Maximum and Time-Dependent Body Mass Index and Breast Cancer Incidence Among Postmenopausal Women in the Black Women’s Health Study

Abstract

While excess weight is an established risk factor for postmenopausal breast cancer, consideration of maximum body mass index (maxBMI; BMI is calculated as weight (kg)/height (m)²) or BMI at a point in time relevant for breast carcinogenesis may offer new insights. We prospectively evaluated maxBMI and time-dependent BMI in relation to breast cancer incidence among 31,028 postmenopausal women in the Black Women’s Health Study. During 1995–2015, a total of 1,384 diagnoses occurred, including 787 estrogen-receptor (ER)–positive (ER+) cases and 310 ER-negative (ER−) cases. BMI was assessed at baseline and 2, 4, 6, and 8 years before diagnosis. Cox proportional hazards regression was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Compared with women with BMI <25, those with BMI ≥35 had increased risk of ER+ breast cancer but not ER− breast cancer. For BMI assessed 2 years before diagnosis, the HRs for ER+ breast cancer associated with maxBMI ≥35 and time-dependent BMI ≥35 were 1.42 (95% confidence interval (CI): 1.10, 1.84) and 1.63 (95% CI: 1.25, 2.13), respectively. The corresponding HR for time-dependent BMI assessed 6 years before diagnosis was 1.95 (95% CI: 1.45, 2.62). These findings suggest strong associations of BMI with risk of ER+ breast cancer in postmenopausal women, regardless of timing of BMI assessment.

Abbreviations

BMI

body mass index

BWHS

Black Women’s Health Study

CI

confidence interval

ER

estrogen receptor

HR

hazard ratio

maxBMI

maximum BMI

WHR

waist:hip ratio

Epidemiologic data suggest that obesity accounts for 12%–15.4% of postmenopausal breast cancer in non-Black women and 28.3% in African-American women (1). Based on 2017–2018 national estimates, African-American women have a disproportionately higher prevalence of obesity (56.9%) than non-Hispanic White women (39.8%) (2). There is growing evidence that weight gain from early adulthood to late adulthood may be a better indicator of breast cancer risk than adult body mass index (BMI) (3–14). There is consistent evidence that elevated BMI is associated with increased risk of postmenopausal breast cancer, especially estrogen-receptor (ER)–positive (ER+) breast cancer (15–19), but studies of recent BMI in African-American women have shown relatively modest associations (20, 21). The use of maximum BMI (maxBMI) in a nonpregnant state has been found to yield stronger associations with mortality (22–24). MaxBMI attained in adulthood is less likely to be influenced by small fluctuations in weight over time and thus may better classify women according to overweight and obesity status. Another metric possible in longitudinal studies with repeated measures is BMI assessed at a point in time considered to be most relevant for breast carcinogenesis, rather than the most recent measure or a baseline measure. However, to our knowledge, the impacts of maxBMI attained in adulthood and time-dependent BMI on risk of incident breast cancer have not been studied.

Most studies have examined baseline BMI or recent BMI in analyses of associations with breast cancer risk. A disadvantage of relying on baseline BMI (i.e., BMI assessed at cohort entry) in follow-up studies is that it represents varying intervals between exposure and outcome for different individuals. Use of recent BMI (e.g., in case-control studies or cohort studies with repeated measures) assumes that the most relevant time period is that close to the time of diagnosis, which may not be a reasonable assumption for studies of cancer etiology. Therefore, in the present analysis, we assessed the impact of varying exposure lag intervals to investigate the relationship between BMI and breast cancer risk among postmenopausal women enrolled in a US cohort study, the Black Women’s Health Study (BWHS). Following the approach of Stokes et al. (22), we also evaluated associations for maxBMI.

