Background

A woman’s risk of developing breast cancer is affected significantly by modifiable risk factors, such as alcohol use, smoking, obesity, and physical activity as well as by non-modifiable risk factors, such as BRCA1 or BRCA2 mutations, family history of breast cancer, and familial breast cancer syndromes [1, 2]. With initiatives aimed at increasing awareness of family medical history, such as National Family History Day, a growing number of Americans will become aware of a family history of breast cancer [3]. Moreover, as the availability of genetic tests grows and consumers’ views on genetic testing become more favorable, more women will pursue genetic tests to assess cancer risk [4]. Women with a family history of breast cancer, a familial cancer syndrome, or a confirmed breast cancer-linked genetic variant, may seek to modify their risk in non-surgical and non-medical ways.

Many women who discover a family history or obtain genetic testing results will receive very little accompanying guidance on strategies for non-medical risk reduction. Even for those who receive test results from a healthcare professional, such as a geneticist or genetic counselor, the vast majority of these practitioners do not provide advice on modifiable risk factors related to lifestyle [5].

The primary objective of this review was to assess the association of modifiable risk factors and breast cancer in women at high inherited risk for breast cancer and highlight the salient information that should be readily available to the growing numbers of women aware of their breast cancer risk due to genetic testing or elucidation of family history. This information could thus provide guidance to the many women seeking to reduce their elevated inherited risk of breast cancer.

Methods

This review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.

Search Strategy

For this systematic review, a medical librarian designed and conducted an electronic search using standardized terms and keywords in Ovid Medline, Embase, Scopus, Cochrane Central Register of Controlled Trials, and Clinicaltrials.gov. A full search strategy is included in the additional files (Additional file 6) and all searches took place from June to August 2019. Results were exported to EndNote. The automatic duplicate finder in EndNote was used to eliminate duplicates and the results were also manually searched for duplicates. The identified 7885 articles were screened first at the level of title then abstract, language, and type of work for exclusion. The remaining studies were screened by full text for exclusion. This was performed by a single reviewer (S.C.). Exclusion was determined by the below criteria. The reference lists of included articles were searched by hand for additional citations.

Eligibility criteria

Eligible studies included published observational studies, including retrospective and prospective case–control and cohort studies evaluating the association of modifiable risk factors on the development of breast cancer in women with a high non-modifiable risk of developing breast cancer. Modifiable risk factors included weight or body mass index (BMI), smoking, alcohol consumption, physical activity, and hormonal contraception (HC) or menopausal hormone therapy (MHT) use. Eligible studies enrolled adults (≥ 18 years old) and identified participants with a non-modifiable risk factor, defined as, first or second-degree family history of breast cancer (FHBC), BRCA1/2 mutations, or a diagnosed familial cancer syndrome. Only studies published in English were eligible.

Studies were excluded if they did not meet the above eligibility criteria and, additionally, if they did not have a control group that was breast cancer-free and had a positive FHBC, BRCA1/2 mutation, or a familial cancer syndrome. In addition, studies were excluded if data were not reported in strata defined by family history or genetic marker status. Studies that evaluated other non-modifiable risk factors, including specific non-BRCA genetic polymorphisms, or other modifiable risk factors, such as dietary patterns were excluded. In addition, studies were excluded if risk factors were only measured in aggregates, such as the effect of smoking and alcohol but neither individually. Studies involving women that had undergone chemoprevention or prophylactic oophorectomy were excluded. Studies were excluded if they did not use incident breast cancer risk as a primary outcome, including if they measured the outcome of the age of onset of breast cancer. Finally, studies were excluded if only an abstract was available for retrieval.

Quality

Quality was assessed across the body of evidence based on the aspects of quality used by the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) approach (risk of bias, inconsistency, indirectness, imprecision, and publication bias). One reviewer (S.C.) reviewed all papers for their quality based on the GRADE guidelines.

