All study protocols, which included provisions for future analyses of HIV-1 transmission risk factors, were approved by the University of Washington Human Subjects Division as well as ethics review committees at each study site. The study was also reviewed by the U.S. Centers for Disease Control and Prevention (CDC), which deemed CDC personnel to be non-engaged, as they had no contact with study participants or access to personal identifiers. All study participants were adults or emancipated minors under Kenyan and Uganda law and provided written informed consent. Additionally, this study was conducted with approval from the Kenyatta National Hospital—University of Nairobi Ethics and Research Committee (KNH-UON ERC), which requires that we release data from Kenyan studies (including de-identified data) only with their written approval for additional analyses.

We utilized data obtained from individuals who experienced HIV-1 seroconversion while enrolled in one of four prospective cohort studies conducted in Kenya and Uganda. Three of these cohorts enrolled heterosexual HIV-1 serodiscordant couples: the Partners in Prevention HSV/HIV Transmission Study [25], the Couples Observational Study [26], and the Partners PrEP Study [27]. In total, these three cohorts followed more than 8,500 couples for 12 to 36 months, and 297 individuals seroconverted during follow-up. The fourth study was the Mombasa Cohort, an open cohort of female sex workers (FSW) that began enrollment in 1993. Through 2014, 3,471 women had enrolled and 332 had experienced HIV-1 seroconversion.

Individuals who HIV-1 seroconverted during study follow-up and who had a blood sample collected prior to seroconversion available for schistosomiasis testing were eligible for the present analysis. In addition, for the Partners in Prevention HSV/HIV Transmission Study and Couples Observational Study, we only included participants enrolled at study sites in Kenya and Uganda, where schistosomiasis is endemic [3]. Screening or mass-treatment for schistosomiasis were not included in the study protocol for any of the four cohorts. All HIV-1 seroconverters in the four cohorts were invited to continue their cohort enrollment. Seroconverters in the serodiscordant couples cohorts were invited to attend quarterly study visits, while seroconverters in the Mombasa Cohort were invited to monthly visits; at these visits, a blood sample was collected to measure plasma viral loads. Additionally, women in the Mombasa Cohort had vaginal and cervical swabs collected to measure genital viral loads. Vaginal swab collection was performed by rolling a Dacron swab for three full turns against the lateral vaginal wall, while cervical swabs were inserted into the cervical os and rotated two full turns (study procedures described previously [28]).

Schistosome infections were identified via a three-stage testing algorithm. Samples collected prior to HIV-1 seroconversion were initially tested by soluble egg antigen (SEA) ELISA to detect antischistosomal antibodies [29]. Since antibodies persist beyond the period of active infection, SEA ELISA-positive serum samples were tested for circulating anodic antigen (CAA) using the SCAA20 assay (detection threshold of 10 pg/ml) [30]; CAA becomes undetectable within two weeks after successful treatment with praziquantel [31]. To identify the schistosome species causing infection, CAA-positive samples had species-specific immunoblots performed for S. mansoni and S. haematobium [32, 33]. SEA ELISAs and species-specific immunoblots were performed by the Parasitic Diseases Branch of the CDC and CAA testing was performed at Leiden University Medical Center.

The assay used to measure plasma HIV-1 RNA levels varied across studies (Partners HSV/HIV: COBAS AmpliPrep/COBAS TaqMan HIV-1 RNA assay, version 1.0 [Roche Diagnostics]; Couples Observational Study and Partners PrEP Study: Abbott m2000 Real-Time HIV-1 RNA assay [Abbott]; Mombasa Cohort: Gen-Probe HIV-1 viral load assay [Gen-Probe Incorporated]). Genital swabs collected from the Mombasa Cohort were eluted in 1 ml of buffer, and the concentration of HIV-1 RNA was reported as HIV-1 copies/swab. Quantification of genital HIV-1 RNA was conducted using the Gen-Probe HIV-1 viral load assay (Gen-Probe Incorporated).

We estimated the date of HIV-1 acquisition using both serology and plasma viral load results. Study participants were tested for HIV-1 seroconversion at all routine study visits (monthly or quarterly). For the serodiscordant couples cohorts, dual rapid HIV-1 antibody tests were performed, followed by a confirmatory HIV-1 enzyme immunoassay and Western blot. For the Mombasa Cohort, HIV-1 ELISA testing was performed, with positive results confirmed by a second ELISA. In all cohorts, once a positive serology was confirmed, HIV-1 RNA levels were measured in plasma samples collected at the seroconversion visit and preceding visits until a visit prior to HIV-1 acquisition was identified.