METHODS

Study population

Self-identified African-American women (n = 59,000) aged 21–69 years were enrolled in the BWHS in 1995 and have been followed biennially ever since (25). At baseline, participants provided information on their medical history as well as on demographic, reproductive, lifestyle, dietary, and early-life factors via mailed self-administered questionnaires. Notices of deaths were obtained from next-of-kin, the US Postal Service, and annual searches of the National Death Index.

For this analysis, women were excluded if they had been diagnosed with breast cancer or any other type of cancer (except nonmelanoma skin cancer) before the start of follow-up. Only women who were postmenopausal (i.e., those whose periods had stopped because of natural causes for at least 1 year or because of surgery (hysterectomy with bilateral oophorectomy or bilateral oophorectomy alone)) were included in analyses. The final analytical cohort included 31,028 African-American women who were postmenopausal at enrollment in 1995 or became postmenopausal during follow-up. The study protocol was approved by the Boston University Institutional Review Board.

Case ascertainment

Incident cases of primary breast cancer in the BWHS were ascertained through self-report on biennial follow-up questionnaires (95% of cases) or identified through death records and linkage to 24 cancer registries in states covering 95% of participants (5% of cases). Diagnoses were confirmed by review of medical records, pathology reports, and cancer registry records. Data on tumor characteristics were abstracted from these records. Of cases for which pathology records have been received to date (>80%), 99% were confirmed. Through 2015, a total of 1,384 incident cases of invasive breast cancer were identified among postmenopausal women; 310 tumors were ER-negative (ER−) and 787 were ER+. ER status was unknown for the remaining 287 cases.

Data collection

Data were collected from biennial mailed or Web-based self-administered questionnaires. Women reported their current height and weight on the 1995 questionnaire; weight was updated on each biennial follow-up questionnaire. BMI was calculated as weight in kilograms divided by the square of height in meters at each follow-up cycle. Waist and hip circumference were also assessed on the 1995 questionnaire. Self-reports of weight, height, waist circumference, and hip circumference were highly correlated with technician measurements in a validation study (26).

The main exposures of interest were time-dependent BMI, including recent BMI, defined as BMI in the follow-up cycle prior to outcome assessment; lagged BMI, defined as BMI in the second, third, or fourth follow-up cycle prior to outcome assessment; and maxBMI attained during adulthood, defined as the maximum BMI assessed from the beginning of follow-up to the cycle prior to outcome assessment.

Covariates we considered were those that are established and putative risk factors for breast cancer, including BMI at age 18 years, family history of breast cancer, smoking history, alcohol consumption, Alternative Healthy Eating Index score (27, 28), vigorous exercise, and several reproductive factors: age at menarche, oral contraceptive use, number and timing of births, and use of menopausal female hormone supplements. Except for adult height, age at menarche, and weight at age 18 years, all variables were updated on follow-up questionnaires.

Statistical analyses

Mean values (standard deviations) and percentages were calculated to detect differences in the characteristics of study subjects according to categories of maxBMI and BMI as defined by the World Health Organization: underweight/normal (<25.0), overweight (25.0–29.9), obese class I (30.0–34.9), or obese class II or greater (≥35) (29, 30). Cox proportional hazards regression, stratified by age in 1-year intervals and questionnaire cycle such that age in years was the underlying time scale, was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for maxBMI and time-dependent BMI in relation to breast cancer incidence. Each exposure was considered as a time-varying variable in analyses, such that it was recalculated in each questionnaire cycle. HRs were calculated for overall breast cancer incidence, as well as for ER+ and ER− breast cancer separately. Follow-up included only postmenopausal person-time: Person-years were calculated from 1995 (or the earliest year in which a woman became postmenopausal, if she was premenopausal at the time of enrollment) to the midpoint of the calendar year of diagnosis of breast cancer, death, or the end of follow-up (March 2015), whichever occurred first.

