Population-based breast (female) and cervix cancer rates in the Gambia: Evidence of ethnicity-related variations



Since 1987, the Gambia National Cancer Registry has provided nationwide cancer registration for the Gambia. We used data from 1998 to 2006 to assess age-standardized incidence rates (ASRs) of 2 common cancers in women, breast and cervix. With an ASR of 15.42 (95% CI [14.18–16.66]) for cervix and 5.86 (95% CI [5.12–6.59]) for breast per 105 person-years, these cancers ranked first and third, respectively, among Gambian women (the second most common being liver, ASR 14.90). Incidence of both cancers, breast and cervix, increased rapidly at young ages to reach a peak at ages 40–44 years. Significant differences were observed in relation to ethnicity. Using the Mandinka (42% of the population) as a reference, breast cancer incidence rates were 2.16-fold higher (95% CI [1.33–3.52]) in Jola (10% of the population), specially at early-onset ages (before 40 years). For cervix cancer, highest rates were observed in Fula (18% of the population; risk ratio (RR): 1.84 (95% CI [1.44–2.36])). In contrast, a significantly lower risk was observed in the Serrahuleh (9% of the population; RR: 0.54 (95% CI [0.31–0.96]). This study revealed a preponderance of early-onset breast cancer among Gambian women similar to that seen in African women in more developed countries but also demonstrates large ethnic variations. It points to the need for further studies on cancer determinants to improve prevention, early detection and therapeutic management of these diseases in a low-resource setting in West Africa.

Cancer is increasingly recognized as a major public health problem in sub-Saharan Africa.1, 2 In most of these countries, the exact burden of cancer has long been overshadowed by the focus on infectious and other diseases and has not been helped by the lack of reliable data on cancer incidence. Collecting and analyzing such data in the context of low-resource countries is a daunting task because of the lack of infrastructure for reporting and collecting data on cancer occurrence.3 The lack of structured programs for cancer diagnosis, management and treatment poses enormous problems for the correct evaluation of cancer burden in most areas, with the exception of a few urban centers. Furthermore, the rapidly changing population structure and the lack of detailed demographic data make it very difficult to evaluate cancer frequencies and variations over time.

The Gambia National Cancer Registry (GNCR), established in 1986 in the context of the Gambia Hepatitis Intervention Study (GHIS) (a long-term nationwide trial of efficacy of childhood hepatitis B vaccine (HBV) against chronic liver disease and liver cancer in adults), is the only nationwide, population-based cancer registry in Sub-Saharan Africa. The Gambia is the smallest country of the African continent, with a population of 1,360,681 million (2003 demographic census), including about a dozen distinct ethnic groups, all of which are also present in other countries of West Africa. The largest groups are the Mandinka, the Wollof, the Fula, the Jola and the Serrahuleh, who together represent about 95% of the Gambian population of whom about 50% resides in rural areas (http://www.accessgambia.com/information/demographic-statistics.html).

Because of the high endemicity of HBV chronic infection and widespread exposure to aflatoxins, liver cancer is by far the most common cancer in both genders in the Gambia. This high representation of rural populations distinguishes the GNCR from most other cancer registries in Sub-Saharan Africa, which are based on large hospitals in urban centers. The GNCR is the main instrument for evaluation of GHIS endpoint, namely, the incidence of liver cancer, and it actively collects cancer data across the country.4 In this study, data registered by the GNCR from 1998 to 2006 were used to assess the age-standardized incidence rates (ASRs) of 2 common cancers in Gambian women, namely cancers of the breast and of the cervix. Variations in the incidence rates by ethnic group were also investigated.

