Overall and central adiposity and risk of contralateral breast cancer: The California Breast Cancer Survivorship Consortium.

INTRODUCTION
In survivors of a first primary invasive breast cancer (FBC), the risk of developing a contralateral second primary breast cancer (CBC) is at least two-fold higher compared with the risk of developing a FBC in women from the general population.1 Risk factors for CBC include younger age at FBC diagnosis, FBC with negative hormone receptor and other tumor characteristics, radiation therapy, family history of breast cancer, and genetic susceptibility (e.g., BRCA1 or BRCA2 germline pathogenic variants).1–4 Furthermore, CBC risk is higher in African American, Asian American, and Hispanic survivors compared with non-Hispanic White survivors,5–11 particularly those first diagnosed at younger ages.8, 9, 12 Finally, given the increasing incidence of early-onset FBC and hormone receptor-positive FBC in the population,13 CBC is a growing concern. Little is known about the prevention of CBC and the role of potentially modifiable risk factors.14 Given advances in treatment and early detection, a better understanding of CBC prevention is essential for the growing number of FBC survivors.
Obesity is a well-established, potentially modifiable risk factor for FBC in postmenopausal women15 and is associated with higher breast cancer-specific and all-cause mortality.16 For other outcomes such as breast cancer recurrence and CBC, data on associations with obesity are limited, with suggestive, but not consistent, evidence for higher CBC risk16 (Supplementary Table 1). In premenopausal women, obesity was associated with higher CBC risk in one small study,17 but not in another study.18 Findings are also inconsistent in studies that included FBC survivors across a broad age range and did not stratify the analyses by age or menopausal status, with higher risks with obesity observed in some studies,19–21 but not others.22, 23 In three meta-analyses of BMI and CBC risk that included up to nine partly overlapping studies, between-study heterogeneity was high.16, 24, 25 For FBC, BMI associations are modified by menopausal status and hormone receptor status. Higher risks associated with higher BMI are limited to postmenopausal women who never used menopausal hormonal therapy (HT) and those with hormone receptor-positive FBC.15
For CBC, potentially modifying factors have not been adequately investigated. It remains uncertain whether BMI associations with CBC differ by menopausal status or FBC hormone receptor status18, 26, 27 and the potentially modifying effect of HT use has not been examined. Central adiposity, which reflects visceral fat stores and is a strong predictor of health outcomes,28 also has not been evaluated in relation to CBC risk. Furthermore, prior studies included primarily non-Hispanic White women, thus it is not known whether associations with obesity differ across racial and ethnic groups. To address these knowledge gaps, we evaluated associations of overall and central adiposity with risk of CBC, by menopausal status, HT use, FBC hormone receptor status, and race and ethnicity in a large, racially and ethnically diverse cohort of women diagnosed with FBC and followed for CBC.12

MATERIALS AND METHODS

Study sample and outcomes
The present study was conducted using data from the California Breast Cancer Survivorship Consortium (CBCSC)29 that has been updated with second primary cancer diagnoses and vital status through 2020.12 We analyzed harmonized data on clinical, personal, and lifestyle characteristics collected in six studies, including three case-control studies, a family-based study of breast cancer, a breast cancer survivor cohort, and a prospective general population cohort (see details and references in Supplementary Table 2). Women with FBC were diagnosed at age 18–79 years from 1993 through 2009 in Northern or Southern California and ascertained through regional cancer registries that are part of the California Cancer Registry and the Surveillance, Epidemiology, and End Results (SEER) program. For women with FBC who participated in more than one study, the data were retained from one study only. The remaining cohort (N=11,269) was linked with the California Cancer Registry, with follow-up through December 31, 2020 (n=8,045 FBC) or through December 31, 2019 for the Multiethnic Cohort (MEC, n=1,594 FBC). After exclusions (Supplementary Figure 1), analyses for CBC were based on 9,479 FBC survivors without a bilateral mastectomy at FBC diagnosis. As in other studies,7, 30 we limited the analysis to invasive CBC diagnosed more than 180 days after FBC diagnosis to minimize the inclusion of metastases that are misclassified as CBC.