Multivariable models included adjustment for established breast cancer risk factors and potential confounders: first-degree family history of breast cancer (yes, no), BMI at age 18 years (continuous), height (inches; continuous), age at menarche (<11, 11, 12–13, or ≥14 years), parity (nulliparous, 1 birth, 2 births, or ≥3 births), age at first birth (<25 years or ≥25 years), breastfeeding history (ever/never), oral contraceptive use (never, <5 years, 5–9 years, or ≥10 years), duration of estrogen-plus-progestin use (never, <5 years, or ≥5 years), Alternative Healthy Eating Index score (possible range, 0–110; quintiles), education (≤12 years, 13–15 years, 16 years, or ≥17 years), smoking status (never, current, or past smoker), vigorous physical activity (never, <5 hours/week, or ≥5 hours/week), and alcohol consumption (not a current drinker, 1–6 drinks/week, or ≥7 drinks/week). Missing indicator categories were used to account for missing information in covariates (generally 2%–4%).

For comparison, we conducted similar analyses using recent BMI (BMI updated in each 2-year follow-up cycle) as the exposure variable. In addition, in separate models, we allowed each exposure variable to lag by 4, 6, or 8 years, and we also evaluated associations for baseline BMI (i.e., BMI assessed at the start of follow-up, not updated). These analyses were performed to accommodate varying plausible latency periods for risk of breast cancer in relation to BMI. In sensitivity analyses, we conducted analyses restricted to never smokers; we also repeated analyses with a uniform sample size (i.e., restricting analyses of lagged BMI to women who also had recent measures of BMI available). Finally, we examined the association between waist:hip ratio (WHR) (<0.80, 0.81–0.85, or ≥0.86; assessed in 1995), a measure of central adiposity, and breast cancer incidence; updated data on WHR were not available.

All analyses were performed using SAS 9.3 (SAS Institute, Inc., Cary, North Carolina). All significance tests were 2-sided, and a P value of <0.05 was considered statistically significant.

RESULTS

The baseline characteristics of women enrolled in the BWHS are summarized in Table 1 according to maxBMI attained by the last cycle of follow-up. Nearly half of the cohort (48%) attained a maxBMI of 30 or more by the last cycle of follow-up, including 24% with maxBMI ≥35 (obese class II or greater). Women who had a maxBMI greater than or equal to 35 were similar to those with lower maxBMI with respect to age, height, family history of breast cancer, parity, alcohol consumption, and education; they were more likely to report young age at menarche, never use of oral contraceptives, and no physical activity and less likely to report current smoking and older age at first birth. Those in the high BMI categories were more likely to have a high BMI at age 18 years and at baseline and a high WHR in 1995. Among cases, the mean age at breast cancer diagnosis was 62.8 (standard deviation, 7.9) years.

Table 1.

Age-Adjusted Baseline Characteristics of Postmenopausal Women in the Black Women’s Health Study According to Maximum Body Mass Index Status at the End of Follow-up, 1995–2015

	Maximum BMI a  at End of Follow-up
	Underweight/Normal   (BMI <25.0)   (Mean = 22.9; n = 5,652)		Overweight   (BMI 25.0–29.9)   (Mean = 27.5; n = 10,432)		Obese Class I   (BMI 30.0–34.9)   (Mean = 32.3; n = 7,615)		Obese Class II or Greater   (BMI ≥35.0)   (Mean = 41.3; n = 7,329)
Characteristic	Mean (SD)	%	Mean (SD)	%	Mean (SD)	%	Mean (SD)	%
Age, years	52.5 (6.7)		53.4 (6.1)		53.2 (6.0)		52.8 (5.9)
aHEI score	45.7 (10.5)		44.7 (10.1)		43.3 (9.9)		41.3 (10.0)
Height, inchesb	65.1 (2.8)		64.9 (2.7)		64.6 (2.7)		64.6 (2.8)
BMI at age 18 years	18.8 (3.0)		19.8 (3.1)		21.0 (3.7)		23.8 (5.3)
Baseline BMI	22.6 (1.8)		27.0 (1.7)		31.5 (2.0)		39.4 (6.0)
Obese at baseline						80		92
Waist:hip ratio (in 1995)
≤0.80		66		54		45		42
0.81–0.85		11		15		15		15
≥0.86		11		16		23		25
Missing data		12		15		17		18
Current smoker		20		17		14		13
Family history of cancer		10		11		11		11
Age at menarche ≤11 years		19		24		29		37
Age at first birth ≥25 years		27		25		24		21
Nulliparous		22		18		17		21
≥7 alcoholic drinks/week at baseline		8		7		5		4
Never use of oral contraceptives		25		25		25		28
No physical activity		41		45		51		62
Education ≤12 years		17		19		20		20

