Project Topics
Get Complete Project Material File(s) Now! »
Protocol and registration
This review was conducted in accordance with the preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines (38) and has been registered in the International Prospective Register of Systematic Reviews (PROSPERO; registration number: CRD42020165801).
Literature search strategy
An electronic search was conducted using the following seven databases: MEDLINE, Cochrane Library, Web of Science, Cinahl, PsycINFO, ProQuest and Open Grey. Search included two themes using the Boolean operator ‘AND’, and was adapted for each aforementioned database. The first theme was ‘bariatric surgery’. The second theme was ‘food preference’ with a focus on the following main conditions: quantitative measurement of consumption, frequency of consumption, hedonic evaluation, wanting or desire to eat measurement, evaluation of food choices or food preferences. The following is an example of the search syntax used for searching MEDLINE: (((« Bariatric surger* » or « Weight loss surger* » or « Obesity surger* » or « Metabolic surger* » or « Bariatric surgical procedure* » or « Bariatric procedure* » or « Bariatric operation* » or « Gastric bypass » or « Stomach bypass » or RYGB or RYGBP or « Sleeve gastrectomy » or « Gastric sleeve » or LSG) .ab,hw,kf,ti.) OR (« Bariatric Surgery »/ or « Gastric Bypass »/ or Gastrectomy/)) AND (((« Food preferenc* » or « Food choic* » or « Food reward » or « Reward system » or Liking or Wanting or « Food intake » or « Diet* record* » or « Diet* assessment* » or « Diet* survey* » or « Nutrition* record* » or « Nutrition* assessment* » or « Nutrition* survey* »).ab,hw,kf,ti.) OR (Food preferences/)). The detailed search strategy is available as Appendix 2.A. Limits were set to include articles published in English after 1960, which corresponds to the decade when bariatric surgery was developed as a treatment for patients with severe obesity. The final search was performed on 1st June 2020.
Inclusion and exclusion criteria
All observational and interventional studies were considered if they involved adult patients with severe obesity (BMI ≥ 35 kg.m-2) undergoing RYGB or SG. Studies had to compare pre- and post-operative measures of food preferences or to involve a control group of non-operated patients. Studies were excluded if they involved animals, children or adolescents. Results related to other types of surgeries (gastric band, omega-loop gastric bypass, biliopancreatic diversion with duodenal switch, etc.) were excluded.
Study selection and data extraction
All materials were retrieved and exported in Covidence systematic review software, (Veritas Health Innovation, Melbourne, Australia. Available at www.covidence.org). Duplicates were removed. Title and abstracts were screened independently by two reviewers (E.G., and S.B.). Any disagreement between them was resolved by a third reviewer (S.I.). The full text of retained studies was accessed and further screened according to the eligibility criteria by the two reviewers. Any disagreement was resolved through discussion between the two reviewers. For each study, the following data were extracted: name of the first author, year of publication, country where the study took place, study design, number of participants, age, preoperative Body Mass Index (BMI) and BMI at the end of the follow-up or if not available any weight loss related outcome, method of assessment, and measurement times. In this review, food preference measures (i.e., method of assessment) were considered as a relative choice a) between different macronutrients (e.g., the percentage energy intake from macronutrients and the frequency of foods consumption); b) between different foods, or food groups; and c) as measures related to food reward (e.g., liking, wanting or desire to eat). A narrative synthesis of the findings from the including studies were performed.
Quality assessment
The quality of the included studies was assessed independently by two reviewers (E.G., and S.B.) using the ‘Quality assessment tool for quantitative studies’ developed by the Effective Public Health Practice Project (EPHPP) (39,40). This tool is recommended for quality assessment in systematic reviews, which include uncontrolled studies (41). For each study, selection bias, study design, confounders, blinding, data collection method and withdrawals-dropouts were rated as strong, moderate or weak. The overall methodological quality was then rated as strong if there were no weak rating sub-domains, moderate if there was one weak rating and weak if there were two or more weak ratings. Any inconsistencies were resolved by communication between the two authors. We presented the results using a summary bar plot weighted by the study size (42). Such bar plot represents the proportion of information with a given level of bias for each sub-domain of the quality assessment scale. We also used a traffic light plot to represent the risk of bias for each sub-domain of the scale for each study. Quality data processing was done using the software Excel (Microsoft Corporation, 2018. Microsoft Excel, Available at: https://office.microsoft.com/excel).
Statistical analysis
We performed meta-analyses comparing the daily mean percentage energy from proteins, carbohydrates and lipids between preoperative and postoperative time points. If the percentage energy from the macronutrients was not available, it was calculated based on the quantity of macronutrients consumed and the total energy intake (4kcal/g for proteins, 4kcal/g for carbohydrates and 9kcal/g for lipids). When data were presented in graphs, we contacted the corresponding author to obtain the numerical data. When no response was received, we estimated the results using WebPlotDigitalizer (43). When the standard deviation (SD) was not available for either the amount consumed or the percentage of energy from the different macronutrients, it was calculated from the standard error of the mean (SEM) or the 95% confidence intervals (CI) (44). When results were available only for subgroups of patients (e.g., for women and men separately, or based on the type or the type of surgery), we combined the subgroups into a single group (44). Because our data were continuous, we used standardized mean differences (SMD) (45). As recommended for studies in the field of health sciences (46), we performed a random-effects-model to pool the effect size. Due to an important number of small size studies, we used the Hedges’ g estimator (47,48). Hedges’g estimator can be interpreted similarly to Cohen’s d: an effect size of 0.2 is considered as small, 0.5 is considered as medium, and 0.8 is considered as large (49). Data has been encoded on a positive effect size indicating that the proportion of the corresponding macronutrient was greater after bariatric surgery (or among postoperative participants than control) and a negative effect size indicating the opposite. A subgroup analysis was performed according to the type of surgery and the type of assessment method, if there were at least 3 studies per subgroup. We presented the results of the meta-analyses using forest plots. Heterogeneity was assessed using the I2 statistics. I2 statistic describes the percentage of variation across studies that is due to heterogeneity rather than chance. A percentage between 0%–40% represents a low heterogeneity; between 30%– 60% a moderate heterogeneity; between 50%–90% a substantial heterogeneity and between 75%–100% a considerable heterogeneity (50,51). In case of high heterogeneity, we ran the meta-analyses excluding outliers (i.e., studies for which the confidence interval overlap with the confidence interval of the pooled effect) and we performed influence analyses with the Leave-One-Out method (52) looking at the impact of each study on the results. Publication bias was studied by visual inspection of the funnel plot and with Egger’s tests of the intercept (53) to test for funnel plot asymmetry. In case of asymmetry, we performed Duval & Tweedie’s trim-and-fill procedure (54), which imputes the missing studies (e.g., small studies with low effect sizes), to correct the funnel plot for asymmetry and to pool the effect size with the imputed studies. All the statistics were performed using the software R (R Core Team (2019). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/) and the R package ‘dmetar’ (55).
Studies and patients’ characteristics
The details of the included studies are presented in Table 2.1. Forty-seven studies were prospective, 8 were cross-sectional and 2 were longitudinal retrospective. Fifty-one studies out of 57 were published since 2010, which corresponds to the increasing use of bariatric surgery as a treatment for obesity. The studies were carried out in 16 countries, of which the first 3 were the USA (N=13), Brazil (N=13), and France (N=5). This echoes with the world ranking of countries performing the most bariatric surgeries. The pooled population included in this systematic review was 2271 patients with a RYGB and 903 patients with a SG. Food preferences were assessed by 16 different methods, with the majority being based on food records (N=24), Food Frequency Questionnaires (FFQ) (N=12), and food recalls (N=11). The time points of assessment ranged from 6 days (d) to 10 years (yr), with the majority of measures being performed preoperatively (N=47), and at 1 (N=10), 3 (N=16), 6 (N=20), 12 (N=25) and 24 (N=8) months (mo) postoperatively.