Data extraction and analysis

A data extraction form was developed and used for each study to collect any data (confidence intervals and p values) pertaining to the above-specified modifiable risk factors. Study characteristics, sample size, time to follow-up, subjects’ breast cancer type (ductal carcinoma in situ (DCIS) vs. lobular carcinoma in situ (LCIS) vs. ductal carcinoma (DC) vs. lobular carcinoma (LC)) were also extracted from each paper and recorded on the form. Data and study information were extracted by S.C. and all included studies were reviewed at least twice. Each relevant data point was classified based on the type of risk factor exposure (e.g., current, past, adolescent, etc.) as well as based on whether it indicated a significant increase in risk, decrease in risk, or no association with risk of developing breast cancer.

If a study presented both adjusted and unadjusted data, only the most adjusted data were included. For studies that reported data separately on women with 1st- and 2nd-degree FHBC, only data on women with 1st-degree FHBC were included. Finally, if a study separated women with invasive and in situ breast cancer, only data on women with invasive cancer were included.

The significance of the data was determined by the confidence interval (CI) or the p value, if the CI was not given. For data that included multiple levels of a single modifiable risk factor exposure (e.g., increasing glasses of alcohol per day), the p-trend was used to determine significance, if available. Otherwise, the CI of the highest level was included.

After classifying data based on exposure categories and association with breast cancer risk, if an exposure category contained more than one article (before division by subgroup, e.g., BRCA1 vs. BRCA2) it was included in the corresponding bar chart. If two articles used the same study population and reported data on the same exposure, only the article with the larger sample size was included in the chart.

Results

The literature search (Fig. 1) identified 7885 citations. In total, 7521 articles were excluded by title or abstract leaving 364 articles for full review. The full text of two of these articles could not be obtained. Of the 362 articles reviewed, 95 studies met the inclusion criteria. Four studies were excluded from the review because they reported on the same exposure in the same study population as other included studies but utilized smaller sample sizes. Two additional studies were added based on screening the citations of relevant articles. Thus, a total of 93 studies were included in the systematic review.

Fig. 1
figure 1

Systematic review chart

Full size image

Thirty studies included participants with BRCA1/2 mutations and 3 additional studies included women with BRCA1/2 mutations and women with family history. Ten studies were prospective (Table 1) and 23 were retrospective (Table 2). The 33 studies represented 27 distinct study populations and none of the included studies using the same study population reported on the same risk factor.

Table 1 Characteristics of included studies on BRCA in prospective studies
Full size table
Table 2 Characteristics of included studies on BRCA in retrospective studies
Full size table

Sixty studies included women with a positive family history of breast cancer (FHBC) and 3 additional studies included women with BRCA1/2 mutations and women with FHBC for a total of 63 studies. Twenty-five studies were prospective (Additional file 4: Table S1), and 38 studies were retrospective (Additional file 5: Table S2). The sixty-three studies represented fifty-two study populations. Included studies that used the same study populations investigated different modifiable risk factors or different categories of exposure for the same modifiable risk factor (e.g., alcohol/day vs. lifetime). Of the 12 studies looking at BMI/weight in women with FHBC, 3 studies used the same cohort; of the 32 studies looking at MHT/HC in women with FHBC, 2 pairs and 1 triplet of studies used the same cohorts; and of the 10 studies looking at alcohol in women with FHBC, 2 studies used the same study population.

Six studies investigated the effect of alcohol consumption on breast cancer risk in BRCA 1/2 mutation carriers (Fig. 2A), 10 alcohol consumption in women with FHBC (Fig. 2B), 9 smoking in BRCA 1/2 mutation carriers (Fig. 3A), 9 smoking in women with FHBC (Fig. 3B), 16 hormonal contraception or menopausal hormone therapy in BRCA 1/2 mutation carriers (Additional file 1: Figure S1A and B), 32 hormonal contraception or menopausal hormone therapy in women with FHBC (Additional file 1: Figure S1C), 9 BMI or weight in BRCA 1/2 mutation carriers (Additional file 2: Figure S2A), 12 BMI or weight in women with FHBC (Additional file 2: Figure S2B), 5 physical activity in BRCA 1/2 mutation carriers (Additional file 3: Figure S3A), and 16 physical activity in women with FHBC. Several of these studies included more than one modifiable risk factor.

