Analytical approach:
A Markov model was used to synthesise the data from the literature and to simulate a patient's health from diagnosis of low-grade dysplasia to death. Each cycle of the model lasted for three months and the time horizon was lifetime. The authors stated that the perspective was societal.
Effectiveness data:
The estimates of the sensitivity and specificity of colonoscopy for the detection of low-grade dysplasia, the complications from ileal pouch anal anastomosis, the proportion of patients after anastomosis who required supportive care, and the positive predictive value for colonoscopy were from published literature. Cohort studies provided the estimates of the incidence rate for high-grade dysplasia or cancer in patients with low-grade dysplasia. The stage-specific cancer mortality was estimated from the Surveillance, Epidemiology and End Results database. The distribution of cancer stage, in patients who developed cancer, and the distribution of high-grade dysplasia, in patients who had progressed to advanced neoplasia, were estimated from observational studies of surveillance for low-grade dysplasia.
Monetary benefit and utility valuations:
The health utilities, for the various health states in the Markov model, were based on relevant literature.
Measure of benefit:
The measure of benefit was quality-adjusted life-years (QALYs) and these benefits were discounted at 3% per annum.
Cost data:
The cost categories were those of the direct cost of medical care. The costs of colectomy, small-bowel obstruction, and pouch failure were derived from the Nationwide Inpatient Sample (NIS) database. Physician fees, ostomy supplies, ostomy nurse time, colonoscopy, and associated pathology expenses were estimated using reimbursement schedules from the Centres for Medicare and Medicaid Services. The drug costs, such as those of oral mesalazine, were estimated from the 2005 Red Book. All costs were reported in US dollars ($) at 2005 prices and were discounted at 3% per annum.
Analysis of uncertainty:
A probabilistic sensitivity analysis was performed to assess whether the results were robust. This involved 1,000 simulations of the data and the results were displayed in a scatter plot. Univariate sensitivity analysis was also performed, on each model variable, which was varied to its lower and upper limits.
