Breast cancer risk factors that are identical and differential associated with menopause status of women in Northeast Brazil


 Background: This case-control study was aimed at the identification of factors that modify breast cancer risk in a population from Northeastern Brazil. Breast cancer risk factors in this population are poorly understood. Methods: Data of 291 patients with invasive breast cancer and 291 healthy controls were obtained from medical records and personal interviews. Odds ratios (OR) and confidence intervals (CI) were determined using univariate and multiple regression analysis. Results: Oral contraceptive use increased the risk of breast cancer among pre- and post-menopause women 3.661 (95% CI: 1.81-7.39; p < 0.001) and 3.086 (95%CI: 1.85-5.79; p < 0.001) times. Family history increased the risk of breast cancer among pre- and post-menopause women 3.791 (95%CI: 1.85-7.77; p < 0.001) and 3.538 (95%CI: 2.16-5.79; p = 0.000) times. Risk of breast cancer among post-menopause women who regularly consumed alcohol was 4.21 (95%CI: 2.08-8.50; p < 0.001) times increased compared to women who did not consume alcohol. Obese and overweight post-menopause women had 2.117 (95%CI: 1.15-3.88) and 1.217 (95%CI: 0.74-2.19) times increased risk, compared to women with normal weight ( p = 0.044). Waist-to-hip ratio > 0.8 increased the risk of breast cancer among pre- menopause women 2.026 (0.94-4.38; p = 0.067) times in the univariate analysis. Nulliparous post-menopause women had 2.667 (95%CI: 1.30-5.50; p = 0.070) times increased risk. Univariate analysis of nulliparity among pre-menopause women indicated 2.59 (OR= 0.386; 95%CI: 0.15-0.98; p = 0.038) times decreased risk. Conclusions: Oral contraceptive use, family history and alcohol consumption increased the risk of breast cancer among pre- and post-menopause women. High body mass index increased the risk only among post-menopause women, whereas unadjusted high waist-to-hip ratio increased the risk of pre- menopause women. Nulliparous pre-menopause women tended to have decreased risk, whereas nulliparous post-menopause women tended to have increased risk of breast cancer.

Since 2003, Brazil has a public program for early detection of breast cancer and mammography screening is propagated in the month of October ("Outubro rosa"). In Northeastern Brazil, most women rely on information about BC on television and very few information is given about preventable risk factors, respectively modi able factors that decrease the risk of BC [16]. There is an urgent need to identify BC risk factors in this population and if modi able, to propagate their prevention in health campaigns. The present case-control study focused on identi cation of lifestyle-related and reproductive risk factors in a population from Northeastern Brazil. The identi cation of several risk factors indicates remarkable differences of pre-and post menopause women that were not reported in previous studies with any Brazilian population.

Study population
Patients were eligible if diagnosed within 36 months from recruitment with invasive BC and aged 18 years or older. Patients with in situ tumors were excluded from the study. Patients with breast cancer recurrence were also excluded from the study. Data from breast cancer patients were sampled in two reference centers for breast cancer treatment in the state of Paraíba, Northeastern Brazil: The "Fundação Assistencial da Paraíba" Hospital (FAP) in Campina Grande and "Napoleão Laureano" Hospital (HNL) in João Pessoa. Both public hospitals receive mainly low-income patients and treat more than 95% of all BC cases in the state. Patients seeking for treatment come from regions as far as 600 km from the reference centre. João Pessoa, capital of the state of Paraíba, has about 0.8 million inhabitants and is located on the coast. Campina Grande, with about 0.4 million inhabitants, the second most populated urban centre in Paraíba, is located about 120 km away from the capital in the inland of the state. Similarly to other states of Northeastern Brazil, Paraíba has mixed-ethnicity population composed of Indigenous, African and European ancestry.
Data of controls were obtained from both public hospitals and in three public health care centers. Many patients were from rural areas. Therefore, a comparable part of controls was also recruited from rural areas: Two of the three health care centers were in rural areas, about 40km away from Campina Grande. Women with any type of cancer were excluded from the study as controls. Women with any other chronic disease, like diabetes and heart disease were also excluded as controls. Furthermore, it was avoided to interview more than one healthy women of the same family. Controls were randomly selected in health care centers and directly asked by one of the authors to participate as volunteers. Most controls (>60%) accompanied children or grandchildren for vaccine or other medical examination. Other controls visited the health care center mainly because of injuries, colds or gastrointestinal infections. Controls interviewed in hospitals were healthy companions of non-cancer patients. Controls were eligible for the study if they were aged over 18 years old and agreed to participate in the study. Participation rate of patients and controls was about 90% and 85%.

