Error bars indicate 95% CIs.
eTable. Distribution of Deaths and Death Rates per 10,000 Person-years, According to Study Group and Age Categories
Customize your JAMA Network experience by selecting one or more topics from the list below.
Davidson LE, Adams TD, Kim J, Jones JL, Hashibe M, Taylor D, Mehta T, McKinlay R, Simper SC, Smith SC, Hunt SC. Association of Patient Age at Gastric Bypass Surgery With Long-term All-Cause and Cause-Specific Mortality. JAMA Surg. 2016;151(7):631–637. doi:10.1001/jamasurg.2015.5501
Copyright 2016 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Bariatric surgery is effective in reducing all-cause and cause-specific long-term mortality. Whether the long-term mortality benefit of surgery applies to all ages at which surgery is performed is not known.
To examine whether gastric bypass surgery is equally effective in reducing mortality in groups undergoing surgery at different ages.
Design, Setting, and Participants
All-cause and cause-specific mortality rates and hazard ratios (HRs) were estimated from a retrospective cohort within 4 categories defined by age at surgery: younger than 35 years, 35 through 44 years, 45 through 54 years, and 55 through 74 years. Mean follow-up was 7.2 years. Patients undergoing gastric bypass surgery seen at a private surgical practice from January 1, 1984, through December 31, 2002, were studied. Data analysis was performed from June 12, 2013, to September 6, 2015. A cohort of 7925 patients undergoing gastric bypass surgery and 7925 group-matched, severely obese individuals who did not undergo surgery were identified through driver license records. Matching criteria included year of surgery to year of driver license application, sex, 5-year age groups, and 3 body mass index categories.
Roux-en-Y gastric bypass surgery.
Main Outcomes and Measures
All-cause and cause-specific mortality compared between those undergoing and not undergoing gastric bypass surgery using HRs.
Among the 7925 patients who underwent gastric bypass surgery, the mean (SD) age at surgery was 39.5 (10.5) years, and the mean (SD) presurgical body mass index was 45.3 (7.4). Compared with 7925 matched individuals not undergoing surgery, adjusted all-cause mortality after gastric bypass surgery was significantly lower for patients 35 through 44 years old (HR, 0.54; 95% CI, 0.38-0.77), 45 through 54 years old (HR, 0.43; 95% CI, 0.30-0.62), and 55 through 74 years old (HR, 0.50; 95% CI, 0.31-0.79; P < .003 for all) but was not lower for those younger than 35 years (HR, 1.22; 95% CI, 0.82-1.81; P = .34). The lack of mortality benefit in those undergoing gastric bypass surgery at ages younger than 35 years primarily derived from a significantly higher number of externally caused deaths (HR, 2.53; 95% CI, 1.27-5.07; P = .009), particularly among women (HR, 3.08; 95% CI, 1.4-6.7; P = .005). Patients undergoing gastric bypass surgery had a significantly lower age-related increase in mortality than severely obese individuals not undergoing surgery (P = .001).
Conclusions and Relevance
Gastric bypass surgery was associated with improved long-term survival for all patients undergoing surgery at ages older than 35 years, with externally caused deaths only elevated in younger women. Gastric bypass surgery is protective against mortality even for older patients and also reduces the age-related increase in mortality observed in severely obese individuals not undergoing surgery.
The opportunity to explore the equivocal association between voluntary weight loss and long-term mortality1,2 has been made possible by studying patients who have undergone bariatric surgery, a treatment that results in significant and sustained weight loss.3,4 A number of controlled retrospective studies5- 11 on bariatric surgery mortality, a few retrospective studies12- 14 without control groups, and 1 prospective matched control study15 have been reported. In addition, 2 bariatric surgery meta-analyses16,17 related to long-term mortality have been published. From these data, the general consensus has been that long-term all-cause mortality and, most often, cardiovascular disease– and cancer-specific mortality are favorably affected in patients who have previously undergone bariatric surgery, even in those who have already experienced a cardiovascular event18 or were diagnosed as having diabetes mellitus before bariatric surgery.19 Furthermore, at least 2 retrospective mortality studies6,20 have reported increased rates of externally caused deaths among patients undergoing bariatric surgery compared with individuals not undergoing surgery. External deaths have been defined as unintentional or undetermined injuries or poisonings and intentional injury or poisonings (eg, suicide, homicide, assault, or legal intervention).
