Abstract
Purpose:
Inflammatory breast cancer (IBC) is a rare, poorly understood and aggressive tumor. We extended prior findings linking high body mass index (BMI) to substantial increased IBC risk by examining BMI associations before and after adjustment for well-characterized comorbidities using medical record data for diabetes, insulin resistance, and disturbances of cholesterol metabolism in a general community healthcare setting.
Methods:
We identified 247 incident IBC cases diagnosed at Kaiser Permanente Northern California between 2005–2017 and 2470 controls matched 10:1 on birth year and geographic area and with ≥13 months of continuous enrollment prior to diagnosis/index date. We assessed exposures from 6 years up to one year prior to the diagnosis/index date, using logistic regression to calculate odds ratios (ORs) with 95% confidence intervals (CIs).
Results:
Before adjustment for comorbidities, ORs (95% CIs) for BMI of 25-<30, 30-<35, and ≥ 35 compared to <25 kg/m2 were 1.5 (0.9–2.3), 2.0 (1.2–3.1), and 2.5 (1.4–4.4), respectively. After adjustment for pre-diabetes/diabetes, HDL-C and triglyceride levels, and dyslipidemia, corresponding ORs were 1.3 (0.8–2.1), 1.6 (0.9–2.9), and 1.9 (1.0–3.5). The OR for HDL-C levels <50 mg/dL compared to ≥65 mg/dL was 2.0 (1.2–3.3) in the adjusted model. In a separate model the OR for a triglyceride/HDL-C ratio ≥2.50 compared to <1.62 was 1.7 (1.1–2.8) after adjustment for BMI, pre-diabetes/diabetes, and dyslipidemia. Results did not differ significantly by estrogen receptor status.
Conclusions:
Obesity and measures of insulin resistance independently increased IBC risk as did obesity and low HDL-C levels. These findings, if confirmed, have implications for IBC prevention.
Keywords: Inflammatory Breast Cancer, Obesity, Insulin Resistance, High-density Lipoprotein Cholesterol, Diabetes, Pre-diabetes
Background
Inflammatory breast cancer (IBC) is a rare, poorly understood and particularly aggressive form of breast cancer. It is characterized by diffuse erythema, edema, and peau d’orange involving the majority of the breast, often with no underlying tumor mass [1–2]. IBC constitutes approximately 2% of breast cancer cases in the United States, but disproportionately accounts for 7% of breast cancer deaths [3].
IBC and obesity have been called “two of the most perplexing problems in breast cancer” [4]. Several studies have shown substantial increases in risk of IBC with high body mass index (BMI)/obesity [5–7], associations that are greater than for breast cancer in general and do not seem to vary by menopausal status [5]. In 2012, the U.S. National Cancer Institute introduced an initiative to determine how obesity, which has been linked to increase risk of breast cancer overall [8], promotes cancer development [9].
Several causal pathways have been proposed linking obesity to cancer, including estrogen synthesis in adipose tissue, the chronic inflammation associated with obesity, dysregulated insulin signaling and hyperinsulinemia [10], and perturbations in cholesterol levels such as high-density lipoprotein cholesterol (HDL-C) [11–12], which may play an important anti-oxidative and anti-inflammatory role [12]. Pre-diabetes and diabetes, associated with dysregulated insulin metabolism, and disturbances in lipid levels have been linked to small changes in breast cancer risk in general [12–17], whereas serum insulin levels have not shown consistent associations [18–19]. Diabetes and dyslipidemia were associated with a small increase and decrease in risk of IBC, respectively, in the Medicare population, but BMI could not be adequately accounted for because it was not routinely captured in the Medicare database [20].
Here, we extend prior findings by leveraging a large resource with well-characterized medical record exposure data for a rare tumor to evaluate the obesity/IBC association and several associated comorbidities, namely pre-diabetes and diabetes, a measure of insulin resistance, and disturbances of cholesterol metabolism, specifically dyslipidemia, and low HDL-C and high triglyceride levels, before and after mutual adjustment.
Methods
Study population and setting
The study was conducted at Kaiser Permanente Northern California (KPNC), a large integrated health care system with over 4 million members. KPNC is a participant in the Cancer Research Network, which established and maintains a Virtual Data Warehouse of data extracted from electronic health records and administrative databases. KPNC also is a participant in SUPREME – DM (SUrveillance, PREvention, and ManagEment of Diabetes Mellitus), a comprehensive, longitudinal clinical registry of insured patients with diabetes mellitus. The Virtual Data Warehouse was first available in 2000 [21]; implementation of KP HealthConnect™, the electronic health record in which weight was routinely recorded during healthcare encounters, began in 2005. KPNC has used electronic health records to conduct analyses on obesity and obesity-related health conditions, including diabetes and dyslipidemia and breast cancer risk [22].
Ascertainment of study subjects
We identified 247 females diagnosed with inflammatory breast cancer (Site and Morphology Histologic Type ICD-0–3 = 8530 or Extent of Disease – CS.CS extension (2004+) = 71, 73, 600, 710, 715, 720, 725, 730, 750, 780 or American Joint Committee on Cancer Tumor Node Metastasis = T4d) between 2005 and 2017 at KPNC with at least 13 months of continuous enrollment prior to the diagnosis date (a 92 day gap period was considered as continuous enrollment). Cases were identified through the KPNC Cancer Registry (KPNCCR), which reports to the National Cancer Institute Surveillance, Epidemiology, and End Results Program registries for the San Francisco Bay Area and Greater California. Those with a prior in situ or invasive breast cancer or cancer of unknown type were excluded. The index date for cases was the date of diagnosis in the KPNCCR. We identified 2470 female controls, which were matched 10:1 to cases on birth year and geographic area (facility of diagnosis for cases vs. primary care physician facility/medical center for controls). Controls had to be alive at the index date (defined as July 1st of the year of diagnosis of the matched case cancer) with no history of in situ or invasive breast cancer or unknown cancer type and have at least 13 months minimum continuous enrollment prior to the index date. Age categories for cases and controls based on full dates of birth and diagnosis/selection are shown in Table 1.
Table 1.