Data sampling
Breast cancer patients and controls were interviewed between March 2017 and January 2019. The study included 291 women with invasive operable breast cancer, diagnosed and treated between 2016 and 2019. No signi cant differences were observed between patients of both reference centers. A structured questionnaire was applied to interview patients in chemotherapy and radiotherapy units of both hospitals. Clinical and histopathological data of patients were obtained from medical records. Height and weight had been measured in both reference centers before any therapeutic treatment and were also obtained from medical records. The same questionnaire was applied to interview healthy controls in waiting rooms of health care centers. All interviews were performed by one of the authors.
Minimum wage and multiple values were used to characterize income. This is a popular and well-known method to de ne economic level among low-and middle-class subjects. Information about ethnic origin was self-reported by participants. Educational level was de ned as follows: 1. Incomplete elementary school with duration of less than nine years was de ned as "Basic"; 2. Elementary school with duration of nine years was de ned as "Low" 3. High school with duration of 12 years was de ned as "middle". 4. Higher educational levels were de ned as "high". Data about ethnic origin were self-reported by participants. Women were asked if they used ever or never oral contraceptives. When asked about family history, interviewed women often confounded BC with other types of cancer. Therefore present information about family history referred to different types of cancer and not to BC alone. Family history included BC and other types of cancer among rst-degree relatives. Physical activity was de ned as any type of regular low, moderate or strong body movement.

Statistical analysis
All statistical analyses were performed using the SPSS STATISTICS™ software (SPPS; IBM company; version 24). The t-test was applied to compare continuous variables. Pearson's Chi-Square (χ2) test was applied to compare categorized variables. To quantify associations among single risk factors for BC, nominal logistic regression analysis was applied. Data of controls served as reference group. Results were presented as adjusted odd ratios (OR), 95% con dence interval (CI) and p-value of likelihood ratio tests. Signi cant univariate regression analysis variables were used for regression modeling with multiple adjusted variables: Variables with signi cance level less than 0.2 in the univariate analysis were included in the model. Then, variables with signi cance level less than 0.05 were kept in the model. Backward selection was used when signi cant variables were selected. The nal model was tested for tness using the likelihood ratio test.
In the univariate regression analysis, pre-menopause nulliparous women had 2.59 (OR = 0.386; 95%CI: 0.15-0.98) decreased risk of BC compared to parous women (p = 0.038; Table 2). The comparison of WHR among pre-menopause women indicated that those with value > 0.8, had 2.026 (95%CI: 0.94-4.38) times increased risk of the disease, compared to women with WHR of ≤ 0.8 (p = 0.067; Table 2). In the adjusted multiple analysis model, only variables oral contraceptive use, family history and alcohol consumption remained heterogeneously distributed between pre-menopause cases and controls ( Table 2): Oral contraceptive use and family history increased the risk of BC by 3.610 (95%CI: 1.77-7.38) and 3.791 (95%CI: 1.85-7.77) times, compared to women who claimed not to have used oral contraceptives before disease diagnosis, respectively, without family history (p < 0.001; p < 0.001; Table 2). Consumption of alcohol increased risk of BC 2,059 (95%CI: 0.90-4.73) times, compared to women who never consumed it (p = 0.086; Table 2).
In the group of post-menopause women with age at menarche < 12 years, the risk of the disease increased 1.918 (95%CI: 1.01-3.66) times compared to women with age at menarche ≥ 12 years (p = 0.044; Table 2). In the adjusted multiple analysis model, variables nulliparity, oral contraceptive use, family history, BMI and alcohol consumption remained heterogeneously distributed between post-menopause cases and controls (Table 2): In contrast to pre-menopause women, nulliparity tended to increase the risk of BC among postmenopause women. Nulliparous women had 2.667 (95%CI: 1.30-5.50) times increased risk of the disease compared to parous women (p = 0.070; Table 2). Post-menopause women, who used oral contraceptives and had positive family history of BC, had 3.086 (95%CI: 1.85-5.79) and 3.576 (95%CI: 2.21-5.79) times increased risk of the disease (p < 0.001; p < 0.001; Table 2). In contrast to pre-menopause women, increased BMI among post-menopause women was positively associated with risk of the disease: Obese and overweight women had 2.117 (95%CI: 1.15-3.88) and 1.271 (95%CI: 0.74-2.19) increased risk of the disease compared to women with normal weight (p = 0.048; Table 2). Post-menopause women who consumed any type of alcohol had 4.21 (95%CI: 2.08-8.50) times increased risk of the disease compared to those who did not consume alcohol (p < 0.001; Table 2).
Breastfeeding had no signi cant impact on risk of BC among parous women (p = 0.081 Table 2). Parous cases and controls had on average 3.27 (s = 2.21) and 3.12 (s = 2.56) children (p = 0.479), respectively. In the multiple analysis model, women with one and two, respectively three children, had a 2.52 (OR = 0.397; 95%CI: 0.21-0.76) and 1.61 (OR = 0.621; 95%CI: 0.38-1.02) times decreased risk of disease compared to women with more than three children (p = 0.016; Table 2). Similarly, as in the case of post-and premenopause women, oral contraceptive use, family history and alcohol consumption were associated with increased risk of the disease (p < 0.001; p < 0.001; p < 0.001; Table 2).