From these long-term mortality data, an important clinically related question arises. Does the reported mortality benefit of long-term all-cause and cause-specific deaths extend to all patients undergoing bariatric surgery regardless of the age at which they underwent surgery? Drawing on a previously published long-term mortality study6 of Utah-based patients undergoing Roux-en-Y gastric bypass surgery, this study examined whether the overall reduction in mortality after gastric bypass surgery was found for each of 4 different age groups when patients had their surgery. One might expect that the earlier in life a severely obese person undergoes gastric bypass surgery, the greater the reduction in mortality. The degree to which gastric bypass surgery is effective in reducing mortality when performed on older patients (after 55 years of age, for example) is unknown. We examined age group–specific comparisons of mortality and trends in increasing mortality risk across age groups between severely obese individuals undergoing or not undergoing gastric bypass surgery. We also examined whether our prior suggestion of increased mortality from external causes after gastric bypass surgery differed according to age at surgery and sex.
This investigation was approved by the institutional review board of the University of Utah with a waiver of consent granted. The data set was not deidentified primarily because the record search would not have been possible without the identifiers. After death status was obtained, however, all statistical analyses were performed on deidentified data. The waiver of consent was approved for this approach. Selection of participants for this age-stratified retrospective study from January 1, 1984, through December 31, 2002, has been previously described.6 Data analysis was performed from June 12, 2013, to September 6, 2015. In brief, from a consecutive series of 9949 patients undergoing gastric bypass surgery and 9628 severely obese (ie, those with a body mass index [BMI] ≥33 [calculated as the weight in kilograms divided by height in meters squared]) Utah driver license and identification card applicants, group matching was performed for 7925 patients undergoing surgery and 7925 individuals not undergoing surgery, after exclusion of patients undergoing gastric bypass surgery who could not be adequately matched to driver license records. Matching criteria included sex, BMI, age, and year of surgery matched with year of identification card application. The presurgical BMI of the patients undergoing surgery was used to match with the adjusted, self-reported BMI of the individuals not undergoing surgery (sex-specific regression equations to adjust driver license–reported BMI to measured BMI have been previously described6). All data were linked to the Utah Cancer Registry and to the National Death Index to identify death status and causes of death as categorized by International Classification of Diseases, Ninth Revision (ICD-9), and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10), classifications. All ICD external-caused death codes were combined for analysis. Patients undergoing gastric bypass surgery and individuals not undergoing surgery were systematically divided into 4 specific age categories: younger than 35 years, 35 through 44 years, 45 through 54 years, and 55 through 74 years.
With the use of Cox proportional hazards regression analysis, the risks (hazard ratios [HRs]) of death were estimated by comparing the patients undergoing Roux-en-Y gastric bypass surgery and the individuals not undergoing gastric bypass surgery within each of the 4 age categories. For each age group, the survival time was computed as the difference between the date of death for decedents or January 1, 2003, for survivors and the baseline date, defined as the date of bariatric surgery for the patient group and the date of driver license or identification card application for the group not undergoing surgery.
Baseline age, year of bariatric surgery or the year of driver license or identification card application, and a cubic polynomial of BMI at baseline were used to adjust the HR in the statistical models for each age group. Stratification by sex within the overall Cox proportional hazards regression model was performed to allow different baseline hazards for each sex to meet the proportional hazards assumption because mortality differs by sex. In addition to this sex-stratified model, sex-specific analyses were performed.
A further model was defined to test for different age-related increases in mortality risk between the 2 study groups. Using the individuals younger than 35 years who did not undergo surgery as the reference group, we estimated the mortality risk for each of the remaining 7 age group– and study group–specific categories. Tests for linear age-related increases in mortality risk for each group and the difference in the linear trends were performed using appropriate linear contrasts of the group-specific β coefficients.
Discrete-time survival models were also fit to the data for comparison to the Cox proportional hazards regression models. The results were similar and are not reported. The unadjusted absolute death rates are represented as deaths per 10 000 person-years of follow-up. The P values and the 95% CIs are all 2-sided, and the criterion used for statistical significance was P < .05. Statistical analyses were conducted using SAS statistical software, version 9.3 (SAS Institute Inc).
