Analytical approach:
A Markov model was used to combine published evidence to assess the costs and outcomes for the two interventions. The time horizon was the lifetime of the patient. The authors reported that a societal perspective was adopted.
Effectiveness data:
The effectiveness data were from published studies. Based on those comparing aripiprazole with olanzapine, in patients with schizophrenia,their efficacy was found to be equivalent. The same was assumed for patients with bipolar disorder, as there were no head-to-head trials for these patients and studies had shown that the risk of metabolic syndrome was comparable for patients with either schizophrenia or bipolar disorder. Published risk prediction models were used to estimate the risk of developing coronary heart disease or diabetes. The main effectiveness estimate was the risk of developing metabolic syndrome after one year of treatment. This risk was from a pooled analysis of three randomised controlled trials.
Monetary benefit and utility valuations:
The literature was reviewed to identify the health utility data. Studies that used the European Quality of life (EQ-5D) questionnaire were selected.
Measure of benefit:
The measures of benefit were life-years and quality-adjusted life-years (QALYs) gained. Future benefits were discounted at an annual rate of 3%.
Cost data:
The direct costs were those of the drugs and the monitoring and treatment of metabolic syndrome, diabetes, and coronary heart disease. This included office visits to the doctor, nurse or dietician, tests, medications, and in-patient care. A clinical expert estimated the resource use for the monitoring and treatment of metabolic risk factors, and this was valued using Swedish unit costs. The drug costs were from national pharmaceutical price lists. A literature review was undertaken to identify the costs of treating diabetes and coronary heart disease. Productivity lost due to metabolic syndrome, diabetes, and coronary heart disease was included. The number of sick days was from published studies, and was valued using average national gross income. All costs were updated to 2009 prices, using Swedish health care consumer price indices, and presented in Swedish kronor (SEK). Future costs were discounted at an annual rate of 3%.
Analysis of uncertainty:
One-way sensitivity analyses were undertaken to determine which parameters had the most impact on the results. A probabilistic sensitivity analysis was performed to analyse the overall uncertainty in the model; every parameter was assigned a statistical distribution and the results were presented in cost-effectiveness scatter plots.