Among United States adults, prevalence of severe obesity (body mass index (BMI) ≥ 35 kg/m²) has doubled from 1999–2000 (4.7%) through 2017–2018 (9.2%).¹ Population studies observed that patients with severe obesity have increased risks of cardiometabolic diseases² and mortality³. While bariatric surgery is the most successful treatment for the severely obese population⁴, it remains under-utilized.⁵ Patients who have undergone bariatric surgery have demonstrated significant improvement in obesity-related comorbidity such as remission of type 2 diabetes mellitus and decreased long-term mortality when compared to non-surgical subjects with severe obesity,^4,6,7 stimulating intense interest in discovery of causal pathophysiologic mechanisms that may facilitate non-surgical treatment(s) of obesity.⁸

This retrospective study compared long-term mortality of bariatric surgical patients and matched non-surgical subjects identified from driver licenses. While mortality after gastric bypass surgery was previously reported from 1984 through 2002,⁹ this study extends mortality follow-up through 2021. Additional gastric bypass patients and patients who had gastric banding, sleeve gastrectomy or duodenal switch from 1982 through 2021 have been included. Mortality outcomes were also analyzed stratified by sex, types of bariatric surgeries, and patients’ ages at surgery.

This study was approved by the University of Utah’s Resource for Genetic and Epidemiologic Research. Due to the deidentification, the study was considered non-human subjects research by the University Institutional Review Board (IRB #00095902). IRB approval was also obtained at Intermountain Healthcare.

Table 1 indicates basic characteristics of the 21,837 matched surgical and non-surgical pairs. The majority of subjects were female (79%). The most common procedure utilized was Roux-en-Y gastric bypass (n=15,110; 69.2%), followed by gastric sleeve (n=3,050; 14.0%), adjustable gastric banding (n=2,629; 12.0%), and duodenal switch (n=1,048; 4.8%). Over 40 years (median=10.8, interquartile range, 5.0 to 19.1 years) there were 2,943 (13.5%) deaths for surgical patients and 3,181 (14.6%) deaths for non-surgical subjects. Death rates during the first year following index date were similar between groups, 111 deaths (0.5%) and 89 deaths (0.4%) in the surgical and nonsurgical groups, respectfully, p=0.14 (Table 1). Mean age at index date was 42.2 years (SD: 11.7 years) and 42.3 years (SD: 11.9 years, p=0.60, for surgical and non-surgical subjects, respectively. Although mean BMI was significantly different between groups, the difference was only 0.2 kg/m², indicating close BMI matching. Of the surgical and non-surgical subjects, 94.4% and 87.6% (P<0.001) were Caucasian, respectively, while 7.8% and 12.4% (p<0.001) of surgical and non-surgical subjects were Hispanic, respectively. Numbers of patients undergoing each type of bariatric surgery and year of surgery are shown in Supplementary Appendix Table 3. In addition, the numbers of subjects in each group meeting study inclusion and exclusion criteria are presented in Figure 1.

Distribution of deaths and death rates per 10,000 person years for all matched subjects and for sex-specific matched subjects are presented in Table 2. Rate of death from all causes was 16% lower in the surgery compared to non-surgery groups (hazard ratio (HR), 0.84; 95% confidence interval (CI), 0.79–0.90; P<0.001; Figure 2). Total mortality was also significantly lower for surgical than non-surgical subjects for both females (HR, 0.86; 95% CI, 0.80 – 0.93; P<0.001) and males (HR, 0.79; 95% CI, 0.69–0.90; P<0.001). Non-external deaths were lower among surgical versus non-surgical subjects for combined sexes (HR, 0.74; 95% CI, 0.69–0.80; P<0.001), females (HR, 0.76; 95% CI, 0. 70–0.83; P<0.001) and males (HR, 0.68; 95% CI, 0. 59–0.79; P<0.001; Figure 2). Death rates were significantly lower for surgical compared to non-surgical subjects (combined sexes) for diabetes (72% lower, P<0.001), major cardiovascular disease (29% lower, P<0.001), cancer (43% lower, P<0.001), and chronic lung disease (39% lower; P=0.04). However, deaths from chronic liver disease were significantly greater among surgical than non-surgery subjects (combined sexes) (83% higher, P=0.02). For combined sexes and for females and males, external-caused death rates were significantly greater in the surgery group compared to non-surgery group. Accidents and adverse effects were 92% higher (P=0.03) and suicide was 140% higher (P<0.001) (Figure 2). Suicide after bariatric surgery was also significantly greater for both females and males.

