Increased risk for major depressive disorder in severely obese patients after bariatric surgery — a 12-year nationwide cohort study

Abstract

Background: Bariatric surgery is associated with a significant improvement in depressive mood in the initial postoperative years, but the maintenance of the improvement is under debate.

Aim: To explore the association between bariatric surgery and major depressive disorder (MDD) in a 12-year nationwide cohort study.

Method: Using the National Health Insurance Research Database of Taiwan, we identified 2302 patients who underwent bariatric surgery in 2001–2009. These patients were matched by propensity score to 6493 obese patients who did not receive bariatric surgery. We followed the surgical and control cohorts until death, any diagnosis of MDD or 31 December 2012. We used Cox proportional hazard regression models to calculate the relative risk of MDD in those who received bariatric surgery.

Results: Overall, there was a 1.70-fold (95% CI: 1.27–2.27) higher risk of MDD in the surgical group. Subjects receiving malabsorptive procedures showed a higher risk of MDD (3.01, 95% CI: 1.78–5.09) than those receiving restrictive procedures (1.51, 95% CI: 1.10–2.07). Stratified by follow-up period, there was a higher risk of MDD in the surgical group (2.92, 95% CI: 1.75–4.88) than in the restrictive group four years after bariatric surgery.

Conclusions: Bariatric surgery was significantly associated with an elevated risk of MDD.

KEY MESSAGES

• Bariatric surgery is associated with a significant improvement in depressive mood in the initial postoperative years, but the improvement is not maintained. Less is known about the relationship between bariatric surgery and risk of major depressive disorder.

• This was the first nationwide cohort study which found that bariatric surgery was significantly associated with an elevated risk of MDD (aHR: 1.70; CI: 1.27–2.27), mainly with malabsorptive procedures (aHR: 3.01; CI: 1.78–5.09) and at time points more than four years after surgery (aHR: 2.92; CI: 1.75–4.88) compared with the risk in matched controls. These findings imply an association between long-term malabsorption and the postoperative incidence of MDD.

• Long-term malabsorption might be related to the incidence of major depressive disorder after bariatric surgery. The possible causal relationship between nutritional deficiency after bariatric surgery and major depressive disorder warrants further investigation.

Keywords:

• Bariatric surgery
• depression
• obesity

Introduction

Obesity is an important public health issue globally. The worldwide prevalence of overweight and obesity in adults was ∼36.9% in men and 38.0% in women as of 2013 [1]. The estimated prevalence of obesity among the middle-aged is 15%–20% in Europe [2]. The most effective and sustained treatment for severe obesity is bariatric surgery, which has increased more than seven-fold worldwide in the last decade [3]. Moreover, bariatric surgery is associated with a long-term reduction in overall mortality and with decreased incidences of diabetes, myocardial infarction, and stroke [4,5]; but an increase in risk of fracture [6]. Depression has been another important health issue in the last few decades. Approximately 20% of US adults have at least one episode of major depressive disorder (MDD) in their lifetime [7]. Additionally, the World Health Organization has ranked depression as the fourth leading cause of disability worldwide [8]. Obesity and depression interact closely. Unsurprisingly, depression is the most common mental health condition among patients seeking and undergoing bariatric surgery [9]. Bariatric surgery is associated with a significant reduction in depressive mood over the initial postoperative years; but this improvement is not maintained [10–16]. One study examined depression before and after bariatric surgery using the Beck Depression Inventory (BDI) questionnaire and observed dramatically lower in the BDI scores during the first year after bariatric surgery than 4 years later [11]. In an intervention study of obesity using 487 Swedish obese subjects matched 1:1 with conventionally treated controls, the surgical group had a greater improvement in quality of life than the conventional group. Although the improvement was maintained for 2 years after surgery, the peak value of quality of life in the surgical group was observed at 6–12 months [12]. Similar studies subsequently examined the effect of bariatric surgery on depression over a longer period [13–16]. In a 3-year follow-up study, depressive symptoms in severely obese subjects were persistently improved in the postoperative compared to the preoperative condition [15], whereas there was no difference in BDI scores between obese patients who had undergone gastric bypass and obese patients without gastric bypass in a 7-year follow-up study [14].

