Self-reported and observed female genital cutting in rural Tanzania: associated demographic factors, HIV and sexually transmitted infections
Elise Klouman, Solvn.14, 1458 Fjellstrand, Norway. Tel.: +47 9576 9476; E-mail: email@example.com (corresponding author)
Rachel Manongi, Kilimanjaro Christian Medical College, Moshi, Tanzania. E-mail: firstname.lastname@example.org
Knut-Inge Klepp, Department of Nutrition, Faculty of Medicine, University of Oslo, Norway. E-mail: email@example.com
Objectives To determine (i) the prevalence and type of female genital cutting (FGC) in a rural multi-ethnic village in Tanzania, (ii) its associated demographic factors, (iii) its possible associations with HIV, sexually transmitted infections (STIs) and infertility and (iv) to assess the consistency between self-reported and clinically observed FGC.
Method The study was part of a larger community-based, cross-sectional survey with an eligible female population of 1993. All were human immunodeficiency virus (HIV)-tested and asked whether they were circumcised (n = 1678; 84.2%). Participants aged 15–44 years were interviewed (n = 636; 79.7%), and 399 (50.0%) were gynaecologically examined to screen for STIs and determine the FGC status.
Results At a mean age of 9.6 years, 45.2% reported being circumcised. In the age-group 15–44 years, 65.5% reported being cut, while FGC was observed in 72.5% and categorized as clitoridectomy or excision. The strongest predictors of FGC were ethnicity and religion, i.e. being a Protestant or a Muslim. FGC was not associated with HIV infection, other STIs or infertility. A positive, non-significant association between FGC and bacterial vaginosis was found with a crude odds ratio of 4.6. There was a significant decline of FGC over the last generation. An inconsistency between self-reported and clinically determined FGC status was observed in more than one-fifth of the women.
Conclusion The data indicate that both women and clinicians might incorrectly report women's circumcision status. This reveals methodological problems in determining women's circumcision status in populations practising the most common type of FGC. The positive association between FGC and bacterial vaginosis warrants further investigation.
Female genital cutting (FGC), or non-therapeutic surgical modification of female genitalia, is an ancient tradition in large parts of Africa. While the practice seems to be on the decline in some areas (Nkwo & Onah 2001; Msuya et al. 2002a; Snow et al. 2002), Leonard (2000) has shown that in the south of Chad, the practice has recently been adopted. In East Africa, FGC is practised in central Tanzania and northwards (Hosken 1993; Toubia 1995; Obermeyer 1999; Shell-Duncan & Hernlund 2000). The practice is usually classified in two broad categories: clitoridectomies and infibulations (Toubia 1994). A World Health Organization (WHO) (1996) technical working group divided the practice into four types: clitoridectomies (I), excision (II) where both the clitoris and part or all of the labia minora are cut, infibulation (III) where of all of the external genitalia are cut with stitching of the vaginal opening, and non-classifiable forms like pricking, piercing, etc. (IV). The first two types affect up to 80–85% of women who are circumcised (Toubia 1994; Dorkenoo 1996).
It has been difficult to reach a consensus regarding the naming of this practice (Obermeyer 1999; Gruenbaum 2001). We consider FGC so far the best descriptive term, at least for type I and II of female genital surgeries (Eliah 1996). In communities where traditional surgical modifications of female genitalia are commonplace, it is often referred to as female circumcision (Toubia 1995), and we will use this term too.
Female genital cutting is one of the harmful practices affecting the health of women and children (Koso-Thomas 1987; Dorkenoo & Elworthy 1992; Toubia 1994; American Medical Association & Council on Scientific Affairs 1995; WHO 1996; Wright 1996). However, the practice is still highly valued in a number of local communities. According to Snow (2001), the focus of the public health attention has been more political than scientific. During the 1990s, several countries have included questions regarding FGC in their national Demographic and Health Surveys (DHS) (Carr 1997). Prevalence estimates of FGC have almost entirely been based on self-reported circumcision status. Few studies base their prevalence estimates on clinically determined FGC status and self-report and thus provide information regarding the accuracy of self-report vs. clinical observation (Odujurin et al. 1989; Asali et al. 1995; Population Council 1996; Adinma 1997; Morison et al. 2001; Snow et al. 2002).
