Increased sensitivity to alcohol after gastric bypass has been described. The aim of this study was to investigate whether bariatric surgery is associated with alcohol problems.
Increased sensitivity to alcohol after gastric bypass has been described. The aim of this study was to investigate whether bariatric surgery is associated with alcohol problems.
The prospective, controlled Swedish Obese Subjects (SOS) study enrolled 2,010 obese patients who underwent bariatric surgery (68% vertical banded gastroplasty (VBG), 19% banding, and 13% gastric bypass) and 2,037 matched controls. Patients were recruited between 1987 and 2001. Data on alcohol abuse diagnoses, self-reported alcohol consumption, and alcohol problems were obtained from the National Patient Register and questionnaires. Follow-up time was 8-22 years.
During follow-up, 93.1% of the surgery patients and 96.0% of the controls reported alcohol consumption classified as low risk by the World Health Organization (WHO). However, compared to controls, the gastric bypass group had increased risk of alcohol abuse diagnoses (adjusted hazard ratio [adjHR] = 4.97), alcohol consumption at least at the WHO medium risk level (adjHR = 2.69), and alcohol problems (adjHR = 5.91). VBG increased the risk of these conditions with adjHRs of 2.23, 1.52, and 2.30, respectively, while banding was not different from controls.
Alcohol consumption, alcohol problems, and alcohol abuse are increased after gastric bypass and VBG.
Bariatric surgery is currently the most effective treatment to obtain sustained weight loss and it is becoming a common treatment for severe obesity. Bariatric surgery reduces the incidence of metabolic disturbances , cardiovascular events , and cancer [5, 6], and decreases overall mortality in obese subjects [6, 7]. However, complications and side-effects of bariatric surgery have been described , including self-reported increased sensitivity to alcohol after gastric bypass [9, 10]. Studies have suggested that peak alcohol levels  and measured alcohol elimination times [11, 13] are increased after gastric bypass. Conflicting results have been obtained in patients undergoing sleeve gastrectomy [14, 15], while gastric banding does not seem to alter alcohol metabolism .
Three recent studies have addressed the question of alcohol use after bariatric surgery. Davis et al. show that alcohol use was decreased in patients who had undergone gastric bypass surgery on average 6 months earlier . King et al. also found a decrease in alcohol intake in the first year following surgery, and no change in the prevalence of symptoms of alcohol use disorder (AUD) . However, the prevalence of AUD symptoms was higher in the second postoperative year . Conason et al. has also reported an increase in frequency of alcohol use 2 years after gastric bypass surgery . However, changes in alcohol consumption and alcohol problems beyond 2-years after bariatric surgery have not been examined. In this report, we therefore investigate very long-term changes in alcohol consumption and incidence of alcohol problems and alcohol abuse diagnoses after bariatric surgery in the non-randomized, prospective, controlled Swedish Obese Subjects (SOS) study.
Seven regional ethics review boards approved the study protocol. Informed consent was obtained from all participants. The SOS study enrolled 4,047 obese patients at 25 surgical departments and 480 primary health care centres in Sweden between September 1, 1987, and January 31, 2001. ClinicalTrials.gov Identifier: NCT01479452. The study design has been previously described [1, 4, 5, 7]. In brief, 6,905 subjects participated in a matching examination, and 5,335 were eligible. Among them, 2,010 individuals preferring surgery constituted the surgery group, and a contemporaneously matched control group (N = 2,037) was created using 18 matching variables [5, 19]. The matching variables were sex, age, weight, height, waist and hip circumferences, systolic blood pressure, serum cholesterol and triglyceride levels, smoking status, diabetes, menopausal status, four psychosocial variables with documented associations with the risk of death, and two personality traits related to treatment preferences. Although a surgery patient and his or her corresponding control patient always started the study on the day of surgery, the matching was not performed at an individual level. Instead the matching algorithm selected control patients so that the current mean values of the matching variables in the control group became as similar as possible to the current mean values in the surgery group according to the method of sequential treatment assignment .
