Association of lifestyle factors with breast cancer incidence: An overview of systematic reviews

Abstract

Objective

Heterogeneity in the evidence of association between lifestyle factors and breast cancer (BC) incidence hampers initiatives to modify BC risk. This overview aims to synthesise evidence from systematic reviews (SRs) to inform lifestyle-related modifications for BC prevention.

Methods

We systematically searched (MEDLINE, EMBASE, and CINAHL) from January 2013 to August 2023 for SRs of the association between lifestyle factors [alcohol consumption, physical activity (PA), body mass index (BMI), smoking, breastfeeding, oral contraception (OC), hormone replacement therapy (HRT), and sedentary behavior (SB)] and BC incidence. A narrative data synthesis was performed.

Results

Sixty-six SRs met the eligibility criteria. Evidence from 40 SRs indicated consistent associations between the risk of BC and postmenopausal BMI increase (relative risk increase: 2%−21%), use of HRT (risk increase: 23%−33%), smoking (risk increase: 4%−86%), and alcohol consumption (risk increase: 4%−61%). Additionally, evidence from 23 SRs suggested protective associations with PA (risk decrease: 10%−39%), breastfeeding (risk decrease: 9%−53%), and healthy lifestyle scores (protective about 20%−26%). However, inconsistent and/or statistically non-significant associations were found between BC incidence and premenopausal BMI increase [relative risk (RR): 0.78−1.08], SB (RR: 1.01−1.20), and OC use [odds ratio (OR): 1.01−1.35].

Conclusions

This overview identifies lifestyle factors associated with BC incidence, highlighting both harmful and protective factors. Our summary findings can support information and interventions related to modifying these factors, including limiting alcohol and smoking, or avoiding postmenopausal BMI increase and HRT.

Introduction

Breast cancer (BC) is one of the most frequently diagnosed cancers worldwide, accounting for 1 in 4 cancers occurring in women (1). In 2020, there were approximately 2.3 million new cases of BC, accounting for 11.7% of all cancer diagnoses. BC is the second most common cause of cancer-related deaths among women (2). With population aging and changes in lifestyle risk factors, the incidence of BC is projected to increase dramatically by 2040, reaching over 3 million new cases and causing 1 million deaths annually (3).

Lifestyle risk factors contribute significantly to cancer burden, accounting for 42% of all cases (4). Several lifestyle factors, such as overweight/obesity, poor diet, excess alcohol and tobacco consumption, and sedentary behavior (SB), have been associated with BC risk (5). Breastfeeding, use of hormone replacement therapy (HRT), and use of oral contraceptive (OC) are also modifiable lifestyle risk factors for BC (6-8). However, inconsistencies exist regarding the association between some lifestyle factors and BC incidence, which warrants further study (9). BC is a heterogeneous disease with molecular subtypes having distinct biological and clinical features that may modify these associations. For instance, luminal A/B subtypes are hormone receptor-positive tumors and strongly influenced by hormonal factors such as reproductive factors. However, triple-negative BC is often interlinked with genetic predispositions such as BRCA mutations and may be affected by metabolic or inflammatory factors (10,11). Furthermore, the combination of these factors as a “lifestyle pattern” may influence BC risk more than each isolated factor.

Heterogeneity in the strength and direction of the association between lifestyle factors and BC incidence in primary studies, systematic reviews (SRs) and meta-analyses (8,9) remains a challenge for clinicians, consumers and other stakeholders in the intervention and prevention of BC. Therefore, we undertook a comprehensive overview of SRs on the association between lifestyle factors and BC incidence to synthesise evidence from SRs and to inform future policy and research relating to lifestyle modifications and BC prevention.

Materials and methods

Protocol registration

A protocol for this overview detailing predefined eligibility criteria, search strategy, quality assessment, data extraction process, data synthesis and reporting was registered with the International Prospective Register of Systematic Review (PROSPERO-CRD42023451291).

Search strategy

A comprehensive search of MEDLINE, EMBASE and CINAHL from 1 January 2013 to 17 August 2023 was conducted. We focussed the search on this ten-year period to capture contemporary SRs of exposure. Keywords and index terms included breast neoplasms combined with lifestyle factors such as health behaviour, alcohol consumption, physical activity (PA), body mass index (BMI), smoking, breastfeeding, diet and nutrition, HRT, and OC. We used the MEDLINE/EMBASE SR filters from the BMJ Toolkit (12), which were adapted for CINAHL. The full search terms and results are given in the Supplementary Tables S1−S5. The references of included studies were screened to identify potentially eligible SRs that may have been missed by our search.

