Photographic Food Recognition and Meal Size Estimation Before and After Roux-en-Y Gastric Bypass

Overview

Direct measurements of changes in food selection in humans after RYGB have been limited by the unreliability of patients, which poses significant methodological and conceptual challenges to researchers and study design. Self-monitoring requires time and effort, and many find tracking of dietary intake tedious, which contributes to attrition. Direct measurements, however, represent an essential component in the attempt to understand how RYGB alters eating and food preferences, but laboratory settings preclude a real-life environment. The aim of this study is to investigate changes in food preferences, total energy intake of the three primary macronutrients and meal patterns between obese women (BMI ≥ 35) before and after Roux-en-Y gastric bypass and lean (BMI ≤ 25) and obese (BMI ≥ 35) controls by means of photographic food recognition with a mobile application.

Introduction Obesity is one of the most important public health conditions worldwide. The morbidity can largely reduce life expectancy. Bariatric surgery for severe obesity is an effective treatment that results in the improvement and remission of many obesity-related comorbidities, cardiovascular benefits, as well as sustained weight loss and improvement in quality of life. Different surgical procedures for weight loss have been developed, with the most common ones being sleeve gastrectomy (SG) and Roux-en-Y Gastric Bypass (RYGB). RYGB is the most commonly used procedure for weight loss in Switzerland, even though in the last decades there has been a global shift towards SG. Independently from the applied procedure, bariatric surgery remains the most effective and durable intervention for obesity. Modern bariatric surgery has an excellent safety profile, even though complications in the short and long term have been reported. Extensive research also increased the understanding of the wide range of mechanisms underlying the effectiveness of bariatric surgery. The relative contributions of these mechanisms vary according to the chosen procedure, where altered satiety and hunger are considered important factors after RYGB. Other mechanisms have also been proposed, including changes in gut hormones, reduced appetite, increased satiation, increased energy expenditure, alteration in the gut microbiota and bile acids levels and composition, as well as changes in vagal nerve signalling. Changes in food preferences have also been implicated as an important candidate mechanism as patients prefer low-sugar low-fat food and report finding food, in general, less enjoyable after RYGB. Both humans and rats eat less and display altered food preferences after RYGB, but the underlying behavioural mechanism remains unclear. The aim of public health policy in the treatment of obesity should be the development of effective prevention and better therapies which would eliminate the need for surgery altogether. Until then, bariatric surgery is safe, effective, affordable, and with large room for improvement. Furthermore, a deeper understanding of the physiological changes underlying the improvement of the morbidity may help in developing new therapies. So far, direct measurements of changes in food selection in humans after RYGB have been limited by the unreliability of patients, which poses significant methodological and conceptual challenges to researchers and study design . Self-monitoring requires time and effort, and many find tracking of dietary intake tedious, which contributes to attrition . Direct measurements, however, represent an essential component in the attempt to understand how RYGB alters eating and food preferences, but laboratory settings preclude a real-life environment. Consequently, obesity researchers have attempted to develop new easier tracking methods such as photography of food items to monitor dietary intake both in the clinical setting and everyday life. These new methods seem more effective compared to others. For instance, mobile phone applications of photographic food recognition (PFR) such as SNAQ enable patients to record and submit consumed food items, portion sizes and frequency of consumption during a pre-defined period. However, this technology is still in its infancy and assessment procedures need to be validated by trained professionals. The doubly labeled water (DLW) method has been proposed as a gold standard to assess free-living total energy expenditure and can be used to monitor the body composition (BC) and total energy expenditure (TEE) of an individual under circumstances of energy balance based on dilution equations of the labelled isotopic elements in the body . Direct segmental multifrequency bioelectrical impedance analysis (DSM-BIA) is considered a practical, non-invasive, and relatively inexpensive method to quickly assess the BC27 and the estimations of BC with DSM-BIA are reported to be consistent with those of the DLW method also for obese patients. The joint measurement of TEE using the estimations of the PFR and the DLW method would serve the validation of PFR for the measurement of TEE in patients undergoing RYGB. PFR would become then a powerful tool for research and provide useful information for the design of appropriate individualized obesity treatment programs. Research question The aim of this study is to investigate changes in food preferences, total energy intake of the three primary macronutrients and meal patterns between obese women (BMI ≥ 35) before and after Roux-en-Y gastric bypass and lean (BMI ≤ 25) and obese (BMI ≥ 35) controls by means of PFR with the mobile application SNAQ. Methods This will be a single-centre observational case-control prospective cohort study on female human subjects with morbid obesity, living in Switzerland and undergoing RYGB (n=20), obese controls (n=20), and lean controls (n=20) before, 3, 6, and 12 months after surgery and beginning of the study respectively. The optimal number of participants included can be based on power calculations with an equivalence design and on the hypothesis that patients will have similar caloric intake to the lean control group one year after RYGB. The equivalent caloric intake between the two groups is defined by an equivalence range of 210 kcal/day with a SD of ±200 kcal. For achieving a power of 90% and a p=0.05 in order to exclude a difference of means \> 210 kcal/day, n=20 participants are