Background: The aim of this investigation was to assess clinical predictors of the glycosylated hemoglobin (HbA1C) response after the addition of a thiazolidinedione (TZD) to a biguanide, a sulfonylurea, or both in subjects with type 2 diabetes.
Methods: Chart review (n = 68 physicians) was used to identify consecutive subjects started on a TZD. Qualifying subjects had been treated with pioglitazone (> or = 4 mg/day) or rosiglitazone (> or = 30 mg/day) for > or = 12 weeks. Clinical characteristics and HbA1C responses were assessed for the purpose of creating an initial predictive response model (Study 1). A separate sample from a managed care database was used to independently validate the model (Study 2).
Results: Data were collected from 4085 subjects (1365 in Study 1; 2720 in Study 2). In Study 1, baseline HbA1C was 8.2 +/- 0.1%. Forty-five percent (611 of 1365) and 55% (754 of 1365) were prescribed pioglitazone and rosiglitazone, respectively. In multivariate regression, baseline HbA1C (beta = -0.693%), age (beta = -0.006%), and use of multiple agents at baseline (referent = single agent, beta = 0.189%) were significant (P < 0.05) predictors, explaining 49% of the variance in HbA1C response in Study 1 and 44% of the variance in HbA1C response in the Study 2 sample. The model showed no material evidence of bias across the range of baseline HbA1C values.
Conclusions: These results suggest that readily available clinical information, particularly baseline HbA1C, explains a substantial proportion of the variance in the response to TZD therapy.