Relative choice between foods or food categories
Table 2.2 summarizes the modifications of the relative choices between foods or food categories after bariatric surgery. A detailed table showing the results for each study is available in Appendix 2.C. There was a postoperative decrease in preference for red meat early after bariatric surgery, which persisted up to 1 yr postoperatively (92,97). It was also the case for other of proteins, such as hot dog, deli meat products, or hamburger (10,92). However, preference for other types of meat did not seem to be impacted by bariatric surgery (70,73,75,98), especially white meat (92) and ham (7,10,98). Two studies even showed an increase in preference for poultry in the longer term after bariatric surgery (70,98). Generally, the results related to fish and eggs were less consistent. For instance, studies showed a decreased preference (58,74), an increased preference (10,61,70,98) or a similar food preference (75,77,92,99) for fish after compared to before bariatric surgery. There was an increased preference for eggs (97) in the first postoperative period, however whether this is maintained up to 6 mo postoperatively remains less consensual (7,70,92,98,99). Concerning dairy products, the results are heterogeneous (58,73,75,77,92,99). However looking at specific types of dairy products, there was a decreased preference for high fat dairy products (e.g., whole milk, sundae, ice cream, cheese, ricotta cheese) (10,60,69,78), while there was an increased preference for the low fat dairy products (e.g., low fat milk, yogurt) in the first postoperative months (10,97). The fruit and vegetable categories are also very broad food categories, which is reflected by conflicting results concerning food preference modifications after bariatric surgery (7,10,58,61,73–75,77,92,97,99). Interestingly, the preference for fresh fruits did not seem to be altered postoperatively, while the preference for fruits with added fat and sugar (i.e., baked and fried bananas, fruit salad) was decreased (10). Regarding vegetables, there was an increased preference for cooked vegetables (70,98), while the preference for raw vegetables remained stable postoperatively (70,98). One study explored food preference for cooked beans, green peas and lentils and showed a decreased preference 6 mo after bariatric surgery. Most of the studies found a lower postoperative preference for starchy foods (10,58,61,69,70,73,77,92,97– 99), yet there were some with no postoperative modification (7,61,70,92,98), especially after 12 mo postoperatively. This trend was also found for drinks, miscellaneous (e.g., mayonnaise, soybean oil, etc.) and sweets.
Overall, when looking at broad food categories, the results are often conflicting between the studies. However, considering specific foods, there were some trends toward a higher preference for healthier food products to the detriment of unhealthy products (i.e., foods high in sugar, fat, salt and energy density (103)) in the first 6 to 12 months after bariatric surgery. The improvement in the quality of the diet towards healthier products has also been explored using the NOVA food classification (104) assigning foods into four categories according to their degree of processing. The higher the food processing, the less healthy it is considered. These studies highlighted a decreased intake per day of all food categories, but a slight increase in the contribution of the unprocessed or minimally processed foods in total energy intake (71,80).