Data collection
Variables obtained from the California Cancer Registry included tumor characteristics for FBC [histology, American Joint Committee on Cancer stage, grade, estrogen receptor (ER) status, progesterone receptor (PR) status, lymph node involvement, tumor size], initial treatment (surgery type, radiation therapy, chemotherapy), age and year of diagnosis, and marital status. Personal and lifestyle factors from questionnaires included self-reported race and ethnicity (categorized as non-Hispanic African American, non-Hispanic Asian American, Hispanic (Black or White), or non-Hispanic White) and education level. Data on height, weight, and other risk factors were collected retrospectively after FBC diagnosis in the case-control studies [Asian American Breast Cancer Study (AABCS), Women’s Contraceptive and Reproductive Experiences Study (CARE), San Francisco Bay Area Breast Cancer Study (SFBCS)] and family-based study [Northern California Breast Cancer Family Registry (NC-BCFR)] and assessed up to the year before FBC diagnosis. In the survivor cohort [Life After Cancer Epidemiology Study (LACE)], data were collected after FBC diagnosis and assessed up to the time of questionnaire completion (on average 1.8 years after FBC diagnosis); additionally, information was collected on weight in the year before FBC diagnosis. In MEC, the questionnaire data used in this analysis were collected at study enrollment (on average 6.7 years before FBC diagnosis). Data on HT use before FBC diagnosis were available for a subset of 5,587 FBC survivors from three studies for this specific analysis (AABCS, SFBCS, and NC-BCFR).
Pre-diagnosis BMI, a measure of overall adiposity, was calculated as self-reported weight (kg) divided by self-reported height (m) squared. In AABCS and SFBCS, missing self-reported height was replaced with measured height. BMI (kg/m2) was categorized as normal weight (<25), overweight (25-<30), or obesity (≥30) for all women except Asian American survivors for whom BMI was categorized as normal weight (<23), overweight (23-<27.5), or obesity (≥27.5) according to cut-points suggested for Asian populations.31 Underweight (BMI <18.5 kg/m2, n=177) was combined with the normal BMI category. Waist circumference (WC) and height measurements were available for a subset of 2,760 FBC survivors from two studies (AABCS and SFBCS). We analyzed two measures of central adiposity at FBC diagnosis: WC, categorized according to the tertile distribution in our study sample and clinical cut-points of <80, 80-<88, and ≥88 cm.32 The second measure, waist-to-height ratio (WHtR), was calculated as measured waist (cm) divided by measured height (cm) and classified according to clinical cut-points of ≤0.5, >0.5–0.6, and >0.60.33 We also calculated the body roundness index, a newer measure of visceral body fat distribution34 which has been associated with risk of colorectal cancer,35 other outcomes and all-cause mortality.36 The body roundness index was calculated as 364.2 − 365.5 × the square root of [1 – (WC in centimeters / 2π) squared / (height in centimeters / 2) squared],34 and categorized according to the tertile distribution in our study sample. Lastly, we examined the association with a composite measure of the joint classification by BMI (overweight or obesity vs. normal weight) and WC (tertiles 2 and 3 vs. tertile 1).

Statistical analysis
For the FBC cohort and CBC, we present distributions of clinical, personal, and lifestyle characteristics. For categorical body size variables, we assessed differences in distributions across racial and ethnic groups using chi square tests. Differences in mean BMI, WC, WHtR, and body roundness index across racial and ethnic groups were assessed using Student’s t-distribution with Bonferroni’s adjustment for multiple pairwise comparisons (critical p value 0.008, based on a significance level of 0.05 for six pairwise comparisons). We calculated the proportion of overweight or obesity by year of diagnosis and age categories and used Mantel-Haenszel chi-square to test for linear trend across the years of diagnosis within each age category. We assessed changes in BMI over time using linear regression models and calculated means for BMI by year of FBC diagnosis (1993–1997, 1998–2002, 2003–2009), stratified by age at FBC diagnosis (<40, 40–49, 50–59, 60–69, ≥70 years), adjusted for age (continuous) and race and ethnicity. We used Fine and Gray competing risk regression models, with death and diagnosis of other primary cancers as competing events, and estimated hazard ratios (HR) and 95% confidence intervals (CI) for the association of BMI, WC, WHtR, and body roundness index with risk of CBC. Follow-up extended until diagnosis of CBC, diagnosis of another primary cancer (n=577, comprising 291 ipsilateral breast cancers, 5 breast cancers with unknown laterality, and 281 other cancers), death (n=3,559), date of last contact, or study end, whichever was earliest. Deaths and other primary cancers were considered competing events. Models were adjusted for study (stratification variable), age at FBC diagnosis, and race and ethnicity, and for potentially confounding factors previously identified, including FBC hormone receptor status and first-degree family history of breast cancer.12 We performed BMI analyses for all FBC survivors, and stratified by menopausal status, HT use in postmenopausal survivors, FBC hormone receptor status, and race and ethnicity. For WC, WHtR, and body roundness index, we fit competing risk regression models for all survivors combined and stratified the analyses by menopausal status, FBC hormone receptor status, and race and ethnicity. Heterogeneity in associations were assessed by including interaction terms in the models and using the Wald test to assess statistical significance. None of the variables included in the multivariable models violated the proportional hazards assumption. Two-sided p-values <0.05 were considered statistically significant. Analyses were performed using SAS software version 9.4 (SAS Institute, Inc., Cary, NC, USA).