Number of inflammatory breast cancer cases and controls according to age, calendar period of diagnosis/selection, location, and other demographic factors before imputation
| Inflammatory Breast Cancer Cases - N = 247 N (%) | Controls N = 2470 N (%) | |
|---|---|---|
| Matching factors | ||
| Age groups | ||
| 25–34 | 10 (4.05) | 105 (4.25) |
| 35–39 | 10 (4.05) | 105 (4.25) |
| 40–44 | 14 (5.67) | 141 (5.71) |
| 45–49 | 21 (8.50) | 197 (7.98) |
| 50–54 | 30 (12.15) | 274 (11.09) |
| 55–59 | 31 (12.55) | 326 (13.20) |
| 60–64 | 34 (13.77) | 350 (14.17) |
| 65–69 | 30 (12.15) | 307 (12.43) |
| 70–74 | 20 (8.10) | 206 (8.34) |
| 75–79 | 16 (6.48) | 147 (5.95) |
| 80–84 | 17 (6.88) | 170 (6.88) |
| 85–95 | 14 (5.67) | 142 (5.75) |
| Year of diagnosis/selection | ||
| 2005–2007 | 64 (25.91) | 640 (25.91) |
| 2008–2010 | 71 (28.74) | 710 (28.74) |
| 2011–2013 | 49 (19.84) | 490 (19.84) |
| 2014–2017 | 63 (25.51) | 630 (25.51) |
| KPNC service area | ||
| Central Valley | 17 (6.88) | 170 (6.88) |
| Diablo | 20 (8.10) | 200 (8.10) |
| East Bay/Fresnoa | 21 (8.50) | 210 (8.50) |
| Greater San Francisco | 18 (7.29) | 180 (7.29) |
| Greater Southern Alameda | 17 (6.88) | 170 (6.88) |
| Marin/Sonoma | 33 (13.36) | 330 (13.36) |
| Napa/Solano/Redwood Citya | 23 (9.31) | 230 (9.31) |
| Roseville | 18 (7.29) | 180 (7.29) |
| Sacramento | 18 (7.29) | 180 (7.29) |
| San Jose | 18 (7.29) | 180 (7.29) |
| Santa Clara | 29 (11.74) | 290 (11.74) |
| South Sacramento | 15 (6.07) | 150 (6.07) |
| Other characteristics | ||
| Race | ||
| White non-Hispanic | 173 (70.04) | 1534 (62.11) |
| White Hispanic | 27 (10.93) | 121 (04.90) |
| Black | 26 (10.53) | 159 (06.44) |
| Asian | 21 (8.50) | 370 (14.98) |
| Other | 0 (0.00) | 286 (11.58) |
| Screening mammogram during exposure period | ||
| No | 116 (46.96) | 754 (30.53) |
| Yes | 131 (53.04) | 1716 (69.47) |
| Median household income by census tract | ||
| <$40,000 | 24 (9.72) | 223 (9.03) |
| $40,000-$59,999 | 47 (19.03) | 543 (21.98) |
| $60,000-$79,999 | 59 (23.89) | 632 (25.59) |
| $80,000-$99,999 | 53 (21.46) | 433 (17.53) |
| >$100,000 | 64 (25.91) | 638 (25.83) |
| Unknown | 0 (0.00) | 1 (0.00) |
| Mean (median) | $80,185 ($76,167) | $79,385 ($73,304) |
| Education (median percent) | ||
| Less than high school | 10.00 | 9.59 |
| High school | 20.00 | 20.00 |
| Some college | 30.00 | 31.00 |
| College graduate | 24.00 | 22.00 |
| Graduate level | 9.00 | 11.00 |
| KPNC Encountersb | ||
| Mean, Median (Min, Max) | 35.5, 22 (1, 268) | 37.5, 27 (1, 382) |
| Missing - N (%) | 15 (6.1) | 74 (3.0) |
Service areas of East Bay and Fresno were combined and Napa/Solano and Redwood City were combined for presentation in the table due to numbers of IBC cases < 10 in a service area.
Outpatient, emergency room, hospitalization in a six-month period.
Exposure ascertainment
We included exposures from 6 years up to one year prior to the index date for cases and controls. We collected the number of encounters (outpatient, emergency room, hospitalization) for each 6-month period over which exposures were ascertained. Electronic laboratory data, pharmacy data, and diagnosis data were used to obtain the following primary measures, based on algorithms previously developed at KPNC [22], when available.
BMI: Heights of less than 4 feet or 7 feet, 2 inches or greater were considered implausible and excluded from the analysis. Weight data below 30 or > 1000 or more pounds were excluded. BMI was calculated using the first plausible height and weight measurements in the electronic health record system, as weight (kg) divided by the square of height (m2) [22]. We additionally excluded BMI values below 15 or 55 or greater kg/m2 as implausible. The median number of months between BMI measurement and index date was 58 for both cases and controls.
Triglyceride levels (mg/dL): an average of available values during the exposure period (medians [ranges] of available values were 2 [1,29] and 2 [1,22] for cases and controls, respectively).
HDL-C levels (mg/dL): an average of available values during the exposure period (medians [ranges] of available values were 2 [1,20] and 2 [1,22]) for cases and controls, respectively).
Triglyceride/HDL-C ratio: ratio of the average values for HDL-C and triglycerides.
Type 2 diabetes mellitus: (1) elevated glucose levels (fasting glucose >126 mg/dL or random glucose ≥200 mg/dL) measured on two separate occasions; (2) elevated hemoglobin A1C (>7.0%); (3) two or more outpatient visits with International Classification of Disease, 9th modification (ICD) codes (250.XX) for diabetes mellitus or new ICD-10 codes that include type 2 diabetes in the definition; (4) one or more diabetes-related hospital discharge codes; or (5) one or more fills for oral or injected diabetes-specific medications, excluding metformin or any thiazolidinedione if no other criterion was met. Patients who were pregnant during the enrollment period were considered diabetic if they met the criterion outside of the pregnancy time period [22].
Type 1 diabetes: at least two mentions of ICD-9 or ICD-10 codes with type 1 diabetes in the definition.
Pre-diabetes: fasting plasma glucose level of 100 to 125 mg/dL (impaired fasting glucose), a 2-h plasma glucose level after a 75-g oral glucose tolerance test of 140 to 199 mg/dL (impaired glucose tolerance), or hemoglobin A1c(A1C) 5.7 to <6.5% (39 to < 48 mmol/mol) [23].
Dyslipidemia: (1) at least one prescription for lipid-lowering medication plus an outpatient diagnosis of dyslipidemia (ICD-9 272.0–272.4, ICD-10 E78.00, E78.01, E78.1, E78.2, E78.3, E78.4, E78.5); (2) at least two outpatient diagnoses of dyslipidemia; (3) at least one prescription for lipid-lowering medication plus at least one elevated low-density lipoprotein cholesterol (≥160 mg/dL); or (4) at least one outpatient diagnosis of dyslipidemia plus at least one elevated low-density lipoprotein cholesterol [22]. Familial hypercholesterolemia was also included.
We also defined other variables as follows:
Screening mammogram during the exposure period: CPT codes 07500, 07504, 07506, 76083, 76092, 77052, 77057, 77067, 79802, G0202.
We also included 2010 census block-level variables on the education level of the adult population (aged ≥25 years) and the median household income linked to member addresses via geocoding. These variables are derived from the American Community Survey 5-year Summary File for 2006–2010, which was conducted as part of the United States Census Bureau’s Decennial Census Program [24].
Statistical analysis
Information on BMI was missing for 27% of IBC cases and 24% of controls. Information on HDL-C and triglycerides was missing for 30 and 22% of cases and 27 and 19% of controls, respectively. Information on estrogen receptor (ER) status was missing for 2% of cases (Tables 2). We therefore imputed missing data using the sequential regression imputation method [25] implemented with IVEware (http://www.isr.umich.edu/src/smp/ive). We obtained five imputed data sets from the imputation models, which included the following variables without missing values: case-control status, age at diagnosis/selection, matching location, race/ethnicity, screening mammogram, type 2 diabetes, type 1 diabetes, pre-diabetes, and dyslipidemia. The models also included the following variables with missing values: estrogen receptor (ER)-status for cases, number of encounters, BMI, HDL-C levels, and triglyceride levels. The imputation model included interaction terms between case-status and BMI, HDL-C, and triglycerides, and age interactions for HDL-C and triglycerides. Missing values for ER were only imputed for the IBC cases. For all statistical analyses, each of the five imputed datasets was first analyzed separately, and then the results were combined using SAS version 9.3 Proc MIANALYZE.
