African youth are disproportionately affected by the global HIV pandemic. Of 10 million HIV-infected persons aged 15-24 years, 62% live in sub-Saharan Africa; the majority are female. [1] The 2007 Kenya Aids Indicator Survey (KAIS) demonstrated a national HIV prevalence of 7.4% among Kenyans aged 15-49 years; 8.8% in females and 5.5% in males. [2] Among adolescents aged 15-19 years, HIV prevalence was 3.5% in females and 1.0% in males, increasing to 7.4% and 1.9% respectively among those aged 20-24 years. [2] Nyanza Province in western Kenya continues with Kenya's highest HIV prevalence of 15.0%; 17.6% in women and 11.4% in men. [2] The KAIS HIV prevalence data varied slightly, a non-statistically significant difference, from the 2003 Kenya Demographic Health Survey (KDHS). [3]

In Nyanza Province, the Kenya Medical Research Institute (KEMRI) and the U.S. Centers for Disease Control and Prevention (CDC) have conducted infectious disease research since 1988 in Asembo, a rural community. [4] An ongoing demographic surveillance system (DSS) in this area reports a life expectancy at birth of 38 years and high adult mortality rates likely attributable to AIDS. [5] In 2002 the Institute of Tropical Medicine in Antwerp, Belgium (ITM), with KEMRI/CDC, initiated a multi-component adolescent prevention intervention in Asembo. This intervention was in response to a needs assessment, conducted by ITM and KEMRI/CDC among Nyanza youth, and to findings from a 1997-1998 study [6] demonstrating an HIV prevalence of 23% and 3.5% among females and males aged 15-19 years, respectively, in Kisumu, Nyanza's capital. [7] Little was known about HIV epidemiology in rural areas, where the majority of Nyanza's population resides.

In 2003, KEMRI, CDC, ITM, and the London School of Hygiene and Tropical Medicine (LSHTM) initiated the first HIV-specific study in Asembo to measure the prevalence of HIV and sexually transmitted infections (STIs), identify factors associated with HIV infection, provide a baseline to evaluate a new adolescent prevention intervention, and inform sample size calculations for a prospective HIV-incidence cohort. We present HIV and HSV-2 prevalence and associated factors for 13-34 year old Asembo residents.

Of 3960 randomly selected individuals, 2606 (66%) were eligible for enrollment; almost all others, 1261/1354 (93%), were ineligible because they had left Asembo. Of the eligible 2606, 1822 (70%) enrolled, 447 (17%) refused, and 337 (13%) consented at home but did not come to the study site. Enrollment rates were similar for males and females (70%) but higher among those aged <20 years (75%) than those aged ≥20 years (63%, p<0.0001). Primary reasons for refusal included not wanting blood taken (37%), partner/parental objection (19%), and not wanting to be tested for or learn one's HIV status (18%).

Among 13-19 year olds in the sample, 43.9% of females and 50.2% of males reported they had ever had sexual intercourse. The median age of sexual debut was 16.5 years for females and 15.5 years for males. The median age of first pregnancy was 18 years. The median age of first marriage was 18 years for females and 23 years for males.

Among currently married participants in the sample, 72/328 (22.0%) of females and 8/189 (4.2%) of males, were in polygamous marriages. Additionally, 63.9% (39/61) of widows had been inherited, and 2.8% (6/214) of ever-married men had inherited a widow. Of sexually active participants, 102/1202 (8.4%) had participated in ritual sex. Only 13.2% (77/583) of sexually active males were circumcised. There was high mobility with only 18% of female and 48% of male sexually active participants having lived their entire lives in Asembo.

The median number of lifetime sexual partners for sexually active participants in our sample was 3 for females and 4 for males. Over half of sexually active females, 327/619 (52.8%), had sex in exchange for gifts; 45/619 (7.3%) reported having ever been forced to have sex. Condom use during the last sexual intercourse was equally low between never-married males and females (25%). Among ever-married participants, condom use during the last sexual intercourse with spousal or non-spousal partners was rare in females (3.3%) and males (6.9%). Of currently married individuals, 3.7% (12/328) of females and 23.8% (45/189) of males reported having had sex with a non-spousal partner in the previous six months.

