An analysis is presented that promotes the use of uric acid levels as an indicator of patients didanosine (ddI) use. Logistic regression techniques were used on data from two test groups that best represent patients entered on ACTG protocols 116A and 116B/117. Two classification functions resulted that are based on serial uric acid measurements and were used to classify patients to one of two groups: those treated with zidovudine (ZDV) and those treated with ddI. These functions correctly classified well over 70% of the patients in each of these two studies. Implications of these results with respect to the assessment of ddI compliance and limitations in the use of these classification functions are discussed.