Table 2.
Number (percentage) of inflammatory breast cancer (IBC) cases and controls by exposure variable categories before imputation
| Variable | Total |
Age ≥50 years |
Estrogen Receptor (ER) -IBC cases |
|||
|---|---|---|---|---|---|---|
| IBC Cases N = 247 N (%) | Controls N = 2470 N (%) | IBC Cases N = 192 N (%) | Controls N = 1922 N (%) | ER+ N = 140 N (%) | ER− N = 101 N (%) | |
| Body mass index | ||||||
| < 25 | 24 (9.72) | 488 (19.76) | 14 (7.29) | 356 (18.52) | 11 (7.86) | 13 (12.87) |
| 25-<30 | 62 (25.10) | 707 (28.62) | 48 (25.00) | 578 (30.07) | 36 (25.71) | 25 (24.75) |
| 30-<35 | 52 (21.05) | 425 (17.21) | 41 (21.35) | 343 (17.85) | 30 (21.43) | 22 (21.78) |
| ≥ 35 | 41 (16.60) | 252 (10.20) | 34 (17.71) | 200 (10.41) | 26 (18.57) | 15 (14.85) |
| Missing | 68 (27.53) | 598 (24.21) | 55 (28.65) | 445 (23.15) | 37 (26.43) | 26 (25.74) |
| Pre-diabetes or diabetes | ||||||
| No | 131 (53.04) | 1405 (56.88) | 91 (47.40) | 940 (48.91) | 72 (51.43) | 55 (54.46) |
| Yes | 116 (46.96) | 1065 (43.12) | 101 (52.60) | 982 (51.09) | 68 (48.57) | 46 (45.54) |
| Triglyceride/HDL-C ratio | ||||||
| Low-1.61 | 41 (16.60) | 659 (26.68) | 32 (16.67) | 543 (28.25) | 20 (14.29) | 20 (19.80) |
| 1.62–2.49 | 40 (16.19) | 529 (21.42) | 33 (17.19) | 449 (23.36) | 22 (15.71) | 17 (16.83) |
| ≥2.50 | 92 (37.25) | 727 (29.43) | 81 (42.19) | 626 (32.57) | 56 (40.00) | 34 (33.66) |
| Missing | 74 (29.96) | 555 (22.47) | 46 (23.96) | 304 (15.82) | 42 (30.00) | 30 (29.70) |
| Dyslipidemia | ||||||
| No | 159 (64.37) | 1543 (62.47) | 109 (56.77) | 1035 (53.85) | 99 (70.71) | 57 (56.44) |
| Yes | 88 (35.63) | 927 (37.53) | 83 (43.23) | 887 (46.15) | 41 (29.29) | 44 (43.56) |
| HDL-C (mg/dL) | ||||||
| ≥ 65.00 | 26 (10.53) | 614 (24.86) | 23 (11.98) | 544 (28.30) | 16 (11.43) | 9 (08.91) |
| 50.00–64.99 | 77 (31.17) | 781 (31.62) | 66 (34.38) | 658 (34.24) | 37 (26.43) | 38 (37.62) |
| < 50.00 | 77 (31.17) | 607 (24.57) | 62 (32.29) | 472 (24.56) | 50 (35.71) | 26 (25.74) |
| Missing | 67 (27.13) | 468 (18.95) | 41 (21.35) | 248 (12.90) | 37 (26.43) | 28 (27.72) |
| Triglycerides (mg/dL) | ||||||
| <100.00 | 46 (18.62) | 691 (27.98) | 33 (17.19) | 551 (28.67) | 24 (17.14) | 21 (20.79) |
| 100.00–149.99 | 58 (23.48) | 648 (26.23) | 50 (26.04) | 556 (28.93) | 35 (25.00) | 21 (20.79) |
| ≥150.00 | 69 (27.94) | 576 (23.32) | 63 (32.81) | 511 (26.59) | 39 (27.86) | 29 (28.71) |
| Missing | 74 (29.96) | 555 (22.47) | 46 (23.96) | 304 (15.82) | 42 (30.00) | 30 (29.70) |
HDL-C high-density lipoprotein cholesterol.
We used a general linear model among the controls (Proc GLM) to calculate least square means and 95% confidence intervals for some risk factors by others. We also calculated Pearson correlation coefficients among controls using Proc CORR. We calculated odds ratios (ORs) and 95% confidence intervals (CIs) for risk factors in relationship to IBC compared with control subjects using unconditional logistic regression (Proc LOGISTIC), with the matching factors of age (with the categories shown in Table 1 treated as an ordinal trend) and location (with original categories before they were collapsed for presentation as shown in Table 1) included in the models. Using unconditional rather than conditional logistic regression allowed us to use polytomous logistic regression for analyses that separated cases according to tumor ER-status. We evaluated whether ORs for ER+ and ER- IBC compared to controls were statistically significantly different by examining p-values from linear hypothesis tests from the TEST statement in Proc MIANALYZE. We used a two-sided p-value of 0.05 to determine statistical significance.
We included all cases and controls in the logistic regression analyses. We did a sub-analysis among women aged ≥ 50 years as a surrogate for post-menopausal status; there were too few IBC cases aged < 50 years (N = 55) to do a separate analysis among them. To assess the robustness of our methods, we also conducted two sensitivity analyses evaluating the association between BMI and IBC risk: 1) unconditional logistic regression using non-imputed data with unknowns included in a separate category for each variable, and; 2) conditional logistic regression using imputed data to accommodate the matched design more fully.
For purposes of analysis, we consolidated pre-diabetes and diabetes, either type 1 or type 2, into one variable as a measure of dysregulated insulin levels [26–29]. Of IBC cases with either pre-diabetes or diabetes, 64% had pre-diabetes and 36% type 2 diabetes (of whom 21% also had an indication of type 1 diabetes). Among controls, 66% had pre-diabetes, 34% type 2 diabetes (of whom 7% also had an indication of type 1 diabetes); only 2 had an indication of type 1 diabetes without an indication of type 2 diabetes. We also used a triglyceride/HDL-C ratio ≥ 2.5 as a marker of insulin resistance and possible hyperinsulinemia [30]. In analyses of HDL-C and triglycerides separate from their ratio, we considered HDL-C levels < 50 mg/dL and triglyceride levels ≥ 150 mg/dL to be indicative of metabolic abnormalities [30]. We classified Asians and other races together in the analyses.
