A serosurvey was conducted in a random sample of 259 women and 231 men in 12 rural communities in Mwanza Region, Tanzania, using a type-specific ELISA for Herpes simplex virus type 2 (HSV-2) infection. Seroprevalence rose steeply with age to ∼75% in women ≥25 years old and 60% in men ≥30. After adjusting for age and residence, HSV-2 prevalence was higher in women who were married, in a polygamous marriage, Treponema pallidum hemagglutination assay (TPHA)-positive, had more lifetime sex partners, or who had not traveled. Prevalence was higher in men who were married, had lived elsewhere, had more lifetime partners, had used condoms, or were TPHA-positive. HSV-2 infection was significantly associated with recent history of genital ulcer. The association between HSV-2 infection and lifetime sex partners was strongest in those < 25 years old in both sexes. This association supports the use of HSV-2 serology as a marker of risk behavior in this population, particularly among young people.
By the end of 1997, an estimated 30.6 million adults and children worldwide were living with human immunodeficiency virus (HIV) infection or AIDS, 20.8 million of them in sub-Saharan Africa , where heterosexual intercourse is the main mode of transmission. In view of the HIV epidemic and the great burden of ill health caused by other sexually transmitted diseases (STDs) in the region, the study of sexual behavior, and the design, evaluation, and implementation of interventions to decrease sexual risk behavior have assumed increasing importance. However, the study of sexual behavior is problematic. By its nature it cannot be directly observed, and consequently only indirect information may be obtained from questionnaires, interviews, focus group discussions, and other qualitative methods. Due to the sensitive nature of the issues raised, these methods are subject to considerable bias and can be difficult to reproduce [2–4]. An inexpensive and reliable biologic marker of sexual activity would be an invaluable tool in HIV and STD research.
Herpes simplex virus (HSV) type 2 infection is the major cause of genital herpes worldwide . Because it is almost exclusively sexually transmitted and leads to the production of lifelong antibodies and because seropositivity is associated with higher risk sexual behavior in selected populations [6, 7], it has been suggested that HSV-2 antibody could be used as a biologic marker of risk behavior . HSV-2 is also of interest in sub-Saharan Africa for other reasons: It increases the risk of HIV transmission and acquisition , and it is increasingly recognized that its role in the etiology of genital ulceration in African populations has been underestimated .
Most previous studies of prevalence and risk factors for HSV-2 infection have been done in selected groups, such as antenatal or STD clinic attenders and blood donors in industrialized countries [6, 7, 11]. Few population-based surveys have been reported [12, 13] and only one in Africa . We are not aware of any studies that have examined risk factors for HSV-2 infection in a representative sample of an African population.
Research on the epidemiology of HSV-2 in sub-Saharan Africa has been hampered by deficiencies in diagnostic technology. Genital herpes is characterized by long periods of latency with episodic symptoms. Because culture and antigen detection techniques are insensitive in the absence of ulcers [15, 16], they cannot be used for large-scale epidemiologic studies. In addition, HSV-2 shares extensive sequence homology with HSV-1. This has caused a high level of cross-reactivity and made typespecific antibody determinations unreliable . There is no known cross-reactivity between the glycoprotein G (gG) in HSV-1 (gG-1) and that in HSV-2 (gG-2) . Sensitive typespecific Western blot assays (WBA) developed over the last decade depend largely on reactivity to gG-2 for the identification of HSV-2 antibody . However, WBA is expensive, cumbersome, not widely available, and difficult to reproduce.
Slomka et al.  developed monoclonal antibody (MAb)—blocking RIAs that detect type-specific HSV-1 and HSV-2 antibodies. The HSV-2 antibody assay was validated against WBA using sera from culture-proven recurrent genital HSV-2 cases and had equivalent sensitivity (96%). Furthermore, no HSV-2 antibodies were detected in patients with first-episode genital HSV-1 or in children, indicating high specificity. The assay has been converted to an ELISA format with similar test characteristics. The HSV-2 antibody ELISA, when validated against the Chiron assay, had similar sensitivity and specificity (94% and 92%, respectively) . When compared with the WBA on 473 sera collected in rural Uganda, sensitivity was 91% and specificity 93%, demonstrating its suitability for use in seroepidemiologic studies (unpublished data).
