A test of the stress-buffering hypothesis of social support among bariatric surgery patients

Surgery for Obesity and Related Diseases - (2019) 1–9

Original article

A test of the stress-buffering hypothesis of social support among bariatric surgery patients Erica Ahlich, M.A.a,*, Jordana B. Herr, B.S.b, Katryna Thomas, B.S.b, Daniel T. Segarra, M.S.b, Diana Rancourt, Ph.D.a b

a Department of Psychology, University of South Florida, Tampa, Florida Morsani College of Medicine, University of South Florida, Tampa, Florida

Received 29 June 2019; accepted 18 October 2019

Abstract

Background: The buffering effect of social support against a range of stress-related health outcomes has been well-documented; however, no previous work has examined the applicability of this model to bariatric surgery outcomes. Objectives: The present study sought to address whether social support interacts with stress in predicting postsurgical outcomes, as well as whether these associations may vary by sex. Setting: Teaching hospital, United States. Methods: Data were collected using retrospective chart review (n 5 548). Stress, patient sex, and social support were explored as predictors of curvilinear weight loss trajectories during the first year after surgery using growth curve modeling. Results: Attendance at follow-up appointments was poor, with 250 patients at 6 months and 187 at 12 months. On average, these patients lost 27% of their total weight between baseline and the 12month follow-up. Overall, weight-related emotional support appeared to be most relevant to weight loss/maintenance in this population; cohabitating with a spouse or significant other and attendance at support group meetings did not predict weight loss or show any significant interactions with stress. Conclusions: The present study found only partial support for the stress-buffering model of social support among bariatric surgery patients. Such findings have important implications for assessment and follow-up care after bariatric surgery, as well as for future research in this area. (Surg Obes Relat Dis 2019;-:1–9.) Ó 2019 American Society for Bariatric Surgery. Published by Elsevier Inc. All rights reserved.

Key words:

Bariatric surgery; Obesity; Weight loss; Social support; Stress

The majority of work investigating the role of individual, environmental, and social factors in predicting bariatric surgery outcomes have received mixed support, underscoring the importance of exploring potential moderating effects [1,2]. Stress may represent an important predictor of surgical outcomes that is influenced by a variety of individual and

* Correspondence: Erica Ahlich, Department of Psychology, University of South Florida, 4202 East Fowler Ave, PCD 4118 G, Tampa, FL 33620. E-mail address: [email protected] (E. Ahlich).

environmental factors. Stress is associated with unhealthy eating behaviors, including disrupted eating patterns and increased consumption of snacks and fast food [3,4], less physical activity, and more sedentary behavior [5]. Likewise, stress is associated with other physiologic (e.g., gut microbiota) and behavioral (e.g., sleep disturbances) changes that have been linked to difficulties with weight regulation [6,7]. Yet, in studies examining the association between stress and postoperative outcomes, findings have been inconsistent [8,9]. A primary goal of this study was to investigate stress as a predictor of weight loss, from the

https://doi.org/10.1016/j.soard.2019.10.020 1550-7289/Ó 2019 American Society for Bariatric Surgery. Published by Elsevier Inc. All rights reserved.

2

Erica Ahlich et al. / Surgery for Obesity and Related Diseases - (2019) 1–9

perspective that these associations are likely influenced by several factors. Specifically, the focus of the present study will be on social support and sex as potential moderators of the association between stress and weight outcomes. The buffering effect of social support against the negative health outcomes associated with stress is well-documented in other areas [10–13]. This has been termed the “stressbuffering hypothesis” of social support [14]. Researchers have theorized that social support buffering against negative outcomes will depend upon the extent to which the kind of support matches the specific needs of a person undergoing a stressful event [14]. The present study focused on the following 3 sources of support most likely to match the needs of patients undergoing bariatric surgery: attendance at support group meetings after surgery (functional support), cohabitating with a partner/spouse (structural support), and weight-related emotional support. Based on descriptive information, bariatric patients report finding social support valuable in coping with stress after surgery [15]; however, no previous studies have investigated the stress by social support interaction quantitatively. Theoretic and empirical work suggests that the stressbuffering effects of support may vary by sex. Historically, research on stress responses was primarily conducted on nonhuman mammals and described a “fight-or-flight” response. More recently, social psychologists described the “tend-and-befriend” response to stress [16]. Specifically, researchers observed that in times of stress, humans both seek out and provide support to others. That humans show an affiliative nature after experiencing stress has been well-documented and has received neurobiological support [16–18]. This process, however, may differ depending on sex. Evolutionarily, social support was particularly important to survival for women. Among early humans, women were largely responsible for child rearing and foraging. When faced with a threat, women would have been more likely to protect both themselves and their offspring. Likewise, women were thought to form close ties with one another to share responsibilities. Therefore, biological processes that encourage an affiliative response, like oxytocin release, are thought to be particularly relevant to women in times of stress. Research on coping behaviors supports that women do rely more on social support than men to effectively downregulate the stress response [19]. Empirical work also suggests the effects of social support on outcomes may indeed depend upon the sex of the patient. In particular, research on the behavioral treatment of obesity suggests men may benefit less from social support in some instances. Wing et al. [20] randomized patients to a behavioral weight control program in which they either participated alone or with their spouses. Women participating with their spouse achieved greater weight loss than women participating alone, whereas men participating alone had better outcomes than men participating with their spouse.