Several studies have aimed to determine why the positive impact on depressive mood after bariatric surgery disappears after a few years. The most common cause of this decline is unclear, but candidates include disappointment from expectations of weight loss, the occurrence of weight regain, and recurring comorbidities [9,17]. However, the association between preoperative mental health conditions and the degree of postoperative weight loss remains unclear [9,18]. In addition, poor nutrient absorption could lead to depression, as noted in updated clinical guidelines [19]. A decline in the serum level of antidepressants was also found in patients after bariatric surgery, supporting the hypothesis of malabsorption [20].

The aim of this study was to compare the incidence of MDD between severely obese patients undergoing bariatric surgery and matched severely obese controls. We also examined the relationship between bariatric surgery and the risk of MDD for different surgical types and observation periods.

Materials and methods

Data source

We used the National Health Insurance Research Database (NHIRD) to conduct this study. The National Health Insurance program, which began in 1995, has enrolled up to 99% of Taiwan’s 23 million people [21]. The present study was approved by the institutional review board of the National Health Research Institutes in Taiwan (EC1030701-E).

From the NHIRD, we selected those who had a diagnosis of morbid obesity (ICD-9-CM 278.01, n = 21465) between 2001 and 2009. A detailed flowchart of the selection process and the matching of controls is shown in Figure 1. After excluding those who received bariatric surgery before the diagnosis of morbid obesity (n = 287), we divided the candidate subjects into 2 groups based on whether they received bariatric surgery during the follow-up period. For the surgical group (n = 2689), we also excluded those who had a prior diagnosis of depression (ICD-9-CM 296.2 and 296.3, n = 157) before surgery and those who had missing data (n = 2). At the end, the remaining subjects who had morbid obesity but did not receive bariatric surgery served as the control pool for the subsequent matching procedures. The first date of bariatric surgery for each subject was defined as the index date for that subject in the surgical group. The propensity scores of all study subjects were calculated by using multivariable logistic regression adjusting for age, sex, Charlson Comorbidity Index (CCI) [22], diabetes, hypertension, hyperlipidemia, and the year when morbid obesity was diagnosed. To select individuals who did not receive bariatric surgery as a comparable control group, we used the 1:3 propensity score matching method [23] and the nearest neighbor algorithm with a perfect proportion of 0.995–1.0 [24]. We assigned the index date of each surgical subject to the corresponding matched controls. In the control group, we also excluded 312 subjects who had a prior diagnosis of depression and 15 subjects with missing data. After the matching procedures were complete, 2302 subjects who received bariatric surgery and 6493 controls who did not receive bariatric surgery were included in further analyses. We followed the study subjects until death, any diagnosis of major depressive disorder, or 31 December 2012, whichever came first.

Figure 1. Diagram showing the flow of subjects through the process of selection and matching.

The clinical procedure codes for bariatric surgery used in this study were divided into 2 subgroups: the malabsorptive procedures included ICD-9-CM 44.31 (high gastric bypass), 44.38 (laparoscopic gastroenterostomy), and 44.39 (other gastroenterostomy without gastrectomy), while the restrictive procedures included ICD-9-CM 43.82 (laparoscopic sleeve gastrectomy), 43.89 (open and another partial gastrectomy), 44.68 (laparoscopic gastroplasty), 44.69 (another repair of stomach), 44.95 (laparoscopic gastric restrictive procedure), and 44.99 (other operations on stomach).

Definition of research variables

The main outcome of this study was the occurrence of MDD defined as an admission diagnosis of MDD or the use of antidepressant drugs (ICD-9-CM code 296.2 or 296.3). The main comorbidities controlled in this study include diabetes (ICD-9-CM code 250), hypertension (ICD-9-CM code 401), and hyperlipidemia (ICD-9-CM code 272). Study subjects were considered to have comorbidities if they had at least 2 ambulatory visits and 1 hospitalization with the corresponding diagnosis within 3 years before the first diagnosis of morbid obesity.

