Abstract
Background Persons living with HIV (PLHIV) who are also infected with Kaposi sarcoma-associated herpesvirus (KSHV) constitute a group among the highest risk for Kaposi sarcoma (KS). As such, understanding KSHV prevalence amongst PLHIV is important for the control of KS. To date, data on KSHV prevalence amongst PLHIV in East Africa — one of the world’s hotbeds for KS — is both sparse and variable.
Methods In a cross-sectional design, we studied consecutive adult PLHIV identified just prior to starting antiretroviral therapy at an ambulatory HIV clinic in Mbarara, Uganda. Results from two enzyme immunoassays (with synthetic K8.1 and ORF 65 antigens as targets) and one immunofluorescence assay (using induced BCBL cells) to detect antibodies to KSHV were combined to classify KSHV antibody positivity.
Results We evaluated 727 PLHIV between 2005 to 2013; median age was 34 years (interquartile range (IQR): 28-40), 69% were women, and median CD4 count was 167 cells/µl (IQR: 95-260). Prevalence of KSHV antibody positivity was 42% (95% CI: 38%-46%), with little substantive change after several correction approaches, including Rogan-Gladen. Adjusted prevalence of KSHV antibody positivity was 1.6 times (95% CI: 1.3-1.9) higher in men than women; adjusted absolute prevalence difference was +0.20 (95% CI: +0.11 to +0.30). Lack of formal education (prevalence ratio=1.6 comparing no school to ≥ 4 years of secondary school; 95% CI: 1.1-2.3) was also associated with KSHV infection. We found no strong evidence for a role for age, alcohol use, or other measurements of sexual behavior, SES, or well-being in the occurrence of KSHV antibody positivity.
Conclusion Among adult PLHIV in western Uganda, KSHV prevalence is estimated at 42%, with little change after several approaches to correction for antibody detection inaccuracy. This estimate differs from several others in the region (up to 83%), highlighting need for inter-assay comparison studies using identical local specimens. To the extent HIV does not influence KSHV acquisition, the findings may also represent KSHV prevalence in the general population. The large-magnitude effect of sex and education on KSHV acquisition motivates an accelerated search for mechanisms. The sex effect, in part, may explain the higher incidence of KS among men.
Even before the AIDS epidemic, portions of East Africa had amongst the highest incidences of Kaposi sarcoma (KS) in the world.1 Widespread HIV infection in the region accelerated2-4 this already high incidence and transformed KS into a much more aggressive condition with poorer survival.5, 6 The advent of HIV-related KS also speeded the search for the long sought-after infectious agent of KS,7 culminating in the discovery of KSHV.8 Soon thereafter, it became recognized that KSHV is necessary but not sufficient for development of KS,9, 10 and HIV infection is one of the strongest complementary co-factors.7, 11 As such, persons living with HIV infection (PLHIV) form a group among highest risk for KS if they are coinfected with KSHV.12 Accordingly, knowledge of KSHV prevalence as well as determinants of KSHV infection among PLHIV represent critical yardsticks to understand the threat of KS in communities and factors to target in an attempt to mitigate its occurrence.
Initial research on KSHV prevalence in PLHIV in East Africa has featured extensive variability in terms of persons evaluated13-19 and assays13, 20, 21 used to document KSHV antibody positivity. Findings have also been variable, with prevalence estimates ranging between 29% in the rural Rakai district of Uganda22 and 83% in the Kalungu district of Uganda.14 Regarding determinants of KSHV infection among PLHIV, very little work has been performed in East Africa, with the closest analogous research emanating from other parts of Africa.23, 24
To better understand the prevalence and determinants of KSHV infection among PLHIV in East Africa, we focused on PLHIV identified just prior to starting antiretroviral therapy (ART) in a prototypic municipal HIV clinic caring for a diverse spectrum of adults in Uganda. We took advantage of a large and well-characterized research cohort, previously assembled to study clinical course following initiation of ART. In addition, to the extent that HIV infection per se does not influence KSHV prevalence,25-27 we also argue that our estimate of KSHV seroprevalence made in this representative sample of PLHIV provides insights on KSHV seroprevalence in the general adult population in this region.
Methods
Overall Design
Using a cross-sectional design, we determined the prevalence and determinants of KSHV antibody positivity among adult PLHIV in Uganda just prior to their initiation of ART.
