Repeated measures of serum glucose and insulin in relation to postmenopausal breast cancer
Article first published online: 2 JUN 2009
Copyright © 2009 UICC
International Journal of Cancer
Volume 125, Issue 11, pages 2704–2710, 1 December 2009
How to Cite
Kabat, G. C., Kim, M., Caan, B. J., Chlebowski, R. T., Gunter, M. J., Ho, G. Y.F., Rodriguez, B. L., Shikany, J. M., Strickler, H. D., Vitolins, M. Z. and Rohan, T. E. (2009), Repeated measures of serum glucose and insulin in relation to postmenopausal breast cancer. Int. J. Cancer, 125: 2704–2710. doi: 10.1002/ijc.24609
- Issue published online: 24 SEP 2009
- Article first published online: 2 JUN 2009
- Accepted manuscript online: 2 JUN 2009 12:00AM EST
- Manuscript Accepted: 19 MAY 2009
- Manuscript Received: 2 APR 2009
- Institutional funds
- serum glucose;
- serum insulin;
- breast neoplasms;
- postmenopausal women
Experimental and epidemiological evidence suggests that circulating glucose and insulin may play a role in breast carcinogenesis. However, few cohort studies have examined breast cancer risk in association with glucose and insulin levels, and studies to date have had only baseline measurements of exposure. We conducted a longitudinal study of postmenopausal breast cancer risk using the 6% random sample of women in the Women's Health Initiative clinical trials whose fasting blood samples, provided at baseline and at years 1, 3 and 6, were analyzed for glucose and insulin. In addition, a 1% sample of women in the observational study, who had glucose and insulin measured in fasting blood samples drawn at baseline and in year 3, were included in the analysis. We used Cox proportional hazards models to estimate hazard ratios and 95% confidence intervals for the association of baseline and follow-up measurements of serum glucose and insulin with breast cancer risk. All statistical tests were 2-sided. Among 5,450 women with baseline serum glucose and insulin values, 190 incident cases of breast cancer were ascertained over a median of 8.0 years of follow-up. The highest tertile of baseline insulin, relative to the lowest, was associated with a 2-fold increase in risk in the total population (multivariable hazard ratio 2.22, 95% confidence interval 1.39–3.53) and with a 3-fold increase in risk in women who were not enrolled in the intervention arm of any clinical trial (multivariable hazard ratio 3.15, 95% confidence interval 1.61–6.17). Glucose levels showed no association with risk. Analysis of the repeated measurements supported the results of the baseline analysis. These data suggest that elevated serum insulin levels may be a risk factor for postmenopausal breast cancer. © 2009 UICC
Risk of postmenopausal breast cancer is increased in association with obesity and diabetes, both of which are characterized by increased insulin resistance, with consequent increases in circulating levels of insulin and glucose.1, 2 Insulin promotes cell proliferation3, 4 and enhances breast tumor growth in animal models,5, 6 whereas glucose may exacerbate insulin resistance,7 favor the selection of malignant clones8 and provide a growth advantage to cancer cells.9 Therefore, it is plausible that relatively high levels of glucose and/or insulin may play a role in the etiology of breast cancer. The few prospective studies to date that have directly investigated the association between fasting glucose and insulin levels and risk of incident postmenopausal breast cancer have yielded conflicting results.10–18 However, each of these studies was based on a single measurement of glucose and/or insulin at baseline only. Analysis of repeated measurements obtained during follow-up may provide greater insight into the role of these factors in the development of breast cancer and increase the precision of estimates of association. Therefore, we conducted a longitudinal study of breast cancer risk in which fasting serum glucose and insulin levels were measured at baseline, and at years 1, 3 and 6 of follow-up in a 6% sample of participants in the Women's Health Initiative clinical trial (WHI-CT) and at baseline and year 3 in a 1% sample of the observational study (WHI-OS).
Material and methods
The Women's Health Initiative is a large, prospective, multicenter study of factors affecting the health of postmenopausal women. It includes an observational study (N = 93,676) and 3 clinical trials (N = 68,132) of hormone therapy, dietary modification, and calcium plus vitamin D supplementation.19 Women were recruited at 40 clinical centers throughout the United States, largely via direct mailings, and were eligible to participate if they were postmenopausal, aged 50–79, likely to reside in their current residence for at least 3 years, and provided written informed consent. Enrollment took place between October 1, 1993 and December 31, 1998. The clinical trials had a number of additional eligibility requirements.20 In general, eligible women were first invited to enroll in the clinical trial component. Women who did not wish to be randomly assigned to an intervention or who were ineligible for the clinical trial component were then invited to participate in the observational study.
The present analysis is based on a 6% random sample of women in the clinical trials (N = 4,396) who provided fasting blood samples at baseline and at years 1, 3 and 6 of follow-up and a 1% sample of women in the observational study (N = 1,054) who provided fasting blood samples at baseline and at year 3.21 The 6% random sample was stratified by age, clinical center and hysterectomy status, with over-sampling of minority groups to increase the numbers of Black, Hispanic and Asian-Pacific women. Approval for the WHI was obtained from Institutional Review Boards at all clinical centers. All participants signed informed consent forms. All protocols and procedures were approved by Institutional Review Boards at participating institutions.