Study selection

We adopted population, exposure, control, outcomes, study design (PECOS) (Supplementary Table S6) to formulate the overview research question and assess the eligibility of the SRs. English language SRs published in peer-reviewed journals that investigated the association between lifestyle factors (BMI, alcohol, PA, smoking, OC, HRT, breastfeeding, SB) and BC incidence in women with no personal history of BC or any other cancers were included. SRs of composite risk factors (e.g. healthy lifestyle indices) were also included. SRs of diet/nutrition were excluded but identified for inclusion in a future overview of this lifestyle factor. SRs not reported in full-text or restricted to high-risk gene mutations were excluded.

One author (SR) initially assessed eligibility by screening titles and abstracts against the inclusion and exclusion criteria, with potentially eligible citations retrieved for full-text assessment. Two independent reviewers (SR and JB) evaluated 20% of randomly selected SRs in both titles and abstracts and full-text screening to ensure consistency in eligibility criteria. Disagreements were resolved through discussions or arbitration by a third reviewer (MLM) when required.

Data extraction and synthesis

Two reviewers (SR and JB) independently extracted data from the selected SRs using a predefined form, including characteristics of SRs (author, year, number of studies, number of participants, participants’ age, region, settings, estimates, publication bias and heterogeneity), primary studies (including years and designs), exposures and outcome of interest (BC incidence). Disagreements were resolved through discussion or with the participation of a third reviewer (MLM). We used a narrative synthesis approach due to the diverse methodologies, exposures, and outcomes reported across SRs. Within each exposure, we tabulated data, displayed review-level estimates graphically as forest plot when appropriate, and qualitatively synthesized key findings.

Quality assessment

A comprehensive and validated tool for rigorous, transparent and standardized appraisal of SRs, a MeaSurement tool to assess systematic reviews (AMSTAR 2), was used to assess the quality of the included SRs (13). The first reviewer (SR) assessed all studies, and another reviewer (MLM) checked a random subset of 10% to ensure reliability, consistency and accuracy. We ranked confidence in the findings of SRs as critically low if there were more than one critical flaw; low if there was one critical flaw; moderate if there was more than one non-critical issue and no critical flaws; and high if there was at most one non-critical flaw and no weakness in any critical elements.

Results

Search results

The preferred reporting items for SRs and meta-analyses (PRISMA) flow chart of SRs selection with reasons for exclusion are provided in Figure 1. The search retrieved 1,578 non-duplicate citations, of which 66 met eligibility criteria.

Figure 1.

Preferred reporting items for systematic reviews and meta-analyses (PRISMA) flow chart of SRs selection. SR, systematic review.

The characteristics of the included SRs are presented in Table 1. The median number of studies and participants in each SR is 27 (range: 2−214), and 709,564 (range: 2,487−22,728,674), respectively. Included SRs incorporated geographically diverse populations, involving a median of 31,250 (range: 839−236,955) BC cases. SRs assessed the following lifestyle factors: BMI (N=23) (8,14-35), PA (N=15) (8,14,17,20,34,36-45), smoking (N=9) (8,14,33-35,41,46-48), alcohol consumption (N=9) (8,14,34,40,49-53), OC (N=11) (8,16,33,54-61), HRT (N=9) (8,39,41,61-66), breastfeeding (N=11) (8,16,31,41,61,67-72), SB (N=5) (68-72), and healthy lifestyle index (N=1) (73). The majority of SRs included mixed study designs (N=35), while fewer were limited to cohort (N=18) or case-control (N=8) designs.

Table 1. Characteristics of included SRs in this overview.

