Risk of De-novo Inflammatory Bowel Disease Among Obese Patients Treated With Bariatric Surgery or Weight Loss Medications

Gursimran S. Kochhar; Aakash Desai; Aslam Syed; Abhinav Grover; Sandra El Hachem; Heitham Abdul-Baki; Preethi Chintamaneni; Elie Aoun; Sowjanya Kanna; Dalbir S. Sandhu; Siddharth Singh; Bo Shen; Edward V. Loftus Jr; Parambir S. Dulai


Aliment Pharmacol Ther. 2020;51(11):1067-1075. 

In This Article


Study Design and Database

This was a retrospective cohort study utilising a prospectively maintained EMR dataset (IBM Corporation), Explorys. Explorys is a multi-institution de-identified electronic health record database with an estimated 60 million covered lives spanning from 1999 to 2018. De-identified data from virtual healthcare systems (EPIC, Amalga, Eclipsys, etc) are standardised and normalised by the Explorys web-based platform. Specific diagnoses, procedures and medications are mapped into the systematised nomenclature of medicine clinical terms (SNOMED-CT). Each unique search allows users to further specify associated medications, comorbidities, demographics and vital signs. All data are live, and updated once every 24 hours.[14] Each participating organisation has access to a password-protected web-based application (https://popex.explorys.com). Explorys is a Health Insurance Portability and Accountability Act (HIPAA) and Health Information Technology for Economic and Clinical Health Act (HITECH) compliant platform, and institutional review board approval is not required. This study is reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational cohort studies.[15] (Supplementary STROBE checklist).

Study Aims

Our primary aim was to determine the prevalence of de-novo CD and UC in obese patients after bariatric surgery and obese patients on weight loss medications (intervention groups); and compare this to the prevalence of de-novo IBD among obese patients who were not exposed to bariatric surgery or weight loss medications (control group). Secondary outcomes were to compare the prevalence of de-novo IBD after exposure to bariatric surgery and weight loss medications and to individual types of surgeries or weight loss medication exposures, and to explore the associations between changes in BMI and risk of de-novo IBD.

Patient Selection and Exposure

Study groups were created using the Explorys search tool and SNOMED-CT procedure codes, and classified as:

  1. Bariatric surgery cohort : Obese (BMI > 30) patients exposed to bariatric surgery (Roux-en-Y gastrojejunostomy, laparoscopic sleeve gastrectomy or gastric banding). Vertical banded gastroplasty (stomach stapling) has been falling out of favour compared with other forms of bariatric surgery and this was reflected in the small sample size seen in the database; hence, we excluded this cohort from the study.

  2. Weight loss medication cohort : Obese (BMI > 30) patients exposed to FDA-approved weight loss medications (orlistat, phentermine/topiramate, lorcaserin, bupropion/naltrexone and liraglutide).

  3. Obese control cohort : Obese (BMI > 30) patients without exposure to bariatric surgery or FDA-approved weight loss medications. These patients had to have a BMI > 30 for 2 continuous years in order to be included in this control cohort to avoid confounding by weight loss through dietary or exercise-based interventions.

Exposure to bariatric surgery and weight loss medications had to have occurred after patient enrolment in Explorys to ensure that patients with historic treatment or exposure were not included.

Outcome Ascertainment

Our outcome of interest was the development of de novo IBD, defined as new cases of IBD occurring during enrolment in Explorys. The diagnosis of IBD was made through IBD-related diagnostic codes in combination with prescriptions for IBD-associated medications (biologic agents, immunomodulators and/or corticosteroids) in the electronic health record.[16] We specifically excluded patients who were only exposed to 5-ASA therapy to minimise the risk for capturing other colitis conditions (diverticular-associated or microscopic colitis) not related to IBD. We excluded patients with existing diagnoses of IBD using a temporal relationship methodology.

Temporal Relationship

Explorys has the unique ability of defining index events and implementation of temporal relationships.[17,18] Using this temporal relationship, we excluded all patients with any diagnosis of IBD, CD, UC or indeterminate colitis prior to exposure to the interventions (weight loss surgery or medications), or at the time of enrolment in Explorys (for the obese control cohort). This allows for capturing of true de-novo IBD cases occurring after exposure to these interventions as opposed to existing prevalent IBD cases. To avoid duplication of patients who leave and enter the healthcare system, Explorys utilises a patient matching engine which ensures that a patient is captured only once throughout their life within the health platform.[19]

This temporal relationship methodology was applied to our primary and secondary analysis as follows:

  • For weight loss surgery, we looked at the prevalence of de-novo IBD (UC and CD) at least 30 days after bariatric surgery to avoid overlap of diagnoses and avoid potentially undiagnosed IBD prior to surgery.

  • For weight loss medications, we used the temporal relationship methodology to ensure patients were on these medications for at least 6 months before examining the prevalence of de-novo IBD.

  • Control population: Our control population cohort included obese patients (BMI > 30) who did not receive bariatric surgery or exposure to any of the FDA-approved weight loss medications. Using the temporal relationship in Explorys we made sure this population had at least 2 years of having BMI > 30.

Statistical Analysis

Explorys allows for searches within selected cohorts as defined above. This includes searches on demographics, medications, vital signs (including BMI) and procedures. Specific searches included evaluation of age groups (18–65 years vs >65 years), gender, race and co-morbid conditions (diabetes mellitus and cardiovascular diseases). Descriptive characterisation of baseline demographics was made, and prevalence rates were calculated as number of de-novo IBD cases divided by total population at risk within each cohort and subgroup. Chi-square test of independence was used to analyse if there was a difference between the type of bariatric surgery or weight loss medication and the risk of CD or UC. Bonferroni correction was used for statistically significant results (P = 0.05/10). Rates were reported standardised to 1000 persons and compared between groups using prevalence odds ratios (OR) with 95% Confidence Intervals (CI).[17,19,20]

Explorys does not provide de-identified aggregated data to calculate total person-years and allow for calculation of incidence rates or standardised age-adjusted incidence rates. Event rates within Explorys are also rounded to the nearest 10 to minimise risk re-identification for rare diseases.