The two study groups had identical inclusion and exclusion criteria, and all controls were eligible for surgery. Inclusion criteria were age between 37 and 60 years and BMI of ≥ 34 kg/m2 for men and ≥ 38 kg/m2 for women. The exclusion criteria were previous bariatric surgery or surgery for gastric or duodenal ulcer, gastric ulcer during the past six months, ongoing malignancy, active malignancy during the past five years, myocardial infarction during the past six months, bulimic eating pattern, psychiatric or cooperative problems contraindicating bariatric surgery, other contraindicating conditions (such as chronic glucocorticoid or anti-inflammatory treatment). Patients were excluded from the study if the total self-reported alcohol consumption per day exceeded 33.9 g (corresponding to approximately 2.5 standard drinks per day, 14 g alcohol/ standard drink), or if the patients reported current alcohol problems considered to be serious by a doctor in the study administration. Drug abuse was also an exclusion criterion. All patients met these inclusion and exclusion criteria at the matching examination undertaken on average 12 months before study start. Approximately four weeks before the start of the intervention, a baseline examination was carried out.
The treating surgeons determined the type of surgery. In the surgery group, 265 patients underwent gastric bypass, 376 underwent nonadjustable or adjustable banding (banding), and 1,369 underwent vertical banded gastroplasty (VBG) . Patients in the control group were given the customary treatment for obesity at their center of registration, ranging from advanced life-style advice to no treatment.
Physical examinations were carried out and questionnaires concerning life style, nutrition, and health were completed at matching, baseline, and after 0.5, 1, 2, 3, 4, 6, 8, 10, 15, and 20 years of follow up. Centralized laboratory examinations were performed at matching and baseline examinations and after 2, 10, 15, and 20 years. The SOS study database was cross-checked against public registers annually.
The primary end point of the SOS study was overall mortality . In the original SOS study protocol from 1987 it was stated that negative effects of medical and surgical treatments of obesity should be reported. Outcomes of the current article are self-reported alcohol consumption and problems, and alcohol abuse diagnoses during hospitalization (see Supporting Information Table S1).
Information about alcohol consumption was collected from the validated SOS dietary questionnaire  which covers habitual intake of food and beverages during the last three months. Beverages included a range of non-alcoholic drinks as well as beer, wine, dessert wine, and liquor. From the responses, the total, average alcohol intake in grams per day was calculated (beer, 2.25% alcohol by volume (ABV), 3.5% ABV or 5% ABV, wine 12% ABV, dessert wine 20% ABV, and liquor 40% ABV).
Medium risk alcohol consumption has previously been defined by the World Health Organization (WHO) as 40-60 g of pure alcohol per day in men and 20-40 g in women . Patients reporting alcohol consumption above 40 g of alcohol per day for men (corresponding to approximately three standard drinks per day) or above 20 g per day for women (corresponding to approximately 1.5 standard drinks per day) were classified as having at least medium risk alcohol consumption. Sixty-six patients consuming alcohol above the WHO high risk level (more than 60 g per day for men and 40 g per day for women) were pooled with the medium risk group.
The straightforward question “Do you think you have alcohol problems” with answer options of yes or no was also included in the questionnaire. A yes answer to this question was used to identify patients with self-reported alcohol problems. This question has not previously been validated to other alcohol abuse assessments scales.
The cut-off date for the analysis was July 1, 2011. For self-reported data on alcohol consumption and alcohol problems, the follow-up time was up to 20 years, with a median of 10 years (range 0-20). The follow-up rates at 2, 10, 15, and 20 years were 87 (3,509 out of 4,027), 71 (2,741 out of 3,845), 52 (1,233 out of 2,393) and 50 percent (321 out of 645), respectively, taking into account mortality and patients not yet having reached their 15 and 20 years examinations. Information on diagnoses related to alcohol abuse among patients who were hospitalized for any reason were obtained from the National Patient Register containing data on hospital discharges using International Classification of Disease (ICD) 9 and 10 codes. The ICD codes were selected to detect active alcohol abuse or any acute or chronic disease related to abuse of alcohol (see Supporting Information Table S1). At the time of register linkage, the National Patient Register contained complete information until the end of year 2009 and covers 99% of all hospital admissions and discharges. Follow-up time with respect to these data was up to 22 years, with a median of 15 years (range 8-22 years).
