Rachel Snow Reproductive Health, Department of Tropical Hygiene and Public Health, University of Heidelberg, Im Neuenheimer Feld 324, 69120 Heidelberg, Germany. Fax: +49-6221-565 039; E-mail: email@example.com
Despite growing public resistance to the practice of female genital cutting (FGC), documentation of its prevalence, social correlates or trends in practice are extremely limited, and most available data are based on self-reporting. In three antenatal and three family planning clinics in South-west Nigeria we studied the prevalence, social determinants, and validity of self-reporting for FGC among 1709 women. Women were interviewed on social and demographic history, and whether or not they had undergone FGC. Interviews were followed by clinical examination to affirm the occurrence and extent of circumcision. In total, 45.9% had undergone some form of cutting. Based on WHO classifications by type, 32.6% had Type I cuts, 11.5% Type II, and 1.9% Type III or IV. Self-reported FGC status was valid in 79% of women; 14% were unsure of their status, and 7% reported their status incorrectly. Women are more likely to be unsure of their status if they were not cut, or come from social groups with a lower prevalence of cutting. Ethnicity was the most significant social predictor of FGC, followed by age, religious affiliation and education. Prevalence of FGC was highest among the Bini and Urhobo, among those with the least education, and particularly high among adherents to Pentecostal churches; this was independent of related social factors. There is evidence of a steady and steep secular decline in the prevalence of FGC in this region over the past 25 years, with age-specific prevalence rates of 75.4% among women aged 45–49 years, 48.6% among 30–34-year olds, and 14.5% among girls aged 15–19. Despite wide disparities in FGC prevalence across ethnic, religious and educational groups, the secular decline is evident among all social subgroups.
An estimated 130 million girls and women alive today have undergone some form of genital cutting or circumcision. Circumcision among both males and females is an old practice referred to in pharaonic writings, but while male circumcision consists of cutting foreskin, even the mildest form of female genital cutting (FGC) includes removal of the prepuce, and many forms of FGC go further. FGC practice varies by ethnicity and region in the type of practitioner employed, age at cutting, association with social and religious rituals, and the form of cutting. WHO (1997) has classified the forms of cutting into Type I: excision of the prepuce, with or without excision of part or all of the clitoris; Type II: excision of the clitoris with partial or total excision of the labia minora; Type III: excision of part or all of the external genitalia and stitching/narrowing of the vaginal opening (infibulation); and Type IV: other forms. Ambiguous types, or combinations of the above, are also reported.
Prevalence data by type of FGC, or details about the social clustering of the practice are available for only a few regions (De Silva 1989; EFCS 1996; Adinma 1997; Mbacke et al. 1998; Jones et al. 1999; Morison et al. 2001), and authors have bemoaned the paucity of good data on determinants or consequences of FGC (Obermeyer 1999). Since 1990 Demographic and Health Surveys (DHS) have added a module on FGC for selected countries, and these data are providing the first comparable estimates of prevalence. The emerging DHS data already underscore variations in prevalence between regions within the same country, and hence the dangers of generalizing from small samples. A summary of prevalence data from recent DHS is provided in Table 1.
Table 1. Self-reported FGC prevalence data from Demographic and Health Surveys (DHS) (includes all FGC Types)
There are several measurement gaps that the DHS cannot address, such as the prevalence of FGC by clinically affirmed type. Assessing the prevalence of FGC by type requires a clinical examination, training and interviewing women about classification, or research with FGC practitioners. From a combination of such studies Toubia (1994) offers a rough estimate that 15–20% of all FGC is Type III, apparently localized in North-east Africa, particularly Somalia, Sudan, Djibouti and Mali. Types I and II appear to comprise the majority of FGC, but studies providing clinical verification of types are rare, and available estimates have wide confidence intervals. Given the suspected disparities in both health and social consequences of FGC by type, reliable prevalence data are needed.
Likewise, there is reason to question the reliability of self-reporting, upon which the DHS is based. Self-reporting has only been evaluated in two well-designed studies to date. A study in Egypt under the auspices of the DHS reported that 94% of women accurately reported their status, as verified by physical examination (EFCS 1996). While this study has been widely quoted, and indeed appears reassuring, a smaller study from South-east Nigeria found self-reporting accurate among only 57% of women (Adinma 1997). A third study, also from Nigeria, checked only the validity of positive FGC status and found 75% of these cases to be accurate (Odujinrin et al. 1989), but the study failed to report the proportion of false positive cases. Notably, in the DHS study in Egypt, there was no indication that any interviewee was unable to answer the question, or did not know her status, while uncertainty was relatively common in the Nigerian studies. These variant results raise questions about the basis of self-reporting that warrant further assessment.
Nigeria is the most populous country in sub-Saharan Africa, with approximately 120 million inhabitants. Twenty-five percent of reproductive age women report having undergone FGC in the most recent DHS (1999), but validity of self-reporting was not assessed. A movement for the eradication of FGC has been underway in Nigeria for more than a decade, but documentations of social correlates are limited to case reports and small studies, and evidence on trends in incidence are not available. Case studies suggest that all three types of FGC are practiced, with disparities by region and ethnicity. Indeed, Nigeria is characterized by social diversity: it has more than 250 ethnic groups, women’s educational attainment and status vary dramatically across the country, and there is considerable religious diversity. Approximately 50% of the population are Muslims, predominating in the north. The traditional Christian churches (Catholic, Methodist, Anglican) are well-represented, but so too are traditional indigenous religions, and more recently, the Pentecostal churches. Despite common assumptions regarding the association between Islam and FGC, closer study reveals that the role of religion in FGC practice is not at all consistent (Obermeyer 1999), and without theological foundation. Within southern Nigeria, little is known about the association between religious culture and FGC, or whether FGC is diminishing or increasing in certain social groups. To address such questions, we report on the validity of self-reporting, and the prevalence, trends and social correlates of FGC among women attending public health services in Edo State, Nigeria. Even in this geographically small region, we find the prevalence of FGC varies considerably by ethnic and religious affiliation. Despite this diversity, there is evidence of a consistent secular decline in FGC among all social and ethnic groups.
A cross-sectional study was conducted between August 1998 and March 1999 among women aged 15–49 years attending antenatal and family planning clinics in three large urban and peri-urban hospitals in Edo State, Nigeria. Successive clinic attendees were approached for consent to join the study, and then interviewed by nurse-midwives about their social and demographic background, reproductive history, experience of genital cutting (FGC), and their views on FGC. The interview was followed immediately by a routine vaginal examination conducted by an obstetrician/gynaecologist (OB/GYN). The collaborating OB/GYN was unaware of the participants’ self-reported FGC status, but provided a verification of the status and type of FGC for each woman.
Two of the hospitals, University of Benin Teaching Hospital (UBTH), and Central Hospital, are located in Benin City, the largest city in Edo State. The third, Specialist Teaching Hospital, is located in Irrua, a large Edo State town about 70 km north of Benin City. UBTH and Central Hospital serve a more urban population, while Specialist Hospital in Irrua was chosen to increase coverage of the peri-urban population. All three hospitals were chosen on the basis of their large size, their central locations, and their status as public hospitals serving less affluent populations. Antenatal services are attended by more than 85% of women in this region of Nigeria [Nigeria, Federal Office of Statistics and IRD/Macro International Inc. (Nigeria & IRD) 1992].
The OB/GYN departments in each hospital collaborated in the study, providing two female staff members (nurse-midwives) to conduct the interviews, and one physician (an OB/GYN) to conduct the physical examination. These research collaborators (six nurses/midwives and three physicians) received initial training of approximately 3 h, followed by practice surveys, role-plays, and a follow-up training session of 1.5 h. The questionnaire and examination procedures were pilot-tested with 50 women attending antenatal services at UBTH and followed by revisions. Age was assessed from two questions, one on current age, and a second on year born; internal consistency was checked against additional questions concerning age at marriage, year married, age at first birth, and year of first birth. Education was measured from a question on completed years of schooling. Religion and ethnicity were determined by open-ended direct questions, and coded from a comprehensive list on the interview sheet.