SR author (year)	Year published	Studies include (N)	Population size	Study designs	Study region	Exposure
SR, systematic review; HRT, hormone replacement therapy; BF, breastfeeding; BC, breast cancer; PC, prospective cohort study; RC, retrospective cohort study; PreM, premenopausal; PostM, postmenopausal; C, cohort study; CC, case-control study; MA, meta-analysis; ET, estrogen therapy; EPT, estrogen + progestin therapy; CT, clinical study; RCT, randomized control trial; ME, Middle East; BMI, body mass index; NCC, nested case-control study; OC, oral contraception; CS, cross-sectional study; PS, prospective study; ES, epidemiological study; USA, United State; LA, Latin America; EU, Europe; NA, North America; SA, South America; WC, waist circumference; WHR, waist-to-hip ratio; PA, physical activity; SB, sedentary behavior; EMR, Eastern Mediterranean Region.
BMI
Byun et al. 2022 (27)	Up to June 2020	21 for BC	1,849,875 (39,733 cases)	PS (C)	USA, EU, Asia, Australia	BMI (early life BMI)
Soltani et al. 2021 (24)	Inception of database-2020	3 (2 PC, 1 RC)	43,352 with T2D (839 cases)	C	USA, Sweden, China	BMI
Torres-de la Roche et al. 2020 (19)	2008−2018	21	55,580 cases	C, CC, CS	USA, EU, Asia	BMI
Hidayat et al. 2018 (28)	1989−2016	24	33,606 BC cases among over 2 million participants	C	NA, EU, Asia, Australia	BMI
Liu et al. 2018 (29)	2004−2014	12	22,728,674 (19,480 cases)	C	USA, EU, Asia, Australia	BMI
Chen et al. 2017 (25)	1997−2014	31	3,318,796 (42,271 cases)	C	EU, USA, Asia, Australia	BMI
Zahmatkesh et al. 2017 (32)	2004−2014	7	33,552 individuals	CC	Iran	BMI
Chen et al. 2016 (23)	1966−2016	18	953,245 (20,602 cases)	PS	USA, EU, Canada, Australia, China	Central obesity (WC or WHR)
De Ridder et al. 2016 (26)	1974−2015	2 BC	167,689 (7,672 cases)	C	NA, EU	BMI, abdominal adiposity
Wang et al. 2016 (15)	1985−2015	39 BC (61 datasets)	12,107,874 (41,406 cases)	C, CC	NA, EU, Asia, Australia	BMI
Keum et al. 2015 (22)	1997−2014	13 BC	709,564 (24,271 cases)	C	USA, Canada, Australia, EU, Japan	Adult weight gain and adiposity
Xia et al. 2014 (30)	1997−2014	32 (12 preM and 20 postM)	PreM: 426,199 (4,699 cases); PostM: 1,155,110 (22,809 cases)	C	NA, EU, Asia	BMI
Amadou et al. 2013 (18)	2000−2010	30 (11 C and 19 CC)	C: 1,489,288 (5,966 cases); CC: 14,940 cases and 23,247 controls	C, CC	Caucasian, African-American, Asian, African, Asian-American	BMI (WHR, height, WC)
Physical activity
Hidayat et al. 2020 (45)	1987−2017	54 BC	C: 1,090,823 (30,533 cases); CC: 93,506 cases, 106,015 controls	C,CC	NA, EU, Asia, Australia	PA
Chen et al. 2019 (44)	1994−2017	38	68,416 cases	C	NA, EU, Asia	PA
Paudel et al. 2019 (36)	2000−2018	2 BC	2,487 (1,399 cases, 1,088 controls)	Matched CC	South Asia (India)	PA
Neilson et al. 2017 (42)	1994−2015	80	1,580,487 participants	C, CC	NA, EU, Asia, Australia	PA
Liu et al. 2016 (43)	Up to 30 December 2014	33 BC	4,098,181 (38,700 cases)	C	USA, EU, Asia	Leisure time PA
Goncalves et al. 2014 (38)	2000−2010	21 (7 C and 14 CC)	810,750 samples (745,584 C, 65,166 CC)	C, CC	USA, EU, Asia	PA
Wu et al. 2013 (37)	1994−2012	31	2,463,599 (63,786 cases)	PS	USA, EU, Asia	PA
Alcohol consumption
Sun et al. 2020 (50)	1990−2018	22	45,350 cases	C	NA, Asia, EU	Alcohol
Choi et al. 2018 (51)	Up to April 2016	34 articles (27 studies)	3,626,225 samples	C	Asia, NA, EU	Alcohol
Jayasekara et al. 2016 (49)	1977−2013	16 (3 C, 13 CC)	35,898 cases	C, CC	NA, EU	Alcohol
Bagnardi et al. 2015 (52)	1956−2012	118 (43 C, 75 CC)	117,317 (48,433 cases)	C, CC	NA, EU, Asia, Australia	Alcohol
Bagnardi et al. 2013 (53)	Inception of databases to 2010	110 (39 C, 71 CC)	Cases in ref. cat: 41,995, cases in light drinking cat: 72,902	C, CC	NA, EU, Asia	Alcohol (light drinking)
Oral contraception
Fitzpatrick et al. 2023 (56)	1995−2022	12 previous MA	9,498 cases and 18,171 controls	OS	USA, EU, Asia, Africa	OC
Barańska et al. 2022 (60)	1999−2018	19	246,152 (31,250 cases and 214,902 controls)	CC	NA, EU, Asia, Australia	OC
Barańska et al. 2021 (54)	2009−2020	42	110,580 [cases: 30,778 (15,722 OC users), controls: 79,802 (38,334 OC users)]	CC	USA, Africa, ME, Asia, EU, multicentre.	OC
Kanadys et al. 2021 (58)	1960–2010	79	72,030 BC cases and 123,650 controls	CC	NA, EU, SA, Asia, Africa, Australia	OC
Samson et al. 2016 (59)	2000−2015	6	217,355 (12,189 cases)	ES	South Africa, USA, EU	OC
Gierisch et al. 2013 (57)	2000 onward	44 BC	38,682 cases, 317,341 control, 3,981,072 person-years	C, CC	NA, EU, Asia, Australia, Africa	OC
Zhu et al. 2013 (55)	1960−2012	13	859,894 (11,722 cases)	C	USA, EU, Asia	OC
HRT
Goldstajn et al. 2023 (64)	2003−2009	7 BC (6 C, 1 CC)	Not mentioned	C, CC	Not mentioned	HRT
Gartlehner et al. 2022 (63)	2016−2021	5 (ET only user), 6 EPT user)	>13,000 ET and >25,000 EPT	C, CT	Not mentioned	HRT
Kim et al. 2018 (62)	1994−2016	25 (23 C, 2 RCTs)	3,795,509 (69,597 cases)	C, RCT	USA, EU, Asia, Australia	HRT
Asi et al. 2016 (65)	2007−2013	3 (2 C and 1 CC)	86,881 postM women	C, CC	France	HRT (progesterone vs. synthetic progestins)
Bae et al. 2015 (66)	2001−2012	6 (1 C and 5 CC)	6,679 cases and 4,302 controls	C, CC	Korea	HRT
Smoking
He et al. 2022 (48)	1988−2022	77	2,326,987 participants	C, CC	NA, EU, Asia	Smoking
Macacu et al. 2015 (47)	Inception of database-2015	86	235,935 cases (26,170,011 person-years)	C, CC	NA, EU, Asia	Smoking
Chen et al. 2014 (46)	1992−2013	51 (3 C and 48 CC)	17,369 cases and 92,967 controls	C, CC	China	Smoking
BF
Unar-Munguía et al. 2017 (67)	2005−2015	65	Not mentioned	C, CC	USA, LA, EU, Asia, Africa	BF
Chowdhury et al. 2015 (69)	Not mentioned	100	Not mentioned	OS, RCT	Not mentioned	BF
Islami et al. 2015 (68)	C: 2007−2014 CC:1983−2014	27 (8 C and 19 CC)	36,881 cases and 31,304 controls	C, CC	NA, Asia, Australia, EU	BF
Zhou et al. 2015 (71)	2008−2014	24 (27 studies)	13,907 cases	C, CC, CS	Asia, EU, Africa	BF
Sedentary behavior
Chong et al. 2021 (73)	2001−2019	8	426,506 (17,048 cases)	C	USA, Japan, China, Netherland, Canada	SB
Lee et al. 2021 (74)	Inception of database-2020	31 (13 C and 18 CC)	3,382,206 (110,249 cases)	C, CC	USA, EU, Japan, China, Malaysia, Canada	SB (Sedentary work)
Zhou et al. 2015 (75)	1997−2014	21 (7 C, 14 CC studies)	2,625,772 (82,630 cases)	C, CC	NA, EU, Asia	SB
Schmid et al. 2014 (76)	Inception of databases-2014	12	326,738 (18,162 cases)	OS	USA, EU, Asia	TV viewing or other SB
Shen et al. 2014 (77)	2010−2013	3 BC	172,818 (4,699 cases)	C	USA, China	SB
Healthy lifestyle score
Armenta-Guirado et al. 2023 (78)	2011−2021	31 (29 in MA)	2,064,214 (77,470 cases)	C, CC	NA, EU, Iran, New Zealand	Healthy lifestyle indices
Multiple individual factors
Khoramdad et al. 2022 (33)	2008−2020	24	12,460 (5,675 cases and 6,785 controls)	Matched CC	Iran	BMI, smoking, OC
Manouchehri et al. 2022 (70)	2002−2019	27 (8 studies on BF)	BF: 4,209 (1,694 cases and 2,515 controls)	CC	Iran	BF, menopausal status (PostM vs. PreM)
Poorolajal et al. 2021 (8)	1981−2018	197	19,413,702 participants	C	NA, EU, Asia, Australia	Alcohol, PA, BMI, smoking, BF, OC, HRT
Shamshirian et al. 2020 (41)	2002−2019	39	54,347 participants	CC, CS	Iran	PA, smoking, BF, HRT
Youn et al. 2020 (35)	2013−2015	214	Not mentioned	Any design except review and MA	Asia	BMI, PA, smoking, HRT, BF
Momenimovahed et al. 2019 (34)	Not mentioned	148	Not mentioned	Not declared	Not mentioned	Alcohol, PA, BMI, smoking
Vishwakarma et al. 2019 (31)	1991−2018	24	21,511 (9,889 cases and 11,622 controls)	CC	India	BMI, BF
Hardefeldt et al. 2018 (20)	Inception of database-2017	139	236,955 cases and 3,963,367 controls	C, CC, CS	Not mentioned	PA, BMI (weight loss)
Neil-Sztramko et al. 2017 (17)	1990−2016	29 (18 C and 11 CC)	C [1,156,200 (28,571 cases)], CC (15,540 cases and 41,238 controls)	C, CC	NA, EU, Australia, Asia	PA, BMI
Nindrea et al. 2017 (16)	1997−2017	15 (1 C and 14 CC)	5,381 cases and 9,682 controls	C, CC	Southeast Asia	BMI, BF, OC, menopausal status
Lambertini et al. 2016 (72)	2007−2014	11 BC	21,941 cases and 864,177 controls	C, CC	USA, EU, Asia	BF, HRT
Pizot et al. 2015 (39)	1987−2014	38	4,124,275 (116,304 cases)	PS	NA, EU, Asia	PA, HRT
Munsell et al. 2014 (21)	1980−2012	89 (57 BMI, 37 EPT)	88,214 cases	CT, C, CC	NA, EU	BMI, HRT
Namiranian et al. 2014 (14)	From the inception of databases to 2012	30 (25 CC, 1 NCC, 4 C)	44,909 (7,427 cases and 9,014 controls)	C, CC, NCC	EMR	Alcohol, PA, BMI, smoking,
Anothaisintawee et al. 2013 (61)	1985−2010	73 (65 CC and 8 C)	60,380 cases	C, CC	NA, EU, Asia, Australia	BF, OC, HRT
Gao et al. 2013 (40)	Up to December 2012	39	100,452 (13,204 cases and 87,248 controls)	C, CC	China	Alcohol, PA

Most AMSTAR 2 items were adhered to the quality assessment of included SRs (Figure 2, Supplementary Table S7). However, due solely to the absence of protocol registration (defined by AMSTAR 2 as a “critical” item), the majority of SRs were rated critically low or low for confidence in the results. If this item were excluded (since encouragement of protocol registration has occurred relatively recently), more than 90% of included SRs in our overview would be assessed as medium or high quality.

Figure 2.

Quality assessment according to the measurement tool to assess systematic reviews. RoB, risk of bias; NOS, Newcastle Ottawa Scale; SIGN, Scottish Intercollegiate Guidelines Network; MMAT, Mixed Methods Appraisal Tool; PICO, patients, intervention, comparison, outcomes; MA, meta-analysis.

BMI

Nineteen SRs reporting the association between BMI and BC incidence conducted meta-analysis and four (19,26,34,35) summarized information descriptively. Sixteen SRs showed a statistically significant positive association between postmenopausal BMI increase and BC incidence (Figure 3). Dose-response analyses showed that postmenopausal BC risk increased by 2%−12% for every 5 kg/m² BMI increase (15,22,24,28,29). Moreover, overweight/obesity was associated with higher incidence of BC [odds ratio (OR): 1.19−2.21] (14,16,31-33); there was a tendency for stronger association for obesity compared with overweight, supporting a dose-response relationship (Figure 4). On the other hand, premenopausal BMI increase (measured as abdominal body fatness; waist-circumference or waist-to-hip ratio) showed significant or non-significant protective associations in 8 of 12 SRs (Figure 4). Every 5 kg/m² of young age (≤30 years) BMI increase was associated with 12% (28) and 16% (27) premenopausal (Figure 4) and 17% postmenopausal (Figure 3) reduced risk of BC. Obesity was associated with lower hormone receptor-positive (HR-positive) premenopausal BC risk and greater postmenopausal BC risk, but no such relationship was found for hormone receptor negative (HR-negative) BC (21). However, one study (18) found a statistically significant positive association in premenopausal BC risk for each 0.1 unit increase of waist-to-hip ratio [relative risk (RR): 1.08, 95% CI: 1.01−1.16].

Figure 3.

Association between postmenopausal BMI and breast cancer incidence. BMI, body mass index; 95% CI, 95% confidence interval; WHR, waist-to-hip ratio; WC, waist circumference; HR+ BC, hormone receptor positive breast cancer. ^*, odds ratio.

Figure 4.

Association between premenopausal BMI and breast cancer incidence. BMI, body mass index; 95% CI, 95% confidence interval; WHR, waist-to-hip ratio; WC, waist circumference; HR+ BC, hormone receptor positive breast cancer.

Alcohol consumption

Eight SRs examining the association between alcohol consumption and BC incidence reported a significant positive association (Supplementary Table S8). One study showed that any type of alcohol intake (relative to none) was associated with greater BC incidence, including very light drinking (≤0.5 drink/d; RR: 1.04), light drinking (≤1 drink/d; RR: 1.09), and moderate drinking (1−2 drinks/d; RR: 1.13) (51). The magnitude of association was greater for heavy drinking [up to 61% RR increase (52)]. Increases of 10 g/d total alcohol or wine consumption were associated with 10.5% (RR: 1.10) and 8.9% (RR: 1.09) greater BC incidence, respectively (50). One SR restricted to Chinese populations found a non-significant inverse association between alcohol consumption and BC risk (RR: 0.85) (40).

Physical activity (PA)

Fifteen SRs (Supplementary Table S8) assessed the association between PA and BC, and all found protective associations, irrespective of PA type and menopausal status. Similar associations between PA and BC incidence were observed for occupational and non-occupational PA [statistically significant 13% and 10% risk reduction (37)]. The strength of the protective effect of PA varied by study design: risk reduction of BC associated with PA in cohort studies was 39%, and in case-control studies was 16% (38). PA categorized as high intensity, low intensity, premenopausal, and postmenopausal had protective associations of 27%, 21%, 21% and 18%, respectively (20). Leisure-time PA was inversely associated with BC incidence of similar magnitude for premenopausal and postmenopausal women (12% risk reduction) (43). PA in postmenopausal women with different BMI groups showed a significant protective association with BC incidence, but this association was less evident in the higher BMI group (17). However, for any BMI category, no significant association was found in premenopausal women between PA and BC risk. Higher PA levels at a young age (age ≤30 years) and lifetime PA were also inversely associated with BC incidence (RR: 0.81, and 0.79, respectively) (45). Another SR reported that while PA (occupational or non-occupational) was associated with reduced risk of BC, but for HRT users the association is not significant (RR: 0.97) (39).

Smoking

All nine SRs assessing the association between smoking and BC (Supplementary Table S8) established positive associations, and the relationship was stronger for passive smokers than active smokers. A SR including 3 cohort and 48 case-control studies published between 1992 and 2013 in China found a non-significant association of active smoking and BC (OR: 1.04), but a strong significant positive association for passive smoking (OR: 1.62) (46). Other studies also established passive smoking as a risk factor for BC (OR: 1.68−1.86 and RR: 1.17) (33,41,48). Analysis stratified by study design showed that active smoking had comparable results in prospective (RR: 1.10) and retrospective (RR: 1.08) studies. For ever-passive smoking, both types of study design reported positive associations, with a larger magnitude in retrospective (RR: 1.30) than prospective studies (RR: 1.07) (47).

OC use

Eleven SRs investigated the association between OC use and BC. Ever- vs. never-use of OC was associated with greater BC risk in six of eleven SRs (16,33,54,56,61,79), but in some cases, this association was not statistically significant (8,55,57,58) (Supplementary Table S8). Current/recent use of oral progestogen-only and combined hormonal contraceptives had strong association compared to past use (56). A meta-analysis of 19 case-control studies from Europe, North America, Asia and Australia reported that OC use was positively associated with triple negative BC risk (OR: 1.37) and estrogen receptor (ER)-negative BC risk (OR: 1.20) but showed an inverse association to risk of ER-positive BC (OR: 0.92) (60).

HRT

Of nine SRs examining HRT (Supplementary Table S8), five found a significant positive association between HRT and BC (8,41,61-63), although the magnitude of the association varied (RR: 1.23−5.48). Only one SR, focused on Korean women, found a OR smaller than one (0.98) between HRT and BC, with overlapping confidence intervals, suggesting that the study did not find a statistically significant association between HRT and an increased risk of BC (66). HRT use was found to be positively associated with BC incidence irrespective of whether the therapy was administered orally or through the transdermal route (64). A SR based in France reported an inverse association between BC and progesterone, which was more protective than synthetic progestin when used in combination with estrogen (RR: 0.67) (65).

Breastfeeding

Eight of eleven SRs found that ever- versus never-breastfeeding was associated with significantly lower BC risk (8,31,61,67-69,71,72) (Supplementary Table S8). Moreover, longer duration of breastfeeding was associated with stronger protective estimates (41,67,69,71). Three SRs found a non-significant inverse association between breastfeeding and BC incidence (OR: 0.68−0.84) (16,67,70). Furthermore, breastfeeding was associated with lower risk of triple negative BC (OR: 0.78−0.79) (68,72) as well as ER/progesterone receptor (PR) negative BC (OR: 0.90) (68). One SR found that exclusive breastfeeding was inversely associated with BC risk, and that ≥1-year extension of breastfeeding further strengthened this association (67).

Sedentary behavior (SB)

Five SRs (Supplementary Table S8) focused on SB (sedentary work, TV viewing). Three SRs demonstrated a statistically significant positive association with BC risk of varying magnitudes. Overall, there was a 16% increased risk of BC associated with sedentary work; the estimates of RR for cohort and case-control studies were 1.20 and 1.12, respectively (74). Two SRs reported a significant positive association between SB and BC risk (OR: 1.08, RR: 1.17) (75,77). Two SRs found a non-significant association between SB (defined as TV viewing for a prolonged time or 1-hour daily increment of TV viewing) and BC risk (73,76).

Healthy lifestyle index

One SR (Supplementary Table S8) assessed the relationship between the healthy lifestyle index and the risk of BC. Women in the highest category of the lifestyle index compared to those in the lowest category experienced a 20% (HR: 0.80, 95% CI: 0.78−0.83) and a 26% (HR: 0.74, 95% CI: 0.63−0.86) risk reduction in prospective and retrospective studies respectively. This protective association remained after the analysis was conducted by molecular subtypes (78).

Discussion

This overview of SRs synthesised evidence on lifestyle factors and BC risk in women. We focused on SRs rather than individual studies to provide a comprehensive overview of current knowledge on this topic, inform public health practice and policy, and assist discussions with women about lifestyle factors and the risk of BC. We found evidence from 40 SRs of consistent moderate to strong associations between postmenopausal BMI, HRT use, smoking, and alcohol consumption and BC risk. Evidence from 23 SRs showed protective associations for PA, breastfeeding, and healthy lifestyle scores. There was inconsistent evidence of association with BC risk for premenopausal BMI, SB, and OC use.

This overview highlights patterns in the relationship between different lifestyle factors and BC risk. Elevated postmenopausal BMI emerged consistently as a risk factor for BC, emphasising the importance of weight management interventions aimed at postmenopausal women. The impact of variations in body fat and distribution on BC is complex, especially in early life and young adulthood. Nevertheless, these effects are probably linked to how adipose tissue influences the levels of sex hormones (80). In postmenopausal women, the progress of ER-positive BCs is caused by higher circulating estrogen levels, which is influenced by adipose tissue — the main source of estrogen through aromatization of androgens. Obesity is also associated with chronic inflammation, insulin resistance, and changes in adipokine secretion, marked by elevated leptin and reduced adiponectin. These interconnected factors collectively stimulate cell proliferation, blood vessel formation, and tumor development. In premenopausal women, obesity may reduce the production of ovarian progesterone, while in postmenopausal women, it increases the production of oestradiol — both hormones associated with BC (81). While obesity was found to increase postmenopausal HR-positive BC risk and protect premenopausal individuals, it is not associated with HR-negative BC risk (21). In addition, postmenopausal women experienced increased BC risk, particularly at higher BMI levels, while premenopausal women did not (30). Our study suggests that interventions to promote postmenopausal healthy weight management are a priority public health policy for preventing BC incidence. The impact of BMI and PA on BC risk varied by menopausal status, with high activity potentially reducing risk in postmenopausal women with moderate BMI but offering less benefit at higher BMIs or in premenopausal women (17). Further research is needed to explain the underlying biological mechanisms driving these interactions between lifestyle and non-modifiable factors in BC aetiology.

Similarly, tobacco smoking, alcohol drinking, and HRT were found to be associated with increased BC risk. This could inform the development of widespread public health initiatives in promoting responsible alcohol drinking (including accurate information on safe levels), smoking cessation programs, and guidance on using hormone therapy. Findings from our overview may also support the development of more nuanced communication about lifestyle-related risk, for example, by highlighting the significant association between passive smoking and BC risk (which has policy implications), or by targeting information about the protective effect of PA to younger women.

Alcohol consumption may increase BC risk through several pathways, including facilitating the entry of carcinogens, enhancing DNA damage, and elevating oestrogen levels. Genetic differences in alcohol metabolism can also influence BC risk (82). Similarly, tobacco smoke, containing numerous carcinogens, is classified as a class one carcinogen and can lead to cancers in multiple organs (83). Targeting alcohol for risk reduction is gaining interest with recent research piloted specifically in breast screening programs to increase awareness of alcohol as a risk factor for BC (84), and in “warning labels” (85) research with implications for cancer risk reduction more broadly.

Conversely, some lifestyle factors have a recognized protective effect against BC. We found consistent evidence of risk reduction from maintaining regular PA, a healthy lifestyle score, and practicing breastfeeding. These modifiable lifestyle factors present promising potential for primary BC prevention. Promoting and supporting their adoption throughout a woman’s lifespan holds significant potential for reducing BC incidence. Public health policies can advocate for promoting the integration of PA into everyday routines and develop support systems for breastfeeding. Our results suggest that premenopausal weight gain is inversely associated with BC risk, which requires further exploration. Assessing the underlying mechanisms further may bring innovative risk assessment tools and preventive interventions customized for individuals with varying weight trajectories.

This overview revealed that magnitudes of association in prospective studies were consistently lower than those in retrospective studies (passive smoking, healthy lifestyle score). This may be attributable to a lower potential for selection bias and measurement error in exposure and outcome ascertainment in prospective designs. There was also a tendency for estimates in cohort studies to be larger than those in case-control studies (PA, SB, HRT). Although a practical design for studies of disease with low incidence, associations from case-control studies may be underestimated or overestimated due to the selection of controls or recall bias in measuring exposure. Nevertheless, consistency between study types in the direction of associations was observed.

Multiple lifestyle factors are likely to have complex inter-relationships in potentially impacting women’s individual risk of BC. Despite this, we found relatively few SRs exploring interactions between lifestyle factors. SRs found the association of PA and BC was modified by obesity (17) and HRT (39), and HRT modified the association for BMI (19,21), suggesting that instead of individual lifestyle factors, modifications of multiple factors in combination could be targeted in BC prevention strategies. Exploration of interactions between lifestyle factors may be challenging in meta-analysis where there is likely to be methodological heterogeneity between studies; however, this represents an important direction of research for future studies. Further evidence on the association between healthy lifestyle index (a composite measure of multiple lifestyle factors) and BC may also be warranted, given that this was examined in only one SR.

Our overview also points to gaps in knowledge and uncertainties that need to be considered. Heterogeneity in study design highlights the importance of standardized protocols and large-scale prospective studies, facilitating individual participant data meta-analysis to strengthen causal relationships and increase evidence generalizability. Despite these efforts, limitations including variations in study design and reliance on self-reported data are likely to persist, demanding care in interpreting results. We observed heterogeneity in risk estimation, which may be attributable to variation in study design, exposure measurements, and geographical regions of source studies. Many of the reported risk estimates were adjusted for different sets of covariates, including other measured risk factors, potentially reducing the comparability and consistency of findings across studies. The association between diet/nutrition and BC risk is complex and not yet completely understood; due to this complexity and the associated methodological heterogeneity in studies of this risk factor, diet was not considered in this overview. Additionally, uncovering the possible impact of environmental exposures, sleep patterns, and sociodemographic factors on BC risk could lead to a more holistic understanding of disease etiology and inform comprehensive prevention strategies. Future research could explore other influential factors like the gut microbiome and personalized genetic profiles (86).

For BC prevention, healthcare professionals may consider individualized risk assessment and evidence-based recommendations for lifestyle modifications such as BMI, PA, smoking, alcohol consumption, HRT, OC, and breastfeeding. However, we acknowledge that information on these associations (on its own) is insufficient to change behavior, and scalable supported interventions are required to encourage uptake and behavior change (87).

Conclusions

This overview highlights the associations between lifestyle factors and BC risk, addressing both risk-inducing and protective factors. Public health efforts should prioritize healthy lifestyles, including healthy weight management by targeting weight-loss programs especially for postmenopausal women; promote PA and longer duration of breastfeeding. By integrating smoking and alcohol counselling in breast screening communication pathways, and providing guidance on OC and HRT use, public health initiatives have the potential to support modification of individual BC risk.

SUPPLEMENTARY DATA

Supplementary data to this article can be found online.

Funding Statement

MSR is an awardee of “The Daffodil Centre Postgraduate Research International Scholarship”, the University of Sydney. NH declares funding via a National Breast Cancer Foundation (NBCF) Chair in Breast Cancer Prevention grant (No. EC-21-001) and a National Health and Medical Research Council (NHMRC) Investigator (Leader) grant (No. 1194410). MLM was supported by funding from a NBCF Investigator Initiated Research Scheme grant (No. 2023/IIRS0028).

Author contributions

N Houssami, XQ Yu, ML Marinovich, MM Rahman and MS Rana contributed in conceptualization and protocol development of the overview. MS Rana and ML Marinovich developed the search terms and conducted the literature search. MS Rana screened titles and abstracts. MS Rana and J Baker conducted full text screening and data extraction, ML Marinovich arbitrated disagreements. MS Rana and ML Marinovich undertook quality assessment and data synthesis. MS Rana wrote the first draft. All authors critically revised the draft for important intellectual content. All the authors read and approved the final manuscript.