Participants
Participants were all those with a baseline visit in the Uganda AIDS Rural Treatment Outcomes (UARTO) research cohort. UARTO was a consecutive sample of consenting ambulatory adult PLHIV (18 years or older) initiating ART at and residing within 60 kilometers of the Immune Suppression Syndrome (ISS) Clinic of the Mbarara Regional Referral Hospital in southwestern Uganda. While the clinic is located in the urban setting of Mbarara (population estimated at 82,000 in 2010), the patients cared for are derived from a wide catchment of urban, peri-urban and rural residences. Patients were deemed eligible to begin ART by local clinic providers, not affiliated with the research cohort, according to clinical indications following Ugandan Ministry of Health and WHO guidelines. In the first wave of enrollment into UARTO, patients were recruited between June 2005 and November 2008, and a second wave of enrollment occurred between June 2011 and July 2013. These periods featured CD4+ T cell- and other clinical indication-derived thresholds for ART initiation; this was prior to the current “Treat-All” era. All participants provided written informed consent.
Measurements
Questionnaire-based
Participants completed standardized interviewer-administered surveys detailing demographic characteristics, asset holdings (via the Filmer-Pritchett asset index28), frequency of hunger, HIV-related clinical history, health-related quality of life (via the Medical Outcome Study-HIV29), alcohol use (via the Alcohol Use Disorder Identification Test30), and sexual behavior.
Laboratory-based
Venous blood samples were drawn from participants prior to initiation of ART. Two enzyme immunoassays (EIA) (with synthetic K8.1 and ORF 65 antigens as target) and one indirect immunofluorescence assay (IFA) (using an induced KSHV-infected BCBL cell line expressing lytic-phase KSHV antigens) were performed at the Centers for Disease Control and Prevention to detect antibodies to KSHV.31-34 The findings from these three assays were interpreted according to a previously described algorithm, leading to a final overall result.35 In this algorithm, specimens reactive in any two tests or the immunofluorescence assay alone were classified as KSHV-antibody-positive. This algorithm has an estimated specificity of 97.5% for the diagnosis of KSHV infection, as evidenced by testing blood donors in the U.S.,36 and sensitivity of 96.3% based on testing patients with KS.37 Plasma HIV RNA level was measured using the Amplicor® HIV Monitor version 1.5 or the COBAS® TaqMan HIV-1 version 1.0 assays (Roche, Branchburg, NJ). CD4+ T cell counts were assessed using the FACSCalibur™ system (Becton Dickinson, San Jose, CA).
Literature search
To identify all published literature regarding KSHV seroprevalence in East Africa, we searched Medline (PubMed), EMBASE, SCOPUS and Web of Science from 1998–2022, according to the PRISMA guidelines, for all articles providing data on the prevalence of KSHV infection by antibody testing in Djibouti, Eritrea, Ethiopia, Kenya, Somalia, Tanzania, and Uganda. The search strategies are listed in the Supplementary Material.
Statistical Analysis
Because the measurement of KSHV antibodies is imperfect, we calculated, in addition to the base KSHV antibody positivity derived by the combined assay algorithm, a range of corrected antibody positivity values using the Rogan-Gladen approach, which takes into account presumed assay sensitivity and specificity imperfections.38, 39 We also determined overall KSHV seroprevalence using a rule in which reactivity on any individual assay equates to overall antibody positivity.
To evaluate determinants of KSHV antibody positivity, we used antibody status derived from the combination assay algorithm. We focused on three constructs as the primary proximal exposures of interest: a) non-sexual horizontal contact with other humans in childhood; b) non-sexual horizontal contact in adulthood; and c) sexual contact in adulthood. For the first two constructs, we did not have direct measurements and instead have measurements that are upstream causes of the underlying constructs but may not be highly associated with the underlying construct. For the third construct (sexual contact in adulthood), we have several direct measurements. Our conceptualization of the system was depicted in a directed acyclic graph (DAG; Figure 1). Attempts to identify the causal effects of the various exposures of interest, including which variables to control for to remove confounding, was informed by the DAG.40, 41 This process resulted in several different regression models, each focusing on a primary exposure of interest, rather than a single final model.42 Relationships between exposure variables and outcome were estimated in a causal framework with prevalence ratios (PR) and prevalence differences, which were obtained with the margins command in Stata following logistic regression.43 This command uses parametric G-computation to average across counterfactual contrasts formed for each participant in the study population.44 In the regression models, continuous exposure variables were specified as restricted cubic splines. When a continuous variable was the primary exposure of interest, our measures of association contrasted participants who had values at the 75th percentile of the distribution to participants with values at the 25th percentile. The regression models were assessed for influential observations by comparing the coefficient value when an observation is included versus the coefficient when the same observation is excluded.45 For goodness-of-fit, the Hosmer-Lemeshow test was used.46 All analyses were performed in Stata (version 17; Stata Corporation, College Station, Texas).
