CRC indicates colorectal cancer.
eAppendix. The French National Health Insurance Information System
eTable. Propensity-Matched Hazard Ratios for CRC and Colorectal Benign Polyps in Matched Groups of Bariatric Surgery vs No Bariatric Surgery Patients With Morbid Obesity (Referent)
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Bailly L, Fabre R, Pradier C, Iannelli A. Colorectal Cancer Risk Following Bariatric Surgery in a Nationwide Study of French Individuals With Obesity. JAMA Surg. Published online March 11, 2020. doi:10.1001/jamasurg.2020.0089
What is the risk of colorectal cancer following bariatric surgery?
In a nationwide cohort study in France (2009-2018), the risk of colorectal cancer was estimated using standardized incidence ratios for 1 045 348 inpatients with obesity aged 50 to 75 years. The standardized incidence ratio was 1.0 following bariatric surgery vs 1.34 for individuals with obesity who did not undergo bariatric surgery.
Following bariatric surgery, individuals with obesity appear to share the same risk of colorectal cancer as that of the general population.
Although bariatric surgery is effective against morbid obesity, the association of this surgery with the risk of colorectal cancer remains controversial.
To assess whether bariatric surgery is associated with altered risk of colorectal cancer among individuals with obesity.
Design, Setting, and Participants
This retrospective, population-based, multicenter, cohort study based on French electronic health data included 1 045 348 individuals with obesity, aged 50 to 75 years, and free of colorectal cancer at baseline. All inpatients with obesity having data recorded during a hospital stay between 2009 and 2018 by the French national health insurance information system database were followed up for a mean (SD) of 5.3 (2.1) years for those who did not undergo bariatric surgery and 5.7 (2.2) years for those who underwent bariatric surgery. Two groups of patients comparable in terms of age, sex, body mass index, follow-up, comorbidities, and conditions who did or did not undergo surgery were also obtained by propensity score matching.
Bariatric surgery (n = 74 131), including adjustable gastric banding, sleeve gastrectomy, gastric bypass; or no bariatric surgery (n = 971 217).
Main Outcomes and Measures
Primary outcome was incident colorectal cancer. Standardized incidence ratios were calculated using age-, sex-, and calendar year–matched colorectal cancer incidence among the general French population during the corresponding years. Secondary outcome was incident colorectal benign polyps.
Among a total of 1 045 348 patients, the mean (SD) age was 57.3 (5.5) years for the 74 131 patients in the surgical cohort vs 63.4 (7.0) years for the 971 217 patients in the nonsurgical cohort. The mean (SD) follow-up was 6.2 (2.1) years for patients who underwent adjustable gastric banding, 5.5 (2.1) years for patients who underwent sleeve gastrectomy, and 5.7 (2.2) years for patients who underwent gastric bypass. In total, 13 052 incident colorectal cancers (1.2%) and 63 649 colorectal benign polyps were diagnosed. The rate of colorectal cancer was 0.6% in the bariatric surgery cohort and 1.3% in the cohort without bariatric surgery. In the latter cohort, 9417 cases were expected vs 12 629 observed, a standardized incidence ratio of 1.34 (95% CI, 1.32-1.36). In the bariatric surgery cohort, 428 cases were expected and 423 observed, a standardized incidence ratio of 1.0 (95% CI, 0.90-1.09). Propensity score–matched hazard ratios in comparable operated vs nonoperated groups were 0.68 (95% CI, 0.60-0.77) for colorectal cancer and 0.56 (95% CI, 0.53-0.59) for colorectal benign polyp. There were fewer new diagnoses of colorectal cancer after gastric bypass (123 of 22 343 [0.5%]) and sleeve gastrectomy (185 of 35 328 [0.5%]) than after adjustable gastric banding (115 of 16 460 [0.7%]), and more colorectal benign polyps after adjustable gastric banding (775 of 15 647 [5.0%]) than after gastric bypass (639 of 20 863 [3.1%]) or sleeve gastrectomy (1005 of 32 680 [3.1%]).