Material and Methods

Cancer incidence data

For the period 1998–2006, incidence rates were calculated for breast and cervix cancer using, as the numerator, primary cases diagnosed during this period according to the GNCR database. The methods and procedures of data collection of the GNCR have been previously described.4 Notification of cancer is voluntary in the Gambia, and the registration process is active. GNCR clerks are posted in 4 main hospital locations throughout the country: Royal Victoria Teaching Hospital (RVTH), Banjul; Medical Research Council, Fajara; Bansang Hospital and AFPRC General Hospital, Farafeni. In- and outpatients clinics of other smaller hospitals are monitored on a regular basis, each clerk covering primary and secondary medical centers in their hospital catchment area. The main basis of diagnosis is clinical, with histological confirmation performed at the National Histopathology Laboratory. Data are centralized at the GHIS office in Fajara, verified and managed using the CanReg 4.0 software.5 For the purpose of this study, incidence cancer data from 1998 to 2006, the most recent year with complete data at the time of extraction, were extracted from the CanReg 4.0 database. Table 1 summarizes the GNCR data completeness exploited for this study.

Table 1. The Gambia national cancer registry completeness
inline image

Information on ethnicity

Information on ethnicity was self-reported at diagnosis; each subject identified herself based on the language and culture of the biological parents or was obtained during a structured interview by GNCR clerks who are fluent in the language of each major ethnic group. Of note, the information on ethnicity was completed in 77% for breast cancer and 79% for cervix cancer (Table 1).

Population data

Population data were based on national demographic censuses of 1993 and 2003. During this period, the population grew from 1,031,145 to 1,310,681, an overall change of 27% (http://www.accessgambia.com/information/demographic-statistics. html). Figure 1 represents the age structure of the population according to the 2003 census. Population age structure of these censuses was linearly interpolated for period from 1998 to 2002 and extrapolated for period from 2004 to 2006. Population distributions for the 5 main ethnic groups were based on 2003 census in which the Mandinka, Fula, Wollof, Jola and Serrahuleh represented 42, 18, 16, 10 and 9% of the total population, respectively (http://www.accessgambia.com/information/people-tribes.html; https://www.cia.gov/library/publications/the-world-factbook/geos/ga.html#People). It was assumed that age and sex population structure in each ethnic group was identical to that of the whole population.

Figure 1.

The Gambia pyramid of age 2003 census.

Age-standardized rates and statistical analysis

Crude and ASRs were calculated for cervix and breast cancers. ASRs were standardized to the world population by 5-year age group6 as described by Boyle and Parkin.7 To do this, the total population and total number of relevant cases by 5-year age group were calculated for the whole study period, i.e., calendar year was not taken into account. To investigate ethnic variations in incidence rates, for each cancer, Poisson regression models were fitted with the number of cases as the response variable and categorical 5-year age group and ethnic group as explanatory variables. Population size was taken into account in modeling by using it as an offset term in Poisson models. All analyses were carried using R software.3 The Mandinka, who represents the most numerous ethnic group, was used as the reference group.


Incidence of breast and cervix cancer

During the period 1998–2006, GNCR recorded a total of 3,843 cases of cancer, 1,993 in men and 1,849 in women. The 3 most common cancers recorded in females of the Gambia were cancers of the cervix, liver and breast (Table 2). With an ASR of 15.42 (95% CI [14.18–16.66]) per 105 person-years, cervix cancer was the most frequent cancer among women, representing 29% of the all registered female cancers. Most cases (77%) were diagnosed at RVTH, the main referral hospital in the capital city, and histological confirmation was available in only 8% of cases. This ASR was almost identical to the one of female liver cancer (14.90 95% CI [13.62–16.17]), a cancer which is common in both genders in the Gambia and represents 27.69% of all female cancers. Breast cancer ranked third, with an ASR of 5.86 (95% CI [5.12–6.59]) per 105 person-years, representing 12% of all registered cases and was confirmed histologically in 24% of the cases with most cases (82%) diagnosed at RVTH. All other cancers were relatively uncommon (Table 2). The number of cancers cases registered annually was stable during the study period (Table 3). Table 3 summarizes the pathological, clinical and individual information available in the GNCR for cervix and breast cancer cases. Among the 150 breast patients for whom information on parity was available, 39% had a maximum of 2 children (including 23% nulliparous). Among cervix cancer patients, 27% had 2 children or less (including 17% nulliparous). The difference in parity between breast and cervix cancer patients was statistically significant (p = 0.005).

