Benign breast disease and breast cancer risk in African women: A case-control study

Purpose To examine the association between benign breast disease (BBD) and breast cancer (BC) in a heterogeneous population of African women. Methods BC cases and matched controls were enrolled in three sub-Saharan African countries, Nigeria Cameroon, and Uganda, between 1998–2018. Multivariable logistic regression was used to test the association between BBD and BC. Risk factors dually associated with BBD and BC were selected. Using a parametric mediation analysis model, we assessed if selected BC risk factors were mediated by BBD. Results Of 6418 participants, 55.7% (3572) were breast cancer cases. 360 (5.7%) self-reported BBD. Fibroadenoma (46.8%) was the most reported BBD. Women with a self-reported history of BBD had greater odds of developing BC than those without (adjusted odds ratio [aOR] = 1.47, 95% CI: 1.13–1.91). Biopsy-confirmed BBD was associated with BC (aOR = 3.11, 95% CI: 1.78–5.44). BBD did not significantly mediate the effects of any of the selected BC risk factors. Conclusions In this study, BBD was associated with BC and did not significantly mediate the effects of selected BC risk factors.


Introduction
Breast cancer is the second leading cause of cancer-related deaths worldwide, with a lifetime risk of 1 in 23 among Eastern African women and about 1 in 46 among Western African women [1,2].The association between Benign Breast Disease (BBD) and breast cancer is documented in the literature, and the magnitude of the relationship varies by BBD lesion type [3][4][5][6].While the consensus is that BBDs are not precursor lesions for breast cancer, research has suggested that BBDs may indicate a background proliferative state of the breast that could herald a cancerous process [3,7], with some studies indicating that the presence of BBD might sometimes re ect a hyperestrogenic state [4,8].

Data collection and measurements
After providing written informed consent, participants completed a structured interviewer-administered questionnaire that collected information on demographics, anthropometrics (i.e., height, waist-hip ratio, and body mass index [BMI]), history of BBD, family history of breast cancer, menstrual and reproductive history (i.e., age at menarche, age at thelarche, age at rst live birth, duration of breastfeeding), physical activity, past medical history, and hormonal contraceptive use.History of BBD was self-reported, physician-diagnosed, and/or biopsy-con rmed.In the current analysis, the history of BBD was considered positive if it was diagnosed at least a year before breast cancer diagnosis (for cases) or study recruitment (for controls) to keep the correct time sequence for causal interpretation.Histologic types of BBD were also enquired, including broadenoma, breast cysts, intraductal papilloma, and atypical hyperplasia, as selectable options, and an open-ended option for other histologic types.In ammatory conditions such as mastitis and breast abscesses were not considered positive BBD history for this study.We also collected data on whether BBD was biopsy-con rmed, age at BBD, date of BBD diagnosis, breast side of BBD, and whether surgical resection was performed.

Statistical analysis
Demographic characteristics were compared between cases and controls using chi-square or Fisher's exact tests for categorical variables and Student's t or Wilcoxon rank-sum tests for continuous variables.Logistic regression examined the relationships between BBD and demographics, anthropometric measurements, and breast cancer risk factors.Three multivariable logistic regression models were t to assess the association between BBD and breast cancer.Model 1 included demographics (age, level of education, study country, ethnicity, and menopausal status).Model 2 included reproductive factors (total duration of breastfeeding, parity, the age at rst live birth) and variables in Model 1. Model 3 included all covariates in Model 2, anthropometries (height, waist-hip ratio), and family history of breast cancer.Both crude odds ratio (OR) and adjusted OR (aOR), with 95% con dence intervals (95% CI), were calculated.We evaluated BBD as a possible mediator of breast cancer and selected risk factors which were based on dual association with breast cancer and BBD in our sample and the existing literature.We followed recommendations of the AGReMA statement in reporting mediation analysis results [27].Alcohol intake, family history of breast cancer, nulliparity or low parity, late menopause, higher adolescent growth velocity, and taller height have been associated with an increased risk of both BBD and breast cancer [26][27][28][29].However, obesity and oral hormonal contraceptives, risk factors for breast cancer, are associated with a reduced risk of BBD [7,27,30,31].Based on signi cant associations found in our analysis and published literature on risk factors related to an increased risk of BBD as well as breast cancer [5, 6, 10, 18, 21-23, 30, 35], we selected age at menarche, parity, age at rst live birth, mean duration of breastfeeding per live birth, hormonal contraceptive use, BMI, alcohol use, and family history of breast cancer as risk factors for breast cancer which BBD could mediate.Causal mediation analysis was performed using parametric regression methods developed by VanderWeele and Vansteelandt [42,43] and implemented in the module paramed in STATA [41].We estimated the natural direct effect (NDE), the effect of selected risk factors on breast cancer risk that is not through BBD, and the natural indirect effect (NIE), the effect of selected risk factors on breast cancer risk that is through BBD.The total effect is the product of NDE and NIE for case-control studies.From NDE and NIE on the OR scale, we estimated the proportion mediated (PM) using the Eq. ( 1) below:

Patient characteristics
Table 1 shows the characteristics of study participants.Of 6,418 participants, 55.7% were breast cancer cases.Most (83.5%) of the women were from Nigeria, 8.9% from Cameroon, and 7.6% from Uganda.
Overall, the mean age at the enrollment/interview was 46 years (SD = 12.7), and 61.2% were premenopausal.The mean age at breast cancer diagnosis was 48 years (SD = 12.0).Compared to controls, cases were more likely to have a known family history of breast cancer, be postmenopausal, have a waist-hip ratio of > 0.85, and have consumed alcohol at least once a week for ≥ 1 year, however, were less likely to have used hormonal contraceptives, were 2cm taller and had about 0.3 kg/m 2 lower BMI on average.The duration of breastfeeding (total and mean per live birth) and age at menarche were also signi cantly associated with breast cancer (Table 1).See Supplementary Table 1 for the study participant description by country and the distribution of ethnic groups.a The age at the time of diagnosis for cases, and at the time of interview for controls.
c At least one alcoholic beverage a week for one year or longer.
d Hormonal contraceptives include oral, injectable, implant, and intrauterine contraceptive devices.a The age at the time of diagnosis for cases, and at the time of interview for controls.
c At least one alcoholic beverage a week for one year or longer.
d Hormonal contraceptives include oral, injectable, implant, and intrauterine contraceptive devices.a The age at the time of diagnosis for cases, and at the time of interview for controls.
c At least one alcoholic beverage a week for one year or longer.
d Hormonal contraceptives include oral, injectable, implant, and intrauterine contraceptive devices.Characteristics of study participants with and without self-reported BBD (Table 3) were compared in the control group.After adjusting for age group and menopausal status among controls, younger age at menarche, older age at rst live birth, shorter duration of breastfeeding, positive family history of breast cancer, and a waist-hip ratio of ≤ 0.85 were signi cantly associated with higher odds of BBD.Higher education was also associated with a positive history of BBD.Higher-order parity (≥ 4) was associated with lower odds of BBD in adjusted analysis.There was no clear relationship between the odds of BBD and reported adult height.a The age at the time of interview for controls.
c At least one alcoholic beverage a week for one year or longer.a The age at the time of interview for controls.
c At least one alcoholic beverage a week for one year or longer.a The age at the time of interview for controls.
c At least one alcoholic beverage a week for one year or longer.a The age at the time of interview for controls.
c At least one alcoholic beverage a week for one year or longer.a The age at the time of interview for controls.
c At least one alcoholic beverage a week for one year or longer.
a The age at the time of interview for controls.
c At least one alcoholic beverage a week for one year or longer.
d Hormonal contraceptives include oral, injectable, implant, and intrauterine contraceptive devices.
e Adjustment for age category and menopausal status.

