Can modeling of health outcomes facilitate regulatory decision making? The benefit-risk tradeoff for rosiglitazone in 1999 vs. 2007.

摘要

Rosiglitazone was initially approved for type 2 diabetes monotherapy. We tested health-outcomes modeling as an aid to regulatory decision making by quantifying the incremental net benefit (INB) value of rosiglitazone (relative to a comparator), both at the time of first approval (1999) and at the FDA advisory committee review (2007). Using 1999 data, rosiglitazone was projected to provide an additional 0.639 years of life (0.373 quality-adjusted life years (QALYs)) relative to placebo but a loss of 0.312 years (0.173 QALYs) relative to glyburide, with uncertainty in reduction of hemoglobin A(1c) (HbA(1c)) level having the greatest impact on the benefit-risk profile. By 2007, rosiglitazone was projected to provide an additional 0.222 years (0.091 QALYs) vs. glyburide and 0.026 years vs. metformin (0.009 QALYs). Modeling suggested that the use of rosiglitazone as monotherapy was not initially warranted, given the uncertainty with regard to benefit. Despite similar net benefit (NB) as metformin shown in postmarketing data, residual cardiovascular (CV) concerns did not support the use of rosiglitazone as first-line therapy. We adapted a mathematical diabetes model to estimate NB and uncertainty of diabetes monotherapy.