Table 2. Most frequent female cancers in the Gambia 1998–2006
inline image
Table 3. Cancer registry detail information, the Gambia 1998–2006
inline image

Figure 2 shows the age-specific incidence rates of cases of breast and cervix cancers in the Gambia. Cervix cancer increased from the age 20 to 24 and peaked in the age groups 40–44 and 55–59, followed by a decrease in older age groups. Although breast cancer had a low ASR, it was characterized by early onset and biphasic profile; 71% of cases was under age 50. Breast cancer was detected from young adult ages and rates peaked in the age group 40–44. It then decreased in the age groups 45–54 and increase again after the age of 55 (Fig. 2a). Figure 2b shows the age distribution of cervix and female breast cancer cases, using the standard world population as a reference. It can been seen that if the age-specific incidence observed in the Gambia occurred in the world standard population, the highest peak of occurrence of breast cancer would occur in the 40–44 year age group.

Figure 2.

Age distribution of cases of breast and cervix cancer in the Gambia, 1998–2006. (a) Age-specific incidence rates; (b) expected age distribution of incident cases in 100,000 Gambian women with the world standard population.

Differences according to ethnic groups

There were significant differences in ASRs for both breast and cervix cancer according to ethnicity (as defined by self-identification of the patient, see Methods) (Table 4). For breast cancer, the highest cumulative ASR was observed in the Jola and Wollof, followed by the Fula with an intermediate cumulative rate, whereas the Serrahuleh and the Mandinka had the lowest cumulative rate (Fig. 3). The Jola, who represents about 10% of the Gambian population, had a 2.16-fold increased risk of breast cancer (95% CI [1.33–3.52]) when compared with the Mandinka, who represent 42% of the population. A smaller but still significant increased risk was detected among the Wollof (16% of the population; risk ratio (RR): 1.96 (95% CI [1.25–3.08])). The Jola had the highest rate of early-onset breast cancer (in the age group 20–44 years old. RR: 2.64 (95% CI [1.49–5.06])). The increased risk of breast cancer in the Jola when compared with the Wollof was mainly due to early-onset breast cancers (Fig. 3). The risk of breast cancer in the Serrahuleh (9% of the population) was similar to the Mandinka (RR: 0.83 (95% CI [0.41–1.69])).

Table 4. Risk ratio female breast and cervix cancer associated with ethnic groups, the Gambia 1998–2006
inline image
Figure 3.

Cumulative incidence rate by age group for female breast cancer in 5 ethnic groups, the Gambia 1998–2006.

For cervix cancer, differences between ethnic groups were also observed, albeit not similar to those observed for breast cancer. The Fula (18% of the population) and the Jola had the highest cumulative ASR (Fig. 4). They had a significantly increased risk of 1.84 (95% CI [1.44–2.36]) and 1.60 (95% CI [1.17–2.17]), respectively, when compared with the Mandinka. The Wollof presented a cumulative rate similar to the one of Mandinka (RR: 1.22 (95% CI [0.92–1.62]). Conversely, the Serrahuleh had a significantly lower risk than the Mandinka to develop cervix cancer (RR: 0.29 (95% CI [0.31–0.96])) (Table 4).

Figure 4.

Cumulative incidence rate by age group for cervix cancer in 5 ethnic groups, the Gambia 1998–2006.

Because ethnic groups have historically distinct geographic distribution in the country and because access to diagnosis may differ between urban and rural areas, residence may represent confounding factors. The Gambia comprises 6 divisions (from east to west, Banjul, Brikama, Kerewan, Massa Konko, Central River Division and Basse). Variations in breast and cervix cancer incidence are detectable between divisions (Table 5). Taking the divisions with the lowest incidence rates as a reference (Kerewan for breast cancer and Basse for cervix cancer), significant variations between divisions were observed for Banjul (breast and cervix), Brikama and Kerewan (cervix) but not for other divisions (Table 5). However, these differences did not correlate with the geographic distribution of ethnic groups.