However, this sex difference has not been consistently replicated [11,21,22], and more work is needed to understand the extent to which sex may moderate the associations between stress, social support, and weight loss outcomes, particularly related to bariatric surgery. The present study sought to address whether social support interacts with stress in predicting postsurgical outcomes and examined whether these outcomes may vary by sex. It was hypothesized that during the first year after surgery high levels of stress would be associated with worse outcomes when social support was low for women. Among men, it was anticipated that the association between stress and postsurgical outcomes would not differ by level of social support. Methods Participants Data were collected using retrospective chart review at a large teaching hospital in the southeastern United States, including records from 2012 through 2015. Adult patients undergoing bariatric surgery for the first time were included (n 5 548 patients; Table 1). The sample was majority female (81.2%) and non-Hispanic (81.1%). The majority of patients self-identified as white (67.6%), black (17.9%), or other/multiracial (13.1%). The average age was 43.86 (standard deviation 5 11.86) and the average presurgical body mass index (BMI) was 47.75 kg/m2 (standard deviation 5 8.69). Patients reported an average of 13.75 years of education and 65.6% were employed. The majority of patients had a sleeve gastrectomy (57.2%) or Roux-en-Y gastric bypass (40.6%) performed, with a small minority receiving a laparoscopic adjustable band (2.2%). Materials and procedure After receiving institutional review board approval, data collection was performed through the hospital’s electronic medical record system. After initial collection, 20% of files were reentered to examine reliability. The criterion was 80% agreement and all variables in the present study surpassed this threshold. Measures of stress and social support were collected from the bariatric center’s comprehensive intake packets, completed at the first preoperative visit. Stress As part of a modified version of the weight and lifestyle inventory patients were asked about perceptions of current stressors [23]. Eight areas were included as follows: work, health, relationship with spouse or significant other, activities related to children, activities related to parents, legal or financial trouble, school, and moving. Stress was operationalized as a count of the items that patients endorsed as current stressors. Number of stressors endorsed ranged from 0 to 8 in the current sample.

Erica Ahlich et al. / Surgery for Obesity and Related Diseases - (2019) 1–9

3

Table 1 Patient characteristics at time of surgery Total (n 5 548) Procedure Roux-en-Y gastric bypass Laparoscopic adjustable band Sleeve gastrectomy Age at time of surgery, yr Mean Range Race/ethnicity % Hispanic % Minority race Marital status Married or domestic partner Single Divorced/separated Widowed Employed Years of education Mean Range BMI at time of surgery* Mean Range Stress Mean SD Structural support Frequency SD Emotional support Mean SD Support group attendance Mean SD

Men (n 5 103)

Women (n 5 445)

40.6% 2.2% 57.2%

46.6% 3.9% 49.5%

39.2% 1.8% 59.0%

43.87 17–74

46.55 24–71

43.24 17–74

18.6% 31.6%

13.6% 19.6%

19.8% 34.7%

59.8% 20.4% 17.9% 1.9% 65.6%

81.9% 9.1% 9.1% .0% 62.3%

54.7% 23.0% 19.9% 2.4% 66.3%

13.75 1–20

14.60 8–20

13.57 1–20

47.75 33–95

49.08 34–78

47.43 33–95

1.95 1.52

2.10 1.71

1.91 1.48

64.9% .48

83.3% .37

60.8% .49

2.14 1.93

1.69 1.34

2.24 2.02

.30 1.49

.35 1.67

.29 1.45

Effect size (d/4c)

P value

.09

.129

.29

.011

.06 .13 .22

.154 .003 ,.001

.13 .40

.025 .001

.18

.119

.12

.319

.18

,.001

.32

.033

.04

.735

BMI 5 body mass index; SD 5 standard deviation. Sex differences were investigated with independent samples t tests or c2 tests as appropriate. * Equal variances not assumed.

