Type II Diabetes, Obesity, and Breast Cancer Risk: The Multiethnic Cohort

The rising burden of type II diabetes affects populations around the world, but the incidence rates are particularly high in African Americans, Native Americans, Pacific Islanders, Latinos, and also in several Asian populations despite their relatively low obesity rates (1). Obesity is considered the most important modifiable risk factor for type II diabetes (2–5) and, depending on the population, one of the most important factors for breast cancer besides physical activity, hormone treatment, and alcohol intake (6–8). Rates for these two conditions are increasing at a rapid rate in all parts of the world (9, 10). The association between type II diabetes and a higher risk of developing breast cancer has been investigated widely in primarily white populations (11, 12). For example, a meta-analysis of 20 studies reported a 20% higher risk for women with type II diabetes (13) and a later meta-analysis based on 39 independent studies showed a summary relative risk of 1.23 [95% confidence interval (CI), 1.12–1.35] for prospective cohorts (11). Adjustment for BMI reduced the association between type II diabetes and breast cancer (11), for example, the summary estimate was 1.16 (95% CI, 1.08–1.24) in studies adjusting for BMI as compared with 1.33 (95% CI, 1.18–1.51) in studies without BMI adjustment. Because of the strong influence of BMI on the risk of both diseases that are increasing worldwide, the exploration of the combined impact of type II diabetes and obesity on breast cancer risk is very important. It appears that the associations of obesity with type II diabetes (12) and breast cancer (2, 14) differ by ancestry and that type II diabetes confers differential breast cancer risk across geographic locations (15). Therefore, we investigated the relative importance of obesity and type II diabetes alone and in combination as independent risk factors for breast cancer in African American, Native Hawaiian, Japanese American, Latino, and white female participants of the Multiethnic Cohort (MEC).

The MEC was established in 1993–1996 to study the association of lifestyle and genetics with cancer and other chronic diseases among different ethnic groups in Hawaii and California (16). More than 215,000 members ages 45–75 years at recruitment entered the cohort (26% Japanese American, 23% white, 22% Latino, 16% African American, 7% Native Hawaiian, and 6% Other). The Institutional Review Boards at the University of Hawaii and the University of Southern California approved the study protocol.

Breast cancer cases were identified through regular linkages with the Los Angeles County Cancer Surveillance Program, the State of California Cancer Registry, and the statewide Hawaii Tumor Registry, all part of the NCI's Surveillance, Epidemiology, and End Results (SEER) program, which provided information on date of diagnosis, tumor stage, and hormone receptor status. Vital status was identified by routine linkages with California and Hawaii vital records and the National Death Index. Complete case and death ascertainment was available up to December 31, 2010. For the current analysis, all women from the five major ethnic groups represented in the MEC and free of breast cancer at cohort entry were included (Table 1). Preexisting cases of breast cancer as reported at baseline and identified through the SEER registries, as well as women with missing BMI information at cohort entry, were excluded. The final dataset for analysis included 103,721 women of whom 6,692 had been diagnosed with breast cancer, including in situ and invasive tumors.

At cohort entry, participants completed a self-administered, 26-page questionnaire (QX1) that collected self-reported demographic characteristics including place of birth, height and body weight, medical history, family history of breast cancer, physical activity, a dietary history, and a reproductive history in women. Between 1999 and 2002, approximately 85% of eligible MEC members answered a brief follow-up questionnaire (QX2) that updated information on medical history and self-reported body weight. During 2003–2007, a full questionnaire (QX3) was returned by approximately 50% of cohort members providing an update on BMI, medical conditions, and information on current or past use of diabetes pills or insulin shots. Ethnic-specific response rates of those who were alive at the time of QX2 and QX3 were 85%/60% for whites, 74%/37% for African Americans, 84%/57% for Native Hawaiians, 89%/60% for Japanese Americans, and 75%/42% for Latinas. Each questionnaire included the question “Has your doctor ever told you that you had diabetes?” A biorepository was created in 2001–2006 on about 70,000 MEC members; a specimen QX was administered at the time of blood and urine collection when participants were asked to provide an inventory of current medications but no information about start date and years of use was collected.