Formal enrolment of subjects began in August 1998 and continued for 7 months, through March 1999. Completed questionnaires were collected twice monthly. Data collection was closed in March 1999 at 1861 surveys, just short of the intended sample size of 2000, because a resident doctors’ strike was leaving the hospitals understaffed, and collaborating senior physicians were over-worked. Of the 1861 women approached, 10 refused to participate, and 17 were outside the targeted age range of 15–49 years. As it became clear in the analysis that age was a critical covariate of our key outcomes (i.e. the propability of being cut, the validity of self-reporting), an additional 125 records were dropped from the analysis because they either provided no information regarding age or year of birth (n=117), or they provided discordant information on age that could not be verified from other responses (n=8). The remaining 1709 records are the basis of analyses presented in this paper.
Subjects were informed of the purpose and format for the study when approached in the waiting room, and assured both that the data was confidential, and refusal would not compromise clinical care. In this manner verbal approval was obtained prior to initiation of the survey. Ethical permission for the study was given by the Committee for Ethical Review at the University of Benin.
We summarized the associations between FGC status and independent variables with odds ratios (OR) and 95% confidence intervals. The association between age and FGC was assessed by stratified analysis using 5-year birth cohorts; the association between education and FGC was assessed by stratified analysis. To adjust for multiple determinants of FGC simultaneously, multivariate analyses were performed using logistic regression. All statistical analyses were performed using SAS software, version 6.0 (SAS Institute 2000, Cary, NC).
A total of 1709 women aged 15–49 years completed both the interview and the clinical examination and 45.9% (n=789) had undergone genital cutting based on clinical examination. A percentage of 32.6 were classified by clinicians as having Type I cuts, 11.5% had Type II cuts, and 1.9% had Type III or IV cuts (Figure 1). Of these 1.9% with the most extreme cuts, 24 women were Type III, having excision of part or all of the external genitalia and stitching/narrowing of the vaginal opening (infibulation), and eight women had Type IV (unclassified cuts).
Validity of self-reporting
Seventy-nine percent of women correctly identified whether or not they had undergone genital cutting, while 14% reported that they were unsure of whether or not they were cut, and 7% reported their status incorrectly (4% false positives, 3% false negatives).
There was no association between the validity of self-reporting and the type of cutting, reported age at circumcision, education or ethnicity (Table 2). However, validity was higher among cut than noncut women, lower for the youngest (15–19 years) age cohort, and lower among Catholics.
Table 2. The validity of self-reported FGC status by socio-demographic characteristics
The overwhelming majority (85%, n=238) of the 14% of women who did not know whether or not they were cut, were not cut. Hence again, as apparent for validity, women who were not cut were less sure of their status. Answers of ‘don’t know’ were also significantly more frequent among the youngest (15–19 years) age cohort, and among Catholics and Protestants (adjusted ORs 1.7 and 2.5, respectively, vs. Pentecostals, P < 0.03 and <0.0001). The probability of answering ‘don’t know’ for FGC status was inversely associated with education, being most common among those with the highest level of education (adjusted OR 4.7 tertiary vs. no education; P < 0.1).
Social correlates of FGC
Most women (84.3%) in the sample were between 20 and 39 years old, consistent with their attendance at antenatal and family planning services. Cut women were older than uncut women by an average of approximately 4 years (P < 0.0007). Cut women were slightly, but significantly, younger than uncut women at both their age at marriage, and first delivery (Figure 2).
While simultaneously controlling for ethnicity, age, religion and education, each of these factors retained a significant, independent association with FGC (Table 3). The change in the chi-square for covariates indicates that with regard to their relative importance in predicting the probability of FGC, ethnicity was the strongest predictor of FGC, age was second, followed by religion, and then educational attainment.
Table 3. Prevalence of female genital cutting (FGC) by socio-demographic characteristics among 1709 women 15–49 years, Edo State Nigeria, 1998–1999
The two largest ethnic groups in the sample were Esan and Bini, followed by Igbo, Yoruba and Urhobo. There were stark differences in the prevalence of FGC by ethnicity (Table 3): the highest prevalence of cutting was found among the Bini (69%), followed by Urhobo with 60.8% prevalence; the Urhobo also had the highest prevalence of Type II cutting (24.8%). The lowest prevalence of cutting was found among the Yoruba (28.9%) and Esan (32.5%).