Research on the consequences of FGC for reproductive health and sexuality is seen as insufficient (Dorkenoo 1996; Obermeyer 1999; Shell-Duncan & Hernlund 2000; Snow 2001). It has been hypothesized that female circumcision is a risk factor for human immunodeficiency virus (HIV) infection (Kun 1997). Possible modes of infection are (i) use of insufficiently sterilized equipment during the cutting procedure, (ii) sexual intercourse before the wound is healed, and (iii) easy bruising of the genital mucosa after FGC, enhancing susceptibility to HIV infection. However, the relationship between FGC and HIV infection is still largely unexplored (Kapiga et al. 2002; Msuya et al. 2002a). Taking into account the high prevalence rates of FGC and STIs in the same areas of Africa, it is remarkable that the consequences of FGC for STIs only recently have gained research interest (Jones et al. 1999; Morison et al. 2001; Msuya et al. 2002a; Okonofua et al. 2002).
In Tanzania, a national DHS conducted in 1996 reported the highest prevalence of circumcision in the north; i.e. 81% in Arusha region and 37% in Kilimanjaro region (Bureau of Statistics 1997). The common types of female circumcision practised were clitoridectomy and excision. The objectives of this study were (i) to determine the prevalence and type of FGC in a rural village in Kilimanjaro region in northern Tanzania, (ii) the associated demographic factors and (iii) to examine if FGC was associated with higher risk of HIV infection, other sexually transmitted infections (STIs) and infertility. In addition, we investigated the consistency between self-reported and clinically observed FGC.
This study was part of a larger community-based, cross-sectional survey. It was conducted in a rural village in the Kilimanjaro region of the Tanzanian highlands situated at the foot of Mount Kilimanjaro in the Kahe ward 30 km south of the regional centre, Moshi town. It was carried out at a temporary survey centre established in the village in 1991–92. Detailed study methods and results of the HIV/STI/RTI (reproductive tract infection) investigations are presented elsewhere (Klouman et al. 1997; Klouman 2004). All age-groups in the entire village population were eligible both for HIV testing and a short medical history including questions regarding circumcision status and age of circumcision. Participants aged 15–44 years were also interviewed using a standardized questionnaire assessing demographic information, past and present symptoms and signs of STIs, and sexual behaviour. Women in this age-group were furthermore offered screening for STIs. Trained nurses interviewed the participants, and medical doctors performed the gynaecological examinations. The clinician had access to the interview before the performance of the gynaecological examinations.
A total of 1993 girls and women in all age-groups were registered, of whom 1678 participated (84.2%). Among the 798 eligible women in the age-group 15–44 years, 79.7% (n = 636) participated in the interview and consented to provide a blood sample. A total of 50.0% (n = 399) of the eligible women aged 15–44 years agreed to a pelvic examination. Reasons given for not consenting to the pelvic examination were: being a virgin (29.2%), presently pregnant (11.9%), disabled (1.7%), or having their menstrual period (17.2%); the remaining 40.0% did not provide any reason.
Human immunodeficiency virus-1 screening was conducted with recombinant enzyme-linked immunosorbent assay (ELISA; Wellcome Diagnostics, Dartford, UK). All positive or doubtful sera were tested twice with ELISA, and confirmed by Western blot (WB; Organon, Epitope, Beaverton, OR, USA). Syphilis serology included a rapid plasma reagin test (RPR; Syphacard-R, Murex Diagnostics Ltd, Dartford, UK), and a Treponema pallidum haemagglutination test (TPHA; Wellcosyph HA, Murex Diagnostics Ltd). If both RPR and TPHA were reactive, this was considered to be active syphilis. A wet mount examination of vaginal fluid in saline was performed for the detection of Trichomonas vaginalis, pus and clue cells. Vaginal fluid in 10% potassium hydroxide was examined for fishy smell and yeast, and pH of the vaginal fluid was recorded. Amsel et al.'s criteria (1983) for the diagnosis of bacterial vaginosis (BV) were followed. Pelvic inflammatory disease (PID) was diagnosed by the clinical criteria referred to by Dallabetta et al. (1998). Infertility was defined as not being able to become pregnant within a year in women living with their partners. In the analyses, primary (3.1%) and secondary infertility (7.2%) were combined to the variable ‘infertility’. Endocervical swabs were obtained for the culture of Neisseria gonorrhoea, and for Chlamydia enzyme immunoassay (EIA) antigen testing (Wellcome Diagnostics).