Fig. 2
figure 2

Alcohol and breast cancer risk in women with BRCA mutations (n = 6) and family history (n = 9). Each bar is divided based on the proportion of included studies that demonstrated an increased risk, decreased risk, or no association with risk of breast cancer due to the specified alcohol exposure. Within each alcohol exposure, each study is represented only once. However, because the category “all alcohol behaviors” combines the results of all other exposure categories, studies may be represented more than once, if the results differ by exposure (e.g., decreased risk with current and no association with total/lifetime alcohol consumption). Numbers on the “all alcohol behaviors” bars indicate the range of risk estimates from studies when reported as a ratio measure (OR/RR/HR). Results from studies reporting only p values or other measures that did not indicate magnitude of effect are not included in these ranges. A Each bar in the figure represents all of the included studies (n = total number of studies) that reported results on the specified alcohol exposure in women with BRCA mutations, separated by BRCA mutation, if provided. Of the studies assessing alcohol type, one looked at wine drinking and the other did not specify the type of alcohol assessed. Please see Tables 1 and 2 for all studies cited. B Each bar in the figure represents all of the included studies (n = total number of studies) that reported results on the specified alcohol exposure in women with family history. Please see Additional file 4: Tables S1 and Additional file 5: Table S2 for all studies cited

Full size image
Fig. 3
figure 3

Smoking and breast cancer risk in women with BRCA mutations (n = 9) and family history (n = 9). Each bar is divided based on the proportion of included studies that demonstrated an increased risk, decreased risk, or no association with risk of breast cancer due to the specified smoking exposure. Within each smoking exposure, each study is represented only once. However, because the category “all smoking behaviors” combines the results of all other exposure categories, studies may be represented more than once, if the results differ by exposure (e.g., increased risk with current smoking and no association with past smoking). Numbers on the “all smoking behaviors” bars indicate the range of risk estimates from studies when reported as a ratio measure (OR/RR/HR). Results from studies reporting only p values or other measures that did not indicate magnitude of effect are not included in these ranges. A Each bar in the figure represents all of the included studies (n = total number of studies) that reported results on the specified smoking exposure in women with BRCA mutations, separated by BRCA mutation, if provided. Please see Tables 1 and 2 for all studies cited. B Each bar in the figure represents all of the included studies (n = total number of studies) that reported results on the specified smoking exposure in women with family history. Please see Additional file 4: Table S1 and Additional file 5: Table S2 for all studies cited

Full size image

Alcohol consumption in women with BRCA 1/2 mutations

Figure 2A includes data on the association between alcohol consumption and breast cancer (BC) in women with BRCA mutations. The data are divided into categories for ever, current, lifetime, age at first alcohol use, current alcohol/day, past alcohol/day, alcohol type, and a combined measure based on these seven exposure categories. Drinks/week was investigated in one study and found to be associated with a decreased risk of BC in women with BRCA1 mutations and no association with BC risk in women with BRCA2 mutations [6]. Of the six studies that looked at alcohol and BC risk, two found that a few of the alcohol exposures investigated decreased risk and the remaining studies found no association with BC risk [6,7,8,9,10,11].

Although the alcohol exposures investigated by each study were organized into a few categories, within those categories there were still multiple definitions of these exposures, as can be seen in Table 3. Ever and current alcohol consumption was defined consistently across studies; however, the other exposures varied.