For individuals with detectable HIV-1 RNA prior to HIV-1 seroconversion, we estimated the date of HIV-1 acquisition to be 17 days prior to the date of the first positive HIV-1 RNA result. For those whose first detectable HIV-1 RNA occurred at the seroconversion visit, the date of HIV-1 infection was estimated to be at the midpoint between the last seronegative and first seropositive visit. For this study we excluded participants without detectable plasma HIV-1 RNA prior to seroconversion who had >1 year between their last seronegative and first seropositive visit, since those individuals’ date of HIV-1 acquisition could not be estimated with adequate precision.

We defined the set point period as being between 4–24 months after HIV-1 infection. Thus, we included all eligible plasma or genital viral load samples collected 4–24 months after the estimated date of HIV-1 acquisition. We excluded any samples collected after the participant reported initiating antiretroviral therapy (ART); information on ART use was collected at all study visits. This approach is consistent with previous analyses conducted with these cohorts [34, 35].

We log₁₀-transformed plasma and genital HIV-1 RNA concentrations to approximate a normal distribution. Samples with undetectable RNA levels were set equal to half the lower limit of quantification (120 copies/ml for plasma and 25 copies/swab for genital samples). Linear mixed-effects models with a random intercept for each individual evaluated if schistosomiasis was associated with plasma or genital set point viral load levels. Multivariable models evaluating set point plasma viral loads adjusted for a set of a priori confounders identified through a review of the literature: age [36], sex [37, 38], study, and four year bands of calendar year (Mombasa Cohort only, to account for time trends in HIV-1 transmission in that community) [36]. Multivariable models for genital viral loads included this same set of confounders with additional adjustment for plasma viral load levels obtained at the same study visit as genital viral loads. At the request of a reviewer, we also evaluated if hormonal contraceptives or sexually transmitted infections confounded the association between schistosomiasis and plasma or genital viral loads, and we found that their inclusion did not meaningfully change effect estimates (change of ≤0.02 log₁₀ copies/ml).

We used analogous statistical models with the same sets of covariates to perform subgroup analyses by schistosome species. We included indicator variables for S. mansoni, S. haematobium, and infections caused by an undetermined species (CAA ≥10 pg/ml but negative immunoblot results for both species) into a single statistical model, with uninfected individuals (CAA negative results) as the reference group. For plasma viral loads, where we had more statistical power, we also performed exploratory subgroup analyses by sex, study population, and schistosome infection intensity levels, using CAA levels to classify infections as low (10–99 pg/ml), medium (100–999 pg/ml), or high burden (≥1000 pg/ml) since CAA levels correlate with worm burden [30]. Additionally, to address the possibility that schistosome infection status changed over time, we performed a sensitivity analysis, restricting the selection criteria to include only samples collected 4–12 months after HIV-1 infection. All analyses were conducted using Stata version 13.1 (Stata Corporation, College Station, TX).

A total of 245 Kenyan and Ugandan participants from the three serodiscordant couples cohorts and 332 FSWs from the Mombasa Cohort experienced HIV-1 seroconversion, and all but two individuals had a pre-seroconversion blood sample available for schistosomiasis testing (Fig 1). From these 575 individuals, 370 (64%) had at least one plasma viral load result from the set point period obtained prior to ART initiation (Table 1). In the serodiscordant couples cohorts 21% (36/168) of HIV-1 seroconverters had schistosomiasis, compared to 23% (46/202) among members of the Mombasa Cohort. The median number of days from collection of the sample used for schistosomiasis testing to the estimated date of HIV-1 acquisition was 14 (interquartile range [IQR], -17 to 46) for the serodiscordant couples cohorts and 74 (IQR, 47 to 127) for the Mombasa Cohort; negative numbers indicate that the participant acquired HIV-1 prior to the date of schistosomiasis testing. The median number of eligible set point viral load results per individual was 4 (IQR, 3 to 6) in the serodiscordant couples cohorts and 3 (IQR, 2 to 5) in the Mombasa Cohort.