A directed acyclic graph (DAG) depicting our hypothesized conception of the system under investigation. We seek to estimate the causal contribution of age, biologic sex, alcohol use, socioeconomic status (SES; proxy measured by education, asset index, and frequency of hunger), well-being (proxy measured by a mental health quality-of-life index, physical health quality-of-life index; peripheral CD4+ T cell count; and history of major clinical co-morbidity in the form of tuberculosis and cryptococcosis), and sexual contact with humans during adulthood (measured by age of sexual debut, lifetime number of penile-vaginal sexual partners, circumcision history, and number of prior 3 month penile-vaginal partners) on the occurrence of KSHV infection (as measured by KSHV antibody positivity) in HIV-infected adults in Uganda just prior to initiation of antiretroviral therapy. The DAG was used to guide which variables to control for when assessing the independent contribution of the various constructs. In this rendition of the DAG, as depicted by the “?”, we are focusing on investigating the relationship between sexual contact with other humans in adulthood and KSHV infection. Other perspectives on this DAG can be used to investigate potential causal roles for the other variables (e.g., alcohol use).
Results
Characteristics of the Study Population
To investigate the prevalence and determinants of KSHV antibody positivity, we evaluated 727 ambulatory adult PLHIV initiating ART in Uganda. Median age was 34 years (interquartile range (IQR): 28 to 40), 69% were women, and 92% were ever married (Table 1). A median monthly income of $14.60 U.S. dollars (IQR: $5.90 to $41) and high frequency of hunger were evidence of the impoverished context of the region. The median CD4+ T cell count was 167 cells/µl (IQR: 95 to 260), and 100% of population had detectable plasma HIV RNA.
Characteristics of HIV-infected adults in southwestern Uganda just prior to initiation of antiretroviral therapy.
Prevalence of KSHV Antibody Positivity
When evaluating the three antibody assays in isolation, reactivity was higher in the IFA (41%) and ORF K8.1 EIA (40%) compared to the ORF 65 EIA (25%) (p-value <0.001 for both using McNemar’s test) (Table 2). Although both yielded approximately 40% reactivity, the IFA and ORF K8.1 EIA were not fully concordant. A total of 247 (34%) participants were reactive on both assays, 364 (50%) were non-reactive on both; 38 (5.3%) reactive on the IFA but non-reactive on the ORF K8.1; 39 (5.4%) were reactive on the ORF K8.1 but not IFA; and 39 (5.4%) had a variety of combinations including an indeterminate value on one of the two assays. When evaluated in the laboratory’s previously established algorithm that combines results across all three assays, overall KSHV antibody positivity was 303/727 (42%; 95% CI: 38% to 45%). Among these 303 participants, KSHV antibody positivity was achieved by reactivity on all three assays in 142 (47%), reactivity in the IFA and one of the two EIAs in 112 (37%), reactivity in the IFA alone in 43 (14%), and reactivity in the two EIAs but not in the IFA in 6 (2%).
Reactivity on three individual assays for detection of antibodies to KSHV (A) and patterns of reactivity (B) among HIV-infected adults in southwestern Uganda just prior to initiation of antiretroviral therapy.
Because the current state of the field in KSHV antibody detection does not feature a comprehensive spectrum of true positive and true negative gold standards, all estimated values of sensitivity and specificity from individual assays or combined assays/algorithms cannot be considered definitive. To address this, we used the approach of Rogan-Gladen to estimate the expected prevalence of KSHV antibody positivity after correcting for a range of assay sensitivity and specificity imperfections. Using the previously estimated 96.3% sensitivity and 97.5% specificity for the combined algorithm, the Rogan-Gladen correction was nearly unchanged, compared to the raw prevalence, at 42% (95% CI: 38% to 46%). When we concurrently varied estimates of sensitivity and specificity from 85% to 100%, we derived a range of corrected estimates from 31% (95% CI: 24% to 38%) to 49% (95% CI: 43% to 55%) (Figure 2). Finally, using a rule in which reactivity on any individual assay equates to overall antibody positivity, 371 (51%) were KSHV-antibody-positive.
Contour plot depicting the range of corrected estimates, according to sensitivity and specificity of the testing algorithm, of the prevalence of KSHV antibody positivity among HIV-infected adults in southwestern Uganda just prior to initiation of antiretroviral therapy. The corrected estimates are calculated using the Rogan-Gladen formula, which combines the observed prevalence (0.42) and an assumed sensitivity and specificity of the testing algorithm to obtain the true prevalence, sampling error notwithstanding.