Table 3 Variations in the definitions of alcohol exposures
Full size table

Alcohol consumption in women with family history of breast cancer

Figure 2B represents data on the impact of alcohol consumption on the development of BC in women with FHBC, divided into categories for past alcohol/day, current alcohol/day, cumulative alcohol, and all alcohol behaviors, a combined measure based on the other three categories. Additional alcohol exposures that appeared in only one article and thus are not presented in Fig. 2B include duration of alcohol use, age at first alcohol consumption, and current alcohol use (with no indication of quantity per day), which all had no association with BC risk as well as current or ever binging, blacking out, or drinking to the point of hurting one’s health, which all increased risk of BC [12, 13]. Overall, all studies demonstrated that alcohol consumption increased or had no association with BC risk (see Additional file 4: Table S1 and Additional file 5: Table S2 and Fig. 2B) [14,15,16,17,18,19,20,21].

Although the alcohol exposures investigated by each study were organized into a few categories, within those categories the definitions of these exposures still vary significantly as illustrated in Table 4. Table 4 demonstrates the various ways the particular alcohol exposure was defined by the studies included.

Table 4 Variations in the definitions of alcohol exposures
Full size table

Cigarette smoking in women with BRCA 1/2 mutations

Figure 3A depicts data on the association of cigarette smoking with BC risk in women with BRCA mutations. Data are separated into exposure categories: ever, current, past, age at smoking initiation, pack-years, and packs/week. Additional exposures that appeared in single studies included total years of smoking, which was associated with increased risk of BC and time since last cigarette use, which found no association with BC [11, 22]. Overall, two studies reported that a few of the smoking exposures were associated with decreased risk of BC and the remaining studies had no association with BC risk [9,10,11,12, 23,24,25,26,27].

Analogous to alcohol exposures, smoking exposure definitions also varied across studies as demonstrated in Table 5. The definitions of ever, current, and past were largely consistent across studies, and the remaining exposure categories varied widely.

Table 5 Variations in the definitions of smoking exposures
Full size table

Cigarette smoking in women with family history of breast cancer

Figure 3B represents data on the association between cigarette smoking and the development of BC in women with FHBC, divided into ever, current, past, age at initiation of, and duration of smoking as well as a combined measure based on the other four categories. One study additionally looked at cigarettes/day and age at first cigarette use and found no association for either with BC risk [28]. All studies reported that smoking increased or had no association with BC risk (see Additional file 4: Tables S1 and Additional file 5: Table S2 and Fig. 3B) [19, 20, 28,29,30,31,32,33,34].

Similar to alcohol exposures, although study findings were organized into categories of smoking exposure, the definitions of some of these exposures remain varied, as shown in Table 6. Ever, current, and past smoking were defined similarly across studies with no specification beyond those labels, whereas duration and age at smoking initiation varied.

Table 6 Variations in the definitions of smoking exposures
Full size table

Menopausal hormone therapy, hormonal contraception, weight, and physical activity

Tables 3, 4, 5, and 6 illustrate the myriad ways that exposure to a modifiable factor was defined in the included studies. Although not featured in a table format, this variability was only greater in the remaining modifiable risk factors (MHT/HC, BMI/weight, and physical activity).

Menopausal hormone therapy and hormonal contraception in women with BRCA 1/2 mutations

Data on the association of MHT/HC with BC risk in women with BRCA mutations are included in Additional file 1: Figures S1A, B. Most studies indicated no association with BC, a few indicated an increased risk for MHT/HC use in women with BRCA1, and a little less than one-third indicated a decreased risk of BC.

Menopausal hormone therapy and hormonal contraception in women with family history of breast cancer

Data on the association of MHT/HC with BC risk in women with FHBC are included in Additional file 1: Figure S1C. Most studies indicated no association with BC, a few indicated an increased risk, and one indicated a decreased risk of BC.

BMI/weight in women with BRCA 1/2 mutations

Data on the association of BMI/weight with BC risk in women with BRCA mutations are included in Additional file 2: Figure S2A. Most studies that looked at the relationship between elevated BMI/weight and pre- or post- menopausal BC showed no association, a few studies reported increased risk, and a few decreased risk.

BMI/weight in women with family history of breast cancer

Data on the association of BMI/weight with BC risk in women with FHBC are included in Additional file 2: Figure S2B. Most studies that looked at the relationship between elevated BMI/weight and pre-menopausal BC showed no association, one reported increased risk and one decreased risk. A little over half of the studies that looked at postmenopausal BC showed an increased risk, two studies showed decreased risk, and the remaining studies showed no association.

