Institutional members access full text with Ovid®

Share this article on:

The relationship between male circumcision and HIV infection in African populations.

Bongaarts, John; Reining, Priscilla; Way, Peter; Conant, Francis


Infection with HIV is widespread in sub-Saharan Africa, but considerable regional variation exists [1]. Lack of male circumcision is considered by some to be one of the most important factors responsible for the geographic spread of HIV infection in Africa [2]. The initial evidence was derived from macro-level analyses of the geographical distribution of HIV infection and circumcision in Africa [3,4]. HIV prevalence appeared to be much higher among cultural groups with no tradition of ritual circumcision than that among circumcising ethnic groups.

The existence of an ecological correlation between lack of male circumcision and HIV prevalence suggests that circumcision could potentially be an important determinant in the prevention of the spread of HIV, but individual-level data are needed to demonstrate such a relationship. A review of the epidemiological studies that have evaluated the association between male circumcision status and risk of HIV infection included 21 cross-sectional and two prospective studies in sub Saharan Africa [5]. Of these studies, 14 were conducted among patients recruited at clinics for sexually transmitted diseases (STD). Most of these studies of male STD patients showed a protective effect of circumcision against HIV infection, with odds ratios (OR) between 0.12 to 0.77. It has been noted that most of the studies were conducted in urban populations, and that several had methodological limitations [6].

Three general population surveys have assessed the role of male circumcision in HIV prevalence. In a rural area of Masaka District, Uganda, no data on circumcision were collected but Muslim men and women had significantly lower HIV prevalence than others [7]. In Rakai District, Uganda, where 15% of men had been circumcised, HIV prevalence was lower among circumcised men although not significantly so, but after controlling for age, residence, sexual behaviour and history of STD, circumcised men had significantly lower risks of HIV infection than non-circumcised men [8]. In a population-based study in Tanzania, the univariate association showed a significant excess risk of HIV among circumcised men, but after adjustment for confounding factors the OR was close to one [9].

Adequate adjustment for confounding factors is clearly of great importance in assessing the association between male circumcision and HIV status, since the practice of circumcision is determined by a range of cultural and socio-demographic factors, some of which are strongly associated with risk of HIV infection.

This paper investigates the association between male circumcision and the risk of HIV infection and other STD among men in Mwanza Region, north-western Tanzania. This area provides a good opportunity for studying this association, since Mwanza lies close to the boundary of the circumcision belt in eastern Africa. The Sukuma are the dominant ethnic group in this region, and are traditionally non-circumcising [10]. About one-quarter of adult men in the region are now circumcised, however, owing to immigration of other ethnic groups and changing practices among the Sukuma themselves. The population includes both Muslims and Christians.

During the past few years, several population-based studies have been conducted in the region, including the early study referred to above [9, 11–15]. These have provided a large quantity of data, which are reanalysed in this paper in order to further investigate the effect of circumcision on the risk of HIV and other STD. Male circumcision may influence the geographic spread of HIV infection by increasing men's susceptibility to infection, their infectivity, or both. This analysis focuses primarily on the effect of male circumcision on susceptibility.

All males providing valid data concerning circumcision and HIV status were included from five studies conducted in Mwanza Region (Table 1). The complete methods of the studies have been described elsewhere, but a brief description of each study follows. Study 1 was a stratified random-cluster sample from the whole of Mwanza Region, carried out in 1990 [9]. Study 2 is an ongoing community study in six villages about 20 km west of Mwanza town, which started in 1994 [11]. Study 3 was a cluster sample survey of four fishing villages in one district, considered to be representative of fishing villages along the shores of Lake Victoria [12]. Study 4 is based on the intake data of a cohort study among factory workers in Mwanza town [13]. Study 5 was the baseline round of a longitudinal cohort study, designed to assess the effects of community-level STD control measures on the incidence of HIV and other STD [14,15]. The cohort was randomly selected from 12 communities, and in-depth data on sexual behaviour were collected from a subsample of the cohort between 4 and 14 months after the baseline study [16]. The subsample formed a nested, unmatched case-control study, the controls being a random selection from the whole cohort, and three sets of cases: HIV-positive individuals; cases of active syphilis [positive Treponema pallidum haemagglutination (TPHA) and rapid plasmin reagin (RPR) tests]; and reported STD cases.

In all studies, respondents were interviewed using structured questionnaires to obtain basic demographic and socio-economic background information and to identify risk factors for HIV infection and other STD. These included a question on whether the respondent had been circumcised. Age at circumcision was also recorded in Studies 4 and 5. The presence of an STD during the past year was taken from the responses to direct questions about symptoms in the questionnaire. Free treatment was offered using syndromic treatment algorithms and, in addition, an RPR test was performed in the field; if this test was positive, syphilis treatment was given immediately. A blood sample was collected in all studies and sera were transported to a central laboratory, in a referral hospital in Mwanza, for HIV and TPHA serological testing, using test procedures described previously [9,11–13,15]. Sera were rendered anonymous using code numbers, according to Tanzanian government guidelines, and the linkage of code numbers with subject identity was kept strictly confidential. Parallel counselling and testing services for HIV were offered in all studies, but were taken up by very few men.

