Pramipexole and Levodopa in Early Parkinson's Disease

BACKGROUND AND OBJECTIVE: In chronic disease, treatment effects and costs accumulate over time; hence, the choice of time horizon in cost-effectiveness analysis can be particularly important. In this article we analyse the dynamic changes in cumulative costs, effects and incremental cost effectiveness of two competing drug strategies in patients with early Parkinson's disease (PD). The incremental cost-effectiveness ratio (ICER) and the net monetary benefit (NB) [using $US50 000, $US100 000 and $US150 000 as the value of a QALY] were estimated, and were bootstrapped to calculate the standard errors. Cost-effectiveness acceptability curves (CEAC) were built to estimate the probability that pramipexole was cost effective given different societal values of QALY, for various study horizons. We conducted sensitivity analyses on the ICER and the NB to test their robustness to various assumptions about missing data, for various subpopulations and under changes in the drug prices. METHODS: Three hundred and one subjects with PD were randomised to initial pramipexole or levodopa and followed every 3 months over a 4-year period. Healthcare resource use was recorded in patient diaries and valued using a variety of sources at year 2002 $US values. Health-related quality of life (HRQoL) was measured using the EuroQoL EQ-5D. The study was conducted from a US societal perspective. Missing data were imputed using a multivariate fixed-effects model. Additional quality adjusted life years (QALY) gained by using pramipexole compared with levodopa were estimated as the area between the normalised treatment HRQoL profiles. The QALYs and costs for each treatment arm were calculated for various study horizons. The incremental cost-effectiveness ratio (ICER) and the net monetary benefit (NB) [using $US50 000, $US100 000 and $US150 000 as the value of a QALY] were estimated, and were bootstrapped to calculate the standard errors. Cost-effectiveness acceptability curves (CEAC) were built to estimate the probability that pramipexole was cost effective given different societal values of QALY, for various study horizons. We conducted sensitivity analyses on the ICER and the NB to test their robustness to various assumptions about missing data, for various subpopulations and under changes in the drug prices. RESULTS: Under the base-case assumptions, the ICER for pramipexole was $US42 989 per QALY. Using the CEAC approach, the probability that pramipexole was cost effective relative to levodopa over the first 4 years was 0.57, 0.77 and 0.82 when a QALY was valued at $US50 000, $US100 000, and $US150 000, respectively. Over time, the ICER for pramipexole improved and uncertainty around the ICER decreased. If, after treatment withdrawal, HRQoL improved in pramipexole subjects and declined in levodopa subjects (best-case scenario for pramipexole), the probability of pramipexole being cost effective increased to 0.88, 0.96 and 0.98, respectively. Factors that improved the ICER of pramipexole were a decrease in the relative price of pramipexole and having low HRQoL or depression at baseline. CONCLUSIONS: The cost effectiveness of pramipexole compared with levodopa in the treatment of early PD increased as the time horizon of the clinical trial extended from 2 to 4 years. Our results suggest that pramipexole is more cost effective for patients with depression and low baseline HRQoL than in other patient subgroups.