Abstract
Background:
The Nigerian Breast Cancer Study (NBCS) model is a new risk assessment tool developed for predicting risk of invasive breast cancer in Nigeria. Its applicability outside of Nigeria remains uncertain as it has not been validated in other sub-Saharan Africa populations.
Methods:
We conducted a case-control study among women with breast cancer and controls ascertained in Cameroon and Uganda from 2011 to 2016. Structured questionnaire interviews were performed to collect risk factor characteristics. The NBCS model, the Gail model, the Gail model for Black population, and the Black Women’s Health Study model were applied to the Cameroon and Uganda samples separately. Nigerian as well as local incidence rates were incorporated into the models. Receiver-Operating Characteristic analyses were performed to indicate discriminating capacity.
Results:
The study included 550 cases (mean age 47±11.9) and 509 Controls (mean age 46±11.7). Compared to the other three models, the NBCS model performed best in both countries. The discriminating accuracy of the NBCS model in Cameroon (age-adjusted C-index = 0.602, 95% CI 0.542-0.661) was better than in Uganda (age-adjusted C-index = 0.531, 95% CI, 0.459-0.603).
Conclusion:
These findings demonstrate the potential clinical utility of the NBCS model for risk assessment in Cameroon. All currently available models performed poorly in Uganda, which suggests that the NBCS model may need further calibration before use in other regions of Africa.
Impact:
Differences in risk profiles across the African diaspora underscores the need for larger studies and may require development of region-specific risk assessment tools for breast cancer.
Introduction
Breast cancer is the leading cause of cancer death in women worldwide (1,2), with 684,996 deaths due to breast cancer in 2020 (2). Although sub-Saharan Africa (SSA) has one of the lowest incidence rates for breast cancer in the world, the mortality rate is comparable to that of high-incidence countries, partly due to more advanced tumor stage at diagnosis in women in SSA (2,3). Additionally, the burden of cancer is expected to grow in less economically developed countries due to changes in lifestyle behaviors that are known to increase cancer risk and the growth and aging of the population (4,5).
To address the disproportionately high breast cancer mortality in SSA, there are multiple strategies, such as community-level education campaigns on cancer awareness and improved access to cancer diagnosis and treatment. Identification of women at high risk of developing breast cancer is another practical and cost-effective approach with the potential to save lives, as these cancers have a high cure rate when detected early (6,7). Risk stratified screening is an especially important solution in low-resource settings and in countries with low incidence rates, including Cameroon and Uganda, because cancer is often detected at a late stage due to low awareness and insufficient resources for universal screening (3,8,9). In addition, risk assessment could help community health workers to understand risk factors of breast cancer and promote primary prevention. Risk prediction models are commonly used in high-income countries to identify women in need of increased surveillance (10). The NCI Breast Cancer Risk Assessment Tool (Gail model), developed initially for white populations, is one of the most commonly used risk-prediction model in the United States (11). To address ethnicity related variation in breast cancer risk factors, the Gail model was adapted for African Americans (CARE model) (12) and researchers from the Black Women’s Health Study (BWHS) developed a breast cancer risk prediction model specifically for African American women (13).
As incidence rates and risk factor profiles for breast cancer differ significantly between populations, risk prediction models created for one population are not always translatable to other populations (10,14,15). Compared to women in high income countries, women in SSA have higher parity, which has been demonstrated to be protective against breast cancer (5). Additionally, the low incidence of breast cancer and higher rates of triple negative breast cancer in SSA women may affect the accuracy of risk prediction models created for white populations or high-income countries (16,17). Thus, development of a risk prediction model specific for SSA women is a fundamental step in introducing risk stratified screening in SSA. In response, a risk prediction model was recently developed by researchers from the Nigerian Breast Cancer Study (NBCS) in Nigerian women (https://bcrisktool.uchicago.edu) (18).
However, the NBCS model was developed using data exclusively from a study of Nigerian women and should therefore be validated in different regions of SSA to avoid biased risk projections outside of Nigeria. Although the literature regarding the variation in risk factor profiles for breast cancer in SSA women remains sparse, the genetic and cultural diversity that exists within SSA is known to influence breast cancer risk. Some pathogenic variants of the BRCA1/2 gene, for example, have been found to be unique to Nigerian women (19,20). Additionally, alcohol consumption has been found to be significantly associated with an increased risk of developing breast cancer among women from Nigeria, Cameroon, and Uganda, countries with drastically different alcohol consumption rates (5.0%, 34.6%, and 50.0%, respectively) (21,22). Acknowledging the variation in breast cancer risk factors throughout SSA, the effectiveness of the NBCS model in SSA populations outside of Nigeria must be investigated before its use in other countries. The aim of this study was to investigate the performance of the NBCS model in predicting individual breast cancer risk in Cameroon and Uganda.
