27.4%)	198 (32.9%)	1.53^***(1.39–1.68)	0.94(0.8–1.12)
	Missing	2 (0.1%)	0 (0%)
^* - p < 0.05, ^**- p < 0.01, ^***- p < 0.001. n - number of participants, OR – odds ratio, aOR – adjusted odds ratio for clustering by community, CI – confidence interval, Ref – reference group, ^a – univariate logistic regression, ^b – multivariable logistic regression, ^c – the test performed in the study counts towards total tests thus all participants had at least 1 HIV test, ^d – partners are listed as HIV positive or negative based on the index partner’s knowledge of their status, ^e – “condom use” was excluded from multivariable analysis due to interactions with condom frequency.

Behavioural predictors of undiagnosed HIV infection at baseline

Individuals who tested for HIV more frequently were more likely to have undiagnosed infections (Table 2, aOR = 3.96, p < 0.001). Participants with three or more sexual partners within the previous year had more than double the odds of having undiagnosed HIV infection compared to those who had one partner (OR = 3.18, p < 0.001), and individuals with concurrent partners also had increased odds (OR = 1.42, p < 0.001). However, these variables (number of sexual partners and concurrent partners) were not significant (p > 0.05) in the multivariable analysis. Individuals who reported knowing their partner’s HIV status to be negative were more likely to have undiagnosed HIV infection (p < 0.001). In addition, condom use in the past year was a significant predictor of undiagnosed infection in the univariate logistic regression (OR = 2.16, p < 0.001), but it was excluded from the multivariable analysis due its interaction with condom frequency. Participants who never used condoms, and those who used them only sometimes, were more likely to have undiagnosed infection than those who always used them (aOR = 4.47, p < 0.001 and aOR = 1.76, p < 0.001, respectively). Increased time away from home increased the risk of undiagnosed infection in the univariate (OR = 1.21 and 1.53, p < 0.001) but not the multivariate analysis.

Differences in assigned timing of HIV infection

Table 3

Timing of HIV infection among all participants
Variables	Incident cases N = 136^#	New cases N = 284^#	Known cases N = 1447^#	p-value
Recent, n(%)	102(75.0%)	105(37.0%)	29(2.0%)
Chronic, n(%)	34(25.0%)	179(63.0%)	1418(98.0%)	< 0.001
^#Analysis for participants with available data, n – number of participants. The Pearson χ² test for categorical data was used to generate the p-value.

Results were aligned with those expected: incident cases were nearly all classified as recent infections (75%), whereas 37% of newly diagnosed cases and only 2% of known cases were classified as recent (Table 3, p < 0.001 based on the Pearson χ² test). The classifier produces a probability of recency for each individual, rather than assigning a categorical stage, and we observed the same pattern when we compared the probability distributions across the three groups (Fig. 1). Incident cases had the highest probability of recency (P = 0.847) while known cases had the lowest probability (P = 0.005). New cases had an intermediary probability of being recently infected (P = 0.252).

Differences in terminal branch lengths

The statistical analysis for terminal branch lengths across three groups and three genes amounted to nine non-independent comparisons, therefore we set our threshold for significance to p < 0.0055. We observed no consistent pattern of difference in terminal branch lengths between groups.

Among a large representative population of adults with HIV in rural/peri-urban Botswana, we compared the demographic and behavioural characteristics of those with known, diagnosed HIV-infection compared with those newly diagnosed with HIV at the start of a clinical trial. Participants with undiagnosed HIV infections were more likely to be male, younger, married, and to be more educated. Undiagnosed HIV infection was also more common among persons who had tested repeatedly, who did not consistently use condoms, and whose primary partners were HIV-negative.

We compared the viral characteristics between those two groups, and a third: those diagnosed with incident infections during the course of the Botswana Combination Prevention Project (BCPP) trial (n = 147). Because this latter group were negative at the start of the trial, then tested yearly, we knew their infections were < 1 year when they were diagnosed. We wanted to use our two reference groups (known older cases, and incident cases) to evaluate how long those diagnosed at the start of trial were likely to have been infected before they were diagnosed. Viral genetic sequences are informative regarding stage of HIV infection because individuals are usually infected with a single virus and genetic diversity then increases with time [36–38]. Results from the genetic diversity-based classifier were concordant with epidemiological data: most known cases were classified as chronic (97.9%) while most incident cases were classified as recent (75.0%, demonstrating a high false negative rate nonetheless). The fraction of recent infections among newly diagnosed cases fell between those two, at 37.0%. Given the high false negative rate of the classifier this proportion may be an underestimate – therefore a sizeable proportion of newly diagnosed cases were likely to be recent infections. Another genetic signature of time between infection and diagnosis is the length of the terminal branch leading to a sequence [38–40]. Short terminal branch lengths indicate that sampling occurred shortly after the last transmission event. However, we found no consistent differences in the distribution of terminal branch lengths among our three groups. This lack of signal may be due to the low sample proportion of our population as a whole: the relationship between terminal branch length and time to diagnosis will be disrupted if too many transmissions are missed in the phylogeny [41].