In the urban cohort study (Study 4), the question on circumcision status was repeated 2 years after the original report in order to assess reliability of reporting. On this occasion, the circumcision status was also determined independently by physical examination. Factory workers who came for their routine medical examination were asked to pull down their foreskin and then to release it. If the glans penis was covered completely by the foreskin, the subject was recorded as not circumcised; if it was partly covered, the subject was recorded as partially circumcised; while if there was no foreskin at all to cover the glans penis, the person was recorded as completely circumcised. The subsample of men in study 5 who were questioned in detail about their sexual behaviour were also asked twice about their circumcision status, once at the baseline survey of the full cohort, and once at the questionnaire survey of the subsample. In both studies, the interviewer or physician were blind to the reported circumcision status at the original interview.

The analysis included only men with known circumcision status (self-reported) and unequivocal HIV status. The first part of the analysis focused on patterns of circumcision in the study populations. Subsequently, the risk of HIV infection was examined by circumcision status for each study separately in a univariate analysis and in a logistic regression model: after adjustment for age and residence; after additional adjustment for education, ethnicity, occupation and religion; and after additional adjustment for sexual risk behaviour, as measured by reported number of sex partners during the year preceding the interview, and if condoms had ever been used. Age adjustment was performed using age-groups 15–19 years, 20–24 years, 25–29 years and 30+ years, since there was little variation in the proportion circumcised above age 30. Residence was classified as urban (Mwanza town), rural, roadside or island. In Study 5, the analysis was limited to the case-control study data as only those men had complete information on background characteristics and sexual behaviour.

A pooled analysis was conducted for the three studies (Studies 2, 4, and 5) that had the most complete information on background and sexual behaviour variables. The association between circumcision and HIV was first examined in a univariate model, followed by stepwise adjustments for age and residence, for all sociocultural and demographic variables, and for markers of sexual behaviour. Subsequently, models were tested including interaction terms between circumcision and residence. In addition, models with interaction effects between study and residence were run to adjust for the individual communities, but this adjustment had no effect on the association of circumcision and HIV. The same procedures were used to examine the effects of circumcision on TPHA seropositivity, indicating past or present syphilis, and on reported history of STD during the past year.

The reliability and validity of self-reported circumcision status were evaluated using kappa scores [17] to assess agreement between first and second report (Study 4 and 5), and between second report and physical examination results (Study 4).

Stata version 3.1 was used for all analyses (Stata Corporation, College Station, Texas, USA).

Circumcision rates were highest in the urban area, followed by the islands, roadside villages and rural areas (Table 2). Men under 20 years of age were less likely to be circumcised than older men, but there was little age-related variation over 20 years of age. Skilled and manual work and increasing level of education were associated with higher levels of circumcision; 70% of men with secondary education were circumcised.

Religious denomination was strongly associated with circumcision status. The majority of Muslims were circumcised, although only 68% of rural Muslims reported themselves as circumcised compared with that of 92% in the urban area. Among Christian men, 43% had been circumcised in the urban area and 21% in the rural area. The Sukuma are the predominant ethnic group in Mwanza Region. Traditionally, Sukuma men are not circumcised, but 34% of Sukuma men in the urban factory population and 13% in the rural cohort reported themselves as circumcised. It was also common to find circumcised men among other traditionally non-circumcising ethnic groups, such as the Haya, Nyamwezi and Zinza. Ethnic groups which are traditionally circumcising, such as the Jita, are usually fishermen and many live on the islands. Adjustment for other demographic variables modified the strength of the association between circumcision and ethnic group, residence, education and religion, but the associations persisted.

Circumcised men in the urban factory population (Study 4) and the subsample of the rural cohort (Study 5) were asked to recall the age at which they were circumcised (Table 3). Almost half of the urban men had been circumcised before their 15th birthday, compared with 34% of men in the rural study. Muslim men and men of traditionally circumcising ethnic groups were circumcised earlier than other men. Of Sukuma men who reported themselves as circumcised, only one-third were circumcised before age 15.

The association between HIV prevalence and circumcision was first examined for each of the studies separately (Table 4). In the univariate analyses, there was a significantly lower risk of HIV among circumcised men in Study 4, but the other four studies had OR ranging from 0.98 to 1.66. After adjustment for age, residence, the other background variables and sexual behaviour, all studies showed OR of less than 1, but the only study to achieve statistical significance was Study 4.

The data from studies 2, 4 and 5 were pooled (Table 5). There was no reduced risk of HIV infection among circumcised men in univariate analysis [unadjusted OR, 1.04; 95% confidence interval (CI), 0.84–1.29]. Adjustment for age and residence and further adjustment for other confounding factors revealed a modest, yet significant, protective effect of circumcision (OR, 0.63; 95% CI, 0.49–0.81) which was similar after adjustment for sexual behaviour markers (OR, 0.62; 95% CI, 0.48–0.81).

Since HIV prevalence and circumcision levels varied markedly by residence, interaction terms were included to assess the effect of circumcision separately in each residential stratum. In the final model, the interaction was not statistically significant [x2 = 3.3; 3 degrees of freedom (df); P = 0.3]. The effect of circumcision on HIV was statistically significant only in the urban stratum (OR, 0.46; 95% CI, 0.32–0.68; P < 0.001) and nearly significant at the 5% level in the roadside villages (OR, 0.65; 95% CI, 0.42–1.01; P = 0.06). Little or no protective effect was observed in the rural and islands strata.

The overall prevalence of positive TPHA test among 4984 men with complete risk factor data was 18.4%. TPHA prevalence was 16.7% in circumcised men and 19.1% in non-circumcised men (OR, 0.85; 95% CI, 0.72–1.00). After controlling for background characteristics (as for HIV) there was almost no difference between the two groups of men (OR, 0.95; 95% CI, 0.79–1.15). Additional adjustment for sexual behaviour markers to the equation had little effect.

The proportions of men reporting an STD during the past year were 10.0% among non-circumcised men and 9.8% among circumcised men (OR, 0.88; 95% CI, 0.72–1.07). Adding all background variables and sexual-behaviour markers resulted in an OR close to one (OR, 0.98; 95% CI, 0.77–1.26). Separate analyses of history of genital discharge syndrome (GDS) and genital ulcer syndrome (GUS) were carried out, but there were no significant effects in either case.

In Study 4, during the first interview 117 men said they were circumcised, and 105 (90%) repeated the same response at the second interview. Of the 166 men who reported being non-circumcised at the first interview, 112 (67%) repeated that report at the second interview. In Study 5, 197 men said they were circumcised during the first interview and 174 (88%) repeated the same response at the second interview. At the first interview, 844 men reported that they were not circumcised and 768 (91%) repeated the same response at the second interview. The agreement between the first and second reports was 77% for Study 4 (kappa score, 0.52; P < 0.001), and 90% for study 5 (kappa score, 0.72; P < 0.001).

The validity of reporting in comparison with physical examination was assessed in 202 factory workers in Study 4. Among these men, eight were recorded as partially circumcised, all of whom reported themselves as circumcised. Among 111 men who had reported they were circumcised, 77 (69%) were found to be circumcised (fully or partially) on examination. Among 91 men who reported themselves as not circumcised, 86 (94%) were found to be not circumcised on examination. The agreement between actual and reported circumcision status was 81% (kappa score, 0.62; P < 0.001).

Data from men in Mwanza region suggest that there is a modest protective effect of male circumcision on the risk of HIV infection among men, with an adjusted OR of 0.62. A stronger protective effect was observed in urban Mwanza (adjusted OR, 0.46) and to a lesser extent in roadside villages (OR, 0.65), while in rural areas and islands there was no significant reduction in HIV prevalence among circumcised men. The overall results from the urban areas and roadside villages are in agreement with findings from other studies in Africa, although the protective effect is smaller than reported in most studies of STD patients [5].

The initial unadjusted analysis showed no association between circumcision and HIV infection (OR, 1.04). However, circumcision tended to be more common in urban areas, in educated men and in those with skilled and manual work, as well as in men who reported the use of condoms and multiple sexual partners. There is also a higher prevalence of HIV infection among the same groups, thus obscuring the protective effect of circumcision, which emerged only after adjustment for these factors. It is likely that our adjustment for confounding factors was incomplete, particularly given the difficulty of measuring sexual exposure accurately, in which case the actual protective effect may be somewhat stronger than that reported here. The confounding effect was also seen in the separate analyses of four of the five studies.

Circumcision was found to be strongly associated with the level of education, and was very common among those with secondary education. Higher levels of education may imply social contact with a broader mix of ethnic and religious groups, perhaps resulting in peer pressure to be circumcised. A qualitative study to investigate reasons for changing circumcision practices is being conducted. Two reasons for the higher circumcision rates in urban areas are the presence of more Muslims and of more educated men. The association between circumcision and these variables persisted after adjustment for other socio-demographic variables.

Male circumcision may have a direct effect on the risk of HIV infection, perhaps because it reduces the risk of preputial lesions and renders the glans penis less vulnerable [6]. It may also have an indirect effect on HIV risk by reducing the risk of STD, particularly genital ulcers [18], which are thought to facilitate HIV transmission [19]. Circumcision may be associated with better penile hygiene, which may reduce the risk of HIV, other STD and penile cancer [6, 20–22].

GUS and other STD have been shown to be very common in Mwanza region [23]. We were unable to demonstrate a protective effect of circumcision against STD in this study. We have found no association with TPHA and self-reported STD during the past year, in agreement with previous work in Mwanza [24] but not with other studies [20]. We cannot, however, exclude a protective effect against STD, particularly as self-reported STD is an unreliable indicator of STD occurrence. Anecdotal reports suggest that some men may be circumcised as a traditional remedy for STD, especially genital ulcers, and this may further distort the association.

Why is the effect of circumcision on HIV and STD less persuasive in this population than in some other studies? First, it may be that the effect of circumcision is diluted because many men are circumcised after initiation of sexual intercourse. Sexual debut usually occurs between 15 and 20 years of age in this population [16,25], while this study has shown that many men are circumcised after the age of 15 years. When men circumcised before and after age 15 were compared, there was little difference in the OR for HIV infection (data not shown), but the number of infections was too small for definitive conclusions.

Secondly, many previous studies were undertaken in men at high risk of STD (STD patients or contacts of commercial sex workers). If the effect of circumcision on HIV is mediated partly by a protective effect against STD, then the effect in the general population is likely to be weaker than in these high-risk groups, since the incidence of STD will be lower.

Thirdly, most studies showing a strong protective effect have been conducted in urban centres [5]. In our study, the protective effect of circumcision was greater in the urban population than in rural areas. This may result from a higher incidence of STD in urban areas, although previous studies in Mwanza Region have failed to show a substantial difference in STD incidence or prevalence between urban and rural areas [23].

Misreporting of circumcision status is a further bias which may obscure a protective effect of circumcision. In our study, repeated questioning of the same subjects gave generally good agreement, although there was a higher rate of discrepancy in the urban factory cohort, with more men reporting that they were circumcised at the second interview. Since there was a 2-year interval between the interviews, and since circumcision in this population often occurred in early adulthood, it is possible that some of these men were circumcised between the two surveys. Unfortunately, this question was not addressed by the interview.

Several studies have validated self-reported circumcision status against subsequent physical examination. In an Australian study, 98.5% of 1284 men gave an accurate report of their circumcision status [26], but other studies have found less satisfactory agreement [27]. Part of this variation may be due to differences in the amount of foreskin removed, and this may vary between populations or by ethnic group [10]. It is possible that the protective effect of circumcision against HIV infection is less pronounced if circumcision is partial. In our study, we found some misreporting, mostly in men who reported themselves as circumcised but who were found to be not circumcised on examination. This may be a reflection of changing norms in this society, favouring male circumcision.

In ecological analysis of the association between circumcision and HIV prevalence, anthropological data have been used to map circumcising and non-circumcising ethnic groups. Our data from Mwanza show that, at least where ethnic groups with diverse circumcision practices are living in the same region, such distinctions may be inadequate at the individual level. Traditionally non-circumcising ethnic groups have significant proportions of men circumcised, and, conversely, not all men of circumcising ethnic groups are circumcised. Similarly, religion cannot necessarily be regarded as a proxy for circumcision status, as assumed in some previous studies [7]. Although circumcision occurs more commonly and earlier among Muslims than among Christians, not all Muslims are circumcised and many Christian men, even of traditionally non-circumcising ethnic groups, are circumcised. In common with many other aspects of sexual and social behaviour, circumcision and HIV infection are affected by personal decisions which may not be totally governed by religion or ethnicity.

In summary, our analysis of the association between male circumcision and risk of HIV and STD has shown a modest protective effect of circumcision against HIV infection. Circumcision status is related to many other demographic, sociocultural and behavioural variables that need to be taken into account in population-based studies.

We thank the Principal Secretary, Ministry of Health and the Director General, National Institute for Medical Research for permission to publish the results of these studies. We thank the regional, district, ward and community leaders for their support and the population in these communities for their participation in these studies.

AIDS. 11(3):73-80, March 11, 1997.

AIDS: June 1989
Orginal papers: PDF Only

The relationship between HIV seroprevalence and the proportion of uncircumcised males in African countries is examined to determine whether circumcision practices play a role in explaining the large existing variation in the sizes of African HIV epidemics. A review of the anthropological literature yielded estimates of circumcision practices for 409 African ethnic groups from which corresponding national estimates were derived. HIV seroprevalence rates in the capital cities were used as indicators of the relative level of HIV infection of countries. The correlation between these two variables in 37 African countries was high (R = 0.9; P < 0.001). This finding is consistent with existing clinic-based studies that indicate a lower risk of HIV infection among circumcised males.

(C) Lippincott-Raven Publishers.