The New Jersey Ovarian Cancer Study is a population-based case–control study and has been described elsewhere [17–19]. In brief, eligible women were older than 21 years, able to speak English and/or Spanish, and residents of six contiguous counties in New Jersey (Essex, Union, Morris, Middlesex, Bergen, and Hudson). Cases were newly diagnosed, histologically confirmed cases of invasive epithelial ovarian cancer, identified by rapid case ascertainment by the New Jersey State Cancer Registry, a SEER Registry. Population controls from the EDGE (Estrogens, Diet, Genetics, and Endometrial Cancer) Study served as controls for this study and are described elsewhere [20, 21]. Briefly, controls were identified via random digit dialing (RDD) if under 65 years of age and Centers for Medicare and Medicaid Services (CMS) and area sampling if age 65+ years and 55+ years, respectively. Recruitment of cases and controls occurred between July 2001 and May 2008. Women who had a hysterectomy or those who had a bilateral oophorectomy were not eligible as controls in the NJ Ovarian Cancer Study. Informed consent was obtained from all participants. This study has been approved by the Institutional Review Boards of the New Jersey Department of Health and Senior Services, Memorial Sloan-Kettering Cancer Center and University of Medicine and Dentistry of New Jersey (UMDNJ) Robert Wood Johnson Medical School.
Same study procedures and materials were used for cases and controls. Informed consent was obtained before the phone interview. Cases and controls completed a phone interview during which a questionnaire was administered ascertaining demographic characteristics and major risk factors for the disease such as hormone use, family history of cancer, reproductive history, medical history, and lifestyle factors up to a year prior to diagnosis (or date of interview for controls). A food frequency questionnaire (FFQ), the Block 98.2 FFQ (110 food items), was self-administered and returned by mail, along with waist and hip measurements (a tape measure and instructions were provided), and a mouthwash sample for DNA extraction.
We initially identified 682 eligible cases, of whom some were excluded as they were either deceased (n=61) or physicians advised us not to contact them (n=9). Additional cases were excluded if they could not be reached or no longer met eligibility requirements, such as a communication barrier or medical conditions that precluded participation (n=119). In total, 233 of the remaining 493 cases (47%) and 467 controls (40%) completed the phone interview. Participants were excluded from the analysis if their menopausal status was unknown or if they were missing other major covariates. Those who were postmenopausal but did not know their age at menopause were included in the analysis. Of the remaining cases and controls, 205 cases (88%) and 398 controls (85%) completed both the interview and FFQ. Eight of these controls were excluded from these analyses because both of their ovaries had been removed. There were no significant differences in major characteristics between those who did and did not complete the food frequency questionnaire.
Processing of dietary data
Participants’ responses were converted to number of servings per day based on their reported frequency and portion sizes for sugary foods and beverages. Frequency was measured as ‘never’, ‘a few times per year’, ‘once per month’, ‘2-3 times per month’, ‘once per week’, ‘2 time per week’, ‘3-4 times per week’, ‘5-6 times per week’, and ‘everyday’ for most food items. For a few foods, ‘never’ and ‘a few times per year’ were combined into one choice: ‘never or a few times per year’ and the choice of ‘2+ times per day’ was added. Portion size for food items was measured in teaspoons, tablespoons, ounces, pounds, cups, pieces, patties, bowls or slices. Portion size for beverages was measured as number of cups, glasses, cans or bottles consumed.
Serving sizes were based on the guidelines listed in Reference Amounts Customarily Consumed (RACC) Per Eating Occasion: General Food Supply by the Food and Drug Administration (FDA) . This document provides the amount of food typically consumed per eating occasion, and is based on the 1977-1978 and 1987-1988 Nationwide Food Consumption Surveys. When making assumptions about participants’ portion sizes consumed, we used the FDA’s assigned RACC values as a guideline. For example, we assumed that one doughnut (RACC=55 grams) is equivalent to one serving. Therefore, participants who reported usually eating one doughnut per occasion were assigned as eating one serving for this food item.
Next, we computed the number of servings of dessert foods with added sugars, non-dessert foods with added sugars, sugary drinks and total sugary foods and drinks for each participant. Total and added sugar intakes (g/day) were calculated for each relevant food item by multiplying the frequency of intake by the total/added sugar content per 100 grams of food.