Conclusion and Relevance
The results of this nationwide cohort study suggested that following bariatric surgery, patients with obesity share the same risk of colorectal cancer as the general population, whereas for patients with obesity who do not undergo bariatric surgery, the risk is 34% above that of the general population.
Obesity is a major and increasingly global public health issue.1 More than half of the European population is overweight or obese.2 In 2016, the World Health Organization estimated the prevalence of obesity as 21.6% (95% CI, 17.4%-25.9%) in France.3 The causal association between obesity and cancer has been well documented,4 and the obesity epidemic is suspected to accelerate the development of several cancers.5 Bariatric surgery is considered the most effective treatment of morbid obesity, providing sustained long-term weight loss6 and remission or improvement of obesity-related comorbidities.7,8 Bariatric surgery has also been shown to reduce cancer incidence in women with morbid obesity in a randomized clinical trial,9 and to be associated with reduced mortality in a controlled cohort study.10 Recent observational studies have suggested that bariatric surgery may lower the risk of incident obesity-associated cancer.11,12 However, unexpectedly, other studies do not confirm this benefit. On the contrary, the results of some observational studies have suggested an increased risk of colorectal cancer (CRC) following bariatric surgery;13,14 1 study observed this increase only among patients who were 50 years or older,15 and another only after gastric bypass.16 However, there were few cases of CRC in these reports likely because of the younger age of the participants relative to the median age of patients receiving a CRC diagnosis and the time of follow-up required before a tumor develops and becomes clinically manifest.17,18 Because the prevalence of obesity increases after 60 years of age in European countries2 as in other Western countries,19 we decided to follow-up all French patients with a diagnosis of obesity confirmed during a hospitalization and of CRC screening age, which in France is 50 to 75 years of age,20 invited since 2009 to a large national CRC screening program. In 2018, the cumulative number of patients who underwent bariatric surgery in France was estimated to exceed 500 000 individuals21 for whom long-term follow-up is recommended.22 However, because the mean (SD) patient age was 40 (12) years for those who had undergone bariatric surgery in 2013,23 a large database was required to include sufficient patients within the CRC screening age range. The French national health insurance information system database is among the largest medical databases worldwide.24 Thus, the aim of the present study was to use this database to investigate the risk of CRC among patients who were morbidly obese and who underwent bariatric surgery and to compare this risk with that among individuals with morbid obesity who did not undergo this surgery.
We conducted a nationwide retrospective cohort study of patients with a diagnosis of obesity hospitalized in France in any public or private hospital, who underwent bariatric surgery or not, as collected by the French national health insurance information system (Système National Des Données de Santé) database between January 1, 2009, and December 31, 2018. Data were collected from this existing electronic health record database, which has already been used for previous studies on bariatric surgery,25-27 and the collection details are given in the eAppendix in the Supplement. Eligible patients were aged 50 to 75 years, free of CRC at baseline, and had been admitted to hospital at least twice between 2009 and 2018, with a minimum interval of 2 years between the first and the last hospitalizations. Patients were then placed in 1 of 2 groups: the operated cohort who underwent bariatric surgery (BS) and the nonoperated cohort who had no bariatric surgery (NBS). Patients who had a history of CRC at baseline or patients with obesity who were diagnosed as having CRC within 2 years of their first hospital stay with a diagnosis of obesity were excluded to avoid detection bias (eg, hospitalization for CRC with a diagnosis of obesity during the same episode, hospitalization for obesity with standard investigation revealing CRC). This investigation is registered at the Institut National des Données de Santé (National Institute for Health-Related Data)28 and declared at the Commission Nationale Informatique et Liberté (National Data Protection Committee). This study was approved by the institutional review board of the University of Côte d’Azur with a waiver for the requirement of obtaining informed patient consent granted because all retrieved data are deidentified. No one received compensation or was offered any incentive for participating in this study.
