Eligibility
Participants must be referred by their clinical oncologist, have a pathology report confirming the resected stage of breast cancer and documentation of the type of systemic adjuvant therapy, as well as have a BMI in the overweight or obese class I range (BMI 25-34.9 kg/m2). Eligible participants: must be at least 4 months post radiation or chemotherapy treatment for breast cancer with no evidence of metastatic disease; must not anticipate having surgery over the study duration period; must not follow a special diet excluding foods or food groups; have not lost 2 or more kg of body weight over the month preceding study initiation; must not be taking pharmaceuticals or supplements for weight management; are not being treated for diabetes or blood glucose control; have no history of eating disorders; do not have digestive issues that may interfere with dietary intake, such as irritable bowel syndrome, Crohn's disease, or diverticulitis; have never had surgery involving constriction or removal of any portion of the gastrointestinal tract; have not been diagnosed with hepatitis B, C, or HIV; do not have implanted electronic devices such as a pacemaker; and do not use tobacco products. Participants must also be willing to follow a dietary plan prescribed for the duration of the study; adhere to American Cancer Society alcohol guidelines (≤ 1 standard drink per day); maintain or increase physical activity as prescribed to achieve negative energy balance required for 0.5-1.0 kg weight loss per week; wear a pedometer and record daily activity; wear an accelerometer/heart rate monitor for 2 weeks during the study (1 week at the beginning and 1 week at the end of the study); wear a body or swim suit and cap for body composition testing; record food intake daily; and attend up to ten one-on-one clinic visits and 5 group visits with seven fasting blood and first-void urine samples in the intervention groups, or three one-on-one clinic visits and two fasting blood and first-void urine samples in the control group over 26 weeks.
Dietary Plans
Dietary patterns are composed of opposing fat and carbohydrate contents but balanced in protein (Table
1). Six-weeks of meal plans were designed for five calorie levels in each diet arm. The meal plans developed included interchangeable meal options with home-prepared recipes and meal instructions. Supporting materials are provided to facilitate adherence including eating out and frozen meal options, food brand options consistent with the plan, meal planning tools and shopping lists. Educational materials were developed based on a systematic review of strategies supporting weight loss maintenance, incorporating program components (e.g. self-monitoring tools) and core competencies reinforcing weight loss behaviors. These strategies are taught to participants through the one-on-one and group sessions in order to promote high levels of dietary adherence.
Table 1
Mean Proposed Macronutrient Composition by Diet Group
Carbohydrate (%) | 32 | 64 |
Fat (%) | 48 | 16 |
Protein (%) | 20 | 20 |
Participants are instructed to increase their physical activity by increasing steps or step equivalents to contribute to a 500 calorie deficit each day in combination with caloric restriction. Calorie goals are determined based on resting metabolic rate and energy expenditure from activity.
Glucose Homeostasis
As hyperglycemia and insulin resistance are associated with an increased risk for breast cancer [
75], factors related to these metabolic processes will be measured including homeostasis model assessment (HOMA). HOMA is a method used to quantify insulin resistance and pancreatic beta cell function and is a calculated index based on fasting levels of insulin and glucose, giving an integrated view of glucose utilization. HOMA has been shown to be associated with increased breast cancer incidence [
75], and higher breast cancer mortality in the HEAL study [
76]. Fasting insulin alone has also been shown to be predictive of survival in breast cancer patients [
77]. In addition, insulin-like growth factor-1 (IGF-1), IGF binding protein-3 (IGFBP-3) and the ratio of IGF-1:IGFBP-3, which provides an estimate of biologically available IGF-1, a more useful indication of overall, longer term control of glucose homeostasis in relation to breast cancer risk, will be measured [
78]. Serum IGF-1 is positively associated with breast cancer risk, as well as changes in response to caloric restriction and nutritional alterations [
79]. Measuring the effects of two differing dietary patterns on endogenous production of IGF-1 will help to further characterize dietary effects on breast cancer risk.
Cellular Oxidation
The inflammatory response stimulated by obesity and insulin resistance also increases oxidative stress in the body [
80]. During oxidative stress, byproducts of nucleic acid metabolism including reactive oxygen species (ROS) promote cancer development by causing genetic mutations and DNA damage [
55,
56,
81]. Secondary measures of cellular oxidation that will be measured include 8-hydroxy-deoxy-guanosine (8-OH-dG) reflecting defects in DNA repair capacity, and markers of whole body lipid peroxidation, 8-isoprostane-F2-alpha (8-iso-PGF
2α), which have been shown to play a role in breast carcinogenesis [
49,
82,
83]. These byproducts of oxidation are more common in cancerous breast tissue compared to normal breast tissue [
49,
55,
56], and may also be altered with reducing energy intake [
51,
52] and changes in dietary macronutrient composition [
84].
Overweight postmenopausal women have elevated concentrations of circulating estrogens and lower concentrations of sex hormone binding globulin (SHBG), which promotes cell growth, putting them at more than twice the risk for breast and endometrial cancers, as evidenced in the Healthy Eating and Lifestyles (HEAL) study and the European Prospective Investigation into Cancer and Nutrition (EPIC) study [
12,
85,
86]. Adipose tissue exhibits aromatase enzyme activity, converting androgenic precursors to estradiol and estrone. Estrogens may promote tumorigenesis through direct or indirect induction of free radical-mediated DNA damage, genetic instability, cell mutations and cell proliferation. Agents such as tamoxifen (selective estrogen receptor modulators) or the aromatase inhibitors have been shown to reduce breast cancer incidence and recurrence [
87]. Although warranted, there are currently no published randomized, controlled studies of the effect dietary pattern has on estrogens or SHBG and other candidate mechanisms [
88‐
90]. Secondary hormone metabolism outcomes will include estradiol, estrone, progesterone, and SHBG.