Abbreviations: aHEI, Alternative Healthy Eating Index; BMI, body mass index; SD, standard deviation.

^a Weight (kg)/height (m)².

^b 1 inch = 2.54 cm.

In multivariable-adjusted models for overall breast cancer, the HR for updated maxBMI ≥35 versus maxBMI <25 was 1.24 (95% CI: 1.02, 1.50) (Table 2). This association reflected a positive association for ER+ breast cancer (HR = 1.42, 95% CI: 1.10, 1.84) (Table 3). In contrast, there was no apparent association for ER− breast cancer (HR = 0.77, 95% CI: 0.51, 1.16). When the maxBMI measure was taken from successively earlier periods (lagged analyses), the HRs for ER+ breast cancer for the highest category of maxBMI versus the lowest increased up to an exposure lag of 6 years (HR = 1.72, 95% CI: 1.31, 2.28) (Table 3).

Table 2.

Hazard Ratios for Breast Cancer Incidence in Relation to Maximum and Time-Dependent Body Mass Index Among Postmenopausal Women in the Black Women’s Health Study, 1995–2015

	Maximum BMI				Time-Dependent BMI
BMI a  Metric	No. of Cases	No. of   Person-Years	HR b	95% CI	No. of Cases	No. of Person-Years	HR b	95% CI
Updated BMI (no lag)	1,384	351,726			1,197	291,209
Underweight/normal (<25.0)			1.00	Referent			1.00	Referent
Overweight (25.0–29.9)			0.95	0.81, 1.12			1.01	0.86, 1.18
Obese class I (30.0–34.9)			0.97	0.81, 1.17			1.12	0.94, 1.34
Obese class II or greater (≥35.0)			1.24	1.02, 1.50			1.38	1.13, 1.68
4-year-lagged BMI	1,235	310,919			1,082	259,730
Underweight/normal (<25.0)			1.00	Referent			1.00	Referent
Overweight (25.0–29.9)			1.03	0.87, 1.22			1.11	0.94, 1.32
Obese class I (30.0–34.9)			1.01	0.84, 1.22			1.06	0.87, 1.28
Obese class II or greater (≥35.0)			1.30	1.06, 1.59			1.42	1.15, 1.75
6-year-lagged BMI	1,122	285,363			966	239,208
Underweight/normal (<25.0)			1.00	Referent			1.00	Referent
Overweight (25.0–29.9)			1.05	0.88, 1.26			1.09	0.92, 1.31
Obese class I (30.0–34.9)			1.16	0.96, 1.41			1.15	0.94, 1.41
Obese class II or greater (≥35.0)			1.38	1.12, 1.72			1.46	1.17, 1.83
8-year-lagged BMI	1,020	256,319			885	215,547
Underweight/normal (<25.0)			1.00	Referent			1.00	Referent
Overweight (25.0–29.9)			1.02	0.85, 1.23			0.97	0.81, 1.16
Obese class I (30.0–34.9)			1.20	0.99, 1.47			1.16	0.95, 1.43
Obese class II or greater (≥35.0)			1.44	1.16, 1.80			1.59	1.26, 2.00
Baseline BMI					1,376	348,769
Underweight/normal (<25.0)			N/A	N/A			1.00	Referent
Overweight (25.0–29.9)			N/A	N/A			1.12	0.98, 1.29
Obese class I (30.0–34.9)			N/A	N/A			1.26	1.06, 1.48
Obese class II or greater (≥35.0)			N/A	N/A			1.42	1.16, 1.73

Abbreviations: BMI, body mass index; CI, confidence interval; N/A, not applicable.