RESULTS
During follow-up of FBC survivors of up to 26.7 years (mean: 15.5 years, median: 16.9 years) 610 CBC were diagnosed, an incidence rate of 0.4% CBC cases per person-year. Mean time between FBC and CBC diagnosis was 9.7 years. In the five studies that administered questionnaires after FBC diagnosis, median time from FBC to questionnaire completion was 2.0 years with an interquartile range of 1.5–2.7 years. Of 9,479 FBC survivors, 2,110 (22%) were African American, 1,865 (20%) were Asian American, 2,330 (25%) were Hispanic, and 3,174 (33%) were non-Hispanic White (Table 1). Distributions of clinical, personal, and lifestyle factors are presented for FBC and CBC. High proportions of survivors had pre-diagnosis overweight or obesity (59%) or large (>0.5) WHtR (66%), and distributions varied by race and ethnicity (Supplementary Table 3). In age groups 40–49, 50–59, and 60–69 years, both mean BMI and proportion of overweight or obesity increased over time (Supplementary Figure 2).
In the analysis of the pooled dataset from six studies that did not consider HT use, BMI was not associated with risk of CBC, neither overall nor in premenopausal or postmenopausal survivors (Table 2). In the dataset from three studies with available data on HT use (5,287 FBC survivors, 342 CBC), a nearly two-fold higher risk of CBC was associated with obesity (vs. normal weight) in postmenopausal survivors without pre-diagnosis HT use (HR=1.89, CI=1.10–3.24, per 5 kg/m2 increase HR=1.22, CI=1.07–1.40). Analyses stratified by FBC hormone receptor status showed more than two-fold higher risk of CBC only in postmenopausal survivors of hormone receptor-negative FBC without pre-diagnosis HT use (HR=2.55, CI=1.07–6.09, per 5 kg/m2 increase HR=1.36, CI=1.12–1.65) (Table 3). Heterogeneity by hormone receptor status, however, was not statistically significant (p-het=.66).
In analyses of six studies stratified by race and ethnicity, no associations between risk of CBC and BMI were observed in any group, neither overall, nor in survivors with hormone receptor-positive or hormone receptor-negative FBC (Supplementary Table 4). Positive associations per 5 kg/m2 increase in BMI were observed in postmenopausal African American (HR=1.17, CI=0.999–1.37) and Asian American (HR=1.26, CI=1.001–1.59) survivors, but not in Hispanic (HR=0.86, CI=0.71–1.05) and non-Hispanic White (HR=1.01, CI=0.88–1.16) survivors. Sample sizes were too small to consider HT use in analyses stratified by race and ethnicity.
In the subset of survivors from two studies with available data on WC (2,760 FBC survivors, 192 CBC), a higher risk of CBC was associated with larger WC (tertile 1 vs. tertile 3: HR=1.59, CI=1.05–2.39, p-trend=.02), which was driven by two-fold higher risk observed in postmenopausal survivors (p-trend=.07) (Table 4). Sample size was too small to further stratify the analysis in postmenopausal survivors by HT use. No dose response trends were observed for WHtR or body roundness index. For the composite measure of BMI and WC, in survivors overall and postmenopausal survivors, risk was higher in those with larger WC, regardless of BMI, when compared to survivors with normal weight and smaller WC, with HRs ranging from 1.75 to 2.37. In premenopausal survivors, a higher risk of CBC (HR=1.68, 1.003–2.80) was also observed in survivors with overweight or obesity and larger WC compared with survivors with normal BMI and smaller WC. In survivors of hormone receptor-positive FBC, larger WC in postmenopausal survivors was the only central adiposity measure associated with higher risk of CBC (tertile 2 or 3 vs. tertile 1: HR=2.15, 1.03–4.49). In survivors of hormone receptor-negative FBC, two-fold higher risks were associated with larger WC (tertile 2 or 3 vs. tertile 1: HR=2.21, 1.02–4.81) or larger body roundness index (tertile 2 or 3 vs. tertile 1: HR=2.53, 1.17–5.46) (Supplementary Table 5).