On the basis of the information from the three questionnaires (QX1–3), as well as and three sources of administrative data, that is, Medicare claims (17), California hospital discharges (18), and a Hawaii health plan linkage (3), we developed a strict definition of type II diabetes. Women with at least one self-reported type II diabetes diagnosis, which was confirmed by one or more administrative data source, were considered cases. The first report of a type II diabetes diagnosis was considered as the year of discovery as type II diabetes often remains latent for several years and exact dates of diagnosis are unavailable. A Medicare linkage for 1999–2012 provided information on 114,309 fee-for-service beneficiaries (17) as part of the Chronic Condition Warehouse (CCW) that supplies information on common diseases using specific algorithms based on Part A institutional claims and Part B institutional and noninstitutional claims (19). We note that for participants enrolled in Medicare-managed care plans, comparable information cannot be obtained. The State of California hospital discharge data base provided information for each patient, including those in managed care, who is treated as an inpatient in a licensed general acute care hospital in California using the same algorithm as Medicare (18). In Hawaii only, records for surviving MEC members were linked in 2007 to the diabetes care registries of the two major insurers in Hawaii that cover at least 90% of the population; multiple claims for diabetes-related treatment services were used to identify cases (3).

Cox regression with age as the time-metric and ethnicity included as a strata variable (African American, Native Hawaiian, Japanese American, Latina, or white) was applied to examine the association between type II diabetes, obesity, and breast cancer incidence. Type II diabetes status was modeled as a time-varying exposure, but only the years with type II diabetes since cohort entry were counted. All models included potential confounders selected a priori based on established risk factors and previous findings in the MEC (2, 20). They included age at cohort entry (continuous), family history of breast cancer (no, yes, missing), educational level (≤12, 13–15, ≥16 years), BMI at cohort entry (<20, 20–<25, 25–<30, ≥30 kg/m²), alcohol consumption (<1/month, ≥1/month to <1/day, ≥1/day, missing), physical activity (<30 or ≥30 minutes of moderate/vigorous activity, missing), age at menarche (≤12, 13–14, ≥15 years, missing), age at first childbirth (no children, ≤20, 21–30, ≥31 years, missing), number of children (0, 1, 2–3, ≥4, or missing), type of menopause (premenopausal, natural, oophorectomy, or hysterectomy, missing), age at menopause (<45, 45–49, 50–54, ≥55 years, missing), and postmenopausal hormone therapy (HT; never, past estrogen with or without progestin, current estrogen without progestin, current estrogen with past or current progestin, missing). Missing covariate information was included into the models as a separate category (Table 1).

To explore whether BMI and type II diabetes are independent predictors of breast cancer risk, type II diabetes models with and without BMI at cohort entry, as well as BMI models with and without type II diabetes, were compared. We examined three potential interactions, that is, type II diabetes/ethnicity, type II diabetes/BMI, and BMI/ethnicity, using global Wald tests of the cross-product terms. As sensitivity analysis, we restricted the models to invasive cases only, postmenopausal women, to observations without missing covariates, and to type II diabetes cases reported after cohort entry. Separate models by ethnicity, and also by birth place for Latinas (United States vs. foreign-born), BMI category, stage of disease at diagnosis, estrogen receptor (ER)/progesterone receptor (PR) status, and medication use as reported from the specimen QX or QX3 were examined to compare effect sizes across subgroups.

Among 103,721 (25,146 white, 20,255 African American, 7,681 Native Hawaiian, 28,012 Japanese American, 22,627 Latina) female participants (Table 1), 18,178 women self-reported type II diabetes and 14,558 (80%) type II diabetes cases were confirmed by administrative data sources. The major reasons for lack of confirmation were not being part of Medicare due to young age or managed care plan (N = 2,987), not member of a linked health plan in Hawaii (N = 825), and deaths (1,218 by 2010).

During 14.8 ± 4.1 years of follow-up, 6,692 breast cancer cases were identified in the entire cohort; among women with type II diabetes, 989 developed breast cancer but only 896 cases occurred after type II diabetes had been diagnosed. The mean time period between type II diabetes and breast cancer diagnosis was 7.3 ± 4.4 years. The prevalence of obesity (≥30 kg/m²) differed by ethnicity: 19% in whites, 38% in African Americans, 36% in Native Hawaiians, 7% in Japanese American, 32% in U.S.-born and 26% in foreign-born Latinas (P < 0.0001). The respective percentages of women with type II diabetes were 2%, 19%, 35%, and 44% among low, normal, overweight, and obese participants. Women with type II diabetes were older at cohort entry and more likely to be African American, Native Hawaiian, or Latina; they also reported lower education, less physical activity, and higher parity.

