ORIGINAL CONTRIBUTIONS

The Relationship of Pre-Operative Health Status to SustainedOutcome in Gastric Bypass Surgery

Richard I. Lanyon & Barbara M. Maxwell &Rebecca E. Wershba

Published online: 13 October 2013# Springer Science+Business Media New York 2013

Abstract The task of sustaining initial weight loss after gas-tric bypass surgery has been identified as the area of greatestconcern in this intervention. The present study investigatedthe role of good vs. poor pre-operative health as a moderatorvariable in identifying useful pre-operative predictors of con-tinued weight loss. Follow-up data at a mean of 12.8 monthsand again at 3.2 years post-operatively were available for 79patients on 227 interview variables and four psychologicalassessment instruments. Thesemeasures were studied for theirsuccess in predicting continued weight loss over the 1–3-yearperiod separately for patients who were in good and in poorgeneral pre-operative health. Previous findings showed thatthe overall mean simple weight loss to 12.8 months was45.61 kg, but additional weight loss to 3.2 years was only0.28 kg. The good and poor pre-operative health groupsdiffered little on these figures. However, the significant pre-dictors of continued weight loss for good-health patients (highanxiety and distress, low self-esteem, poor eating habits,strong expectations of life improvement, and good achieve-ment and coping skills) were quite different from those forpoor-health patients (good psychological health and happi-ness, strong personal support and life satisfaction, good eatinghabits, and little knowledge about their health). Thus, pre-operative health status served as a powerful moderator inpredicting continued weight loss from pre-operative charac-teristics. These findings offer a means of making more accu-rate predictions as to which patients are the best candidates forsurgery, and also suggest that different psychological andother interventions should be selected according to pre-operative health status.

Keywords Gastric Bypass . Long-TermOutcome .

Pre-OperativeHealth Status . Moderated Outcome Prediction

In a recent review, Powell et al. [1] concluded that, for personswho are obese, the task of losing weight and sustaining the lossis a difficult proposition, and that the area of greatest concernand needed focus is the sustainability of the initial weight loss.

This conclusion is dramatically illustrated by the findingsof the present authors [2, 3]. In these studies, the initial meanweight loss over the first post-operative year after gastricbypass surgery (GBS) was found to be 45.61 kg (BMI de-crease of 16.42); but the mean change over the subsequent2 years was essentially zero (0.28 kg; BMI decrease of 0.15).Even more interesting was the finding that the changes overthe 0–1-year period and the 1–3-year period were uncorrelated(0.00 for weight change; 0.08 for BMI change).

The best individual predictor of 0–1-year weight loss washealth status prior to surgery, defined in terms of the absenceof physical problems commonly associated with obesity(number of blood pressure medications, number of pain med-ications, regular or frequent hip joint pain, and severity oftypical illness). Other significant predictors of weight lossover the first year were composite variables representing goodpsychological health, having told friends and/or coworkers ofthe coming operation, and functional (versus dysfunctional)eating behaviors.

Over the second and third post-operative years, when asmany patients regained weight as lost additional weight, themain pre-operative predictors (of 1–3-year weight loss) werequite different from those for 0–1-year weight loss. They wereexpectation of increased overall confidence as a result ofsurgery, amount of present and expected future informationalsupport from close friends or family, and overall coping skills.The strongest predictor of 0–1-year loss (pre-operative healthstatus) was not significantly related to 1–3-year change.

R. I. Lanyon (*) : R. E. WershbaArizona State University, Tempe, AZ, USAe-mail: rlanyon@asu.edu

B. M. MaxwellSouthwest Medical Psychology, Phoenix, AZ, USA

OBES SURG (2014) 24:191–196DOI 10.1007/s11695-013-1085-x

Aims of the Present Study

The finding that although themeanweight loss in the 1–3-yearinterval was essentially zero, the pre-operative predictors ofweight loss over this period were quite different from those for0–1-year loss, draws attention to the importance of gatheringas much information as possible about the pre-operative var-iables that influence 1–3-year loss. Such information couldhopefully suggest pre-operative interventions that would pro-mote more successful long-term changes. An overall consid-eration of the previous findings of the present authors sug-gested that it would be fruitful to focus on the patients whowere in good pre-operative general health separately fromthose who were in poor pre-operative general health, withthe expectation that the variables that predicted weight losswould be rather different for each group. Thus, it was hypoth-esized that the accuracy of predicting continued weight loss(i.e., over the 1–3-year period) could be improved by consid-ering good- and poor-health patients separately. Stated differ-ently, it was hypothesized that pre-operative health statuswould be shown to serve as a moderator in predictor-outcome relationships.

