OBJECTIVES: The DPP 3 year trial examined 3 interventions, placebo (PBO), metformin (MET) and individual lifestyle (ILS) added to baseline lifestyle modification advice for diabetes prevention. It found the two active interventions effective, with ILS being more effective. A trial-based cost-effectiveness analysis (CEA) indicated that both ILS and MET were cost-effective. Our work examined the conclusions of that CEA and sought to clarify misleading conclusions.

METHODS: We reanalyzed the CEA results, adding, as in the original paper, a group LS intervention (GLS) – modelled only (with assumed equal efficacy and reduced costs), not observed. We also examined (as in original) a generic MET alternative, MET25 (where efficacy is assumed identical, but at a cost of only 25% of the branded version), but do not use them in the same analyses as MET 25 dominates MET.

RESULTS: Instead of presenting ICERs for MET, MET25 and ILS, GLS interventions all vs. PBO as in the original CEA, we used the original CEA data according to a process outlined in Glick (2007), clarifying conclusions by eliminating dominated alternatives. In cases with MET, GLS dominates both ILS and MET with an ICER vs. PBO within acceptable WTP thresholds ($8982/QALY). If GLS is excluded (as it was only a modeled intervention), ILS dominates MET and has an ICER likely to be acceptable ($31,512/QALY). The same qualitative results were seen for MET25.

CONCLUSIONS: The conclusion of the original CEA was that both ILS and MET were cost-effective. We have shown this to be false, using the exact data as in the original CEA. Based on these data, there is no place for MET in diabetes prevention. We suggest that graphical depictions of results in the cost-effectiveness plane or in a less common cost-disutility plane might have driven home the correct conclusion more clearly than the numerical results.