Figure 3 indicates relative mortality risk in the surgery patients compared with non-surgery subjects by age category at the time of surgery or driver license application/renewal. While the 18 through 34 and 35 through 44-year age categories were the only groups to have significantly greater risk for external causes of death (HR, 3.63; 95% CI, 2.37–5.59; P<0.001 and HR, 2.19; 95% CI, 1.46–3.31; P<0.001, respectively), only the 18 through 34-year group had significantly greater suicide mortality (HR, 5.08; 95% CI 1.97–13.09; P=0.001).

Except for the 18-to-34-year group, all age categories had significantly lower mortality from all causes of death for surgery compared to non-surgery subjects (Figure 3). Significantly greater all causes of death risk for surgical versus non-surgical subjects in the 18 through 34-year category appeared related to greater risk for external causes of death among surgical patients in this age group. Chronic liver disease death risk was significantly greater among surgery versus non-surgery subjects in the 18 through 34-year age category, but the number of deaths was small leading to large confidence intervals.

Analyses of mortality risk between surgical patients and non-surgical subjects were stratified by type of bariatric surgical procedure (Figure 4). Gastric bypass, gastric banding, and sleeve gastrectomy each showed significantly lower total mortality risk for surgical compared to non-surgical subjects: gastric bypass (HR, 0.85; 95% CI, 0.79–0.91; P<0.001); gastric banding (HR, 0.72; 95% CI, 0.55–0.94; P=0.017); and sleeve gastrectomy (HR, 0.49; 95% CI, 0.30–0.79; P=0.004). Patients undergoing duodenal switch were the fewest in number and these procedures were performed later in the follow-up time-period. Cancer, diabetes, and cardiovascular disease mortality were all significantly reduced after gastric bypass surgery. The number of deaths for other surgery procedures for cause-specific mortality were small, although diabetes mortality was significantly lower after sleeve gastrectomy.

Unadjusted Kaplan-Meier survival curves were estimated for non-external-, external-, and all-cause mortality risk by surgical and non-surgical subjects (Supplemental Figure 1). Restricted mean survival time for non-external deaths was 1.72 years longer (95% CI 1.35–2.09; P<0.001) for surgical compared to non-surgery subjects (Supplemental Figure 1a). Survival time for external deaths was 0.4 years less (95% CI, −0.56 to −0.25; P<0.001) for surgery patients versus non-surgery patients (Supplemental Figure 1b). Combined, survival time for all causes of death was 1.3 years (95% CI, 0.93–1.67; P<0.001) longer for surgery compared to non-surgery subjects (Supplemental Figure 1c).

Advancing the understanding of clinical efficacy of bariatric surgery has focused on short- and long-term outcomes of obesity-related comorbidities, including type 2 diabetes mellitus (T2DM),^4,12–14 primary cardiovascular (CVD) risk factors,^15–17 CVD events such as myocardial infarction and stroke,^18–20 cancer,^21,22 as well as all-cause mortality.^7,9,23 Multiple association studies relating bariatric surgery and mortality outcomes have been reported, with wide-variation in study design pertaining to: subject number; control cohorts; mean follow-up; procedure type; age at surgery; clinical endpoints (i.e. life expectancy and death rates for all cause and cause specific); and presence or absence of prevalent diabetes.^3,6,24–36 Study type has been predominately retrospective in nature, with the Swedish Obesity Subjects (SOS) study prospectively studied.^7,23

This 40-year retrospective Utah study extends the mortality risks after bariatric surgery from the previously reported matching of 7,925 surgical and non-surgical pairs to 21,837 matched pairs. This study now includes patients who had undergone the most common bariatric procedures performed today, doubling the number of previously reported gastric bypass patients (n = 15,110), and including 2,629 gastric banding, 3,050 sleeve gastrectomy, and 1,048 duodenal switch patients. Mortality benefits related to bariatric surgery were shown to remain durable for multiple decades following surgery, with a significant 16% lower all-causes of death for surgical versus matched non-surgical subjects, and in addition, a significant 14% and 21% reduction in all-cause mortality for female and male surgical patients, respectively, compared to non-surgical subjects. Previously reported results did not show reduced mortality among the male surgical population but were based on much smaller numbers of males. Further, improved cause-specific mortality (cardiovascular, diabetes, and cancer) was shown for surgical patients compared to matched non-surgical subjects. Cardiovascular and diabetes mortality risk was lower for both male and female surgical patients, and lower for cancer for female surgical patients compared to the respective non-surgical subjects. Finally, new to this study is a reported increase in deaths from cirrhosis of the liver, occurring primarily among the surgical patients who underwent bariatric surgery between ages 18 through 34 years. Clinical information concerning details of liver cirrhosis were not available, preventing us from further investigation of alcohol use, viral hepatitis, or fibrosis before or after baseline.