Method

Participants

As described in the authors' previous work [2, 3], participantswere a subset of 243 patients who had received GBS at a largemedical center in metropolitan Phoenix. Of these, weight datawere obtained on 131 after approximately 1 year, and on 79after approximately 3 years. The present study utilized data onthe 79 patients for whom 3-year data were available.

Measures

Extensive pre-operative information had been obtained for bothclinical and research purposes, utilizing a comprehensive pro-cedure developed by the second author [2]. These data included273 interview items over 21 content areas in a structuredformat, plus four psychological inventories: the MinnesotaMultiphasic Personality Inventory-2 (MMPI-2) [4], the BasicPersonality Inventory (BPI) [5], the two sections of the Multi-dimensional Health Profile (MHP) [6], and a structured ques-tionnaire reflecting the Diagnostic and Statistical Manual ofMental Disorders, Fourth Edition, Text Revision (DSM-IV-TR) research criteria for binge-eating disorder [7]. Some ofthe interview areas of inquiry (such as extended details ofmedical history and parental histories) were not consideredrelevant for the present study. Omitting these areas, a total of227 itemswere considered as potential predictors, including testscores, plus the individual items of the MHP.

As described previously, a factor-analytic study of severalpossible weight loss measures found that the two most represen-tativewere simpleweight change and simple change in BMI.Dataon continuing weight loss were obtained from an abbreviated 1-year follow-up questionnaire and a 3-year follow-up questionnaire.For participants who did not return the questionnaire, weight lossdata were sought in the context of telephone interviews.

Pre-Operative Health Status The measure of general pre-operative health status was developed as follows. A review ofthe available pre-operative variables identified seven items withcontent related to general health status. Scores on these itemswere subjected to a principal components factor analysis withvarimax rotation, which yielded a single factor. The three highestloading variables on the factor, which represented 34.6 % of thetotal variance, were combined into a factor score to be used in thepresent study. These items were as follows: “Compared to otherpeople your age, how good has your health been over your adultlife?”; “Compared to other people your age, how good has yourhealth been over the past six months?”; and “Do you have achronic illness (illness lasting longer than 6 months)?” The firsttwo items had been rated on 5-point scales, and the third hadbeen answered yes or no. All three items were on theMHP; theyhad been developed for that instrument through objective test-construction procedures, as described in the MHP Manual [6]which also reports validity data. This measure of pre-operativehealth can be considered to represent the quality of the person'scontinuing general health status on a stable basis. The factorscores for the 79 patients were dichotomized as close as possibleto the median; this yielded 41 good pre-operative health patientsand 38 poor pre-operative health patients.

Correlations were then computed, separately for good-health and poor-health patients, between 1–3-year weightloss and all available pre-operative predictor variables, includ-ing individual items of the MHP where appropriate, but fo-cusing primarily on those variables whose content was relatedin any way to the previously determined predictors of 0–1-and 1–3-year weight loss.

Results

Four overlapping patient groups can be identified in the pres-ent work: those who underwent GBS (n =243), those forwhom 1-year follow-up data were available (n =131); thosefor whom 1-year follow-up data were not available (n =112),and those for whom both 1- and 3-year follow-up data wereavailable (n =79). To test for potential biases in the results dueto attrition, these four groups were compared on basic demo-graphic and other pre-operative characteristics [3]. Univariatecomparisons using t tests showed that the patients with com-plete follow-up data differed pre-operatively from the other

192 OBES SURG (2014) 24:191–196

groups only as being older and better educated, and havinglower pre-operative weight. There were no differences on anyof the medical or psychosocial variables that had been found topredict 0–1- and 1–3-year weight loss. It was therefore con-cluded that the attrition was unlikely to have biased the results.