Social support Attendance at support group meetings (functional support) was extracted from electronic medical record and was measured as a count variable. Attendance ranged from 0 to 20 meetings in the current sample. The second and third measures of social support were extracted from the weight and lifestyle inventory [23]. Living arrangements/marital status captured structural support and responses were coded dichotomously (living with a spouse/partner/significant other, or not). Last, patients were asked the open-ended question “How many people do you talk to about your weight when you are upset by it?” followed by “How many of these people are helpful to you?” A count measure was extracted from responses to the latter question, with higher counts indicating greater weight-related emotional support. Responses ranged from 0 to 21 in the current sample.

Weight loss The outcome of interest was postoperative weight loss. BMI (kg/m2) was selected as the primary outcome. However, percent total weight loss (TWL) and percent excess BMI loss (based on BMI of 25) were also used to describe weight loss in the sample, in accordance with recent recommendations [24]. Weight at time of surgery was coded as Time 0 with 4 additional time points, corresponding to every 3 months postsurgery (e.g., Time 1 5 3 mo, Time 2 5 6 mo) until 1 year. Data analysis Independent samples t tests, for continuous measures, and c2 analyses, for categoric measures, were used to explore sex differences in the covariates (i.e., age, employment status, procedure type, race, and education) and key study variables at time of surgery (Table 1). A correlation matrix is

4

Erica Ahlich et al. / Surgery for Obesity and Related Diseases - (2019) 1–9 Table 2 Correlation matrix with patient demographic characteristics and key study variables

1. Age 2. Racey 3. Years of education 4. Stress 5.Structural support 6. Emotional support 7. Support group attendance 8. BMI at surgery

1

2

3

4

5

6

7

8

— 2.15* .05 2.14* .07 2.06 .06 2.22*

2.07 — 2.09 2.05 2.17* 2.07 2.08 .14*

.17 .04 — 2.03 .03 2.05 .02 2.06

2.12 2.05 .01 — .03 2.07 2.08 .07

.06 2.14 .05 2.05 — 2.01 2.03 2.12z

2.15 .19 2.08 2.02 2.19 — 2.06 .00

.01 .00 .11 .10 .10 .02 — 2.05

2.40* .08 2.17 .16 2.19 .19 .09 —

BMI 5 body mass index. Correlations for males are presented above the diagonal, while correlations for females are presented below the diagonal. * P , .01. y 1 5 identified as belonging to minority racial group. z P , .05.

shown in Table 2. Multilevel modeling was used to test hypotheses. Degree of data missingness varied by time point, with nearly every patient having a weight measurement at time of surgery (Time 0; 96%), and a smaller number of patients having data 6 (Time 2; 46% follow-up visit information available) and 12 months later (Time 4; 34% available). Demographic characteristics and responses to primary variables were compared between those who did and did not attend the 6-month visit, as well as those who did and did not attend the 12-month visit. No significant differences were found for any variable, except those who missed the 6-month appointment were more likely to be cohabitating with a spouse/partner (P 5 .039). However, the effect was small (4c 5 .10) and no association was found for missing the 12-month appointment (P 5 .465, 4c 5 .03). The following are percentages of data available for the predictors: 81% current stressors, 99.5% support group meeting attendance, 84.3% structural support, and 69.5% weightrelated emotional support. Missing data were handled using Blimp [25], which generated 20 imputed data sets (1000 burn-in; 1000 thinned; Gibbs option). Blimp uses Markov Chain Monte Carlo algorithms, which produce less bias in samples with a large amount of missing data. Additionally, to reduce bias, all available time points were used to impute missing data. All predictors were level 2, time-invariant predictors. Dichotomous codes included 0 5 male; 0 5 unemployed; 0 5 did not identify as racial minority; and 0 5 not living with a spouse/partner/significant other. Age was centered at the mean. Time was coded 0 to 5 and was the only level 1, repeated-measures variable. All analyses were conducted using SAS 9.4 (Cary, NC, USA) with PROC MIXED/MIANALYZE, restricted maximum likelihood, and unstructured errors. Maximum likelihood was used for model comparisons. Linear and quadratic changes in weight were modeled. Models were built beginning with covariates at step 1, main effects at step 2, 2-way interaction terms at step 3, and so on; however, only the final models are described and shown in