Table 5. Crude rate rations for female breast and cervix cancer for divisions of the Gambia 1998–2006
inline image


Cancer burden estimation remains a challenge in low-resource countries in Western Africa, and there is a lack of population-based information on cancer incidence in particular in rural areas. In this study, we have taken advantage of the settings of the National Cancer Registry of the Gambia, which provides nationwide coverage for cancer registration since 1987. Although data compiled in the registry confirm that liver cancer is by far the main cancer diagnosed throughout the country, it became apparent that female cancers, including breast and cervix cancer, were relatively common in particular in young women. This prompted us to analyze the information available on these cancers in the GNCR. As the Table 1 shows, although tumor-related variables are almost complete, there is insufficient information on many risk factors (such as HPV, HIV and breast feeding status), which may help to understand these cancers. On the other hand, the low percentage of cases with histology confirmation may be due to the fact that most patients present with late-stage disease, when there is no more justification for biopsy. This late presentation may explain the high percentage of the clinical diagnostic (85% on an average) despite the existence of a National Pathology Service laboratory.

A recent survey of cancer registration data since the inception of the GNCR have showed that cancer incidence is relatively stable in the Gambia. During the 20 past years, the incidence of breast cancer only marginally increased by 6.5%. The ASR for 1987–1997 was 5.5 per 105 person-years4 compared with 5.86 in 1998–2006, the 9 years covered by our study. However, it should be noted that there is no national program for active detection and reporting of breast cancer and that the ascertainment of this disease may be incomplete, in particular in more rural areas. In contrast, during the same period, the incidence of cervix cancer had decreased by 18%. The ASR for the period 1987–1997 was 18.9 per 105 person-years4 compared with 15.45 per 105 person-years during the period covered by this study. These variations may be artifactual, because of causes such as changes in infrastructure and personnel in the Gambian health care system. During the period studied, the number of cervix cancer cases was less stable than breast cancer and may have been underestimated in some years. However, for breast cancer, the increase is greater than the 1.5%, which is the worldwide average increase for breast cancer incidence.8 Although the observed incidence is remarkably low when compared with many industrialized countries, breast cancer has become the second most frequent cancer after liver cancer before the age of 25 in the Gambia.

The age-specific incidence curve of breast cancer showed a biphasic variation with a progressive increase until the age 40–44 followed by a strong decrease in the age group 50–54 and by a second increase at later ages (60–64 years). For cervix cancer, there was also a peak in the 40–44 year age group, although the highest incidence was observed in the 55–59 year age group. The average age at diagnosis was 43 years for breast cancer and 44 years for cervix cancer. Both cancers were typically detected at late stages, and the survival after diagnosis was very poor (less than 6 months on an average).

The age-specific incidences rate for breast cancer showed a predominance of premenopausal breast cancer, as commonly observed in other African populations. Several studies and available cancer registration data have shown that in sub-Saharan Africa the main peak of incidence of breast cancer occurs between 35 and 45 years of age with an average age at diagnosis of about 40 years.9–13 The same trend toward premenopausal breast cancer is observed in the populations of African origin in the US, Caribbean Barbados14, 15 and British women of African origin.16 It has been suggested that breast cancer in women of African descent may show different clinical and biological characteristics than in women of Caucasian or Asian origin.16–19 There is an evidence that women of African origin tend to present with larger primary tumors at more advanced stages than Caucasians. Furthermore, their cancers are more frequently negative for estrogen and/or progesterone receptors (ER and PR).16–18, 20, 21 Although data are scarce, 3 studies have shown that the proportion of patients with ER and PR negative cases varies between 50 and 66% among sub-Saharan African women,22–24 in contrast to 30–40% in women of Caucasian populations.25, 26

Parity has been reported to be a protective factor for postmenopausal breast cancer.27–30 Thus, a protective effect of parity may contribute to explain the relative decrease in age-standardized incidence in postmenopausal women. Of note, 77% of breast cancer patients was parous (with ≥3 children in 61% of the cases) with on an average of 5 children (range 1–13) per women. It should be noted, however, that the size of population groups in the postmenopausal age range is small and that breast cancer may be underestimated in older age groups. It has also been reported that the protective effect of parity is restricted to hormone-dependent breast cancer (ER- and PR-positive cases).27, 31, 32 Further studies are needed to assess the stage and hormonal status of breast cancer in Gambian women as well as the contribution of parity as a risk factor, particularly at young ages.