Results
Analyses before imputation
The mean age of the 247 IBC cases and 2470 controls was 60.7 years (range 28 to 95 years). The years of diagnosis/selection ranged from 2005 to 2017 over 14 service areas, several of which were combined for presentation due to small numbers of cases (Table 1). A higher percentage of controls than cases were classified as “other race”. A lower percentage of cases had a screening mammogram during the exposure period (53 vs. 69%). Median household income and education based on census data were similar among cases and controls; therefore, these variables were not included in subsequent analyses. The mean number of encounters in the Kaiser healthcare plan was similar in cases and controls, but the range was greater in controls than cases. Five percent of cases had 2 or fewer encounters and five percent of controls had four or fewer encounters. The mean numbers of encounters for study subjects with and without BMI measurements were 38.0 and 34.8, respectively.
Numbers of cases and controls in total, aged ≥ 50 years, and by case ER status are shown by exposure variable categories before imputation in Table 2. Seventy-eight percent of IBC cases were ≥ 50 years of age at diagnosis. Fifty-eight percent of IBC cases with known ER-status were ER-positive (+) (N = 140) and 42% ER-negative (−) (N = 101). The percentages of cases with unknown BMI, HDL-C, and triglycerides data were greater than for controls.
Analyses with imputed data
Among controls, least square means and median levels of BMI and triglycerides were higher in those with pre-diabetes/diabetes, and dyslipidemia than without (Table 3), whereas mean and median levels of HDL-C were lower in those with these conditions. BMI was negatively correlated with HDL-C (r = −.32) and positively correlated with triglycerides (r = .21), and the triglyceride/HDL-C ratio (r = .25). HDL-C and triglycerides were negatively correlated (r = −.40).
Table 3.
Associations among the exposure variables in the controls after imputation
| Least square mean (95% CI)a | ||||
|---|---|---|---|---|
| Body mass index | HDL-C | Triglycerides | Triglyceride/HDL-C ratio | |
| Pre-diabetes or diabetes | ||||
| No | 27.8 (27.3, 28.2) | 61.3 (60.4, 62.1) | 118.4 (112.9, 123.9) | 2.1 (2.0, 2.3) |
| Yes | 30.3 (30.0, 30.7) | 54.8 (53.9, 55.7) | 151.5 (147.2, 155.8) | 3.1 (3.0, 3.2) |
| Dyslipidemia | ||||
| No | 28.2 (27.8, 28.6) | 60.9 (60.1, 61.7) | 118.5 (113.4, 123.6) | 2.2 (2.0, 2.3) |
| Yes | 30.0 (29.6, 30.3) | 54.4 (53.4, 55.3) | 156.7 (152.2, 161.2) | 3.2 (3.1, 3.3) |
| Pearson Correlation Coefficients (95% CI) | ||||
| Body mass index | HDL-C | Triglycerides | Triglyceride/HDL-C ratio | |
| Body mass index | 1.0 | −0.32 (−0.35, −0.28) | 0.21 (0.16, 0.26) | 0.25 (0.21, 0.30) |
| HDL-C | - | 1.0 | −0.40 (−0.44, −0.37) | −0.60 (−0.62, −0.57) |
| Triglycerides | - | - | 1.0 | 0.93 (0.92, 0.94) |
| Triglyceride/HDL-C ratio | - | - | - | 1.0 |
HDL-C high-density lipoprotein cholesterol.
From general linear model adjusted for age categories shown in Table 1 treated as an ordinal trend.
After adjustment for matching variables, race, number of encounters, and mammographic screening, increasing BMI was associated with increased IBC risk (e.g., OR = 2.5; 95% CI 1.4–4.4 for BMI ≥35 compared to BMI < 25; OR = 1.3; 95% CI 1.1–1.5 for each 5 BMI unit increase) (Table 4). IBC risk was also increased with pre-diabetes or diabetes (OR = 1.4; 95% CI 1.1–1.9), the triglyceride/HDL-ratio ≥ 2.5 (OR = 2.0; 95% CI 1.3–3.1), HDL-C levels < 50 mg/dL (OR = 2.6; 95% CI 1.7–4.0), and triglyceride levels ≥ 150 mg/dL (OR = 1.8; 95% CI 1.2– 2.6); none of these results differed statistically significantly by case ER-status. Associations were generally slightly stronger when analyses were limited to women ≥ 50 years of age. Dyslipidemia was not associated with risk overall, but the heterogeneity in the risk associations for ER+ (OR = 0.7; 95% CI 0.5–1.1) versus ER- IBC (OR = 1.7; 95% CI 1.0–2.8) was statistically significant (p-het = .007).
Table 4.
ORs (95% CIs) for the relation of body mass index and other exposure variables with all inflammatory breast cancer (IBC) cases, those ≥ 50 years at diagnosis, and estrogen receptor (ER) status (positive (+) and negative (-)) cases with imputed data adjusted for matching and other demographic variables
| Variable | Total ORs (95% CI)a | Age ≥50 Years ORs (95% CI)a | ER+ IBC ORs (95% CI)a | ER- IBC ORs (95% CI)a | P-value for comparing ER+ to ER− |
|---|---|---|---|---|---|
| Body mass index | |||||
| < 25 | 1.0 | 1.0 | 1.0 | 1.0 | |
| 25-<30 | 1.5 (0.9–2.3) | 1.6 (0.9–2.8) | 1.7 (0.8–3.6) | 1.3 (0.6–2.7) | |
| 30-<35 | 2.0 (1.2–3.4) | 2.4 (1.3–4.4) | 2.4 (1.1–5.5) | 1.6 (0.8–2.9) | |
| ≥ 35 | 2.5 (1.4–4.4) | 2.9 (1.4–6.0) | 2.9 (1.1–7.9) | 1.8 (0.9–3.7) | |
| Increase per 5 BMI unit | 1.3 (1.1–1.5) | 1.3 (1.1–1.6) | 1.3 (1.1–1.6) | 1.2 (1.0–1.4) | .37 |
| Pre-diabetes or diabetes | |||||
| No | 1.0 | 1.0 | 1.0 | 1.0 | |
| Yes | 1.4 (1.1–1.9) | 1.3 (1.0–1.8) | 1.5 (1.0–2.2) | 1.3 (0.8–2.0) | .53 |
| Triglyceride/HDL-C ratio | |||||
| Low-1.61 | 1.0 | 1.0 | 1.0 | 1.0 | |
| 1.62–2.49 | 1.3 (0.8–2.0) | 1.4 (0.8–2.4) | 1.5 (0.8–2.8) | 1.0 (0.5–2.0) | |
| ≥2.50 | 2.0 (1.3–3.1) | 2.3 (1.4–3.6) | 2.4 (1.5–4.1) | 1.5 (0.8–2.7) | |
| Categorical variable as continuous | 1.4 (1.2–1.8) | 1.5 (1.2–1.9) | 1.6 (1.2–2.1) | 1.2 (0.9–1.7) | .25 |
| Dyslipidemia | |||||
| No | 1.0 | 1.0 | 1.0 | 1.0 | |
| Yes | 1.1 (0.8–1.5) | 1.1 (0.8–1.5) | 0.7 (0.5–1.1) | 1.7 (1.0–2.8) | .01 |
| HDL-C (mg/dL) | |||||
| ≥ 65.00 | 1.0 | 1.0 | 1.0 | 1.0 | |
| 50.00–64.99 | 2.0 (1.3–3.0) | 2.1 (1.4–3.3) | 1.8 (1.0–3.1) | 2.2 (1.1–4.6) | |
| < 50.00 | 2.6 (1.7–4.0) | 2.7 (1.7–4.4) | 2.9 (1.6–5.2) | 2.2 (1.1–4.5) | |
| Categorical variable as continuous | 1.6 (1.3–1.9) | 1.6 (1.3–2.0) | 1.7 (1.3–2.2) | 1.4 (1.0–1.9) | .58 |
| Triglycerides (mg/dL) | |||||
| <100.00 | 1.0 | 1.0 | 1.0 | 1.0 | |
| 100.00–149.99 | 1.4 (0.9–2.0) | 1.6 (1.0–2.5) | 1.6 (0.9–2.8) | 1.1 (0.5–2.4) | |
| ≥150.00 | 1.8 (1.2–2.6) | 2.1 (1.4–3.3) | 1.9 (1.2–3.1) | 1.6 (0.9–3.1) | |
| Categorical variable as continuous | 1.3 (1.1–1.6) | 1.4 (1.2–1.8) | 1.4 (1.1–1.7) | 1.3 (0.9–1.8) | .78 |
OR Odds ratio, CI Confidence interval, IBC Inflammatory breast cancer, ER Estrogen receptor, BMI Body mass index, HDL-C High-density lipoprotein cholesterol
Adjusted for age categories treated as an ordinal trend, service area, number of encounters, race, mammographic screening.
ORs for each exposure variable additionally adjusted for each other exposure variable in relation to total IBC are shown in Table 5. ORs were slightly attenuated, but generally robust to adjustment by other risk factors, except for the association with triglycerides, which was substantially attenuated by adjustment for HDL-C.
Table 5.
ORs (95% CIs) for the relation of exposure variables with all inflammatory breast cancer (IBC) adjusted individually for each other exposure variable as well as for matching and other demographic factors with imputed data
| Adjustment Variables | ||||||
|---|---|---|---|---|---|---|
| Body mass index ORa (95% CI) |
Pre-diabetes or diabetes ORa (95% CI) |
Triglyceride/HDL-C ratio ORa (95% CI) |
Dyslipidemia ORa (95% CI) |
HDL-C (mg/dL) ORa (95% CI) |
Triglycerides (mg/dL) ORa (95% CI) |
|
| Body mass index | - | |||||
| < 25 | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | |
| 25-<30 | 1.4 (0.9–2.3) | 1.3 (0.8–2.2) | 1.4 (0.9–2.3) | 1.3 (0.8–2.1) | 1.4 (0.9–2.3) | |
| 30-<35 | 1.9 (1.1–3.2) | 1.7 (1.0–3.0) | 1.9 (1.1–3.2) | 1.6 (0.9–2.8) | 1.8 (1.0–3.1) | |
| ≥ 35 | 2.3 (1.4–3.8) | 2.0 (1.2–3.5) | 2.3 (1.4–3.8) | 1.9 (1.1–3.3) | 2.2 (1.3–3.8) | |
| Pre-diabetes or diabetes | - | |||||
| No | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | |
| Yes | 1.3 (0.9–1.7) | 1.3 (0.9–1.8) | 1.5 (1.1–2.0) | 1.3 (0.9–1.7) | 1.3 (1.0–1.8) | |
| Triglyceride/HDL-C ratio | - | - | - | |||
| Low-1.61 | 1.0 | 1.0 | 1.0 | |||
| 1.62–2.49 | 1.1 (0.7–1.8) | 1.2 (0.8–1.9) | 1.3 (0.8–2.0) | |||
| ≥2.50 | 1.7 (1.1–2.6) | 1.9 (1.3–2.8) | 2.0 (1.3–3.0) | |||
| Dyslipidemia | - | |||||
| No | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | |
| Yes | 1.0 (0.7–1.3) | 0.9 (0.7–1.3) | 0.9 (0.7–1.3) | 0.9 (0.7–1.3) | 1.0 (0.7–1.3) | |
| HDL-C (mg/dL) | - | - | ||||
| ≥ 65.00 | 1.0 | 1.0 | 1.0 | 1.0 | ||
| 50.00–64.99 | 1.8 (1.2–2.8) | 1.9 (1.3–2.9) | 2.0 (1.3–3.0) | 1.9 (1.3–2.8) | ||
| < 50.00 | 2.2 (1.4–3.5) | 2.5 (1.6–3.8) | 2.6 (1.7–4.0) | 2.3 (1.4–3.8) | ||
| Triglycerides (mg/dL) | - | - | ||||
| <100.00 | 1.0 | 1.0 | 1.0 | 1.0 | ||
| 100.00–149.99 | 1.2 (0.7–1.9) | 1.2 (0.8–2.0) | 1.3 (0.8–2.0) | 1.1 (0.7–1.7) | ||
| ≥150.00 | 1.5 (0.9–2.4) | 1.6 (1.0–2.5) | 1.7 (1.1–2.6) | 1.3 (0.8–2.1) | ||
OR Odds ratio, CI confidence interval, HDL-C high-density lipoprotein cholesterol
Also adjusted for age categories treated as continuous, service area, number of encounters, race, mammographic screening
Because HDL-C was more highly correlated with BMI than was the triglyceride/HDL-C ratio, we present multivariable results, except for the triglyceride/HDL-C ratio, from a model with BMI, pre-diabetes/diabetes, dyslipidemia, HDL-C and triglycerides (Table 6). Results for the triglyceride/HDL ratio are adjusted for BMI, pre-diabetes/diabetes and dyslipidemia. In general, associations shown in Table 4 were somewhat attenuated, but most remained statistically significant after adjustment for multiple other exposure variables (e.g. the OR per 5 unit increase in BMI was 1.2; 95% CI 1.0–1.4 as compared to 1.3; 95% CI 1.1–1.5 as shown in Table 4). Only associations with pre-diabetes or diabetes and triglyceride levels were no longer statistically significant. Again, associations did not vary statistically significantly by ER-status, except for those for dyslipidemia, such that risk was reduced for ER+ but elevated for ER- IBC (p-het=.002).
Table 6.