Mean values and standard deviations were used to describe baseline characteristics of the participants. Differences between group means were analyzed with one-way ANOVA (continuous variables) or Exact Fishers test (dichotomous variables). Time to first event (the first time a patient reported alcohol problems or at least medium risk alcohol consumption or was diagnosed with an ICD-code related to alcohol abuse) was calculated from the date of inclusion into the study. Those never reporting medium risk alcohol consumption or alcohol problems were treated as censored observations at end of follow-up. Time of progression to first event after inclusion was compared between the four treatment groups (gastric bypass, VBG, banding, and control) with Kaplan-Meier estimates of cumulative incidence rates. Log-rank test was used to analyze differences in cumulative incidence. Cox proportional-hazards models based on baseline data were also used to evaluate time to an event while adjusting for preselected risk factors for overconsumption of alcohol (sex and age, daily smoking, alcohol consumption, and total calorie intake at baseline). Patients reporting at least medium risk alcohol consumption or alcohol problems or missing data on these variables at baseline were excluded from the corresponding cumulative incidence analysis. A total of 98 patients were excluded. Sixty-seven patients were excluded from the analyses of medium risk alcohol consumption and 38 from analyses of alcohol problems (the number of patients does not add up to 98 as some patients reported having both at least medium risk alcohol consumption and alcohol problems). Follow-up time in the Kaplan-Meier figures is truncated at 15 years, because number of persons at risk beyond this point was low. However, all available follow-up data are used to calculate hazard ratios and 95% CIs. To identify independent predictors for alcohol abuse, preselected factors of sex, age, daily smoking, alcohol intake, and BMI at baseline were considered simultaneously in a multivariate Cox model.
In addition, cross-sectional prevalence estimates and 95% CIs for medium risk alcohol consumption and alcohol problems were calculated for each follow-up time. Primarily, the intention to treat principle was applied in that each participant remained in the original treatment group. In a secondary analysis, 580 subjects who had undergone surgery that resulted in a change of treatment group were excluded from the analysis. For all analyses, a P-value < 0.05 was considered significant. Statistical analyses were carried out using the Stata statistical package 10.1 (Stata-Corp. 2007. Stata Statistical Software: Release 10.1. College Station, TX; StataCorp LP.).
At baseline, the patients in the surgery groups were on average heavier (P < 0.001), younger (P < 0.001), and were more frequently smokers (P < 0.001) than patients in the control group (see Supporting Information Table S2). The mean alcohol consumption did not differ between the surgery and control groups at baseline (P = 0.161), and there was no difference in prevalence of at least medium risk alcohol consumption as defined by the WHO (more than 20 and 40 g of alcohol per day for women and men, respectively, P = 0.137) , self-reported alcohol problems (P = 0.804), or the number of patients reporting any alcohol consumption (P = 0.268).
The alcohol consumption during ten years in the SOS study is shown in Figure 1. In addition, the figure shows the WHO cut-off levels for medium risk alcohol consumption for men and women . During the follow-up, 96.0% of control patients and 93.1% of the surgery patients consistently reported an alcohol consumption that was classified as low risk. The corresponding figures for gastric bypass, VBG, and banding were 89.2%, 94.0%, and 92.4%, respectively. The percentage of the individuals in each group that reported intake of alcohol at least at the WHO medium risk level at years 1, 2, 3, 4, 6, 8, and 10 is shown in Figure 2. At all time points, the proportion of individuals with at least medium risk alcohol intake was highest among those that had undergone gastric bypass surgery. The cumulative incidence of medium risk alcohol consumption over the entire observation period is shown in Figure 3. The gastric bypass group had increased risk of self-reported alcohol consumption above the WHO medium risk level (unadjusted hazard ratio (HR) = 2.63; P < 0.001; Table 1). When the model was adjusted for sex and baseline parameters (age, total caloric intake, daily smoking, and alcohol consumption), the adjusted hazard ratio (adjHR) was 2.69 (P < 0.001; Table 1). After adjustments, the HR for surgery vs. control group was significantly elevated also for VBG (adjHR = 1.52, P = 0.013) but not for banding (adjHR = 1.22, P = 0.415; Table 1). Gastric bypass was associated with a significantly increased risk compared to VBG (adjHR = 1.77, P = 0.026) and banding (adjHR = 2.21, P = 0.013) (Table 1).