^a Weight (kg)/height (m)².

^b All models adjusted for age, first-degree family history of breast cancer, age at menarche, parity, age at first birth, breastfeeding history, oral contraceptive use, duration of estrogen-plus-progestin use, Alternative Healthy Eating Index score, education, smoking, physical activity, alcohol consumption, height, and BMI at age 18 years (continuous).

Table 3.

Hazard Ratios for Estrogen-Receptor–Positive and Estrogen-Receptor–Negative Breast Cancer Incidence in Relation to Maximum and Time-Dependent Body Mass Index Among Postmenopausal Women in the Black Women’s Health Study, 1995–2015

	ER-Positive Breast Cancer								ER-Negative Breast Cancer
	Maximum BMI				Time-Dependent BMI				Maximum BMI				Time-Dependent BMI
BMI a  Metric	No. of   Cases	No. of   PY	HR b	95% CI	No. of   Cases	No. of   PY	HR b	95% CI	No. of   Cases	No. of   PY	HR b	95% CI	No. of   Cases	No. of   PY	HR b	95% CI
Updated BMI (no lag)	787	351,084			685	290,660			310	350,591			273	290,244
Underweight/normal (<25.0)			1.00	Referent			1.00	Referent			1.00	Referent			1.00	Referent
Overweight (25.0–29.9)			0.91	0.72, 1.14			1.05	0.84, 1.31			1.00	0.72, 1.40			0.96	0.70, 1.32
Obese class I (30.0–34.9)			1.08	0.85, 1.37			1.34	1.06, 1.71			0.79	0.54, 1.15			0.80	0.55, 1.17
Obese class II or greater (≥35.0)			1.42	1.10, 1.84			1.63	1.25, 2.13			0.77	0.51, 1.16			0.86	0.57, 1.32
4-year-lagged BMI	733	310,369			644	259,256			267	309,894			235	258,835
Underweight/normal (<25.0)			1.00	Referent			1.00	Referent			1.00	Referent			1.00	Referent
Overweight (25.0–29.9)			1.05	0.84, 1.32			1.23	0.98, 1.54			1.12	0.79, 1.59			1.03	0.73, 1.46
Obese class I (30.0–34.9)			1.16	0.91, 1.49			1.32	1.02, 1.69			0.86	0.58, 1.29			0.73	0.48, 1.11
Obese class II or greater (≥35.0)			1.56	1.19, 2.03			1.75	1.33, 2.31			0.85	0.54, 1.34			0.90	0.57, 1.42
6-year-lagged BMI	675	284,876			575	238,782			243	284,443			218	238,423
Underweight/normal (<25.0)			1.00	Referent			1.00	Referent			1.00	Referent			1.00	Referent
Overweight (25.0–29.9)			1.15	0.91, 1.46			1.34	1.05, 1.70			1.06	0.73, 1.53			0.82	0.58, 1.17
Obese class I (30.0–34.9)			1.30	1.01, 1.69			1.42	1.08, 1.86			1.07	0.72, 1.61			0.84	0.56, 1.26
Obese class II or greater (≥35.0)			1.72	1.31, 2.28			1.95	1.45, 2.62			0.84	0.52, 1.36			0.76	0.47, 1.24
8-year-lagged BMI	629	255,895			552	215,188			222	255,473			200	214,833
Underweight/normal (<25.0)			1.00	Referent			1.00	Referent			1.00	Referent			1.00	Referent
Overweight (25.0–29.9)			1.03	0.81, 1.30			1.05	0.82, 1.33			1.12	0.77, 1.63			0.97	0.67, 1.40
Obese class I (30.0–34.9)			1.31	1.01, 1.69			1.35	1.03, 1.76			1.08	0.71, 1.66			0.91	0.58, 1.40
Obese class II or greater (≥35.0)			1.69	1.27, 2.24			1.95	1.45, 2.61			0.92	0.56, 1.53			0.95	0.58, 1.58
Baseline BMI					781	348,128							309	347,645
Underweight/normal (<25.0)			N/A	N/A			1.00	Referent			N/A	N/A			1.00	Referent
Overweight (25.0–29.9)			N/A	N/A			1.18	0.97, 1.42			N/A	N/A			0.98	0.74, 1.30
Obese class I (30.0–34.9)			N/A	N/A			1.39	1.12, 1.74			N/A	N/A			0.92	0.65, 1.30
Obese class II or greater (≥35.0)			N/A	N/A			1.68	1.30, 2.18			N/A	N/A			0.75	0.48, 1.17