ORs (95% CIs) for the relation of body mass index and other exposure variables with all inflammatory breast cancer (IBC) cases, those ≥ 50 years at diagnosis, and estrogen receptor (ER) status (positive (+) and negative (−)) cases with imputed data adjusted for matching and other demographic variables and other comorbidities in the table as noted in the footnotes
| Variable | Total ORs (95% CI) |
Age ≥50 Years ORs (95% CI) | ER+ IBC ORs (95% CI) | ER− IBC ORs (95% CI) | P-value for comparing ER+ to ER |
|---|---|---|---|---|---|
| Body mass index (BMI)a | |||||
| < 25 | 1.0 | 1.0 | 1.0 | 1.0 | |
| 25-<30 | 1.3 (0.8–2.1) | 1.4 (0.8–2.5) | 1.5 (0.7–3.2) | 1.0 (0.5–2.1) | |
| 30-<35 | 1.6 (0.9–2.9) | 1.8 (0.9–3.6) | 2.0 (0.8–4.6) | 1.2 (0.6–2.2) | |
| ≥ 35 | 1.9 (1.0–3.5) | 2.2 (1.0–4.8) | 2.3 (0.8–6.9) | 1.3 (0.6–2.8) | |
| Per 5 BMI unitsc | 1.2 (1.0–1.4) | 1.2 (1.0–1.5) | 1.2 (1.0–1.5) | 1.1 (0.9–1.3) | .44 |
| Pre-diabetes or diabetesa | |||||
| No | 1.0 | 1.0 | 1.0 | 1.0 | |
| Yes | 1.2 (0.9–1.7) | 1.1 (0.7–1.5) | 1.4 (0.9–2.1) | 1.0 (0.6–1.6) | .25 |
| Triglyceride/HDL-C ratiob | |||||
| Low-1.61 | 1.0 | 1.0 | 1.0 | 1.0 | |
| 1.62–2.49 | 1.2 (0.7–1.9) | 1.3 (0.7–2.3) | 1.4 (0.7–2.9) | 1.0 (0.5–1.9) | |
| ≥2.50 | 1.7 (1.1–2.8) | 1.9 (1.1–3.3) | 2.2 (1.2–4.2) | 1.3 (0.7–2.3) | |
| Categorical variable as continuousd | 1.4 (1.1–1.7) | 1.4 (1.1–1.8) | 1.5 (1.1–2.1) | 1.1 (0.8–1.6) | .15 |
| Dyslipidemiaa | |||||
| No | 1.0 | 1.0 | 1.0 | 1.0 | |
| Yes | 0.8 (0.6–1.1) | 0.8 (0.6–1.2) | 0.5 (0.3–0.8) | 1.5 (0.9–2.6) | .002 |
| HDL-Cl (mg/dL)a | |||||
| ≥ 65.00 | 1.0 | 1.0 | 1.0 | 1.0 | |
| 50.00–64.99 | 1.7 (1.1–2.7) | 1.9 (1.1–2.8) | 1.6 (0.9–2.9) | 2.1 (1.0–4.2) | |
| < 50.00 | 2.0 (1.2–3.3) | 2.0 (1.2–3.4) | 2.3 (1.1–4.6) | 1.8 (0.9–3.8) | |
| Categorical variable as continuousc | 1.4 (1.1–1.8) | 1.4 (1.1–1.8) | 1.5 (1.1–2.0) | 1.3 (0.9–1.7) | .39 |
| Triglycerides (mg/dL)a | |||||
| <100.00 | 1.0 | 1.0 | 1.0 | 1.0 | |
| 100.00–149.99 | 1.1 (0.8–1.7) | 1.3 (0.8–2.0) | 1.3 (0.7–2.3) | 0.9 (0.4–2.2) | |
| ≥150.00 | 1.3 (0.8–2.0) | 1.5 (0.9–2.3) | 1.3 (0.8–2.3) | 1.2 (0.6–2.3) | |
| Categorical variable as continuousc | 1.1 (0.9–1.4) | 1.2 (1.0–1.5) | 1.2 (0.9–1.5) | 1.1 (0.8–1.6) | .82 |
OR Odds ratio, CI Confidence interval, IBC Inflammatory breast cancer, ER Estrogen receptor, BMI Body mass index, HDL-C High-density lipoprotein cholesterol
Variables included in the model: age categories treated as an ordinal trend, service area, race, number of encounters, mammographic screening, BMI, pre-diabetes or diabetes, dyslipidemia, HDL-C, triglycerides.
Variables included in the model: age categories treated as an ordinal trend, service area, race, number of encounters, mammographic screening, BMI, pre-diabetes or diabetes, dyslipidemia, triglyceride/HDL-C ratio.
Variables included in the model: age categories treated as an ordinal trend, service area, race, number of encounters, mammographic screening, BMI per 5 units treated as continuous, pre-diabetes or diabetes, dyslipidemia, HDL-C (categorical variable treated as continuous), triglycerides (categorical variable treated as continuous).
Variables included in the model: age categories treated as an ordinal trend, service area, race, number of encounters, mammographic screening, BMI per 5 units treated as continuous, pre-diabetes or diabetes, dyslipidemia, triglyceride/HDL-C ratio (categorical variable treated as continuous).
Sensitivity Analyses
Sensitivity analyses (Table 7) for the BMI/IBC association with unimputed data with missing BMI included as a separate category and with imputed data based on conditional logistic regression yielded results similar to those in Table 4.
Table 7.
Sensitivity Analyses using Unconditional Logistic Regression with Unimputed data and Conditional Logistic Regression with Imputed Data for the Association of Body Mass Index with Inflammatory Breast Cancer
| Unconditional logistic regression using unimputed data | Conditional logistic regression using imputed data | P-value for comparing estimates for ER+ to ER− IBCb | |||
|---|---|---|---|---|---|
| Total IBC | Total IBC | ER+ IBC | ER− IBC | ||
| OR (95% CI)a | OR (95% CI)a | OR (95% CI)a | OR (95% CI)a | ||
| Body mass index | |||||
| < 25 | 1.0 | 1.0 | 1.0 | 1.0 | |
| 25-<30 | 1.7 (1.1–2.8) | 1.5 (0.9–2.4) | 1.9 (1.0–3.6) | 1.1 (0.5–2.3) | |
| 30-<35 | 2.3 (1.4–3.8) | 2.0 (1.2–3.5) | 2.4 (1.2–4.8) | 1.6 (0.7–3.7) | |
| ≥ 35 | 2.9 (1.7–4.9) | 2.7 (1.6–4.5) | 3.8 (2.0–7.4) | 1.6 (0.7–4.0) | .36 |
| Missing | 1.7 (1.1–2.7) | ||||
OR Odds ratio, CI Confidence interval, IBC Inflammatory breast cancer, ER+ Estrogen receptor positive, ER− Estrogen receptor negative
Adjusted for age categories treated as continuous, service area, number of encounters, race, mammographic screening
3 degree of freedom chi-square test comparing the ER+ and ER− body mass index estimates across the three categories, accommodating multiple imputation, and assuming proportionality between/within covariance matrices
Discussion
This retrospective analysis of women treated in a general community health care plan confirms a substantial increase in risk of IBC associated with higher BMI. We have extended prior findings by testing associations using electronic medical records to comprehensively assess IBC risk associated with a variety of comorbid conditions, namely pre-diabetes and diabetes, a measure of insulin resistance, and disturbances of cholesterol metabolism. We found that BMI-associated IBC risk increases were attenuated but still substantial after adjustment for measures of hyperinsulinemia/insulin resistance and disturbances in lipid metabolism. Furthermore, a high triglyceride/HDL-C ratio, one measure of insulin resistance, and low levels of HDL-C substantially increased IBC risk, before and after adjustment for BMI and other related comorbid conditions. These associations did not differ significantly for ER+ and ER- IBC, although power to distinguish was limited. Pre-diabetes or diabetes and high triglyceride levels exhibited associations with increased IBC risk, but after adjusting for other correlated comorbidities, the associations did not meet the threshold for statistical significance.
