Methods: Individual patient data from a randomized controlled clinical trial was used to populate a Markov model to forecast long term costs and outcomes. The RCT compared six month exercise programs with a control group with no organized program. Three exercise programs were considered; a resistance program, an aerobic program and a combination of both. Each program involved patients exercising on a regular basis with up to eleven sessions with a personal trainer. Data was collected on clinical outcomes at twelve months as well as resource use during the twelve month period. Outcomes at twelve months were inputted for individual patients into the UKPDS economic model for type 2 diabetes which had been adapted for the Canadian context. From this, expected life years, QALYs and costs were estimated for all patients within the trial (discounted at 5% per annum). Uncertainty was assessed in terms of sampling uncertainty with respect to the clinical trial population (through non parametric bootstrapping) and model uncertainty (through Monte Carlo simulation).
Results: Aerobic exercise alone was the most expensive program ($13675) followed by combined exercise ($13614), no exercise ($13207) and resistance exercise ($12613). QALYs was highest for combined (8.9), followed by aerobic (8.8), resistance (8.7) and control (8.7). The incremental cost per QALY gained for the combined exercise program was dominant compared to aerobic alone, $5004 compared to resistance alone and $2035 compared to no program. Analysis focusing on sampling uncertainty found the probability that the combined program was the most cost effective when a QALY was worth $50 000 was 99.9%. Incorporating model uncertainty, the probability was 92.4%.
Conclusions: A combined program providing training in both resistance and aerobic exercise was the most cost effective of the alternatives compared. Within a public health care system, the funding of exercise training for diabetics can be considered an efficient use of resources.