Table 1 highlights the descriptive data for all patients undergoing gastric bypass surgery, participants not undergoing surgery, and age-specific categories. The mean (SD) age at surgery was 39.5 (10.5) years, and the mean (SD) presurgical BMI was 45.3 (7.4). Although no significant differences were found between the 2 groups for mean follow-up time for all ages combined and for age-specific categories, the mean BMI was consistently 1.2 to 1.5 greater (P < .001) for the group not undergoing surgery compared with the group undergoing surgery. During the total follow-up period of 18 years (mean, 7.2 years), there were 213 and 321 total deaths for the groups undergoing and not undergoing surgery, respectively (eTable in the Supplement). Total deaths and death rates per 10 000 person-years (ie, all-cause mortality) were consistently greater among the participants not undergoing surgery within the age-specific groups, with the exception of the subgroup of those younger than 35 years, in which the group undergoing surgery reported 69 all-cause mortality deaths compared with 50 all-cause mortality deaths for the participants not undergoing surgery. The increased mortality among the groups undergoing and not undergoing surgery in those younger than 35 years was primarily owing to externally caused deaths, where the number of deaths was 35 vs 12 in those undergoing and not undergoing surgery, respectively. Exclusion of externally caused deaths in this youngest group resulted in a nonsignificant (P = .37) reduced mortality for nonexternal causes (34 vs 38 deaths among those undergoing and not undergoing surgery, respectively). The large age-related increases in mortality occurred in those not undergoing surgery, but there was a much lower rate of increase across age for the patients undergoing gastric bypass surgery (eTable in the Supplement).
Compared with individuals not undergoing surgery, adjusted all-cause mortality after gastric bypass surgery was significantly lower for the patients 35 through 44 years old (HR, 0.54; 95% CI, 0.38-0.77), 45 through 54 years old (HR, 0.43; 95% CI, 0.30-0.62), and 55 through 74 years old (HR, 0.50; 95% CI, 0.31-0.79; P < .003 for all) but was not lower for the those younger than 35 years (HR, 1.22; 95% CI, 0.82-1.81; P = .34) (Table 2). Dividing the participants who were 55 through 74 years old into groups of 55 through 64 years old and 65 through 74 years old (only 78 individuals and 12 deaths) revealed similar benefits in both subgroups (HR, 0.52; 95% CI, 0.32-0.86, and HR, 0.34; 95% CI, 0.07-1.52, respectively). The unadjusted death rates per 10 000 person-years in those undergoing surgery vs not undergoing surgery were 93 vs 172 in the group aged 55 through 64 years and 193 vs 447 in the group aged 65 through 74 years.
To investigate the higher but nonsignificant HR of 1.22 (95% CI, 0.82-1.81) in those younger than 35 years, we analyzed only those deaths classified as externally caused. The age category of younger than 35 years for those undergoing vs not undergoing surgery revealed a mortality HR for externally caused death of 2.53 (95% CI, 1.27-5.07; P = .009), whereas patients undergoing surgery in the 3 older age groups did not have significantly elevated mortality from external causes compared with those not undergoing surgery. The combined age group for externally caused deaths was also significantly greater for patients undergoing gastric bypass surgery compared with those not undergoing surgery (HR, 1.58; 95% CI, 1.02-2.45). Repeating the all-cause mortality analyses after coding the externally caused deaths as being censored at their death rather than as an event, we found that the HR for those younger 35 years was reduced from 1.22 to 0.79 (95% CI, 0.47-1.32) (Table 2). We also excluded, rather than censored, the participants who died of external causes from the analysis: the results were almost identical.
With all ages combined, deaths from cardiovascular disease (HR, 0.51; 95% CI, 0.36-0.73) and cancer (HR, 0.40; 95% CI, 0.25-0.64) were significantly lower for those undergoing gastric bypass surgery compared with those not undergoing surgery. The HRs for each group of age at surgery for these 2 end points were also universally decreased but were not significant for many of the age categories because of the small numbers of deaths.
The all-cause mortality analyses for the age groups were repeated separately for men and women. The benefit of surgery was significant or nearly significant for both sexes in each of the 3 oldest age groups (Table 3). In the youngest age group, only women had an elevated all-cause mortality risk (HR, 1.42; 95% CI, 0.90-2.26) seen in the sex-adjusted model, although this finding was not significant (P = .13). The men younger than 35 years had a 21% reduction of risk, but this reduction was not statistically significant (P = .56). Compared with those younger than 35 years not undergoing surgery, the mortality rate from external causes in women who underwent surgery before 35 years of age was 3.08 (95% CI, 1.4-6.7; P = .005), whereas for men after surgery compared with men not undergoing surgery, the mortality rate was only 0.98 (95% CI, 0.2-4.8; P = .98).
