Association of Bariatric Surgery With Cancer Risk and Mortality in Adults With Obesity

Abstract

Importance: Obesity increases the incidence and mortality from some types of cancer, but it remains uncertain whether intentional weight loss can decrease this risk.

Objective: To investigate whether bariatric surgery is associated with lower cancer risk and mortality in patients with obesity.

Design, setting, and participants: In the SPLENDID (Surgical Procedures and Long-term Effectiveness in Neoplastic Disease Incidence and Death) matched cohort study, adult patients with a body mass index of 35 or greater who underwent bariatric surgery at a US health system between 2004 and 2017 were included. Patients who underwent bariatric surgery were matched 1:5 to patients who did not undergo surgery for their obesity, resulting in a total of 30 318 patients. Follow-up ended in February 2021.

Exposures: Bariatric surgery (n = 5053), including Roux-en-Y gastric bypass and sleeve gastrectomy, vs nonsurgical care (n = 25 265).

Main outcomes and measures: Multivariable Cox regression analysis estimated time to incident obesity-associated cancer (a composite of 13 cancer types as the primary end point) and cancer-related mortality.

Results: The study included 30 318 patients (median age, 46 years; median body mass index, 45; 77% female; and 73% White) with a median follow-up of 6.1 years (IQR, 3.8-8.9 years). The mean between-group difference in body weight at 10 years was 24.8 kg (95% CI, 24.6-25.1 kg) or a 19.2% (95% CI, 19.1%-19.4%) greater weight loss in the bariatric surgery group. During follow-up, 96 patients in the bariatric surgery group and 780 patients in the nonsurgical control group had an incident obesity-associated cancer (incidence rate of 3.0 events vs 4.6 events, respectively, per 1000 person-years). The cumulative incidence of the primary end point at 10 years was 2.9% (95% CI, 2.2%-3.6%) in the bariatric surgery group and 4.9% (95% CI, 4.5%-5.3%) in the nonsurgical control group (absolute risk difference, 2.0% [95% CI, 1.2%-2.7%]; adjusted hazard ratio, 0.68 [95% CI, 0.53-0.87], P = .002). Cancer-related mortality occurred in 21 patients in the bariatric surgery group and 205 patients in the nonsurgical control group (incidence rate of 0.6 events vs 1.2 events, respectively, per 1000 person-years). The cumulative incidence of cancer-related mortality at 10 years was 0.8% (95% CI, 0.4%-1.2%) in the bariatric surgery group and 1.4% (95% CI, 1.1%-1.6%) in the nonsurgical control group (absolute risk difference, 0.6% [95% CI, 0.1%-1.0%]; adjusted hazard ratio, 0.52 [95% CI, 0.31-0.88], P = .01).

Conclusions and relevance: Among adults with obesity, bariatric surgery compared with no surgery was associated with a significantly lower incidence of obesity-associated cancer and cancer-related mortality.

Figure 1.. Identification of Eligible Patients and Development of Cohorts in the SPLENDID Study

To create a comparable control group, dates for bariatric surgery were randomly assigned to a pool of 666 451 patients with a body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) of 35 or greater. Patients who had not undergone bariatric surgery were then removed from the pool if they failed to meet inclusion criteria on the assigned date, at which point the patients could be seen as potentially eligible for bariatric surgery. Using this algorithm, 128 119 comparable patients who had not undergone surgery were identified to be considered for matching. With propensity matching of each patient who underwent bariatric surgery to 5 patients who had not undergone surgery (nonsurgical control), 5053 patients in the bariatric surgery group and 25 265 matched patients in the nonsurgical control group were enrolled in the study. The diagnosis and procedure codes appear in eTables 1-2 in the Supplement. SPLENDID indicates Surgical Procedures and Long-term Effectiveness in Neoplastic Disease Incidence and Death. ^aA patient may have met more than 1 exclusion criterion; therefore, the total excluded exceeds the sum of the individual reasons for exclusion.

Figure 2.. 10-Year Cumulative Incidence Estimates (Kaplan-Meier) for the Primary and Secondary End Points

A, The primary composite end point was the first occurrence of 1 of the 13 types of obesity-associated cancer. The median observation time was 5.9 years (IQR, 3.4-8.9 years) for patients in the bariatric surgery group and was 6.1 years (IQR, 3.9-9.2 years) for patients in the nonsurgical control group. B, The risk for the primary end point was assessed separately after RYGB and SG. The median observation time was 6.8 years (IQR, 3.4-10.0 years) for RYGB and was 4.6 years (IQR, 3.0-6.8 years) for SG. C, The occurrence of all cancer types was a secondary end point. The median observation time was 5.8 years (IQR, 3.4-8.8 years) for patients in the bariatric surgery group and was 6.1 years (IQR, 3.9-8.9 years) for patients in the nonsurgical control group. D, Cancer-related mortality was a secondary end point. The median observation time was 6.0 years (IQR, 3.4-9.0 years) for patients in the bariatric surgery group and was 6.3 years (IQR, 4.0-9.1 years) for patients in the nonsurgical control group. HR indicates hazard ratio.

Figure 3.. Association of Bariatric Surgery vs Nonsurgical Controls for the Primary End Point in Key Subgroups in the Fully Adjusted Cox Models

Testing for interaction revealed no heterogeneity in the association of bariatric surgery with the primary end point in a wide range of patients including both men and women, young and old patients, Black and White patients, and patients with or without diabetes. The adjusted hazard ratios (HRs) were obtained after individually removing the original variable from the fully adjusted Cox model and replacing it with the dichotomous subgroup variable as well as its interaction with the treatment variable. For example, the continuous age covariate was replaced by the dichotomous version and its interaction with the treatment. Age and body mass index were categorized based on their median values. ^aCalculated as weight in kilograms divided by height in meters squared. ^bAmerican Indian or Alaska Native, Asian, multiracial, Native Hawaiian or Other Pacific Islander, and not reported.

Figure 4.. Weight Loss and Cumulative Incidence of Primary End Point Stratified by Maximum Weight Loss Quartile

A, The data were smoothed and are mean trends for the percentage change in body weight from baseline in patients in the bariatric surgery group and the nonsurgical control group during follow-up. The shaded areas indicate 95% CIs. The mean between-group difference at 10 years from baseline was estimated from a flexible regression model with a 4-knot restricted cubic spline for the time × treatment interaction. The median observation time was 5.9 years (IQR, 3.4-9.0 years) for patients in the bariatric surgery group and was 6.3 years (IQR, 4.0-9.2 years) for patients in the nonsurgical control group. B, The data are Kaplan-Meier estimates for incidence of obesity-associated cancer types by the quartile of maximum (the largest) weight loss in the bariatric surgery group (P < .001 from log-rank test). The findings suggest that weight loss in the bariatric surgery group was associated with lower risk of incident cancer cases in a dose-dependent response.