Abbreviations: BMI, body mass index; CI, confidence interval; ER, estrogen receptor; HR, hazard ratio; N/A, not applicable; PY, person-years.

^a Weight (kg)/height (m)².

^b All models adjusted for age, first-degree family history of breast cancer, age at menarche, parity, age at first birth, breastfeeding history, oral contraceptive use, duration of estrogen-plus-progestin use, Alternative Healthy Eating Index score, education, smoking, physical activity, alcohol consumption, height, and BMI at age 18 years (continuous).

Associations for time-dependent BMI were somewhat stronger than those observed for maxBMI. For ER+ breast cancer, the HR for updated BMI ≥35 versus BMI <25 was 1.63 (95% CI: 1.25, 2.13) in the time-dependent analysis. The comparable HR was 1.95 (95% CI: 1.45, 2.62) when a 6-year lag was applied. When baseline BMI was used as the exposure measure, the HR for BMI ≥35 vs. BMI <25 was 1.68 (95% CI: 1.30, 2.18) (Table 3). In sensitivity analyses that used a uniform sample, associations for time-dependent BMI and ER+ breast cancer were similar in updated and lagged analyses (see Web Table 1, available at https://doi.org/10.1093/aje/kwac004).

Results of analyses restricted to never smokers were similar (data not shown). Finally, WHR assessed in 1995 was positively associated with risk of ER− breast cancer risk but not ER+ breast cancer. Comparing women with WHR ≥0.86 with those with WHR <0.80, HRs for ER− and ER+ breast cancer were 1.52 (95% CI: 1.00, 2.30) and 1.03 (95% CI: 0.78, 1.36), respectively.

DISCUSSION

The findings of this novel study show that high BMI was associated with increased risk of postmenopausal ER+ breast cancer but not ER− breast cancer, regardless of whether the exposure metric was baseline BMI, maxBMI, or time-dependent BMI. While maxBMI was associated with increased risk of ER+ breast cancer among postmenopausal African-American women, associations were not stronger in magnitude than when time-dependent BMI was considered as the exposure variable. Our findings are in accord with earlier analyses carried out within the BWHS with 10 years of follow-up, in which Palmer et al. (20) reported a positive but nonsignificant association between recent BMI and ER+ breast cancer in postmenopausal women (for BMI ≥30 vs. BMI <25, HR = 1.66, 95% CI: 0.86, 3.21).

Many aspects of weight have been evaluated with respect to breast cancer risk, including current/recent BMI and weight gain/changes. The evidence supports excess weight and weight gain as risk factors for postmenopausal breast cancer (15–18), including among African-American/Black women (20, 21, 31–34). The findings of our study confirm the established associations for recent BMI (20, 21) and also show that maxBMI is associated with increased risk of ER+ postmenopausal breast cancer while accounting for BMI at age 18 years and other breast cancer risk factors. We found no association of BMI with ER− postmenopausal breast cancer; results from prior studies evaluating ER− breast cancer have been mixed (21, 34, 35).