Surveys across sub-Saharan Africa have consistently demonstrated that HIV testing uptake is higher among women than among men [2, 13, 42]. One reason women are more likely to know their HIV status is that most countries, including Botswana, screen for HIV during pregnancy as part of prevention of mother to child transmission [3, 43, 44]. This strategy may explain why in our study, women with more children were less likely to have undiagnosed HIV infections. Furthermore, fewer undiagnosed infections were observed as participants increased in age, in agreement with others [2, 13, 19, 42], and testing increased with age. HIV testing rates among men below 25 years are low [17]. In Botswana, people above 50 living with HIV are more likely to be aware of their HIV status and to be on antiretroviral therapy [36]. Older women in particular are more likely to be aware of their positive HIV status than older men [45].

The association between marriage and undiagnosed HIV infection has been noted previously [10, 11, 46]. A survey conducted in South Africa showed that married people living with their spouses were less likely to be HIV positive [47], but even in that study, HIV infections were highest among those who were married but who spent extended periods away from home. Similarly in our study being married increased the risk of an undiagnosed HIV infection. Knowing one’s partner to be HIV-positive decreased the odds of an undiagnosed HIV infection. This finding could indicate that sero-discordant couples take precautions to prevent transmission and to get diagnosed rapidly but is most likely to demonstrate the success of treatment as prevention. In our study, we found that those with a higher education were more likely to have undiagnosed HIV infections, despite education improving HIV testing behaviours [10, 43, 46]. Previous cross-sectional analyses have demonstrated a positive association between educational attainment and HIV positivity across sub-Saharan African countries[48]; educational attainment may increase the likelihood of sexual opportunities or risk-taking.

The behavioural factors associated with undiagnosed HIV were repeated HIV testing and inconsistent condom use. At first glance, this first finding seems contradictory: frequent testing increases the chances of diagnosing HIV within the early stages of infection and should decrease the probability that a person has an undiagnosed HIV infection. Most participants (72.8% of known, previously diagnosed cases and 44.6% newly diagnosed cases) were tested for the first-time at enrolment or had tested only once previously. It is possible that participants who test frequently do so because they are aware of being at risk for HIV. For example, they may use HIV testing as a greenlight for unprotected sex, or they may get tested following a risk event. Further research is required to make specific recommendations regarding optimal HIV testing timings and frequency for this group, who already test regularly. We note that once a person tests HIV positive, they will not continue to get tests. Therefore, there is some censoring in our data on number of tests, however this censoring should not bias our interpretation. Health seeking behaviours such as frequent HIV testing and safe sexual practices by using a condom reduce HIV transmission [12]. Consistent condom use offered protection against HIV infection and participants who never used condoms were 4 times more likely to have undiagnosed HIV than those who always did.

Despite the tremendous success of the “Test and Treat all” strategy in Botswana, 16.7% (601/3596) of BCPP participants with HIV at enrolment did not know their positive status. Taken together, our results indicate that adults with undiagnosed infections are likely to be young, educated, married, men who do not consistently use condoms with their partners. Notably, these men had been tested for HIV repeatedly, and a sizeable proportion had infections likely to be recent based on a genetic-diversity-based classifier, suggesting they are aware of their risk. Our results stress the importance of targeting interventions towards these men in places where they will be most useful; for example, by offering HIV self-testing or testing in workplaces, sporting events, barber shops or places where men congregate regularly. Clearer recommendations may be needed as to how frequently, and under what circumstances, this group should get tested for HIV. With high rates and coverage of HIV testing and antiretroviral therapy initiation, incorporating the identified predictors to prioritize HIV testing will help reduce national HIV incidence in Botswana.