Discussion
In the present study, oral contraceptive use, family history and alcohol consumption were positively associated with risk of BC. Their effect was similar among pre-and post-menopausal parous women, respectively. Anthropometric measures BMI and WHR in contrast, were differently associated with risk of BC between pre-and post-menopause women. Reproductive risk factors age at menarche and nulliparity also showed remarkable differences between pre-and post-menopause women.
Age at menarche <12 years increased the risk of the disease among post-menopause women, but not among pre-menopause women. However, this effect remained not signi cant after data adjustment. In agreement with present results, recent studies carried out in Thailand, Brazil and Norway revealed increased risk of BC associated with younger age at menarche [14,19,20].
Post-menopause nulliparous women of the present study had increased risk of the disease. Among premenopause women in contrast, nulliparity as a single variable, decreased the risk of BC, but had no signi cant effect after data adjustment. Parity is a well-established modulator of risk of BC [21]. Previous studies have associated nulliparity with an overall increased risk of the disease [20,22]. Nulliparity as a risk factor for BC was also identi ed in a previous study carried out in Brazil [14]. Furthermore, the effect of parity depends on menopause status: Parous pre-menopause women, mainly at the age of 35 or older, had increased risk of the disease until 15 years after delivery [23]. Among post-menopause women in contrast, nulliparity was associated with increased risk of the disease, explaining the general protective effect of parity, as most BC cases are found among post-menopause women [21]. The dual effect of parity that depends on menopause status, may be also present in the present population, reducing risk among premenopause women, increasing the risk among post-menopause women, respectively. However, present data indicated the effect of parity, comparing nulliparous and parous women, but did not establish a positive association between high number of children and reduced risk among post-menopause women.
Furthermore, in contrast to present results the protective effect of parity among parous women generally increases with higher number of children [21].
The use of oral contraceptives increased the risk of BC between pre-and post-menopause women in the present study. Previous ndings support the idea that the use of oral contraceptives increases the risk of BC: In literature, oral contraceptive use is a well-established risk factor for BC [5,19,22]. A previous Brazilian study also identi ed oral contraceptive use for more than ve years as risk factor for BC among women aged ≥45 years [24]. Studies have indicated that the risk of BC increases only during or after recent use: Recent use of oral contraceptives was associated with increased risk of the disease among young women [25]. An observational study conducted for more than four decades in the United Kingdom showed that the risk of BC decreased approximately ve years after stopping use of oral contraceptives [26]. Present data also indicated that positive family history increased the risk of BC of both pre-and postmenopause women. In agreement with present results, recent studies performed in Thailand and Brazil, also identi ed rst-degree family history of BC as a variable that increased the overall risk of the disease [10,19].
According to the World Health Organization (WHO), between 2010 and 2014, overweight and obesity increased from 51.10% to 54.10%, respectively, from 17.80% to 20.00% among Brazilian adults [17]. In the present study, antrophometric measures BMI and WHR showed differences between pre-and postmenopause women: Results have indicated that obesity and overweight, determined as BMI, was positively associated with risk of BC among post-menopause women, but not among pre-menopause women. Obesity and overweight are considered the most important modi able lifestyle-related risk factor among women, especially due to its impact on BC [6]. In studies that used anthropometric measures, obesity and overweight increased the overall risk of BC [15,19,20,22]. A recent study performed in Southern Brazil, associated obesity, respectively high BMI, to increased risk of BC among post-menopause women and to decreased risk of BC among pre-menopause women [11]. Similarly, recent studies, all based on BMI, identi ed obesity and overweight as risk factors for BC among post-menopause women, whereas results for pre-menopause women remained con icting [5,27]. A recent study performed in the United Kingdom revealed that low BMI and low waist-to-hip ratio increased the risk of pre-menopause women, whereas both measures were positively associated with risk of BC among post-menopause women [6]. In a metaanalysis, Amadou and colleagues (2013) pointed out that increased BMI of post-menopause women was positively associated with BC, whereas high WHR increased the risk of both pre-menopause and post-menopause women [18]. The present results about BMI, were consistent with most previous studies.
However, in the case of WHR, univariate analysis of present data indicated borderline signi cance among pre -menopause women, but not among post-menopause women. Only obese pre-menopause women tended to have increased risk of the disease.
Pre-and post-menopause women of the present study, who consumed some type of alcohol, had increased risk of the disease. This effect was strong for post-menopause women, but week for pre-menopause women. A recent study carried out in Rio de Janeiro, Brazil, also identi ed alcohol consumption as a risk factor for BC [12]. A previous study indicated that even low doses of alcohol could increase the risk of BC and that there is a positive relationship between dose of alcohol and risk of the disease [28]. The effect of alcohol on the etiology of BC could be by epigenetic mechanisms [29]. A recent study on pre-menopause women suggested that alcohol had negative effect only on individuals who also had family history of BC and lower folic acid intake [30]. Therefore, the effect of alcohol depends on the context of other risk factors and this may also explain why in other previous studies, alcohol consumption had no impact on the risk of BC [30].
The low availability of data may have impaired the establishment of a positive association between dose of alcohol consumption and risk of BC. It was also not possible to identify any possible protective effect of breastfeeding and high number of children for parous and post-menopause women, respectively. Low number of data generally leads to lower resolution and therefore may have obscured signi cant associations between BC and such factors. The study established an overall positive association between use of oral contraceptives and risk of BC, but important data about the time interval between last use of oral contraceptives and disease diagnosis were missing. The study included patients who received diagnosis of disease within the last three years. Therefore, another limitation of the present study may have been recall biases of interviewed women and information about risk factors may have been confounded.
Furthermore, a selection bias of controls and patients cannot be excluded.

