We conducted a prospective, randomized, cross-sectional study to develop and validate a new model to predict oxycodone in urine that can be used to help evaluate whether patients are complying with their oxycodone dosing regimens. We studied 20 patients: eight black women, two white women, six black men, and four white men; ages 48 ± 10 years (mean ± SD); weight 97 ± 32 kg. Pain levels before treatment averaged 9.5 ± 0.9 out of 10. We prescribed oral oxycodone for each patient, tailoring the dosing regimen using clinical pharmacokinetics and measured the oxycodone concentration in each patient's urine 10 to 14 days after starting the dosing regimen. For each patient, we predicted oxycodone in their urine using our model, checked the actual concentration, and compared predicted with actual concentrations. For 18 of 20 patients (90%), actual results fell within ±10% of our model's prediction. One patient was 35% below the prediction; the other was 51% above. Our model accurately predicts oxycodone in urine (±10% for 90% of the patients). The model appears clinically useful for evaluating the results of a quantitative urine test, since it objectively discriminates between (1) a "normal" patient complying with their oxycodone dosing regimen, and (2) a patient who may require genetic testing to distinguish between unusual metabolism or abuse.