HIV: human immunodeficiency virus, BCPP: Botswana Combination Prevention Project, NGS: next generation sequencing, KS: Kolmogorov–Smirnov, n: number of participants, OR: odds ratio, aOR: adjusted odds ratio, CI: confidence interval, Ref: reference group

Ethics approval and consent to participate

The BCPP study was approved by the Human Research Development Committee (HPDME 13/18/1) (the Institutional Review Board based at the Botswana Ministry of Health and Wellness), and the US Centers for Disease Control and Prevention (Protocol 6475). The study is registered at ClinicalTrials.gov (NCT01965470). All participants provided written informed consent and participants aged 16 to 18 years provided written assent (with parental or guardian written permission). The data collection procedures were performed in accordance with the Declaration of Helsinki.

Consent for publication

Not applicable

Availability of data and materials

Sequence data and basic demographics for this study are available upon request to the PANGEA HIV consortium (www.pangea-hiv.org). BCPP protocols and collected data are made available at https://data.cdc.gov/Global-Health/Botswana-Combination-Prevention-Project-BCPP-Publi/qcw5-4m9q. For full access, please use the data request form.

Competing Interests

SM is an editorial Board Member for BMC Infectious Diseases. All other authors report no competing interests.

Funding

We acknowledge funding from the MRC Centre for Global Infectious Disease Analysis (reference MR/R015600/1), jointly funded by the UK Medical Research Council (MRC) and the UK Foreign, Commonwealth & Development Office (FCDO), under the MRC/FCDO Concordat agreement and is also part of the EDCTP2 programme supported by the European Union. PANGEA consortium is funded by the Bill & Melinda Gates Foundation (OPP1084362 and OPP1175094). Additional support was provided by the Research England Global Challenges Research Fund to MRC. The President’s Emergency Plan for AIDS Relief (PEPFAR) funded the implementation and evaluation of the BCPP through the Centers for Disease Control and Prevention under the terms of cooperative agreements U01 GH000447, U2G GH001911, U2G GH000073, U2G GH002027, and U2G GH000419.

SG was partially supported by H3ABioNet. H3ABioNet is supported by the National Institutes of Health Common Fund [U41HG006941]. H3ABioNet is an initiative of the Human Health and Heredity in Africa Consortium (H3Africa) programme of the African Academy of Science (AAS). SM & SG were supported by the Fogarty International Center and National Institute of Mental Health, of the National Institutes of Health under Award Number D43TW010543. SM & SG were partly supported through the Sub-Saharan African Network for TB/HIV Research Excellence (SANTHE), a DELTAS Africa Initiative [grant # DEL-15-006]. The DELTAS Africa Initiative is an independent funding scheme of the African Academy of Sciences (AAS)’s Alliance for Accelerating Excellence in Science in Africa (AESA) and supported by the New Partnership for Africa’s Development Planning and Coordinating Agency (NEPAD Agency) with funding from the Wellcome Trust [grant #107752/Z/15/Z] and the U.K. government. The views expressed in this publication are those of the authors and not necessarily those of AAS, NEPAD Agency, Wellcome Trust, or the U.K. government. SL received funding from US NIH K24AI131928. The funders had no role in the study design, data collection and decision to publish, or in the preparation of the manuscript.

Disclaimer

The views expressed in this publication are those of the authors and not necessarily those of NIH, Bill & Melinda Gates Foundation, MRC UK, AAS, NEPAD Agency, Wellcome Trust, PEPFAR, or the U.K. government. The funders had no role in the study design, data collection and decision to publish, or in the preparation of the manuscript.

Authors’ contributions

LB, MRC, SM, conceived and designed the study. TG, EK, MM and JM conducted field work and quality assurance. SM and SG performed the lab work and generated viral data. PB, FU, SM, SG, MPH, EKY, MM, TG, JM, KEW, SL, ME, VN collected, cleaned, and compiled data. LB and MRC performed analyses and interpreted results. SM, SG, SL and VN provided support for analysis. LB wrote the first draft of the manuscript. All authors reviewed and approved the final manuscript.

Acknowledgements

We acknowledge the Botswana Combination Prevention Project study team, the research staff at Botswana Harvard AIDS Institute Partnership and, most importantly, the study participants. We thank all members of the PANGEA consortium, and in particular Tanya Golubchik for bioinformatic support.