Only three cases of gonorrhoea among women were identified, and gonorrhoea was therefore not included in the further analyses. We did not test for herpes simplex virus 2 (HSV2) infection.
Research and ethical clearance was obtained both from the Ministry of Health in Tanzania and the Norwegian Committee for Medical Research Ethics. Participation was based on oral informed consent. Parents gave consent on behalf of children up to the age of 14 years, and provided information on the medical history of the under-aged girls.
All analyses were conducted using SPSS (version 11). Chi-square test, Fisher's exact test and logistic regression analyses were used when testing the relationships between categorical variables. Selection criteria for inclusion of independent variables in the multivariate analyses were P-values <0.2 in bivariate analyses. P-values, odds ratios (OR) and 95% confidence intervals (CI) are provided. anova was used for comparing mean values. Kaplan–Meier analysis was used to calculate the median age when 50% of all females in the entire village population were circumcised. The kappa value was calculated for comparable categorical data obtained by self-report and by observation, as a measure of agreement between the two methods (Altman 1991).
The age and self-reported prevalence of all girls and women who participated in this study are presented in Table 1. Approximately half of the participating women, aged 15–44 years, were Muslims and the other half Christians. While a total of 64 ethnic groups were registered in the survey village, 73.0% of the women aged 15–44 years belonged to one of three local ethic groups: Chagga, Pare and Kahe. The education level among women in this village was generally low; only 3.8% had secondary school or higher education. The demographic characteristics of the 15–44-year age-group are presented in Table 2 for all interviewed, and for the subgroups who consented to and refused gynaecological examination.
Table 1. Age-specific prevalence (%) of self-reported female genital cutting (FGC) among all participating girls and women in a rural population, Kilimanjaro, Tanzania (n = 1660)
Table 2. Demographic characteristics among the interviewed women (all); and those who accepted and refused gynaecological examination (gyn. exam.) aged 15–44 years, in a rural population, Kilimanjaro, Tanzania*
|Average age||27 ± 8.0|| ||29 ± 7.4|| ||23.8 ± 7.9|| ||<0.001|
| No formal||59||9.4||48||12.1||11||4.7||0.001|
| Stand. 1–4||123||19.6||90||22.7||33||14.2|
| Stand. 5–7||423||67.2||246||62.0||177||76.3|
| Low. Sec.||18||2.9||9||2.3||9||3.9|
| High. Sec.||4||0.6||3||0.8||1||0.4|
|Self-reported female genital cutting|
Prevalence and type of FGC
Of the 1660 participating girls and women who answered the question regarding FGC, 45.2% reported being circumcised at a mean age of 9.6 years (±5.1). The prevalence increased with age (Table 1). The median circumcision age when 50% of all females in the entire village population were circumcised was 15 years, and by the age of 8 years, 25% were circumcised. Women ≥35 years were significantly more at risk of being circumcised than women aged 20–34 years (P = 0.001, OR = 1.7, 95% CI 1.2–2.3), indicating a change in practice. In the age-group 15–44 years, 65.5% self-reported being circumcised, while the observed prevalence of FGC among women who consented to gynaecological examination was 72.5% (n = 399). Among those who were clinically determined to be circumcised, clitoridectomy was found in 44.9% (type I), excision in 53.1% and absence of most of the external genitals in 2.1% (type II). Infibulation or removal of the foreskin only, was not found. Clitoridectomy was often partial, leaving a part of the proximal clitoris behind both in the type I and II cutting. Total clitoridectomy without removal of parts or most of the inner lips was not registered.