Comparison of Good and Poor Pre-Operative Health Patients

The variables were identified that significantly predicted 1–3-year weight loss for good-health patients and those that sig-nificantly predicted 1–3-year weight loss for poor-health pa-tients. Results for good-health patients are shown in Table 1,and for poor-health patients in Table 2.

Listed in the top section of Table 1 are the variables forgood-health patients whose correlations significantlyexceeded those for poor-health patients at p <0.05. Analogousdata for poor-health patients are shown in the top section ofTable 2, which lists those variables whose correlations signif-icantly exceeded those for good-health patients.

Tables 1 and 2 show clear differences between good pre-operative health and poor pre-operative health patients. Forgood-health patients, the following four general characteristicswere most predictive of 1–3-year change.

(1) A high level of personal discomfort, as shown by chronicanxiety and relevant medications, a poor self-concept, andexcessive sensitivity to what other people think about them.

(2) Strong expectations of improved self-confidence, sociallife, life in general, and spouse relationships in particular,as a result of surgery.

(3) Some poor eating habits, such as binge/purge/fasting andfailure to monitor food intake.

(4) Good high school achievements and good overall skillsfor coping with stress.

For poor pre-operative health patients, the following fourgeneral characteristics were most predictive of 1–3-year change:

(1) Health related, no use of pain medication or other med-ications, more frequent doctor visits, and little knowl-edge regarding own health.

(2) Good psychological adjustment, happy and upbeatmood, low anxiety/distress, normal self-presentation,not impulsive, and able to focus thoughts easily.

(3) High level of support and advice from family andfriends, and satisfied with this support.

(4) No eating-related behaviors that are characteristic ofbinge eating.

A moderator is defined as a variable that interacts with therelationship between two other variables; so that, for example, apredictor might predict the criterion more accurately at lowlevels of the moderator than at high levels [e.g., 8]. As shownin Tables 1 and 2, this definition applies to pre-operative healthstatus for 13 predictor/criterion relationships in the presentstudy: the first six variables listed in Table 1 and the first sevenvariables listed in Table 2. Thus, pre-operative health status canbe viewed as a moderator of these relationships.

The presence of moderator effects can also be shown by thedemonstration of a significant interaction between the third(moderator) variable and the relationship between the first twovariables. Moderated regression analysis was therefore

Table 1 Significant predictors of1–3-year weight loss for pre-op-erative good-health patients andcomparison with poor-healthpatients

Correlations for one or both of theweight loss measures are signifi-cant at p <0.05a Differences between good-health and poor-health patientsare significant at p <0.05

Predictor Correlation

Good health Poor health

Weight change BMI change Weight change BMI change

Significantly stronger for good-health than poor-health patientsa

BPI: anxiety score 0.42 0.43 −0.30 −0.30History of medication for anxiety 0.54 0.51 −0.08 −0.09MMPI-2: paranoia (Pa) score 0.47 0.44 −0.25 −0.25BPI: self-depreciation score 0.36 0.37 −0.23 −0.24Expects improved spouse relationship 0.37 0.33 −0.38 −0.38MHPH: does not track food intake 0.32 0.29 −0.23 −0.22

Not significantly stronger

Expects increased social life 0.41 0.41

Expects increased self-confidence 0.41 0.41

High performance in high school 0.33 0.31

Binge-and-purge, fast 0.40 0.35

Expects positive life changes in general 0.41 0.41

Uses alcohol 0.34 0.33

MHPP: total coping 0.30 0.28

OBES SURG (2014) 24:191–196 193

performed for the 13 predictor/criterion relationships indicatedabove. The results confirm the presence of the moderatingeffect of pre-operative health status for most of these relation-ships. Specifically, in regard to Table 1, significant interactionswere shown for the BPI anxiety score, history of medication foranxiety, BPI self-depreciation score (all p <0.01), MMPI-2paranoia score, and expectation of improved spouse relation-ship (all p <0.05). In regard to Table 2, significant interactionswere shown for low use of pain medications, low use of othermedications, and normal patient presentation (all p <0.05).