In the clinical trial, cancer outcomes were ascertained through semiannual, self-administered questionnaires, and then confirmed by centralized review of pathology reports, discharge summaries, operative and radiology reports, and tumor registry abstracts. In the observational study, ascertainment of outcomes was performed on an annual basis.
Fasting bloods were collected with minimal stasis and maintained at 4°C until plasma/serum was separated. Plasma/serum aliquots were then frozen at −70°C and sent on dry ice to the central repository (Fisher BioServices, Rockville, MD), where storage at −70°C was maintained. Serum glucose was measured using the hexokinase method on the Hitachi 747 (Boehringer Mannheim Diagnostics, Indianapolis, IN).22, 23 Monthly interassay coefficients of variation were <2% for mean concentrations of 84 and 301 mg/dL. Serum insulin was measured in a stepwise sandwich ELISA procedure on an ES 300 (BMD, Indianapolis, IN).24 Monthly interassay coefficients of variation were 4.7%–9.5% and 3.2%–7.9% at mean concentrations of 26.6 and 80.6 μIU/mL, respectively.
Cox proportional hazards models were used to estimate hazard ratios and 95% confidence intervals (95% CI) for the associations of serum glucose and insulin with risk of breast cancer, with duration of follow-up (days) as the time scale. For these analyses, study participants were considered to be at risk from their date of enrollment until the date of diagnosis of their breast cancer, termination of follow-up (September 12, 2005), loss to follow-up, withdrawal from the study, or death, whichever occurred first. Event times of participants who had not developed breast cancer by the end of follow-up, who had died, or who had withdrawn from the study before the end of follow-up, were censored.
In the first stage of the analysis, we estimated the risk of breast cancer in association with baseline glucose and insulin levels, and with a measure of insulin resistance, the homeostasis model assessment-insulin resistance (HOMA-IR) index ([fasting insulin (μIU/mL) × fasting glucose (mg/dL)]/22.5).25 Tertiles of the 3 study variables were created based on their distribution in the total study population. Established risk factors and potential confounding variables included in multivariable analyses were: age (continuous), education (less than high school graduate, high school graduate/some college, college graduate, post-college), ethnicity (white, black, other), body mass index (kg/m2—continuous), waist circumference (continuous), oral contraceptive use (ever, never), hormone therapy (ever, never), age at menarche (continuous), age at first birth (<20, 20–29, ≥30, missing), age at menopause (<50, ≥50, missing), alcohol (servings per week—continuous), family history of breast cancer (yes, no), history of breast biopsy (ever, never), physical activity (METs—continuous), energy intake (continuous), randomization status (for women in the clinical trial) in hormone therapy, calcium plus vitamin D, and dietary modification trials, and diabetes (yes, no). Women were classified as having diabetes if they reported taking diabetes medication or had a fasting glucose level of ≥126 mg/dL. Tests for trend were performed by assigning the median value to each category and modeling this variable as a continuous variable. Given that intervention status may have affected the postbaseline measurement of glucose and insulin and may also have influenced the risk of breast cancer, analyses were carried out in the total study population and in women who were not randomized to any intervention (control group in the dietary modification and placebo groups in the hormone therapy and calcium plus vitamin D trials, plus women in the observational study).
Additional analyses examined the associations for glucose and insulin by age (<65, ≥65), by ethnicity (white, African-American) and by estrogen receptor status. Estrogen receptor status was available for 118 invasive cases (102 ER+, 16 ER−).
Because glucose and insulin levels vary by body mass index26 and by hormone therapy use,18, 27 we performed stratified analyses to determine the association of glucose, insulin and HOMA-IR (categorized into tertiles) with breast cancer by levels of these exposures. We formally tested for interaction between glucose/insulin/HOMA-IR and hormone therapy/body mass index by comparing models with and without the product terms representing the variables of interest with a likelihood ratio test. Hormone therapy was categorized as never, past and current use; body mass index was categorized as <25, 25–<30, 30+ kg/m2.
Analysis of repeated measures
In the second stage of the analysis, the repeated measurements of the different biomarkers were analyzed by modeling them as time-dependent covariates in Cox proportional hazards models.28 With this approach, we evaluated the predictive value of the most recent measurement, measurements obtained 1–3 years, 2–4 years and 3–5 years before the date of diagnosis of breast cancer, and the mean of all available measurements. Among cases, measurements that were obtained within 1 year of diagnosis were excluded from all analyses, since these values may have been influenced by the presence of subclinical disease. The time-dependent analyses were not adjusted for multiple testing, because the results were used primarily to confirm trends observed with the baseline data. All p-values were 2-sided.
During a median follow-up period of 8.0 years, a total of 190 breast cancer cases (153 invasive and 37 in situ) were ascertained. Approximately two-thirds of the women (136 cases and 3,637 noncases) were not in any of the clinical trial intervention groups or were in the observational study.