Materials and Methods
Study population
In 1998, we initiated a case-control study of breast cancer in Nigeria (5). In 2011, using the same protocols and questionnaires, the study was expanded to include Cameroon and Uganda. Together, we labeled the study as African Breast Cancer Study. The design of the study is described in detail previously (21,23). At the Uganda site, breast cancer cases were identified and recruited through the department of surgery’s breast and endocrine unit at Mulago Hospital, which is a national referral hospital in Kampala, Uganda, that serves a population of 1.3 million people. Controls were randomly selected from admissions to the surgical ward and general outpatient clinics of the same hospital. They were matched to cases for ethnicity and age, and consisted of either healthy individuals who accompanied their sick relatives or friends (about 60%) or patients with hernias (about 20%), lipoma (about 5%), bone fractures (about 5%) or other conditions. At the Cameroon site, cases were enrolled through the department of oncology at Yaounde General Hospital, which is a hospital in Yaoundé, Cameroon, serving a population of 2.5 million people. Hospital based controls were randomly enrolled at the same hospital through general medicine clinics and obstetrics and gynecology departments. They were matched to cases by ethnicity and age and were almost all healthy women who accompanied their sick friends or relatives. At both the Uganda and Cameroon sites, controls were absent of breast cancer and unselected for their medical conditions, while cases were defined as females ages 18 years or older with either a clinical or histologic breast cancer diagnosis. Recruitment was highly successful in both patient and control groups, with response rates >90%. By April 2016, 1,059 participants were recruited. Written informed consent was provided by all participants before their interview. Institutional review boards of the two study centers and the University of Chicago approved the study protocol.
Data collection and measures
At both sites, trained nurse interviewers measured height and weight, and administered structured questionnaire interviews that included questions about demographics, menstrual and reproductive history, history of benign breast disease, lifestyle factors, history of alcohol consumption, and family history of breast cancer. We defined having a family history of breast cancer as having a first degree female relative with a breast cancer diagnosis. We defined regular alcohol consumption as drinking alcoholic beverages at least once a week for 5 or more years. Due to the rare use of estrogen plus progestin therapy as treatment for symptoms of menopause in SSA, its use among participants was not addressed in the questionnaire.
Statistical analyses
To compare demographic factors and potential confounders for breast cancer between cases and controls, we used student t-test or Wilcoxon rank-sum test if the variable was continuous and chi-square test or Fisher’s exact test if the variable was categorical. The adjusted odds ratio and confidence interval were estimated using multivariable logistic regression. The NBCS model, the Gail model for White population, the Gail model for Black population (i.e. CARE model), and the BWHS model were applied to the Cameroon and Uganda samples separately. The NBCS model was applied to the Cameroon and Uganda samples using the Nigerian breast cancer incidence rates utilized in the original NBCS model (18). Because breast cancer incidence rates vary across African countries, we also modified the NBCS model using breast cancer incidence rates in Uganda and Cameroon, obtained from the Kampala Cancer Registry (24) and the Yaounde Cancer Registry (Dr. George Enow Orock, personal communication), respectively. The incidence rates of breast cancer used in the models can be found in Supplemental Table S1. We performed Receiver-Operating Characteristic (ROC) analyses, and use area under the ROC curve to indicate discriminating capacity. Age-adjusted concordance index (C-index) was calculated to account for the effect of age. Two-sided P-value <0.05 was considered statistically significant. All statistical analysis was conducted using Stata 16.0 (StataCorp) and SAS 9.4 (SAS Institute Inc.).
Results
This study included a total of 1,059 women, 298 cases and 273 controls in Cameroon and 252 cases and 236 controls in Uganda. The mean age was 46.6 years old for both Cameroon and Ugandan participants. Significant differences were noted in the distribution of several risk factors between study participants in Cameroon and Uganda (Supplemental Table S2). Compared to Uganda, participants in Cameroon were significantly more likely to have a family history of breast cancer, a lower parity, a later age at first live birth, shorter total duration of breastfeeding, taller, and a higher BMI.
Table 1 shows demographics and risk factors of cases and controls, stratified by study site. In both study sites, age was similar between cases and controls. Having a benign breast disease was associated with breast cancer risk in both Cameroon and Uganda sites. Higher body mass index was significantly associated with breast cancer risk in Cameroon, and same trend was observed in Uganda but it was not statistically significant. Regular alcohol intake was significantly associated with breast cancer risk in Cameroon, but only trend toward significance in Uganda.
Table 1.