|Self-reported medium risk alcohol consumption, n = 3,980a||Self-reported alcohol problems, n = 4,009b||Alcohol abuse diagnosis during hospitalization|
|HR (95% CI)||P value||HR (95% CI)||P value||HR (95% CI)||P value||HR (95% CI)||P value||HR (95% CI)||P value||HR (95% CI)||P value|
|GBP vs. Controls||2.63 (1.71-4.03)||<0.001||2.69 (1.58-4.57)||<0.001||5.13 (2.99-8.82)||<0.001||5.91 (3.40-10.39)||<0.001||4.29 (2.37-7.79)||<0.001||4.97 (2.70-9.15)||<0.001|
|VBG vs. Control||1.36 (1.00-1.86)||0.051||1.52 (1.09-2.11)||0.013||2.06 (1.32-3.22)||0.002||2.30 (1.45-3.66)||<0.001||2.00 (1.27-3.14)||0.003||2.23 (1.38-3.59)||0.001|
|Banding vs. Control||1.64 (1.07-2.52)||0.022||1.22 (0.76-1.97)||0.415||1.45 (0.71-2.95)||0.303||1.44 (0.69-3.01)||0.328||1.54 (0.75-3.13)||0.237||1.57 (0.73-3.35)||0.246|
|GBP vs. banding||1.60 (0.95-2.67)||0.075||2.21 (1.18-4.12)||0.013||3.54 (1.68-7.45)||0.001||4.10 (1.92-8.73)||<0.001||2.79 (1.26-6.21)||0.012||3.17 (1.40-7.20)||0.006|
|GBP vs. VBG||1.93 (1.26-2.95)||0.003||1.77 (1.07-2.93)||0.026||2.49 (1.51-4.13)||<0.001||2.57 (1.56-4.22)||<0.001||2.15 (1.21-3.82)||0.009||2.23 (1.24-4.01)||0.007|
The unadjusted cumulative incidence of self-reported alcohol problems over the entire observation period is shown in Figure 4. There was a significant difference (P < 0.001, log-rank test) between the different treatment groups. The gastric bypass group had an adjHR of 5.91 (P < 0.001) and the VBG group had an adjHR of 2.30 (P < 0.001) compared to controls (Table 1). The adjHR for banding vs. control was not significant (P = 0.328). When different surgical procedures were compared, the gastric bypass group had an increased risk compared to VBG (adjHR = 2.57, P < 0.001) and banding (adjHR = 4.10, P < 0.001) (Table 1). With the exception for year 3, the gastric bypass group had the highest proportion of self-reported problems at different follow-up times (Figure 5).
Problems with alcohol use after bariatric surgery were also analyzed using data on alcohol abuse diagnoses during hospitalization from the National Patient Register. The unadjusted cumulative incidence rates of alcohol abuse diagnosis during hospitalization over the entire observation period are shown in Figure 6. There was a significant difference (P < 0.001, log-rank test) between the different treatment groups.
The gastric bypass group had an adjHR of 4.97 (P < 0.001; Table 1) and the VBG group had an adjHR of 2.23 (P = 0.001; Table 1) for postoperative alcohol abuse diagnoses as compared with the controls. The adjHR for banding vs. control was not significant (adjHR = 1.57, P = 0.246). When different surgical procedures were compared, the gastric bypass group had higher incidence of postoperative alcohol abuse diagnoses compared both to VBG (adjHR = 2.23, P = 0.007) and banding (adjHR = 3.17, P = 0.006) (Table 1).