Higher risk of CBC was associated with larger WC (tertile 2 or 3 vs. tertile 1: HR=1.71, CI=1.08–2.72) and larger body roundness index (tertile 2 or 3 vs. tertile 1: HR=1.75, CI=1.08–2.84) in Asian American women, whereas no associations were observed in Hispanic women (Supplementary Table 6). In survivors of hormone receptor-positive FBC, no statistically significant associations with any central adiposity measure were observed, with HR estimates that were somewhat lower than those observed overall. Sample sizes were too small for separate analyses in survivors of hormone receptor-negative FBC.

DISCUSSION
In this pooled cohort of California FBC survivors, a higher risk of CBC was associated with obesity in postmenopausal survivors who had never used HT before FBC diagnosis. To our knowledge, our study is the first to assess BMI associations with CBC risk in the context of HT use in postmenopausal survivors. Leveraging multiple measures of body size, our study is also the first to consider central adiposity, identifying that larger WC was associated with higher risk of CBC. Importantly, the association with WC was independent of BMI. More than half of women diagnosed with FBC in our study had overweight or obesity and a large WC (≥78.8 cm), suggesting that a large proportion of FBC is at higher risk (estimated at 78%) of developing CBC compared with survivors with normal weight and lower WC. Given the continued rise of overall and central obesity seen in the U.S.37–39 and in our survivor cohort, the prevalence is likely higher in more recently diagnosed patients, highlighting the importance of considering enhanced surveillance and risk-reduction strategies in this population.
We observed an association between CBC risk and BMI only in postmenopausal survivors without pre-diagnosis HT use, although heterogeneity by menopausal status was not statistically significant. This result parallels findings on the BMI association with risk of FBC in postmenopausal women, where BMI associations are limited to or stronger in women who never used HT.15, 40 For FBC, the higher risk associated with BMI in postmenopausal women has been linked to hormone-related mechanisms,15 such as conversion of androgens to estrogens in peripheral adipose tissue, which is the major source of estrogen in postmenopausal women.41 Higher estrogen concentrations are associated with higher risk of FBC.42 The modifying effect of HT use has been attributed to the higher hormone concentrations in HT users which override the effects of endogenous estrogens derived from adipose tissue, thereby masking BMI associations in HT users.40 Our findings suggest that HT use also may mask BMI associations with CBC in current or former HT users, and may contribute to the inconsistent findings for postmenopausal survivors in the current literature (Supplementary Table 1). Depending on the prevalence of HT use in different study populations, risk estimates associated with BMI in postmenopausal women vary accordingly. When we did not consider HT use, no BMI association in postmenopausal women was observed.
The epidemiological evidence on the BMI association with risk of CBC is sparse and inconsistent for younger women. In agreement with another study,18 we found no association of BMI with risk of CBC in premenopausal women. In a third study in survivors diagnosed with FBC at age <45 years, higher BMI (≥30 vs. ≤19.9 kg/m2) was associated with a two-fold higher risk of CBC.17 That study, however, included only 77 CBC. Several studies in survivors of FBC across a broad age range also produced mixed results, with reports of higher risk of CBC in survivors of FBC overall,19 ER-positive FBC,20 and in situ FBC,21 and higher risk of ipsilateral and contralateral SBC combined.43 Other studies found no association with BMI.22, 23 The summary estimates from three meta-analyses are difficult to interpret, as the data were combined across studies despite numerous differences in study design.16, 24, 25 Potentially modifying factors, including HT use, were not investigated, and some recent studies22, 43, 44 were not included in the meta-analyses.
We found that the association of WC with risk of CBC remained after adjustment for BMI. Our finding of higher CBC risk associated with larger WC regardless of BMI suggests that central adiposity may be a stronger risk factor for CBC than overall adiposity. Larger studies need to confirm this finding. Visceral fat is associated with insulin resistance and chronic inflammation,28 which are risk factors for the development and progression of FBC.45 These pathways and related biomarkers, however, have not been evaluated for CBC. Interestingly, the association with WC was stronger for WC categorized by tertiles (p-trend=.02) than for WC categorized by clinical cut-points (p-trend=.05). The clinical cut-points have been primarily used for non-cancer outcomes such as hypertension, diabetes, dyslipidemia, and metabolic syndrome.32 The optimal cut-points for CBC remain to be determined. Our findings suggest that the higher cut-points by tertiles yield associations of larger magnitude. Although we found no statistically significant heterogeneity by menopausal status, associations with WC were limited to postmenopausal survivors.