Distribution of selected risk factors in cases and controls in Cameroon and Uganda
| Characteristics | Cameroon | Uganda | ||||||
|---|---|---|---|---|---|---|---|---|
| Case N=298 | Control N=273 | Age-adjusted OR (95% CI) | P | Case N=252 | Control N=236 | Age-adjusted OR (95% CI) | P | |
| N (%) | N (%) | N (%) | N (%) | |||||
| Age group | 0.79 | 0.59 | ||||||
| <30 | 16 (5.4) | 10 (3.7) | 1.58 (0.68-3.68) | 18 (7.1) | 20 (8.5) | 0.94 (0.46-1.91) | ||
| 30-39 | 73 (24.5) | 70 (25.6) | 1.03 (0.66-1.60) | 52 (20.6) | 48 (20.3) | 1.13 (0.68-1.87) | ||
| 40-49 | 89 (29.9) | 88 (32.2) | 1.0 (ref.) | 75 (29.8) | 78 (33.1) | 1.0 (ref.) | ||
| 50-59 | 84 (28.2) | 68 (24.9) | 1.22 (0.79-1.89) | 61 (24.2) | 61 (25.8) | 1.04 (0.65-1.67) | ||
| 60-69 | 24 (8.1) | 27 (9.9) | 0.88 (0.47-1.64) | 36 (14.3) | 22 (9.3) | 1.70 (0.92-3.16) | ||
| ≥70 | 12 (4.0) | 10 (3.7) | 1.19 (0.49-2.89) | 10 (4.0) | 7 (3.0) | 1.49 (0.54-4.11) | ||
| Age at menarche | 0.77 | 0.80 | ||||||
| <12 | 11 (3.8) | 13 (4.8) | 0.76 (0.27-2.14) | 3 (1.2) | 5 (2.2) | 0.68 (0.14-3.34) | ||
| 12 | 40 (13.9) | 27 (10.0) | 1.33 (0.59-2.97) | 10 (4.1) | 12 (5.3) | 1.17 (0.41-3.36) | ||
| 13 | 46 (16.0) | 51 (18.8) | 0.80 (0.38-1.71) | 34 (14.1) | 32 (14.1) | 1.36 (0.61-2.99) | ||
| 14 | 75 (26.1) | 73 (26.9) | 0.96 (0.47-1.96) | 80 (33.2) | 74 (32.6) | 1.41 (0.70-2.84) | ||
| 15 | 59 (20.6) | 60 (22.1) | 0.92 (0.44-1.91) | 56 (23.2) | 48 (21.1) | 1.47 (0.70-3.05) | ||
| 16 | 35 (12.2) | 28 (10.3) | 1.16 (0.52-2.60) | 40 (16.6) | 31 (13.7) | 1.74 (0.80-3.78) | ||
| >16 | 21 (7.3) | 19 (7.0) | 1.0 (ref.) | 18 (7.5) | 25 (11.0) | 1.0 (ref.) | ||
| Benign breast disease | 42 (14.2) | 13 (4.8) | 3.40 (1.76-6.56) | <0.001 | 29 (11.6) | 6 (2.5) | 5.43 (2.19-13.47) | <0.001 |
| Family history of breast cancer | 29 (9.8) | 16 (5.9) | 1.85 (0.98-3.47) | 0.089 | 6 (2.6) | 13 (6.0) | 0.40 (0.15-1.05) | 0.06 |
| Parity | 0.31 | 0.09 | ||||||
| 0 | 31 (10.5) | 24 (8.8) | 1.0 (ref.) | 22 (9.0) | 12 (5.1) | 1.0 (ref.) | ||
| 1-2 | 77 (26.0) | 74 (27.1) | 0.77 (0.40-1.48) | 55 (22.4) | 49 (20.9) | 0.58 (0.25-1.32) | ||
| 3-4 | 85 (28.7) | 64 (23.4) | 1.01 (0.52-1.96) | 48 (19.6) | 67 (28.5) | 0.37 (0.16-0.84) | ||
| ≥5 | 103 (34.8) | 111 (40.7) | 0.68 (0.35-1.32) | 120 (49.0) | 107 (45.5) | 0.54 (0.24-1.19) | ||
| Age at first live birth | 0.88 | 0.41 | ||||||
| No birth | 31 (10.5) | 24 (8.8) | 1.25 (0.65-2.39) | 22 (9.7) | 12 (5.3) | 1.93 (0.88-4.22) | ||
| <20 | 111 (37.6) | 107 (39.2) | 1.0 (ref.) | 109 (48.2) | 107 (47.3) | 1.0 (ref.) | ||
| 20–24.9 | 89 (30.2) | 85 (31.1) | 0.98 (0.65-1.48) | 60 (26.5) | 68 (30.1) | 0.87 (0.56-1.36) | ||
| 25–29.9 | 43 (14.6) | 35 (12.8) | 1.19 (0.70-2.02) | 20 (8.8) | 25 (11.1) | 0.81 (0.42-1.58) | ||
| ≥30 | 21 (7.1) | 22 (8.1) | 0.90 (0.46-1.77) | 15 (6.6) | 14 (6.2) | 1.04 (0.47-2.31) | ||
| Total months of breastfeeding | 0.71 | 0.46 | ||||||
| <12 | 60 (20.5) | 49 (18.0) | 1.0 (ref.) | 32 (13.4) | 18 (7.9) | 1.0 (ref.) | ||
| 12–23 | 44 (15.0) | 50 (18.4) | 0.74 (0.42-1.31) | 21 (8.8) | 18 (7.9) | 0.60 (0.24-1.46) | ||
| 24–35 | 39 (13.3) | 30 (11.0) | 1.11 (0.59-2.08) | 14 (5.9) | 25 (10.9) | 0.29 (0.12-0.74) | ||
| 36–47 | 30 (10.2) | 29 (10.7) | 0.87 (0.45-1.69) | 18 (7.6) | 15 (6.6) | 0.69 (0.27-1.72) | ||
| 48–59 | 27 (9.2) | 19 (7.0) | 1.19 (0.57-2.45) | 23 (9.7) | 22 (9.6) | 0.52 (0.22-1.25) | ||
| 60–71 | 21 (7.2) | 26 (9.6) | 0.66 (0.32-1.38) | 13 (5.5) | 14 (6.1) | 0.51 (0.19-1.37) | ||
| 72–83 | 22 (7.5) | 21 (7.7) | 0.83 (0.39-1.75) | 16 (6.7) | 19 (8.3) | 0.45 (0.18-1.12) | ||
