OBJECTIVES: To explore whether SAS’s new HPSPLIT procedure within SAS/STAT generates the same output as Salford System’s Predictive Modeler when computing Categorization and Regression Trees (CART). METHODS: Data was collected on a group of 744 patients treated for asthma and followed for 48 weeks. We hypothesized that the treatment (placebo vs. drug) would predict physician’s global evaluation of treatment effectiveness (GETE) response (response vs. non-response) at the end of the 48 week study controlling for sex (male, female) and race (white, non-white). The same data set was used to compute a CART model with the HPSPLIT procedure and with Predictive Modeler. RESULTS: Both methods produced a CART tree with the same values, splitting first at treatment: placebo, n=369, 54.5% response and drug, n=375, 66.9% response. The placebo group was further split by sex: male, n=107, 49.5% response and female, n=262, 56.5% response. Finally, males are split by race: white n=83, 44.6 response and non-white n=24, 66.7% response. Both software packages produced the same ROC curve (AUC=.579) and the same fit statistics: ASE=.233, Misclassification=.380, Sensitivity=.158, and Specificity=.918. Lastly, both methods produced the same variable relative importance: Treatment=1.00, Race=.561, and Sex=.357. Overall, we find that treatment is associated with the GETE response, but Sex and Race only help further differentiate the placebo group. CONCLUSIONS: As new software and applications become available it becomes paramount that researchers understand variations in approaches. In this exploration, SAS’s new HPSPLIT procedure generated the same trees and tree analysis tools as previously but exclusively available with stand-alone niche software. The results of this comparative investigation illustrated that the underlying statistical techniques of the HPSPLIT procedure and Predictive Modeler are similar and deliver consistent results.