Weight loss in adults following bariatric surgery, a systematic review of preoperative behavioural predictors

Bariatric surgery is effective in treating obesity in many cases, yet as many as 50% of patients may not achieve the desired weight reduction. Preoperative modifiable behavioural factors could help patient selection and intervention design to improve outcomes. Medline, EMBASE, Cochrane Library and PsychINFO were searched to identify studies published between 1 January 2008 and 14 February 2019 reporting on preoperative modifiable behavioural factors associated with postoperative weight loss, with minimum 2 years follow‐up. A total of 6888 articles were screened, 34 met the inclusion criteria. Maladaptive eating behaviours (MEB), preoperative weight loss (PWL), and tobacco use were reported 21, 18, and 3 times respectively. Physical activity and substance abuse were each reported once. Most articles on PWL (72.2%) and MEB (52.4%) reported no association. Positive associations were reported in 22.2% and 14.3% of articles for PWL and MEB respectively. Negative associations were reported in 5.6% and 33.3% of articles for PWL and MEB, respectively. Marked heterogeneity in outcome reporting hindered quantitative synthesis. The current paucity of evidence amenable to synthesis leads to ongoing uncertainty regarding the size and direction of association between PWL and MEB with outcomes following bariatric surgery. Long‐term studies with common reporting of outcomes are needed.


| INTRODUCTION
Bariatric surgery rates have increased following the global rise in patients with obesity, as well as the recent advances in laparoscopic techniques. 1 Evidence supports the efficacy of bariatric surgery to produce safe large-scale weight loss, 2,3 yet outcomes are not always favourable. Reports range from 10% to 50% of patients not achieving the desired weight loss following surgery. 4,5 This results in a reemergence of medical and psychological comorbidities and a decrease in quality of life. 6,7 Identifying predictors of postoperative outcomes has proven difficult. 8 Many predictors have been proposed and investigated including preoperative body mass index (BMI), age, gender, preoperative weight loss, eating behaviours, history of psychiatric disorders and history of sexual abuse. 8 Not all predictors share the same implications for patients. Most factors are not modifiable and can act as barriers to accessing treatment. As such, increasing emphasis is now being placed on identifying modifiable preoperative predictors. 9 A subset of these are behavioural factors such as preoperative weight loss, eating behaviours, physical activity, tobacco use and substance abuse. 8,[10][11][12][13] Findings in this area could subsequently be used to formulate inter- Minimum follow-up intervals are important in evaluating the results of bariatric surgery. Most patients will experience substantial weight loss in the first few months following surgery. 15,16 This trend tends to stop with a plateau of weight loss seen during the first and second year postoperatively. [17][18][19] A 2-year interval after surgery has been proposed as the minimum amount of time before reliably evaluating postoperative weight loss outcomes. 20 The aim of this review is to identify and investigate the modifiable preoperative behavioural factors associated with postoperative weight loss at least 2 years postoperatively in adult patients with obesity undergoing bariatric surgery.

| METHODS
A systematic review of the literature published between 1 January 2008 and 14 February 2019 was conducted using searches of Medline, EMBASE, Cochrane library and PsychINFO. The searches were carried out between January and February 2019. Separate search strategies were developed for each database (Appendix A). Our inclusion criteria encompassed studies reporting on modifiable preoperative behavioural predictive factors of adults with obesity undergoing bariatric surgery. If a study included adults as well as patients <18 years of age, these were also included. We excluded studies published prior to 2008, not published in English, or with a post-operative follow-up of less than 24 months. Case series and case reports were also excluded.
Two independent reviewers performed the screening. Full text articles were retrieved for all screened results. Conference abstracts were accepted only if they reported sufficient data required for extraction. Review articles were not included, but their references were manually searched to identify other studies that met our inclusion criteria. Data regarding type of publication, study and predictive factor characteristics were collected. Predictors included were ones deemed to be modifiable behaviours that could be addressed preoperatively via behavioural interventions. When there was missing data regarding postoperative weight loss, corresponding authors were emailed with requests for that information.
Due to significant heterogeneity in the types of eating behaviours reported and the tools used to measure them, eating behaviours were grouped to allow comparisons to be made. When available, the specific eating behaviours and assessment methods were collected and stated. We compared these in terms of their effect on postoperative weight loss.
Each predictor was outlined according to their associative direction (positive, negative and neutral) and level of statistical significance.
This methodology was selected in line with similar past reviews by Livhits et al 8,12 who faced comparable difficulties in analysing studies with considerable variability in predictor and outcome reporting.
Assessment of study quality was performed using the Newcastle-Ottawa Quality Assessment Tool (NOQAT), see Appendix B for marking criteria. An overall quality assessment was made on the basis of the available criteria to be evaluated. If all were of acceptable quality, the article was deemed of good overall quality. If only one criterion was of low quality, then the study was deemed of fair quality. If more than two criteria were of low quality, then the study was deemed of poor quality.
The protocol for this systematic review was prospectively registered with PROSPERO (CRD42019119358) 21 and reported according to PRISMA guidelines. 22 Ethical approval was not required.
What is already known about this subject?
• Bariatric surgery is an effective therapy in the management of patients with obesity, yet in some patients significant weight loss is not achieved or maintained in the longer term.
• Multiple predictors of weight loss have been investigated, a proportion of which are behavioural and modifiable meaning they could be utilized preoperatively to optimize outcomes.
• Modifiable behaviours are already being used in other surgical disciplines through prehabilitation programmes to optimize outcomes.
What this study adds?
• Highlights ongoing ambiguity and lack of strong evidence regarding behavioural predictors, perhaps challenging the strict use of behavioural factors as barriers to surgical treatments • Reinforces urgent need for use of common outcome measures to allow robust synthesis of findings.
• Demonstrates paucity of literature reporting long term outcomes following bariatric surgery.