Surgical patients undergoing surgery between ages 35 to 44 years, 45 to 54 years and 55 to 80 years had significantly lower mortality from all causes of death compared to non-surgery subjects, suggesting bariatric surgery to be associated with a lower non-external mortality risk (i.e., cardiovascular disease and cancer). Significant results only in the older age groups should not imply patients necessarily postpone surgery until older age, as post-surgical complications have been shown to increase with increasing age at surgery and surgical postponement may result in worsened clinical status related to certain conditions such as orthopedic joint health. Finally, significantly improved all-cause mortality for surgical versus non-surgery subjects was shown for individual surgical procedures (gastric bypass, gastric banding, and sleeve gastrectomy).

Consistent with previous findings,⁹ deaths related to external causes such as suicide and accidents were significantly greater (2.35 times) among bariatric surgery patients compared to matched non-surgical subjects, with most of this increased mortality risk occurring among patients who were ages 18 to 34 years at the time of surgery. Utah State mortality rates for suicide were 1.0/10,000 person-years for females and 3.6/10,000 for males ages 15–74 for the years 1999–2020.(https://ibis.health.utah.gov/) These rates are comparable to the non-surgical subjects of our study with severe obesity suicide rates of 1.3/10,000 person-years (female) and 2.9/10,000 person-years (male). The surgery group rates of 3.0 for females and 8.2/10,000 persons-years for males are clearly higher than the population as a whole.

As mentioned, among the youngest age group there was a significant increase in mortality from cirrhosis of the liver. Mitchell et al have reported that following certain bariatric surgery procedures there is increased disinhibition and impulsivity and increased rates of absorption of alcohol.³⁷ However, a recent study by Aminian, et al., retrospectively followed patients who underwent bariatric surgery (n=650) and non-surgery patients with obesity (n=508) who all had baseline biopsy-proven fibrotic nonalcoholic steatohepatitis (NASH) without cirrhosis.³⁸ After median follow-up of seven years, bariatric surgery was significantly associated with a lower risk of major adverse liver outcomes, with 5 surgical and 40 non-surgical patients experiencing the adverse liver outcomes.³⁸ Alcohol use and clinical liver disease details were not available in our study. One possibility that would explain these discrepant results includes that the increased mortality after bariatric surgery in our study was derived from those with possible liver cirrhosis at baseline, and once cirrhosis has occurred resolution is limited after surgery. Another possibility is that the risk estimates are unstable, despite the statistical significance, due to small sample sizes.

Recently reported mortality outcomes of the Swedish Obesity Subjects (SOS) study, with median follow-up of 20 to 24 years, reported a 23% lower mortality in the bariatric surgery group (n = 2007 patients) compared to the matched control group (HR, 0.77; 95% CI, 0.68 to 0.87; p<0.001). With 69%, 18%, and 13% of surgical patients undergoing vertical banded gastroplasty, banding, and gastric bypass, respectively, their results showed an unadjusted median life expectancy of 2.4 years greater than non-surgical matched controls (95% CI, 1.2 to 3.5; p<0.001). This reduction in mortality was similar to our Utah study in which risk from all causes of death was 16% lower among surgical patients compared to non-surgical subjects (HR, 0.84; 95% CI, 0.79–0.90; p<0.001), with a mean extended life expectancy for all causes of deaths of 1.3 years (95%CI, 0.93–1.67; P<0.001) The SOS study also reported a significantly reduced mortality of 30% and 23% for cardiovascular and cancer, respectively, for the surgery compared to non-surgical groups, similar to our Utah study which reported a significant 29% and 43% lower risk for major cardiovascular diseases and cancer, respectively, when comparing the two groups. Finally, as pointed out by SOS authors,⁷ when considering populations who are at increased clinical risk such as patients with severe obesity, minimal gains in mean life expectancy (i.e., 1.3 years) are meaningful. Even with minimal benefits on overall mortality, studies have shown significantly increased quality of life after bariatric surgery.¹³

