We observed a strong positive association between the risk of breast cancer and fasting insulin levels in postmenopausal women who were neither diabetic nor using HT. In these women, breast cancer incidence rates were 2.4-fold greater among those in the highest quartile compared with the lowest quartile of fasting insulin level, even after controlling for estradiol levels, BMI, free IGF-I level, and established breast cancer risk factors. Estradiol level was associated with ER-positive and PR-positive breast cancers only, and an initially strong association between BMI and postmenopausal breast cancer was greatly attenuated by adjustment for insulin level and, to a lesser degree, for estradiol level. Together, insulin and estradiol levels appeared to largely explain the association between obesity and postmenopausal breast cancer among the women in this study.
Obesity is a well-established risk factor for postmenopausal breast cancer (2
). Results of previous studies have indicated that this association may be partly explained by the high levels of circulating estrogen in obese women. For example, two large prospective studies (50
) demonstrated that the association between BMI and the risk of breast cancer was substantially reduced by controlling for estrogen levels. In this study, the association between BMI and the risk of breast cancer was attenuated more by controlling for insulin level than by controlling for estradiol level. Laboratory data support a direct role for insulin in the etiology of breast cancer. For example, the binding of insulin to the insulin receptor activates the mitogen-activated protein kinase (MAPK) and phosphatidylinositol-3-kinase pathways, which leads to an increase in cell proliferation (52
). In addition, insulin stimulates cell growth in both normal and neoplastic breast tissue (11
) and promotes breast tumor growth in animal models (13
). Insulin may also play a role in breast carcinogenesis via the extensive cross talk that occurs between the insulin–IGF and the estrogen signaling pathways in breast tissue. Insulin and estradiol can act in concert to promote cell cycle progression in breast cancer cells (56
), and insulin activates estrogen receptor alpha–mediated transcription in breast cancer cell lines (57
), even in the absence of estradiol (59
), whereas estrogen stimulates the insulin signaling pathway by enhancing MAPK activation (60
A strength of this study was our ability to control for endogenous estradiol levels, which allowed us to evaluate the association between insulin level and postmenopausal breast cancer independent of estradiol level. A previous investigation conducted among European women that measured C-peptide levels and adjusted for levels of estradiol and other sex hormones found that the association between C-peptide levels and the risk of postmenopausal breast cancer was attenuated after controlling for free estradiol level (21
). By contrast, we observed a positive association between fasting insulin levels and the risk of postmenopausal breast cancer that was unaffected by adjustment for endogenous estradiol levels, supporting a mechanism that is independent of circulating estradiol. Similarly, the positive association of estradiol levels with breast cancer risk was unaltered by controlling for insulin levels, suggesting that hyperinsulinemia and high estrogen may represent two distinct mechanisms that underlie the obesity–breast cancer relationship. Although we directly measured fasting insulin levels, unlike the European study, we did not control for free estradiol level, which may be the more bioactive estradiol component in circulation. However, in previous studies (50
), adjustment for free estradiol level resulted in only a slightly greater attenuation of the obesity–breast cancer association than adjustment for total estradiol level. Moreover, the association between fasting insulin level and the risk of breast cancer observed in this study was much more robust than that between C-peptide level and breast cancer risk reported by the European study.
Our finding that there was no association between total IGF-I level and breast cancer risk was largely consistent with that in earlier reports (31
). However, we found a modest positive association between the level of free IGF-I and the risk of breast cancer among nonusers of HT in our study population. Free IGF-I is purported to be the main bioactive component of circulating IGF-I (30
), and in two previous studies (18
), the level of free IGF-I was more strongly associated with the risk of postmenopausal breast cancer than the level of total IGF-I. We did not find evidence for a linear association between free IGF-I levels and breast cancer risk; however, we observed a possible curvilinear association between free IGF-I levels and the risk of breast cancer in nonusers of HT. However, we note that this association was attenuated following adjustment for insulin levels, suggesting that the strong relation between insulin levels and breast cancer observed in this study may underlie the association between free IGF-I levels and breast cancer risk. Furthermore, given that we had not hypothesized a curvilinear association and the paucity of prospective data regarding the relationship between free IGF-I level and breast cancer risk, these data should be interpreted with caution.
We found no association between the risk of incident breast cancer and BMI, the fasting levels of insulin, or free IGF-I among women who used HT. This finding is consistent with data from several large prospective cohort studies that found that HT use interacts with the association between obesity and postmenopausal breast cancer (4
). However, this interaction was not confirmed in subsequent clinical trials (61
). It is possible that an association between insulin level and the risk of breast cancer may have been obscured by the high estrogen levels in the HT users. However, other physiologic changes that arise as a consequence of HT use may also explain why an association between insulin and breast cancer was not observed among HT users. In particular, orally administered HT exposes the liver to a large bolus of estrogen, which alters hepatic protein synthesis (63
). Indeed, in this study, the levels of IGF-I and IGFBP-3, both of which are produced in the liver (63
), were statistically significantly lower among the women who used HT (which in the WHI-OS is almost exclusively orally administered HT) than in those who did not. Insulin levels were also statistically significantly lower among the HT users than among the nonusers of HT in our dataset. The use of HT therefore complicates the interpretation of our data on the associations between insulin and IGF-I levels and breast cancer risk, and greater knowledge of the effects of orally administered HT on hepatic protein synthesis may be needed to better understand the associations between these factors and the risk of breast cancer among HT users.
Our study has some important limitations. First, we assessed only baseline serum levels of insulin, glucose, total IGF-I, free IGF-I, IGFBP-3, and estradiol. Measuring these factors over time may have allowed us to classify the study participants more precisely with respect to their circulating levels of these factors. However, plasma levels of insulin, total IGF-I, free IGF-I, IGFBP-3, and estradiol in postmenopausal women have been shown to be stable over at least 3 years (65
), suggesting that any misclassification of study subjects according to levels of these factors is likely to have been limited. Second, we cannot exclude the possibility that some case subjects had subclinical breast cancer at WHI-OS recruitment and that the associations we observed were due in part to reverse causality. However, we limited this study to women who developed breast cancer after at least 12 months of follow-up, thereby minimizing the potential bias due to breast cancers that were undiagnosed at baseline. Although the latency period for the breast cancers in this study cannot be known with certainty, we subsequently conducted an exploratory analysis that included only women who developed breast cancer 3 or more years after the baseline blood draw and found that the results were essentially unaltered.
In summary, our data indicate that hyperinsulinemia and high endogenous estradiol levels are independent risk factors for postmenopausal breast cancer and largely explain the association between obesity and the risk of breast cancer in postmenopausal women. Interventions aimed at lowering fasting insulin levels or circulating estrogen levels—either through weight loss or increased physical activity or via pharmacologic approaches—may reduce the risk of breast cancer in postmenopausal women.