The most commonly performed bariatric procedures in France include adjustable gastric banding (AGB), sleeve gastrectomy (SG), and gastric bypass (GB). Each hospital record contains a description of the treatment procedures, based on the Classification Commune des Actes Médicaux (11th edition), which is a national standardized classification of medical procedures. The procedure for extracting the data has already been described elsewhere29 and is detailed in the eAppendix in the Supplement.
Descriptive statistics were used for demographic and clinical characteristics. To measure the risk of developing CRC in each cohort of operated and nonoperated patients with obesity, we calculated the number of expected cases in each cohort, by computing the observed person-time by age (grouped in 5-year age categories), sex, and calendar year–specific cancer incidence rates among the general French population. Historical incidence data are only available in France for the years 2009 to 2013.30-32 For those 5 years, we used mean incidence rates observed in 4 representative cancer registries, located across the cardinal points of the country (Calvados, Hérault, Bas-Rhin, and Isère). For the 2014 to 2018 period no observed incidence rates were available; thus, we applied the latest national estimated CRC incidence rates by age and sex, available in 2019.33 We calculated a standardized incidence ratio (SIR) to compare the CRC risk in the study population with the risk among the French general population, ie, the number of observed cases divided by the number of expected cases, and 95% CIs, following the Breslow and Day statistical methods.34
To compare the risk of developing CRC between patients with obesity who underwent an operation and those who did not, a matching procedure was conducted based on demographic and clinical characteristics. Two comparable groups were obtained by eliminating the selection bias and by matching (1:1) patients with morbid obesity from the 2 cohorts. A logistic regression model was performed using 18 clinical and demographic baseline characteristics to calculate a propensity score for each individual. The selected variables included age, sex, body mass index, time of follow-up, alcohol use, smoking status, cardiometabolic comorbidities (hypertension, myocardial infarction, diabetes, dyslipidemia, nonalcoholic steatohepatitis, and obstructive sleep apnea syndrome) and other digestive diseases and conditions (detailed in the eAppendix in the Supplement). Standardized differences were calculated as suggested by Austin35 to evaluate the comparability of propensity-matched groups. A Cox proportional hazards regression model was built to estimate hazard ratios and 95% CIs matched on propensity score in a 1:1 manner, as described by Parsons.36
The same procedure was applied to estimate the risk of colorectal benign polyps in the BS compared with the NBS cohorts. Individuals were excluded if a hospital discharge diagnosis of colorectal benign polyps was recorded within 2 years after the first hospitalization associated with a diagnosis of obesity or with bariatric surgery. To compare bariatric surgical procedures, a multivariate-adjusted Cox proportional hazards regression model was constructed to estimate adjusted hazard ratios and 95% CIs for the outcomes, controlling for age, sex, total number of surgical procedures, body mass index, and comorbidities (detailed in the eAppendix in the Supplement). Stratification by time of follow-up was performed using periods of 5 or fewer years and more than 5 years. All P values and 95% CIs are 2-sided, and P < .05 was considered statistically significant. All statistical analyses were conducted using SAS, version 5.1 (SAS Institute Inc).
From 2009 to 2018, 2 044 622 patients aged 50 to 75 were hospitalized in France with a hospital discharge diagnosis of obesity. Among this population, 107 774 patients underwent BS. After excluding patients with fewer than 2 years’ follow-up, a history of CRC, or CRC diagnosed during the first 2 years of follow-up, 1 045 348 patients met the inclusion criteria. Among them, 74 131 patients underwent BS (the BS cohort); and 971 217 did not undergo BS (the NBS cohort) (Figure). Between 2009 and 2018, the rate of new cases of CRC in the NBS cohort was 1.3% (12 629 of 971 217 patients) and was 0.6% (423 of 74 131 patients) in the BS cohort (P < .001). Of patients undergoing BS, 0.7% (115 of 16 460) who underwent AGB, 0.5% who underwent GB (123 of 22 343), and 0.5% who underwent SG (185 of 35 328) received a new diagnosis of CRC (P = .04).