Cervix cancer is the most common cancer in Gambian women. The age-standardized incidences show a progressive increase during early adult life to reach a peak in the age group 40–44. The incidence remains stable until age 55–60 and then decreases steadily in older women. Infection by human papillomaviruses, the main etiological agent for cervix cancer, is common among Gambian women.33 However, the prevalence of HPV in cervix cancer cases has not been analyzed so far. A study on cervix HPV infection and squamous intraepithelial lesions (SILs) in an unselected cohort of 1,061 women in a rural Gambian setting reported a prevalence of cervix HPV infection in 13% and SILs in 7%. High-risk HPV was found in 87% of SIL with human papillomavirus-16 being the most prevalent and most strongly associated with SIL.33 When compared with breast cancer, patients with cervix cancers were more frequently multiparous with an average number of 6 children per women.

In relation to ethnicity, for breast and cervix cancers, we observed significant differences in cumulative risk. The highest incidence of breast cancer was detected in the Jola, a group that represents about 10% of the population. When compared with the Mandinka, who are the largest ethnic group (42% of the population), the Jola has a significantly increased risk of breast cancer of 2.16-fold. Moreover, most of this excess risk was due to cancers diagnosed before the age of 40. A lesser but still significant increased risk was observed in the Wollof, the third largest population group (16% of the population). Different variations in relation to ethnicity were observed for cervix cancer, with a risk 1.84-fold higher than in the Mandinka detected in the Fula (representing 18% of the population) and a lower but significantly increased risk in the Jola. In the cancer registry, information on ethnicity was self-reported at interview by trained cancer registrars. Patients are generally interviewed in their local language, which differ among ethnic groups. Thus, the variable “ethnicity” captures different sources of variations including lifestyle, cultural, geographic and possible genetic factors. Studies on genetic polymorphisms in carcinogen metabolism and detoxification genes, as well as on human leukocyte antigen genes, have identified significant differences between these ethnic groups.34, 35 A recent genome-wide association analysis of malaria in the Gambia demonstrated that self-reported ethnicity correlates with genetically defined subpopulations.36 In this study, the greatest differentiation was seen between the Fula and the Jola and the least between the Mandinka and the Wollof. This observation gives weight to the hypothesis that ethnicity-related differences in cumulative cancer risk may be due, at least in part, to genetic susceptibility. On the other hand, differences in lifestyle may also have an impact. In particular, the Fula, who with the highest cumulative risk of cervix cancer, has also been reported as having a much higher prevalence of HPV infection (21% when compared with 14% in the Mandinka and 8% in the Wollof33) as well as a highest prevalence of other sexually transmitted diseases.37, 38 Finally, our analyses show that differences in ethnic groups do not reflect difference in residence; a potential confounder in this analysis plays a relatively minor role. In particular, the Brikama division contains the largest proportion of Jola, but the incidence rate in this division is not significantly higher than in other divisions. The lack of concordance was even clearer for cervix cancer, because the lowest incidence rate was observed in Central River and Basse Divisions, 2 divisions where the Fula population is particularly important. On the other hand, parity is unlikely to be an explanatory variable, because parity is high in all ethnic groups and the differences from one group to the other are small. However, variables associated with reproductive factors are not recorded in detail in the registry, and further studies on the role of these variables are warranted. Finally, there are no data documenting the differences in socioeconomic status between the ethnic groups.

In conclusion, this study provides the first assessment of age-related incidence of breast and cervix cancer based on nationwide, population-based cancer registration in West Africa. These data support the notion that breast cancer may show different biological characteristics in African women than in women of Caucasian descent. They also suggest a possible contribution of ethnicity-related factors, including genetic factors, as a determinant of risk. Future studies are needed to better characterize these risk factors to improve prevention, early detection and therapeutic management of these diseases in a low-resource setting in West Africa.