This was a pharmacological companion study to a randomized Phase III trial comparing 21-day oral versus 3-day i.v. etoposide in combination with i.v. cisplatin in patients with extensive-stage small cell lung cancer. Etoposide plasma concentrations were measured in patients randomized to the 21-day schedule and correlated with toxicity and tumor response. Patients were treated with etoposide (50 mg/m2/day) orally for 21 days and cisplatin (33 mg/m2/day) i.v. for 3 consecutive days every 28 days for 6 courses. Plasma samples before the daily etoposide dose (trough concentrations) and complete blood counts were obtained weekly during treatment. The average of three etoposide concentrations (EC) per course was calculated. Of 158 patients registered to this schedule of the study, 150 were eligible. In 106 patients, etoposide samples were obtained at least in the first course in which the mean EC was 0.39 microgram/ml (SD = 0.29). In 102 patients (missing albumin values in 4 of 106 patients), the concentration of etoposide not bound to protein (Efree) was estimated based on the following equation: percentage unbound = (1.4 x total bilirubin) - (6.8 x albumin) + 34.4. Regression analysis revealed that increasing age was correlated with higher EC (r = 0.27; two-tailed P < 0.01) and Efree (r = 0.31; two-tailed P < 0.01). Higher EC and Efree values were associated with lower WBC counts and absolute neutrophil counts after the first treatment course in 83 patients with nadir counts. Using multiple linear regression, a pharmacodynamic model was developed that included EC or Efree, age, and alkaline phosphatase. An interaction with bone marrow results at diagnosis was found, indicating a sharper decline in nadir counts with increasing EC or Efree when the marrow was involved with small cell lung cancer. This model explained 29% of the variation for WBC nadirs (P < 0.001) and 31% of the variation for absolute neutrophil count nadirs (P < 0. 001). Neither EC nor Efree showed a significant correlation with tumor response. A pharmacokinetic relationship between EC or Efree and age was found. A pharmacodynamic model could be developed for toxicity but not for tumor response.