A recent community-randomized trial conducted in rural Tanzania afforded the opportunity to examine the seroepidemiology of HSV-2 in a representative sample of an African population using the novel ELISA described. The extensive data provided by a detailed behavioral questionnaire enabled us to examine associations between HSV-2 and sexual risk behavior and to assess the role of HSV-2 serology as a marker of sexual risk behavior. The specific objectives were to measure the seroprevalence of HSV-2 infection in this population by age and sex and to examine associations between HSV-2 infection and reported sociodemographic and behavioral factors.
Selection of study subjects
In order to evaluate the effect of improved STD treatment services on HIV incidence in a rural African population, randomly selected persons were recruited in Mwanza Region, Tanzania, and followed for 2 years. The study population, design, and methodology of the trial are described elsewhere . In brief, 12 communities were randomly assigned to intervention or comparison groups. The effect of the intervention was measured in a cohort of ∼1000 adults, aged 15–54 years, selected from each community by random cluster sampling. Sera were taken from the cohort at baseline and again after 2 years, and the incidence of HIV infection was compared between the intervention and comparison groups .
A random subsample from the trial cohort (1500 adults) was selected for a detailed behavioral survey  and served as the controls for a case-control study of risk factors for HIV infection . The homes of all selected subjects were visited. Subjects were interviewed concerning demographic characteristics and behavioral risk factors using a structured questionnaire. The questionnaire was designed in English, translated into Kiswahili, back-translated into English, and pretested in a pilot study. Of subjects eligible for the study, 74% participated. Interviews were conducted in May and June 1993, 6–17 months after the baseline survey.
The random subsample of 1500 subjects formed the sampling frame for the present study. Rates of HSV-2 infection change most rapidly among adolescents and young adults. Consequently, the selection of subjects was stratified by age at baseline survey and weighted to maximize the number of young people. All 290 subjects aged 15–19 years, and simple random samples of 100 subjects aged 20–24, 50 subjects aged 25–34, and 50 subjects ≥35 were selected for the present study. Sociodemographic and behavioral data were taken from the behavioral study questionnaire, and data on reported STD syndromes from the baseline survey of the cohort. Syphilis and HSV-2 serology were done on sera obtained at the baseline survey.
Syphilis serology at baseline was measured by the Treponema pallidum hemagglutination assay (TPHA; Fujirebio, Tokyo).
Type-specific antibodies to HSV-2 were determined using a MAb-blocking ELISA. This is a direct modification of the validated MAb-blocking RIA ; full details of the assay will be reported elsewhere. In brief, Greiner microtiter plates were coated with HSV-2—infected cell lysate at 1:25 dilution in PBS overnight at 4°C. The coated plates were then incubated successively with detergent (1.5% Triton X-100 and 0.5% Nonidet P40 in PBS) for 30 min at room temperature, with 10% fetal calf serum in PBS for 2 h at 37°C, with a 1:4 dilution of test serum for 1 h at room temperature, with a 1:1600 dilution of HSV-2—specific MAb and a 1:1000 dilution of peroxidase-conjugated anti-mouse antibody. Plates were washed three times between each stage using 0.05% PBS-Tween. Finally, tetramethylbenzadene was added to each well, and the reaction was stopped as appropriate by the addition of 2 M H2SO4. The absorbance of each well was measured at 450 nm, and the percentage blocking of each serum specimen was calculated by comparison with diluent control and the mean absorbance of four wells containing positive control sera.
The positive cutoff was determined by fitting a mixture model to the bimodal distribution of percentage blockings obtained as previously described . With the cutoff set at 30% blocking, this gave estimated false-negative and false-positive rates of 0.4%. Fifty sera were further tested by in-house WBA to confirm the sensitivity and specificity of the testing strategy.
HSV-2 seroprevalence by age and sex was estimated using all selected subjects. The distributions of sociodemographic and behavioral risk factors were examined by age and sex in the 73% of selected subjects who participated in the behavioral survey. Because the distribution of risk factors varied between women and men and because gender was strongly associated with risk of infection, data for women and men were analyzed separately. Odds ratios (ORs) were estimated for each risk factor and were adjusted for age in years and residence stratum (rural, island, lake shore, or roadside) using logistic regression. For significant variables and other behavioral factors of interest, ORs were further adjusted for variables significantly associated with HSV-2 infection on univariate analysis. Since HSV-2 infection in young people is more likely to reflect recent sexual exposure, a further analysis was conducted to compare the association of seropositivity and reported numbers of partners in those aged < 25 and ≥25 years. Statistical significance of ORs was assessed using the likelihood ratio test. The analysis was performed using STATA software (Stata, East College Station, TX).