Cases and noncases were similar with respect to age, anthropometric and reproductive/menstrual variables (Table I). Compared to noncases, cases were significantly more likely to be non-Hispanic white and had significantly lower levels of physical activity compared to noncases.
|Cases (n = 190)||Noncases (n = 5,260)||p-value|
|Age1||62.7 ± 7.2||62.6 ± 6.6||0.55|
|Body mass index (kg/m2)1||29.1 ± 5.7||29.1 ± 6.1||0.86|
|Height (cm)1||161.4 ± 6.3||161.0 ± 6.7||0.77|
|Waist circumference (cm)1||88.5 ± 12.9||88.6 ± 14.0||0.88|
|Waist-to-hip ratio1||0.82 ± 0.07||0.82 ± 0.08||0.48|
|Parity1||2.6 ± 1.7||2.7 ± 1.7||0.26|
|Age at menopause1||47.6 ± 6.5||46.7 ± 6.8||0.14|
|Alcohol (servings/week)1||1.6 ± 3.2||1.7 ± 4.0||0.81|
|Physical activity (METs/week2)1||7.3 ± 10.4||10.0 ± 12.9||0.0002|
|Oral contraceptive use (% ever)||37.1||41.7||0.19|
|Hormone therapy use (% ever)||50.3||46.6||0.31|
|Age at menarche (% ≤12 yrs)||51.0||46.9||0.38|
|Age at first birth (% ≥30 yrs)||10.1||9.0||0.27|
|Breast cancer in a first-degree family member (% yes)||16.2||15.4||0.74|
|Diabetes (% yes)||9.6||8.1||0.44|
|Education (% some post-college)||27.0||23.9||0.72|
|Ethnicity (% non-Hispanic white)||61.9||51.8||0.02|
|Smoking (% current smokers)||6.7||8.4||0.65|
Baseline glucose and insulin were modestly correlated (Pearson correlation coefficient 0.31, p < 0.0001). Baseline serum glucose showed a moderately strong association with HOMA-IR (correlation coefficient 0.49, p < 0.0001), whereas insulin was strongly correlated with HOMA-IR (correlation coefficient 0.92, p < 0.0001).
Mean serum glucose and insulin levels were higher in cases compared to noncases at baseline and at years 1 and 3, but the differences were not statistically significant (Table II). In year 6, mean glucose was similar in cases and noncases, whereas mean insulin was lower in cases compared to noncases. Although postbaseline levels of glucose were correlated with baseline levels, the strength of the correlation decreased with increasing interval between measurements, and similarly for insulin (data not shown).
|02||104.7 ± 33.1 (n = 190)||101.5 ± 30.2 (n = 5,260)||0.15|
|1||103.0 ± 31.8 (n = 134)||101.7 ± 30.3 (n = 3,871)||0.63|
|3||103.3 ± 24.8 (n = 96)||101.2 ± 29.5 (n = 3,466)||0.43|
|6||103.4 ± 24.4 (n = 43)||103.5 ± 28.6 (n = 3,172)||0.98|
|02||12.9 ± 7.9 (n = 184)||11.9 ± 9.7 (n = 5,112)||0.08|
|1||14.4 ± 23.3 (n = 132)||11.9 ± 8.5 (n = 3,765)||0.21|
|3||14.5 ± 8.9 (n = 90)||13.1 ± 10.9 (n = 3,282)||0.16|
|6||10.2 ± 5.5 (n = 42)||11.6 ± 34.1 (n = 3,170)||0.19|
Baseline glucose levels were not associated with breast cancer risk. In contrast, baseline insulin levels and HOMA-IR were both positively associated with breast cancer risk, with statistically significant linear trends over increasing tertiles (Table III). The association was stronger in multivariable-adjusted models than in age-adjusted models. Addition of BMI, waist circumference and ethnicity to the model each contributed to the increased hazard ratios. The association of insulin and breast cancer was stronger in women who had not undergone any intervention. For all participants, the multivariable hazard ratio for the highest tertile of serum insulin compared to the lowest was 2.22 (95% CI 1.39–3.53). For women who were not in the intervention arm of any of the WHI trials, the multivariable hazard ratio for the highest tertile was 3.15 (95% CI 1.61–6.17). Corresponding hazard ratios for HOMA-IR were 2.03 (95% CI 1.26–3.25) and 2.99 (95% CI 1.56–5.73), respectively. Hazard ratios were similar for all cancers (invasive + in situ) and invasive cancer. When insulin and glucose were included in the model simultaneously, the risk estimates for insulin were unchanged, whereas those for glucose decreased toward the null. Inclusion of additional lifestyle variables (smoking status and intakes of fat, fiber and vegetables) in the models did not materially affect the results (data not shown).