The mean (SD) age of the BS cohort was 57.3 (5.5) years vs 63.4 (7.0) years of the NBS cohort (P < .001) (Table 1). The mean (SD) observation time (follow-up) was 5.7 (2.2) years in the BS cohort (422 547 person-years) and was 5.3 (2.1) years in the NBS cohort (5 147 450 person-years). Characteristics of the study population by surgical procedure are given in Table 2. The mean (SD) follow-up was 6.2 (2.1) years for patients who underwent AGB, 5.5 (2.1) years for patients who underwent SG, and 5.7 (2.2) years for patients who underwent GB.
In total, 13 052 incident colorectal cancers (1.2%) were diagnosed. The numbers of observed and expected CRC cases according to the French general population incidence rates for each cohort between 2009 and 2018 are given in Table 3. Overall, in the BS cohort, 423 CRC cases were observed and 428 cases were expected, leading to a standardized incidence ratio (SIR) of 1.0 (95% CI, 0.90-1.09), whereas in the NBS cohort 12 629 cases were observed and 9417 cases were expected, providing an SIR of 1.34 (95% CI, 1.32-1.36). In the NBS cohort, the SIR was higher among men (1.36; 95% CI, 1.33-1.39) than among women (1.22; 95% CI, 1.19-1.26), and the SIR reached 1.80 (95% CI, 1.73-1.87) for individuals aged 50 to 59 years vs 1.12 (95% CI, 1.09-1.14) for those aged 66 to 75 years. In the NBS cohort, a lower SIR was observed when follow-up was 5 or fewer years compared with more than 5 years (1.13 [95% CI, 1.10-1.16] vs 1.60 [95% CI, 1.56-1.63]), whereas the SIR was stable in the BS cohort for both follow-up periods (SIR, 1.0 [95% CI, 0.88-1.18] for ≤5 years vs SIR, 1.0 [95% CI, 0.85-1.10] for >5 years). In the BS cohort, the SIR was also higher among men than women (SIR, 1.05 [95% CI, 0.88-1.25] vs SIR, 0.96 [95% CI, 0.85-1.08]), patients aged 60 to 65 years (SIR, 1.11; 95% CI, 0.92-1.32), and following AGB (SIR, 1.12; 95% CI, 0.92-1.34) than after GB (SIR, 0.98; 95% CI, 0.82-1.17) or SG (SIR, 0.92; 95% CI, 0.80-1.07).
A propensity score model was applied to the study population of the BS and NBS cohorts. Patient characteristics of a 1:1 propensity-matched sample and standardized differences are given in Table 4. There were 71 300 patients with morbid obesity in each arm, with the mean age, sex ratio, body mass index, and clinical characteristics comparable in the operated and in the nonoperated cohorts. Regarding the individuals who did not undergo an operation, the 1:1 propensity-matched hazard ratio (HR) for a diagnosis of CRC in all patients was 0.68 (95% CI, 0.60-0.77), and was lower among women who underwent a surgical procedure than among men who underwent a surgical procedure (HR, 0.66 [95% CI, 0.57-0.77] vs HR, 0.72 [95% CI, 0.57-0.89]) (eTable in the Supplement).
