Study population
The present cross-sectional evaluation is based on the Cypriot cohort of the MedWeight study, a registry of weight loss maintainers and regainers [11]. Eligible for participation in the registry were adult men and women aged 18–65 years of Cypriot ethnicity, who reported being at least overweight (Body Mass Index ≥ 25 kg/m²) and experienced an intentional weight loss of ≥ 10% of their maximum weight, at least 1 year before participation in the study. Each participant was classified as “maintainer” if his/her current weight was ≤ 90% of his/her maximum weight or “regainer” if his/her current weight was ≥ 95% of his/her maximum weight. Participants who had a current weight between 91% and 94% of their maximum weight were excluded so as to avoid overlapping between the two groups.
This study was conducted according to the guidelines laid down in the Declaration of Helsinki and the study protocol was approved by the National Bioethics Committee. The recruitment procedure held for two years (2/2018-2/2020) and it was communicated through press releases, advertisements in tv, radio and social media. All eligible participants signed the consent form prior participation to the study and were then advised to access a website platform (http://medweight.hua.gr) to fill in a series of questionnaires. In specific, volunteers were asked to report socio-demographic status such as marital status (single, married/cohabitating, divorced, widowed, then coded for married/cohabitating or not), occupational status (employed or not) and years of education. Eligible volunteers were asked to report physical/personal characteristics such as sex, age, weight, height, BMI, maximum weight, maximum BMI, initial loss, maintenance loss and duration of maintenance.
Assessment of dietary intake
Two telephone 24-h dietary recalls were conducted for each participant in order to assess dietary intake [23]. The recalls were performed by two well-trained dietitians within the period of 10 days for each participant, with weekdays and weekends proportionately represented among participants. Using the multiple-pass method, dietitians asked for all foods and beverages consumed the previous day [24, 25]. Dietitians were blinded regarding the participant’s maintenance status. All data were analyzed in terms of total daily energy intake by using the dietary analysis software SNPRO Nutrition Software (Cheapsoft Softwares, 2017).
Assessment of obesity-related behaviors and development of the Healthy Eating Behavior Index (HEBI)
The frequency of obesity-related eating behaviors was also assessed. We selected specific behaviors that have been previously reported to be associated either with weight loss and/or weight loss maintenance [26–33]. In specific, participants were asked about the frequency of eating out, eating with others, eating breakfast (rarely/never, 1–3 times per month, 3–6 times per week, daily, more than twice a day), the number of meals per day (1–3 meals, 4–5 meals or ≥ 6 meals per day), the number of main meals per day (1–3 main meals per day) and eating visible fat or meat skin (almost all, part of or none). Moreover, other questions related to food supplements (yes or no), eating rate (very fast, fast, medium, slow, very slow), food preparation (yes or no), person responsible for food preparation (mostly you or mostly others) and eating home-cooked meals (almost never, sometimes, often, almost always) were also included.
The eating behavior responses were collectively evaluated through the Healthy Eating Behavior Index (HEBI), a simple to understand and easy to use index developed specifically for the current study. This index consisted of 10 variables as listed above; 9 of these variables were related to eating behaviors and 1 variable was related to the frequency of self-weight measurement. In particular, the variables used for the development of the index were: eating out, eating with others and eating breakfast frequency, meals per day, main meals per day, eating rate, time spend on food preparation, responsible for food preparation, eating home cooked meals frequency and self-weighing frequency. The variables were then coded to dichotomous types. Each variable was scored with 0 or 1 (0 indicates less healthy behavior and 1 indicates more healthy behavior). The scoring system allowed the development of more distinct categories for each variable of the final index. As an example, for the eating out frequency, rarely/never and 1–3 times per month were coded to 1 and 3–6 times per week, daily or more than twice a day were coded to 0. The score range is 0–10: the higher the score is the more the individual is engaged to a behavior that it is not expected to promote weight loss maintenance.
Assessment of physical activity
The short version of the International Physical Activity Questionnaire (IPAQ) validated for the Greek population was used to assess physical activity [34]. Participants were asked to report high, intermediate and low intensity activities lasting ≥ 10min, as well as sedentary activity and time spent during these activities on a weekly basis.
Statistics
Using Q-Q plots we explored normality of distribution of data. Normally distributed values were presented as means and standard deviation (SD), non-normally distributed values as medians and interquartile range (IQR) and data from categorical variables as frequencies (in percentage). We explored differences between maintenance status in participants’ characteristics using independent t test or Mann Whitney rank tests, depending on the normality of the data, and chi-square tests for categorical variables. Differences between maintainers and regainers were tested by logistic regression models for categorical variables (results were expressed as odds ratio [95% confidence interval]).
Logistic regression models were performed using maintenance status as a dependent variable and HEBI total score as independent variables: Model 1 was adjusted for age, sex, and marital status (married or not); Model 2 was additionally adjusted for energy intake; Model 3 was additionally adjusted for physical activity (IPAQ total Met-minutes per week). Data analysis was carried out using SPSS Statistics 22.0; a P-value of 0·05 was considered statistically significant.