Nearly all women (99%) reported a religious affiliation, with the highest proportion being affiliated with various Christian churches: Catholic (36.1%), Pentecostal (33.1%), or Protestant (16.3%). Only 5% of women identified themselves as Muslim, and 3% as followers of traditional religion(s). Among the popular Christian denominations, there was a significant difference in FGC prevalence between more traditional Christian denominations of Catholicism or Protestantism and Pentecostals (Table 3, Figure 4). Of those affiliated with the Pentecostal church 61.4% had been cut, exceeded only by 76.5% in the small group of women (n=51) affiliated with traditional religion(s). A woman affiliated with the Pentecostal church was 2.6 times more likely to be cut than a non-Pentecostal woman (P < 0.0001). Pentecostal and traditional religious affiliations included the highest proportions of women with Type II FGC. Both overall FGC and Type II cutting were lowest among Muslim women.
In this study, 76.1% of women had at least a secondary education, and 45% reported some tertiary training. The highest proportion of FGC (66.6%) was found among women with the least education (primary or less schooling). In contrast, women with some tertiary education were only one-third as likely to have undergone cutting (Table 3).
There has been a highly significant secular decline in the incidence of FGC during the last 25 years, with age-specific prevalence rates declining from 75.4% among women aged 45–49 years, to 48.6% among women aged 30–34 years, and 14.5% among women 15–19 years of age. Plots of FGC prevalence by 5-year birth cohorts within ethnic (Figure 3), religious (Figure 4) and educational (not shown) subgroups showed a consistent pattern of secular decline within all subgroups, with statistically significant tests for trend. The observed secular decline might have merely reflected a gradual cohort change in the mean age of cutting, whereby uncut women in the youngest cohorts are still waiting to be cut. To ensure that this was not the case we tested and affirmed that there has been no measurable change in the mean age of cutting among individuals (P=0.099), or by 5-year cohorts (P=0.499) over the same period. The overall mean age at FGC was 4.9 years, and the age distribution at FGC is presented in Table 2.
Women in our study who reported on their FGC status were overwhelmingly accurate in such reporting (92%). The caveat in the validity of self-reporting was that 14% of women reported they did not know their status, and this group was proportionately more likely to be uncut (85% were uncut). Consistent with this finding, Odujinrin et al. 1989) also reported that 81% of ‘unsure’ women in a Nigerian study were uncut. A recent study from the Gambia reports that 1156 of 1346 women reported their circumcision status, and among these women, agreement with examination was extremely high (97%); regrettably, the authors do not report on the cut status of women who did not know their status (Morison et al. 2001). Our finding suggests that studies which omit women who report they are ‘unsure’ about or ‘don’t know’ their FGC status may overestimate true prevalence.
Uncertainty and validity both appear to bear some association to overall FGC prevalence, i.e. invalid answers were more frequent among uncut, and among both young and Catholic women, each group with a comparatively lower prevalence of cutting. Responses of ‘don’t know’ were also highest among groups with a lower prevalence of FGC: the youngest cohort, the most highly educated, Catholics and traditional Protestants.
Uncertainty over personal FGC status may increase in social groups where cutting is less uniform, either because of less exposure to cut forms, or more ambiguity about an accurate or socially appropriate response. This finding may offer some explanation for the discrepancy of earlier reports on the validity of self-reporting. In Egypt, self-reporting was accurate in 94% of cases, where overall prevalence was 93% (EFCS 1996), and the authors make no reference to participants who were unable to answer the question. By contrast, a prior study in Nigeria found self-reporting valid in only 57% of cases, where overall prevalence was only 48.4% (Adinma 1997).