Essex, M., J. Makhema, and S. Lockman, Reaching 90-90-90 in Botswana. Curr Opin HIV AIDS, 2019. 14(6): p. 442–448.
Lebelonyane, R., et al., To achieve 95-95-95 targets we must reach men and youth: High level of knowledge of HIV status, ART coverage, and viral suppression in the Botswana Combination Prevention Project through universal test and treat approach. PloS one, 2021. 16(8): p. e0255227-e0255227.
Gaolathe, T., et al., Botswana's progress toward achieving the 2020 UNAIDS 90-90-90 antiretroviral therapy and virological suppression goals: a population-based survey. Lancet HIV, 2016. 3(5): p. e221-30.
Makhema, J., et al., Universal Testing, Expanded Treatment, and Incidence of HIV Infection in Botswana. N Engl J Med, 2019. 381(3): p. 230–242.
Statistics, B., Botswana AIDS Impact Survey (BAIS) IV 2013, Statistics Botswana: Gaborone Botswana.
Pao, D., et al., Transmission of HIV-1 during primary infection: relationship to sexual risk and sexually transmitted infections. AIDS, 2005. 19(1).
Brenner, B.G., et al., High Rates of Forward Transmission Events after Acute/Early HIV-1 Infection. The Journal of Infectious Diseases, 2007. 195(7): p. 951–959.
Fisher, M., et al., Determinants of HIV-1 transmission in men who have sex with men: a combined clinical, epidemiological and phylogenetic approach. AIDS, 2010. 24(11).
Ragonnet-Cronin, M., et al., Longitudinal Phylogenetic Surveillance Identifies Distinct Patterns of Cluster Dynamics. JAIDS Journal of Acquired Immune Deficiency Syndromes, 2010. 55(1).
Mtowa, A., et al., Socio-demographic inequalities in HIV testing behaviour and HIV prevalence among older adults in rural Tanzania, 2013. AIDS Care, 2017. 29(9): p. 1162–1168.
Budambula, V., et al., Socio-demographic and sexual practices associated with HIV infection in Kenyan injection and non-injection drug users. BMC public health, 2018. 18(1): p. 193–193.
Auvert, B., et al., Randomized, controlled intervention trial of male circumcision for reduction of HIV infection risk: the ANRS 1265 Trial.
Alwano, M.G., et al., Increasing knowledge of HIV status in a country with high HIV testing coverage: Results from the Botswana Combination Prevention Project. PloS one, 2019. 14(11): p. e0225076-e0225076.
Mkandawire, P., Assessing factors associated with HIV testing among adolescents in Malawi. Global Public Health, 2017. 12(7): p. 927–940.
Shanaube, K., et al., What works - reaching universal HIV testing: lessons from HPTN 071 (PopART) trial in Zambia. AIDS (London, England), 2017. 31(11): p. 1555–1564.
Pettifor, A., et al., Factors Associated with HIV Testing Among Public Sector Clinic Attendees in Johannesburg, South Africa. AIDS and Behavior, 2010. 14(4): p. 913–921.
Marcus, U., et al., Recency and frequency of HIV testing among men who have sex with men in Germany and socio-demographic factors associated with testing behaviour. BMC public health, 2015. 15: p. 727–727.
Billings, J.D., et al., Factors associated with previously undiagnosed human immunodeficiency virus infection in a population of men who have sex with men and male-to-female transgender women in Lima, Peru. Medicine, 2016. 95(42): p. e5147-e5147.
Mustanski, B., et al., Factors Associated With HIV Testing in Teenage Men Who Have Sex With Men. Pediatrics, 2020. 145(3): p. e20192322.
CDC, Botswana Combination Prevention Project (BCPP) - Public Release Data https://data.cdc.gov/Global-Health/Botswana-Combination-Prevention-Project-BCPP-Publi/qcw5-4m9q. Accessed on 09 March 2022.
Biopolymers Facility. https://genome.med.harvard.edu Accessed on 9 March 2022.
Pillay, D., et al., PANGEA-HIV: phylogenetics for generalised epidemics in Africa. Lancet Infect Dis, 2015. 15(3): p. 259–61.
Abeler-Dorner, L., et al., PANGEA-HIV 2: Phylogenetics And Networks for Generalised Epidemics in Africa. Curr Opin HIV AIDS, 2019. 14(3): p. 173–180.
PANGEA HIV. https://www.pangea-hiv.org Accessed on 09 March 2022.
Ratmann, O., et al., HIV-1 full-genome phylogenetics of generalized epidemics in sub-Saharan Africa: impact of missing nucleotide characters in next-generation sequences. AIDS Res Hum Retroviruses, 2017. 33(11): p. 1083–1098.
