OBJECTIVES

: A cost-effectiveness analysis (CEA) of the 3 year Diabetes Prevention Program (DPP) concluded that individual, group lifestyle and metformin (ILS, GLS, MET) interventions were all cost-effective. We reanalyzed that data and showed there was no role for MET or ILS. Heterogeneity of treatment effect (HTE) was not considered. A recent clinical analysis (Sussman, 2015) of DPP used a diabetes risk prediction model, indicating substantial baseline HTE, leaving a possible path for cost-effectiveness for MET in a high risk subgroup (HRS), where its effect was most pronounced.

METHODS

: We used Sussman’s 3 year cumulative incidences and a constant rate assumption for developing diabetes to estimate patients with diabetes for years 1 and 2. Those were combined with information on mean QALYs and costs by diabetes status from the DPP to assess cost-effectiveness in the HRS. We also applied a QALY calculation correction to the original CEA that used end of year utilities as if they applied for the entire year, instead of using averages based on t and t+1 values.

RESULTS

: The QALY correction made only small changes to QALYsThe ICER for GLS versus MET in the high risk group, using corrected QALYs was $155,449, indicating that GLS was unlikely to be cost effective and that MET was. For uncorrected QALYs, the high risk group ICER was only $61,840, potentially in the range of acceptability for GLS over MET. The seemingly small QALY error had important implications.

CONCLUSIONS

: For the highest risk group, MET may be cost-effective. Had the original CEA been done correctly and had it also explored HTE (retaining its QALY error), it might have concluded GLS was cost-effective and MET was not. It is only with our corrected QALYs that there appears to be a place in diabetes prevention for metformin – in the highest risk group.