In total, there were 1,102 set point plasma viral load results from 288 HIV-1 seroconverters without schistosomiasis (mean = 4.50 log₁₀ copies/ml) and 305 results from 82 seroconverters with schistosomiasis (mean = 4.32 log₁₀ copies/ml). After controlling for age, sex, cohort, and year of HIV-1 acquisition, we found that HIV-1 seroconverters with schistosomiasis had set point plasma viral loads that were 0.17 log₁₀ copies/ml lower (95% CI -0.38 to 0.03) than seroconverters without schistosomiasis (Table 2). To address the possibility that schistosomiasis status may have changed over time, we performed a sensitivity analysis including only samples collected 4–12 months after the estimated date of HIV-1 acquisition. In this sensitivity analysis, we obtained a point estimate of similar magnitude that achieved statistical significance (-0.22 log₁₀ copies/ml, 95% CI -0.43 to -0.01). Additionally, we performed subanalyses by schistosome species (Table 2). Compared to individuals without schistosomiasis, individuals infected by S. mansoni had statistically significantly lower set point plasma viral loads (-0.34 log₁₀ copies/ml, 95% CI -0.58 to -0.09). In contrast, individuals infected by S. haematobium had higher plasma viral loads compared to uninfected individuals (+0.33 log₁₀ copies/ml, 95% CI -0.07 to 0.73), though the result did not achieve statistical significance.

Although we did not anticipate that schistosome infection intensity would impact plasma HIV viral load levels, we conducted an exploratory analysis and found that individuals with high intensity infections (CAA ≥1000 pg/ml) had lower set point plasma viral load levels (-0.44 log₁₀ copies/ml, 95% CI -0.75 to -0.13) compared to those without schistosomiasis (S1 Table). We also performed exploratory subanalyses by sex and study population (S2 Table). We found that women with schistosomiasis had a near-significant association with lower set point plasma viral load levels (-0.23 log₁₀ copies/ml, 95% CI -0.46 to 0.00) while among men with schistosomiasis we did not observe a strong association (0.04 log₁₀ copies/ml, 95% CI -0.38 to 0.46). Additionally, we found that FSWs in the Mombasa Cohort with schistosomiasis had lower set point plasma viral load levels (-0.29 log₁₀ copies/ml, 95% CI -0.55 to -0.02) compared to those without schistosomiasis (S2 Table).

Genital viral load results were only available for the Mombasa Cohort. Out of 202 women with eligible plasma viral load results, 152 had at least one eligible cervical viral load result and 153 had at least one eligible vaginal viral load result. The median number of cervical and vaginal viral load results per individual was 2 (IQR, 1 to 4).

We found that individuals with and without schistosomiasis had similar set point genital viral loads (Table 3). Individuals with schistosomiasis had cervical set point viral loads that were on average 0.07 log₁₀ copies/swab (95% CI -0.20 to 0.34) higher than uninfected individuals, adjusting for age, year of HIV-1 acquisition, and plasma viral load. Adjusting for the same covariates, individuals with schistosomiasis had vaginal set point viral loads that were 0.11 log₁₀ copies/swab (95% CI -0.17 to 0.39) higher than uninfected individuals. In a sensitivity analysis including only samples collected 4–12 months after the estimated date of HIV-1 acquisition, we still found no evidence that cervical (-0.02 log₁₀ copies/ml, 95% CI -0.35 to 0.30) or vaginal (-0.14 log₁₀ copies/ml, 95% CI -0.49 to 0.21) viral loads were associated with schistosome infection.

Since ova from S. haematobium more frequently induce genital damage than ova from S. mansoni, potentially leading to inflammation and elevated genital viral load levels, we performed subgroup analyses for each schistosome species (Table 4). After adjusting for age, year of HIV-1 acquisition, and plasma viral load levels, S. mansoni infection was not associated with a statistically significant change in cervical (0.29 log₁₀ copies/swab, 95% CI -0.04 to 0.62) or vaginal (0.16 log₁₀ copies/swab, 95% CI -0.19 to 0.52) set point viral loads. In adjusted models, S. haematobium infected individuals had lower set point cervical viral loads (-0.59 log₁₀ copies/swab, 95% CI -1.11 to -0.06) but similar set point vaginal viral loads (-0.09 log₁₀ copies/swab, 95% CI -0.65 to 0.46) compared to schistosome-uninfected individuals.