Our results are consistent with other studies showing positive associations between high BMI and IBC risk regardless of hormone receptor status [5,7]. Among post-menopausal women, obesity has been associated with increased risk of hormone receptor positive breast cancer in general [31]. We are unaware of other studies of insulin resistance and IBC risk. Meta-analyses have found little evidence of an association between serum insulin or c-peptide levels and overall breast cancer risk [18–19], but small associations with diabetes and pre-diabetes [13–14]. A previous study of IBC in the Medicare population reported a small increase in risk with diabetes [20]. Higher HDL-C and triglyceride levels have been associated with reduced breast cancer risk in general [12,17], but have not been examined for IBC. Note that we considered low levels of HDL-C as exposed, rather than high levels. A previous study found a reduced risk of IBC with dyslipidemia regardless of ER status of the tumors [20], whereas here we found such a reduction only for ER+ IBC. The literature on this issue is too sparse to speculate about reasons for the differences between the two studies.
Our findings suggest that insulin resistance/hyperinsulinemia or other closely associated factors, such as inflammatory processes associated with lipid levels, account for some of the increased risk of IBC. We were unable to distinguish these mechanisms because the ratio of triglycerides/HDL-C (a measure of insulin resistance/hyperinsulinemia) was highly correlated with levels of HDL-C (a possible measure of levels of anti-inflammatory cytokines). We were also unable to evaluate other possible mechanisms for the obesity/IBC association, including estrogen biosynthesis in adipose tissue [10].
IBC is angiogenic and expresses high levels of vascular endothelial growth factor (VEGF), which stimulates new blood vessel formation [32]. In addition, high levels of inflammatory cell infiltration and deregulated inflammatory signaling pathways have been identified in IBC tumors [33]. These characteristics are consistent with our findings of increased risk with insulin resistance, which can lead to hyperinsulinemia and associated increases in VEGF production [10], and low levels of HDL-C, which is associated with decreased production of anti-inflammatory cytokines [12]. Obesity is also associated with increased production of pro-inflammatory cytokines and chemokines, which upregulate VEGF [34].
This manuscript extends previous analyses of BMI and IBC risk by assessing and adjusting for risk associated with comorbidities of BMI. We had access to a large study population from which to identify IBC cases and detailed pre-diagnostic clinical data available from electronic health records. However, the use of electronic health records does present some challenges. Availability of data depends in part on number of encounters a person has with the health care system; in our analysis the average number of encounters was similar in cases and controls with a broader range in controls than cases. Only a small percentage of cases and controls had three or fewer encounters, mitigating the concern of misclassifications of conditions due to the absence of encounters rather than true negatives. To deal with missing data in an unbiased fashion we used multiple imputation; results for BMI were similar using unimputed and imputed data. Due to the relatively short exposure window we were unable to determine duration of disease (e.g. for diabetes). Moreover, distinguishing between a final diagnosis versus a work-up for a condition can be sometimes difficult. It is likely, however, that we truly identified patients with the conditions under study, because multiple data sources such as pharmacy, diagnosis codes, and laboratory results were used [22]. In addition, we did not have information on all standard breast cancer risk factors, particularly menopausal hormone use; associations with BMI and breast cancer in general are stronger in never users of post-menopausal hormones [8]. We also did not evaluate medication use associated with the medical conditions studied. Most type 1 diabetic patients have therapeutic hyperinsulinemia [29]. The majority of therapies for type 2 diabetes aim to increase insulin secretion or increase its effect [28]. Metformin and thiazolidinediones have been associated with reduced breast cancer risk [35]. Lipid-lowering drugs did not account for the reduced breast cancer risk associated with dyslipidemia in one study [36]. It is possible that our results for pre-diabetes or diabetes and dyslipidemia reflected conditions that were both controlled and uncontrolled by medication.
Conclusions
In summary, we have confirmed the substantial increased risk of IBC with obesity and demonstrated that it is independent of several associated comorbidities. We have also shown that several of these comorbidities, namely a measure of insulin resistance and low HDL-C levels, themselves increase IBC risk independently of the BMI association. These findings, if confirmed, have implications for prevention of IBC.
Acknowledgements and Funding Information
Acknowledgements are not applicable. This work was supported by the National Cancer Institute at the National Institutes of Health Intramural Research Program and in part by NCI grant U24 CA171524.
Abbreviations
- IBC
Inflammatory breast cancer
- KPNC
Kaiser Permanente Northern California
- KPNCCR
Kaiser Permanente Northern California Cancer Registry
- BMI
Body mass index
- ER
Estrogen receptor
- ICD
International Classification of Disease
- OR
Odds ratio
- CI
Confidence interval
- HDL-C
High- density lipoprotein cholesterol
- VEGF
Vascular endothelial growth factor
Footnotes
Competing Interests
The authors declare that they have no competing interests.
Disclosure and Declarations
Ethics approval and consent to participate
This project was reviewed and approved by the KPNC Institutional Review Board (Protocol number: 1285916). The requirement to obtain informed consent for this data-only project was waived.
Availability of data and materials
The data in this manuscript were sourced from electronic health records and related clinical and administrative data sources from Kaiser Permanente Northern California and are not generally available to the public. These data may be made available through scientific collaborations under IRB-approved protocols and data use agreements.