the tables for the sake of clarity. Significant interactions were probed using an online interaction utility. Data were screened for violations of relevant assumptions. Results Preliminary analyses Overall weight loss On average, the sample overall lost a significant amount of weight between time of surgery and 12-month followup (mean 5 37.11 kg, standard deviation 5 17.46). This corresponds to an average loss of 13.19 BMI points, 58.91% excess BMI loss, and 27.05% TWL. There was significant variability in weight loss; based on a recommended 25% TWL criterion [26], 38.5% were classified as suboptimal responders. These individuals did not differ on any of the key variables under study, including demographic characteristics, stress, or social support, with the exception of age. Those with ,25% TWL at 12 months were significantly older than those with .25% TWL (P 5 .008). See Table 3 for weight loss information by initial BMI group. Preliminary models Preliminary model building suggested substantial betweenindividual variance in weight (intraclass correlation 5 .56) and the utility of including a quadratic random effect (D-2 log likelihood 5 1794.06, P , .05). While weight initially decreased after surgery (linear b 5 27.90), on average, patients began to show a modest amount of weight regain over time (quadratic b 5 1.22). Covariates were entered into the model as fixed effects. Being younger (P 5 .047), unemployed (P 5 .001), and identifying as a racial minority (P 5.003) were associated with a higher BMI at the time of surgery (Table 4). Support group attendance Neither stress (P 5 .634) nor support group attendance (P 5 .800) significantly predicted starting BMI; however,

Erica Ahlich et al. / Surgery for Obesity and Related Diseases - (2019) 1–9

5

Table 3 Weight loss during short-term follow-up by BMI group Weight loss

Initial BMI Weight 6 mo BMI BMID %EBMIL Weight %TWL 12 mo BMI BMID %EBMIL Weight %TWL

Initial BMI group 30–39.9 M (SD)

40–49.9 M (SD)

50–59.9 M (SD)

60 M (SD)

37.79 (1.72) 104.70 (11.84)

44.75 (2.68) 124.04 (15.42)

53.99 (2.70) 151.94 (17.43)

68.01 (9.23) 197.20 (34.25)

31.35 (3.52) 6.25 (3.14) 50.39 (26.22) 85.65 (13.04) 22.84 (11.17)

34.97 (3.30) 9.98 (3.01) 50.27 (14.62) 97.68 (14.88) 27.12 (8.63)

41.60 (3.31) 12.16 (3.24) 42.10 (10.56) 116.65 (13.76) 27.65 (8.11)

52.99 (9.54) 15.98 (4.39) 37.22 (9.48) 152.26 (29.83) 30.13 (12.43)

30.08 (4.04) 8.08 (4.12) 61.61 (30.03) 83.81 (15.03) 21.10 (10.46)

32.48 (4.23) 12.32 (4.37) 62.29 (20.87) 90.39 (16.63) 27.39 (9.27)

38.74 (5.57) 15.25 (5.31) 52.79 (17.87) 111.21 (17.43) 28.24 (9.59)

49.42 (10.48) 20.81 (6.24) 47.28 (14.42) 137.43 (27.36) 29.88 (8.54)

BMI 5 body mass index; M 5 mean; SD 5 standard deviation; TWL 5 total weight lost; EBMIL 5 excess BMI lost. Descriptive data available for 250 patients at 6 mo and 187 patients at 12 mo, of an initial sample of 548.

women had a lower baseline BMI than men (P 5 .049). Covariates that showed no significant fixed effects in step 1 were removed for parsimony. No 2-way interactions between time, stress, or support group attendance emerged. Subsequently, 3- and 4-way interactions were entered in the model to examine hypothesized moderation effects. Hypotheses were not supported; no significant higher order interactions emerged (linear P 5 .199; quadratic P 5 .316; Table 4). Structural support Controlling for the significant effects of sex and stress (described above) and covariates, the effect of structural support was explored. Results suggested that those with greater structural support had lower starting BMI values (P 5 .021). However, structural support did not predict weight change over time (linear P 5 .792; quadratic P 5 .181), or show any significant 3-way interactions with time and stress. Likewise, patient sex did not moderate these associations, with no significant 4-way interactions (linear P 5 .377; quadratic P 5 .343; Table 4). Weight-related emotional support Weight-related emotional support emerged as a significant predictor of linear weight change over time (P 5 .042). Additionally, a significant 3-way interaction emerged, between time2, stress, and weight-related emotional support (P 5 .030). The inclusion of this 3-way interaction significantly improved model fit (D-2 log likelihood 5 30.17, P , .05). No other significant 2- or 3-way interactions