Prevalence by age and sex
Of 490 subjects selected, 259 (53%) were women and 231 (47%) were men. Eighty percent of the subjects were < 25 years old. There was no difference in age distribution between men and women (median =18 and 19 years, respectively). Overall, 153 subjects (31%) were HSV-2—seropositive. Seroprevalence by age and sex is shown in figure 1 and table 1. The prevalence of infection rose steeply with increasing age to a plateau of ∼75% at age > -25 in women (P< .001 for trend) and ∼60% at age ≥30 years in men (P < .001 for trend). Women had a significantly higher prevalence of HSV-2 infection than men (age-adjusted OR = 3.67; 95% confidence interval [CI], 2.31–5.82). There were no HSV-2—positive males < 17 years old (0/52), whereas 10 of 51 women in this age group were positive. Further analysis in men was restricted to those ≥17. There was no significant difference in HSV-2 prevalence between residence strata (P = .95 adjusted for age and sex; data not shown).
Sociodemographic risk factors
Sociodemographic and behavioral data were available for the 356 study subjects (73%) who participated in the behavioral survey: 180 women (51%) and 176 men (49%). Only 133 men were ≥17 years old. There was no difference in age between respondents and nonrespondents (median for both age groups, 19 years). Detailed information on participation in the behavioral survey is available elsewhere ; however, only 0.5% of those approached refused to be interviewed, and the main reasons for nonparticipation were change of residence or travel outside the community.
HSV-2 prevalence was lower in those with more years of education (grade 4 or more) in both women and men (table 2). Age adjustment removed this effect in women and reversed it in men, although in neither case was it statistically significant. For women, there was an interaction between education and age (P = .047). In women ≥20 years old, the prevalence of HSV-2 infection was lower among those reaching grade 4 (OR = 0.38; 95% CI, 0.15–0.97; P = .04 adjusted for residence stratum). However, there was no significant effect of education among women < 20 years old (OR = 1.63; 95% CI, 0.48–5.50; P = .42).
Most women (93%) and men (75%) identified themselves as farmers (table 2). Occupation was not significantly associated with HSV-2 prevalence in women (P = .79) or men (P = .89).
About a quarter of women and men reported living somewhere other than their present village at some time during the past 5 years. More men (57%) reported traveling away from their village during the past year than women (34%) (table 2). After adjustment for age and residence stratum, the prevalence of HSV-2 was much higher in men who had lived elsewhere during the past 5 years, and this association was highly significant (OR = 3.96; P = .006). Residence elsewhere was not significantly associated with infection in women. However, women who had traveled away from their village were significantly less likely to be infected with HSV-2 (OR = 0.42; P = .02).
Marital risk factors
HSV-2 prevalence by marital status is shown in table 2. Prevalence of infection was higher in married women (OR = 2.37; P = .025) and men (OR = 4.97; P = .007) than in those not married. Among married women, those whose husbands had > 1 wife had a significantly higher prevalence of infection than those in monogamous marriages (OR = 4.90; P = .007). Prevalence was also higher in men with > 1 wife than in men in a monogamous union, but this association failed to reach statistical significance (OR = 3.60; P = .19). The prevalence of infection was higher in those whose spouse had previously been divorced or widowed, but this effect was not statistically significant in women (OR = 2.33; 95% CI, 0.86–6.31; P = .09) or men (OR = 2.50; 95% CI, 0.58–10.84; P = .22). In both sexes, there was a higher prevalence of infection among those whose spouse was > -25 years old compared to those with spouses < 25, although these effects were not significant after adjustment for residence and age of the index case.
Behavioral risk factors in women
The distribution of behavioral risk factors in women is shown in table 3. Analysis of age at first sex is restricted to those ≥ years old due to the censored nature of this variable. Age at first sex was not significantly associated with HSV-2 infection in women. Only 4 women ≥ reported never having had sex, and none were HSV-2—positive. There was a significant trend for increasing prevalence of infection with increasing number of lifetime partners, and this was still significant (P = .028) after adjustment for age and residence stratum. There was a nonsignificant trend (P = .28) for increasing HSV-2 prevalence with increasing numbers of sex partners in the past year. Rates of reported condom use were too low in women to examine their association with HSV-2 infection. Nine percent of women were TPHA-positive, and this was associated with a significantly higher prevalence of HSV-2 infection (OR = 5.18; P = .009).