We also performed an analysis to identify preoperative (baseline) predictors of postoperative alcohol abuse. Male sex (HR = 1.86, 95% CI 1.14-3.05; P = 0.013), baseline smoking (HR = 2.76, 95% CI 1.84-4.13; P < 0.001), and baseline alcohol consumption (per 10 g/day, HR = 1.80, 95% CI 1.52-2.12; P < 0.001) were independently related to an increased likelihood of alcohol abuse diagnoses after surgery whereas baseline BMI and age were not.
To account for possible effects of reoperations in the surgery group or bariatric surgery of patients in the control group, 580 patients who had undergone surgery that constituted a change from the original group were excluded from the analysis. In these per protocol analyses, the risk for at least medium risk alcohol consumption, alcohol problems, and alcohol abuse diagnoses remained significantly elevated for the patients treated with gastric bypass as compared to the control patients (adjHR = 2.75, 5.37, and 4.76, respectively. All P < 0.001; Supporting Information Table S3). The risk for alcohol problems and alcohol abuse diagnosis also remained significantly elevated for the patients treated with VBG as compared to the control patients (adjHR = 1.67 and 2.10, P-value = 0.048 and 0.005, respectively; Supporting Information Table S3). The adjusted HRs for banding vs. control patients remained non-significant for at least medium risk alcohol consumption, self-reported alcohol problems, and alcohol abuse diagnosis (Supporting Information Table S3). When different surgical procedures were compared, patients treated with gastric bypass had increased risk for at least medium risk alcohol consumption, self-reported alcohol problems, and alcohol abuse diagnoses compared both to VBG (adjHRs = 1.91, 3.21, and 2.27, respectively) and banding (adjHRs = 2.28, 4.34, and 4.45, respectively; Supporting Information Table S3).
Among the 98 persons who were excluded from the analyses because of self-reported alcohol problems, high alcohol consumption, or missing data at baseline, there were 12 persons with alcohol abuse diagnosis during follow-up (median follow-up = 14 years), 8 persons with self-reported alcohol problems (median follow-up = 10 years), and 2 persons with medium risk alcohol consumption (median follow-up = 4 years).
In this report, we have for the first time investigated the long-term changes in alcohol consumption and prevalence of alcohol problems after bariatric surgery in a large prospective study. We show that obese patients undergoing gastric bypass and VBG are more likely to start reporting at least medium risk alcohol consumption or alcohol problems and have alcohol abuse diagnoses as compared to controls given usual care. In contrast, none of these alcohol risk parameters were significantly increased after banding in this study. When different surgical procedures were compared, patients treated with gastric bypass had significantly increased risk for alcohol abuse diagnoses, at least medium risk alcohol consumption and self-reported alcohol problems compared both to VBG and banding.
Our finding that gastric bypass surgery increases the risk of alcohol problems and abuse is well in line with a 2-year prospective cohort study showing that the risk is greater after gastric bypass compared to adjustable banding . The long-term follow-up in our study allows us to show that increased alcohol problems after bariatric surgery persist beyond 2 years and result in increased alcohol abuse diagnosis compared to controls given usual care. Our study, as well as three previous reports , shows that alcohol consumption is reduced during the first year after surgery, emphasizing the importance of long-term follow-up.
We also show that male sex and baseline smoking and alcohol consumption increase the likelihood of postoperative alcohol abuse diagnosis. These results are well in line with the study by King et al.  but they also report that age, preoperative AUD, and recreational drug use were independent predictors of postoperative AUD. In our study, age was not a significant predictor and no data on recreational drug use were available in our study.
The biological mechanisms behind the observed differences between the gastric bypass group compared to VBG and banding are unknown. While VBG and banding are restrictive procedures, gastric bypass alters the normal anatomy and physiology of the upper gut and it is possible that this affects alcohol uptake, effects, or metabolism. This idea is supported by previous studies demonstrating higher peak alcohol levels and longer alcohol elimination times after gastric bypass . These findings have been attributed to faster transport of alcohol to the small intestine and reduced first-pass metabolism of ethanol by alcohol dehydrogenase in the stomach . Studies on alcohol uptake after VBG have not been performed and it is unknown if alcohol metabolism in the stomach is affected by restrictive procedures . It is not possible to draw conclusions about additional biological or psychosocial mechanisms underlying the association between increased alcohol-related problems and bariatric surgery from our data.
