Introduction

Weight loss surgery (WLS) is the only intervention that has been shown to have significant long-term efficacy in treating severe obesity, producing meaningful and durable improvements in both weight and associated medical comorbidities, risk factors, and quality of life [1–3]. When evaluating WLS outcomes, it is important to consider both initial weight loss and long-term weight loss/regain, which arguably are two distinct outcomes. Long-term follow-up studies support that 27–31.5 % of total weight is lost, but after achieving nadir (or lowest) weight, regain may occur [2, 3]. Importantly, within 2 to 4 years post-surgery, health care cost savings due to lower health care utilization are estimated to offset the initial cost of the surgery [4], largely attributable to improvement or resolution of obesity-related conditions including diabetes, hyperlipidemia, hypertension, and sleep apnea [5]. However, the durability of these health improvements is dependent upon the maintenance of weight loss over the long-term [6].

Although WLS is the most effective weight loss treatment for severe obesity, not all WLS patients experience optimal outcomes. For example, higher weight class groups have poorer initial weight loss outcomes than lower weight class groups [7]. In addition, initial weight loss may not be strongly associated with long-term weight loss [8], indicating that initial weight loss and weight regain (WR) are distinct post-WLS outcomes. A 5-year prospective study revealed that 46 % of Roux-en-Y gastric bypass (RYGB) patients begin to regain weight by 2 years post-WLS, with the number experiencing WR increasing over time (64 % by 4 years) [9]. A more recent 5-year follow-up of RYGB patients, with better retention (91 %), observed a similar rate and amount of WR [1]. WR is an important consideration because as weight is regained, obesity’s medical comorbidities may reemerge or, if they have persisted, worsen, with an associated decline in health-related quality of life [10]. Therefore, although weight loss has been the primary outcome of interest in most studies to date, WR may well be the outcome that truly matters with respect to the ultimate cost-benefit balance associated with the procedure.

WR is generally presumed to be related to behavioral factors such as suboptimal eating and exercise behaviors [11, 12]. Other data suggest that WR may be due primarily to anatomical and physiological adaptations that occur in the years following surgery (e.g., pouch dilation) [13]. The literature suggests that a number of factors predict WR and they may be either patient-related (poor diet quality, psychopathology, pathological eating behaviors, physical inactivity) or anatomical (dilation of stoma, pouch, or sleeve), generally falling into five domains: nutritional nonadherence, mental health problems, endocrine/metabolic factors, physical inactivity, and anatomical/surgical factors [13], although efforts to empirically establish a link between the latter and WR have yielded mixed results [14]. As such, the study of protective and risk factors for WR is in its infancy, warranting additional research, particularly that of an exploratory nature, to replicate what has been reported and identify potentially important new predictors.

A distinction should be made between pre- and post-surgical factors that may be related to WR. For instance, several pre-operative variables are associated with suboptimal nadir and long-term weight outcomes, including higher baseline BMI, iron deficiency, use of diabetes medication, no history of smoking, aged >50 years, and the presence of liver fibrosis [8]. However, attempts to relate pre-operative psychosocial factors to post-operative weight outcomes have been less conclusive. For example, pre-operative mood symptoms, such as depression, have not been consistently found to be related to post-surgical weight loss outcomes [15–17]. On the other hand, recent data more definitively suggest that a number of post-WLS psychosocial variables may be prognostic of poorer initial WL and/or WR, including depression, increased food urges, binge eating, emotional eating, “grazing” (i.e., continuous snacking), poor well-being, alcohol/drug misuse, dietary disinhibition/restraint, and lack of planned exercise [17–25]. Studies of WR suggest that consuming larger quantities of sweet foods and snacks [13], large portions of soft foods or liquids (e.g., alcohol and sugar-sweetened beverages) [26], grazing [20], and/or nocturnal eating [20, 27] may be contributors to poorer long-term WLS outcomes.

The distinction between factors contributing to initial weight loss vs. WR has been examined more thoroughly in the behavioral weight loss literature, with different types of behaviors found to be associated with initial weight loss vs. weight loss maintenance [28], but this has been examined less in the WLS literature. Poor long-term outcomes following WLS are frequently attributed to poor adherence to pre- and post-operative behavioral recommendations, particularly those related to eating and physical activity. Nonadherence to behavioral recommendations is prevalent following bariatric surgery, with lack of exercise being the most common domain of nonadherence [29, 30]. These findings take on particular significance given that consistent physical activity after surgery is associated with better weight outcomes [31, 32]. Indeed, in one study, WLS patients who performed the recommended amount of physical activity had the lowest risk of WR [12].