|Analyte||All trial arms||No intervention|
|All cases (N = 190)||All cases (N = 190)||Invasive cases (N = 157)||All cases (N = 99)||All cases (N =99)||Invasive cases (N =79)|
|HR (95% CI)||HR (95% CI)||HR (95% CI)||HR (95% CI)||HR (95% CI)||HR (95% CI)|
|<89.5 mg/dL||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref)||1.00 (ref.)||1.00 (95% CI)|
|89.5–<99.5 mg/dL||1.41 (0.98–2.02)||1.31 (0.88–1.94)||1.27 (0.81–1.97)||1.24 (0.76–2.03)||1.21 (0.73–1.99)||1.07 (0.58–1.95)|
|≥99.5 mg/dL||1.37 (0.94–1.98)||1.41 (0.90–2.22)||1.41 (0.85–2.33)||1.26 (0.75–2.10)||1.14 (0.60–2.16)||1.28 (0.61–2.66)|
|<7.95 μIU/mL||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
|7.95–<12.55 μIU/mL||1.50 (1.03–2.20)||1.76 (1.15–2.70)||1.80 (1.12–2.88)||1.91 (1.09–3.35)||2.33 (1.26–4.32)||2.39 (1.21–4.73)|
|≥12.55 μIU/mL||1.71 (1.18–2.48)||2.22 (1.39–3.53)||2.07 (1.23–3.47)||2.47 (1.43–4.26)||3.15 (1.61–6.17)||2.74 (1.29–5.83)|
|<32.9||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
|32.9–<56.06||1.47 (1.00–2.12)||1.57 (1.03–2.39)||1.76 (1.10–2.81)||2.08 (1.19–3.62)||2.33 (1.31–4.13)||2.88 (1.44–5.75)|
|≥56.06||1.62 (1.12–2.34)||2.03 (1.26–3.25)||1.89 (1.11–3.24)||2.38 (1.37–4.11)||2.99 (1.56–5.73)||2.86 (1.28–6.39)|
Exclusion of cases diagnosed within the first 2 years of follow-up did not alter the results (data not shown). Furthermore, exclusion of women with diabetes did not materially affect the results (data not shown).
Serum glucose was not significantly associated with risk either in women <65 years of age or in those aged ≥65 years; however, in both strata serum insulin showed significant associations (HRs for highest vs. lowest tertile 2.12, 95% CI 1.24–3.61 and 2.06, 95% CI 1.06–4.05, respectively). Glucose was not associated with risk in white women, but in African-American women HRs for the middle and highest tertiles were 2.95 (95% CI 1.05–8.29) and 1.98 (95% CI 0.63–6.19), respectively. Insulin was associated with risk in both whites and African-Americans: HR for highest tertile 2.19 (95% CI 1.24–3.82) and 2.61 (95% CI 0.82–8.33), respectively. Among estrogen receptor-positive cases (N = 102), neither glucose nor insulin showed a significant association with breast cancer; however, the HRs for insulin were below 1.0 (middle tertile 0.81, 95% CI 0.39–1.67, highest tertile 0.61, 95% CI 0.28–1.32), but not statistically significant. The small number of estrogen receptor-negative cases (N = 16) did not permit the use of Cox proportional hazards models.
Although the association of baseline glucose, insulin and HOMA-IR with breast cancer varied by hormone therapy user status, tests for interaction were not significant (Table IV). The association of insulin with breast cancer was strongest among lean women (BMI < 25 kg/m2) and weakest among obese women (BMI >= 30 kg/m2), though the test for interaction was not significant. HOMA-IR was also most strongly associated with risk in lean women, and the test for interaction between HOMA-IR and BMI was statistically significant.
|HR (95% CI)1||HR (95% CI)1||HR (95% CI)1|
|Hormone therapy status||Never user||Past user||Current user|
|(Ncases = 93)||(Ncases = 31)||(Ncases = 60)|
|<89.5 mg/dL||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
|89.5–<99.5 mg/dL||1.03 (0.61–1.72)||3.23 (1.02–10.26)||1.58 (0.91–2.75)|
|≥99.5 mg/dL||1.14 (0.67–1.97)||3.50 (1.04–11.76)||1.19 (0.56–2.53)|
|<7.95 μIU/mL||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
|7.95–<12.55 μIU/mL||1.63 (0.92–2.89)||1.77 (0.98–2.96)||1.83 (0.94–3.56)|
|≥12.55 μIU/mL||1.74 (0.92–3.30)||3.57 (1.21–10.54)||2.40 (1.14–5.05)|
|<32.9||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
|32.9–<56.06||1.15 (0.67–1.97)||2.85 (1.01–8.08)||1.51 (0.83–2.74)|
|≥56.06||1.49 (0.84–2.63)||4.06 (1.31–12.58)||1.69 (0.83–3.44)|
|Body mass index||<25||25–<30||≥30|
|(Ncases = 45)||(Ncases = 63)||(Ncases = 75)|
|<89.5 mg/dL||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
|89.5–<99.5 mg/dL||1.91 (1.01–3.59)||1.27 (0.70–2.33)||1.02 (0.56–1.86)|
|≥99.5 mg/dL||1.27 (0.53–3.08)||1.42 (0.71–2.84)||1.08 (0.59–1.95)|
|<7.95 μIU/mL||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
|7.95–<12.55 μIU/mL||1.35 (0.66–2.74)||2.03 (1.08–3.82)||1.39 (0.59–3.25)|
|≥12.55 μIU/mL||2.90 (1.31–6.43)||1.94 (0.94–4.00)||1.68 (0.73–3.86)|
|<32.9||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
|32.9–<56.06||1.81 (0.94–3.47)||1.89 (1.05–3.42)||0.57 (0.29–1.11)|
|≥56.06||2.96 (1.33–6.59)||1.49 (0.72–3.08)||0.98 (0.54–1.79)|
The time-dependent covariate analysis largely supported the results based on analysis of the baseline measures (Table V). Use of the most recent measurement (but excluding measurements obtained within 1 year of diagnosis) and of 3 different time windows prior to diagnosis (1–3 years, 2–4 years and 3–5 years—data shown only for the first time window) as the predictor variable yielded estimates for glucose and insulin that were similar in magnitude to those obtained in the baseline analysis. Use of the average of all available measurements (but excluding measurements made within 1 year of diagnosis) yielded somewhat stronger estimates for insulin and HOMA-IR. In the OS participants and those in the CT who did not receive an intervention, hazard ratios for the associations of average fasting insulin and HOMA-IR with breast cancer for the highest relative to the lowest tertile were 4.34, 95% CI 2.19–8.62 and 3.75, 95% CI 1.89–7.41, respectively (Table V). Although the most recent glucose measurement and the average of all glucose measurements were associated with breast cancer risk in all participants (Table V), these associations decreased toward the null and were no longer significant when insulin was included in the model.
