Association of a Healthy Lifestyle Index with Risk of Breast Cancer among Women with Normal Body Mass Index in the UK Biobank

Breast cancer is the most common cancer among women worldwide, accounting each year for 25% to 30% of all cancer diagnoses and for approximately 15% of cancer-related deaths in this group (1, 2). Breast cancer has a multifactorial etiology, due both to modifiable and nonmodifiable risk factors (3).

The World Cancer Research Fund/American Institute of Cancer Research (WCRF/AICR; ref. 4) has indicated the need to lower excessive weight (5), and in addition, to minimize exposure to lifestyle factors associated with increased cancer risk such as alcohol consumption, tobacco smoking, poor diet (high in fat and sugar, and low in vegetable, fruit and whole-grain consumption), and physical inactivity. Given the modifiable nature of these factors, and hence their potential role in cancer prevention, their association with breast cancer has been evaluated independently (6–9) and in combination, the latter by creating a lifestyle index score based on body mass index (BMI), smoking, alcohol intake, diet quality, and physical activity. An inverse association between a healthy lifestyle index (HLI) and risk of breast cancer has been found in several epidemiologic studies of postmenopausal women, from different ethnic and cultural backgrounds (10–21). The results from two studies on premenopausal women were less consistent, with no association found in one study, although the results were based on a limited number of cases (22), and a significant reduction of the risk of breast cancer associated with a higher HLI that did not include BMI in a second one (17).

Recent reports have shown that normal BMI (18.5–<25 kg/m²) postmenopausal women with high levels of body fat are at increased risk of breast cancer (22–24), and both inflammation and altered metabolism have been suggested as potential underlying mechanisms to explain the increased risk of breast cancer in this group (25, 26). Little is known, however, about the potential to reduce the risk of breast cancer in postmenopausal and premenopausal women with a normal BMI. To this end, in the present study, we evaluated the association between HLI and the risk of breast cancer among women participating in UK Biobank with a BMI within the normal range.

UK Biobank is a prospective cohort study including approximately 500,000 individuals (54.4% women) aged 40 to 69 years at recruitment, living in England, Wales, and Scotland, and registered with the United Kingdom's National Health Service (NHS; ref. 27). Participants were invited to attend one of the 22 study centers between 2006 and 2010. A detailed protocol of the UK Biobank study is available online (https://www.ukbiobank.ac.uk/media/gnkeyh2q/study-rationale.pdf). The UK Biobank received ethical approval from the North West Multi-centre Research Ethics Committee, the National Information Governance Board of Health and Social Care in England and Wales, and the Community Health Index Advisory in Scotland (http://www.ukbiobank.ac.uk/ethics/). All participants provided written informed consent.

At the baseline visit, study participants completed self-administrated touchscreen questionnaires on sociodemographic, health, reproductive, medical, and lifestyle factors (http://www.ukbiobank.ac.uk/resources/). Anthropometric measurements were obtained by trained personnel following standardized procedures (24). BMI was calculated by dividing weight (kg) by the square of height (m²). Waist circumference (cm) was measured at the narrowest part of the torso using a nonstretchable spring tape.

Dietary assessment was based on 29 questions regarding the average frequency of consumption of foods and food groups per week over the past year. For certain items such as bread type and cereals, portions expressed as quantity were indicated, while for others such as meat type, frequencies were specified. The validity of this tool to obtain reliable information on participants' diet habits has been previously evaluated by comparing it with an extensive 24-hour dietary recall of 206 types of food and 32 types of drinks in a subgroup of this cohort. The results showed that the questionnaire reliably categorized participants with respect to their main food group intakes (28).

Alcohol consumption was expressed as the number of drinks per week, while self-reported smoking status was recorded as never, former, and current. Total physical activity, which included various activities over the last 4 weeks was computed by multiplying the hours/week of each reported physical activity by the metabolic equivalent (MET) of that activity, and summing the values across activities (29).

