The two most commonly performed types of economic studies in IBD are (1) cost of illness assessments that quantify the overall cost of the diseases and identify specific cost drivers and (2) cost–utility comparisons that evaluate the relative cost-effectiveness of treatment alternatives.

A classic study performed by Hay and Hay in the 1990s, estimated the cost of IBD care in a cohort of patients registered with an American health maintenance organization [4, 16]. Cost modeling, based on a sample of 100 patients with CD, suggested the annual cost of disease was $6561 per person, with 70 % attributed to hospitalization and surgery, 11 % to medications, 18 % to testing and office visits, and 5.5 % to extra-intestinal manifestations. In UC, the annual cost of care was $1488. Importantly, 80 % of disease-related costs were generated by only 20 % of patients. Several valuable teachings came from this study. First, the overall economic burden of CD was greater than UC, Second, inpatient costs were paramount which suggested that interventions that kept patients well and out of hospital had the potential to be cost-effective even if drug acquisition costs were high. Finally, concentrating on delivering highly effective therapies to the sickest patients had the best chance of reducing overall costs. These data suggested that medical therapy that reduces surgery and hospitalization would have the highest likelihood of reducing the overall cost of IBD care.

In contrast to these historical data, modern studies have shown a dramatic shift away from the predominance of inpatient care costs to outpatient prescription drug costs as being responsible for the financial burden of the two diseases. Representative Canadian data are shown in Fig. 26.2.

The high cost of biologics drugs is directly responsible for this shift. In accordance with this change in treatment patterns, a number of studies indicate that surgery rates now appear to be falling, particularly in CD, likely as a consequence of the introduction of more effective biologic therapy for outpatient management [17–19]. Costa et al. [17] performed a meta-analysis that included 27 studies that evaluated the efficacy of infliximab-based treatment regimens for reducing surgery and hospitalization in both UC and CD. These investigators documented a substantial reduction in the rates risk of hospitalization overall [Odds Ratio (OR) 0.51; 95 % confidence interval (CI) 0.40, 9.65], that was not different between the two diseases. Surgical rates were similarly reduced [the effect was greater in CD (OR 0.31; 0.15, 0.64) than UC (OR 0.57; 0.37, 0.88)].

In summary, dramatic changes have taken place over the past three decades in the cost of managing IBD. Biologic drug costs have risen and there has been an associated drop in the rates of both surgery and hospitalization. Cost–utility studies are therefore needed to determine whether these changes in management have been cost-effective.

Performance of high quality cost–utility comparisons is not a small matter. Ideally these studies should be purpose built and randomize a large number of patients to standard therapy or the novel intervention. All important costs should be identified prospectively and collected meticulously. Relevant outcomes such as surgeries, hospitalizations, and complications should be collected for a fixed duration of time in both groups irrespective of whether patients continue to receive investigational drug. Utility assessments should be prospectively collected using validated instruments. The study should be of sufficient size and duration to collect clinically meaningful differences in costs and utility scores. Appropriate missing data conventions should be specified in the statistical analysis which should also state the expected cost per QALY estimate for the intervention (or incremental cost-effectiveness ratio or ICER). We are currently unaware of any comparative cost–utility study in IBD that meets these rigorous requirements. For a number of reasons, many of which arise from either operational and financial constraints, cost–utility models are often performed post-hoc by using trial outcome data to generate an economic model that draws in cost and utility data from other sources. The most common approach is to utilize Markov modeling. Markov models are based on the concept that a finite number of health states (Markov states) exist for any chronic disease and that the probability of transitioning between any one state and another during a defined period of time (a “cycle” often 3–6 months) can be estimated. These cycles can be rolled out computationally for various periods of time, depending on the specific question being addressed. And a descriptive picture of the probability of an individual being in any state at any specific time over the whole course of the disease from inception until death can be estimated. Assignment of costs and utilities to each state can be used to generate cost–utility estimates that can be differentially modeled to incorporate transition probability differences observed between placebo and active treatment in controlled clinical trials. To illustrate this concept, Fig. 26.3 shows a Markov stimulation of the natural history CD in Olmstead County Minnesota [20]. In this population, the portion of patients in remission decreases rapidly 5 years after diagnosis whereas the number of patients with mild disease increases. Patients transition into a prolonged remission following surgery, but fewer patients entered remission following therapy.

To illustrate the potential value of this approach, consider the following Markov model developed by Bodger and colleagues [21] who wished to evaluate the relative cost-effectiveness of TNF-antagonists for the treatment of CD from the perspective of the UK National health Service. The Olmstead County model previously described was used as the basis for a Markov model that described the usual course of the disease, however the number of health states were reduced to five (full response, partial response, non-response, surgery, and death). Efficacy estimates for the TNF-antagonists were derived from Phase III clinical trials [22, 23]. Costs were derived from UK specific sources, while utilities were estimated using an algorithm that translated differences in CDAI scores into EQ-5D [24] utilities. Both lifetime and 1 year time horizons were evaluated.