During the entire follow-up period, 93.1% of the surgery patients consistently report an alcohol consumption that is classified as low risk by the WHO . In addition, the mean alcohol consumption levels observed in our study are within a range that has been shown to be associated with a reduced relative risk of overall mortality compared to abstainers . Furthermore, bariatric surgery is associated with a number of positive health effects. In the SOS study, bariatric surgery improved quality of life , reduced overall mortality , and reduced the incidence rates of cancer , cardiovascular events , and diabetes [1, 3]. Therefore, the increased risk for alcohol problems and abuse after gastric bypass or VBG should be balanced against the multitude of positive health effects [1, 4, 5, 7] and increased quality of life . However, postoperative AUD is a potential side effect and in addition to the general detrimental health effects of alcohol, bariatric surgery patients with AUD may be at greater risk for nutritional deficiencies or weight regain.
A limitation of the SOS study is that the intervention could not be randomized for ethical reasons and that we have some differences in baseline characteristics (i.e., smoking) between the control and surgery groups. Such differences and other differences that we could not control for could potentially affect our findings. Another limitation is that the Swedish National Patient Register only contains in-patient data and alcohol abuse diagnoses from outpatient clinics were therefore not included. This may result in underestimation of the incidence of alcohol abuse diagnoses. Underestimation of the negative effects of a given alcohol consumption level may occur in the gastric bypass (GBP) patients since they are more sensitive to alcohol . There is also a risk that persons with the proclivity or history of at-risk drinking may improve their behavior in the period leading up to surgery and subsequently relapse after obtaining the surgical procedure. Furthermore, self-reported data prior to surgery may have been biased by concerns for surgery eligibility. However, this does not explain the observed differences between gastric bypass and the other surgical techniques, since the patients chose between surgical and non-surgical treatment but had no influence on the type of surgical procedure selected. An additional limitation of the SOS study is that many of the participants underwent VBG which is an older surgical procedure that is not used today. We only found an increased risk for alcohol problems in the gastric bypass and VBG groups, but it is possible that the lack of significance for the banding patients compared to the controls is due to insufficient statistical power. Other limitations include that the alcohol problem question has not been validated, the WHO alcohol risk consumption levels may not be optimal for GBP patients and that the ICD codes used for identification of alcohol abuse diagnosis during hospitalizations may reflect both current and previous alcohol abuse. Hence, our incidence rates may not be directly translated to incidence of AUD.
Strengths of the study are the large number of well-characterized patients in a controlled setting with long-term follow-up. Furthermore, alcohol related problems were assessed by three different alcohol risk parameters, which generated similar results. In addition, per protocol analysis gave very similar results to the intention to treat analysis.
The patients in the SOS study were recruited from all parts of Sweden and exclusion criteria were minimal and aimed at obtaining an operable surgery group. This indicates that our findings may be generalized to other Caucasian populations. Patients considered to have alcohol or drug-related problems were not allowed to enter the study. Therefore, our cohort may not be fully representative of the overall obese population but our exclusion criteria are in line with current preoperative assessment guidelines  indicating that our cohort is similar to other groups of obese patients eligible for bariatric surgery.
We conclude that gastric bypass and VBG increase the likelihood to report at least medium risk alcohol consumption and alcohol-related problems and these procedures are associated with a higher incidence of alcohol abuse diagnoses compared to control subjects given usual care. Furthermore, gastric bypass is less favourable compared to VBG and banding for all three alcohol risk parameters. Even though only a minority of the patients seem to be affected, the results suggest that patients should be informed about the increased risk before they undergo bariatric surgery and that the postoperative care should include assessment of alcohol-related problems.
Authors thank the staff members from the 25 surgical departments and 480 primary health-care centers in Sweden that participated in the study. Gerd Bergmark is acknowledged for invaluable administrative support in the SOS study.