| 84–95 | 11 (3.8) | 6 (2.2) | 1.53 (0.51-4.62) | 17 (7.1) | 16 (7.0) | 0.52 (0.20-1.34) | ||
| ≥96 | 39 (13.3) | 42 (15.4) | 0.74 (0.39-1.43) | 84 (35.3) | 82 (35.8) | 0.49 (0.24-1.00) | ||
| Height in cm | 0.54 | 0.57 | ||||||
| <160 | 64 (22.1) | 63 (24.4) | 1.0 (ref.) | 83 (40.9) | 72 (36.5) | 1.0 (ref.) | ||
| 160–169 | 173 (59.7) | 143 (55.4) | 1.22 (0.80-1.85) | 83 (40.9) | 84 (42.6) | 0.86 (0.55-1.35) | ||
| ≥170 | 53 (18.3) | 52 (20.2) | 1.00 (0.59-1.69) | 37 (18.2) | 41 (20.8) | 0.74 (0.42-1.31) | ||
| Body mass index in kg/m2 | 0.018 | 0.28 | ||||||
| <18.5 | 5 (1.7) | 1 (0.4) | 3.22 (0.36-28.83) | 15 (7.5) | 9 (4.6) | 1.51 (0.61-3.74) | ||
| 18.5–24.9 | 103 (36.0) | 69 (27.0) | 1.0 (ref.) | 94 (46.8) | 80 (40.8) | 1.0 (ref.) | ||
| 25–29.9 | 109 (38.1) | 100 (39.1) | 0.72 (0.48-1.10) | 64 (31.8) | 72 (36.7) | 0.78 (0.49-1.24) | ||
| ≥30 | 69 (24.1) | 86 (33.6) | 0.54 (0.34-0.84) | 28 (13.9) | 35 (17.9) | 0.66 (0.36-1.21) | ||
| Oral contraceptive | 38 (13.0) | 48 (17.8) | 0.67 (0.41-1.08) | 0.10 | 51 (22.3) | 57 (27.0) | 0.77 (0.48-1.25) | 0.29 |
| Regular alcohol consumption | 119 (40.1) | 78 (28.6) | 1.70 (1.19-2.42) | 0.004 | 89 (36.3) | 64 (27.6) | 1.42 (0.95-2.13) | 0.085 |
OR, odds ratio; CI, confidence interval.
In Cameroon, the NBCS model using Nigerian breast cancer incidence rates has a moderate discriminating capacity, with an age-adjusted C-index of 0.602 (95% CI: 0.542-0.661). The discriminating capacity of the BWHS model in Cameroon was weaker than the NBCS model, while the Gail and CARE models were not statistically significant (Table 2). In Uganda, none of the models evaluated performed well, though the NBCS model has the highest C-index (0.531, 95%CI 0.459-0.603; Table 3). The discriminating accuracy of the NBCS model in Cameroon and Uganda remained consistent when accounting for local incidence rates, with age-adjusted C-index of 0.590 (95% CI: 0.529-0.651) for Cameroon and 0.530 (95% CI, 0.457-0.603) for Uganda (Table 4).
Table 2.
Discriminating accuracy of the NBCS model, BWHS model, Gail model, and CARE model in Cameroon
| Cameroon | NBCS model |
BWHS model |
Gail model |
CARE model |
|||||
|---|---|---|---|---|---|---|---|---|---|
| Age group | Number of participants | C-Index | 95% CI | C-Index | 95% CI | C-Index | 95% CI | C-Index | 95% CI |
| <30 | 26 | 0.631 | 0.400-0.862 | 0.669 | 0.447-0.890 | 0.617 | 0.378-0.856 | 0.656 | 0.435-0.878 |
| 30-39 | 140 | 0.576 | 0.480-0.671 | 0.570 | 0.476-0.665 | 0.533 | 0.437-0.629 | 0.503 | 0.407-0.599 |
| 40-49 | 174 | 0.613 | 0.529-0.698 | 0.614 | 0.531-0.697 | 0.624 | 0.540-0.707 | 0.551 | 0.466-0.636 |
| 50-59 | 150 | 0.583 | 0.491-0.675 | 0.549 | 0.456-0.642 | 0.505 | 0.411-0.599 | 0.563 | 0.469-0.657 |
| 60-69 | 47 | 0.569 | 0.396-0.741 | 0.520 | 0.358-0.682 | 0.550 | 0.389-0.712 | 0.563 | 0.403-0.724 |
| ≥70 | 12 | 0.657 | 0.231-1.00 | 0.263 | 0.000-0.575 | 0.200 | 0.000-0.453 | 0.200 | 0.000-0.440 |
| 35-70 | 457 | 0.599 | 0.547-0.651 | 0.561 | 0.509-0.613 | 0.546 | 0.493-0.598 | 0.531 | 0.478-0.583 |
| Total | 549 | 0.565 | 0.517-0.613 | 0.542 | 0.494-0.590 | 0.530 | 0.482-0.578 | 0.517 | 0.469-0.565 |
|
| |||||||||
| Age-adjusted (35-70) | 457 | 0.610 | 0.547-0.673 | 0.572 | 0.506-0.638 | 0.536 | 0.467-0.606 | 0.514 | 0.445-0.584 |
| Age-adjusted | 549 | 0.602 | 0.542-0.661 | 0.573 | 0.512-0.635 | 0.531 | 0.467-0.594 | 0.506 | 0.441-0.571 |
NBCS, Nigerian Breast Cancer Study; BWHS, Black Women Health Study; CARE, Women’s Contraceptive and Reproductive Experiences Study
Table 3.