The overall prevalence of 46% FGC in our study from Edo State is very close to the 48.4% recently reported from the DHS for the whole south-west region of Nigeria (DHS 1999). Our findings by FGC type, however, were different: among all cut women we found 71% Type I and 25% Type II, while the more representative DHS data for the region report 84% Type I and only 5% Type II. The higher proportion of Type II cuts in our sample was largely attributable to the high numbers of Urhobo participants, but self-reporting by type (the source of DHS data) warrants further assessment.
Our findings on FGC prevalence within specific ethnic groups were close to those reported in several prior reports: Adinma (1997) reported 48% prevalence among Igbo women, consistent with our 47%; Ehigiegba et al. (1998) found 65% prevalence among Binis, consistent with our 69%. Likewise, our findings with regard to the rank order of FGC prevalence among different ethnic groups was similar to studies on mixed groups, specifically the higher rate of cutting among the Bini relative to the Esan (Myers et al. 1985; Omorodion & Myers 1989), and the higher prevalence among the Ibo relative to the Yoruba (Odujinrin et al. 1989).
We found an unexpected distinction between the proportion of FGC among adherents to traditional Christian vs. Pentecostal churches. After testing the possibility of confounding by ethnicity, age, or education, the higher prevalence of cutting among adherents to Pentecostal churches remains an independent association, and highly significant. This is a provocative finding for those interested in the social dimensions of FGC or the ethnology of Pentecostal appeal in Nigeria, raising the question of why this church has such an over-representation of cut women among its members? Given that the growth of the Pentecostal churches in Nigeria is relatively recent, almost all of its adherents will have undergone FGC long before joining this church. Hence the outstanding question is whether and why the church appears to attract more cut women, and whether this can be explained by socio-demographic clustering beyond that explored in this study (i.e. education, ethicity or age). From an intervention perspective, lobbyists for the eradication of FGC may find it worthwhile to work with the Pentecostal churches to discourage a continuation of cutting.
Our finding that higher educational attainment by women was inversely associated with probability of being cut was consistent with similar findings from several studies in Nigeria and elsewhere (Ehigiegba et al. 1998). A recent study in Ghana found educational attainment among females to be the most important social correlate of cutting, with lower rates of cutting with higher education (Adongo et al. 1998; Mbacke et al. 1998). Presumably in both the Ghanaian and our study it is the characteristics of parents that are most decisive in both promoting their daughters education, and avoiding cutting. In general, youth and parental educational attainment are closely correlated, and a recent Nigerian study affirms that higher parental education is associated with a lower prevalence of cutting (Ehigiegba et al. 1998). Likewise, in Egypt it was found that the likelihood of a women having some type of FGC is lower if either of her parents (particularly her mother) has a higher education (EFCS 1996).
Finally, our data indicate a significant and consistent secular decline in the incidence of FGC in this region during the past 25 years both for the total sample and for specific subgroups. The major caveat to this finding would be if the age of circumcision were advancing by cohort, with younger cohorts not yet cut, but there was no evidence of such a trend. In an analysis of DHS reports up to 1997, a comparison between the 20–24-year and 45–49-year age groups suggested a decline in the Central African Republic, but no evidence of decline in Cote d’Ivoire, Egypt, Eritrea, Mali or the Sudan (Carr 1997). In two studies from Nigeria and one from Ghana, stratification of FGC prevalence by age group also suggests a secular decline (Megafu 1983; Adinma 1997; Mbacke et al. 1998), but the authors do not comment on this aspect of their data. The emerging DHS data from Nigeria and other countries will provide an opportunity to evaluate these findings further.
Funding for this research was provided by the BMBF as part of research programme TropMed- Heidelberg. We are grateful to Prof. Dr Heiko Becher for advice and support regarding the statistical analysis. Special thanks are also due to Dr Bruni Ludwig for her contributions during the planning and design of the study. We gratefully acknowledge the participation of colleagues in each of the collaborating institutions: University of Benin Teaching Hospital, Central Hospital, and Specialist Teaching Hospital in Irrua. A special thanks to Mr Dapo Ogunsakin for support with data entry, and to Helen Ajabor for research assistance to Tracy Slanger during field visits. Finally, we thank all the women who took part in the study, and shared their experiences.