The pharmacodynamics of total and unbound etoposide was studied in 28 adult patients with solid tumors who were receiving etoposide and cisplatin combination chemotherapy. Etoposide plasma concentrations were determined by use of an HPLC method, and etoposide plasma protein binding was determined by equilibrium dialysis. Patients with higher systemic exposure experienced greater hematologic toxicity. The sigmoid maximum effect model with unbound systemic exposure performed better (i.e., lower residual sum of squares) than the model using total systemic exposure; in all patients (7043 versus 9755) and in the subset of patients who had not received previous chemotherapy (1986 versus 3664). The model estimates for unbound systemic exposure were more precisely estimated than for total systemic exposure (e.g., coefficient of variation for the area under the concentration-time curve producing half of the maximal effect = 51% for total drug versus 21% for unbound drug). These findings indicate that the hematologic toxicity of etoposide is better correlated with systemic exposure to unbound drug than total drug, which may be of clinical importance because of the variable plasma protein binding of etoposide.