Insulin Resistance Contributes to Racial Disparities in Breast Cancer Prognosis in US Women

Emily J. Gallagher; Kezhen Fei; Sheldon M. Feldman; Elisa Port; Neil B. Friedman; Susan K. Boolbol; Brigid Killelea; Melissa Pilewskie; Lydia Choi; Tari King; Anupma Nayak; Rebeca Franco; Daliz Cruz; Irini M. Antoniou; Derek LeRoith; Nina A. Bickell


Breast Cancer Res. 2020;22(40) 

In This Article


Patient Accrual, Data Collection, and Laboratory Measurements

In this cross-sectional study, women were recruited shortly after the diagnosis of a new primary invasive breast cancer from ten US hospital sites, including five academic medical centers and five community hospitals in five states: New York, New Jersey, Connecticut, Maryland, and Michigan. A survey assessing breast cancer and metabolic syndrome risk factors, anthropometric measures, fasting blood, and tissue samples was collected. Institutional Review Board (IRB) approval was obtained from all participating sites. Recruitment began in March 2013. Women who were eligible for the study were aged 21 years or more and self-identified as being Black women (including Hispanic Black women) or White women (excluding Hispanic White women). Exclusion criteria included women with type 1 or type 2 diabetes being treated with oral or injectable medication; previous bariatric surgery; glucocorticoid treatment within 2 weeks of recruitment for blood testing, biopsy, or surgical resection; end-stage renal disease or hepatic cirrhosis; prior organ transplantation; and receipt of neoadjuvant chemo- or hormonal therapy for breast cancer prior to blood tests or tissue sampling. We excluded women with type 2 diabetes on oral or injectable medications in order to evaluate HOMA-IR in the absence of medication that could affect insulin sensitivity, secretion, or measurement of endogenous insulin levels; however, it led to a lower rate of eligible Black women who had higher rates of type 2 diabetes.

Clinical data recorded included self-reported smoking, alcohol intake, diet, physical activity, education, income, and health insurance. Breast cancer screening history was also recorded and was defined as inadequate if women between the ages of 50–74 years had not had a mammogram in the 2 years prior to the mammogram that led to the current diagnosis of breast cancer. Charlson Comorbidity Index was calculated.[18]

At the study visit, each participant had height (m) and weight (kg) measurements recorded from which body mass index (BMI) was calculated (kg/m2). Waist circumference (cm) was measured using the NHANES procedures.[19] Blood pressure was obtained using a clinical electronic blood pressure monitor. Venous blood was drawn after an overnight fast (minimum 8 h) for plasma glucose, serum insulin, C-peptide, and a lipid panel [total cholesterol, high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, and triglycerides (TG)]. Insulin resistance was calculated by the HOMA-IR equation: [fasting glucose (mg/dL) × fasting serum insulin (μU/mL)]/405.

Definitions of Obesity, Metabolic Syndrome, and Insulin Resistance

Obesity was defined as a BMI of ≥ 30 kg/m2, or by the Adult Treatment Panel (ATP) III waist circumference (WC) cutoff of ≥ 88 cm. The metabolic syndrome was defined as having three or more of the following five criteria: (1) WC ≥ 88 cm; (2) triglycerides ≥ 150 mg/dL, or on treatment for hypertriglyceridemia; (3) HDL < 50 mg/dL; (4) fasting glucose ≥ 100 mg/dL; (5) systolic blood pressure ≥ 130 mmHg or diastolic blood pressure ≥ 85 mmHg, or on treatment for hypertension.[20] Insulin resistance was defined as a HOMA-IR score of > 2.8, the upper quartile of the US population, reported by NHANES III.[21]

Breast Cancer Subtype, Stage, and Prognosis Determination

Clinical pathology reports were obtained from the patients' electronic medical records to classify breast cancers as ER positive, HER2 overexpressing, or TNBC. The Nottingham Prognostic Index (NPI) score was calculated as 0.2 × tumor size (cm) + lymph node (LN) stage (1: LN negative, 2: 1–3 positive LNs, 3: ≥ 4 positive LNs) + histological grade (1, well-differentiated; 2, moderately differentiated; 3, poorly differentiated).[22] Improved NPI (iNPI) was defined as previously described, adding one point for HER2 positivity, and subtracting one point for progesterone receptor (PR) positivity.[23] Tumor grade was defined by the Nottingham combined histological grade (NCHG), as recommended by the American Joint Committee on Cancer (AJCC) criteria.[24] Poor prognosis was defined as an NPI score of > 4.4, or an iNPI > 5.4.[25,26]

Immunohistochemistry Staining and Analysis

The IR and IGF-1R expression was evaluated by immunohistochemistry (IHC) in compliance with the REMARK guidelines.[27] IHC methods with antibody sources and concentrations are detailed in Supplementary file 1. The Allred scoring system was used to assess the intensity of cell staining, and the proportion of tumor stained positive for IR and IGF-1R,[28] as previously described.[29,30] As no standard cutoffs have been determined for IR and IGF-1R staining, 0–4 was considered "low" and > 5 was considered "high." For the IR/IGF-1R ratio, we assigned a score of > 1 if the level of IR expression was greater than IGF-1R, 1 if IR was equal to IGF-1R expression, and < 1 if IGF-1R was greater than IR expression.

Statistical Analysis

Basic statistics were used to describe patient characteristics. Means and standard deviations were presented for continuous variables, and frequencies and proportions were presented for categorical variables. Group comparisons used t tests on continuous, and χ 2 tests on categorical variables. The sample size was determined based on estimated rates of insulin resistance in White women (20%) and Black women (33%) and a main effect OR > 1.5 between insulin resistance and poor prognosis breast cancer between Black women and White women. We used Andrew Hay's INDIRECT macro to estimate the path coefficients in an adjusted mediator model and used bootstrapping to estimate confidence intervals for indirect effects of race on prognosis through HOMA-IR. It allowed for adjustment for the potential influence of covariates not proposed as being mediators in the model. SAS 9.4 software (SAS Institute, Cary, NC) was used for all statistical analyses. All tests were two-sided and statistical significance was set at 0.05 level.