Statistical analysis

The baseline data are presented by frequency, with percentages for categorical variables and the means with standard deviation for continuous variables. t tests and χ² tests are used to describe the differences between the surgical and control groups for categorical variables and continuous variables, respectively. The incidence of MDD is defined as the number of events divided by the number of follow-up person-years, which was calculated from the index date to the diagnosis of MDD, death, or 31 December 2012, whichever occurred first. We used Cox proportional hazards regression to assess adjusted hazard ratios (aHRs) and 95% confidence intervals (CIs) of MDD for the surgical group compared to the controls. The Schoenfeld residuals test [25] and complementary log-log plots showed that the proportional hazards assumption was not violated. In the multivariable analyses, we adjusted for all covariates shown in Table 1. To test the possibility of selection bias, we also followed up the excluded depressive subjects after the index date to compare the differences in readmission for MDD and regular use of antidepressant drugs between the bariatric surgery and control groups. To evaluate the effects of different surgical procedures, we divided the bariatric surgery group into those who received malabsorptive procedures and those who received restrictive procedures. And to clarify the changes in depression risk during the follow-up, we also estimated the HRs of depression in the following time periods: <1, 1–1.99, 2–2.99, 3–3.99 and >4 years after the index date. All p values were 2-sided, and differences with p values <.05 were considered significant. All analyses were conducted using SAS version 9.4 (SAS Institute Inc, Cary, North Carolina).

Table 1. Baseline characteristics of the patients receiving bariatric surgery and the matched controls.

	Matched controls N = 6493	Bariatric surgery N = 2302
Years of follow-up, mean(SD)	4.98 (2.06)	4.89 (2.17)
Age (years), mean (SD)	32.8 (10.7)	32.2 (9.8)
Age group, n (%)
0–19	414 (6.4)	153 (6.7)
20–34	3846 (59.2)	1366 (59.3)
35–49	1825 (28.1)	643 (27.9)
50–64	400 (6.2)	137 (6.0)
≥65	8 (0.1)	3 (0.1)
Sex, n (%)
Male	2498 (38.5)	877 (38.1)
Female	3995 (61.5)	1425 (61.9)
CCI, n (%)
0	4935 (76.0)	1721 (74.8)
1–2	1191 (18.3)	440 (19.1)
≥3	367 (5.7)	141 (6.1)
Year of morbid obesity diagnosis, n (%)
≤2005	1703 (26.2)	594 (25.8)
2006–2007	2046 (31.5)	724 (31.5)
2008–2009	2744 (42.3)	984 (42.7)
Diabetes, n (%)	699 (10.8)	261 (11.3)
Hypertension, n (%)	1280 (19.7)	430 (18.7)
Hyperlipidemia, n (%)	917 (14.1)	337 (14.6)

The propensity score matched controls were matched by age, sex, CCI, history of diabetes, hypertension, hyperlipidemia, and year of morbid obesity diagnosis.

CCI: Charlson Comorbidity Index.

Results

The surgical patients, 61.9% of whom were women, had a mean age of 32.2 years. The matched controls, 61.5% of whom were women, had a mean age of 32.8 years. The two groups were similar in most of the baseline characteristics for which they were matched (Table 1). Since we applied propensity score matching according to age group (treated as a categorical variable rather than a continuous variable), there was a difference in mean age between the surgical and the control groups (p = .0199). The difference would be further adjusted when applying Cox regression multivariate regression analyses.