Predictors of 3-Year Weight Loss

In addition to examining the predictors of 1–3-year weight loss,it is also meaningful to study the pre-operative variables thatpredict overall (0–3)-year weight loss. As with the 1–3-yeardata, the 0–3-year predictions are here presented separately forgood pre-operative and poor pre-operative health patients. Aswith the results for the 1–3-year comparisons, there are cleardifferences in the predictors, suggesting that different psycho-logical mechanisms are at work. The data are presented inTable 3, which shows the predictor variables that were signif-icantly correlated with 0–3-year weight loss for each group.

The one general pre-operative characteristic that is includ-ed for both groups is a clear anticipation/expectation of in-creased self-confidence, self-esteem, and social life. However,the remaining significant predictive characteristics are differ-ent. For the good pre-operative health group, the strongestpredictor is simply pre-operative weight. The other strongpredictors include embarrassment related to binge eating,some lack of trust in the health care system as a whole,significant marijuana and/or alcohol use, and looking forwardto a verywide range of improvements such as a better physicalmovement and an improved work situation. It appears thatthese patients are very much aware of their limitations andhave strong feelings about being viewed in highly negativeways; and that the prospect of changing these aspects providestheir motivation for change.

The characteristics related to improvement for the poor pre-operative health patients (in addition to expectations of im-proved self-esteem and improved social life) are quite differ-ent. For these patients, improvement is highly dependent onthe amount of support, advice, and care given by others, mostlikely family members. Improvers have strong, stable, andongoing positive spouse/family relationships, and they antic-ipate that this care will continue in the future. Interestingly,

Table 2 Significant predictors of1–3-year weight loss for pre-op-erative poor-health patients andcomparison with good-healthpatients

Correlations for one or both of theweight loss measures are signifi-cant at p <0.05a All differences between poor-health and good-health patientsare significant at p <0.05

Predictor Correlation

Poor health Good health

Weightchange

BMIchange

Weightchange

BMIchange

Significantly stronger for poor-health than poor-health patientsa

Uses pain medications −0.53 −0.53 −0.00 −0.01Uses other medications −0.39 −0.40 0.16 0.14

Normal presentation during interview 0.39 0.40 −0.24 −0.23Meets DSM-IV-TR binge-eating disordercriteria

−0.34 −0.33 0.11 0.13

MHPP: psychological distress −0.30 −0.29 0.16 0.16

MHPP: anxiety −0.32 −0.31 0.17 0.19

MHPH: number of doctor visits 0.33 0.32 −0.05 0.02

Not significantly stronger

BPI: impulse expression score −0.40 −0.40BEB: lacks control of eating behavior −0.33 −0.31BEB: regularly binges −0.35 −0.33MHPP: emotional support 0.35 0.35

MHPP: informational support 0.32 0.34

MHPP: satisfied with advice from family andfriends

0.39 0.40

MHPP: feels guilty when things go wrong −0.34 −0.34MHPP: easy to focus thoughts 0.44 0.44

MHPH: knows enough to protect health −0.34 −0.34MHPP: mood happy, upbeat 0.35 0.36

MHPH: high doctor visits 0.33 0.32

194 OBES SURG (2014) 24:191–196

they have little interest in their own health, do not acknowl-edge the importance of good health, and have little knowledgeabout health in general. They tend not to seek medical help oruse medication when sick, or make plans to resolve their ownstresses; rather, other people do this for them. Also, they arehappy and upbeat, and free from psychological difficulties.However, the current presence of a very serious illness didinterfere with weight loss.

How similar are these findings to the predictors of changeover the second and third (1–3)-year period? There are definitesimilarities. For the good pre-operative health patients, thesefactors can be viewed as strong negative feelings/sensitivityabout their obesity, plus strong anticipation of an improvedspouse relationship and general life improvement. For poorpre-operative health patients, change was strongly predictedby being fully taken care of, perhaps in a hyper-dependentmanner, with no responsibility (or even interest in) one'shealth problems, and by the absence of pre-operative binge-eating characteristics.