emerged. Likewise, no significant 4-way interactions emerged between time/time2, stress, weight-related emotional support, and patient sex (linear P 5 .694; quadratic P 5 .782; Table 4). The significant 3-way interaction between time2, stress, and weight-related emotional support was plotted (Fig. 1). Among those with high stress, there was little difference in linear slope between those with high and low levels of weight-related emotional support. However, those with low weight-related emotional support showed greater weight regain toward the end of the 12-month follow-up period, compared with those with high levels of weightrelated emotional support. Among those with low levels of stress, differences in weight regain/stabilization (i.e., quadratic trajectories) were in the opposite direction; in the low-stress group, higher weight-related emotional support was associated with greater weight regain. Discussion Consistent with previous research, patients lost a significant amount of weight during the first year postsurgery. On average, patients lost 27% of their total weight between baseline and 12 months. Weight stabilized or showed modest regain within the first year. There was significant variability in weight loss, with 38.5% classified as suboptimal responders. Findings are partially consistent with the buffering hypothesis of social support, suggesting that support acts as a buffer for those who need it (i.e., high levels of stress), but is not as influential for those who do not. Consistent with the study rationale, those who reported more stressors,

6

Erica Ahlich et al. / Surgery for Obesity and Related Diseases - (2019) 1–9 Table 4 Stress predicting body mass index change over 12 months after bariatric surgery, including social support and patient sexes as moderators Preliminary model Fixed effects Intercept Time Time2 Age Employment status Procedure type LAGB SG Education Race Sex Stress Support 2-way interactions Time–stress Time–support Time–sex Stress–support Stress–sex Support–sex Time2–stress Time2–support Time2–sex 3-way interactions Time–stress–support Time–sex–stress Time–sex–support Stress–sex–support Time2–stress–support Time2–sex–stress Time2–sex–support 4-way interaction Time–stress–support–sex Time2–stress–support–sex Variance components Residual s2 Intercept t00 Time t11 Time2 t22 Covariance time, intercept t10 Covariance time2, intercept t20 Covariance time, time2 t12

Structural support

Emotional support

Support group attendance

47.36* 27.90* 1.22* 2.06y 22.79z

50.15* 27.90* 1.22* 2.05 22.49z

46.08* 26.58* .97* 2.06y 22.63z

48.82* 27.90* 1.22* 2.06* 22.69z

3.69 2.51 .10 1.98z

3.80 2.51 .06 1.85z 21.86y .13 21.69y

— — — 2.03z 2.23 1.35 1.32

3.44 2.46 .06 2.12z 21.57y .11 2.05

2.17 2.09 .04 2.25 2.28 2.75 .02 .11 .01

2.20 2.22y 2.20 2.06 2.36 2.10 .02 .04 2.04

2.17 .02 .06 .05 2.35 2.37 .02 2.01 2.01

2.10 .14 2.24 2.90 .07 2.01 .02

.11 .03 .13 .14 2.03y 2.01 2.01

.11 .08 .32 .28 2.02 2.02 2.05

.34 2.09

.09 .00

2.17 .03

3.41 67.47 4.03 .06 210.89 1.39 2.44

3.41 65.09 3.65 .05 29.57 1.11 2.37

3.41 65.77 3.99 .07 210.08 1.22 2.44

3.41 67.72 4.09 .07 210.78 1.34 2.46

LAGB 5 laparoscopic adjustable gastric band; SG 5 sleeve gastrectomy. * P , .001. y P , .05. z P , .01.

but had a larger emotional support network, had better weight maintenance. Perceiving that others are available to provide emotional support has been shown to be protective against a wide range of stressors [27] and in the present study, this type of social support may have “protected” against the negative effects of stress that many of these patients reported. If so, this type of buffering effect may work through several pathways [27]. For example, support might mitigate the effects of negative emotions on eating behaviors, enhance motivation, or dampen the harmful

physiologic effects of stress (e.g., gut microbiota and hormones [6]). Those with low stress and high levels of weight-related emotional support, however, demonstrated greater weight regain. This finding is inconsistent with the buffering hypothesis of social support, or the belief that social support benefits everyone, regardless of need. Instead, findings indicate the importance of needs matching. If a higher level of support is needed, such as experiencing numerous stressors, support is associated with better weight outcomes. However,