Conclusion
The present results indicated that family history and several lifestyle-related factors increased the risk of BC among women in this study: Family history and oral contraceptive use increased the risk of BC among post-and pre-menopause women. Increased risk of women, who consumed alcohol, was more pronounced among post-menopause than pre-menopause women. Obesity and overweight, measured by BMI, increased the risk of BC among post-menopause women, but not among pre-menopause women. In the univariate analysis, early age at menarche increased the risk of BC among post-menopause women and WHR had borderline signi cance among pre-menopause women. Nulliparity increased risk of postmenopause women and reduced risk of pre-menopause women. These ndings are in agreement with most previous case-control studies about risk factors for BC. Future studies should include larger groups of patients and controls. Furthermore, regarding alcohol consumption, the effect of different doses should be analyzed and in the case of oral contraceptive use, time interval between last use and BC diagnosis should be integrated. Avoiding modi able behaviors such as alcohol consumption and poor diet should be propagated by health authorities mainly during public health campaigns like "Outubro Rosa", that focuses so far nearly exclusively on mammography screening, but not on risk factors.  Odds ratios (OR), confidence intervals (95% CI) and significance level of likelihood ratio tests (P LRT ) of variables are shown for cases stratified for pre and post menopause, respectively, parous women. Control served as reference group. Variables of univariate analysis (OR CRUDE ) were used for modelling (OR ADJUSTED ).