Given the substantial risk of poor initial and/or long-term post-operative weight loss, and attendant suboptimal medical and quality of life benefits of surgery for some patients, it is critical to identify variables related to WLS outcomes. Therefore, the present study evaluated the relationship of well-documented (e.g., health, diet, physical activity) and theoretical (i.e., addiction-like tendencies regarding food and other substances) variables to weight loss nadir and subsequent WR following RYGB surgery.

Methods

A sample of 113 adult bariatric patients was recruited via postal mail from the Bariatric Center of Excellence at St. Vincent Carmel Hospital (Carmel, IN) and by emailing participants from our previous studies—most of whom were recruited through Henry Ford Health Systems (Detroit, MI)—who had given permission to be recontacted for future studies. Because there is some variability in weight loss outcomes between procedures, with RYGB producing significantly greater weight loss in most studies [33, 34], we included only RYGB patients. Participants were surveyed about a host of pre- and post-operative variables including weight history, medical comorbidities, drug and alcohol use, physical activity, food addiction, depression, and other weight-, health-, and eating-related variables. Twelve participants were excluded due to insufficient survey data (e.g., not supplying weight history or other specific measurement data); four participants were excluded because they had undergone procedures other than RYGB, or it could not be confirmed that RYGB was the procedure that they had undergone, resulting in a final sample of 97 participants. Study procedures were approved by the appropriate institutional review boards, and informed consent was obtained from all participants.

Measures

We defined nadir weight loss as the percentage of total weight lost at the patient’s lowest self-reported post-operative weight (%TWL). We chose this definition based on suggestions in the literature that this is the preferred mode of operationalizing weight loss after WLS, as %TWL is significantly less affected by initial BMI [7, 35]. There is no clear consensus on how to define WR. Previous investigators have used kilograms regained, change in BMI from nadir BMI, or percent of lost weight regained to assess WR. We defined WR as

$$ \frac{\left(\mathrm{initial}\kern0.5em \mathrm{maximum}\kern0.5em \mathrm{number}\kern0.5em \mathrm{of}\kern0.5em \mathrm{pounds}\kern0.5em \mathrm{lost}\right)-\left(\mathrm{current}\kern0.5em \mathrm{number}\kern0.5em \mathrm{of}\kern0.5em \mathrm{pounds}\kern0.5em \mathrm{lost}\right)}{\mathrm{initial}\kern0.5em \mathrm{maximum}\kern0.5em \mathrm{number}\kern0.5em \mathrm{of}\kern0.5em \mathrm{pounds}\kern0.5em \mathrm{lost}} $$

In determining a cutoff for clinically significant WR to create a categorical variable, we chose the definition used by Abu Dayyeh, Lautz, and Thompson [36]: >20 % of maximum initial weight lost. As shown in Table 1, approximately half of our participants met this criterion.

Table 1 Participant characteristics
Full size table

Pre- and Post-RYGB Variables

The Alcohol Use Disorders Identification Test (AUDIT) [37] was used to assess alcohol intake patterns and symptoms of alcohol misuse pre-and post-WLS. The AUDIT was developed as a screener for hazardous and potentially harmful alcohol consumption. A total symptom severity score is obtained by summing all scale items. A score of 8 or more is the recommended indicator of hazardous and/or harmful alcohol use. Scores of 20 or more are suggestive of alcohol dependence. The AUDIT’s reliability is supported by high internal consistency (α = 0.83) [38]. The AUDIT has also been found to perform equally well or better in comparison to other alcohol screening instruments in detecting problematic alcohol use [38–40]. To evaluate drug use, additional items assessed whether participants regularly (i.e., more than twice a week) used opiates (illicit or prescribed, not differentiated), hallucinogens, amphetamines, inhalants, marijuana/THC, cocaine/crack, and PCP. The present study used the total number of endorsed substances (ranging from 0 to 7) to operationalize pre- and post-WLS drug use.

Medical comorbidity was assessed in terms of whether participants endorsed pre-surgical and/or current hypertension, diabetes, congestive heart failure, angina, high cholesterol, gout, PCOS, sleep apnea, back pain, and musculoskeletal disease/arthritis. Comorbidity, both pre- and post-WLS, was operationalized as the total number of endorsed medical complaints (ranging from 0 to 10).