Erica Ahlich et al. / Surgery for Obesity and Related Diseases - (2019) 1–9 High stress, High emotional support

55

High stress, Low emotional support Low stress, High emotional support

50

Low stress, Low emotional support

Body Mass Index

45

40

35

30 Baseline Low time

6 mo

12 mo

Fig. 1. Curvilinear trajectory of body mass index over the first 12 months after surgery, as a function of stress- and weight-related emotional support.

if a high level of support is not needed, it may actually be associated with worse weight outcomes. Patients initially reported weight-related emotional support at the presurgery visit, suggesting that at this time, confidants were perceived as helpful. However, it could be that the weight-related discussions with close confidants that were helpful before surgery became less helpful postsurgery (e.g., feelings of pressure, greater focus on weight), unless there was a specific need (i.e., high-stress levels). Alternatively, for those in the low-stress group, having a larger number of confidants may be associated with more opportunities for social influences on eating, making habit maintenance challenging (e.g., more frequent eating related to social occasions, more celebratory eating, etc.). Given the unexpected nature and complexity of the results, future work should seek to replicate these findings. Surprisingly, support group attendance did not emerge as a predictor of weight loss, regardless of stress level and/or sex. This is in contrast to previous work observing positive associations between support group attendance and weight loss [28]. Notably, support group meetings were not always highly structured regarding discussion topics. This is intended to allow for group-specific tailoring. Previous research in this area observing associations with surgical outcomes has described groups with more structured content [29]. It may be that patients in the present study attended 1 to 2 groups (attendance was low), but discontinued if the material was not perceived as relevant. This flexible approach should be directly compared with a more structured support format. Also, investigations of how social support may be associated with different mechanisms, both behavioral (e.g., dietary adherence) and physiologic (e.g., stress regulation), during the weight loss and maintenance phases are needed. This could help bariatric centers tailor meetings to certain groups.

7

That structural support did not interact with stress in predicting weight loss was also unexpected. This was not due to lack of variability, as 65% endorsed living with a significant other, spouse, or partner. The effect of social support is most powerful when the type of support matches the needs of those experiencing a stressor [14]. While it was hypothesized that living with a significant other would translate to more active support, this may not have been the case. In contrast, it is possible that for some patients, having a partner at home made behavior change more difficult, especially if the patient was experiencing stress related to home life. Additional research exploring perceived support postsurgery will be important. Findings may be particularly relevant to patient eligibility for surgery. Physicians are often wary of providing surgery to those who lack sufficient support at home [30]. Support is often assessed by asking patients whether they live alone, who their caretaker will be, or how many people will be supporting them, but this approach neglects the importance of whether these structural supports are experienced as stress-inducing or -buffering by the patient. Indeed, these measures of structural support showed no association with weight loss or weight maintenance in the current data. Determining patient eligibility for surgery based on structural support alone may not be warranted at this time. Based on the tend-and-befriend theory, as well as previous research, it was anticipated that the interactions between stress, social support, and weight loss would differ by sex. However, the interaction between stress- and weight-related emotional support showed the same pattern, regardless of sex. One possible explanation is that the tendand-befriend theory may be more applicable to instances of acute, potentially life-threatening stressors. The development of this model arose from an evolutionary perspective, and has primarily been applied to responses to acute threats (e.g., acute pain, danger). The types of stressors measured in the present study (e.g., work, children, finances, etc.) are dissimilar in many ways and sex differences in the effectiveness of social support for these stressors may be less pronounced. Alternatively, men undergoing surgery who experience a high level of stress may particularly benefit from weight-related emotional support. Last, some of the research linking oxytocin release to sex differences in affiliative responses has relied on administration of oxytocin intranasally. Recent data have called some of these methodologies into question [31,32]. Additional work investigating the role of weight-related emotional support in bariatric surgery, and sex differences in particular, is needed. Strengths and limitations This is the first investigation of the stress-buffering hypothesis applied to bariatric surgery. Others have tested main effects only (e.g., support group attendance and