By the end of the 12-year study period, there were a total of 72 incidences of MDD in the surgical group (mean follow-up 4.89 years, 6.40 cases per 1000 person-years) and 123 incidence of MDD in the matched control group (mean follow-up 4.98 years, 3.81 cases per 1000 person-years) (Table 2). Overall, there was a significant 1.70-fold increase (95% CI: 1.27–2.27) in the risk of MDD in the surgical group compared with the controls after adjusting for age, sex, CCI, diabetes, hypertension, hyperlipidemia and the year morbid obesity was diagnosed. We also followed up the excluded depressive subjects after the index date and there was no significant difference in readmission for MDD or regular use of antidepressant drugs between the bariatric surgery (123.4 per 1000 person-years) and control (133.0 per 1000 person-years) groups (data not shown). Therefore, the exclusion of prior depression did not increase selection bias. With regard to the type of surgery, 337 (14.6%) subjects underwent malabsorptive procedures and 1965 (85.4%) underwent restrictive procedures. The adjusted relative risk of MDD was 3.01 (95% CI: 1.78–5.09) in patients who underwent malabsorptive procedures and 1.51 (95% CI: 1.10–2.07) in those who underwent restrictive procedures compared with controls. Furthermore, we found that the malabsorptive group had an elevated risk of MDD (adjusted HR: 2.27, CI: 1.28–4.03) compared with the restrictive group.

Table 2. Risk of major depressive disorder in patients receiving bariatric surgery compared with matched controls, by type of surgical procedure.

	N	Event	Person-years	Incidence^a	HR	Adjusted HR
Matched controls	6493	123	32304.48	3.81	1.0	1.0
Bariatric surgery	2302	72	11247.87	6.40	1.69 (1.26–2.26)**	1.70 (1.27–2.27)**
Malabsorptive procedure	337	16	1534.69	10.43	2.76 (1.64–4.65)**	3.01 (1.78–5.09)**
Restrictive procedure	1965	56	9713.17	5.77	1.52 (1.10–2.08)*	1.51 (1.10–2.07)*

Adjusted HR was adjusted for age, sex, Charlson Comorbidity Index, history of diabetes, hypertension, hyperlipidemia, and year of morbid obesity diagnosis.

Malabsorptive procedures included ICD-9-CM 44.31 (high gastric bypass), 44.38 (laparoscopic gastroenterostomy), and 44.39 (other gastroenterostomy without gastrectomy).

Restrictive procedures included ICD-9-CM 43.82 (laparoscopic sleeve gastrectomy), 43.89 (open and other partial gastrectomy), 44.68 (laparoscopic gastroplasty), 44.69 (other repair of stomach), 44.95 (laparoscopic gastric restrictive procedure), and 44.99 (other operations on stomach).

^aIncidence was represented per 1000 person-years.

* p < .05; **p < .001.

In Kaplan–Meier analysis (Table 3 and Figure 2), the incidences of MDD were 4.90 per 1000 person-years during the first year, 5.10 per 1000 person-years during the second year, 4.89 per 1000 person-years during the third year, 6.51 per 1000 person-years during the fourth year and 9.26 per 1000 person-years past the first 4 years for the surgical group, compared with 4.84, 3.73, 3.93, 3.26 and 3.33, respectively, for the matched controls. Our findings revealed an elevated risk of MDD in the surgical group (aHR: 2.92, 95% CI: 1.75–4.88) four years after bariatric surgery.

Figure 2. Cumulative incidence of major depressive disorder in patients receiving bariatric surgery and matched controls, using Cox proportional hazards regression analyses with adjustment for age, sex, Charlson Comorbidity Index, history of diabetes, hypertension, hyperlipidemia, and year of morbid obesity diagnosis.

Table 3. Risk of major depressive disorder in patients receiving bariatric surgery and matched controls by follow-up period.