required per group . Patients are enrolled in this study if (1) they are females aged ≥ 18 years, (2) have a body mass index (BMI) ≥ 35 kg/m2 without any previous bariatric surgery, (3) have a planned RYGB, (4) can move autonomously, (5) have a basic technological knowledge, (6) are fluent in German, and (7) are able to give their informed consent. Obese controls respect the same criteria, but have no plan to undergo surgery, which makes this group very hard to recruit. Lean controls have BMI ≤ 25 kg/m. The main purpose of the lean control group is to evaluate test stability over time and across multiple retesting. Patients who conform to the recruitment criteria are enrolled at the outpatient clinic of the University Hospital of Zurich. Obese and healthy controls are recruited by the study team. The study protocol is explained and written informed consent is obtained from all participants. The study protocol was approved by the Ethical Committee of the Canton of Zurich (BASEC-Nr. 2019-00952). TEE under free-living conditions is assessed using the DLW method. On the day before measurement, BIA is performed. On the first day of measurement, subjects perform urine collection at baseline and consume oral administration of DLW. After DLW administration, subjects collect urine respectively 3 and 4 h after administration and again after 7 days at the same time point of the second and third urine sampling . Subjects are instructed not to change their lifestyle or medication during the study period. Under free-living conditions for one week, overall dietary intake is registered by the participants using a mobile app for the photo-recognition of food items. Participants record their dietary intake over 7 days before, 3, 6, and 12 months after surgery. Data analysis The data of each patient will be organized as follows: time of the meal, total calories, total carbohydrates, proteins and fat, food group. This result has to be summed up for the final TEE calculation, in order to have the total amount of each variable per meal, per day and per week. The PFR data will be compared with DWL-based TEE for validation of the dietary assessment. Descriptive statistics will be used to describe demographic data, as well as to present meal-related information over different time periods. Change over time and from baseline will be assessed with ANOVA and t-tests, with correction for false discovery rates. Each meal-related parameter will be correlated with each other and with the 1-year weight loss, expressed as absolute total-body weight loss % excess BMI loss (% EBMIL). Explorative models will be generated using logistic regression to map meal-related information and optimal weight loss. All data will be analysed using RStudio. Limitations The present study has several limitations associated with the methods and the participants. First, the food data is self-recorded and prone to social desirability which may lead to inaccurate data. Indeed, the food preferences could be changed during the week of recording and some food items could may not be recorded. The database of food items and calories embedded in the app also may not be as accurate as expected. To reduce this limitation, the participants are asked to report any missing food items or incongruences reported by the app, and feedbacks are discussed with the app developer. Dietary self-monitoring is a key component in weight loss programs, and frequency of self-monitoring with a smartphone app is strongly correlated with weight loss. However, lean and obese controls of this study are not following any diet with the aim of weight loss and participants will record their food intake for a period that should not be long enough to induce changes in habits. Recent studies that attempted to validate the energy intake estimations by using PFR compared to the DLW method with obese participants found that this method was not accurate in estimating energy intake compared to DLW. However, in one case the TEE was measured in pregnant women who showed low compliance for reporting frequent intake of small meals. In the other one, the TEE was measured in minority preschool children whose intake was recorded by caregivers. Furthermore, both studies were cross-sectional studies. It is also reported that participants who use their own phones compared with participants who use borrowed phones capture more images and have higher accuracy , therefore such variable should also be considered in the data analysis. Technical problems with the phones may also be a source of underestimation. Training of the participants, constant communication and technical support must be provided in order to increase accuracy. Relevance of the expected outcome for research or practice This study would provide for the first time a direct and precise measurement of dietary intake in obese patients before and after RYGB. The direct PFR measurement with the mobile application SNAQ, validated by means of the DLW method, will allow to compare the energy intake of the three primary macronutrients per day (kcal/24h) and per meal (kcal/meal) between RYGB patients and controls. Furthermore, it could produce more fundamental information about human behaviour which will allow to translate for humans behavioural changes previously reported in rats. It is expected that patients will postoperatively and progressively adjust their relative caloric intake and food selection with a decrease of fat and sugars and an increase in complex carbohydrates, while non-operated controls will exhibit a stable, unaltered food selection and intake. The proposed research project is characterised by conceptual and methodological innovation. Food preferences and relative macronutrient intake within and between meals in patients after RYGB will be directly measured in a real life setting with PFR. This approach will overcome or reduce many of the limitations reported so far in the literature. If the expected outcome is confirmed, such a finding has the potential to fundamentally change clinical practice as regards to current post-bariatric nutritional counselling. Moreover, the results may also contribute to the development of less invasive interventions in the treatment of obesity and would justify a neurogenetic approach with animal models of obesity for mapping and studying neural circuits and linking specific brain activities to specific behaviours involved in obesity, as specified by the BRAIN initiative.