|Analyte||Most recent measurement||Measurement 1–3 years prior||Average of all measurements|
|All participants||No intervention||All participants||No intervention||All participants||No intervention|
|HR (95% CI)3||HR (95% CI)||HR (95% CI)3||HR (95% CI)||HR (95% CI)3||HR (95% CI)|
|<89.5 mg/dL||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref)||1.00 (ref.)|
|89.5–<99.5 mg/dL||1.57 (1.09–2.27)||1.25 (0.74–2.09)||1.68 (1.08–2.63)||1.54 (0.77–3.08)||1.44 (0.99–2.10)||1.23 (0.73–2.07)|
|≥99.5 mg/dL||1.50 (1.00–2.25)||1.37 (0.77–2.44)||1.48 (0.90–2.45)||1.58 (0.73–3.43)||1.66 (1.10–2.52)||1.29 (0.71–2.37)|
|<7.95 μIU/mL||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
|7.95–<12.55 μIU/mL||1.91 (1.30–2.82)||2.32 (1.31–4.12)||2.04 (1.25–3.33)||2.27 (1.03–5.04)||2.29 (1.52–3.46)||3.50 (1.87–6.54)|
|≥12.55 μIU/mL||2.08 (1.36–3.18)||2.74 (1.47–5.10)||2.23 (1.30–3.82)||2.96 (1.27–6.92)||2.78 (1.75–4.40)||4.34 (2.19–8.62)|
|<32.9||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)||1.00 (ref.)|
|32.9–<56.06||2.21 (1.48–3.29)||2.63 (1.48–4.68)||2.29 (1.38–3.80)||3.22 (1.40–7.44)||2.11 (1.39–3.20)||2.78 (1.52–5.10)|
|≥56.06||2.49 (1.59–3.90)||2.64 (1.36–5.11)||2.56 (1.45–4.52)||3.28 (1.29–8.37)||3.02 (1.89–4.83)||3.75 (1.89–7.41)|
Our results suggest that baseline serum insulin levels are positively associated with risk of postmenopausal breast cancer. HOMA-IR, an indicator of insulin resistance, was associated with a comparable increase in risk, whereas baseline glucose was not associated with breast cancer risk. The positive associations of insulin and HOMA-IR with breast cancer risk were strengthened when the analysis was restricted to women who were not in the intervention group of any of the WHI clinical trials or who were in the observational study. Results of the repeated measures analysis largely supported those obtained from analysis of the baseline data.
The few previous cohort studies that have examined the association of fasting glucose and/or insulin levels with risk of breast cancer have yielded conflicting results. Three12, 14, 15 of five cohort studies10–15 found glucose to be a statistically significant predictor of increased breast cancer risk, but the associations differed by menopausal status. Muti et al.12 reported that in premenopausal women the highest quartile of serum glucose was associated with a 2.8-fold increase in risk (95% CI 1.2–6.5) compared to women in the lowest quartile; however, in postmenopausal women there was no significant association (relative risk 1.63, 95% CI 0.59–4.46). Stattin et al.15 found that fasting glucose was significantly associated with risk in women less than age 49 (relative risk for highest quartile 2.13, 95% CI 1.2–4.1), but not in women aged 49 or over (RR 0.90, 95% CI 0.68–1.2). In contrast, Rapp et al.14 observed an overall association of plasma glucose with breast cancer that was strongest in women older than 65 years. The remaining cohort studies showed no association between glucose levels and risk.10, 11, 13 Only 1 of the 4 cohort studies11, 16–18 that have examined fasting insulin levels in relation to breast cancer risk reported a significant positive association,18 while the remainder showed no association.11, 16, 17 Using data from the WHI Observational Study, Gunter et al.18 observed a positive association between baseline fasting serum insulin levels and breast cancer risk that was limited to women who were not currently using hormone therapy. Finally, in a large nested case-control study within the European Prospective Investigation into Cancer and Nutrition (EPIC)29 C-peptide levels, a marker of pancreatic insulin secretion, were not associated with increased risk of breast cancer overall, but were associated with increased risk among women above 60 years of age, albeit only among those women who had provided a blood sample under nonfasting conditions. With the exception of the study by Stattin et al.,15 in which a second measurement of plasma glucose was obtained 10 years postbaseline on a 15% subsample, none of the previous studies had repeated measures of insulin and/or glucose during the follow-up period.