Post-RYGB Variables

In addition to the pre- and post-RYGB measures described above, several post-operative measures were included. The Yale Food Addiction Scale (YFAS) [41] was used to assess post-WLS “food addiction.” The 25-item YFAS was developed to operationalize the construct of food addiction by including items corresponding to the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition, Text Revision (DSM-IV-TR) [42] criteria for substance abuse or dependence, applied to food intake rather than substances. The YFAS has demonstrated adequate internal reliability (α = 0.75) and convergent and discriminant validity [41], including validity for use with bariatric patients [43, 44]. The present study used a continuous symptom count index (i.e., number of abuse/dependence symptoms endorsed) with possible scores ranging from 0 to 7.

The Patient Health Questionnaire (PHQ-9) [45] was used to assess post-WLS depression. The PHQ-9 is a nine-item self-report instrument based upon the DSM-IV-TR diagnostic criteria for a major depressive episode. The PHQ-9 assessed participants’ experience with depressive symptoms over the past 2 weeks. Items are scored on a 0–3-point scale, with total scores ranging from 0 to 27. Specific cut points exist for mild (5–9), moderate (10–14), moderately severe (15–19), and severe (20–27) depression. Reliability is supported by good internal consistency (α = 0.89) and test-retest values (ICC = 0.84 within 48 h) [45]. The PHQ-9 was originally validated in medical settings [45], but more recently, it has been validated for use in the general population [46].

The Bariatric Surgery Self-management Questionnaire (BSSQ) [47] assessed adherence to recommended post-WLS eating and physical activity over the course of the previous week. The BSSQ includes seven domains: (1) eating behaviors; (2) fluid intake; (3) physical activity; (4) dumping syndrome management; (5) supplement intake; (6) fruit, vegetable, and whole grain intake; and (7) protein intake. Subscale and total scores were converted to a 0–100 range for ease of interpretation, with higher scores indicating greater adherence to recommended behaviors. Previous research has established good internal reliability coefficients and 2-week test-retest reliabilities for all BSSQ subscales and total score [47, 48].

To assess post-WLS eating behavior and eating disorder pathology, we used items developed for use in follow-up evaluations of bariatric patients [49] and existing measures of eating pathology, including the Questionnaire on Eating and Weight Patterns—Revised (QEWP-R) [50], a validated, widely used measure of binge eating and purging [50, 51]. Our questionnaire also incorporated the 14-item Nocturnal Eating Scale, a validated instrument that assesses night eating syndrome (NES) [52].

In addition to the physical activity item on the BSSQ, the World Health Organization’s Global Physical Activity Questionnaire version 2 (GPAQv2) [53] was used for a more detailed assessment of post-WLS physical activity. The GPAQv2 assesses frequency and duration of physical activity (moderate to vigorous physical activity) across four domains: work-related physical activity, transportation-related activity, leisure time recreational physical activity, and sedentary behavior. This measure has good test-retest reliability [53].

Analytic Plan

Statistical analyses were conducted using the Statistical Package for Social Sciences (SPSS), version 21.0 for Windows. A correlation matrix was examined to assess the association of candidate predictor variables with nadir weight (%TWL) and WR (regain of ≥20 % initial weight loss, or <20 % initial weight loss). Analysis for normality revealed that WR, when functioning as a continuous variable, was positively skewed. Thus, all WR analyses utilized WR as a categorical variable. After identifying a number of significant associations, stepwise linear and logistic regression analyses were conducted to estimate the relationship between predictor variables and weight loss outcomes. Missing data were minimal, but when present, cases were deleted listwise if they did not have complete data for all variables included in any given analysis.

Results

The sample consisted of 97 participants (M age = 56.11 years, SD = 11.26) who were predominantly white (92.8 %) females (77.3 %), all of whom had undergone RYGB a mean of 8.76 years previously (SD = 3.69). Given the very limited gender and ethnic variability, these variables were not be included in our statistical analyses. Additional sample characteristics are summarized in Table 1.