Physical activity in women with BRCA 1/2 mutations

Data on the association of physical activity with BC risk in women with BRCA mutations are included in Additional file 3: Figure S3A. All studies that investigated the association between physical activity and BC showed that increased physical activity decreased or had no association with the risk of BC.

Physical Activity in women with family history of breast cancer

Data on the association of physical activity with BC risk in women with FHBC are included in Additional file 3: Figure S3B. All studies that investigated the association between physical activity and BC showed that increased physical activity decreased or had no association with the risk of BC.

Other non-modifiable risk factors

Only one eligible study included participants with a familial cancer syndrome (Li Fraumeni). The study explored the effect of hormonal contraception on breast cancer risk and found that the use of oral contraception had no impact on risk, however, increasing the duration of use increased risk.

Quality

The quality of the studies included in this review was analyzed using the GRADE approach, which includes an assessment of risk of bias, inconsistency, indirectness, imprecision, and publication bias. We also assessed the racial/ethnic distribution of the study populations used in included studies to assess the generalizability of findings (Table 7).

Table 7 Racial/ethnic distribution of study populations used in included studies
Full size table

Risk of bias

Risk of bias in the included studies was assessed based on the consistency of the measurement of exposure and outcome, the adequacy of accounting for confounding variables, and the sufficiency of follow-up.

Exposure was measured very inconsistently across studies of the same modifiable risk factors. Studies differed in whether the exposure was measured in the past, present, or specific periods of life, or based on duration as well as the level of exposure investigated within these time frames. A few studies did not specify exposure, in which case the exposure was classified as ever use. Tables 3, 4, 5, and 6 illustrate the heterogeneity of exposure definitions for smoking and alcohol. Despite the variability demonstrated, smoking and alcohol exposures are defined more consistently across studies than MHT/HC, BMI/weight, or physical activity exposures. Not only were studies of these modifiable risk factors more variable in the way exposures were defined but also more likely to undertake further subgroup divisions. For example, studies of BMI/weight and HC/MHT commonly, but not universally, assessed the impact of menopausal status or utilized postmenopausal (POM) or pre-menopausal (PRM) participants only.

There was less variation in the way studies measured and classified non-modifiable risk factors. Four studies including women with BRCA mutations included women with BRCA1 mutations only. The remaining studies included women with BRCA1 or BRCA2 mutations and studied them combined or separately. Most studies of women with FHBC included women with a 1st-degree relative with BC. Nine studies included women with 1st- or 2nd-degree relatives with BC, one study included only women with a 2nd-degree relative, and six studies did not specify the degree of relatedness of the relative.

The outcome was reported largely consistently across studies as breast cancer. Though many studies did not specify the type of breast cancer (DCIS vs. LCIS vs. invasive) with which their participants were diagnosed, those studies that did, most commonly included participants with invasive breast carcinoma (IBC) and occasionally additionally included DCIS. Two studies used only participants with DCIS. Breast cancer status was collected by self- or proxy-report, physician report, or health records.

Most of the studies included adjusted for at least some of the other known risk factors such as age, age at menarche, menopausal status, age at first birth, parity, as well as those factors included in this review. In those studies that reported follow-up, it ranged from 4 to 26 years with an average of 9.9 years and a median of 7.8 years.

Imprecision

There was significant heterogeneity in exposures measured as discussed above; thus, it would be very difficult to compare confidence intervals directly or to perform a statistical test of heterogeneity.

Indirectness

Most of the studies included directly assessed the question of how modifiable risk factors affect women with non-modifiable risk of BC in the main analysis or as a subgroup analysis. However, small sample sizes in many of the articles likely reduce the generalizability of their findings. Sample sizes in the included studies ranged from 68 to 4,984 women with BRCA mutations and 24 to 43,713 women with FHBC, including both cases and controls. However, when studies investigated specific exposures, such as glasses/day of alcohol, sample sizes could fall as low as 4 participants.

Publication bias

Although it would not be possible to assess publication bias using a funnel plot, it is generally assumed that unpublished studies are mostly small null studies.

Race/ethnicity data

Studies varied in reporting of the information about the racial/ethnic distribution of the study populations used. While some detailed the demographics of participants by race (such as White, Black, etc.), others reported on the numbers of French-Canadian and Jewish participants. There were also a significant number of studies that did not provide any information on race or ethnicity for the study populations used. The studies included in this review represent a wide number of countries with several using international cohorts, though most were conducted in Western countries such as the USA, Canada, Europe, and Australia. Additionally, a few studies represented Asian countries including Japan and Korea. With the lack of detailed and consistent reporting of racial and ethnic demographics of study populations used, it is difficult to assess the generalizability of findings.

Discussion

All women face some risk of breast cancer and modifiable risk factors have been demonstrated to be an important component of that risk [2]. Some risk factors such as low physical activity and alcohol use have shown a consistent pattern of increasing breast cancer risk [35]. Though trends in breast cancer risk related to other risk factors such as hormonal contraception use and body mass index have been linked to breast cancer risk, the relationship differs based on menopausal status and age. Given the smaller numbers of women with BRCA mutations or family history, necessitating further subdivision of subjects by age or menopausal status makes the elucidation of these trends more challenging.

Despite robust research on modifiable risk factors for breast cancer, it has remained unclear how, if at all, modifiable risk factors affect breast cancer risk in women with underlying high risk due to inherited non-modifiable risk factors. Particularly as women become more aware of both their family histories and commonly tested genetic polymorphisms, such as in BRCA1/2, information on modifiable risk factors for high-risk women will become even more critical. It is also important to consider how these modifiable factors affect the risk of other cancers that patients with BRCA1/2 mutations are at increased risk of, such as ovarian cancer [36].

Although this review included ten or more studies with women with positive family history for every investigated modifiable risk factor, fewer were available for most modifiable risk factors in BRCA1/2 mutation carriers, and only one study investigated women with a familial cancer syndrome, despite the search including several familial cancer syndromes. Furthermore, the studies in this review present risk estimates based on disparate exposures to modifiable risk factors and often suffer from low sample sizes and low quality.

The inconsistency in the classification and measurement of exposures to modifiable risk factors limited the direct comparison of studies and made it problematic to perform a meta-analysis as well as to draw conclusions on the high-priority modifiable risk factors for women at high risk of BC. Summarizing across all exposures for each modifiable risk factor, the data support that women at high risk of BC are similarly or less strongly affected by the commonly accepted modifiable risk factors for breast cancer. In studies on women with positive family history, very few studies indicated that physical activity or elevated pre-menopausal BMI/weight increased risk or that HC/MHT, smoking, postmenopausal elevated BMI/weight or alcohol use decreased risk, rather they were null or associated with increased risk. Studies of women with BRCA1/2 mutations demonstrated similar though more varied trends with smaller numbers of studies.

The discrepancies in modifiable risk factor exposure also increased risk of bias, decreasing quality. Quality was adequate in other areas such as indirectness and participant classification; however, imprecision could not be assessed due to heterogeneity. The ability to compare studies would improve if the same definitions of exposures were utilized. Alternatively, it could also be worthwhile to obtain the original participant data and to apply consistent definitions retrospectively to accurately compare cohorts.

Conclusion

With growing awareness of family history and utilization of direct-to-consumer genetic testing, it will become easier to conduct higher-powered studies on this subject. In order for future studies to be more meaningful to women considering lifestyle changes, standardization of the reporting of exposures would be beneficial. Although this review suggests that women with high risk could benefit from following the same guidelines on modifiable risk used for women without underlying risk, such as maintaining an active lifestyle with low levels of tobacco and alcohol use, due to the limitations and heterogeneity of existing literature, women at high risk of breast cancer would benefit from further investigations into these and other modifiable risk factors.