In contrast to Gunter et al.,18 we found a significant positive association between fasting serum insulin and breast cancer risk regardless of hormone use status. The association was strongest in former users, followed by current users, followed by never users. However, the numbers of cases stratified by hormone use status were small, and the variation in risk by hormone therapy status was not statistically significant. Differences in the subgroup associations between the present study and that of Gunter et al.18 may be explained in part by differences in sample size between the 2 studies, as well as by differences in baseline characteristics of the CT and OS cohorts, reflecting differences in their inclusion criteria.
The suggestion in our data that insulin may be a risk factor for breast cancer in relatively lean women requires confirmation in larger studies. This finding is potentially important because it indicates that insulin may contribute to postmenopausal breast cancer risk, independent of obesity. Both baseline glucose and insulin showed a weak positive association with BMI (Pearson correlation coefficients 0.22 and 0.33, respectively). Furthermore, the association of insulin with breast cancer was only slightly altered when BMI was included in the multivariable model (HR for highest vs. lowest tertile 1.89, 95% CI 1.23–2.92) compared to the model without BMI (HR 1.76, 95% CI 1.17–2.64). This further suggests that the association of insulin with breast cancer risk may be independent of obesity.
Our finding of a differential association of insulin levels with estrogen receptor-positive and estrogen receptor-negative breast cancer warrants further study, as the number of estrogen-receptor-negative cases was extremely small. Gunter et al.18 found no difference in the association of insulin with breast cancer by estrogen receptor status.
HOMA-IR, based on a measurement of fasting insulin and glucose, has been shown to reflect insulin resistance assessed by euglycemic clamp more accurately than fasting insulin alone.30 Thus, the positive association of HOMA-IR with breast cancer risk adds to the evidence of a possible role of insulin resistance in the etiology of breast cancer.
The results of the repeated measures analysis supported the findings of the analyses using baseline values for glucose, insulin and HOMA-IR. Specifically, the association of insulin with breast cancer risk was evident using 3 different scenarios and was stronger when the average of all available measurements was used (Table V). The average value is a more precise estimate of a subject's “true” long-term insulin level than is a single measurement. For example, the intraclass correlation coefficient (ICC) of insulin estimated from the baseline, year 1, and year 3 measurements is 0.59; averaging the 3 repeated measures would increase the ICC to 0.81. Additional correction for nondifferential measurement error using statistical approaches would only further strengthen the observed associations, but not appreciably, given that the average value appears to be highly reliable.31 Moreover, the fact that when glucose and insulin were both in the model insulin remained unchanged but glucose was no longer significant suggests that insulin is the more important factor of the two, even though insulin was less reliable than glucose (ICCglucose = 0.89 for the average value).
Our results are consistent with a biologic role of insulin in breast carcinogenesis, which is supported by experimental studies. Insulin has mitogenic and antiapoptotic activity, induces transformation of normal breast epithelial cells, stimulates cell proliferation in both normal and neoplastic breast tissue,3, 4, 32 and may contribute to breast cancer risk by upregulating the secretion of ovarian hormones.33
The present study has several limitations. First, the number of breast cancer cases was relatively small, limiting our ability to assess variation in analyses by hormone therapy use and body mass index, and, particularly, to stratify by these 2 variables simultaneously. Second, the small number of in situ cases limited our ability to conduct separate analyses for this category of breast cancer. Nevertheless, insulin appeared to be associated with in situ breast cancer as well as with invasive cancer. Third, it is possible that the clinical trial interventions may have affected postbaseline measurements of glucose and insulin. Therefore, intervention status was taken into account in the analysis. To address this further, we carried out a sensitivity analysis by restricting attention to women who were not randomized to any intervention. The observed associations with insulin and HOMA-IR were stronger in this group than in the total population, suggesting that the weaker associations in the total study population may have been due to the effects of the interventions. Finally, although data on free or total estradiol were not available in our study, Gunter et al.18 found that adjustment for total estradiol did not affect their positive association with insulin. Nevertheless, the possibility of residual confounding by estradiol levels, as well as by BMI and hormone therapy, remains.
In conclusion, we found a substantial positive association between fasting serum insulin levels and risk of breast cancer among postmenopausal women in the WHI, even after taking into account the possibility of change in levels over time. Our results, the first from a repeated measures analysis, suggest that elevated insulin levels may contribute to breast cancer risk. The association of insulin with breast cancer requires confirmation in other studies. Current recommendations for reducing breast cancer risk in postmenopausal women by maintaining a healthy weight and engaging in physical activity34, 35 may have a beneficial impact on insulin levels as well as on estrogen levels.