Table 4 shows the ORs for selected risk factors after further adjustment was made for factors that were significantly associated with HSV-2 in table 2 (travel and marital status) and table 3 (lifetime sex partners and TPHA status). After adjustment, there was no association between HSV-2 infection and age at first sex, and the association with lifetime sex partners was weaker and not statistically significant (P = .10 for trend). The effect of partners in the past year was reduced substantially, largely due to the adjustment for marital status and lifetime partners. The association of HSV-2 infection and TPHA remained strong (OR = 3.72; P = .055), and the associations with marital status and travel also persisted after multivariate adjustment (P = .061 and .012, respectively).
In order to investigate in more detail the usefulness of HSV-2 serology as a marker for risk behavior, further analyses were conducted in different age groups. Table 5 shows the distributions of lifetime sex partners and partners in the past year by HSV-2 status, age group, and sex. In women, increased numbers of lifetime partners were associated with an increased prevalence of HSV-2 infection, although the trend was stronger among those < 25 years old (P = .126 for trend) than among those ≥ (P = .54 for trend). However, the interaction between lifetime partners and age was not significant (P = .86). There was no evidence of an association between HSV-2 status and sex partners in the past year in either age group.
Behavioral risk factors in men
The distribution of behavioral risk factors in men is shown in table 3. Analysis of age at first sex was restricted to those ≥ as above. No men ≥ years old reported never having had sex, and age at first sex was not significantly associated with HSV-2 infection. After adjustment for age and residence stratum, there was a trend for increasing risk of HSV-2 infection with increased numbers of lifetime sex partners (P = .057). Men reporting ≥ partners in the past year had a lower HSV-2 prevalence than those reporting < 2, but this effect was not statistically significant.
The prevalence of HSV-2 was significantly higher in men who had ever used condoms (OR = 3.01; P = .03). There was a lower HSV-2 prevalence in circumcised men, but this effect was not statistically significant (OR = 0.66; P = .47). Six percent of men were TPHA-positive, and this was associated with a significantly higher prevalence of HSV-2 infection (OR = 6.57; P = .026).
Table 4 shows the ORs for selected risk factors after further adjustment was made for factors that were significantly associated with HSV-2 in table 2 (lived elsewhere and marital status) and table 3 (lifetime sex partners, condom use, and TPHA status). After adjustment, there was no association between HSV-2 infection and age at first sex. The effect of lifetime partners was reduced, largely due to adjustment for marital status and condom use. The effect of partners in the past year became stronger and more significant: Paradoxically, men with≥ sex partners in the past year had a lower HSV-2 prevalence than those with < 2 partners (OR = 0.27; P = .027).The prevalence of HSV-2 infection remained higher in men who had used condoms (OR = 2.99; P = .065) and lower in men who had been circumcised (OR = 0.39; P = .16). The association of HSV-2 infection and TPHA remained strong (OR = 8.04; P = .042), and the associations with marital status and living elsewhere also persisted after multivariate adjustment (P = .06 and .02, respectively).
Table 5 shows the effect of number of lifetime sex partners separately for those < 25 and ≥ years old (P = .39 for interaction between lifetime partners and age group). In those < 25, there was a significant trend for increasing HSV-2 prevalence with increased lifetime partners (P = .048 for trend); in those ≥, there was no association. The effect of partners in the past year also varied with age (P = .065 for interaction): Those < 25 who had ≥ partners in the past year had a lower HSV-2 prevalence (OR = 0.30; P = .072), whereas those ≥ with ≥ partners had a higher prevalence (OR = 1.77; P = .43).
Reported STD syndromes
Data on reported genital ulceration were available from the baseline survey of the cohort study. In those |>-17, 26 women (15%) and 27 men (20%) reported genital ulceration in the year preceding the baseline survey. HSV-2 prevalence was significantly higher among women (OR = 2.51; 95% CI, 1.00–6.25; P = .03) and men (OR = 2.79; 95% CI, 2.04–7.52; P = .04) who reported genital ulcers than in those who did not.