Discriminating accuracy of the NBCS model, BWHS model, Gail model, and CARE model in Uganda
| Uganda | NBCS model |
BWHS model |
Gail model |
CARE model |
|||||
|---|---|---|---|---|---|---|---|---|---|
| Age group | Number of participants | C-Index | 95% CI | C-Index | 95% CI | C-Index | 95% CI | C-Index | 95% CI |
| <30 | 34 | 0.618 | 0.423-0.813 | 0.594 | 0.399-0.790 | 0.494 | 0.298-0.690 | 0.556 | 0.363-0.748 |
| 30-39 | 97 | 0.555 | 0.439-0.670 | 0.532 | 0.416-0.648 | 0.538 | 0.419-0.656 | 0.536 | 0.419-0.653 |
| 40-49 | 144 | 0.482 | 0.385-0.579 | 0.489 | 0.396-0.581 | 0.482 | 0.389-0.576 | 0.489 | 0.396-0.580 |
| 50-59 | 119 | 0.583 | 0.480-0.687 | 0.502 | 0.398-0.605 | 0.380 | 0.279-0.481 | 0.443 | 0.340-0.545 |
| 60-69 | 55 | 0.371 | 0.213-0.529 | 0.437 | 0.281-0.593 | 0.576 | 0.416-0.737 | 0.446 | 0.291-0.601 |
| ≥70 | 12 | 0.861 | 0.584-1.00 | 0.625 | 0.279-0.971 | 0.611 | 0.252-0.970 | 0.667 | 0.332-1.00 |
| 35-70 | 386 | 0.507 | 0.449-0.565 | 0.496 | 0.440-0.552 | 0.481 | 0.424-0.537 | 0.492 | 0.435-0.548 |
| Total | 461 | 0.525 | 0.472-0.578 | 0.515 | 0.463-0.566 | 0.506 | 0.454-0.557 | 0.512 | 0.461-0.564 |
|
| |||||||||
| Age-adjusted (35-70) | 386 | 0.511 | 0.428-0.594 | 0.485 | 0.403-0.567 | 0.445 | 0.362-0.528 | 0.401 | 0.316-0.487 |
| Age-adjusted | 461 | 0.531 | 0.459-0.603 | 0.506 | 0.434-0.578 | 0.473 | 0.398-0.548 | 0.430 | 0.354-0.506 |
NBCS, Nigerian Breast Cancer Study; BWHS, Black Women Health Study; CARE, Women’s Contraceptive and Reproductive Experiences Study
Table 4.
Discriminating accuracy of the NBCS model in Cameroon and Uganda using local incidence rates
| All | Cameroon | Uganda | |||||||
|---|---|---|---|---|---|---|---|---|---|
|
|
|||||||||
| Age group | N | C-Index | 95% CI | N | C-Index | 95% CI | N | C-Index | 95% CI |
| <30 | 60 | 0.628 | 0.486-0.771 | 26 | 0.644 | 0.414-0.873 | 34 | 0.642 | 0.447-0.838 |
| 30-39 | 237 | 0.556 | 0.483-0.630 | 140 | 0.568 | 0.472-0.664 | 97 | 0.551 | 0.435-0.667 |
| 40-49 | 318 | 0.545 | 0.480-0.609 | 174 | 0.605 | 0.520-0.690 | 144 | 0.484 | 0.387-0.581 |
| 50-59 | 269 | 0.543 | 0.474-0.612 | 150 | 0.561 | 0.468-0.653 | 119 | 0.583 | 0.480-0.687 |
| 60-69 | 102 | 0.556 | 0.442-0.671 | 47 | 0.556 | 0.382-0.729 | 55 | 0.377 | 0.217-0.538 |
| ≥70 | 24 | 0.671 | 0.447-0.896 | 12 | 0.600 | 0.193-1.00 | 12 | 0.889 | 0.660-1.00 |
| Total | 1,010 | 0.562 | 0.518-0.607 | 549 | 0.564 | 0.516-0.612 | 461 | 0.533 | 0.480-0.585 |
|
| |||||||||
| Age-adjusted | 1,010 | 0.562 | 0.517-0.608 | 549 | 0.590 | 0.529-0.651 | 461 | 0.530 | 0.457-0.603 |
Figures 1 and 2 showed the distribution of the 5-year predicted risks of breast cancer for three risk prediction models in Cameroon and Uganda, respectively. The distinguishing ability of the NBCS model was significantly better in Cameroon. The NBCS model with Nigerian incidence rates predicted that 69 (24.4%) cases in Cameroon had a 5-year risk ≥1.66%, compared with 32 (12.0%) controls, yielding a two-fold difference (Supplemental Table S3). If using Cameroon local incidence rate in the NBCS model, the predicted 5-year risk of controls was about two third lower than that from the NBCS model using Nigerian rates. By contrast, the predicted 5-year risk from the BWHS model was higher than those from the NBCS models. These probably reflected the population incidence rates used in the models (Supplemental Table S1). In Uganda, the NBCS model with Nigerian incidence rates predicted that 29 (12.3%) cases had a 5-year risk ≥1.66%, compared with 22 (9.8%) controls. The NBCS model with Uganda rate gave similar risk estimates, while the BWHS model had higher estimates of risks.
