Alcoholics

Generation of items: The 35 items from the CEBQ were changed from the “My child …” format to a self-complete “I ...” format (e.g. “My child loves food” was changed to “I love food”) and the original response options (‘never’, ‘rarely’, ‘sometimes’, ‘often’ and ‘always’) were retained. Ten researchers working in the area of Energy Balance completed the self-report version of the CEBQ and discussed their experiences. The researchers described how the Desire to Drink scale was difficult to complete. Items from the CEBQ such as “My child is always asking for a drink” had been adapted to “I am always asking for a drink” for the AEBQ and it became unclear what type of drink (i.e. alcoholic versus non-alcoholic) was being referred to. Additionally, the item “My child is always asking for food” from the FR construct in the CEBQ, which became “I am always asking for food” in the AEBQ, was difficult for adults to relate to. It was therefore agreed that the 3 items from the Desire to Drink scale, and the “I am always asking for food item” from the FR scale should be eliminated.
Further refinement of the questionnaire took place in 3 group discussions with a panel of clinical psychologists, behavioural scientists, dieticians, and authors of the original CEBQ. The panel initially reviewed the remaining items from the original CEBQ for any obvious gaps or additional problem areas. It was suggested that a measure of hunger experience (H), which could not be captured by the CEBQ because parents are unable to accurately determine their child’s experienced level of hunger, should be added (Wardle et al., 2013 ). It was also agreed that aspects of Food Responsiveness that related to food cues a parent would not have been able to comment on should also be included. Following this discussion, potential items for the Hunger scale were identified for review, and additional items for the Food Responsiveness scale were developed by the authors for piloting. Finally the panel reviewed all included and excluded items to ensure no further additions/removals were felt to be required. A group consensus was reached and the total number of items following these additions, and the removal of the Desire to Drink scale, was 49.
Piloting. The extended version of the AEBQ was piloted online in an opportunity sample of 49 adults (21–73 years old), 36 women (79.6%) and 13 men (20.4%). Colleagues at University College London were asked to circulate a link to the questionnaire to their friends and family from a range of professional backgrounds. Participants were invited to comment on each individual item and on the questionnaire as a whole. Piloting led to changes in the response options from ‘never’, ‘rarely’, ‘sometimes’, ‘often’ and ‘always’, to ‘strongly disagree’, ‘disagree’, ‘neither agree not disagree’, ‘agree’ and ‘strongly agree’ because participants commented that the original response options did not fit with the questions. The new response options were tested with a small convenience sample (two females and three males, aged 31 ± 7 years). This answer format appeared to be more meaningful and better understood by this sample.
Piloting also led to the deletion of the item “Given the choice, I would always have food in my mouth” because several participants commented that it “sounded a bit odd” or was “over the top”. A second item (“I am interested in food”) was eliminated because participants reported they found the meaning ambiguous. The remaining 47 item version of the AEBQ was included in the Principal Component Analysis (PCA).

Data were double-entered in OpenClinica, cleaned, and exported to STATA 17.0 (StataCorp, College Station, TX, United States) for analysis. Descriptive analysis was used to understand baseline group characteristics. We determined the proportion that became pregnant as the number that tested positive for pregnancy during study follow-up divided by the total number of participants who tested negative for pregnancy at enrollment (i.e., 100% of our study population, by design). To estimate the incidence of pregnancy, we used the Kaplan–Meier technique to estimate the time to test positive for pregnancy after enrollment into the study and presented the estimated incidence rate per 100 person-years. To estimate baseline factors associated with incident pregnancy, we used Poisson regression analysis with robust error variance both for unadjusted and adjusted levels; variables for which the association attained p < 0.15 at the unadjusted level were selected for the multivariable model except for age, which was considered an a priori confounder. We tested for collinearity between the baseline audit score and the question “have you had any alcoholic drink in the past month?” The results showed no collinearity (variance inflation factors <5). However, due to the very small number of participants with the outcome for the high-risk alcohol use category, we did not consider the baseline audit score for the multivariable model. Variables with a p-value <0.05 at multivariable analysis were considered significantly associated with incident pregnancy. Results are presented as adjusted risk ratios (aRR) with p-values and 95% confidence intervals (CIs).