8

Erica Ahlich et al. / Surgery for Obesity and Related Diseases - (2019) 1–9

weight loss) or described the relationship qualitatively. The present study drew from well-established models in socialhealth psychology, included a large sample with 30% racial minority and a broad age range, and explored multiple sources of support. Nonetheless, a number of limitations should be considered. While the collaborating hospital is a teaching hospital, it is not an academic medical center. Thus, these data may be more generalizable to community hospitals with bariatric centers. Also, the ratio of males to females, though similar to what is commonly reported among bariatric surgery samples [33], may have impacted our ability to detect sex differences. This study used a retrospective design and relied on data collected as part of the surgery center intake packet. Single-item measures are common in medical intakes and this study was limited to single-item measures for multiple predictors. Low reliability of single-item measures may have attenuated effect sizes. Also, social support may have the largest buffering effect for outcomes, such as quality of life, as opposed to weight change. Likewise, given that social support may be most beneficial when matched with the current stressor, the lack of specificity in the measurement of stress, which captured a range of different stressors, may have attenuated results. For example, social stressors, such as marital stressors or care taking, may predict outcomes and interact with social support differently, compared with other stressors, such as moving or attending school. Additionally, the study did not include any physiologic markers of stress, such as cortisol or heart rate variability, or measurement of eating behaviors that may be influenced by stress. Future work may benefit from inclusion of such measures. Last, the attrition rate in the present study was significant, though commensurate with rates reported by other large studies [34]. The missing data techniques employed are most appropriate when data are missing at random. While comparison testing did not reveal identifiable patterns, it is possible that missingness was related to factors not included in the study.

Conclusions The present study found partial support for the stress buffering model of social support among bariatric surgery patients. Overall, results suggested that weight-related emotional support may be more relevant to weight loss/ maintenance than support group attendance or structural support. Future research should focus on recording postsurgical indicators of such variables to examine whether there are changes in perceived social support after surgery, whether these changes are associated with weight change over time, what the relevant mechanisms are, and whether certain types of social support influence different mechanisms involved in weight loss (e.g., diet changes, physiologic changes, exercise).

Disclosures The authors have no commercial associations that might be a conflict of interest in relation to this article. References [1] Livhits M, Mercado C, Yermilov I, et al. Preoperative predictors of weight loss following bariatric surgery: systematic review. Obes Surg 2012;22(1):70–89. [2] Van Hout GC, Verschure SK, Van Heck GL. Psychosocial predictors of success following bariatric surgery. Obes Surg 2005;15(4):552–60. [3] Barrington WE, Beresford SA, McGregor BA, White E. Perceived stress and eating behaviors by sex, obesity status, and stress vulnerability: findings from the vitamins and lifestyle (VITAL) study. J Acad Nutr Diet 2014;114(11):1791–9. [4] O’Connor DB, Jones F, Conner M, Ferguson E. Effects of daily hassles and eating style on eating behavior. Health Psychol 2008;27(1 S):S20–31. [5] Stults-Kolehmainen MA, Sinha R. The effects of stress on physical activity and exercise. Sports Med 2014;44(1):81–121. [6] Cryan JF, Dinan TG. Mind-altering microorganisms: the impact of the gut microbiota on brain and behaviour. Nat Rev Neurosci 2012;13(10):701–12. [7] Spiegel K, Tasali E, Leproult R, Van Cauter E. Effects of poor and short sleep on glucose metabolism and obesity risk. Nat Rev Endocrinol 2009;5(5):253–61. [8] Buddeberg-Fischer B, Klaghofer R, Sigrist S, Buddeberg C. Impact of psychosocial stress and symptoms on indication for bariatric surgery and outcome in morbidly obese patients. Obes Surg 2004;14(3):361–9. [9] Figura A, Ahnis A, Stengel A, et al. Determinants of weight loss following laparoscopic sleeve gastrectomy: the role of psychological burden, coping style, and motivation to undergo surgery. J Obes 2015;2015:626010. [10] Beeble ML, Bybee D, Sullivan CM, Adams AE. Main, mediating, and moderating effects of social support on the well-being of survivors of intimate partner violence across 2 years. J Consult Clin Psychol 2009;77(4):718–29. [11] Bowen KS, Uchino BN, Birmingham W, Carlisle M, Smith TW, Light KC. The stress-buffering effects of functional social support on ambulatory blood pressure. Health Psychol 2014;33(11):1440–3. [12] Koopman C, Hermanson K, Diamond S, Angell K, Spiegel D. Social support, life stress, pain and emotional adjustment to advanced breast cancer. Psychooncology 1998;7(2):101–11. [13] Peirce RS, Frone MR, Russell M, Cooper ML. Financial stress, social support, and alcohol involvement: a longitudinal test of the buffering hypothesis in a general population survey. Health Psychol 1996;15(1):38–47. [14] Cohen S, Wills TA. Stress, social support, and the buffering hypothesis. Psychol Bull 1985;98(2):310–57. [15] Hafner R, Watts J, Rogers J. Quality of life after gastric bypass for morbid obesity. Int J Obes 1991;15(8):555–60. [16] Taylor SE, Klein LC, Lewis BP, Gruenewald TL, Gurung RA, Updegraff JA. Biobehavioral responses to stress in females: tendand-befriend, not fight-or-flight. Psychol Rev 2000;107(3):411–29. [17] Ditzen B, Schaer M, Gabriel B, Bodenmann G, Ehlert U, Heinrichs M. Intranasal oxytocin increases positive communication and reduces cortisol levels during couple conflict. Biol Psychiatry 2009;65(9):728–31. [18] Cardoso C, Ellenbogen MA, Serravalle L, Linnen AM. Stress-induced negative mood moderates the relation between oxytocin administration and trust: evidence for the tend-and-befriend response to stress? Psychoneuroendocrinology 2013;38(11):2800–4.

