Study design and setting
We studied women enrolled in the Nigerian Integrative Epidemiology of Breast Cancer (NIBBLE) Study, a case-control study of female breast cancer that recruited participants at six government hospitals in Nigeria, five of whom are located in Abuja (National Hospital, University of Abuja Teaching Hospital Gwagwalada, Asokoro District Hospital, Garki Hospital and Wuse General Hospital) and the sixth hospital, the University of Nigeria Teaching Hospital, in Enugu, between January 2014 and July 2016. The details of the study design and setting have been previously published (2).
Participants
Overall, 508 newly diagnosed patients with primary invasive breast cancer aged 18 years and above were identified at their first visit to the clinical sites. Research nurses informed potential participants about the study and obtained their informed consent. Age-matched hospital-based controls (892) were women who did not have cancer or endocrine diseases and were within ±2.5 year of the age of specific breast cancer patients enrolled within one month in the same hospital. Most (94.0%) of the women approached consented to participate. Research nurses conducted face-to-face interviews in the English language (70.6%) or local Nigerian language (29.4%) according to the patient's preference.
Primary exposure
For the LTPA assessment, we used a modification of the U.S. Nurses’ Health Study (NHS) II physical activity questionnaire (4). The questionnaire measures the average amount of time spent per week on moderate and vigorous leisure-time activities. Participants reported the average time per week spent on each of the following moderate or vigorous activities, in the past year: walking, hiking, jogging, running, bicycling, dancing, playing tennis, soccer, squash; golf, swimming, aerobics, weightlifting or resistance exercise. We calculated participants’ metabolic equivalents (METs) - hour/week of total LTPA by multiplying the number of hours per week of each activity with its corresponding MET values and then summarized all the MET values (32). We excluded one participant with an extreme MET value. The final MET score was used to create two categories of participants: ‘Leisure-time physical active’ - participants who met the WHO PA recommendations of at least 150 minutes of moderate-intensity or 75 minutes of vigorous-intensity aerobic LTPA, or an equivalent combination, and ‘Leisure-time physical inactive’ - for those who did not meet the WHO PA recommendations (33). In addition, we created categories of LTPA in quartiles of METs (< 3.00, 3.00-5.49, 5.50-11.49, 11.50 ≤) based on the distribution of the study sample.
Breast cancer and breast cancer subtypes
Needle core biopsies were performed using Bard Magnum Biopsy Gun®. Breast specimens were fixed in 10% neutral buffered formalin and processed within 48 hours of fixation with a minimum fixation time of 8 hours in Leica® automatic tissue processors at the African Collaborative Center for Microbiome and Genomics Research (ACCME) Laboratory at the Institute of Human Virology, Nigeria.
Histology
Sections of Paraffin-embedded blocks were cut at 3-4 μm and stained with hematoxylin and eosin stains. Histological features were classified according to the 2003 WHO classification of breast diseases and graded using the Nottingham modification of the Bloom-Richardson grading (34). Only participants with final histologic confirmation of breast cancer were included in the analysis.
Immunohistochemistry (IHC)
Histologically confirmed invasive breast tumors were stained by immunohistochemical techniques using the Thermo Scientific Lab Vision primary antibodies (clones ER-SP1; PR-SP2; Her2-SP3) and Thermo Scientific™ Ultra Vision™ Quanto HRP DAB detection kit according to manufacturer’s recommended protocol. In brief, formalin-fixed paraffin-embedded (FFPE) tissues were sectioned serially into 4 μm, placed in histogrip coated microscope slides, incubated overnight at 60 C, deparaffinized in series of xylene (three changes), graded alcohol (2 changes 100%, 90%, and 70% ethanol) and rehydrated in distilled water. Antigen retrieval was performed, sections were washed with Phosphate Buffered Saline (PBS) and blocked with hydrogen peroxide for 10 minutes. Then Ultra V was applied to block nonspecific background staining for 5 minutes. Sections were washed with PBS and primary antibodies (ER-SP1; PR-SP2; Her2-SP3) were applied on the sections and incubated at room temperature for 1 hour followed by application of primary antibody enhancer and HRP polymer. Staining was visualized using Diaminobenzidine (DAB) and counterstained with haematoxylin. Sections were dehydrated and cover slipped. We planned to perform immunohistochemistry (IHC) for all participants recruited into the NIBBLE study but in a few cases the core tissue biopsies were too small so IHC was not feasible.
