Introduction: Health outcomes modeling of benefit-risk tradeoffs is a potential regulatory decision aid. We quantified the incremental net benefit of rosiglitazone at the time of approval (1999) for montherapy and advisory committee review (2007). Methods: We adapted a commercial Markov model to estimate net benefit of diabetes monotherapy using various data. Results: With 1999 data, rosiglitazone was projected to extend life by a mean of 0.639 years (0.373 quality-adjusted life years) versus placebo, but shorten life by 0.312 years (0.173 QALYs) versus glyburide. Uncertainty in HbA1c reduction had greater impact on the benefit-risk profile than uncertainty in harm. By 2007, rosiglitazone was projected to add 0.222 years (0.091 QALYs) and 0.026 years (0.009 QALYs) versus glyburide or metformin. Conclusions: Modeling offered new interpretations on rosiglitazone’s tradeoffs.