In models without BMI (Table 2), type II diabetes was significantly associated with breast cancer (HR, 1.15; 95% CI, 1.07–1.23; P = 0.0002), but including BMI lowered the relative risk estimate to 1.08 (95% CI, 1.00–1.16; P = 0.04). Modeling BMI as continuous variable or dividing the highest BMI category into 30–<35 and ≥35 kg/m² did not change the results (HR, 1.08; 95% CI, 1.00–1.16 for both models). Findings restricted to postmenopausal women were similar with respective HRs of 1.16 (95% CI, 1.07–1.25) and 1.09 (95% CI, 1.00–1.18).

Although the interaction between type II diabetes and ethnicity was not statistically significant (P = 0.12), stratification by ethnic group showed significant or suggestive increases in breast cancer risk among Latinas (HR, 1.33; 95% CI, 1.14–1.56 and HR, 1.30; 95% CI, 1.11–1.52), African Americans (HR, 1.18; 95% CI, 1.02–1.37 and HR, 1.14; 95% CI, 0.99–1.33), and whites (HR, 1.18; 95% CI, 0.98–1.43 and HR, 1.10; 95% CI, 0.90–1.34) for models without and with BMI. However, type II diabetes status was not a significant predictor of breast cancer risk among Native Hawaiian and Japanese American women with and without BMI adjustment. The interaction term for place of birth among Latinas was not significant (P = 0.59), but the BMI-adjusted relative risk estimates was lower for U.S.-born (HR, 1.18; 95% CI, 0.96–1.44) than foreign-born (HR, 1.39; 95% CI, 1.08–1.79) women. Among the 11,418 U.S.-born, 21% had a type II diabetes diagnosis, while the proportion in the 11,209 foreign-born women was 16%.

The type II diabetes with BMI interaction was also not significant (P = 0.31) but small differences in risk by BMI category were detected. Only within the group of obese participants was the risk elevated by 12% (95% CI, 1.00–1.26); type II diabetes was not a significant predictor for breast cancer among low, normal, and overweight women. Stratified models by ER/PR status indicated no difference in the association between type II diabetes and breast cancer after adjustment for BMI, whereas stage-specific models showed elevated risks for regional/distant (HR 1.22; 95% CI, 1.03–1.44) and localized disease (HR 1.11; 95% CI, 1.00–1.24) but not for in situ cancers. Excluding the 1,134 in situ breast cancer cases that may have resulted from screening in more health conscious women changed the BMI-adjusted relative risk estimate only to a small degree (HR 1.09; 95% CI, 1.01–1.18).

Of women with type II diabetes, 8,343 (57%) participated in one of the questionnaires asking about medication and 7,046 (48%) reported medication in the specimen QX (N = 5,045) or QX3 (N = 6,848). The remaining 7,512 type II diabetes cases did not report any medication or did not participate in the specimen QX or QX3. Of the 896 incident breast cancer cases, 654 (73%) participated in one of the questionnaires and 556 (62%) reported diabetes medication but timing and length of use in relation to breast cancer diagnosis was unknown. In users of type II diabetes medication, a significant association of type II diabetes with breast cancer was observed after BMI adjustment (HR, 1.27; 95% CI, 1.16–1.38), while the BMI-adjusted relative risk for never users was 0.88 (95% CI, 0.79–0.98; Table 2).

A sensitivity analysis of 79,510 participants without missing covariate values indicated unchanged results with an overall relative risk of 1.08 (95% CI, 0.99–1.17) for the BMI adjusted model. When all type II diabetes cases diagnosed before cohort entry were excluded, the risk estimate also did not substantially change (HR, 1.09; 95% CI, 0.89–1.35).

BMI (Table 3) was an important predictor for breast cancer (P < 0.0001) with a borderline interaction effect across ethnic groups (P = 0.05). The respective HRs for overweight and obesity as compared with the 20–<25 kg/m² category in the total study population were 1.21 (95% CI, 1.14–1.29) and 1.31 (95% CI, 1.22–1.41) when type II diabetes was included. However, these relative risk estimates were basically unchanged in models without type II diabetes. BMI was also a statistically significant predictor in all ethnic groups except Latinas (P = 0.27). The respective HRs for obese women as compared with the 20–<25 kg/m² category in white, African Americans, Native Hawaiians, and Japanese American were 1.33 (95% CI, 1.15–1.53), 1.21 (95% CI, 1.04–1.41), 1.35 (95% CI, 1.09–1.67), and 1.46 (95% CI, 1.23–1.72). Among obese Latinas, the association did not reach statistical significance (HR, 1.17; 95% CI, 0.99–1.38) despite being similar in strength as in African Americans. The interaction term for place of birth among Latinas was not significant (P = 0.24); however, the risk was only elevated in U.S.-born (HR, 1.32; 95% CI, 1.06, 1.64) and not in foreign-born (HR, 0.94; 95% CI, 0.72–1.23) women. Stratification by type II diabetes status indicated strong associations of obesity with breast cancer for women with type II diabetes (HR, 1.41; 95% CI, 1.16–1.71) and without (HR, 1.30; 95% CI, 1.20–1.41).