Discussion

The present study sought confirmation for the hypothesis that thesignificant pre-operative predictors of sustained weight loss aftergastric bypass surgery differ according to pre-operative generalhealth status, whether good or poor. As hypothesized, the pre-dictors for each of these two groups were rather different, and tosome extent, contradictory. For good pre-operative health pa-tients, those who were most successful in losing weight beyondthe first year were psychologically distressed with a poor self-concept, and had poor eating habits and were likely to usealcohol, but had good achievement and coping skills andexpected major and widespread improvements in their personallives as a result of the surgery. The poor pre-operative healthpatients who were most successful in continuing to lose weighthad a high level of interpersonal support, good psychologicaladjustment, an upbeat mood, no eating disorder behavior, a highlevel of personal support, and a greater number of doctor visitsbut little understanding of their ownhealth. Inspection of Tables 1and 2 show that the sharpest differences between the predictorsfor good and poor pre-operative health patients involve variablesrelated to mental health. Poor-health improvers were happy andanxiety-free, made a positive impression on the interviewer, andtended to be satisfied with their lives, while good-health im-provers were the opposite.

Examination of Table 3, showing 0–3-year changes basedon predictions from pre-operative measures, show somewhatsimilar results to the 1–3-year changes. Good pre-operativehealth improvers appeared to be motivated by a high degree ofpersonal discomfort, while the poor pre-operative health pa-tients did best if they could happily rely on somebody else tobe responsible for them.

Several methodological concerns regarding the presentstudy should be noted. The first pertains to the comparabilityof the patient groups for whom all data were available (n =79),those for whom the year 1 follow-up data were available(n =131), and the original group who underwent the surgery(n =243). As stated above, it was shown in a previous paper [3]that these groups were comparable in relevant characteristics;

Table 3 Significant predictors of 0–3-year weight loss for pre-operativegood health patients and pre-operative poor-health patients

Indicator Correlation with0–3-year lossa

Weight BMI

Pre-operative good-health patients (N =41)

Pre-operative weight/BMI 0.48 0.56

Excess weight/BMI 0.43 0.37

Marijuana use 0.43 0.37

Significant alcohol use 0.37 0.36

No expectation of life improvement −0.37 −0.36Expect improved social life 0.36 0.40

Expect improved work situation 0.36 0.36

Importance of good health 0.35 0.33

Expect ergonomic improvement 0.34 0.29

Expect higher self-esteem/confidence 0.32 0.32

Trust in health care system −0.32 −0.31Embarrassed about binge eating 0.32 0.28

Physical abuse as a child −0.32 −0.26Peer teasing −0.31 −0.33

Pre-operative poor-health patients (N =38)

Serious illness within past year −0.50 −0.44Spouse/family relationships good long-term 0.46 0.45

Mood happy, upbeat 0.45 0.45

Number of non-psychiatric medications −0.45 −0.36General psychological distress −0.44 −0.44Extent of health knowledge −0.44 −0.42Memory problems −0.43 −0.40Positive health values −0.42 −0.32Much social support 0.41 0.37

Satisfied with advice received 0.41 0.41

Expects much advice in future 0.40 0.41

Importance of good health −0.40 −0.19Smoker 0.39 0.38

Able to take care of health problems −0.36 −0.30Seeks medical advice when sick −0.36 −0.31Makes plans to resolve stress −0.36 −0.31Expects improved social life 0.34 0.23

Expects improved self-esteem/confidence 0.34 0.25

a Either or both correlations are significant at p<0.05

OBES SURG (2014) 24:191–196 195

thus, the results of the present study were not a function ofbiases due to attrition.

Another concern pertains to the large number of variablesthat were considered and the possibility that the results couldhave been obtained by chance alone. As stated above, a totalof 227 possible predictor/criterion relationships were exam-ined, 35 of which (i.e., 15.4 %) showed significance beyondthe 0.05 level. Thus, chance is unlikely to have been a factor.More important, the patterns of significance showed strongconsistencies in content, supporting the validity of these con-tent areas as predictors despite the large number of variables.

A third limitation of the study involves possible biases dueto the fact that the follow-up data were obtained through self-report. While acknowledging the risk of unreliability in thisprocedure, it is pointed out that one of the authors (BMM) hadhad ongoing relationships with many of the patients thoughconducting follow-up groups and workshops, increasing thelikelihood that patients' self-reports would be made with care.