FGC and demographic characteristics
The mean age of circumcision varied significantly according to ethnicity among women aged 15–44 years. The Chaggas reported being circumcised at an average age of 15.5 years (±4.5). The corresponding age for the two other local ethnic groups were: Kahes 10.6 years (±3.8) and Pares 7.8 years (±3.7). The reported mean age for circumcision among the non-local ethnic groups was 9.8 years (±4.8) (P < 0.001).
The prevalence of self-reported (n = 636) and clinically determined FGC (n = 399) in different groups of the population and the associations with demographic characteristics, sexual behaviour and HIV/STIs/RTIs in crude bivariate analyses are presented in Table 3. Religion, i.e. being Protestant or Muslim; ethnicity, i.e. belonging to two of the three local ethnic groups, Pares and Kahes; and being ever-married were significantly associated with both self-reported FGC and with clinically determined FGC. Education showed an inverse relationship with FGC in both groups, without reaching significance. Significantly more women with no or low education (less than Standard 4) participated in the gynaecological examination (P < 0.001).
Table 3. The prevalence of female genital cutting (FGC) and association with demographic characteristics, no. of sexual partners, HIV/STIs/RTIs among interviewed (n = 636) and clinically examined (n = 399) rural women, aged 15–44 years, Kilimanjaro, Tanzania*
| 25–34||207||66.2||1.4 (0.9–2.0)||152||73.7||1.5 (0.9–2.5)|
| 35–44||154||76.6||2.3 (1.5–3.6)||116||79.3||2.1 (1.2–3.7)|
| Ever married||453||71.7||2.5 (1.7–3.6)||329||74.8||2.0 (1.1–3.4)|
| Protestants||148||72.3||2.3 (1.5–3.7)||100||77.0||2.1 (1.1–3.7)|
| Muslims||297||69.7||2.1 (1.5–3.1)||171||77.7||2.2 (1.3–3.6)|
| Not local||169||42.0||1.0||103||58.3||1.0|
| Chagga||175||61.7||2.2 (1.4–3.4)||116||62.1||1.2 (0.7–2.0)|
| Pare||240||80.8||5.8 (3.7–9.1)||145||88.3||5.4 (2.8–10.2)|
| Kahe||43||93.0||18.4 (5.5–61.8)||29||86.2||4.4 (1.4–13.5)|
| Secondary or higher||24||58.3||1.0||13||69.2||1.0|
| No formal||59||81.4||3.1 (1.1–8.8)||48||83.5||2.2 (0.5–9.0)|
| Standard 1–4||123||68.3||1.5 (0.6–3.8)||90||71.1||1.1 (0.3–3.9)|
| Standard 5–7||422||63.5||1.2 (0.5–2.7)||243||71.2||1.1 (0.3–3.7)|
|Sexual partners last 12 months|
| One partner||437||68.4||1.9 (1.2–2.8)||311||73.0||1.9 (0.9–4.2)|
| Two partners||43||74.4||2.5 (1.2–5.4)||32||78.1||2.5 (0.8–7.7)|
| Three or more partners||24||66.7||1.7 (0.7–4.3)||22||77.3||2.4 (0.7–8.3)|
| HIV neg.||599||66.1||1.0||371||72.8||1.0|
| HIV pos.||27||66.7||1.0 (0.5–2.3)||21||76.2||1.2 (0.4–3.4)|
| Syphilis neg.||612||65.8||1.0||378||72.8||1.0|
| Syphilis pos.||16||56.3||0.7 (0.2–1.8)||15||66.8||0.7 (0.3–2.2)|
| Not infertile||527||66.7||1.0||342||71.6||1.0|
| Infertile||52||71.1||1.2 (0.7–2.3)||48||77.1||1.3 (0.7–2.7)|
| No PID|| || || ||351||72.4||1.0|
| PID|| || || ||42||73.8||1.1 (0.5–2.2)|
| Chlam. neg.|| || || ||363||72.2||1.0|
| Chlam. pos.|| || || ||27||70.4||0.9 (0.4–2.2)|
|Trichomonas vaginalis vaginitis|
| Trich. neg.|| || || ||283||73.5||1.0|
| Trich. pos.|| || || ||94||73.4||1.0 (0.6–1.7)|
|Candida albicans vaginitis|
| Candida neg.|| || || ||357||73.1||1.0|
| Candida pos.|| || || ||17||82.4||1.7 (0.5–6.1)|
| BV neg.|| || || ||381||71.9||1.0|
| BV pos.|| || || ||13||92.3||4.6 (0.6–35.5)|
In the multivariate analyses presented in Table 4, we repeated these analyses with dichotomized variables; ethnic group: local vs. other tribes; education: no formal vs. any education; and kept single vs. ever married, religion and sexual behaviour unchanged. The strongest demographic predictors of circumcision both regarding self-reported and clinically determined were ethnicity, i.e. belonging to local tribes, and religion, i.e. being Protestant or Muslim. In the multivariate analysis, having no education was also significantly associated with FGC.
