Response to Lamotte et al.

In the February 2002 issue of Diabetes Care, Lamotte et al. (1) present a provocative study of the cost-effectiveness of orlistat in obese type 2 diabetic patients. We suspect, however, that the authors have greatly overestimated the cost-effectiveness of this drug, due primarily to inappropriate parameter estimates.

First, the authors estimated the impact of reductions in HbA_1c on the risks of macrovascular complications and death based on a finding from the Diabetes Control and Complications Trial (DCCT)—a 40% decrease in the risk of complications for every 10% reduction in HbA_1c (2,3). The DCCT, however, included only type 1 diabetic patients, and this particular finding applied to microvascular complications only. Nevertheless, Lamotte et al. used this estimate—along with data from the U.K. Prospective Diabetes Study (UKPDS)—to project the risks of microvascular and macrovascular complications, as well as death, among type 2 diabetic patients (4). The authors estimated that a 0.46% absolute decrease in HbA_1c, the assumed reduction with orlistat therapy (versus placebo), would reduce the annual risk of mortality by 27%, myocardial infarction (MI) by 29%, and stroke by 31% (5).

In our opinion, a better source with which to estimate these relationships is the UKPDS study of the association between glycemia and the complications of type 2 diabetes (6). In this study, a 1% absolute reduction in HbA_1c was associated with 14% risk reductions in mortality and MI and a 12% reduction in stroke. Based on these data, estimated risk reductions with orlistat therapy (assuming a 0.46% absolute decrease in HbA_1c) would have been substantially lower; 7% for mortality and MI, and 6% for stroke.

Second, they estimated the relationship between LDL cholesterol level and coronary events among type 2 diabetic patients using data from the Helsinki Heart Study (HHS), which was not adequately powered to accurately estimate treatment effects within subgroups (7,8). In this study, a 21.9-mg/dl reduction in LDL cholesterol was associated with a nominal 68% reduction in MI and cardiac death.

More reliable estimates can be found, we believe, in subgroup analyses from secondary prevention studies. In the Coronary Atherosclerosis and Recurrent Events (CARE) study, a mean 37-mg/dl decrease in LDL cholesterol was associated with a significant 25% reduction in coronary events (9). In the Scandinavian Simvastatin Survival Study (4S), an ∼75- mg/dl reduction in LDL cholesterol was associated with a significant 57% reduction in major coronary events (10). These estimates are substantially lower than the HHS estimate (a reduction in coronary events of <1% per 1 mg/dl decrease in LDL cholesterol, versus 3% in the HHS), even though they occurred in the context of secondary prevention.

In summary, while it is impossible for us to calculate with certainty the effect that questionable parameter estimates had on their results, we believe it is quite likely that if the authors had used more appropriate data sources, their study would have yielded substantially higher cost-effectiveness ratios (i.e., they would have found orlistat to be substantially less cost-effective).