Time since index date (years)	N	Event	Person-years	Incidence^a	HR	Adjusted HR
<1
Matched controls	6493	31	6409.02	4.84	1.0	1.0
Bariatric surgery	2302	11	2245.07	4.90	1.02 (0.51–2.02)	1.01 (0.51–2.02)
1–1.99
Matched controls	6287	23	6170.81	3.73	1.0	1.0
Bariatric surgery	2187	11	2155.36	5.10	1.37 (0.67–2.81)	1.38 (0.67–2.83)
2–2.99
Matched controls	6018	23	5859.23	3.93	1.0	1.0
Bariatric surgery	2102	10	2046.67	4.89	1.25 (0.59–2.62)	1.25 (0.59–2.62)
3–3.99
Matched controls	5606	16	4903.98	3.26	1.0	1.0
Bariatric surgery	1934	11	1688.93	6.51	2.00 (0.93–4.30)	1.97 (0.92–4.25)
≥4
Matched controls	4221	30	9022.03	3.33	1.0	1.0
Bariatric surgery	1470	29	3132.90	9.26	2.81 (1.68–4.68)**	2.92 (1.75–4.88)**

Adjusted HR was adjusted for age, sex, Charlson Comorbidity Index, history of diabetes, hypertension, hyperlipidemia, and year of morbid obesity diagnosis.

^aIncidence was represented per 1000 person-years.

** p < .001.

Discussion

To our knowledge, this is the first nationwide cohort study to compare the incidence of MDD between obese patients who have undergone bariatric surgery and obese controls who have not. We found a significant 1.50-fold increase in the risk of MDD in the surgical group compared with the control group. In addition, the risk of MDD increased 2.36-fold with the malabsorptive type of surgery and 1.38-fold with the restrictive type of surgery compared with controls, implying that malabsorption plays a more important role than restricted food intake in MDD development. It was also found that the incidence of MDD increased 4 years after surgery compared with the incidence in controls. These findings depict the trajectory of MDD development after bariatric surgery.

In previous studies, bariatric surgery was associated with a modest reduction in clinical depressive mood over the initial postoperative years, but this reduction was not sustained. Most studies prospectively compared scores on a depressive scale, using questionnaires prior to and after the surgery, and found that improvement peaked at 6–12 months [12,26] or one year after surgery [15,16]. Nevertheless, the improvement became attenuated and non-significant compared with the preoperative condition after a seven-year follow-up [14]. One British study, defining clinical depression as a medical diagnosis recorded in the British Clinical Practice Research Datalink, found that a sample of 3045 patients had an increased prevalence of depression compared with their preoperative status, although the difference was not statistically significant [10]. Similarly, a Swedish obese intervention study reported that peak improvements in health-related quality of life (HRQL) in the surgical group were observed during the first year of weight loss [27]. Interestingly, a cohort study in Norway compared changes in comorbidities and mood status a group receiving surgery and groups receiving specialized medical care and found that although the surgical groups maintained greater reduction of hypertension and diabetes, they also had a significantly higher incidence of new cases of MDD [28]. A meta-analysis also found that the evidence regarding the association between mental health conditions and postoperative weight loss was conflicting [9]. We wonder what factors might have resulted in new cases of MDD after weight reduction and the remission of obesity-related comorbidities.

Strikingly, the risk of MDD increased 3.01-fold in the malabsorptive surgery group and 1.51-fold in the restrictive surgery group compared with controls, implying that malabsorption plays a more important role than restricted food intake in MDD development. “Malabsorptive” operations combine two methods: restriction and malabsorption. Taking gastric bypass (RYGB: Roux-en-Y gastric bypass) as an example, the superior part of the stomach is cut down to a small pouch to restrict its food-storage capacity and the duodenum and jejunum are bypassed to cause malabsorption. “Restrictive” operations such as sleeve gastrectomy and gastric banding essentially decrease the size of the stomach. Therefore, malabsorptive procedures are more effective than restrictive procedures in weight reduction, but patients tend to have poorer digestion and absorption with the former type [29,30]. Accordingly, clinical guidelines have urged caution against micronutrient deficiency after bariatric surgery [19,31,32]. To date, no study has examined the association between micronutrient deficiency and depression after bariatric surgery, despite their theoretical relation. One review examined the association between nutrition and depression, and found that deficiencies of micronutrients (folate and vitamin B₁₂) and trace minerals (iron, zinc and selenium) tended to be more common among depressed than non-depressed persons [33]. Folate and vitamin B₁₂ are two of the most commonly deficient micronutrients after bariatric surgery [34,35]. The prevalence of folate deficiency was similar among malabsorptive and restrictive operations, whereas vitamin B₁₂ deficiency was more prevalent with malabsorptive operations than with restrictive operations [35]. Iron, zinc and selenium are absorbed by the duodenum and jejunum. Therefore, patients undergoing bariatric surgery, especially those receiving malabsorptive procedures, are recommended to receive postoperative nutritional supplementation [19,30,32]. Currently, dietetic counseling is mandatory during the first year but optional later [19]. There is not yet a consensus on how long nutritional deficiencies should be monitored after surgery, and patients’ needs vary depending on surgical type, categories of nutrients and the year when surgery was received [36]. In our study, the incidence of MDD increased 4 years after surgery compared with the controls, implying that the postoperative nutrient deficiency might progress slowly and is related to the late onset of MDD. Therefore, long-term monitoring for nutritional support after bariatric surgery is necessary.

