Our research utilized data collected during the Diet Intervention Examining The Factors Interacting with Treatment Success (DIETFITS) study, a year-long dietary clinical trial conducted in the San Francisco Bay Area between May 2014–May 2016 . Participants were recruited from the general population and were healthy women and men, 18–50 years of age, with a BMI between 28 and 40 kg/m2. Further information on the DIETFITS study and recruitment has been previously published . Our research includes participants of DIETFITS, starting with the trial’s second through its fifth and final cohort.
The BPA dietary exposure risk questionnaire was developed and piloted in April 2014 based upon literature describing which food sources or food preparation methods are most likely to increase exposure to BPA and methods developed to score this type of exposure [29, 32]. The more packaged and processed foods are, the more BPA a food can contain. Specifically, the BPA dietary exposure risk questionnaire (Supplemental Material- Questionnaire, See Additional file 1) asked about the frequency that participants undertake certain dietary behaviors that increase the risk of BPA exposure: eating canned food [27, 33], microwaving food in plastic food storage containers , drinking beverages in polycarbonate plastic containers , drinking hot beverages from polycarbonate plastic containers , microwaving food covered with polyvinyl chloride-based plastic stretch wrap , eating microwaveable meals [35, 36], and eating packaged food [36, 37]. Questions were scored using frequency per day or week and Likert-scale style coding.
A composite BPA exposure (BPAe) risk score of the seven questions on the dietary exposure risk questionnaire was created to enable bivariate comparisons between the questionnaire and BPA intake reported in dietary recalls. BPAe was created by (i) recoding questionnaire responses such that lower responses indicated lower frequency and higher responses indicated a higher frequency of conducting BPA exposing activities (e.g., 1 = Never and 5 = Most), (ii) choosing the minimum number of principal components (PCs) that explained at least 90% of the variation in the data, and (iii) calculating the quintile of the sum of the PCs weighted by the percentage of variation explained by each PC. For example, if 3 PCs explained 90% of the variation, e.g., PC1 75%, PC2 10%, and PC3 5%, and the values of PC1, PC2, and PC3, were 1, 1, and 1, respectively, BPAe would be the quintile containing 1*(0.75) + 1*(0.15) + 1*(0.05) = 0.9. Thus, BPAe was equal to 1 for the lowest exposure and 5 for the highest exposure. To validate scores, each of the seven questions was compared to levels of BPAe to ensure increasing values of question responses across the BPAe categories.
Quantifying dietary BPA intake in dietary recalls
The DIETFITS study assessed dietary behaviors using 24-hour dietary recalls. The recalls were collected using the Nutrition Data System for Research (NDSR) software versions 2012–2015, developed by the Nutrition Coordinating Center (NCC), University of Minnesota, Minneapolis, MN. Three dietary recalls were administered at each data collection time point, aiming for two weekdays and one weekend day, to estimate typical eating patterns. Dietary data collected with NDSR was coded upon collection into a unique food identification code (Food ID) with accompanying Food Description.
The DIETFITS study questionnaire and the dietary recalls were administered over the same two-week window at baseline, 3, 6, and 12 months. Our BPA dietary exposure risk questionnaire was a subset of questions within the main study questionnaire. In our analyses, we considered only data collected at baseline as participants were explicitly instructed over the course of the study to favor whole foods rather than processed foods, and thus BPA exposure was anticipated to be highest at baseline. The average of the baseline dietary recalls was utilized for our analysis.
We identified foods likely to be high in BPA contamination from the dietary recall data. We searched the NDSR Foods 2017 database, containing information for 32,300 food items, for the keywords “canned”, “packaged”, and “microwave” in the NCC Food Descriptions to develop a subset of food items possibly containing BPA. There were 317 items listed in NDSR as having “canned” in their Food Description. There were 866 items that had the word “packaged” in their Food Description. All but two of these packaged items were described as a type of sweet, including snack cakes, ice cream, or frozen treats. Twelve items had the word “microwave” in their Food Description. These were all packaged foods, either frozen meals, microwave-in-a-cup meals, or microwave popcorn. From this method, we created a list of canned foods, packaged food, and microwave foods (Supplemental Tables S1-S3, See Additional file 1).
To estimate BPA exposure from the dietary recalls, we first recorded the number of specific food items consumed per participant from the lists of canned, packaged, and microwave specific BPA foods created from the NDSR Food Descriptions. Next, we created two variables to describe an estimate of BPA exposure: (i) weighted servings and (ii) total grams. Weighted servings were calculated as the sum of the total number of servings of foods consumed from the canned, packaged, and microwave lists, with each item weighted by the potential concentration of BPA contamination: 1 x canned food items, 0.25 x packaged food, and 0.25 x microwave food, a weighting scale developed and applied in previous research , reflecting the higher BPA contamination potential of consuming canned foods. Our method differed from the original research as we applied the weighing scale to the dietary recall. Total grams were calculated as the sum of grams consumed from all canned, packaged, and microwave food items identified on the lists. Consequently, each study participant had up to eight dietary BPA outcomes: weighted servings and grams of canned, packaged, microwave, and overall BPA.
Study demographics were summarized using means and standard deviations for continuous variables and sample sizes and percentages for categorical variables. The BPAe composite scores were used as a first approach to measuring the association between the dietary exposure risk questionnaire and overall BPA in dietary recall data. Kruskal-Wallis rank-sum tests (K-W) were used to test the null hypothesis that weighted servings and total grams were equal across the five groups of BPAe. Heavily right-skewed outcome data indicated that K-W was more appropriate than one-way ANOVA . If an overall K-W test was significant, a follow-up analysis was conducted using Dunn’s test pairwise comparisons with a Bonferroni adjustment for multiple hypothesis testing.
A second approach to measure the association between the dietary exposure risk questionnaire and dietary recall data involved the use of linear regression models for overall weighted servings and total grams as a function of the seven separate questions on the questionnaire, maintaining each question as a categorical variable. Questionnaire responses were binned into three levels per question to enable easy interpretation of the findings without making an ordinality assumption. For example, the original question for the number of canned foods eaten per week had answer choices none, 1, 2, 3, 4, 5, and > 5, and we binned to none, 1 or 2, and 3 or more per week. Missing data in the questionnaire data were statistically addressed using multiple imputations via chained equations (MICE) to generate 5 imputed data sets, and model estimates and inferences were combined using Rubin’s rules . Furthermore, questions in which the individual participant marked their answer as “I don’t know” were coded as missing and included in the imputation. Pooled adjusted R2-squared and 95% confidence intervals were recorded for each model. In addition, the overall F-statistic was computed using an approximation based on chi-squared statistics [40, 41].
Exploratory analyses were used to measure the association between questionnaire responses to the canned, packaged, and microwave food questions and weighted servings and total grams of analogous food consumption reported in the dietary recalls. All statistical analyses were carried out using R version 3.6.1 .
Visit our sponsors
Wise (formerly TransferWise) is the cheaper, easier way to send money abroad. It helps people move money quickly and easily between bank accounts in different countries. Convert 60+ currencies with ridiculously low fees - on average 7x cheaper than a bank. No hidden fees, no markup on the exchange rate, ever.