Novitsky, V., et al., Long-Range HIV Genotyping Using Viral RNA and Proviral DNA for Analysis of HIV Drug Resistance and HIV Clustering. J Clin Microbiol, 2015. 53(8): p. 2581–92.
Gall, A., et al., Complete Genome Sequence of the WHO International Standard for HIV-1 RNA Determined by Deep Sequencing. Genome announcements, 2014. 2(1): p. e01254-13.
Struck, D., et al., COMET: adaptive context-based modeling for ultrafast HIV-1 subtype identification. Nucleic Acids Res, 2014. 42(18): p. e144.
Team, R.C., R: A Language and Environment for Statistical Computing. 2019, R Foundation for Statistical Computing: Vienna, Austria.
Stamatakis, A., RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics, 2006. 22(21): p. 2688–90.
Volz, E.M. and S.F. Frost, Scalable relaxed clock phylogenetic dating. Virus Evol, 2017. 3(2).
Paradis, E. and K. Schliep, ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics, 2018. 35: p. 526–528.
Revell, L., phytools: An R package for phylogenetic comparative biology (and other things). Methods in Ecology and Evolution, 2012. 3: p. 217–223.
Chen, T. and C. Guestrin. XGBoost: A Scalable Tree Boosting System. in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining 2016. New York, NY, USA.
Ragonnet-Cronin, M., et al., HIV genetic diversity informs stage of HIV-1 infection among patients receiving antiretroviral therapy in Botswana. The Journal of Infectious Diseases, 2021.
Carlisle, L.A., et al., Viral Diversity Based on Next-Generation Sequencing of HIV-1 Provides Precise Estimates of Infection Recency and Time Since Infection. The Journal of Infectious Diseases, 2019. 220(2): p. 254–265.
Ragonnet-Cronin, M., et al., Genetic Diversity as a Marker for Timing Infection in HIV-Infected Patients: Evaluation of a 6-Month Window and Comparison With BED. The Journal of Infectious Diseases, 2012. 206(5): p. 756–764.
Shankarappa, R., et al., Consistent viral evolutionary changes associated with the progression of human immunodeficiency virus type 1 infection. Journal of virology, 1999. 73(12): p. 10489–10502.
Poon, A.F., Impacts and shortcomings of genetic clustering methods for infectious disease outbreaks. Virus Evol, 2016. 2(2): p. vew031.
Moshiri, N., D.M. Smith, and S. Mirarab, HIV Care Prioritization Using Phylogenetic Branch Length. JAIDS Journal of Acquired Immune Deficiency Syndromes, 2021. 86(5).
Lewis, F., et al., Episodic Sexual Transmission of HIV Revealed by Molecular Phylodynamics. PLOS Medicine, 2008. 5(3): p. e50.
Staveteig, L., et al., Demographic patterns of HIV testing uptake in sub-Saharan Africa. DHS Comparative Reports. 2013 Calverton, Maryland, USA.
Singh, K., W. Luseno, and E. Haney, Gender equality and education: Increasing the uptake of HIV testing among married women in Kenya, Zambia and Zimbabwe. AIDS care, 2013. 25(11): p. 1452–1461.
(CDC)., C.f.D.C.a.P., Introduction of routine HIV testing in prenatal care-Botswana. 2004, MMWR Morb Mortal Wkly Rep. p. 1083-6.
Matlho, K., et al., HIV prevalence and related behaviours of older people in Botswana — secondary analysis of the Botswana AIDS Impact Survey (BAIS) IV. African Journal of AIDS Research, 2019. 18(1): p. 18–26.
Brima, N., et al., Factors Associated with HIV Prevalence and HIV Testing in Sierra Leone: Findings from the 2008 Demographic Health Survey. PloS one, 2015. 10(10): p. e0137055-e0137055.
Shisana, O., et al., Does marital status matter in an HIV hyperendemic country? Findings from the 2012 South African National HIV Prevalence, Incidence and Behaviour Survey. AIDS care, 2016. 28(2): p. 234–241.
Fortson, J.G., The gradient in sub-Saharan Africa: socioeconomic status and HIV/AIDS. Demography, 2008. 45(2): p. 303–322.

Competing interest reported. SM is an editorial Board Member for BMC Infectious Diseases. All other authors report no competing interests.

Epidemiological and viral characteristics of undiagnosed HIV infections in Botswana

Status:

Version 1

Abstract

Figures

Background

METHODS

Study population

HIV-1 full genome sequencing

Statistical analysis

Phylogenetic analysis

Inference of stage of HIV infection

Results

DISCUSSION

Conclusion

Abbreviations

Declarations