The prevalence and incidence of MDD vary among different ages, periods, cohorts and countries. In the elderly in Taiwan, for example, the prevalence of MDD was 5.9%, which was the highest among the different age groups [37]. In different periods, the cumulative prevalence of MDD increased from 1.67 to 17.24 per 1000 and the annual incidence of treated MDD increased from 1.89 to 2.58 per 1000 in 10 years [38]. Furthermore, the lifetime prevalence for MDD ranges from 1.5 cases per 100 adults in Taiwan to 19.0 cases per 100 adults in Beirut. In addition, the annual incidence of MDD ranges from 0.8 cases per 100 adults in Taiwan to 5.8 cases per 100 adults in New Zealand [39]. In our current study, the overall incidence of MDD (surgical: control= 6.40:3.81 per 1000 person-years) was higher than the incidence in the Taiwanese population (2.58 per 1000 year-person), but lower than in western countries.

This study has several limitations. First, although we performed propensity score matching and excluded preoperative depression history in both groups to minimize the impact of measured covariates, our findings cannot demonstrate unobserved confounding factors. As our study used a matched controlled cohort, the estimated HRs addressed only the correlation, not the causal relationship, between bariatric surgery and the incidence of MDD. To minimize probable confounders, we excluded subjects with history of depression prior to surgery in both groups, and we matched comorbidities in controls by propensity scoring. More convincingly, compared to previous studies which tended to employ questionnaires, our study sought to minimize self-reported bias by defining MDD using medical records plus antidepressant prescriptions. Second, data on body mass index (BMI) before and after surgery were not available. Using ICD-9-CM diagnostic codes, we included all subjects with severe obesity defined as either BMI ≥ 40 kg/m² or BMI ≥ 35 kg/m² plus at least one comorbidity; however, the NHIRD did not record the severity of obesity. Although there was no record of BMI in our database, the previous investigation and meta-analysis implying the relationship regarding the degree of weight reduction and remission of MDD was inconsistent [9,9,18]. Third, we were unable to assess the use of self-administered medications and nutritional supplements, which may have altered moods and nutrition status. Despite the limitations, the data are longitudinal and nationally representative. There are very few studies that have looked at the association between bariatric surgery and new onset MDD postoperatively.

In conclusion, our study demonstrates that severely obese patients undergoing bariatric surgery had a higher risk of MDD than the matched controls, with the difference at a significant level 4 years after surgery. The elevated risk of MDD was more severe in patients receiving malabsorptive than restrictive procedures. The association between micronutrient deficiencies and depression after bariatric surgery warrants further investigation. Understanding the long-term outcomes of bariatric surgery in terms of MDD should be a priority for public health policy.

Acknowledgements

The authors would like to thank Mr. Marc Anthony, the lecturer at the Academic Writing Education Center, National Taiwan University, for his thoughtful revision of the English writing in this manuscript.

Disclosure statement

No potential conflict of interest was reported by the authors.