A recently published one-stage meta-analyses explored the association of bariatric surgery with long-term survival, including patients with and without diabetes prior to their surgery. The study, which included 174,772 participants from 16 matched cohort studies and one prospective controlled trial, reported that bariatric surgery was associated with a significantly reduced hazard rate of death of 49.2% (95% CI 46.3 to 51.9; p<0.0001), with a calculated mean life expectancy of 6.1 years (95% CI 5.2 to 6.9) longer among the surgical patients when compared to the usual care, non-surgical controls.³⁹ Further, Syn et al.,³⁹ performed subgroup analyses and reported that while surgical patients who did and did not have baseline diabetes had lower mortality rates when compared to non-surgical subjects, long-term mortality and survival was more favorable for the bariatric surgery patients who at surgery had diabetes compared to surgical patients without baseline presence of diabetes. Further evidence of improved long-term survival following bariatric surgery has recently been reported by Homberg et al., ⁴⁰. Investigators compared the all-cause mortality of patients in Sweden and Finland who had undergone gastric bypass with patients receiving sleeve gastrectomy. Among the 61,503 patients who were followed for a mean of 6.8 person-years, both bariatric surgical procedures had similar all-cause mortality over the entire study duration. Sub-group analyses suggested that patients presenting with diabetes status at time of surgery may have more favorable long-term mortality following gastric bypass when compared to sleeve gastrectomy.⁴⁰ Our retrospective study did not have clinical outcomes relating to whether patients and driver license subjects had baseline diabetes.

An important limitation of our study relates to the lack of clinical data at the time of bariatric surgery or application date, as well as absence of clinical surveillance throughout the study. For example, we don’t know how much weight was lost after surgery and throughout the follow-up period or the degree of weight maintenance in the driver license subjects. The possibility that surgical and non-surgical subjects were similar in physical health status at surgery or at driver license application is supported by a long-term prospective study our group has conducted on a subset of these surgical patients.¹³ At baseline of the prospective study, there were no significant differences in clinical endpoints, (smoking status, hypertension, diabetes, dyslipidemia, or sleep apnea) between patients undergoing bariatric surgery (n=420), patients seeking bariatric surgery but did not have surgery (n=415), and subjects randomly selected from the population who were severely obese and not seeking bariatric surgery (n=321).⁴¹ Current smoking prevalence in that study was 8% in the surgical subjects compared with a northern Utah prevalence of 11% in persons ages ≥18 using data from the Behavioral Risk Factor Surveillance System (http://health.utah.gov/opha/OPHA_BRFSS.htm). The low state smoking prevalence and similar prevalence in subjects having bariatric surgery hopefully minimize bias from not having smoking data. Further, at 12-years follow-up of our prospective study, with over 90% follow-up rate, the bariatric surgery group had sustained a mean change in body weight from baseline of −35 kg (−26.9%), while mean weight changes of the two non-surgical groups were −2.9 kg and 0 kg, respectively, minimizing the effects weight change might have over the follow-up period. Nevertheless, whether non-surgical subjects who were part of our mortality analysis were less likely to choose and participate in healthy lifestyle practices or were less likely to engage in medical screening and treatment compared to patients who underwent bariatric surgery remains in question.

Because some of the bariatric procedures were developed more recently, follow-up time of these procedures was much shorter than for gastric bypass surgery. Therefore, the numbers of deaths after these surgeries were smaller, resulting in wider confidence intervals around the hazard ratios. Hazard ratios were separately provided for each procedure to address this limitation.

Multiple methodological-related strengths are associated with our study, specifically extended length of follow-up and different bariatric surgery procedures. This study extends from near the time gastric bypass procedures began in the U.S. to approval and initiation of gastric banding, followed by the rise in gastric sleeve. Close matching of surgical patients to driver license subjects facilitated important physiologic and temporal alignment by which to compare mortality between groups, and the analyses of age at surgery provided unique insight into study results. Finally, the Utah Population Database strengthened the accuracy of mortality data extraction.

In conclusion, results of this study attest to the decades-long durability of bariatric surgery in reducing death from all causes and reducing deaths related to cardiovascular disease, cancer and diabetes when compared to matched subjects with severe obesity. In addition, favorable mortality outcomes were evident for major bariatric surgery procedures. Serious concern, however, continues to be exhibited regarding increased mortality following bariatric surgery in relation to suicide, accidents, and cirrhosis of the liver. This study showed the primary group associated with this untoward mortality outcome are patients choosing to have bariatric surgery between ages 18 to 34 years, suggesting this age group may require more aggressive pre-surgical psychological screening and post-surgery follow-up. Finally, with what appears to be an ever-increasing rise in percentage of individuals with severe obesity, coupled with the realization that in practicality bariatric surgery has limited treatment delivery, there remains an important research need to discover physiologic and biomolecular mechanisms leading to non-surgical treatment that result in weight loss and improved mortality similar to that achieved by bariatric surgery.