This article was supported by institutional funds. The authors thank the WHI investigators and staff for their outstanding dedication and commitment. A list of key investigators involved in this research follows. All persons named as “Dr.” A full listing of WHI investigators can be found at the following website: http://www.whi.org.
Program Office: (National Heart, Lung, and Blood Institute, Bethesda, Maryland) Elizabeth Nabel, Jacques Rossouw, Shari Ludlam, Linda Pottern, Joan McGowan, Leslie Ford, and Nancy Geller.
Clinical Coordinating Center: (Fred Hutchinson Cancer Research Center, Seattle, WA) Ross Prentice, Garnet Anderson, Andrea LaCroix, Charles L. Kooperberg, Ruth E. Patterson, Anne McTiernan; (Wake Forest University School of Medicine, Winston-Salem, NC) Sally Shumaker; (Medical Research Labs, Highland Heights, KY) Evan Stein; (University of California at San Francisco, San Francisco, CA) Steven Cummings.
Clinical Centers: (Albert Einstein College of Medicine, Bronx, NY) Sylvia Wassertheil-Smoller; (Baylor College of Medicine, Houston, TX) Jennifer Hays; (Brigham and Women's Hospital, Harvard Medical School, Boston, MA) JoAnn Manson; (Brown University, Providence, RI) Annlouise R. Assaf; (Emory University, Atlanta, GA) Lawrence Phillips; (Fred Hutchinson Cancer Research Center, Seattle, WA) Shirley Beresford; (George Washington University Medical Center, Washington, DC) Judith Hsia; (Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center, Torrance, CA) Rowan Chlebowski; (Kaiser Permanente Center for Health Research, Portland, OR) Evelyn Whitlock; (Kaiser Permanente Division of Research, Oakland, CA) Bette Caan; (Medical College of Wisconsin, Milwaukee, WI) Jane Morley Kotchen; (MedStar Research Institute/Howard University, Washington, DC) Barbara V. Howard; (Northwestern University, Chicago/Evanston, IL) Linda Van Horn; (Rush Medical Center, Chicago, IL) Henry Black; (Stanford Prevention Research Center, Stanford, CA) Marcia L. Stefanick; (State University of New York at Stony Brook, Stony Brook, NY) Dorothy Lane; (The Ohio State University, Columbus, OH) Rebecca Jackson; (University of Alabama at Birmingham, Birmingham, AL) Cora E. Lewis; (University of Arizona, Tucson/Phoenix, AZ) Tamsen Bassford; (University at Buffalo, Buffalo, NY) Jean Wactawski-Wende; (University of California at Davis, Sacramento, CA) John Robbins; (University of California at Irvine, CA) F. Allan Hubbell; (University of California at Los Angeles, Los Angeles, CA) Howard Judd; (University of California at San Diego, LaJolla/Chula Vista, CA) Robert D. Langer; (University of Cincinnati, Cincinnati, OH) Margery Gass; (University of Florida, Gainesville/Jacksonville, FL) Marian Limacher; (University of Hawaii, Honolulu, HI) David Curb; (University of Iowa, Iowa City/Davenport, IA) Robert Wallace; (University of Massachusetts/Fallon Clinic, Worcester, MA) Judith Ockene; (University of Medicine and Dentistry of New Jersey, Newark, NJ) Norman Lasser; (University of Miami, Miami, FL) Mary Jo O'Sullivan; (University of Minnesota, Minneapolis, MN) Karen Margolis; (University of Nevada, Reno, NV) Robert Brunner; (University of North Carolina, Chapel Hill, NC) Gerardo Heiss; (University of Pittsburgh, Pittsburgh, PA) Lewis Kuller; (University of Tennessee, Memphis, TN) Karen C. Johnson; (University of Texas Health Science Center, San Antonio, TX) Robert Brzyski; (University of Wisconsin, Madison, WI) Gloria E. Sarto; (Wake Forest University School of Medicine, Winston-Salem, NC) Denise Bonds; (Wayne State University School of Medicine/Hutzel Hospital, Detroit, MI) Susan Hendrix.
- 1Nutrition, hormones, and breast cancer: is insulin the missing link? Cancer Causes Control 1996; 7: 605–25..
- 2Adiposity, type 2 diabetes and the metabolic syndrome in breast cancer. Obes Rev 2007; 8: 395–408., , .
- 3Effect of insulin on cell cycle progression in MCF-7 breast cancer cells. Direct and potentiating influence. J Biol Chem 2001; 276: 38023–8., , , , , .
- 4Mitogenic properties of insulin and insulin anologues mediated by the insulin receptor. Diabetologia 1997; 40 Suppl 2: S25–S31., , , , , , .
- 5Insulin receptor levels and magnitude of insulin-induced responses in 7,12-dimethylbenz(a) anthracene-induced mammary tumors in rats. Cancer Res 1981; 41: 826–9., .