In addition to comparing the mortality of patients undergoing gastric bypass surgery with individuals not undergoing surgery within each age group, we selected individuals younger than 35 years who did not undergo surgery as the reference group and compared the mortality of the remaining 7 groups. These mortality risks were used to estimate age-related trends in mortality risk for each study group. Figure 1 shows the expected large significant increases in mortality risk in those not undergoing surgery the older they are when they enter the study (HR, 1.87; 95% CI, 1.33-2.64, for those aged 35-44 years; HR, 4.01; 95% CI, 2.87-5.60, for those 45-54 years; and HR, 7.97; 95% CI, 5.49-11.55, for those 55-74 years; P < .001 for a linear trend of mortality risk for all age groups). After gastric bypass surgery, those younger than 35 years (HR, 1.32; 95% CI, 0.91-1.91), 35 through 44 years (HR, 1.06; 95% CI, 0.72-1.55), 45 through 54 years (HR, 1.68; 95% CI, 1.13-2.51), and 55 through 74 years (HR, 3.85; 95% CI, 2.48-5.99) also had a significant age-related increase in mortality risk compared with those younger than 35 years not undergoing surgery (P < .001 for a linear trend of mortality risk for all age groups). However, the linear age trend of mortality risk for those who underwent surgery was significantly reduced compared with the age trend in those not undergoing surgery (P < .001). Externally caused mortality was only significantly elevated in the patients younger than 35 years undergoing gastric bypass surgery and not in any older patients undergoing gastric bypass surgery or in those not undergoing surgery (Figure 2).
This study further investigated the long-term mortality outcomes of patients undergoing gastric bypass surgery and individuals not undergoing surgery within 4 groups undergoing surgery at different ages: younger than 35 years, 35 through 44 years, 45 through 54 years, and 55 through 74 years. Patients who underwent gastric bypass surgery after 35 years of age had a significant reduction in all-cause mortality compared with those who did not undergo surgery. Before 35 years of age, mortality was not significantly different, and the HR suggested an increased all-cause mortality. However, when deaths from external causes were excluded, the HR suggested a reduced mortality even in those younger than 35 years, although the HR was not significant. Externally caused mortality when gastric bypass surgery occurred before 35 years of age was 2½ times higher than in severely obese individuals not undergoing surgery. This significant increase was seen only in those younger than 35 years and appeared to be confined mostly to women. Furthermore, study results suggest that gastric bypass surgery may be associated with a lower mortality risk for cardiovascular disease– and cancer-related deaths irrespective of age at surgery, although the youngest and oldest age groups had comparatively fewer deaths from these causes and less power to find statistically significant results.
We hypothesized that undergoing gastric bypass surgery at older ages might reduce the effectiveness of the operation owing to long-standing disease that would be harder to reverse after weight loss. However, it appears that severely obese persons aged 55 through 74 years who undergo gastric bypass surgery are likely to have significantly reduced long-term mortality compared with age-matched individuals not undergoing surgery. Although gastric bypass surgery appears to be effective when performed in individuals at any age, these results do not imply that a severely obese person should wait until an older age to undergo surgery. A prior study21 found that the risk of complications early after surgery increases with age. Our results also indicate a significantly slower increase in age-related mortality risk after gastric bypass surgery than in severely obese individuals.
Longer-term mortality studies6- 13,15,19,22,23 of patients undergoing bariatric surgery have generally reported a reduction in all-cause mortality and reduced death rates from certain diseases, such as cardiovascular disease and cancer. An exception is an initial study10 of predominantly male veterans who, after bariatric surgery, did not have a mortality benefit after 6.7 years of follow-up compared with individuals not undergoing surgery. A subsequent analysis11 of an expanded sample of 2500 veterans undergoing bariatric surgery, however, found a significant 5- and 10-year mortality benefit in men after analyzing the first surgical year separately (despite no significant mortality difference at 1 year). From this study and the results of another study,24 it would appear that the survival advantage of bariatric surgery might not be manifest until after the first postsurgical year. The significant long-term mortality benefits for men in the study10 of veterans, along with the results of major long-term observational studies6,7,15,19 in bariatric surgery cohorts predominantly composed of women, are confirmed by this Utah-based age- and sex-stratified report revealing significant longer-term mortality benefit for men and women. This benefit was independent of the year of surgery, which was a nonsignificant predictor of mortality in all models. In addition, our sex-specific analyses revealed a highly significant total mortality reduction of 77% in men who had surgery at 55 through 74 years of age, a greater reduction than in any other age group of men or women.