Figure 1.
The distribution of 5-year predicted risks for breast cancer cases and controls in Cameroon with the Nigerian Breast Cancer Study (NBCS) model and Black Women Health Study (BWHS) model. Risk was truncated at 5%.
Figure 2.
The distribution of 5-year predicted risks for breast cancer cases and controls in Uganda with the Nigerian Breast Cancer Study (NBCS) model and Black Women Health Study (BWHS) model. Risk was truncated at 5%.
Discussion
In this study, we evaluated the applicability of the NBCS risk assessment model (18) in Cameroon and Uganda. To our knowledge, this is the first study to investigate the effectiveness of the NBCS model in populations outside of Nigeria. We found the performance of the NBCS model was moderate in Cameroon but not good in Uganda. We also evaluated other risk assessment models developed in the United States among the two SSA populations.
In Cameroon, we found the NBCS model performed better than the BWHS model and the two Gail models. We found the discriminatory ability, as measured by an age-adjusted C-index of 0.602, was moderate. Additionally, the age-adjusted C-index of the BWHS model was 0.573 in Cameroon, which is similar to 0.574 in Nigeria (18) and slightly lower than 0.59 in African Americans (25). The absolute risk estimates depend on incidence rates being used because risk factors for risk assessment models cannot explain all variation in breast cancer incidences. Therefore, a model calibration using local incidence rates is necessary. The NBCS model with original Nigerian rates and the BWHS model with African American rates may overestimate absolute risk in Cameroon. On the other hand, the NBCS model with Cameroon rates might underestimate absolute risk because the Yaounde Cancer Registry might not capture all breast cancer cases.
In Uganda, we found poor discriminating accuracy of the NBCS model as well as other models evaluated. The poor performance of the NBCS model in Uganda may be partially explained by the variable effect of alcohol consumption on development of breast cancer; the strength of association was weaker in Uganda than in Nigeria and Cameroon observed in this and previous study (21). In addition, we noted significant differences in distribution of multiple risk factors, including age at menarche, parity, family history of breast cancer, length of breastfeeding, use of oral contraceptives, and BMI. This illustrates the heterogeneity of SSA populations, underscoring the importance of developing population-specific breast cancer risk assessment models.
In addition to population-related differences in risk profiles, increased genetic susceptibility for the development of breast cancer and early onset of disease among native African women are also important for risk assessment in SSA populations. African women are more likely to experience early-onset breast cancer, with a median age of diagnosis of 45 years of age compared to 60 years of age in white and African American women (26,27). Zheng et al. found an exceptionally high frequency in BRCA1/BRCA2 mutations among Nigerian breast cancer patients (7.0% and 4.1%, respectively) (20). Adedokun and colleagues found similarly high frequency of BRCA1/BRCA2 mutations among breast cancer patients in Cameroon and Uganda (5.6% and 5.6%, respectively) (28). By contrast, the mutation prevalence of BRCA1 and BRCA2 was 1.6% and 1.9%, respectively, in African American patients (29), and 0.9-1.0% and 1.3-1.5% in predominantly European-ancestry populations (30,31). These findings underscores the need for increased access to genetic testing and counseling in resource-limited settings. It also suggests that integration of genetic risk factors into risk prediction models for native African populations could improve model accuracy.
The main use of the NBCS risk prediction model is to help identify high risk women for mammography screening in SSA where limited medical and economic resources make universal breast cancer screening not practical now. At the individual level, the model can be used for individual risk counseling and promoting a healthy lifestyle. Knowing their own cancer risk may motivate women to stop regular alcohol drinking. On the other hand, we acknowledge that the risk prediction model is not arcuate enough and the major hindrances to reduce breast cancer mortality in SSA are delayed diagnosis and insufficient treatment.
There are several limitations to this study. First, the data was collected in single hospital in Uganda and Cameroon, which could affect the generalizability of the results of this study. Further replication of the model in other populations in Uganda and Cameroon as well as in other SSA countries is necessary to determine the applicability of the NBCS model in SSA. Second, our study is a case-control study, so we are not able evaluate the calibration accuracy of the models. We postulated that the NBCS model with Nigerian rates and the BWHS model with African American rates could overestimate absolute risk in Cameroon, while the NBCS model with Cameroon rates might underestimate absolute risk. However, prospective cohort studies are needed to evaluate model calibration. Third, controls were enrolled from women who visited hospitals for various reasons (most were healthy women accompanying their relatives or friends to other clinics), and may not representative of the underlying population that gave rise to the cases. This may bias the estimates of relative risks. Last, death registrations in Africa may be inaccurate and we used mortality rates from entire Nigeria to determine competing risks in NBCS model. Therefore, the absolute risk estimates for older women may not be accurate.
