Prospective evaluation of a model for predicting etoposide plasma protein binding in cancer patients.

Etoposide protein binding in cancer patients is variable and has been related mathematically to a linear model consisting of serum albumin and total bilirubin [percentage unbound = (1.4 x total bilirubin) - (6.8 x albumin) + 34.4]. In this prospective evaluation of the model, plasma samples were obtained following the administration of etoposide in 31 patients, and the unbound percentage (% unbound) of etoposide in plasma was determined by equilibrium dialysis. The mean measured % unbound was 15.3 +/- 11.6 (SD), and the mean model predicted % unbound was 16.7 +/- 10.1. The relation between predicted and measured etoposide % unbound was highly correlated (r2 = 0.92, P = 0.001). The model was precise but with a slight bias toward overpredicting % unbound (mean prediction error, 1.36%; 95% confidence interval, 0.09 to 2.6%). In patients with abnormal total bilirubin (i.e., greater than 1.5 mg/dl) or with hypoalbuminemia (i.e., less than 3.3 g/dl), the model was both precise and unbiased. These results demonstrate that etoposide % unbound can be predicted using serum albumin and total bilirubin. This model should be useful in prospectively identifying patients at increased risk of experiencing altered pharmacological effects due to altered protein binding of etoposide.