The combined HLI score used in this study has been described in detail previously (14). It consists of a modified score based on the recommendations of the WCRF/AICR (4), and combines diet (including red and processed meat, vegetable, fruit, and whole-grain intake), alcohol consumption, physical activity, and waist circumference. Smoking status is also included in the HLI score, although it was not included in the WCRF/AICR recommendations. Each individual component was indexed separately in three categories, with the highest score representing the healthiest group: never smoked, no alcohol consumption, relatively high physical activity level [≥ 3,000 MET-minutes per week (min/wk) or ≥ 750 min/wk] (29), waist circumference <80 cm, lowest intake of meat, and highest intake of fruits, vegetables, and whole grains. The final score ranged from 0 to 5.5 (Supplementary Table S1; ref. 14).

Incident invasive breast cancer cases were ascertained through the Health & Social Care Information Centre for women resident in England and Wales, and the NHS in Scotland. Code C50 was used to identify cases of invasive breast cancer based on the International Classification of Diseases, Tenth Revision (ICD-10; ref. 30). Breast cancer cases were further classified according to the ICD-0–3 morphologic code (31) as ductal (8500–08, 8022, 8035), lobular (8520-9), or other types. Data for HER2, estrogen receptor and progesterone receptor status were not available.

This study included women with normal BMI (18.5–<25.0 kg/m²; n = 105,680), who were breast cancer–free at baseline (excluded prevalent cases, n = 3,100) and had information on follow-up time (missing, n = 8), leaving a cohort of 102,572 women (33,917 premenopausal and 68,655 postmenopausal). Women were considered to be postmenopausal if they had had a natural menopause, had a bilateral oophorectomy, or had missing information on menopausal status and were at least 53 years old (32). The number of participants with data on HLI and each individual lifestyle component are reported in Supplementary Table S1.

All analyses were performed separately in pre- and postmenopausal women. Baseline characteristics of the cases and noncases were compared using Wilcoxon rank-sum tests for continuous variables and χ² tests for categorical variables. To evaluate the association between the HLI and breast cancer incidence, tertiles of the HLI were created separately by menopausal status based on the distribution of the score among the respective noncases; breast cancer incidence rates (per 1,000 person-years) were calculated for each tertile. Cox proportional hazards models were used to estimate HRs and 95% confidence intervals (CI) for the association of the HLI with risk of breast cancer, using the lowest tertile as the referent category. Follow-up time was used as the underlying time at risk and was measured from the date of enrollment to the date of breast cancer occurrence, or the date of death, withdrawal from the study, loss to follow-up, or to the date of the end of follow-up (October 31, 2015 for Scotland and March 31, 2016 for England and Wales), whichever came first. No violation of the proportional hazards assumption was found on the basis of examination of Schoenfeld residuals. Selection of the variables included in the final model was based on their association with breast cancer and whether these factors acted as confounders by altering the estimates of association for the main exposure by more than 10%. The analyses were adjusted for age at enrollment (age-adjusted model), and for socio-economic status (Townsend deprivation index; ref. 33), race (White, non-White), height (quartiles), family history of breast cancer (first-degree relatives), use of hormone replacement therapy (never, past, current), use of oral contraceptives (never, past, current), number of live births (0, 1, 2–4, >4), history of mammogram screening (yes, no), and age at menopause (in postmenopausal women; multivariable adjusted model). Missing values of categorical variables were retained in the analysis by creating a separate category. To examine the possibility of reverse causality, we repeated the analyses after excluding breast cancer cases that occurred during the first 2 years of follow-up and noncases who were lost to follow-up during this period. We also tested whether any individual index component acted as a determinant of the association of HLI with risk of breast cancer by removing one component at a time from the composite score and adjusting for it in the model. For these new reduced HLI scores, tertiles were created based on their distributions in the noncases. In addition, we also evaluated the associations of the individual HLI components by testing their association with breast cancer risk while adjusting for all of the other components. Women with missing values for any of the HLI components were excluded from the main analysis (81,230 women had complete data on all HLI components), but they were retained in the analyses of individual components for which they did not have missing values.