| RESULTS
A total of 6888 initial records were identified following removal of duplicates. After abstract screening and full text review, 34 studies were included in the final review ( Figure 1). Using the NOQAT, the majority of the included studies were found to be of poor quality (Table S1). There were insufficient studies using comparable preoperative and outcome variables of adequate quality to be combined in a meta-analysis.
The most common procedure reported was Roux-en-Y Gastric  This was not the case in two cohorts with veteran soldiers where there was a majority of male patients. Table 1 summarizes the characteristics of the included studies.
Five factors were reported a total of 44 times within the 34 articles.
They were preoperative weight loss, eating behaviours, tobacco use, physical activity and substance abuse. Eating behaviours and preoperative weight loss were reported 21 and 18 times, respectively. The rest of the reported factors were in the minority with three reports for tobacco use, one for physical activity and one for substance abuse.

| Preoperative weight loss
Eighteen studies reported on associations of preoperative weight loss on postoperative weight loss. Thirteen articles found no association. 23 in Table 4 In clinic Four studies showed a positive association, [36][37][38][39] and one study found a negative association. 40 Table 2  After excluding studies of poor quality, seven of the 18 studies remained. Four of these were deemed good quality 30,32-34 ( Table 2). All of these showed a no association between preoperative weight loss and postoperative weight loss. Three studies were deemed fair quality, 23 The methods of achieving preoperative weight loss among studies varied greatly (Table S2). This ranged from advice at a preoperative appointment to supervised programmes of dieting and physical exercise. Two studies did not specify the method of weight loss but described the number of previous weight loss attempts and maximum weight loss. One study did not specify the method of preoperative weight loss or how it was measured.

| Tobacco use and substance abuse
Tobacco use was reported as a potential predictor in three studies. Two studies found no association between current tobacco use and weight loss after surgery. 11,53 One study found it to be negatively associated with weight loss. 33 Adams et al 11 also reported a negative relationship between substance abuse and weight loss within its cohort (Table 4).
Excluding studies of poor quality, only the study by Andersen et al remained. 33 Using linear regression models in their cohort of 160 patients they found a B of 13.3 (4.3-22.4) (P = .004) between smoking status and postoperative weight loss at 24 months follow-up.

| Physical activity
Only one poor quality study examined the relationship between the level of preoperative physical activity and postoperative weight loss. It did not report on the way physical activity was identified or measured. Increased physical activity levels were reported to be associated with higher postoperative weight loss. 10 The measure of association was not reported. Of note, this article reported findings from a younger study population, with a mean age of 39.9, and the weight loss results were self-reported (Table 4).   [15][16][17][18][19] Of note, in patients with BED a "honeymoon period" between 12 and 18 months postoperatively has been described as the period before differences in outcomes compared to those patients without preoperative BED become apparent. 48   become almost historical. 55 LAGB is also falling out of favour following disappointing weight loss outcomes and significant postoperative complication rates requiring re-intervention. 29,32 LSG is becoming one of the most commonly performed operations. 1 The main limitation was substantial heterogeneity in predictor and outcome reporting which lead to difficulties in data synthesis.

| DISCUSSION
This includes difficulties in performing meta-analyses or calculating formal measures of heterogeneity, such as I. 2 Eight different weight loss-related outcomes were used, the most common being %EWL. In addition to the problem of comparisons and meta-analysis, these varying measurement methods can also lead to different interpretations of results. The same amount of weight loss can be statistically significant using one measurement method and not significant if using another, as seen in Sethi et al. 36 An additional limitation is the risk of confirmation bias in the literature with regards to preoperative weight loss. It is common practice for preoperative weight loss to be a prerequisite for bariatric surgery. This is separate and in addition to a history of weight loss attempts prior to being considered for surgery. This leads to reporting of findings derived only from people that lost weight prior to their operation. 37 Further studies reporting the long-term outcomes of bariatric surgery are needed. In our experience there were more than twice as many studies reporting outcomes with less than 2 years of follow-up.
This is particularly the case in literature reporting on modifiable pre- Finally, more information is required to elucidate the role of alcohol use, tobacco, use, and substance abuse as predictors of postoperative outcomes, especially their relationship with behaviours and eating disorders in modulating bariatric outcomes. 58 In the postoperative period, there is some evidence that preoperative maladaptive eating behaviours is associated with vulnerability to other addiction disorders. 8

| CONCLUSION
The search for preoperative predictors of postoperative outcomes continues within bariatric surgery. This review suggest that preoperative weight loss is likely to be a positive or non-predictor of postoperative weight loss and maladaptive eating behaviours may be a negative or non-predictor. Tobacco use may be a negative predictor.
There was insufficient data to make conclusions on physical activity, and substance abuse as predictors. The main strengths of this review were the inclusion of recent studies, and a minimum follow-up interval of 2 years after surgery. The main limitation was widespread heterogeneity and inconsistent outcome reporting which made it difficult to analyse the currently available evidence. Strong clinical implications are difficult to discern, although the use of preoperative weight loss appears to be beneficial, while the strict use of preoperative maladaptive eating behaviours as barriers to being considered for surgery may not be appropriate in all cases. Further studies investigating these behaviours with >2-year postoperative outcomes are needed. The use of common predictor and outcome measures in further studies is vital for the meta-analysis of future evidence.