Body Fat
BOD POD technology is fundamentally the same as underwater (hydrostatic) weighing, but uses volumetric air displacement versus water displacement. All outcome measures including weight are assessed using validated and standardized measuring equipment and techniques. Air displacement plethysmography (BOD POD, Life Measurement, Inc., Concord, CA) has been shown to measure changes in body composition in response to weight change to the same extent as dual x-ray absorptiometry (DEXA), with similar sensitivity [
92]. The BOD POD measures the volume of air a person's body displaced while sitting inside a comfortable chamber. By using air versus water, the BOD POD offers a fast, safe, and easy-to-use tool for measuring body composition, without sacrificing accuracy. Since it is based on the same whole-body measurement principle as hydrostatic weighing, the BOD POD first measures the subject's mass and volume. From these measurements, whole-body density is determined, and body fat and lean mass calculated using standardized equations.
Monitoring Breast Cancer Recurrence
Women with a history of invasive breast cancer are seen every 3 months for the first 2 years following treatment at the study site clinic. At each visit, a clinical history is updated, a physical examination is performed by the attending oncologist, and serum levels of CA 27.29 and CEA are determined. The frequency of clinical visits decreases to every 6 months for the next three years, and follow up occurs annually thereafter. Following a patient's enrollment into this study, and for the duration over which the project is funded, at each of these clinical visits, the relevant disease recurrence data will be recorded along with height, weight, and body composition. Blood will also be drawn and banked at each visit for subsequent hypothesis testing, although such analyses are beyond the scope of this project.
Data Analysis
All data elements collected will be examined using descriptive statistics (mean, median, SD, range, percentiles, proportion) and graphical summaries (box plots, profile plots by time and diet group). Log transformations will be made before further analysis to stabilize variances as needed.
The primary hypothesis on C-reactive protein will be tested using the following ANCOVA model:
Where Y
4 is the outcome measure at the end of 6 months, Y
1 is the outcome measure at baseline, and, GLYC is a 2-level indicator for dietary pattern. This method of analysis adjusts for any remaining pre-treatment differences between groups (a precaution against imbalance after diet assignment) and reduces variability in the data being analyzed [
97], thus improving the power of the test for the main effect of interest, b
2.
There will be a wealth of information in the repeated measures on each subject; the results for all measures using all available data from all time points will be estimated in a mixed-effects repeated measures model to assess the slopes and between group differences after each month of weight loss [
98]. The power of this approach lies in its ability to incorporate all of the available longitudinal data even in the unbalanced case, that is, when some of the observations are missing for one or more individuals. Observations within a person over time are allowed to be correlated while observations across individuals are assumed to be independent. These models will also be used to explore the effects of breast cancer stage, BMI (a time-varying covariate) and age.
Depending on the appearance of change over time (seen in the profile plots of each outcome measure by time and diet) linear or nonlinear mixed models will be used. If the trend appears to be linear, the following model for the response vector
y
i for the
ith group will be used:
where
b
i
~
N (0,D) and
e
i
~
N (0,R
i
) are independent. X
i
is a fixed effects design matrix that includes indicators for diet group (1, 2), assessment time (after each 1.5 kg fat loss, and potential covariates (age) or confounders (disease stage, BMI).
Z
i
is a design matrix for the random effects that allows random subject deviations from the population average response. The marginal distribution of
y
i
is normal with mean
X
i
β and variance
. Parameter estimation in SAS allows a wide range of specifications for the forms of
D and
R, and combines empirical Bayes and restricted maximum likelihood using the EM algorithm.
If the descriptive graphs suggest a nonlinear model is appropriate, we will estimate:
Where
b
i
~
N (
0,
D) and
e
i
~
N (
0,
R
i
). The marginal distribution of
y
i
is difficult to find in most cases, but its mean and variance can be approximated by
Where
is the partial derivative of
f (⋅) with respect to
b
i
. Parameter estimation in SAS combines a linearization algorithm, such as Gauss-Newton, and the method of Laird-Ware for linear mixed models. We will explore the use of nonlinear models only if it appears that the response trajectory of
Y over time could be fit well by a smooth nonlinear function. Otherwise, simpler piece-wise linear mixed models will be fit using Equation
1 above.
Secondary measures in the 4 families of outcomes will be assessed using the same statistical methods described above for the primary measures. That is, ANCOVA assessment of the measure by diet group at 6 months followed by an exploratory analysis with mixed models using all available data.
Mixed models will also be used to estimate the effects of fat loss on the 4 families of outcomes. Fat loss and weight loss will be modeled with the expectation that differences across diets will be minimal, and success in weight loss better explained by age, initial BMI, and usual level of physical activity.
Missing data are expected to be missing at random, and are unlikely to exceed 5% of all observations. This assumption will be checked during the initial descriptive analysis of the data after database lock, and appropriate sensitivity analyses will be done if there is evidence that the data are MNAR (missing not-at-random) [
99]. Given the nature of the patient population and the incentives to return for monthly evaluations, we expect to collect most endpoint measures regardless of compliance. We will accrue 370 subjects with the expectation of completing at least 135 per diet group and 100 in the control group, that is, 10% loss to follow-up. Analysis will be intent-to-treat. All statistical analysis will be done using SAS version 9 (SAS Institute, Cary, NC).