IHC interpretation
We considered ER and PR were positive if ≥ 1% nuclei of the tumor cells were stained per the American Society College Oncology/College American Pathology (ASCO/CAP) guidelines (35). HER2 staining was scored as 0, 1+, 2+, or 3+ and we considered a staining of 3+ (uniform, intense membrane staining of > 30% of invasive tumor cells) a positive HER2 result (36).
Breast cancer subtypes
Overall, 57% of the cases (292/508) had estrogen, progesterone and human epidermal growth factor 2 test results. Some 32.2% (n=94/292) were classified as HP, and 42.1% (n=123/292) as TNBC. We classified breast cancer subtypes using combinations of the IHC markers as follows (a) hormone receptor positive (HP) were tumors that had positive estrogen and progesterone tests but negative HER2 test and (b) triple-negative breast cancer (TNBC) were tumors that lacked all 3 markers (37).
Covariates
We collected information on age in years, levels of education completed (elementary, completed high school, post-high school with no university degree, completed university), marital status (married, single, separated/divorced/widowed), smoking experience (yes vs. no), alcohol use (yes vs. no), age at menarche, number of pregnancies (0, 1-2, 3-5, 6≤), ever use of oral contraceptive (yes vs. no), menopausal status (premenopausal vs. postmenopausal), and breastfeeding experience of more than one month (yes vs. no). Research nurses measured participants’ height, weight, waist, and hip circumferences and we derived body mass index (BMI kg/m2) and waist-hip ratio (WHR) from these measurements. Participants with extreme values of WHR of less than 0.7 or higher than 1.6 or with a BMI less than 10 kg/m2 or greater than 50 kg/m2 were excluded from the analyses (38). BMI was categorized into < 25, 25 - 29.9, ³30) and WHR was categorized to ≤0.85, and > 0.85. To compute socio-economic status, we calculated the ‘wealth index’ using the following variables - house ownership and type of house owned (e.g. home, apartment, house or duplex); source of drinking water (e.g. from outside, well, borehole, piped or bottled); type of cooking fuel; use of separate room for cooking; type of toilet; and ownership of household goods including car and refrigerator. We used Principal Component Analysis (PCA) with varimax rotation to compute factor scores based on the sum of responses to these variables weighted by their factor loading. We used the first component in the PCA that explained (35%) of the variations in the data, to generate a wealth index (39). The wealth index variable was used to classify participants to low socio-economic status (lowest 40% of the score distribution), middle (middle 30%) and high (highest 30%) socio-economic class.
Statistical Analysis
Overall breast cancer
From the initial study sample (n=1,400), we matched 472 cases with 472 controls based on age (±5 year) using propensity score with the optimal matching procedure making a final sample of 944 participants. To examine bivariate associations between independent variables, primary exposure (LTPA) by cases and controls, we implemented conditional logistic regressions for each independent variable separately. To construct our multivariable models, we selected independent variables with p-value < 0.20 in the bivariate analysis for inclusion. Multiple imputations technique was performed to impute missing values of the independent variables after conducting missing completely at random test (MCAT) (p-value=0.63). For multivariable analyses, we conducted conditional logistic regression models and used Wald tests to identify covariates with significant associations (p-value < 0.05) with risk of overall and molecular subtypes of breast cancer.
Breast cancer subtypes
To examine associations between LTPA and breast cancer subtypes, we used two subsamples – cases classified as HP on immunohistochemistry, and those classified as TNBC. We used unconditional logistic regression models to identify age-adjusted variables with p-value < 0.20 in bivariate analyses. These were included in multivariable unconditional logistic regression models for each breast cancer subtype, separately.
We present the adjusted odds ratios (ORs) and their 95% confidence intervals (CIs) of LTPA variables with breast cancer overall and by subtypes. All analyses were performed using Stata SE version 15.1 (College Station, Texas) and R-Studio Version 1.1.447.