We applied the same procedure as described above for colorectal benign polyps with participants free of polyps at baseline and during the 2 years following the initial diagnosis of obesity or following BS, leading to a total of 975 519 patients with obesity. Of them, 69 190 patients underwent BS and were considered the BS cohort, and 906 329 did not undergo BS and were considered the NBS cohort. Between 2009 and 2018, the rate of colorectal benign polyps in the NBS cohort was 6.8% (61 230 of 906 329 patients) vs 3.5% (2419 of 69 190 patients) in the BS cohort (P < .001). The propensity score included 66 516 participants with morbid obesity in the operated and nonoperated groups. The HR for a diagnosis of colorectal benign polyps among all patients was 0.56 (95% CI, 0.53-0.59) and was higher among women than among men (HR, 0.59 [95% CI, 0.56-0.63] vs 0.47 [95% CI, 0.43-0.52]). In the BS cohort, there were significantly (P < .001) more colorectal benign polyps after AGB (5.0% [775 of 15 647]) than after GB (3.1% [639 of 20 863]) or SG (3.1% [1005 of 32 680]). The adjusted HR for the diagnosis of colorectal polyps was 0.72 (95% CI, 0.64-0.82) after GB and 0.77 (95% CI, 0.69-0.86) after SG, compared with patients who underwent AGB. Regarding duration of follow-up, the same results were obtained for patients who underwent an operation with 5 or fewer years of follow-up (HR, 0.75 [95% CI, 0.61-0.93] vs GB; and HR, 0.77 [95% CI, 0.64-0.92] vs SG) and patients who underwent an operation with longer follow-up (HR, 0.71 [95% CI, 0.61-0.82] vs GB; and HR, 0.77 [95% CI, 0.68-0.89] vs SG).
The results of this study suggested that individuals with morbid obesity undergoing BS carry the same risk of CRC as that of the general population aged 50 to 75 years. However, in our large national cohort of 1 045 348 patients, compared with the general population, individuals of the same age with obesity who did not undergo BS had an increased risk (34%) of developing CRC. For patients aged 50 to 59 years, the NBS cohort had an increased risk (80%) of developing CRC, but our study results suggested that BS may decrease the risk of CRC compared with comparable individuals with morbid obesity not undergoing BS. Finally, our study results were not consistent with the increased risk of CRC observed after BS in studies conducted in Sweden13,14 or after GB performed in England.16 Method issues that strengthened our results included the population-based design, with SIRs calculated on matched CRC incidence among the French general population, which avoids potential selection bias associated with nonrandomized comparison of operated and nonoperated patients with obesity. Another major advantage of our study was the long-term follow-up of a large national cohort who underwent these surgical procedures, providing a much larger sample size than previous studies (eg, 423 patients with CRC out of 74 131 participants in the BS cohort). Aside from previously discussed potential method biases,17,18 in the studies conducted by Östlund and colleagues14 and Mackenzie et al,16 the total number of CRC cases was 35, whereas Derogar et al13 reported 70 cases among 15 095 in the obesity surgery cohort.
A propensity score was also calculated to compare 74 131 individuals in the BS cohort with 971 217 matched participants in the NBS cohort, yielding two 1:1 matched groups of patients with morbid obesity for CRC (N = 71 300) and for colorectal benign polyps (N = 66 516) who had similar demographic and clinical characteristics. This comparison of matched groups showed that participants in the BS group had 32% lower risk of developing CRC and 44% lower risk of developing colorectal benign polyps than those in the NBS group. Moreover, the results of the present study also suggested that there were significantly fewer new cases of CRC and colorectal benign polyps among patients undergoing GB or SG than among those undergoing AGB. Several hypotheses may account for this difference, such as the lower efficacy in achieving long-term weight loss and remission of comorbidities following AGB than following GB or SG.7,37,38 Follow-up was also longer for patients who underwent AGB compared with those who underwent GB or SG (ie, 6.2 years vs 5.7 and 5.5 years, respectively), but adjustment and stratification in the 5-year follow-up led to the same results.