Maximum BMI has been proposed as a relevant metric for evaluating associations of body size with mortality, since it is not affected by recent weight loss (22–24, 36); however, it has not been extensively evaluated with respect to cancer incidence. We hypothesized that maxBMI would be a stronger predictor of breast cancer risk than recent BMI because it may be less influenced by fluctuations in weight over time and therefore may better classify women according to obesity status. Maximum and time-dependent BMI are highly correlated in the BWHS (r = 0.95). Our study did not show a marked difference in breast cancer risk associated with maxBMI compared with time-dependent BMI; however, associations were somewhat stronger for time-dependent BMI than for maxBMI.

To address timing of exposure to adiposity in relation to breast cancer risk, we lagged maxBMI and time-dependent BMI by varying intervals prior to outcome assessment. Our primary analyses suggested that risk estimates may increase with successively longer exposure lags, up to 6–8 years; however, results from a sensitivity analysis with a uniform sample did not support differences in associations by varying lag intervals.

The biological mechanisms linking obesity and breast cancer risk are complex and are hypothesized to include hormonal and nonhormonal pathways. It is believed that obesity can induce aromatase expression and estrogen synthesis, contributing to breast carcinogenesis (37–39). In addition, adipose tissue is a major source of estrogenic hormones after menopause, and both aromatase expression and estrogen synthesis are linked to increased risk of breast cancer (40). Hyperinsulinemia and increased levels of insulin-like growth factor 1, which are common in obesity, also up-regulate aromatase activities and have been linked to cell proliferation and tumor progression (41). Additionally, obesity is characterized by elevated levels of proinflammatory cytokines, which are likely to influence breast tumor development (39, 42, 43). Notably, patterns of associations differed for BMI, a measure of general adiposity, and WHR, a measure of central adiposity. Specifically, while BMI was positively associated with risk of ER+ but not ER− breast cancer, the opposite was observed for WHR. These findings suggest possibly different pathways of carcinogenesis driven by general obesity versus central obesity for ER+ and ER− disease; however, other studies have observed positive associations of WHR with both ER+ and ER− postmenopausal breast cancer (21, 33). Further research incorporating hormonal and inflammatory biomarkers may help to elucidate the biological underpinnings of obesity-related postmenopausal breast cancer.

Study limitations are those inherent to observational studies. A limitation of this study is that BMI, though a universal measure of adiposity, may not accurately capture visceral fat (44, 45). We examined WHR as a proxy for central adiposity; however, only baseline measures of WHR were available, and a relatively high proportion of participants were missing information on this variable. Self-reports of height and weight were shown in a BWHS validation study to have a high degree of accuracy (Spearman correlations were 0.93 and 0.97, respectively) (26); however, some bias in reporting due to concepts of socially desirable weight may have consciously or unconsciously influenced reporting.

The strengths of this study include the large number of incident breast cancer cases, which allowed us to assess associations of maxBMI and time-dependent BMI with breast cancer subtypes and to obtain excellent statistical power. The study measures included BMI at age 18 years, which was controlled for in multivariable models; this is key, because early BMI has been shown to have an inverse association with breast cancer risk (46, 47). Since this was a prospective study where exposure data were collected every 2 years and prior to the outcome, there is little concern about recall bias. The longitudinal nature of these data allowed us to evaluate successive exposure lag periods to identify susceptible windows of risk.

In conclusion, our study suggests that BMI is positively associated with risk of ER+ postmenopausal breast cancer; associations for maxBMI were weaker than those observed for time-dependent BMI. These findings suggest that recent weight loss is not likely to strongly bias results in cohort studies with repeated assessment of BMI, such as the BWHS. Further, maximum BMI may be a reasonable proxy for recent BMI in studies that do not have repeated measures. Finally, our findings suggest that greater adiposity, as measured by BMI, is relevant for breast cancer risk regardless of timing of exposure assessment.