Society: AGA

INTRODUCTION: A number of medications have been implicated as primary risk factors for microscopic colitis (MC) and guidelines recommend discontinuation of these medications as part of MC management. However, there are few epidemiologic studies that have attempted to explore a causal relationship between medications and MC.

METHODS: We conducted a cohort study of all residents in Sweden since 2004 over the age of 60 years to examine the association between select medications and the risk of MC. For each medication, new initiators were identified at the time of their first dispensed drug (index date for users) and matched to non-users at the time of dispense of any other medication (index date for non-users) within 6 months of index date for users. Exact 1:1 matching was done according to age, sex, calendar year, healthcare utilization defined by number of prior dispensed medications and encounters for inpatient visits and outpatient subspecialty care, and GI-specific utilizations, including number of endoscopies and GI-related encounters. End of follow up was defined as death, emigration, diagnosis of MC, ascertained through linkage to nationwide histopathology cohort, or December 31^st, 2017, whichever came first. Proportional hazard models were fit to obtain effect estimates. To assess the validity of our models, use of coumadin and beta blockers were defined as negative control exposures while colonoscopy with normal biopsies was defined as a negative control outcome.

RESULTS: We identified 3,278,547 residents of Sweden over the age of 60 years with at least one drug prescription from 2006-2017 which was used to build nine separate cohorts of drug user versus non-users: proton pump inhibitors (PPI, n = 2,487,228), selective serotonin reuptake inhibitors (SSRI, n = 1,232,510), angiotensin converting enzyme inhibitors (ACE-I, n = 2,151,098), angiotensin receptor blockers (n = 1,422,642), Statins (n = 2,374552), NSAIDs (n = 2,581,372), Anti-Parkinson drugs (APD, n = 289,930), coumadin (n = 708,582), and beta blockers (BBs, n = 2,480,988). In an intention-to-treat analysis, PPIs (HR=1.43 [1.32-1.56]), SSRIs (HR= 2.32 [2.07-2.60]), ARBs (HR= 1.26 [1.13-1.42]), and NSAIDs (1.13 [1.06-1.22]) associated with increased risk of MC but equally so with colonoscopy with normal mucosa (Table 1). Our results were similar in as-treated analyses, where participants were censored after discontinuation of medications.

CONCLUSION: In a nationwide cohort study, we observed an increased risk of MC with initiation of previously implicated medications. However, most of these associations were equally observed with receiving a colonoscopy with normal biopsies suggesting that such medication use was associated with undergoing colonoscopy and not MC. These findings suggest that previously observed associations were likely related to surveillance bias.

BACKGROUND: Ulcerative colitis (UC) is treated with infliximab (IFX). However, up to 45% of patients on IFX have sub-therapeutic drug levels, which is associated with loss of treatment response. Despite improved outcomes with proactive therapeutic drug monitoring (TDM), only 37% of gastroenterologists use proactive monitoring for IBD, with the primary barrier being insurance coverage (78%).

STUDY DESIGN: Cost-effectiveness analysis comparing proactive TDM, reactive TDM, and no TDM in UC patients on IFX.

METHODS: We developed a Markov model for patients on IFX therapy that tracked IFX dose, antibody levels, and IFX drug concentrations over a five-year time horizon. IFX concentration and antibody status were measured every six months or at the time of a flare under proactive TDM, while they were only measured during a flare under reactive TDM. Patients were initialized on IFX monotherapy at 5 mg/kg. Management decisions for each strategy are depicted in Figure 1. Patients entering a flare state incurred a reduced utility score and risk of further complications (hospitalization, surgery, readmission, and/or transition to other agents). The cost and utility values for Markov states, risk of complications, and transition probabilities for second-line agents were estimated from the literature. Transition probabilities for the IFX states are uncertain; thus, we used the incremental mixture importance sampling algorithm and specified plausible ranges to calibrate these parameters. We matched model outputs to specific calibration targets: the proportion of patients who fail IFX and those who are post-surgical at the one- and five-year marks.

RESULTS: Proactive TDM dominated other approaches, with an approximate cost of $135,000 and a utility of 3.87 QALYs per patient (Figure 2). In the proactive TDM strategy, the discovery of anti-drug antibodies at baseline initially resulted in higher rates of IFX discontinuation. However, long-term, proactive TDM allowed a greater proportion of patients to remain on IFX—mitigating the risk and associated costs of flares and complications. At the five-year mark, 27% of patients remained on IFX in the proactive TDM group compared to 21% for patients in both the reactive and no TDM groups. In one-way sensitivity analyses, we varied medication and test costs, surgery and hospitalization costs, risk of flare at different drug levels, and baseline distribution of drug levels and presence of antibodies. In each case, proactive TDM was cost-saving compared to the other TDM strategies for plausible parameter ranges.

CONCLUSION: Proactive monitoring in UC patients on IFX was cost-saving compared to reactive monitoring, primarily by identifying patients with high antibody levels and preventing antibodies that occur at lower IFX levels. Further studies to provide more precision to the benefit of proactive monitoring are warranted.

Paul Dupenloup

Appearances