In summary, the present study demonstrated that the NBCS model performed moderately in Cameroon but poorly in Uganda. We learned two key lessons. First, a re-calibration of the model with local breast cancer incidence rates is necessary, and this requires reliable data from local population-based cancer registries. Second, heterogeneity in risk factor profiles across SSA populations, as seen in the differences in risk factor profiles between Cameroon and Uganda (Table S2), may signify a need for population-specific risk prediction models of breast cancer. Given increased genetic component in SSA populations, future risk prediction model development that incorporates genetic risk factors is desirable.
Supplementary Material
Acknowledgements
The authors wish to thank the dedicated staff members of Yaoundé General Hospital, Cameroon, and Mulago Hospital, Uganda, for their expert technical assistance with this research. They also wish to Dr. George Enow Orock for sharing breast cancer incidence rates of Yaounde Cancer Registry. D. Huo received a grant from National Cancer Institute (R01CA228198). D. Huo and O.I. Olopade received grants from National Institute on Minority Health and Health Disparities (R01MD013452) and Breast Cancer Research Foundation (BCRF-22-071).
Funding:
This research was supported by awards from National Institutes of Health (R01CA228198 and R01MD013452) and Breast Cancer Research Foundation (BCRF-22-071).
Footnotes
The authors declare no potential conflicts of interest.
References
- 1.Fitzmaurice C, Akinyemiju TF, Al Lami FH, Alam T, Alizadeh-Navaei R, Allen C, et al. Global, Regional, and National Cancer Incidence, Mortality, Years of Life Lost, Years Lived With Disability, and Disability-Adjusted Life-Years for 29 Cancer Groups, 1990 to 2016: A Systematic Analysis for the Global Burden of Disease Study. JAMA oncology 2018;4(11):1553–68 doi 10.1001/jamaoncol.2018.2706. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71(3):209–49 doi 10.3322/caac.21660. [DOI] [PubMed] [Google Scholar]
- 3.Jedy-Agba E, McCormack V, Adebamowo C, Dos-Santos-Silva I. Stage at diagnosis of breast cancer in sub-Saharan Africa: a systematic review and meta-analysis. The Lancet Global health 2016;4(12):e923–e35 doi 10.1016/S2214-109X(16)30259-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-Tieulent J, Jemal A. Global cancer statistics, 2012. CA Cancer J Clin 2015;65(2):87–108 doi 10.3322/caac.21262. [DOI] [PubMed] [Google Scholar]
- 5.Huo D, Adebamowo CA, Ogundiran TO, Akang EE, Campbell O, Adenipekun A, et al. Parity and breastfeeding are protective against breast cancer in Nigerian women. British journal of cancer 2008;98(5):992–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Howell A, Anderson AS, Clarke RB, Duffy SW, Evans DG, Garcia-Closas M, et al. Risk determination and prevention of breast cancer. Breast Cancer Res 2014;16(5):446 doi 10.1186/s13058-014-0446-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Warner E Clinical practice. Breast-cancer screening. N Engl J Med 2011;365(11):1025–32 doi 10.1056/NEJMcp1101540. [DOI] [PubMed] [Google Scholar]
- 8.Black E, Richmond R. Improving early detection of breast cancer in sub-Saharan Africa: why mammography may not be the way forward. Global Health 2019;15(1):3 doi 10.1186/s12992-018-0446-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Shieh Y, Eklund M, Madlensky L, Sawyer SD, Thompson CK, Stover Fiscalini A, et al. Breast Cancer Screening in the Precision Medicine Era: Risk-Based Screening in a Population-Based Trial. J Natl Cancer Inst 2017;109(5) doi 10.1093/jnci/djw290. [DOI] [PubMed] [Google Scholar]
- 10.Cintolo-Gonzalez JA, Braun D, Blackford AL, Mazzola E, Acar A, Plichta JK, et al. Breast cancer risk models: a comprehensive overview of existing models, validation, and clinical applications. Breast Cancer Res Treat 2017;164(2):263–84 doi 10.1007/s10549-017-4247-z. [DOI] [PubMed] [Google Scholar]
- 11.Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, Schairer C, et al. Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst 1989;81(24):1879–86. [DOI] [PubMed] [Google Scholar]
- 12.Gail MH, Costantino JP, Pee D, Bondy M, Newman L, Selvan M, et al. Projecting individualized absolute invasive breast cancer risk in African American women. J Natl Cancer Inst 2007;99(23):1782–92 doi 10.1093/jnci/djm223. [DOI] [PubMed] [Google Scholar]