Determinants of KSHV Antibody Positivity
Informed by our DAG, we evaluated several potential causal determinants of KSHV antibody positivity. Biological sex and education exhibited the largest magnitudes of effect (Table 3). After adjustment for several factors mediating indirect paths, the prevalence of KSHV antibody positivity in men was 1.6 times (95% CI: 1.3 to 1.9) higher than in women. This equates to an absolute adjusted prevalence difference of +0.20 (95% CI: +0.11 to +0.30). Lack of formal education (PR=1.6, comparing no school to ≥ 4 years of secondary school; 95% CI: 1.1-2.3, p=0.004 for trend) was associated with higher KSHV infection, equating to a prevalence difference of +0.17 (95% CI: +0.040 to +0.31). Older age at sexual debut (PR = 1.2, comparing 20 years to 16 years; 95% CI: 0.99 to1.3) and higher physical health score (a quality-of-life metric from the Medical Outcomes Study; PR = 1.3 comparing scores at the 75th to 25th percentile; 95% CI: 1.03 to 1.5) were also independently associated with higher KSHV prevalence but at lower magnitude. We found no strong evidence for a role for age, alcohol use, or other measurements of sexual behavior, socioeconomic status, or well-being in the occurrence of KSHV antibody positivity.
Relationship between a variety of participant characteristics (exposures) and KSHV antibody positivity (outcome) among HIV-infected adults in Uganda just prior to initiation of antiretroviral therapy.
Discussion
The advent of the HIV pandemic transformed KS from a rare medical curiosity to a multicontinent public health problem. Forty years later, despite successes in the control of HIV infection, persons living with HIV infection still represent the group at highest risk for KS. Many areas in sub-Saharan Africa have unfortunately lagged behind much of the world in achieving population-wide treatment of HIV infection, leaving the region amongst the world’s hotspots of KS.47-50 As such, to understand the threat of KS in Africa, it is relevant to focus on PLHIV. To address this, we measured the seroprevalence and determinants of the necessary causal agent of KS — KSHV — in a large and well-characterized group of adult PLHIV whose consecutive sampling in a municipal setting bolsters their representativeness of the larger target population. In a measurement domain that has been plagued by lack of agreement,51-53 we used a suite of time-tested KSHV antibody assays whose combination performance has been exemplary in available gold standard reference populations. We found a high prevalence of KSHV antibody positivity in this population — 42% — and applied various correction and sensitivity approaches to yield a range between 30% and 51%. We also found that biologic sex and education level attained had the largest-magnitude influence on KSHV seroprevalence.
Our finding of 42% KSHV antibody positivity in Uganda among adult PLHIV adds another estimate to an already heterogeneous lot. A systematic review of published literature from 7 East African countries (Djibouti, Eritrea, Ethiopia, Kenya, Somalia, Tanzania, and Uganda) reveals 9 other studies yielding a range of KSHV antibody positivity between 29% and 83% among adult PLHIV (Table 4). While not a probability sample of the source population, we believe that our sample of consecutive patients about to initiate ART at a prototypical municipal clinic is closely representative of the community population of adult PLHIV. Several of the prior studies feature equally convincing representative samples, including a door-to-door sample of households,13, 14, 22, 54 a probability sample of an entire country;55 and women attending antenatal clinic.18, 20 Other than geographic location, the main difference, in terms of important known determinants for KSHV infection, between prior studies and ours is the fraction of women. Accounting for sex, however, does not alter the heterogeneity. Among studies from East Africa in which the prevalence of KSHV antibody positivity in women could be isolated, values range from 36% and 70%. Of note, despite being just ∼160 kilometers (∼100 miles) apart and having nearly the same sex distribution, the prevalence of KSHV antibody positivity in our sample differed by an absolute -41% compared to Nalwoga et al.20
Description of the study population, method of detection of antibodies to KSHV, and prevalence of KSHV antibody positivity among studies, including the present work, conducted in East African populations and identified in a systematic review of the published literature regarding KSHV antibody positivity among HIV-infected adults.
Aside from geographic differences across studies, the other difference is the method of antibody detection. First-generation assays for KSHV antibodies had a well-chronicled history of disagreement.53 Many of these assays have ceased being performed, but none of the surviving assays, which are all individual laboratory-derived “in house” protocols, have reached commercialization and accompanying high-fidelity standardization. Furthermore, there exists no study of agreement between the contemporary assays. Including the present study, seven different KSHV antibody detection platforms/algorithms are represented in the ten reports from East Africa to date (Table 4). The assay used by Wawer et al. featured an IFA targeted only at antibodies against KSHV latency-associated antigens (and a less sensitive lytic antigen); this may explain the low KSHV seroprevalence (29%) in this work.22 This explanation cannot, however, account for the difference between our work and all of the other community-based studies given that each study was nominally able to detect antibodies against lytic-phase KSHV antigens. Of note, our IFA, in addition to detecting antibodies to lytic-phase antigens, can also detect participants who solely have antibodies to latency-associated antigens. When we corrected our KSHV antibody positivity estimates to account for a range of assay sensitivity/specificity imperfection as low as 85% or altered our rule to allow reactivity on any single assay to count as overall positivity, our upper bound estimate still only reached 51%. Thus, the difference between our seroprevalence finding and the others is not readily explained by moderate differences in sensitivity or how we constructed our combined assay algorithm. In summary, the different antibody assays used across studies makes formal comparison difficult.
