OBJECTIVES: To describe characteristics of T2DM patient subgroups who were more likely to achieve HbA1c goal < 7% with combination treatment of DPP4i with PIO or with MET using a predictive model.  METHODS: Stepwise logistic regression was applied to MarketScan claims data to develop a predictive model that estimated the probabilities of HbA1c goal achievement in patients receiving DPP4i combinations. Sample selection criteria included: 1) T2DM diagnosis; 2) treatment of DPP4i with PIO or with MET; 3) baseline HbA1c ≥ 7%; and 4) with one-year continuous enrollment. Patients were ranked by the probabilityof achieving HbA1c<7% and grouped into cumulative percentiles; baseline characteristics of the optimal subgroups identified as the first 20th and 80thpercentiles were reported.  RESULTS: A total of 328 patients were included.  The predictive model showed that patients who had neuropathy, cerebrovascular conditions, or higher total medication use at baseline were less likely to achieve goal on DPP4i combinations while patients with self monitoring blood glucose use at baseline were more likely to achieve goal (P<0.05).  The 80th percentile subgroup (n=270) had a goal reaching rate of 57.0%, mean age of 50.3 years old, 44.3%  female, 38.5% on MET, 13.8% on thiazolidinedione (TZD), and HbA1c = 9.13% at baseline. The 20thpercentile subgroup (n=83), achieved goal at the rate of 72.3%, mean age of 50.6 years old, 46.1% female, 53.9%  on MET, 25.7% on TZD, and HbA1c = 8.96% at baseline.    CONCLUSIONS: Predictive factors for reaching goal include: 1) use of self monitoring blood glucose, and 2) lack of neuropathy, cerebrovascular disease, or usage of medications. Subgroups that might benefit the most from DPP4i treatment were identified. These patients exhibited a higher likelihood of having prior use of MET or TZD, and baseline HbA1c less than 9.0%.