The current state of comparative effectiveness research in inflammatory bowel disease

Abstract

As the number of available treatments, and different mechanisms of actions, increases for inflammatory bowel disease (IBD), clinicians are faced with more complex decisions about treatment positioning and/or sequencing. Comparative effectiveness research (CER) allows for the comparison of benefits and harms of treatments and intervention strategies. In 2012, several areas were identified as IBD CER priorities including effectiveness and safety of currently available biologics. The most traditional form of CER is head-to-head comparisons; however, these trials are difficult to perform, and therefore, few exist. Alternatively, CER can be performed through indirect comparisons via network meta-analyses of randomized controlled trials or through direct comparisons using non-randomized observational data which provides a real-world view of comparisons. Importantly, this type of observational data research can include patients who are not traditionally included in clinical trial populations due to disease severity, previous exposure to medications, or comorbidities. It is important to note that indirect comparisons via network meta-analyses are dependent on the strength of data included and therefore may change as trials evolve over time, and observational CER research is not without methodological challenges, including the potential for bias, confounding and missing data, but these can be overcome to a certain degree with evolving methodologies. Network analyses of clinical trials looking at first-line and second-line biologic treatment in patients with moderate to severe ulcerative colitis have demonstrated that depending on the trials included the results of the analysis and comparative estimates may change. Importantly, the indirect estimates from these network analyses are not comparable to those obtained from a recently completed head-to-head trial in this same population, suggesting that these network meta-analyses are not able to overcome differences in trials to obtain true comparative estimates. Claims-based analyses are one form of observational research which offers the advantage of having large patient numbers to overcome population heterogeneity. Using administrative claims data and propensity score matching, Singh et al. found that in anti-tumor necrosis factor (TNF) naïve patients with ulcerative colitis, patients treated with infliximab may have lower corticosteroid use compared to patients treated with adalimumab. However, the authors noted that depending on the statistical analysis used subtle differences in results were seen, again suggesting that this approach does not fully address sources of bias. More recently, CER using observational cohort studies with detailed patient phenotypes has emerged as an opportunity to provide routine practice comparative estimates to guide treatment positioning. A propensity score matched cohort study of vedolizumab versus anti-TNF therapy in ulcerative colitis observed a point estimate for comparative effectiveness that was nearly identical to that achieved by a head-to-head phase 3 double-blind randomized control trial of vedolizumab versus adalimumab. The detailed phenotyping and matching achieved in this study allowed for a more true-to-life comparative estimate, and this may be the most ideal approach going forward as newer biologics are introduced into practice. Regardless of the methodology, limitations remain in CER and there lacks an ability to personalize these population level comparative estimates. An alternative, emerging, approach to CER is to use clinical prediction tools to guide comparative evaluations. A clinical decision support tool (CDST) to predict the outcomes of patients with ulcerative colitis treated with vedolizumab has been developed using data from the GEMINI 1 study and validated using data from the VICTORY consortium. This tool can identify ulcerative colitis patients in whom vedolizumab will be effective but not anti-TNF therapy. In Crohn’s disease, a similar tool has been developed using data from the GEMINI 2 study and also validated using data from the VICTORY consortium. This tool has been shown to similarly identify patients most likely to respond to vedolizumab. Other tools have been developed for ustekinumab in Crohn’s disease, and recently, it has been demonstrated that the ustekinumab and vedolizumab tools for Crohn’s disease are indeed drug specific and allow for discrimination of an individual patient for choosing between these agents. The capacity to include patient specific data (smoking status, C-reactive protein results, etc.), allows modeling to be tailored to the patient and therefore brings forward CER into a new era of personalization. The potential to develop these models to include genetic and serum biomarkers, or to provide response data for multiple therapies, will continue to strengthen these tools. Further, these tools have the potential to educate patients; for example, patients are able to visualize their predicted response to treatment thereby increasing their capacity to make informed treatment decisions. While head-to-head studies continue to be the primary research goal, CER-based research provides clinicians with a greater understanding of the available research. Clinical decision support tools assist the clinician in delivering personalized medicine; however, clinicians must make a treatment decision based on the patient’s presentation and the totality of available evidence.