Total sugars are the sum of both natural and added sugars in the diet . Natural sugars, like fructose or lactose, are found in whole fruit, vegetables, or milk products, which also have nutrients and phytochemicals beneficial to an individual’s health . Added sugars are all caloric sweeteners that have been added to foods or drinks during processing, preparation, and also consumed separately or at the table. Foods and beverages with added sugars tend to be high in calories and lacking essential nutrients . Examples of added sugars are sucrose (i.e. table sugar), high fructose corn syrup, honey, molasses, and syrups [14, 15, 23]. Sugary foods and drinks are foods that have been processed, prepared, or consumed with added sugars . Total and added sugar content values were based on the USDA Database for Added Sugars Content of Selected Foods .
Calculation of percent of calories from sweets and desserts (% kcal from sweets) included the following FFQ items: regular and low-fat ice cream, ice milk or ice cream bars, doughnuts or Danish pastry, regular or low-fat cake, sweet rolls or coffee cake, regular and low-fat cookies, pumpkin pie or sweet potato pie, other pie or cobbler, chocolate candy or candy bars, candy (not chocolate), soft drinks or sweetened bottled drinks like Snapple (not diet), sugar or honey added to coffee/tea, breakfast bars, granola bars or power bars, sweetened cereals, and jelly, jam or syrup. Information about the respondent’s consumption of diet drinks or use of non-caloric sweeteners (within foods or added at the table) was not collected.
Descriptive statistics were computed for total and added sugars and food and drink groups. For all analyses, statistical significance was considered a p-value less than 0.05. To describe our study population, the distribution of major characteristics for cases and controls was tabulated. Two sample t-tests were used to compare cases and controls across continuous variables and chi-square tests were used for categorical variables. Age-adjusted logistic regression models were used to calculate odds ratios (ORs) and 95% confidence intervals (CIs) to compare ovarian cancer risk across major risk factors (except for age).
ANCOVA was used to calculate age-adjusted means to compare mean intake between cases and controls for each food and drink group: dessert foods, non-dessert foods, sugary drinks, total sugary foods and drinks, as well as total and added sugar intakes. Based on the distribution in controls, tertiles for the food and drink groups and total and added sugars intake were created and frequencies calculated across the tertiles. Age-adjusted and multiple unconditional logistic regression models were used to estimate ORs and 95% CIs for the food and drink groups and total and added sugar intakes.
Covariates considered in multiple logistic regression models include age (continuous), years of education (≤12, 13-16, >16), race/ethnicity (White, Black, Other, Hispanic-any race), age at menarche (>13, 12-13, ≤11), menopausal status (pre- or postmenopausal) and age at menopause for postmenopausal women (<40, 41-54, ≥55, age at menopause unknown), parity (0-1, 2, ≥3), oral contraceptive (OC) use (ever vs. never), hormone replacement therapy (HRT) use (never, unopposed estrogen only, any combined HRT), BMI (weight in kg/height in m2; continuous), smoking status (never, past, current) and pack-years (continuous) for ever smokers, physical activity measured in continuous metabolic equivalents (METs), tubal ligation (yes vs. no), dietary intakes of fiber, total fat and saturated fat, and diabetes (yes vs. no). We adjusted for total energy intake using the multivariate nutrient density method . Specifically, we computed density measures for servings of sugary foods and/or drinks per 1,000 kcal of intake, as well as grams of total or added sugars per 1,000 kcal of intake and included daily caloric intake as a continuous variable in the multivariable models. Tests for trend were conducted by assigning to each tertile the median value of servings of sugary foods and/or drinks per 1,000 kcal or total or added sugar intakes (g/1,000 kcal) among controls. In addition, tertiles for percent of calories from sweets and desserts (i.e. dessert foods group) per day were created based on the controls, and frequencies calculated across these tertiles. Odds ratios and 95% CIs were calculated to assess ovarian cancer risk across these tertiles.
Overweight or obesity is a strong determinant of insulin resistance and hyperinsulinemia [26–30]. Additionally, central obesity [31, 32], which is related to insulin resistance , has been shown to significantly increase risk of ovarian cancer. We hypothesized that insulin-related risk factors might modify the relationship between sugar intake and cancer risk. Thus, we explored effect modification by factors capable of affecting the body’s response to insulin production such as BMI (normal weight: <25 kg/m2 vs. overweight or obese: ≥25 kg/m2), waist-to-hip ratio (WHR; ≤0.85 vs. >0.85), or physical activity (< median vs. ≥median for controls). Odds ratios and 95% CIs were calculated for ovarian cancer risk across tertiles for total sugary foods and drinks and total and added sugars, stratified by these factors. Because the number of women with diabetes was too small to conduct separate analyses on them, we also repeated analyses excluding women diagnosed with diabetes. The Wald test was used to calculate p-values.