An important finding of this study is that patients who undergo gastric bypass surgery at a younger age (ie, <35 years) are at greater risk of death from external causes. The sex-specific analyses suggested that this increased mortality risk from external causes was confined to women who underwent surgery before 35 years of age, although men accounted for only 24% of this age group. Even though none of the other age groups had significant increases (or decreases) in mortality owing to deaths from external causes, the number of deaths remains small. The nonsignificant but elevated HR of 1.3 for externally caused deaths for the 2 oldest groups suggests that greater statistical power is required to evaluate the importance of these estimates. There were too few deaths to subdivide by the various causes of death within the external cause definition, such as poisoning or suicide. Therefore, at this time, counseling each patient on expectations about and risks of these external causes of death should probably be given without regard to age at surgery. Finally, with reference to patients who underwent surgery at younger than 35 years, although the HR of 0.79 for patients undergoing surgery compared with those not undergoing surgery was not significant (95% CI, 0.47-1.32; P = .37) for nonexternally caused deaths, there was a favorable trend for reduced mortality. The lack of significance may have been the result of too few deaths related to chronic illnesses (ie, cardiovascular disease– and cancer-related deaths) in this younger-age group. Longer follow-up of these individuals is required for a more definitive assessment of mortality risks in those younger than 35 years.
Although limited, observational studies20,25- 29 suggest that some procedures, such as gastric bypass surgery, may increase the long-term risk of substance use disorders and suicide. However, several studies30- 33 report that individuals presenting for surgery are already at higher risk of psychiatric disease and substance abuse, indicating that deaths after surgery from these causes may not be directly associated with the surgery. At the 6-year follow-up of our prospective Utah study, the Medical Outcomes Study 36-Item Short-Form Health Survey assessment of physical health–related quality of life was improved among the surgical group, but the mental health status was not.4,34 As Livingston35 has pointed out, there is a great deal of overlap between mental health disorders and obesity. In addition to the potential increased risk of substance use disorders, postsurgical dissatisfaction related to presurgical expectations may exist. Clearly, additional research related to the behavioral and psychological aspects before and after bariatric surgery, perhaps especially in the younger population, is warranted.
Limitations related to this study included the lack of clinical information before or after surgery or driver license application, with the exception of presurgical BMI or BMI at the time of driver license or identification card application, age, and sex. As a result, whether the patients undergoing gastric bypass surgery had presurgical health statuses similar to those of the individuals not undergoing surgery or whether greater medical attention was provided for patients undergoing surgery as a result of their entry into the surgical medical system was not known. There were too few cardiovascular disease– or cancer-related deaths to find statistically significant results in 3 of the 4 age groups when the diseases were analyzed separately. However, the protective HRs suggest a likelihood that they would become statistically significant if longer follow-up could be performed. Potential bias related to self-reported height and weight among driver license and identification card applicants may have been present, despite the sex-specific regression correction. Inadequate statistical adjustment for this self-reporting of BMI may be an explanation of the significantly different mean BMIs of 1.2 to 1.5 reported in Table 1 between the 2 study groups. Other potential biases related to the National Death Index matching have been previously described.6 Strengths of this study included the large number of patients undergoing gastric bypass surgery retrospectively followed up for up to 18 years and the opportunity to divide mortality results into separate age categories and by sex.
This study found that the increased mortality from external causes in patients undergoing gastric bypass surgery was limited to the subgroup of women who underwent surgery before 35 years of age. Importantly, this study implies that gastric bypass surgery is protective against mortality even for patients who undergo surgery at an older age. Gastric bypass surgery also reduces the age-related increase in mortality risk compared with severely obese individuals who do not undergo surgery.
Accepted for Publication: November 10, 2015.
Corresponding Author: Lance E. Davidson, PhD, Department of Exercise Sciences, Brigham Young University, 271 Smith Fieldhouse, Provo, UT 84604 (firstname.lastname@example.org).
Published Online: February 10, 2016. doi:10.1001/jamasurg.2015.5501.
Author Contributions: Drs Hunt and Adams had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Davidson, Adams, Jones, Hunt.
Acquisition, analysis, or interpretation of data: Davidson, Adams, Kim, Hashibe, Taylor, Mehta, McKinlay, Simper, Smith, Hunt.
Drafting of the manuscript: Davidson, Adams, Hunt.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Kim, Hashibe, Hunt.
Obtained funding: Adams, Hunt.
Administrative, technical, or material support: Davidson, Adams, Simper, Smith, Hunt.
Study supervision: Davidson, Adams, Jones, McKinlay, Smith, Hunt.
Conflict of Interest Disclosures: Dr Mehta reported being a paid consultant with Gjording Fouser PLLC. No other disclosures were reported.
Funding/Support: This study was supported by grants DK-55006 and P30DK-056336 from the National Institute of Diabetes and Digestive and Kidney Disease and grant M01-RR00064 from the National Center for Research Resources (Drs Davidson, Adams, Mehta, and Hunt) and Ethicon Endo-Surgery (Dr Adams).
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.