Table 4. The association of female genital cutting (FGC) with demographic characteristics and number of sexual partners among interviewed (n = 636) and the association of FGC with demographic characteristics, number of sexual partners and bacterial vaginosis (BV) among clinically examined (n = 399) rural women, aged 15–44 years, Kilimanjaro, Tanzania*
| Ever married||453||2.2 (1.3–3.8)||327||1.6 (0.8–3.1)|
| Protestants||148||2.2 (1.3–3.6)||100||1.9 (1.0–3.5)|
| Muslims||292||3.0 (1.9–4.6)||168||2.6 (1.5–4.6)|
| Not local||169||1.0||101||1.0|
| Local||454||5.1 (3.3–7.7)||289||3.0 (1.8–5.0)|
| Any education||564||1.0||343||1.0|
| No formal education||59||2.5 (1.2–5.4)||47||2.5 (1.0–6.3)|
| BV neg.|| || ||377||1.0|
| BV pos.|| || ||13||3.7 (0.5–31.7)|
|Sexual partners last 12 months|
| One partner||435||1.0 (0.6–1.8)||307||1.3 (0.6–3.2)|
| Two partners||43||2.1 (0.9–5.1)||32||2.1 (0.6–7.0)|
| Three or more partners||24||1.3 (0.5–3.7)||22||2.0 (0.5–7.7)|
HIV, STIs and infertility
For the full sample, we have data on both self-reported circumcision and HIV infection in 81.4% of the participants (n = 1623). In the crude analysis, the association between HIV infection and FGC was significant at the P = 0.05 level, OR = 2.0 (95% CI 1.0–4.1). However, both HIV infection and FGC were highly significantly associated with age at P < 0.001 for both variables. Correcting for age, the significant association between HIV and FGC disappeared (P = 0.4, OR = 1.4, 95% CI 0.6–3.0). In the age-group 15–44 years, in which most of the HIV infections occurred, there was no association between HIV infection and self-reported FGC, neither in the crude nor the age-adjusted analyses (Table 3).
The diagnosis of syphilis was determined by serological tests, allowing a larger sample to be analysed in the age-group 15–44 years (n = 636) than only the group who consented to gynaecological examination (n = 399). However, there was neither an association with self-reported FGC in the crude (Table 3) or in age-adjusted analyses, nor with clinically determined FGC (Table 3). This group of women had also answered the question regarding infertility, and no association with self-reported or clinically determined FGC was found (Table 3).
Among the women who underwent gynaecological examination (n = 399), 46.9% had an ongoing reproductive tract infection or sexually transmitted disease. Those who tested positive and negative for the different STIs/RTIs/infertility and prevalence of FGC in these groups are presented in Table 3. We did not find any significant difference regarding infertility, HIV or any other infections among clinically determined circumcised and uncircumcised women. We did, however find a positive, non-significant association between BV and circumcision. BV was thus the only clinical condition that merited being included in the multivariate logistic regression analyses on factors predictive of FGC in this population.
Participants, aged 15–44 years, were asked about the number of sexual partners they had during the last 4 weeks, 12 months and 5 years, and mean numbers of partners reported among the sexually active were 1.0 (±0.2); 1.2 (±0.5); 1.5 (±0.9) for these time periods, respectively. In bivariate analyses, the reported numbers of sexual partners the last 12 months were significantly associated with self-reported FGC (Table 3), marital status, education, HIV-infection, PID, infertility, and trichomoniasis. This sexual behaviour variable was therefore used in the multivariate analyses to control for possible confounding. Infertility was significantly associated with all sexual behaviour variables. T. vaginalis-caused vaginitis was also significantly associated with the number of sexual partners the last 5 years, as were chlamydial infection and Candida albicans-caused vaginitis. None of the sexual behaviour variables were significantly associated with clinically observed FGC (Table 3) syphilis or BV, and sexual behaviour did not influence the positive association between FGC and BV. The outcome of multivariate logistic regression analyses on the association between FGC and demographic variables and BV was not significantly altered by the inclusion of ‘sexual behaviour’.
Consistency between self-reported and observed circumcision
Data regarding circumcision are available both from interview and clinical examination in 396 cases, i.e. in 62.4% of participating women in the age-group 15–44 years. When self-reported circumcision was compared with clinical findings, agreement was found in 77.5% (n = 307). However 14.1% (n = 56) reported not to be circumcised and 8.3% (n = 33) answered that they were circumcised, while the contrary was found during the gynaecological examination. The kappa value for the agreement between self-reported and observed circumcision was 0.47. Among the 33 women who reported to be circumcised, but were not found circumcised by the clinicians, only three women did not report the age of circumcision. The reported mean age for circumcision for these women was 12.7 years (±5.4). When testing for demographic differences between the groups that ‘under-reported’, ‘over-reported’ and the group with agreement between self-report and observation, we found statistically significant differences regarding ethnicity only (P = 0.02). In the group ‘over-reporting’, there were more Chaggas and Kahes, the ethnic groups that circumcise their daughters at an advanced age (15.5 and 10.6 years, respectively). In the ‘under-reporting’ group, there were more Pares and non-local ethnic groups, which circumcise at an earlier age (7.6 and 9.8 years, respectively).
In this study, nearly three-quarters of rural Tanzanian women were circumcised. A decline in the prevalence of FGC was observed over the last generation. Ethnicity and religion were strong predictors of circumcision, while FGC was not associated with increased risk of infertility, HIV infection or other STIs, although a non-significant positive association was found for BV. There was an inconsistency between self-reported and clinically observed FGC status in more than one-fifth of the women.
Prevalence and type of FGC
We report a higher prevalence of FGC, 66% of self-reported circumcision in the age-group 15–44 years, than presented in any other study from the Kilimanjaro region (Bureau of Statistics 1997; Chuguli & Dixey 2000; Msuya et al. 2002a). This might reflect a decline in the prevalence of FGC in the urban part of Kilimanjaro region, and that the prevalence is higher in rural areas, although it might be on the decline among the younger generations as seen in our study. Other studies also indicate a shift in practice (Adinma 1997; Mbackéet al. 1998; Nkwo & Onah 2001; Snow et al. 2002), while the national DHS in many other countries in sub-Saharan Africa do not provide any evidence of decline (Carr 1997).
In this study, circumcision was characterized as mainly belonging to type I and II based on the WHO classification (1996). Shell-Duncan and Hernlund (2000) emphasize that a sharp distinction between clitoridectomy (type I) and excision (type II) is difficult to draw. Our observations are consistent with their description.
FGC and demographic characteristics
Circumcision age differed significantly between the various ethnic groups. Obermeyer (1999) states that within each ethnic group, circumcision is performed at a certain age with relatively narrow age limits, and is usually performed on pre-pubertal girls. However, one of the local ethnic groups in this study, the Chaggas, circumcise girls late, at a mean age of 15 years. Age of female genital surgeries varies across the African continent from infancy to adulthood (Mbackéet al. 1998; Orubuloye et al. 2000; Shell-Duncan & Hernlund 2000).
The strongest predictor of FGC in this study was ethnicity, i.e. belonging to local ethnic groups. Obermeyer (1999) points out that the greatest variation in the prevalence and types of genital surgeries appear to be associated with regional and ethnic differences. According to Caldwell et al. (1997), circumcision was carried out in most sub-Saharan African societies with major ritual ceremonies. However, in some societies, a girl's circumcision is not regarded to be of particular ceremonial importance. In Somalia, it is a matter of concern mainly to the women of the family, and has rather an aura of secrecy and privacy (Talle 1993). Circumcision in Kahe was traditionally part of an initiation ceremony, also including teaching, preparing girls for marriage (Manongi 1998). However, both the ceremonies and the camps have disappeared more rapidly than circumcision itself.
Religion was significantly associated with FGC in our study. Catholics practice FGC less than Protestants and Muslims. Snow et al. (2002) also found a diversity regarding FGC within Christian groups with the prevalence of FGC being particularly high among Pentecostals. Although there is no scriptural support for FGC either in the Koran or the Bible, women who practice circumcision often interpret this as a religious custom and link it to their religious identity (Johnson 2000).
In our study, having no formal education was significantly associated with increased rate of FGC in the multivariate analyses. A similar relationship between FGC and education has been observed also in other studies, both in Tanzania (Msuya et al. 2002a; Carr 1997) and elsewhere in Africa (Adinma 1997; Mbackéet al. 1998; Obermeyer 1999; Shell-Duncan & Hernlund 2000; Snow et al. 2002). In the demographic literature, the inverse relation between schooling and indicators of high fertility and ill-health is one of the most robust correlations (Obermeyer 1999).
HIV, STIs and infertility
Female genital cutting has been hypothesized to be a risk factor for HIV infection (Kun 1997). We did not find circumcised women to be more at risk for HIV infection than uncircumcised women. Our study was performed in the first phase of the HIV epidemic, and only one under-aged girl was HIV-infected (Klouman et al. 1997), the possibility of being infected during the cutting procedure is therefore nearly ruled out. This possibility will increase as more children are HIV-infected in communities practising circumcision of girls. Nevertheless, our finding is consistent with the results of two recent studies from Kilimanjaro (Kapiga et al. 2002; Msuya et al. 2002a). We found a positive, non-significant association between BV and FGC. BV has been associated with increased risk of HIV infection (Sewankambo et al. 1997; Kapiga et al. 2002; Msuya et al. 2002b). In a population-based study from rural Gambia, Morison et al. (2001) found a significantly higher prevalence of BV and HSV2 infection among circumcised compared with uncircumcised women. In a study from Nigeria (Okonofua et al. 2002), circumcised women were more than four times more likely to have genital ulcer disease (GUD), compared with uncircumcised women. HSV2 is presently considered an important risk factor for HIV infection in sub-Saharan Africa, especially for women (Weiss et al. 2001). The associations between FGC and BV and HSV2/GUD may indicate that circumcised women may be more susceptible to HIV infection, and further studies are needed to ascertain this.
Furthermore, no statistically significant association between FGC, infertility or any of the infections investigated was seen in our study. Shandal (1976) found that PID was three times more common in patients with infibulation compared with patients with sunna circumcision (type I). Okonofua et al. (2002) report significantly more episodes of lower abdominal pain among women with type I and II circumcision, compared with uncircumcised women. This might indicate a higher prevalence of PID in cut women. Msuya et al. (2002a) found that infertility rates were similar in circumcised and uncircumcised women. This supports the findings of Larsen and Sharon (2000), that circumcised women did not have significantly different odds of infertility from uncircumcised women in three African countries, including Tanzania.
De Silva (1989) found a significantly higher prevalence of candidiasis among circumcised women compared with controls. In two studies, RTIs were assessed syndromically; i.e. based on women's self reporting of symptoms in peri-urban Nigeria (Okonofua et al. 2002) and the observation of clinical signs associated with RTIs in rural Burkina-Faso (Jones et al. 1999). In these clinic-based studies, vaginal discharge was more common in circumcised compared with uncircumcised women. It is well documented that the symptom of vaginal discharge is more a symptom of vaginal than cervical infection (Steen & Dalabetta 1999). Thus, studies using standard laboratory testing and studies assessing the syndrome of vaginal discharge have documented higher risk of BV and vaginitis in circumcised compared with uncircumcised women, although not consistently (Msuya et al. 2002a). As results are not conclusive, the documented associations between FGC and BV and vaginal infections warrant further investigations.
Consistency between self-reported and observed circumcision
Only a fair relationship between self-report and clinically observed circumcision was seen in this study. In studies investigating the validity of self-report vs. clinically observed circumcision status, the clinically determined circumcision status is regarded as the gold standard (Odujurin et al. 1989; Population Council 1996; Adinma 1997; Morison et al. 2001; Snow et al. 2002). This might well be the case when circumcision is substantial. Taking into consideration the wide variation that naturally exists regarding the width of women's inner genital lips, clinicians might evaluate the circumcision status wrongly when cuts are minimal. Most women who were determined by the clinicians not to be circumcised provided an age for their circumcision. The reported ages were so high that we expect these women to remember their circumcision, and it fits also with a higher representation of Chaggas and Kahes, who circumcise girls late. Because of the social pressure to be circumcised, some might also have reported falsely that they were circumcised.
Clinically, it is probably easier to overlook a minor circumcision than to incorrectly classify a woman to be circumcised. The clinicians reported 56 women to be circumcised, while these women reported the contrary. This group had an overrepresentation of Pares and non-local ethic groups who tend to circumcise girls during their childhood. Especially if the cuts are small and they were performed at an early age, the women may not remember their circumcision.
There may also be reasons why women consciously under-report their circumcision status. Manongi (1998) undertook a qualitative study to understand the social and cultural factors motivating the continuation of FGC. The most common reason given for the circumcision practice was to strengthen the bond with the ancestors. Family members took illness or deaths within the family as disapproval of misconduct by their forefathers. In such situations, any uncircumcised woman in the family would be urged to be circumcised, to retrieve the relationship with her ancestors. In this cultural environment, some might deny this practice in interviews with people representing modern medicine. Among initiated women, it was traditionally taboo to tell the ‘secret’ behind circumcision not only to men, but also to women and girls who were not circumcised (Manongi 1998). As the government was regarded to oppose female circumcision, people opted to do it secretly. Both the old and this new tradition to keep female circumcision a secret, may add to the unwillingness to reveal one's circumcision status to strangers who might be associated with the government.
In studies comparing data on circumcision assessed both by self-report and clinical examination, the agreement varies: 97% in the community-based study from Gambia (Morison et al. 2001); and 94% in the Egyptian fertility care study (Population Council 1996). In Nigerian studies, the agreement was 79% and 57% (Adinma 1997; Snow et al. 2002), respectively, while Odujurin et al. (1989) found that 54.6% of women claimed to be circumcised while examination revealed that 75.5% were cut. In Nigeria, girls are usually circumcised during infancy. This might explain why many do not know their circumcision status. In future surveys on FGC, inclusion of a ‘do not know’ option should be considered, and further research is required to assess inter-observer reliability especially regarding type I FGC.
This study calls for caution when relying on women's self-reported FGC status as the exact FGC prevalence in a survey population, and questions that the clinicians’ determination of FGC status should be considered the gold standard in populations which practice minor circumcisions. The strongest predictors of FGC in this study was ethnicity and religion, while FGC was not associated with increased risk of infertility, HIV infection or other STIs/RTIs, although a non-significant positive association was found for BV. As BV is considered to be a risk factor for HIV infection, cut women may be more susceptible to HIV infection, and this relationship should be further explored.
We thank the women of the survey village and their village government for support and participation in the project. We further thank the field team; the MUTAN (Mwradi wa Ukimwi wa Tanzania na Norway – The Tanzanian Norwegian AIDS Project) office staff; and the laboratory research staff at Kilimanjaro Christian Medical Centre (KCMC) with the heads of laboratories; Drs N.E. Sam and W.M.M.M. Nkya; and Dr E.J. Masenga, dermato-venereologist at KCMC. We are grateful to the District and Regional Health Authorities in Kilimanjaro and the Ministry of Health in Tanzania for supporting the project. The study was supported by grants from the Norwegian Agency for Development Cooperation through the Tanzanian-Norwegian AIDS Project (MUTAN), the Norwegian Research Council, and the Norwegian University Committee for Development Research and Education.