- 13.Boggs DA, Rosenberg L, Pencina MJ, Adams-Campbell LL, Palmer JR. Validation of a breast cancer risk prediction model developed for Black women. J Natl Cancer Inst 2013;105(5):361–7 doi 10.1093/jnci/djt008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Wang X, Huang Y, Li L, Dai H, Song F, Chen K. Assessment of performance of the Gail model for predicting breast cancer risk: a systematic review and meta-analysis with trial sequential analysis. Breast Cancer Res 2018;20(1):18 doi 10.1186/s13058-018-0947-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Han Y, Lv J, Yu C, Guo Y, Bian Z, Hu Y, et al. Development and external validation of a breast cancer absolute risk prediction model in Chinese population. Breast Cancer Res 2021;23(1):62 doi 10.1186/s13058-021-01439-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Huo D, Ikpatt F, Khramtsov A, Dangou JM, Nanda R, Dignam J, et al. Population differences in breast cancer: survey in indigenous african women reveals over-representation of triple-negative breast cancer. J Clin Oncol 2009;27(27):4515–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Eng A, McCormack V, dos-Santos-Silva I. Receptor-defined subtypes of breast cancer in indigenous populations in Africa: a systematic review and meta-analysis. PLoS medicine 2014;11(9):e1001720 doi 10.1371/journal.pmed.1001720. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Wang S, Ogundiran T, Ademola A, Olayiwola OA, Adeoye A, Sofoluwe A, et al. Development of a Breast Cancer Risk Prediction Model for Women in Nigeria. Cancer Epidemiol Biomarkers Prev 2018;27(6):636–43 doi 10.1158/1055-9965.EPI-17-1128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Zhang J, Fackenthal JD, Huo D, Zheng Y, Olopade OI. Searching for large genomic rearrangements of the BRCA1 gene in a Nigerian population. Breast Cancer Res Treat 2010;124(2):573–7. [DOI] [PubMed] [Google Scholar]
- 20.Zheng Y, Walsh T, Gulsuner S, Casadei S, Lee MK, Ogundiran TO, et al. Inherited Breast Cancer in Nigerian Women. J Clin Oncol 2018;36(28):2820–5 doi 10.1200/JCO.2018.78.3977. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Qian F, Ogundiran T, Hou N, Ndom P, Gakwaya A, Jombwe J, et al. Alcohol consumption and breast cancer risk among women in three sub-Saharan African countries. PLoS ONE 2014;9(9):e106908 doi 10.1371/journal.pone.0106908. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Martinez P, Roislien J, Naidoo N, Clausen T. Alcohol abstinence and drinking among African women: data from the World Health Surveys. BMC Public Health 2011;11:160 doi 10.1186/1471-2458-11-160. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Hou N, Ndom P, Jombwe J, Ogundiran T, Ademola A, Morhason-Bello I, et al. An epidemiologic investigation of physical activity and breast cancer risk in Africa. Cancer Epidemiol Biomarkers Prev 2014;23(12):2748–56 doi 10.1158/1055-9965.EPI-14-0675. [DOI] [PubMed] [Google Scholar]
- 24.Parkin DM, Jemal A, Bray F, Korir AR, Kamaté B, Singh E, et al. Cancer in Sub-Saharan Africa. Geneva, Switzerland: The Union for International Cancer Control; 2019. [Google Scholar]
- 25.Boggs DA, Rosenberg L, Adams-Campbell LL, Palmer JR. Prospective approach to breast cancer risk prediction in African American women: the black women’s health study model. J Clin Oncol 2015;33(9):1038–44 doi 10.1200/JCO.2014.57.2750. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Sighoko D, Ogundiran T, Ademola A, Adebamowo C, Chen L, Odedina S, et al. Breast cancer risk after full-term pregnancies among African women from Nigeria, Cameroon, and Uganda. Cancer 2015;121(13):2237–43 doi 10.1002/cncr.29305. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.DeSantis CE, Ma J, Gaudet MM, Newman LA, Miller KD, Goding Sauer A, et al. Breast cancer statistics, 2019. CA Cancer J Clin 2019;69(6):438–51 doi 10.3322/caac.21583. [DOI] [PubMed] [Google Scholar]
- 28.Adedokun B, Zheng Y, Ndom P, Gakwaya A, Makumbi T, Zhou AY, et al. Prevalence of Inherited Mutations in Breast Cancer Predisposition Genes among Women in Uganda and Cameroon. Cancer Epidemiol Biomarkers Prev 2020;29(2):359–67 doi 10.1158/1055-9965.EPI-19-0506. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Palmer JR, Polley EC, Hu C, John EM, Haiman C, Hart SN, et al. Contribution of Germline Predisposition Gene Mutations to Breast Cancer Risk in African American Women. J Natl Cancer Inst 2020;112(12):1213–21 doi 10.1093/jnci/djaa040. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Hu C, Hart SN, Gnanaolivu R, Huang H, Lee KY, Na J, et al. A Population-Based Study of Genes Previously Implicated in Breast Cancer. N Engl J Med 2021;384(5):440–51 doi 10.1056/NEJMoa2005936. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Breast Cancer Association C, Dorling L, Carvalho S, Allen J, Gonzalez-Neira A, Luccarini C, et al. Breast Cancer Risk Genes - Association Analysis in More than 113,000 Women. N Engl J Med 2021;384(5):428–39 doi 10.1056/NEJMoa1913948. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.