This is the first community-based study to examine the association between reported sociodemographic characteristics, sexual behavior, and HSV-2 infection in an African population. By nesting the study within the baseline survey of a large randomized trial, we were able to study a representative sample of this rural population. By oversampling those 15–24 years old, we were able to increase the power of the study to examine correlates of HSV-2 infection in the subgroup most at risk of acquisition of infection and in whom such correlates are likely to be most meaningful. However, because this was a prevalence study, we were unable to establish the temporality of events and consequently cannot ascribe a causal role to the associations we identified. In addition, several methodologic issues need further consideration.
Potential sources of bias
Bias may have been introduced due to noninclusion in the baseline serosurvey or to nonparticipation in the questionnaire survey. No information is available for subjects who did not participate in the initial serosurvey, although the participation rate (80%) was high. Of the subsample selected for the HSV-2 study, 73% participated in the questionnaire survey. It is possible that the 27% who did not participate differed from the remaining sample in important respects, since the main reasons for nonparticipation were travel or change in residence. In this study, both of these factors were associated with risk of HSV-2 infection, and in other studies they have been associated with higher risk sexual behavior . If such mobile subgroups were underrepresented, this suggests that HSV-2 seroprevalence may have been underestimated.
The validity of our results also depends on the accuracy of measurement both of HSV-2 serostatus and sexual behavior. The sensitivity and specificity of the ELISA in the detection of culture-proven cases was comparable to WBA, which is the accepted reference standard. However, as with many ELISAs, some uncertainty remains about borderline values. The cutoff estimated using the mixture model technique is dependent on the overall prevalence of infection in the sample. The prevalence in our population was not uniform, and by using a fixed cutoff we risked overestimation in low prevalence subgroups. This could potentially affect younger age groups and would tend to dilute observed associations. However, the absence of positives in males < 17 suggests that the false-positive rate using the derived cutoff was low.
It is well-established that responses to sexual behavior questionnaires can be subject to substantial bias. Although the interviewers were rigorously trained, and the overall response rate for each question was high, some responses are likely to have been influenced by perceived cultural norms of behavior. In addition, some variables, such as number of lifetime sex partners, are subject to recall bias.
Prevalence by age and sex
While the trend for prevalence to increase with age reflects patterns seen worldwide, the absolute prevalence of HSV-2 infection in this population is high in all age groups when compared with data from industrialized countries. Cowan et al.  showed prevalences of < 5% in British male and female blood donors below age 25 compared with 19% and 33% among men and women of the same age group in our study. Comparably high rates have been demonstrated among urban ethnic minorities in North America and STD clinic attenders in urban centers in the developing world [8, 11–13]. The only other study to examine age-specific prevalence in a rural African population was performed by Wagner et al.  in Masaka, Uganda. Their study found higher rates of infection in all age groups than we found. For example, prevalence was 43% (compared with 27%) in 15- to 19-year-old girls, rising to a plateau of ∼80% after age 20, although their estimates in the latter age group were based on a smaller sample.
The higher prevalence of HSV-2 infection among women than in men in Mwanza agrees with observations elsewhere and may reflect a greater probability of transmission from male-to-female than from female-to-male . The high prevalence among young women is remarkable and, given that 93% of women < 20 years old reported ≥ partners, is indicative of a high risk of infection per partner even at an early age. Young women are often in partnerships with older men who have a high prevalence of HSV-2 infection, and this may help explain the high incidence in this age group.
Sociodemographic risk factors
We found an increased HSV-2 prevalence in men who had lived elsewhere in the past 5 years. Such men may have lived in roadside settlements or large towns, where the prevalence of HSV-2 infection may be higher . Migration and travel may also be associated with riskier behavior. The negative association with travel among women is more difficult to explain. It may be related to factors, such as financial independence, which were not adjusted for in the analysis, but which are related both to risk of infection and travel. Compared with women who had not traveled, those who traveled were younger, better educated, and more likely to be married, but they had the same number of lifetime sex partners and the same TPHA prevalence. Unfortunately, we did not collect data on reasons for travel.
The strong association between HSV-2 infection and marital status is plausible. Marriage or cohabitation is effectively a guarantee of sexual activity and, because it is unlikely to be misclassified, is a reliable proxy marker of sexual activity. The strength of the effect in men is remarkable and suggests that a substantial amount of HSV-2 transmission occurs from women to their husbands during marriage.