The analyses were conducted using STATA version 16.1 (Stata Corp LP). All P values were two-sided.

A total of 1,796 incident cases of breast cancer (560 premenopausal, 1,236 postmenopausal) were diagnosed in the study population over an average follow-up period of 7.0 (SD = 1.1) years. Table 1 summarizes the baseline characteristics of the cohort by menopausal status. Among premenopausal women, compared with noncases, breast cancer cases were older at recruitment and at the time of their first-born child, had higher frequency of a family history of breast cancer, were more likely to be current users of oral contraceptives, and had slightly higher BMIs. Among postmenopausal women, compared with noncases, breast cancer cases were also older at enrollment, more likely to be nulliparous, more likely to be older at the time of their first born, had a higher frequency of a family history of breast cancer, were more likely to use hormone replacement therapy, and had higher BMIs. In addition, postmenopausal breast cancer cases were more likely to have reported a poor diet and a lower level of physical activity, and to have a larger waist circumference at enrollment into the study.

Among premenopausal women, there was no association between the HLI and risk of breast cancer (Fig. 1), while in postmenopausal women, a relatively high HLI was associated with a reduced risk of breast cancer overall (adjusted HR_{3rd tertile} = 0.76; 95% CI, 0.64–0.91), and by histologic subtypes (ductal: adjusted HR_{3rd tertile} = 0.78; 95% CI, 0.64–0.94; and lobular: adjusted HR_{3rd tertile} = 0.71; 95% CI, 0.51–0.99). Analyses excluding women with less than 2 years of follow-up (leaving 1,217 cases and 100,474 noncases) showed results similar to those obtained in the main analyses (Supplementary Table S2). In premenopausal women, exclusion of individual components of the HLI one factor at a time showed that a relatively high HLI without the smoking score was associated with reduced risk of breast cancer (adjusted HR_{3rd tertile} = 0.71; 95% CI, 0.56–0.90), while the exclusion of other factors produced results similar to those obtained with the full HLI score (Fig. 2A). Among postmenopausal women, higher HLI remained associated with a reduced risk of breast cancer independently of the component excluded (Fig. 2B).

We also evaluated the individual components of the HLI score for their associations with breast cancer risk (Table 2). In premenopausal women, higher scores, indicating healthier behaviors, showed statistically nonsignificant reductions in risk of breast cancer for all of the factors except for smoking, for which there was a statistically nonsignificant increase in risk. Among postmenopausal women, higher scores for each HLI component were associated with reduced risk of breast cancer; however, the results were significant only for waist circumference (HR_{<80 cm} = 0.64; 95% CI, 0.48–0.86) and physical activity (for MET_{≥3,000 (min/wk)}: HR = 0.76; 95% CI 0.63–0.92).

In this large prospective study of women with normal BMI, we found that a high HLI score, reflective of a relatively low waist circumference, a diet low in red and processed meat, and high in vegetable, fruit, and fiber, no alcohol consumption, no cigarette smoking, and moderate to high levels of physical activity, was associated with a reduced risk of breast cancer among postmenopausal women. The results were not specifically linked to any particular component of the index, as indicated when individual factors were excluded in turn from the total score. Among premenopausal women, the association between the HLI and risk of breast cancer showed a similar trend, although it did not reach statistical significance. The latter results may be due in part to menopause-related hormonal and physiologic differences (34) or potentially to the smaller sample size for this group compared with the postmenopausal group. Of note, however, when smoking (which showed an inverse association with breast cancer risk) was excluded from the composite score, a relatively high HLI was associated with a significant risk reduction in premenopausal women.

Previous case-control and cohort studies have explored the association of the HLI with risk of breast cancer in postmenopausal women regardless of BMI, consistently showing that higher scores were associated with a reduced risk of breast cancer (10–21). While previous data on premenopausal women are limited, they have also shown an inverse association between HLI and breast cancer risk regardless of BMI (14). No data have been previously published on HLI and breast cancer risk by histologic subtypes.