Biologic sex and level of education were the largest-magnitude determinants of KSHV antibody positivity. Higher prevalence of KSHV infection among men has been observed in several other studies from East Africa13, 55-58 but not all.22, 54, 59 The absolute difference we estimated (+20% among men), a value that is adjusted for many factors and represents a marginal measure of association, is the largest estimate of sex-derived differential for KSHV transmission of which we are aware in East Africa. The mechanism underlying the effect of male sex, which likely occurs in large part prior to puberty, is not understood but possibilities include sex-related differences in innate immune response60, 61 and behavioral exposures. A protective effect of higher education has also been observed in several other studies from East Africa16, 18, 24, 55, 57 but not all.22, 62 We speculate that education is a marker of socioeconomic status which influences some heretofore undocumented behavioral practice that transmits KSHV (e.g., some practice in which saliva is spread63). Regarding a role for sexual practices in the transmission of KSHV, the only association we found was a paradoxical higher KSHV seroprevalence with older age of sexual debut. We are unable to explain this low-magnitude association, but it fails to support sexual transmission of KSHV among adults. A lack of evidence for sexual transmission of KSHV has also been concluded in several other studies from this region.13, 22, 24, 54
Several limitations merit mention. First, the facility-based nature of our sampling precludes our ability to claim a pristinely representative sample of adult PLHIV. Specifically, we missed those who became HIV-infected but did not seek care, and, to the extent this is related to sex or education, our seroprevalence estimate may be biased. Second, regarding the measurement of KSHV antibodies, we earlier demonstrated how our antibody positivity estimate would be altered if assay sensitivity and specificity were as low as 85%. It is conceivable assay accuracy could even be lower but proving this will require a transformative finding in the field regarding assay gold standard construction. Finally, regarding confounding, residual confounding (particularly unmeasured) is a threat to some of our weak-magnitude analytic findings (e.g., older age at sexual debut). Notably, because sex is genetic in origin, there are few proximal causes, and, as depicted by our DAG, confounding of any relationships influenced by sex is highly unlikely.
There are several implications of our findings. The most prominent stems from the sizeable difference between our seroprevalence estimate (42%) and several others (up to 83%; Table 4) in the region. This disparity is explained either by differences in antibody assay performance or in the distribution of some very influential causative factor(s) for KSHV transmission. Resolving this will be critical and informative for the field. Inter-assay comparison studies using identical local specimens from East Africa are the requisite next step. Another implication is that, to the extent HIV infection does not influence KSHV acquisition (as most data indicate),18, 21, 27, 54, 64 our prevalence findings may, after standardization for sex distribution, also depict KSHV prevalence in the general population of East Africa. In turn, serial monitoring of KSHV seroprevalence in HIV-infected clinic populations could be an efficient way to monitor general population trends. In terms of mechanisms for KSHV transmission, our findings provide further evidence against any substantial magnitude of transmission in adulthood (including via sexual contact) in Africa. This is in contrast to the sexual transmission of KSHV that occurs among adults (men who have sex with men) in the U.S., Europe, and Australia.65-70 Instead of sexual transmission in adulthood, attention should be focused on KSHV transmission in childhood, particularly the mechanism underlying higher risk among boys. Finally, higher KSHV seroprevalence in men may explain, in part, the higher incidence of KS among men in Africa, especially HIV-uninfected men.1, 71-73
In summary, among adult PLHIV in western Uganda, we found that KSHV seroprevalence is ∼40%, with little change in estimate after various correction approaches accommodating uncertainty in antibody assay accuracy. Biological sex and education are the largest-magnitude determinants of KSHV infection; these findings motivate continued research regarding mechanism. The discrepancy between our KSHV seroprevalence and others in the region calls for formal inter-assay comparison studies using identical local East African specimens.
Data Availability
All data produced in the present study are available upon reasonable request to the authors.
Footnotes
Funding: National Institutes of Health (R01 MH054907, U54 CA190153, U54 CA254571, K43 TW011987 and P30 AI027763)
References