Our results are consistent with previous studies on obesity-associated cancer incidence following BS: the Utah Obesity Study10 found 38% lower risk for both sexes, and the Swedish Obese Subjects study9 found 33% lower risk for women than for men. More recently, Schauer et al11 have shown lower risk of cancer in women than in men. We also observed lower risk of CRC following BS, with 44% decreased risk for women and 38% for men. MacKenzie et al16 highlighted in their study that a 2-fold increased risk of CRC following GB was in line with the findings of increased proliferation of the rectal mucosa39 and changes in the gut microbiome40 after GB surgery. Our large sample population study did not support the hypothesis that GB might increase the risk of CRC. In general, patients aged 50 to 75 years who undergo an operation not only lose a substantial amount of weight but also increase their ability to move41 and to practice regular physical activity.42,43 Moreover, potential mechanisms linking the reduced cancer risk observed after weight loss are considered to be mainly endocrine-based, and hormonal changes after bariatric surgery are believed to lower CRC risk, for example, in tumors expressing the estrogen receptor β.44 Otherwise, prolonged hyperinsulinemia attributable to increased adiposity leads to reduced production of insulin-like growth factor binding protein (IGFBP)-1 and IGFBP-2, which in turn results in increased levels of free IGF-I.45,46 This may play a role in CRC risk because the IGF-I receptor is overexpressed in CRC. Finally, in a large sample such as the present one, all positive and preventive outcomes associated with physical activity,47-49 hormonal change,50,51 and weight loss appeared to counterbalance potential adverse GB outcomes, as has already been suggested for patients aged 55 to 74 years52 or in patients older than 65 years.53
Our study has several limitations. This is a retrospective, observational study based on electronic administrative data. Indeed, it may be speculated that patients in the NBS cohort avoided surgery because of an increased operative risk linked to a poorer state of health, resulting in a major selection bias. Patients with obesity undergoing BS were younger than those who did not (57 vs 63 years of age; P < .001) and were mostly women, suggesting that younger women were less likely to develop CRC. However, by calculating the SIR for each cohort separately, our results for the NBS are coherent and consistent with the elevated SIR observed in patients with obesity.13,54 Moreover, our results were also observed in 2 comparable operated and nonoperated groups of patients with morbid obesity, with individuals matched 1:1 for age, sex, body mass index, follow-up duration, and clinical characteristics. The analysis of this propensity-matched population also found 32% risk reduction of developing CRC in the BS group, consistent with the result obtained using the SIR in the BS cohort. Finally, the risk associated with developing CRC or colorectal benign polyps among patients who underwent AGB was higher than that among patients who underwent GB or SG. For the latter 2 operations, a lower risk observed for both outcomes is in favor of a proportional effect associated with BS because reduced efficacy in achieving sustained weight loss is suspected for AGB compared with GB and SG.7,37,38
To conclude, the results of this nationwide, long-term observational study of individuals with obesity aged 50 to 75 years suggested that BS is associated with decreased risk of CRC compared with no operation, exposing the individuals who undergo BS to the same CRC risk as that for the general population. Clinicians may consider these data when discussing BS with patients with obesity.
Accepted for Publication: December 26, 2019.
Corresponding Author: Laurent Bailly, MD, PhD, Département de Santé Publique, Centre Hospitalier Universitaire de Nice, Hôpital Archet 1, Niveau1 151 Route Saint Antoine de Ginestière CS 23079 06202, Nice Cedex 3, France (email@example.com).
Published Online: March 11, 2020. doi:10.1001/jamasurg.2020.0089
Author Contributions: Dr Bailly had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Bailly, Pradier, Iannelli.
Acquisition, analysis, or interpretation of data: Bailly, Fabre, Pradier.
Drafting of the manuscript: Bailly, Iannelli.
Critical revision of the manuscript for important intellectual content: Fabre, Pradier, Iannelli.
Statistical analysis: Bailly, Fabre, Pradier.
Administrative, technical, or material support: Pradier.
Supervision: Pradier, Iannelli.
Conflict of Interest Disclosures: None reported.
Additional Contributions: Brigitte Dunais, MD, formerly employed by the Nice University Public Health Department, made substantial contributions to the writing and editing of this article. The Agence Technique de l’Informatique Hospitalière provided equipment used for data collection. No one received any compensation for the stated contribution.
Create a personal account or sign in to: