Male Circumcision and Risky Sexual Behavior in Zimbabwe: Evidence from the 2010-11 Zimbabwe Demographic and Health Survey

In 2009, Zimbabwe adopted voluntary medical male circumcision (VMMC) as an additional method of HIV prevention. The promotion of VMMC has raised a major concern that it might lead to an increase in high-risk sexual behavior, a phenomenon known as risk compensation or behavioral disinhibition. This study sought to test whether circumcised men in Zimbabwe are more likely to have engaged in risky sexual behavior. The study used data collected from 7,480 men age 15-54 who were interviewed during the 2010-11 Zimbabwe Demographic and Health Survey (ZDHS). Logistic regression was used to assess the association between circumcision and risky sexual behaviors. The study found no statistically significant association between male circumcision and risky sexual behavior. These results suggest a need to continue monitoring the relationships between ongoing VMCC campaigns and men’s risky sexual behavior. Information dissemination on VMMC should emphasize caution in messages promoting medical male circumcision to avoid giving the impression that it provides immunity against HIV.


Introduction
Zimbabwe has one of the highest HIV prevalence rates in the world, currently estimated at 15.2 percent among adults age 15-49 (ZIMSTAT and ICF International, 2012). While HIV can spread through many channels, it has been suggested that most HIV infections (92 percent) in Zimbabwe occur through heterosexual contact, followed by mother-to-child transmission (Zimbabwe National AIDS Council, 2005). Existing approaches to preventing the spread of HIV have mainly been based on the "ABC" approach to behavior change, which promotes abstinence, faithfulness and correct use of condoms. In devising policy guidelines on the effective prevention of new HIV infections, Zimbabwe came up with a Behavior Change Strategy, which focuses on increasing the age at sexual debut, increasing correct condom use, and reducing the average number of sexual partners. The approach also encouraged individuals and couples to get tested for HIV. However, the ABC approach has had limited success in most sub-Saharan African countries (Kerwin et al, 2011). In 2009, Zimbabwe adopted voluntary medical male circumcision (VMMC) as an additional HIV prevention strategy to the existing ABC behavior change model. The campaign for scaling up VMMC for HIV prevention focuses on men from the age of 13 who are HIV-negative. The procedure is offered free of charge at designated public hospitals and health centers throughout the country.

Literature review
Medical male circumcision has become a significant dimension of HIV prevention interventions, after the results of three randomized controlled trials (RCTs) in Uganda, South Africa and Kenya demonstrated that circumcision has a protective effect against contracting HIV of up to 60 percent for men having sex with women (Auvert et al., 2005;Bailey et al., 2007;Gray et al., 2007). Pursuant to these findings, the World Health Organization (WHO) and UNAIDS recommended that VMMC be implemented immediately as an additional HIV prevention strategy in countries with high HIV prevalence rates but low male circumcision rates (WHO and UNAIDS, 2007).
Notwithstanding the WHO and UNAIDS recommendation, and the promotion of VMMC as an additional HIV prevention strategy in a number of sub-Saharan African countries, including Zimbabwe, fundamental questions remain about whether promoting VMMC as an HIV prevention intervention will result in decreases in HIV incidence. A major concern is that promoting VMMC may lead circumcised men to develop a false sense of effective protection against HIV and engage in risk compensation or behavioral disinhibition. The Zimbabwe National AIDS Council (NAC) highlighted concern over the risk compensation behaviors among circumcised men (Yikoniko, 2012). This phenomenon involves either halting or decreasing previous protective behaviors such as condom use or partner reduction, thus reducing or nullifying the protective effects of male circumcision (Westercamp et al., 2012).
However, studies have not found evidence of this phenomenon. Mattson et al (2008) developed an 18item measure of risky sexual behavior for their RCT in Kisumu, Kenya, and found no significant differences in the risk propensity scores between circumcised and uncircumcised men. Another study in Kenya by Westercamp et al (2012) followed up circumcised men after 24 months and compared their perception of risk and their sexual behavior with that of uncircumcised males. The findings supported the earlier findings by Mattson et al (2008). In a related study in Malawi, Godlonton et al (2011) concluded that after learning that circumcision lowers the risk of HIV infection, circumcised men did not become riskier in their sexual behavior. Despite this evidence, there is still debate on the issue of behavioral disinhibition. In light of the discourse on male circumcision and risky sexual behavior, the present study instead sought to investigate the relationship between male circumcision status and engaging in risky sexual behavior.

Conceptual framework
The conceptual framework designed for this study shows the possible individual and household factors associated with male circumcision on one hand, and the link between male circumcision status and indicators of risky sexual behavior on the other (see Figure 1).
It is not clear how or when male circumcision began, but it is known that religion has an influence on circumcision. Among Jews and Muslims, male circumcision is practiced as a symbol of keeping the covenant with God 1 , although circumcision is not mentioned in the Holy Koran (Aldeeb Abu Sahlieh, 1999). Thus, in the present study we have included religion as determinant of circumcision. Another significant determinant of circumcision is ethnicity. Male circumcision has been practiced for many decades as a rite of passage into adulthood among some groups of the Xhosa of South Africa and the Aborigines of Australia (Vincent, 2008;Emery, 2005). However, the present study did not include ethnicity as a determinant of circumcision because most national surveys in Zimbabwe, including the ZDHS, do not collect statistics on ethnicity as a national policy.
Other studies have linked male circumcision to demographic and socioeconomic variables, such as level of education. The 2004-05 Tanzania DHS shows higher rates of circumcision among men with higher levels of education compared with men with lower levels of education. Tanzanian males with above primary school education had significantly higher odds of being circumcised than those with incomplete primary school education (Wambura et al., 2009). The 2005-06 ZDHS found a weak relationship between education and circumcision; men with primary school education were the most likely to be circumcised (CSO and Macro International, 2007). In Central Uganda the willingness to undergo MMC was found to have an inverse relationship with education (Asiimwe, 2012).
The influence of age as a determinant of male circumcision cannot be understood outside of the context of ethnicity and the generally accepted culture. In many Western countries, and among Jewish people, circumcision is generally performed at birth or within the first week of life (Gallaher, 2000). Among Muslims, circumcision may be performed in early childhood or adolescence. Studies in the United Kingdom have found that circumcision is mostly done by the upper middle class (Gallaher, 2000). In Zimbabwe, traditional circumcision is done at different ages depending on the ethnic group performing the ritual.
Several major surveys have explored the measurement of risky sexual behaviors. Consecutive studies in the British National Survey of Sexual Attitudes and Lifestyles (Natsal) have used proxies of risky sexual behavior such as a person's number of sexual partners and condom use during intercourse (Mercer, 2010). Data for common indicators of risky sexual behaviors used by agencies such as UNAIDS 2 are collected by MEASURE DHS. As well as number of sexual partners and condom use, these indicators also include having paid sex and higher-risk sex (defined as recent sex with a non-marital or noncohabiting partner). In the conceptual framework of this study, shown in Figure 1, male circumcision is hypothesized to be a determinant of these risky sexual behaviors.

Data and methods
This study used data on men from the 2010-11 ZDHS. The ZDHS is designed to produce nationally and provincially representative estimates of data on population and health indicators. The survey interviewed men age 15-54 who are usual residents of a household or were visitors who stayed in the household the night before the survey. In total, 7,480 men were interviewed in the survey. Of particular interest for this study were the data on men's circumcision status and sexual behavior. All interviewed men (7,480) were included in the analysis of factors associated with circumcision, regardless of whether or not the men reported having ever had sex. Individual and household characteristics included in the analysis were men's age, level of education, religion, province, and urban-rural residence. These were considered as determinants of male circumcision. The categories of religion, education and province were collapsed into summary categories according to the characteristics considered to be most salient for the analysis. Proxies of risky sexual behavior were developed and compared between circumcised and uncircumcised men interviewed in the 2010-11 Zimbabwe Demographic and Health Survey (ZDHS). The key variables in the present analysis were defined as highlighted below.

Background Variables
All of the respondents were grouped into 10-year age groups due to their relative homogeneity. Education was defined as highest level of education attended. Secondary and higher categories were combined into the "Secondary and above" category. Similarly, marital status categories were also combined to produce "never married", "married/living together" and "formerly married" categories. The latter category included all men who reported that they were divorced, separated or widowed at the time of the survey. Wealth was regrouped into terciles at the household level. For the purposes of this study, Zimbabwe's 10 provinces were collapsed into five groups.

Mashonaland
Central, Mashonaland East, Mashonaland West and Masvingo provinces were combined into "Mashonaland"; Matabeleland North and South and Midlands were combined into "Matabeleland/Midlands"; while Manicaland, Harare and Bulawayo were left as is because of their uniqueness in relation to the rest of the country. The ZDHS asked respondents to state their religion according to nine categories. For the purposes of this study, respondents were reclassified into four groups: Christian ZDHS defined circumcision as the complete removal of the foreskin. This was necessary to achieve a common understanding of the concept between interviewers and respondents. Thus for this study, circumcision status was determined by the responses to the question: "Some men are circumcised, that is, the foreskin is completely removed from the penis. Are you circumcised?" There were three possible responses to this question: "yes", "no" or "I don't know". A small proportion of men who reported that they did not know their circumcision status were assumed to be uncircumcised.

Risky Sexual Behavior
Four indicators of risky sexual behavior were considered for this study, as follows: (1) Paid sex: In the present study, all men who reported to have ever paid for sex were considered to have engaged in risky sexual behavior.
(2) Multiple sexual partners in the past 12 months: Having had two or more sex partners in the 12 months preceding the survey. While this indicator is often reported as a percentage of sexually active men, we considered all men to be 'at risk'; hence the denominator is all men.
(3) Higher-risk sex in the past 12 months: Higher-risk sex is defined as any sexual intercourse with a non-marital or noncohabiting partner, among those who had sex in the 12 months before the survey.
(4) Not using a condom during last higher-risk sex: Condom use during sexual intercourse is considered a preventive measure against the transmission of sexually transmitted infections (STIs). The ZDHS collected data on condom use in the past 12 months during last sexual intercourse with up to three different partners. Not using a condom at the last sexual intercourse with a high-risk partner was considered a risky behavior. In eight cases, men reported higher-risk sex in the past 12 months, but use of a condom with that partner could not be determined because their last three partners were spousal; these men were excluded from the denominator.

Research Question and Hypothesis
Risk compensation theory suggests that as the risk of an undesirable outcome is reduced, people may compensate by behaving in riskier ways that keep their overall level of risk constant (cf. Hedlund, 2000). This study was guided by the need to answer a question on the relationship between male circumcision and high-risk sexual behavior. The analysis tests the hypothesis that in Zimbabwe men who are circumcised engage in riskier sexual behaviors. The corresponding null hypotheses for this study are that there is no significant association between male circumcision status and risky sexual behavior, and that there is no significant association between wanting a circumcision and risky sexual behavior.

Data Analysis
The analysis was done using STATA version 12. Models were used primarily for two reasons, first to determine the strength of the association between circumcision and background variables, and second, to determine the association between circumcision status and proxies of risky sexual behavior. In the data analysis, we took into account the complex multi-stage sampling design and the weights that were calculated for the ZDHS. Table 1 summarizes the characteristics of the 7,480 men age 15-54 interviewed by the 2010-11 ZDHS. Forty-two percent were age 15-24, while 30 percent were age 25-34. Seventy-six percent of the respondents had at least some secondary education, 51 percent were married or living together, 74 percent gave their religion as Christian, and 63 percent resided in a rural area at the time of the survey.  Table 2 shows the proportion of men reporting risky sexual behavior. Eighteen percent reported having ever paid for sex in their lifetime. Eleven percent of all men reported having had two or more partners in the past 12 months, while 33 percent of sexuallyactive men had sex with a higher-risk partner, that is, a non-marital non-spousal partner, in the past year. Of the men who reported having had a higherrisk partner in the past year, 22 percent did not use a condom during their last sexual encounter with that partner. a Higher-risk sex is defined as sex with a non-marital, non-cohabiting partner in the past 12 months among men who had sex in the past 12 months. b In eight weighted cases, men who had higher-risk sex in the past 12 months were excluded because their last 3 partners were spouses; condom use with non-spousal partner could not be determined

Characteristics of Circumcised and Uncircumcised Men
Of the 7,480 men interviewed, 9 percent (685) reported being circumcised. An additional 37 percent of men responded that they would want to be circumcised if it were available for free and were safe. Table 3 shows the proportion of men who are circumcised, want circumcision, or do not want circumcision by their background characteristics. Men age 35-44, men who attended only primary education, Muslim men, men who are married, and men from the Manicaland region are most likely to be circumcised. The proportion of circumcised men is lowest in the Mashonaland region, where only 7 percent of men are circumcised. The 2010-11 ZDHS did not ask circumcised men whether they were circumcised as a result of Zimbabwe's voluntary medical male circumcision (VMMC) campaign, but did ask about their age at circumcision and who performed their circumcision. Nearly half of men (47 percent) were circumcised before age 5, while 31 percent of respondents were circumcised at age 15 or above. As mentioned, Zimbabwe's male circumcision campaign was scaled up in 2009; few of the respondents in the 2010-11 ZDHS had been circumcised in the two years preceding the survey. 3 Moreover, only health professionals are employed as part of the VMMC program, whereas 58 percent of the circumcised men were circumcised by a traditional practitioner or a family friend, while 24 percent were circumcised by a health professional. In all, the findings indicate that relatively few men interviewed by the ZDHS could have been circumcised as a result of the VMMC campaign.
3 The circumcision scale-up began in 2009 and the survey was fielded in 2010-2011. In the ZDHS 58 men (weighted=62.6 cases) reported having been circumcised in the past two years.

Factors Associated with Male Circumcision
The present study analyzed background factors that are associated with male circumcision. We fitted logistic regression models and computed unadjusted and adjusted odds ratios. Table 4 presents the results of the analysis. Unadjusted odds ratios of being circumcised increase with men's age, which is consistent with later ages of circumcision overall in Zimbabwe. Ever-married men are significantly more likely to be circumcised. Education and urban-rural residence are not significantly associated with circumcision.
After controlling for other factors, age is not significantly associated with circumcision; only religion, province, and marital status are significant in the adjusted odds ratios (see Table 4). Married men are more likely to be circumcised compared with never-married men. Men in Mashonaland and Harare are significantly less likely to be circumcised compared with men in Manicaland (the reference group). Muslim men are overwhelmingly more likely to be circumcised compared with other religions.

Male Circumcision and Risky Sexual Behavior
The major objective of this study was to determine if there are significant differences among circumcised and uncircumcised men in engaging in risky sexual behavior. We examined unadjusted and adjusted odds of the association between male circumcision and risky sexual behaviors, as shown in Tables 5-8. The unadjusted odds ratios show that circumcised men are significantly more likely to have ever paid for sex (Table 5) and to have had multiple sexual partners in the past 12 months (Table 6). However, these associations are not statistically significant after adjusting for other background factors. Being circumcised is not significantly associated with higher-risk sex in the past 12 months (Table 7) nor with condom use at last higher-risk sex ( Table 8).
The adjusted models for each of these dependent variables indicate that older men and men who have ever been married are more likely to have ever paid for sex (Table 5). Ever-married men also have more than twice the relative odds of having multiple sexual partners in the past 12 months, while men's age has an inverse association with having multiple sexual partners (Table 6). Currently and formerly married men are significantly less likely to have had higherrisk sex in the past 12 months. Wealth and education have no significant relationship with higher-risk sexual behaviors. The relationship between province and risky sexual behavior is inconsistent across indicators.   Higher-risk sex is defined as sex with a non-marital, non-cohabiting partner in the past 12 months. Exponential coefficients; 95 percent confidence intervals in brackets *p<0.05, **p<0.01, ***p<0.001 + based on <50 weighted cases "Mashonaland Province" includes Mashonaland East, Mashonaland West, Mashonaland Central and Masvingo "Matabeleland/Midlands" includes Matabeleland North, Matabeleland South and Midlands Not using a condom at last sexual encounter with a non-marital, non-cohabiting partner in the past 12 months, among men who had such a partner. Eight cases were excluded because men's last three partners were spousal, so they were not eligible for questions about condom use with the last high-risk partner. Exponential coefficients; 95 percent confidence intervals in brackets *p<0.05, **p<0.01, ***p<0.001 ‡ eliminated due to perfect collinearity; all 12 Muslim men who had higher-risk sex in past year used a condom during their last encounter. "Mashonaland Province" includes Mashonaland East, Mashonaland West, Mashonaland Central and Masvingo "Matabeleland/Midlands" includes Matabeleland North, Matabeleland South and Midlands

Discussion and conclusion
The present study investigated the association between male circumcision and risky sexual behavior among men who participated in the 2010-11 ZDHS. The study is set against the backdrop of the recent promotion in Zimbabwe of voluntary medical male circumcision as an additional HIV prevention method. VMMC has raised fears among critics of the program on the basis that circumcision would lead to an increase in high-risk behavior. These fears emanate from the concern that circumcised men might experience a decline in the perception of their risk of contracting HIV, a phenomenon known as risk compensation or behavioral disinhibition. The study results support the null hypothesis that there is no association between male circumcision and risky sexual behavior. These results are in line with earlier findings from studies in Kenya (Mattson et al., 2008, Westercamp et al., 2012. While the current study is limited to cross-sectional data and thus cannot establish causality, the lack of a relationship between circumcision and risky sexual behavior likely reflects the fact that men are circumcised for many reasons besides HIV prevention. In fact, given the timing of the ZDHS fieldwork  in relation to the scale-up of the VMMC campaign (2009), few men surveyed are likely to have been circumcised as a result of the campaign.
All in all, the findings of this study do not support the argument that circumcised men are prone to engage in risky sex behaviors compared with uncircumcised men. The lack of observed relationship between medical male circumcision and risky sex may be due to the short period between the start of the male medical circumcision in Zimbabwe which started in 2009 and the collection of ZDHS data (2010)(2011). Subsequent surveys in Zimbabwe may indeed find a significant association between circumcision and sexual risk-taking behavior.
The results of this study lead us to make the following recommendations. First, there is a need to continue observing the relationship between ongoing medical male circumcision campaigns and risky sexual behavior. Second, communication and information dissemination policy must emphasize the need to take caution in messages promoting medical male circumcision, to avoid giving the false impression that circumcision provides immunity against HIV. Finally, we recommend that there is a need to continue studies on the acceptability and feasibility of medical male circumcision among noncircumcised populations with high incidence of HIV.