Table of contents :
Chapter I. Introduction
1.1 The global burden of obesity
1.2 Obesity and the regulation of food intake
1.2.1. Obesity and the homeostatic model of appetite control
Appetite and the “Psychobiological” system
Obesity alterations of the psychobiological system
1.2.2. Obesity and the hedonic model of appetite regulation
Sensory domain
Reward domain
Alteration of the hedonic regulation of appetite in the case of obesity
Obesity and food preferences
1.3 Bariatric surgery as a treatment for obesity
1.3.1. Bariatric surgery: insight on its effectiveness and associated challenges
1.3.2. Bariatric surgery, appetite control and food preferences
1.4 Aim and objective of the thesis
1.5 References
Chapter II. A systematic review of food preference modifications after bariatric surgery
Abstract
3.1 Introduction
3.2 Methods and procedures
2.2.1. Protocol and registration
2.2.2. Literature search strategy
2.2.3. Inclusion and exclusion criteria
2.2.4. Study selection and data extraction
2.2.5. Quality assessment
2.2.6. Statistical analysis
3.3 Results
2.3.1. Study selection
2.3.2. Studies and patients’ characteristics
2.3.3. Relative choice between macronutrients, foods and food groups
Relative choice between macronutrients
Relative choice between foods or food categories
2.3.4. Choices related to hedonics
2.3.5. Quality of the total body of evidence
Quality
Publication bias
3.4 Discussion
2.4.1. Main findings
2.4.2. Comparison with other studies
Clinical studies and reviews
Pre-clinical animal studies
2.4.3. Strengths and limitations
2.4.4. Implications of findings and future research
3.5 Conclusion
3.6 References
Chapter III. Food preferences and their perceived modifications before and after bariatric surgery: a cross sectional study
Abstract
3.1 Introduction
3.2 Material and methods
3.2.1 Study population
3.2.2 Data collection and calculations
3.2.3 Statistical analysis
3.3 Results
3.3.1 Sample characteristics and surgery outcomes
3.3.2 Food preferences following bariatric surgery and perceived post-operative modifications
3.3.3 Food preferences between participants with and without taste or smell alterations
3.3.4 Food preferences according to the postoperative follow-up duration
3.3.5 Food preferences between patients in success and failure of bariatric surgery
3.3.6 Food rejections
3.4 Discussion
3.5 Conclusion
3.6 References
Chapter IV. Alterations in food reward following bariatric surgery
Abstract
3.7 Introduction
3.8 Material and Methods
3.8.1 Subjects
3.8.2 Data collection and calculations
Food reward
Behavioral assessment
Other covariates
3.8.3 Sample size calculation
3.8.4 Statistical analyses
3.9 Results
3.9.1 Characteristics of the subjects
3.9.2 Relationship between explicit liking for foods and bariatric surgery
3.9.3 Relationship between explicit wanting for foods and bariatric surgery
3.9.4 Relationship between implicit wanting for foods and bariatric surgery
3.9.5 Relationship between food choices and bariatric surgery
3.9.6 Relationship between fat and sweet appeal bias and behavioral traits
3.9.7 Relationship between food preference for ‘high-fat – sweet’ foods and behavioral traits
3.9.8 Comparison of food reward according to the follow-up periods
3.10 Discussion
3.11 Conclusion
3.12 References
Chapter V. Study of food preferences and eating behavior after bariatric surgery in a real context eating environment: conception and settings of the study
Abstract
5.1. Introduction
5.2. Subjects and methods
5.2.1. Study design
5.2.2. Sample size
5.2.3. Subjects
5.2.4. Recruitment
5.2.5. Primary and secondary endpoints
5.2.6. General organization of the study
5.2.7. Data collection
Anthropometric data and medical history
Gustatory function
Olfactory function
Eating behaviors assessment using TFEQ
Assessment of levels of hunger, desire to eat, prospective consumption and
satiation
Assessment of food preferences
5.2.1. Summary of the course of the experiment (see Figure 5.8)
5.2.2. Statistical analyses
Treatment of missing subjects
Descriptive statistics
Analytical statistics
5.3. Expected results
5.4. Conclusion
5.5. References
Chapter VI. General discussion and conclusion
8.1 Main results
8.2 Reflection about the model of homeostatic and hedonic regulation of appetite following bariatric surgery
8.3 Methodological considerations related to the measure of food preferences in the context of bariatric surgery
8.4 Strengths and limitation
8.5 Perspectives and recommendations
2.4.5. Direction for future research
Study food preferences in an ecological setting
Study other determinants of food preferences
Build a predictive model of weight loss success and failure
A consumer study to improve acceptance of healthy products
2.4.6. Recommendations for clinical practice and public health
Prevent protein energy malnutrition after bariatric surgery
Adherence to nutritional recommendations, barriers and facilitators for the longerterm follow-up
8.6 Conclusion
8.7 References