The population, HIV prevalence for the study area, based on 1762 participants who provided blood specimens and complete data, weighted by age group and sex was 15.4% overall, 20.5% among females, and 10.2% among males. Table 1 presents HIV prevalence of all study participants by demographic characteristics. Females became infected several years younger than males, and HIV prevalence was higher among females than males until the fourth decade. (Figure 1) Weighted HSV-2 prevalence was 40.0% overall, 53.0% among females, 25.8% among males, and remained higher in females for all age groups. (Figure 2) Although the study was not powered for single year of age-specific estimates, HIV prevalence in adolescent females increased from 3.1% [95% Confidence Intervals (CI), 0.0, 6.3] at 16 years to 12.8% (95% CI, 5.7, 19.9) at age 17.

Of 309 females who denied ever having had sexual intercourse in our sample, 4 (1.3%) were HIV-positive and 5 (1.6%) were pregnant; 2/249 (0.8%) males who denied having had sexual intercourse were HIV-positive.

After adjustment for age group and stratification by sex, several demographic and risk variables were associated with HIV prevalence among sexually active participants (Table 2). Among sexually active females, HIV infection was significantly associated with the following factors: cash income through regular/intermittent employment, being married or widowed, higher number of lifetime sexual partners, ritual sex, having received an injection in the previous six months, scarification, HSV-2 infection, and past treatment for an STI. In multivariate analysis, HIV infection remained significantly associated with the following factors: older age, cash income, higher number of lifetime sexual partners, having received an injection in the prior six months, and scarification. Adding STI-related covariates to the model attenuated the positive association of HIV infection with age and lifetime number of sexual partners and rendered non-significant the positive association between HIV infection and scarification or prior STI treatment. HSV-2 infection remained strongly associated with HIV infection.

Among sexually active males (Table 2), factors significantly associated with HIV infection included the following: being married or a widower, higher number of lifetime sexual partners, having received an injection in the previous six months, HSV-2 infection, and previous STI treatment. In multivariate analysis, older age and higher number of lifetime sexual partners remained significantly associated with HIV infection. STI-related covariates in the model attenuated the association of both age and number of sexual partners with HIV infection, but HSV-2 infection and prior STI treatment remained associated with HIV infection.

Marital status was an effect modifier necessitating separate analyses by gender and marital history. Because few never-married individuals were HIV-infected, we were unable to conduct adequate multivariate analyses in this group. For ever-married participants, univariate analyses of factors associated with HIV infection are presented in Table 3 and multivariate analyses in Table 4. In multivariate analyses of ever-married females, widowhood, cash income, having received an injection in the prior six months, and HSV-2 infection were significantly associated with HIV infection. For ever-married males, a higher number of lifetime sexual partners and HSV-2 infection was significantly associated with HIV infection.

Limited data were available on circumstances around injections. Of 458 sexually active individuals who reported receiving an injection in the previous six months in our sample, 87% received it from clinicians, 9% from community health workers, and 4% from traditional healers/herbalists. Those aged 30-34 years comprised the highest proportion receiving an injection (42.1%); 15-19 year olds were the lowest (35.2%).

This population-based study in rural Nyanza Province, Kenya, found a high prevalence of HIV infection among those aged 13-34 years. Despite a decreasing trend in Kenya's national HIV prevalence, [18] similarly high rates were reported in the 2003 KDHS for Nyanza Province [3] and the Kenya AIDS Indicator Survey 2007. [2] In our study, HIV prevalence among females aged 15-19 years was 12 times higher than their male counterparts. This substantial gender disparity is found in Kenya nationwide [3], [19] and other African countries. [6], [18], [20] Social [21]-[24] and biological factors may explain this difference. [25] Young women may have sex with older, more sexually experienced men, who are more likely to be HIV-infected. [26] Young women in our study, however, described most of their sexual experiences to be with peers. Having a first sexual partner much older in age was not significantly associated with HIV infection. [27] Biological factors including immature genital tracts, cervical ectopy, [28] and more efficient male-to-female HIV transmission [29] likely contribute to the vulnerability of young women to HIV infection.

HSV-2 may also play a role. The strong association we found between HIV and HSV-2 infection confirms findings from other studies. [7], [30]-[37] A meta-analysis of longitudinal studies with established temporal relationships between HIV and HSV-2 acquisition found that HSV-2 infection increases the risk for HIV acquisition three-fold. [31] HSV-2 infection could be a driving force for HIV transmission in our study population.

Including HSV-2 infection in the multivariate model attenuated the positive association between HIV and a higher lifetime number of sexual partners. Although under-reporting of lifetime number of sexual partners is a possible explanation, HSV-2 infection, particularly in young people, may be a more accurate marker of sexual activity than reported sexual behavior. [7] Because HSV-2 infection is such a strong co-factor in sexual transmission of HIV, differences in lifetime number of sexual partners, depending on HIV prevalence among such partners, may translate into relatively small differences in risk for HIV infection in the presence of HSV-2 infection.