We are not aware of any studies that examined BMI associations with CBC risk across racial and ethnic groups. Although the prevalence of overweight and obesity varied widely across the four racial and ethnic groups, we found no evidence of statistically significant heterogeneity in associations with BMI, likely because of limited sample size. As has been observed for FBC,46 it is possible that the impact of BMI and WC on risk of CBC varies across racial and ethnic groups, given observations of borderline positive associations per 5 kg/m2 increase in BMI in postmenopausal African American and Asian American survivors, but not in Hispanic and non-Hispanic White survivors, and a positive association with larger WC in Asian American survivors, but not in Hispanic survivors. Compared with non-Hispanic White women, Asian women have a higher prevalence of central adiposity at every level of BMI47, 48 and a higher proportion of visceral than subcutaneous adipose tissue,49 whereas African American women have less visceral adipose tissue,50, 51 and Hispanic women have more subcutaneous adipose tissue than non-Hispanic White women.51 Given evidence that Asian populations have different health risks associated with obesity, we used BMI cut-points that have been recommended for Asian populations.31
Few studies have examined whether BMI associations with CBC differ by FBC hormone receptor status.18, 20, 26, 27, 43, 44 Risk estimates were of similar magnitude for ER-positive FBC and FBC overall in some studies,20, 43, 44 whereas other studies,26, 27 including ours, suggest that the BMI association may differ by FBC hormone receptor status, with a higher obesity-associated risk in postmenopausal survivors with ER-negative FBC than those with ER-positive FBC. Similarly, we found that WC was associated with higher risk only in survivors with hormone receptor-negative FBC. It is possible that treatment with endocrine therapy reduced CBC risk in postmenopausal women with hormone receptor-positive FBC,52 even in the presence of obesity. We did not have information available on endocrine therapy use. Other potential mechanisms that are independent of estrogen, such as inflammatory pathways, may also play a role. The sample size of hormone receptor-negative FBC was small in our pooled dataset, warranting confirmation in larger studies of CBC. Interestingly, some studies of FBC also observed BMI associations in postmenopausal women for hormone receptor-negative tumors,15, 53 although in most studies, the BMI associations were limited to hormone receptor-positive tumors.15, 40, 53
Our study has some limitations. Foremost, despite a large cohort and long follow-up, the sample sizes for CBC in subgroup analyses tended to be small and rendered some HR estimates that were marginally statistically significant, although in the expected direction. BMI was based on self-reported height and weight and may therefore be underestimated.54 The analysis by HT use was limited to data from three studies, and data on central adiposity measures were available only for two studies. Like most prior studies, we had no data available on weight change after FBC diagnosis. Unlike some other risk factors, weight is a time-dependent variable, and weight or BMI at FBC diagnosis alone is likely not an optimal body size measure. We did not have information available on adjuvant endocrine therapy for FBC which has been associated with lower CBC risk in women with a hormone receptor-positive FBC.52 Lastly, we performed a large number of statistical comparisons. We can therefore not rule out the possibility that some associations may be due to chance.
Despite these limitations, our study has several strengths, including predominantly population-based studies with uniform ascertainment of CBC through the California Cancer Registry and a racially and ethnically diverse study population. Unlike most previous studies that evaluated associations with BMI, we also examined associations with three measures of central adiposity (WC, WHtR, body roundness index). The availability of both clinical and harmonized epidemiologic characteristics allowed us to adjust the multivariable models for potential confounding factors.
Our findings underline the importance of considering HT use in the assessment of BMI associations in postmenopausal women. We found that obesity was associated with a nearly two-fold higher risk of CBC in survivors who never used HT, with an even higher risk for survivors of hormone receptor-negative FBC. Thus, obesity may be a more important risk factor for CBC than gleaned from the current epidemiologic literature that did not consider HT use in the analysis. Higher risk also was associated with larger WC which needs to be confirmed in larger studies. Our findings suggest that weight and waist circumference in patients newly diagnosed with FBC should be considered in prevention strategies for CBC. Among patients newly diagnosed with FBC, those with overweight or obesity and with central adiposity may benefit from more frequent surveillance screening after FBC diagnosis and communications about potential risk-reducing strategies, such as weight loss through lifestyle modifications and/or use of weight loss medications such as GLP-1. Few modifiable risk factors have been identified for CBC, thus more research efforts should be directed towards investigating the role of overall and central adiposity and weight gain in CBC.

Supplementary Material
supplementary materials