In seeking possible interpretations for the findings, it is firstnoted that the patients on average showed negligible furtherweight loss over the second and third post-operative years.Thus, approximately one half of the patients were completelyunsuccessful in continuing to lose weight or actually regainedweight. Computations showed that this was not much differentfor the good and the poor pre-operative health patients (mean1–3-year loss of 2.69 and mean gain of 2.15 kg, respectively).

The task of making any significant and permanent change inbehavior, such as sustaining and enhancing one's weight loss,requires commitment to a regimen involving personal determi-nation and structured planning, plus mental and physical effort.It does not just happen. Thus, explanations for the differentialeffects of the predictors can be sought both in terms of factorspromoting continued weight loss and factors interfering with it.In regard to the good pre-operative health patients who showedthe greatest sustained improvement, it is noteworthy is that theywere unhappy, anxious, depressed, perhaps embarrassed, andgenerally miserable; but they strongly expected that their liveswould be greatly improved. These factors illustrate the presenceof strong, positive internal motivation.

Now, considering the poor pre-operative health patients whoimproved the most, the factors predictive of success for themwere good psychological adjustment and a great deal of exter-nal personal support—a comprehensively warm, supportive,and accepting environment. Perhaps the fact of poor healthwas a sufficient burden that these assets—good emotionalhealth and a high level of personal support—were needed inorder to continue improving. The low pre-operative healthpatients who did not continue to improve had lower levels ofmental health and self-esteem, and in addition, lacked a

supportive environment. Perhaps the task of sustaining changeswithout these assets and with poor health too was simplybeyond their capacity.

These findings have implications for improving the long-term success rate of GBS. First, consideration of patients interms of their pre-operative health status would enable theprediction of long-term weight loss to be made more accu-rately. Second, there is the clear possibility that pre-operativepersonal changes made using psychotherapeutic or relatedprocedures could improve the probability of meaningfullong-term weight loss. For patients with poor pre-operativehealth, an obvious example would be to help arrange for amore supportive interpersonal environment; while for patientswith good pre-operative health, the probability of greaterlong-term change would be improved by helping them togenerate a stronger expectation of overall life improvement.

In conclusion, it is apparent that good and poor pre-operative health patients should be viewed as rather differentpopulations, and their psychological and behavioral prepara-tions for surgery should involve different tasks and goals. It isnot unlikely that other pre-operative characteristics besidehealth status can serve to moderate prediction-criterion rela-tionships, and that identifying them could lead to even greatereffectiveness in outcome after GBS.

Conflict of Interest Richard I. Lanyon receives royalties from the saleof the Multidimensional Health Profile. Barbara M. Maxwell andRebecca E. Wershba have no conflicts of interest to declare.

References

1. Powell LH, Calvin III JE, Calvin Jr JE. Effective obesity treatments.Am Psychol. 2007;62:234–46.

2. Lanyon RI, Maxwell BM. Predictors of outcome after gastric bypasssurgery. Obes Surg. 2007;17:321–8.

3. Lanyon RI, Maxwell BM, Kraft AJ. Prediction of long-term outcomeafter gastric bypass surgery. Obes Surg. 2009;19:439–45.

4. Butcher JN, Dahlstrom WG, Graham J, et al. MMPI-2: manual foradministration and scoring. Minneapolis: University of MinnesotaPress; 1989.

5. Jackson DN. Basic Personality Inventory: manual. Sigma AssessmentSystems: Port Huron, MI; 1989.

6. Ruehlman LS, Lanyon RI, Karoly P. Multidimensional Health Profile:professional manual. Psychological Assessment Resources: Odessa,FL; 1998.

7. American Psychiatric Association. Diagnostic and statistical manual ofmental disorders–IV–TR. Washington, DC: APA; 1964.

8. McGrath RE,MitchellM, KimBH, et al. Evidence for response bias asa source of error variance in applied assessment. Psychol Bull.2010;136:450–70.

196 OBES SURG (2014) 24:191–196