Clozapine (CLZ) and metabolites norclozapine and clozapine-N-oxide were assayed with a new, sensitive (2 pmol), and selective method in 68 serum samples from 44 psychotic subjects, 20 to 54 years old, ill 16 years, and treated with CLZ for 2.2 years (currently at 294 mg, 3.4 mg/kg daily). CLZ levels averaged 239 ng/ml (0.73 microM; 92 ng/ml per mg/kg dose) or 48% of total analytes (norclozapine = 41% [91% of CLZ] and clozapine-N-oxide = 11%); metabolite and CLZ levels were highly correlated (rs = 0.9), and CLZ levels varied with daily dose (rs = 0.7). Sampling twice yielded similar within-subject analyte levels (r = 0.8 to 0.9; difference = 24% to 33%). Range and variance narrowed when levels were expressed per weight-corrected dose (ng/ml per mg/kg). Levels per dose were 40% higher in nonsmoking women than men, despite a 60% lower milligram per kilogram dose in women, and did not vary by diagnosis or age in this limited sample. Fluoxetine increased serum CLZ analytes by 60%; valproate had less effect. Patients rated treatment very positively; observer-assessed benefits typically were more moderate. Common late side effects were sialorrhea (80%), excess sedation (58%), obesity (55% > 200 lb), mild tachycardia (51%), constipation (32%), and enuresis (27%); there were no seizures or leukopenia. There was little evident relationship of drug dose or serum level to current clinical measures or side effect risks.