Another aspect to be considered is the relative insensitivity of men to infection. It may be that the sustained exposure of marital sexual relationships provides sufficient cumulative exposure for men to be infected. In this population, men tend to marry women 5 or so years younger than themselves. Thus, it is likely that the prevalence of HSV-2 infection in younger women tends to equilibrate with that in older men.
We also showed that polygamy of the spouse places married women at increased risk of infection. This may indicate a means by which infection is transmitted between cowives via the husband, the wives forming part of an extended sexual network. Finally, the increased risk in women whose husbands have previously been married supports the role of men in transmission of infection to their wives.
Behavioral risk factors
Several studies have demonstrated an association between HSV-2 infection and reported sexual behavior [6, 7, 11–13]. In contrast to studies in industrialized countries, we found no significant association with age at first intercourse. In both sexes, however, HSV-2-infected persons were more likely to report higher numbers of sex partners, and the trend for increasing prevalence of infection with increased number of lifetime partners was statistically significant for both men and women, even after adjustment for age and residence stratum. The strongest associations were observed in young men and women (< 25), although there were too few observations for the interactions with age to be statistically significant. The high cumulative incidence of infection suggests that most persons are infected before age 25, so it is not surprising that the correlation between infection and lifetime partners is strongest in this age group. The negative association between HSV-2 infection and recent sex partners among young men was unexpected, but may be explained by a reduction in sexual activity during primary episodes of genital herpes.
In both sexes, there was a strong association between TPHA seropositivity and HSV-2 infection, even after adjusting for other variables. This may reflect risk factors that are common to herpes and syphilis, or it may be that syphilitic ulcers enhance HSV-2 transmission or vice versa. We also found an increased prevalence of infection in men who had used condoms. However, condom use in this population appears to be a marker for high-risk partnerships, for example with sex workers, as suggested by its association with HIV infection . We were unable to determine in this study whether condoms provide any protection against HSV-2 infection.
There was a negative association between circumcision and HSV-2 infection that became stronger after adjustment for other variables. The decreased prevalence of HSV-2 among circumcised men in this population is of interest and in agreement with the previously observed lower prevalence of HIV infection in this group . The data support a protective role for circumcision in the acquisition of ulcerative STDs, but these results must be interpreted with caution: The sample size was small relative to the number of variables in the model, and it is possible that some of the observed effects arose by chance.
Reported STD syndromes
Participants with a history of recent genital ulceration had a significantly higher prevalence of HSV-2, suggesting that HSV-2 may be an important cause of genital ulceration in this population, although it is possible that this association reflects common risk factors for HSV-2 and other STDs. The high overall prevalence of infection together with the nature of genital herpes is cause for concern. HSV-2 and HIV are believed to interact so that one enhances the transmission of the other [9, 31, 32]. Furthermore, asymptomatic transmission of HIV and HSV-2 is common [5, 33]. Given the lifelong nature of HSV-2 infection, it is possible that HSV-2 may be responsible for a significant proportion of HIV transmissions in this population. It is unfortunate that our study lacked the power to sufficiently examine this relationship, due to the small number of HIV-infected subjects in this study sample.
In conclusion, this study showed a high prevalence of HSV-2 infection among both women and men in rural Mwanza. We found a strong association between HSV-2 seropositivity and an increased number of lifetime sex partners, especially among young people (< 25). It has been suggested that HSV-2 serology may be used as a biologic marker for sexual risk behavior. These data from a rural African population suggest that HSV-2 serology may be of limited value for this purpose in older age groups: In a cross-sectional study, seropositivity may reflect sexual exposure many years earlier, and this may explain the relatively poor correlation with recent or lifetime sex partners among older adults, whereas in a longitudinal study, HSV-2 seroincidence could only be recorded in the relatively few persons not already infected. However, our data support the use of HSV-2 serology as a proxy for risk behavior among persons ≤25 years old.
We thank the principal secretary, Ministry of Health, the manager of the National AIDS Control Programme, and the director general of the National Institute for Medical Research, Tanzania, for permission to carry out and publish the results of this study. We are also grateful to the field research team for their hard work and for the support given by the regional medical officer, Mwanza, the director of the National Institute for Medical Research, Mwanza, the director of the African Medical and Research Foundation, Mwanza, and regional, district, ward, and community leaders. We especially thank the study population, particularly those who gave their time to respond to the detailed behavioral questionnaire. We are also grateful to the referees for their helpful suggestions.