References:
- 1.Anderson WF, Schairer C, Chen BE, Hance KW, Levine PH (2005, 2006) Epidemiology of inflammatory breast cancer (IBC). Breast Dis 22:9–23 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Greene FL, Page DL, Fritz A, Balch CM, Haller DG, Breast Morrow M. (2002) In: AJCC Cancer Staging Manual. 6th ed New York: Springer-Verlag; p. 225–226 [Google Scholar]
- 3.Hance KW, Anderson WF, Devesa SS, Young HA, Levine PH (2005) Trends in inflammatory breast cancer incidence and survival: The Surveillance, Epidemiology, and End Results Program at the National Cancer Institute. J Natl Cancer Inst 97:966–974 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Amiri-Kordestani L, Kamangar F, Zujewski JA (2013) Inflammatory breast cancer: yet another risk of the obesity epidemic? J Natl Cancer Inst 105:1340–1342 [DOI] [PubMed] [Google Scholar]
- 5.Schairer C, Li Y, Frawley P, Graubard BI, Wellman RD, Buist DS, Kerlikowske K, Onega TL, Anderson WF, Miglioretti DL (2013) Risk factors for inflammatory breast cancer and other invasive breast cancers. J Natl Cancer Inst 105:1373–1384 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Chang S, Buzdar AU, Hursting SD (1998) Inflammatory breast cancer and body mass index. J Clin Oncol 16:3731–3735 [DOI] [PubMed] [Google Scholar]
- 7.Atkinson RL, El-Zein R, Valero V, Lucci A, Bevers TB, Fouad T, Liao W, Ueno NT, Woodward WA, Brewster AM (2016) Epidemiological risk factors associated with inflammatory breast cancer subtypes. Cancer Causes Control 27:359–366 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Suzuki R, Orsini N, Saji S, Key TJ, Wolk A (2009) Body weight and incidence of breast cancer defined by estrogen and progesterone receptor status – A meta-analysis. Int J Cancer 124:698–712 [DOI] [PubMed] [Google Scholar]
- 9.Varmus H, Harlow E (2012) Provocative questions in cancer research. Nature 481:436–437 [DOI] [PubMed] [Google Scholar]
- 10.Matthews SB, Thompson HJ (2016) The obesity-breast cancer conundrum: an analysis of the issues. Int J Mol Sci 17:989. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Rose DP, Gracheck PJ, Vona-Davis L (2015) The interactions of obesity, inflammation and insulin resistance in breast cancer. Cancers 7:2147–2168 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ni H, Liu H, Gao R (2015) Serum lipids and breast cancer risk: a meta-analysis of prospective cohort studies. PLoS One. November 10;10(11):e0142669. doi: 0.1371/journal.pone.0142669. eCollection [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.De Bruijn KM, Arends LR, Hansen BE, Leeflang S, Ruiter R, van Eijck CH (2013) Systematic review and meta-analysis of the association between diabetes mellitus and incidence and mortality in breast and colorectal cancer. Br J Surg 100:1421–1429 [DOI] [PubMed] [Google Scholar]
- 14.Huang Y, Cai X, Qiu M, Chen P, Tang H, Hu Y, Huang Y (2014) Prediabetes and the risk of cancer: a meta-analysis. Diabetologia 57:2261–2269 [DOI] [PubMed] [Google Scholar]
- 15.Boyle P, Koechlin A, Pizot C, Boniol M, Robertson C, Mullie P, Bolli G, Rosenstock J, Autier P (2013) Blood glucose concentrations and breast cancer risk in women without diabetes: a meta-analysis. Eur J Nutr 52:1533–1540 [DOI] [PubMed] [Google Scholar]
- 16.Esposito K, Chiodini P, Capuano A, Bellastella G, Maiorino MI, Rafaniello C, Giugliano D (2013) Metabolic syndrome and postmenopausal breast cancer: systematic review and meta-analysis. Menopause 20:1301–1309 [DOI] [PubMed] [Google Scholar]
- 17.Touvier M, Fassier P, His M, Norat T, Chan DS, Blacher J, Hercberg S, Galan P, Druesne-Pecollo N, Latino-Martel P (2015) Cholesterol and breast cancer risk: a systematic review and meta-analysis of prospective studies. Br J Nutr 114;347–357 [DOI] [PubMed] [Google Scholar]
- 18.Autier P, Koechlin A, Boniol M, Mullie P, Bolli G, Rosenstock J, Boyle P (2013) Serum insulin and c-peptide concentration and breast cancer: a meta-analysis. Cancer Causes Control. 24:873–883 [DOI] [PubMed] [Google Scholar]
- 19.Hernandez AV, Guarnizo M, Miranda Y, Pasupuleti V, Deshpande A, Paico S, Lenti H, Ganoza S, Montalvo L, Thota P, Lazaro H (2014) Association between insulin resistance and breast carcinoma: a systematic review and meta-analysis. PLoS ONE 9(6): e99317.doi 10.1371/journal.pone.0099317 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Schairer C, Gadalla SM, Pfeiffer RM, Moore SC, Engels EA (2017) Diabetes, abnormal glucose, dyslipidemia, hypertension, and risk of inflammatory and other breast cancer. Cancer Epidemiol Biomarkers Prev 26:862–868 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Nichols GA, Desai J, Lafata JE, Lawrence JM, O’Connor PJ, Pathak RD, Raebel MA, Reid RJ, Selby JV, Silverman BG, Steiner JF, Stewart WF, Vupputuri S, Waitzfelder B, SUPREME-DM Study Group (2012) Construction of a multisite datalink using electronic health records for the identification, surveillance, prevention, and management of diabetes mellitus: the SUPREME-DM Project. Prev Chronic Dis doi: 10.5888/pcd9.110311 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Koebnick C, Smith N, Huang K, Martinez MP, Clancy HA, Kushi LH (2012) The prevalence of obesity and obesity-related health conditions in a large, multiethnic cohort of young adults in California. Ann Epidemiol 22:609–616 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.McKeever K, Saydah SH, Imperatore G, Cowie CC, Gregg EW, Geiss LS, Cheng YJ, Rolka DB, Williams DE, Caspersen CJ (2013) Secular changes in U.S. prediabetes prevalence defined by hemoglobin A1c and fasting plasma glucose. Diabetes Care 36:2286–2293 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Schmittdiel JA, Dyer WT, Marshall CJ, Bivins R (2018) Using neighborhood-level census data to predict diabetes progression in patients with laboratory-defined prediabetes. Perm J 22:18–096 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Raghunathan TE, Lepkowski JM, van Hoewyk M, Solenberger PW (2001) A multivariate technique for multiply imputing missing values using a sequence of regression models. Survey Methodol 27:85–95 [Google Scholar]
- 26.Priya G, Kalra S (2018) A review of insulin resistance in type I diabetes: is there a place for adjunctive metformin? Diabetes Ther 9:341–361 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Hostalek U, Gwilt M, Hildemann S (2015) Therapeutic Use of Metformin in Prediabetes and Diabetes Prevention. Drugs 75:1071–1094 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Ojha A, Ojha U, Mohammed R, Chandrashekar A, Ojha H (2019) Current perspective on the role of insulin and glucagon in the pathogenesis and treatment of type 2 diabetes. Clinical Pharmacy: Advances and Applications 11;57–65 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Scappaticcio L, Maiorino MI, Bellastella G, Giugliano D, Esposito K (2017) Insights into the relationships between diabetes, prediabetes, and cancer. Endocrine 56:231–239 [DOI] [PubMed] [Google Scholar]
- 30. https://ncbi.nlm.nih.gov/books/NBK507839.
- 31.Garcia-Estevez L, Moreno-Bueno G (2019) Updating the role of obesity and cholesterol in breast cancer. Breast Cancer Research 21:35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Vermeulen PB, van Golen KL, Dirix LY (2010) Angiogenesis, lymphangiogenesis, growth pattern, and tumor emboli in inflammatory breast cancer: a review of current knowledge. Cancer 116 (Suppl 11):2748–2754 [DOI] [PubMed] [Google Scholar]
- 33.Morrow RJ, Etemadi N, Yeo B, Ernst M (2017) Challenging a misnomer? The role of inflammatory pathways in inflammatory breast cancer. Mediators Inflamm doi. 10.1155/2017/4754827 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Kolb R, Sutterwala FS, Zhang Weizhou (2016) Obesity and cancer: inflammation bridges the two. Curr Opin Pharmacol 29:77–89 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Ahmadieh H, Azar ST (2013) Type 2 diabetes mellitus, oral diabetic medications, insulin therapy, and overall breast cancer risk. ISRN Endocrinology. 10.1155/181240 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Schairer C, Freedman DM, Gadalla SM, Pfeiffer RM (2018) Dyslipidemia, lipid-lowering drug use and breast cancer risk in the SEER-Medicare database. Breast Cancer Res Treat 169:607–614 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data in this manuscript were sourced from electronic health records and related clinical and administrative data sources from Kaiser Permanente Northern California and are not generally available to the public. These data may be made available through scientific collaborations under IRB-approved protocols and data use agreements.