Further stratification by ER/PR status (Table 4) indicated that obesity was a risk factor for breast cancer only in ER⁺/PR⁺ tumors across ethnic groups with HRs as high as 1.71 for Japanese Americans although the risk estimates were not significant after separation of Latinas by birth place.

Among a population of 103,855 participants with 14,558 confirmed type II diabetes and 6,692 breast cancer cases, type II diabetes was modestly associated with breast cancer before but not after BMI adjustment. When stratified by ethnicity, we detected a 30% higher risk with and without BMI adjustment for Latinas, a small 10%–14% increase among African American and white women but no significant associations among Japanese American and Native Hawaiian women. In contrast, the association with obesity was not significant in Latinas as was observed in an earlier report from the MEC (2). Interestingly, stratification by place of birth among Latinas suggested that obesity was a risk factor for breast cancer in U.S.-born but not foreign-born women, while type II diabetes predicted a higher risk among foreign-born but not U.S.-born Latinas. However, the differences between U.S.-born and foreign-born Latinas need to be interpreted with caution given the lack of significant interaction terms. As in previous reports (21), obesity was only associated with ER⁺/PR⁺ tumors.

The 8% increase in risk for the entire MEC population and the 10% nonsignificant increase in risk for whites associated with type II diabetes is lower than reported in the literature, for example, a relative risk of 1.23 (95% CI, 1.12–1.35) for a meta-analysis in prospective cohorts (11), although the risk was also lower when type II diabetes was included in the model (RR, 1.16; 95% CI, 1.08–1.24) than in studies without BMI (RR, 1.33; 95% CI, 1.18–1.51). The results by geographic location are contradictory. Whereas the association between type II diabetes and breast cancer risk was higher in Europe (RR, 1.88; 95% CI, 1.56–2.25) and America (RR, 1.16; 95% CI, 1.12–1.20) than Asia (RR, 1.01; 95% CI, 0.84–1.21) in one report (15), an older meta-analysis with a larger number of studies (13) found a higher relative risk estimate in Asia (RR, 1.45; 95% CI, 1.07–1.97) than in North America (RR, 1.12; 95% CI, 1.06–1.18) and Europe (RR, 1.19; 95% CI, 1.08–1.31). The current study did not confirm the higher risk for ER⁺ than ER⁻ tumors reported from the Nurses' Health Study although the results looked suggestive (22).

For African Americans, our BMI-adjusted results agree with the only published study of 1,128 breast cancer cases (23) that reported no significantly higher breast cancer risk associated with type II diabetes. For women of Asian ancestry, four previous cohorts found significantly elevated risks for breast cancer in women with type II diabetes (24–29), but our results are consistent with three cohorts that detected no significant increase of breast cancer associated with type II diabetes (30–32) or type I diabetes (33). For example, the most recent study with 182,542 Japanese women reported a risk estimate of 0.98 (95% CI, 0.69–1.38; ref. 32). The difference in relative risk estimates for obesity and breast cancer, adjusted for type II diabetes, between Japanese Americans at 1.46 versus 1.33 for whites in the current analysis is somewhat smaller than in a summary report that indicated a 2-fold higher risk for obese as compared with normal weight among women of Asian ancestry (14) as compared with the risk estimate of 1.27 (95% CI, 1.03–1.55) reported for mostly white populations (34).