Erica Ahlich et al. / Surgery for Obesity and Related Diseases - (2019) 1–9 [19] Tamres LK, Janicki D, Helgeson VS. Sex differences in coping behavior: a meta-analytic review and an examination of relative coping. Pers Soc Psychol Rev 2002;6(1):2–30. [20] Wing RR, Marcus MD, Epstein LH, Jawad A. A “family-based” approach to the treatment of obese type II diabetic patients. J Consult Clin Psychol 1991;59(1):156–62. [21] Knoll N, Schwarzer R. Gender and age differences in social support: a study of East German migrants. In: Weidner G, Kopp M, Kristenson M, editors. Heart disease: environment, stress and gender, Amsterdam: IOS Press; 2002. p. 198–210. [22] Heitzmann CA, Kaplan RM. Interaction between sex and social support in the control of type II diabetes mellitus. J Consult Clin Psychol 1984;52(6):1087–9. [23] Wadden TA, Foster GD. Weight and lifestyle inventory (WALI). Obesity 2006;14(S3):99S–118S. [24] van de Laar A. The% EBMIL/% EWL double booby-trap. A comment on studies that compare the effect of bariatric surgery between heavier and lighter patients. Obes Surg 2016;26(3) :612–3. [25] Enders CK, Keller BT, Levy R. A fully conditional specification approach to multilevel imputation of categorical and continuous variables. Psychol Methods 2018;23(2):298–317. [26] van de Laar AW, van Rijswijk A, Kakar H, Bruin SC. Sensitivity and specificity of 50% excess weight loss (50% EWL) and twelve other

[27] [28]

[29]

[30]

[31] [32]

[33]

[34]

9

bariatric criteria for weight loss success. Obes Surg 2018;28(8):2297–304. Cohen S. Social relationships and health. Am Psychol 2004;59(8):676–84. Livhits M, Mercado C, Yermilov I, et al. Is social support associated with greater weight loss after bariatric surgery? A systematic review. Obes Rev 2011;12(2):142–8. Orth WS, Madan AK, Taddeucci RJ, Coday M, Tichansky DS. Support group meeting attendance is associated with better weight loss. Obes Surg 2008;18(4):391–4. Santry HP, Lauderdale DS, Cagney KA, Rathouz PJ, Alverdy JC, Chin MH. Predictors of patient selection in bariatric surgery. Ann Surg 2007;245(1):59–67. Leng G, Ludwig M. Intranasal oxytocin: myths and delusions. Biol Psychiatry 2016;79(3):243–50. Walum H, Waldman ID, Young LJ. Statistical and methodological considerations for the interpretation of intranasal oxytocin studies. Biol Psychiatry 2016;79(3):251–7. Buchwald H, Estok R, Fahrbach K, et al. Weight and type 2 diabetes after bariatric surgery: systematic review and meta-analysis. Am J Med 2009;122(3):248–56. Spaniolas K, Kasten KR, Celio A, Burruss MB, Pories WJ. Postoperative follow-up after bariatric surgery: effect on weight loss. Obes Surg 2016;26(4):900–3.