Obesity is associated with an increased risk of cancer for at least 13 anatomical sites including the breast in postmenopausal women (35), as well as for chronic conditions such as type-2 diabetes, hypertension, cardiovascular disease, and kidney disease (36–38). Public health recommendations have often advised that individuals reduce their weight if overweight/obese or maintain a healthy weight to lower their risk of such adverse health outcomes. This is the first study that has evaluated the HLI association with breast cancer risk among women with normal BMI, a group usually considered to be at lower breast cancer risk after the menopause. We have shown here that among postmenopausal women, a high HLI score, with or without the inclusion of excessive abdominal adiposity measured as waist circumference, is associated with reduced risk of breast cancer in women with a normal BMI. This result suggests that following a healthy lifestyle with moderate to high levels of physical activity, a diet with high content of fruit, vegetable, and whole grains, and low in red and processed meat, with no alcohol and no smoking, might provide benefit against breast cancer risk beyond that obtained by weight control and a low level of abdominal fat. Importantly, although premenopausal women with a normal BMI have been shown to be at increased risk of breast cancer (39), we found that with the exception of smoking, adopting a combination of healthy lifestyle habits was associated with a reduced risk of breast cancer at this stage of life, although no individual modifiable behavior was independently associated with a lower risk.

Our study was conducted in a large population with detailed information on health and reproductive characteristics collected at the time of enrollment, when the participants were breast cancer–free. All the analyses were adjusted by several important risk factors and potential confounders. Although the information used to calculate the score was based on a single point in time, previous analysis has demonstrated the reproducibility of the study variables (28); nevertheless, changes in healthy behaviors and BMI may have occurred after the baseline measurements. The HLI used in the study was based on five of the eight components recommended by the WCRF/AICR guideline (4), given that some of the dietary information such as polyunsaturated, saturated, and trans fats, and glycemic load, was not available; therefore, we could not estimate the full impact of these factors. Following the guideline of the Cancer Research UK, additional modifiable risk factors more specific to breast cancer risk, such as oral contraceptive and hormone replacement therapy for postmenopausal women, were included in the HLI score in one study (40). In the present study we have decided to include these factors in the final model as confounders to remove their effects in other to better compare the results obtained using the WCRF/AICR guideline with those present in the literature. Future analyses should evaluate the association of a HLI score with the inclusion of other modifiable factors (e.g., oral contraceptive and hormone replacement therapy use). Other studies on premenopausal women have reversed the score for waist circumference, assigning the highest score to the group with the highest waist circumference values (14). In this group of premenopausal women, we did not find a reverse association between waist circumference and risk of breast cancer, therefore, following current public health recommendations, we did not reverse the score for waist circumference (4). Estimates of total energy intake were not available in the present study and therefore could not be taken into account as a potential confounding factor. We did not have information on breastfeeding history, which is included in the HLI guidelines of modifiable factors for the prevention of breast cancer. However, given the age of the women in the study, it is unlikely that this exposure can be considered modifiable. We also did not have access to information on the hormone receptor status of the breast cancer cases and therefore we were unable to address potential heterogeneity in the association by tumor subtype. In addition, given the average age of the premenopausal group, it is possible that some of the women had changed their menopausal status by the date that the breast cancer was diagnosed. Finally, the UK Biobank study includes volunteer participants who are, on average, a healthier sample compared with the UK population at large (41).

Overall, the results of this study show that healthy lifestyle behaviors may significantly reduce the risk of breast cancer among normal weight postmenopausal women, and possibly among normal weight premenopausal women as well. Recommendations to follow a healthy lifestyle, even for normal weight women, may contribute to a reduction in breast cancer risk.

No disclosures were reported.

R. Peila: Data curation, formal analysis, methodology, writing–original draft, writing–review and editing. R.S. Arthur: Data curation, writing–review and editing. A.J. Dannenberg: Writing–review and editing. T.E. Rohan: Conceptualization, resources, funding acquisition, writing–review and editing.

T.E. Rohan is supported in part by the Breast Cancer Research Foundation (BCRF-21-140).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.