Marriage has been associated with risk for HIV infection in sub-Saharan Africa. [3], [7] In our study, widowhood was strongly associated with HIV infection, presumably because many widows lost their husbands to AIDS. Being currently married was significantly associated with HIV infection. HIV prevention efforts must target married couples. The 2003 KDHS found 17.1% of couples in Nyanza Province to be HIV discordant and 9.8% concordantly HIV-infected. [3] The low frequency of condom use within marriage and the high frequency of extramarital sex among men in our study suggest that a substantial proportion of married people in this community will be exposed to HIV infection through unprotected sexual intercourse with their HIV-infected spouse.

The potential contribution of unsafe injections to HIV transmission has been raised [38]-[42], disputed [43]-[45], and continues to be examined. [46] In our study, having received an injection in the prior six months was associated with HIV infection for sexually active or ever-married females but not for males or never-married participants. Additionally, for all participants there was no association between having received an injection in the prior six months and having ever had STI treatment. Given this study's cross-sectional design, this finding may represent reverse causality or confounding for which we could not control. Further investigation through a prospective incidence cohort is warranted.

HIV infection in Africa has been associated with urban-to-rural and intra-rural human mobility. [47]-[49] Females in our study reported more mobility than males; HIV prevalence was lower among females and males who had lived their entire lives in Asembo than those who had spent more than one night away from home in the previous six months. Mobility may be a risk factor for HIV infection. The 2002 DSS data for Asembo showed substantial migration. In-migration exceeded out-migration, adults younger than 30 years were the most migratory, and women migrate more than men. [5] Although primary reasons for migration included marriage and seeking employment, health-related reasons cannot be ruled out. [5] Further investigation is warranted to better elucidate the roles of mobility and migration in Asembo's HIV epidemic.

Recent randomized clinical trials of male circumcision resulted in an approximate reduction of HIV acquisition by half. [15]-[17] Luo males are not routinely circumcised, [3], [50] which could explain the higher HIV prevalence in Asembo compared to non-Luo areas of Nyanza Province and Kenya. [3], [51]

Ritual sex around funerals, polygamy, and widow inheritance are traditional Luo practices that can facilitate HIV transmission. [11]-[14] The low rates of reported ritual sex participation in this study may reflect a decreasing prevalence of such practices or underestimate their relevance, due to the young study population. Traditionally, widow inheritance requires a widow to become the wife of her late husband's brother, assuring her and her children economic and social stability. [11]-[14] Although we lacked sufficient power to detect associations between HIV infection and polygamy, widow inheritance, or ritual sex, the strong association between HIV infection and widowhood and the high prevalence of inheritance among widows in our study suggest that widow inheritance could play a role in this epidemic. [11]-[14]

Other data regarding HIV prevalence in Nyanza Province include a second cross-sectional study with different sampling methodology and age distribution was conducted from 2004-2005 in the neighboring community of Gem. This study of volunteers aged 15-34 years showed a weighted HIV prevalence of 28.2% among females and 14.8% among males. [52] Cohen et al. present HIV prevalence data from a 2006 general population survey in Kisumu city, the capital of Nyanza Province, conducted to evaluate antiretroviral therapy-related attitudes and beliefs. Although an urban setting, they found a similar over HIV prevalence of 25% in women and 16% in men aged 15-49 years. [53]

Study limitations exist. Risk factor data in cross-sectional studies are associations and cannot identify causality. There is potential for selection bias from migration, because 30% of selected individuals had moved out of Asembo between the DSS survey round and study enrollment (<8 months). Ill individuals may have left the area seeking health care or returned home to die. Healthy individuals may have left seeking employment or returned to care for ill relatives. By requiring participants to visit a nearby site, we may have selected for healthier individuals. Since study participation took up to four hours, we may have selected for individuals who were wealthier, unemployed, or had time. As the first HIV-specific research conducted in Asembo, individuals with self-perceived low HIV risk may have self-selected to participate. Most of these potential biases would have resulted in lower HIV prevalence estimates. However, if the individuals aged ≥20 years who refused study participation did so because they engaged in risky sexual behavior and did not want to know their HIV status, their non-participation might have resulted in an underestimation of HIV prevalence.

Recall bias due to illness, gender, or age can occur when describing sexual histories or risk behaviors. Sexual behavior interviews are limited by the accuracy of self-reporting, and women, particularly adolescent females, frequently underreport sexual experiences. [25]