Association between BBD and breast cancer
Overall, breast cancer was associated with higher odds of a positive history of BBD (crude OR 1.43, 95% CI: 1.14-1.78).After adjusting for multiple covariates (model 3 in Table 4), the association between BBD and breast cancer remained statistically signi cant (aOR 1.47, 95% CI: 1.13-1.91).The positive association existed in all three countries, although it was only tend towards signi cant in Nigeria.Overall, biopsied BBD was statistically signi cantly associated with increased odds of breast cancer in the multivariable analysis, and the strength of association was stronger than that for self-reported BBD (aOR 3.11, 95% CI 1.78-5.44)(Table 4).b Adjusted for all covariates in model 1, total breastfeeding duration, parity, and age at rst live birth.
c Adjusted for all covariates in models 1 and 2, height, waist-hip-ratio, and family history of breast cancer *P < 0.05; **P < 0.01; ***P < 0.001.† P = 0.082 Evaluating BBD as a mediator of selected risk factors We identi ed breast cancer risk factors that were signi cantly associated with BBD in controls.Age at menarche, age at rst live birth, mean duration of breastfeeding, parity, and family history of breast cancer were signi cantly associated with BBD in controls (Table 3) and are known breast cancer risk factors in the literature with associations con rmed in our study (Table 1).Waist-hip-ratio was also associated with BBD but had an opposite relationship with breast cancer in our study, thus was not included in causal mediation analysis.Although BMI, alcohol consumption, and hormone contraceptive use were not signi cantly associated with self-reported BBD in our dataset (Table 3), they were selected for mediation analysis because of their association with BBD according to existing literature [5,6,10,21].
None of the selected breast cancer risk factors were signi cantly mediated through BBD (Table 5), and BBD accounted for less than 10% of the association of any one of the selected risk factors with breast cancer.a At least one alcoholic beverage a week for one year or longer.
b Hormonal contraceptives included oral, injectable, implant, and intrauterine contraceptive device.
c De ned as having at least one alcoholic beverage a week for the past 6 months or longer.