In contrast to the current study, a population-based case–control study in the U.S. Southwest (35) with Hispanic and non-Hispanic women found no association of a type II diabetes history with pre- (OR, 0.89; 95% CI, 0.64, 1.26) or postmenopausal breast cancer (OR, 0.96; 95% CI, 0.78, 1.19) despite the higher prevalence of type II diabetes among Latina than white women. Interestingly, one study showed that Latina women with type II diabetes who exercised had a significantly reduced breast cancer risk (OR, 0.41; 95% CI, 0.21–0.83) as compared with women with type II diabetes who did not exercise (OR, 0.96; 95% CI, 0.63–1.48); women without a type II diabetes history and reporting no physical activity served as the reference group (36). As to the weak association between obesity and breast cancer among Latina women, similar findings have been reported previously for the MEC (2) and for Hispanic women in Los Angeles (37) and in the U.S. Southwest, which detected no association of obesity and weight gain in Hispanic women as compared with a significant positive association with breast cancer risk in whites (38). The point has been made that the association with various established breast cancer risk factors, not only BMI, appears to be different in Latinas (39) and that weight gain appears to be a better predictor breast cancer than recent BMI among Latinas with a low BMI in early adulthood (14). One explanation could be that admixed genetic components lead to diverse metabolic consequences of obesity in Hispanic women, for example, a high degree of genetically defined American Indian ancestry predicted a lower risk of breast cancer associated with high weight/obesity and fat distribution (38). Other possible explanations are that the increase in BMI associated with migration and acculturation to the U.S. lifestyle (40–42) affects the metabolic profile associated with obesity or body fat distribution leading to higher visceral and liver fat that may increase breast cancer risk more than BMI (43). The presence of type II diabetes may also influence the uptake of cancer screening in a way that recent immigrants to the United States may be less likely to be diagnosed with early stage type II diabetes as indicated by the lower proportion of type II diabetes in foreign versus U.S.-born Latinas in the MEC (21% vs. 16%).

The major strength of the current study is the availability of multiple data sources for type II diabetes status. This makes it possible to create a robust and consistent definition of diagnosis across ethnic groups, provides high specificity, and avoids misclassification. In addition, the MEC not only includes large numbers of type II diabetes and breast cancer cases from ethnic groups with disparate risk for both conditions, but the participants also represent a wide range of BMIs. These diverse data offered the opportunity to investigate the association of type II diabetes with breast cancer with increased power than previous reports. One of the limitations of the MEC data is the lack of detailed information about type II diabetes treatment as diabetes medication in all type II diabetes cases, in particular the insulin glargine as a possible risk factor and metformin as a protective agent, have received a lot of attention (44–46). Medication use may also be an indicator of disease severity and, therefore, explain the stronger association between type II diabetes and breast cancer in the MEC (HR 1.27 for users vs. HR 0.88 for nonusers) similar to a reported risk of 2.23 for insulin-treated women and 0.72 for those with oral medication (36).

As possible mechanisms for a link between type II diabetes and breast cancer, insulin resistance, hyperinsulinemia, hyperglycemia and dyslipidemia, inflammatory cytokines, and adipokines have been discussed as they are all associated with increased tumor growth and metastasis (47). Hyperglycemia may contribute to the development and progression of cancers by promoting transformation of cancer cells and providing energy for cell growth. The possible effects of insulin are supported by a report from the Women's Health Initiative showing that elevated fasting insulin levels were associated with breast cancer risk (48). It is possible that the balance of glucose and fasting insulin is more important than adiposity in the development of breast cancer (45). Foreign-born Latinas with similar weight status may, therefore, be more metabolically healthy than U.S.-born Latinas and/or those with higher acculturation (38–41). To clarify the factors that contribute to the associations between diabetes, hyperglycemia, diabetes treatment, and cancer, it may be helpful to investigate the influence of glycemic control in future studies (45).

As to the relative importance of type II diabetes and obesity, the current data suggest that excess body weight is a strong predictor of breast cancer development in women with and without type II diabetes in all ethnic groups except foreign-born Latinas, possibly a chance finding. This observation agrees with a study of type II diabetes patients, in which obesity conferred a significant 19% higher breast cancer risk (4). On the other hand, type II diabetes increased breast cancer risk by only 8% in this multiethnic population after taking into account BMI status and the increased risk was limited to foreign-born Latinas, more advanced stage breast cancers, type II diabetes cases using medication, and obese women. Conversely, BMI was an important predictor for breast cancer independent of type II diabetes status except in foreign-born Latinas suggesting a different relation between obesity and metabolic health in these women.

No potential conflicts of interest were disclosed.

Conception and design: G. Maskarinec

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.R. Wilkens, L. Le Marchand

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Maskarinec, S. Jacobs, V.W. Setiawan

Writing, review, and/or revision of the manuscript: G. Maskarinec, S. Jacobs, S.-Y. Park, C.A. Haiman, V.W. Setiawan, L.R. Wilkens, L. Le Marchand

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V.W. Setiawan

Study supervision: L. Le Marchand

The work within the Multiethnic Cohort was supported by the following grants from the NIH: R37CA54281 (to L.N. Kolonel), U01CA164973 (to L. Le Marchand, L.R. Wilkens, C.A. Haiman), and R21 DK73816 (G. Maskarinec). S. Jacobs was supported by a postdoctoral fellowship from the German Research Foundation (DFG, JA 2564/1-1). The tumor registries are supported by NCI contracts N01 PC 35137 and N01 PC 35139.

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.