- 6Role of hormones in the growth and regression of human breast cancer cell (MCF-7) transplanted into athymic nude mice. J Natl Cancer Inst 1981; 67: 51–6., .
- 7Discovery of a metabolic pathway mediating glucose induced desentitization of the glucose transport system. Role of hexosamine biosynthesis in the induction of insulin resistance. J Biol Chem 1991; 266: 4706–12., , .
- 8On the origin of cancer cells. Science 1956; 123: 309–14..
- 9Oncogenic alterations of metabolism. Trends Biochem Sci 1999; 24: 68–72., .
- 10Risk of breast cancer in relation to anthropometry, blood pressure, blood lipids and glucose metabolism: a prospective study within the Malmö Preventive Project. Eur J Cancer Prev 2001; 10: 33–42., , , .
- 11Serum insulin and glucose levels and breast cancer incidence: the Atherosclerosis Risk in Communities Study. Am J Epidemiol 2002; 156: 349–52., , , , .
- 12Fasting glucose is a risk factor for breast cancer: a prospective study. Cancer Epidemiol Biomarkers Prev 2002; 11: 1361–8., , , , , , , , , , .
- 13Fasting serum glucose level and cancer risk in Korean men and women. JAMA 2005; 293: 194–202., , , , , .
- 14Fasting blood glucose and cancer risk in a cohort of more than 140,000 adults in Austria. Diabetologia 2006; 49: 945–52., , , , , , , .
- 15Prospective study of hyperglycemia and cancer risk. Diabetes Care 2007; 30: 561–7., , , , , , , .
- 16Obesity, weight change, fasting insulin, proinsulin, C-peptide, and insulin-like growth factor-1 levels in women with and without breast cancer in the Rancho Bernardo Study. J Womens Health Gend Based Med 1999; 8: 1265–72., .
- 17Prospective study of IGF-I, IGF-binding proteins, and breast cancer risk, in Northern and Southern Sweden. Cancer Causes Control 2002; 13: 307–16., , , , , , , , , , .
- 18Insulin, insulin-like growth factor-I, and risk of breast cancer in postmenopausal women. J Natl Cancer Inst 2009; 101: 48–60., , , , , , , , , , , , et al.
- 19The Women's Health Initiative Study Group. Design of the Women's Health Initiative clinical trial and observational study. Control Clin Trials 1998; 19: 61–109.
- 20Ethnicity and breast cancer: factors influencing differences in incidence and outcome. J Natl Cancer Inst 2005; 97: 439–48., , , , , , , , , , , .
- 21Insulin resistance and weight gain in postmenopausal women of diverse ethnic groups. Int J Obes 2004; 28: 1039–47., , , , , , , , , .
- 22Determination of hexokinase and glucose-6-phosphate dehydrogenase. In: BergmeyerHU, ed. Methods of enzymatic analysis,2nd Engl. edn. New York, NY: Academic Press, 1974, 1196–1201., , , .
- 23Evaluation of the hexokinase-glucose-6-phosphate dehydrogenase method of determination of glucose in urine.Anal Biochem 1968; 23: 301–316., .
- 24Fundamentals of Clinical Chemistry, 3rd edn. Philadelphia, PA: WB Saunders Co, 1987. 544 p..
- 25Homeostasis model assessment: insulin resistance and β-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985; 287: 412–19., , , , , .
- 26High insulin levels in newly diagnosed breast cancer patients reflect underlying insulin resistance and are associated with components of the insulin resistance syndrome. Breast Cancer Res Treat 2008; 114: 517–25., , , , , , , .
- 27Glycemic effects of postmenopausal hormone therapy: the Heart and Estrogen/Progestin Replacement Study. Ann Int Med 2003; 138: 1–9., , , , , , , .
- 28Evaluating serial cancer marker studies in patients at risk of recurrent disease. Biometrics 1981; 37: 67–78..
- 29Serum C-peptide levels and breast cancer risk: results from the European Prospective Investigation into Cancer and Nutrition (EPIC). Int J Cancer 2006; 119: 659–67., , , , , , , , , , , , et al.
- 30Clinical significance of the insulin resistance index as assessed by homeostasis model assessment. Endocr J 2001; 48: 81–6., , , , .
- 31Covariate measurement errors and parameter estimation in a failure time regression model. Biometrika 1982; 69: 331–42..
- 32Insulin receptor expression and function in human breast cancer cell lines. Cancer Res 1992; 52: 3924–30., , , , , , , .
- 33Estrogen and insulin/IGF-1 cooperatively stimulate cell cycle progression in MCF-7 breast cancer cells through differential regulation of c-Myc and cyclin D1. Mol Cell Endocrinol 2005; 229: 161–73., , , , , .
- 34Effects of weight loss vs aerobic exercise training on risk factors for coronary disease in healthy, obese, middle-aged and older men: a randomized controlled trial. JAMA 1995; 274: 1915–21., , , , , .
- 35Insulin, physical activity, and caloric intake in postmenopausal women: breast cancer implications. J Clin Oncol 2004; 22: 4507–13., , , , , .