Discussion
In this large multi-site case-control study of breast cancer in Africa, we report the association between BBD and breast cancer and examine whether BBD could be a mediator of the effects of selected risk factors on breast cancer.We observed that 6.6% of cases had physician-diagnosed BBD as compared with 4.7% in healthy controls.The median interval between BBD and breast cancer diagnosis was 9 years.We found that self-reported BBD was associated with a 47% increased odds of breast cancer and biopsy-con rmed BBD was associated with a 211% increased odds of breast cancer.The association was signi cant in all study countries except in Nigeria, where it was marginally signi cant.
Among controls, lower waist-hip ratio, younger age at menarche, higher age at rst live birth, and a family history of breast cancer were associated with higher odds of BBD after adjusting for age and menopausal status in our study; this is in agreement with the existing literature on BBD-predisposing risk factors [5,18,21,28].Previous studies have not shown a clear relationship between breastfeeding practices and BBD [7,18].However, we found that a longer (both total and mean-per-live-birth) duration of breastfeeding was inversely associated with BBD.We found that a higher level of education was positively associated with BBD, similar to ndings by Dorjgochoo et al. [13], but this may be due to detection bias since more educated women are more likely to consult a physician for breast conditions and/or examine their breasts more frequently [18].
In the present study, we observed a positive association between breast cancer and self-reported physician-diagnosed (aOR = 1.47) and biopsy-con rmed BBD (aOR = 3.11  [16,38].A study of 1,406 African American women failed to show an association between "proliferative BBD without atypia" and subsequent breast cancer risk, although "proliferative BBD with atypia" conferred over three times greater risk of subsequent breast cancer compared to women with nonproliferative lesions [12].Another study in a multi-ethnic cohort of 4,970 women (1,341 African Americans) showed that women with proliferative BBD were 1.7 and 3.8 times (with and without atypia respectively) at greater risk for breast cancer [44].
To our knowledge, our study is the rst to assess the mediating effects of BBD on breast cancer risk in a heterogeneous population of African women.Our analysis suggested that BBD did not signi cantly mediate any selected breast cancer risk factors although it may account for about 10% of the association between age at menarche and breast cancer and 8% between mean breast-feeding duration and breast cancer.Although we did not nd any published studies assessing BBD as a mediator of breast cancer risk, a published study examining the mediating effect of mammographic density (MD) on breast cancer risk found that about 17% of the risk conferred by BBD was mediated by MD [37].
Our study is not without limitations.First, the relatively low frequency of reported non-in ammatory BBD may indicate that BBDs were underreported and thus probably underestimated.This could be partly because women in the African countries are not routinely screened for breast cancer, and as such, majorly palpable breast disease will present to the physician for evaluation.Education is signi cantly associated with health-seeking behavior in African women [39].With about 39% of our study population at or below elementary education, a sizeable proportion of BBD may have gone unreported because a physician did not evaluate them.Further, delay of diagnosis and misdiagnosis is a massive problem in resource-limited settings, increasing the possibility that self-reported BBD may have represented misdiagnosis.Our observation of a stronger association of biopsied BBD than the association of any BBD suggests that misreporting might dilute a true association.Secondly, since we did not extract data on speci c histologic types of BBD but instead relied on participant recall, we could not tease out proliferative vs. nonproliferate, simple vs. complex, and typical vs. atypical BBD lesions, each of which has varying degrees of association with breast cancer [16,38].In developing countries, it is common for physicians to offer tentative histologic diagnoses based on clinical presentations and epidemiologic patterns because of the limited access to pathological diagnostics [19].As self-reported histologic subtypes could not be entirely relied on for accuracy, we did not analyze the association of the reported BBD subtypes with breast cancer to avoid misleading results.Lastly, participants of this study might not be representative of all women in the three African countries and other women in Africa, limiting the generalizability of our ndings.
In conclusion, our ndings highlight that BBD, especially when biopsy-con rmed, is an important risk factor for breast cancer and that BBD does not appear to mediate the effects of traditional breast cancer risk factors in African women.Assessing the differences in breast cancer risk for distinct histologic subtypes is worth further investigation in this population.The effects of uni-vs.multi-focal lesions and surgical treatment for BBD need also to be examined in future studies.Nonetheless, our ndings suggest that when women in these regions present with BBD, they are at an elevated risk for breast cancer and should be considered in breast cancer risk assessment.Future studies using modern imaging technologies and molecular pathology are needed to con rm our ndings and assess the contributions of distinct histologic subtypes of BBD to breast cancer risk. Declarations

Table 1
Characteristics of study participants in the African Breast Cancer Study by case-control status c At least one alcoholic beverage a week for one year or longer.dHormonal contraceptives include oral, injectable, implant, and intrauterine contraceptive devices.

Table 2
Distribution of benign breast disease by case-control status and by study country BBD interval: the years between the date of BBD diagnosis and the date of breast cancer diagnosis for cases; and between the date of BBD diagnosis and the date of interview for controls.
Abbreviations: BBD, benign breast disease; IQR, interquartile range.aBBD interval: the years between the date of BBD diagnosis and the date of breast cancer diagnosis for cases; and between the date of BBD diagnosis and the date of interview for controls.Abbreviations: BBD, benign breast disease; IQR, interquartile range.aBBD interval: the years between the date of BBD diagnosis and the date of breast cancer diagnosis for cases; and between the date of BBD diagnosis and the date of interview for controls.a

Table 3
Distribution of risk factors for self-reported physician-diagnosed BBD among controls in the African Breast Cancer Study Abbreviations: CI, con dence interval; SD, standard deviation; BMI, body mass index.

Table 4
Multivariable logistic regression of the association between benign breast disease and breast cancer, overall and by country a Adjusted for age, level of education, study site (only for all-country model), ethnicity, and menopausal status.

Table 5
Mediation analysis of benign breast disease in the associations between selected risk factors and breast cancer