To verify that WR groups did not differ at time of surgery, we compared them on their reports of time since surgery, age at surgery, age first felt had a weight problem, highest BMI prior to surgery, lowest BMI prior to surgery, and BMI at time of surgery; none of these comparisons were statistically significant. A number of variables, however, were associated with either nadir weight (%TWL) or WR (regain of ≥20 % initial weight loss vs. <20 % initial weight loss), or both variables with significant or nearly significant associations with either weight outcome are depicted in Table 2. Notably, correlations for eating pathology based on QEWP-R scores could not be computed due to undimensional scores across the sample. Predictors for the two types of weight loss outcomes differed. For nadir weight, the number of post-WLS medical comorbidities and pre-WLS drug use were significant negative and positive predictors, respectively. In contrast, multiple post-operative variables were positively associated with the likelihood of significant WR, such as the number of medical comorbidities, and symptoms of depression, alcohol abuse, food addiction, and nocturnal eating. Additionally, several post-operative behavioral adherence variables (eating enough protein, fruits, vegetables, and whole grains; adequate fluid consumption; and engaging in physical activity) were inversely associated with the likelihood of significant WR. All significant bivariate predictors were entered as candidates for final stepwise linear and logistic regression models predicting nadir weight and WR, respectively. In the final model predicting nadir weight, only the number of post-WLS medical comorbidities remained significant (R 2 = 0.09, F(1, 95) = 9.49, p < 0.01). In the final model predicting WR, only post-operative depression and avoiding sweets remained significant (R 2 = 0.22, p < 0.01).

Table 2 Correlates of nadir weight (%TWL) and WR (>20 % from nadir) in bivariate analysis
Full size table

Discussion

The present study evaluated the relationship of relatively well-documented (e.g., health, diet, physical activity) and theoretically relevant but previously unexamined variables (e.g., addictive elements of food and substances) to nadir weight and WR following RYGB. Echoing findings of a previous study in which 59 % of RYGB patients reported significant WR [36], approximately half of our participants reported significant WR (>20 % of maximum weight loss). Only pre-WLS drug use and number of post-WLS medical comorbidities were related to nadir weight. WR, on the other hand, was associated with several variables in addition to number of post-operative medical comorbidities. These included night eating; adequate protein, fruit, vegetable, and whole grain intake; avoiding sweets; fluid consumption; physical activity; depression; and post-operative alcohol abuse.

Importantly, predictors were different for the two different weight loss outcomes. In particular, it is noteworthy that almost all of the predictors of WR are modifiable factors. Thus, the results of our study constitute support for the proposition that more systematic and longer-term monitoring of (and, if needed, interventions for) post-operative behavioral adherence, mood, and alcohol use may have some utility in minimizing long-term WR after WLS.

There are some limitations to the present study. We relied on self-report to assess variables that may be more accurately measured objectively or via interviews, particularly psychiatric and medical comorbidities. This is evident in some of our findings; for example, the physical activity variables derived from different measures (e.g., BSBQ vs. GPAQ) were not consistent in their associations with the WR outcome variable. Emerging research suggests that within the bariatric population, self-reported estimates of physical activity demonstrate very little correlation with data obtained by objective measures (e.g., accelerometer) [54]. Inclusion of accelerometer or other objective measures of physical activity would represent an improvement for future studies. Additionally, several variables were not assessed pre-operatively, simply due to concerns about the sample’s ability to validly self-report about certain subjective variables for a time period of approximately 8.76 years previously. Additionally, certain measures (e.g., BSSQ) are based upon the assumption of undergoing WLS and therefore could not have been completed regarding behaviors pre-WLS. Finally, findings from our relatively small sample, comprised largely of female, Caucasian, and middle-aged patients may not generalize to the bariatric surgery population at large.

The present study looked at both well-documented and theoretically relevant novel variables associated with weight nadir and significant WR. We have focused predominantly on psychosocial and behavioral factors, but we readily acknowledge that physiological factors may also play a role, and future research should explore those potentially complex relationships. We focus on “significant WR” (>20 % of maximum weight loss) because more modest amounts of WR may be associated with maintenance of health improvements and sustained health-related quality of life. In general, however, WR over the long-term is related to worsening of comorbidities, with the possible exception of infertility [6], making it an important outcome worthy of further investigation. In our preliminary study, predictors of significant WR were factors that can be targeted and modified, whereas initial weight loss was more dependent upon nonmodifiable physiological factors. Results highlight the importance of adherence to the recommended post-operative behavioral regimen in attenuating long-term WR, as well as appropriate monitoring and treatment for mental health problems, such as depression or substance misuse.