Methodology for monitoring free valproic acid (VPA) and experimental proof of intrapatient in vivo diurnally and disproportionately variable free VPA fractions has recently been reported. An inherent assumption in accepting therapeutic VPA plasma levels to be 50-100 micrograms/ml is that free fractions remain constant. This assumption is no longer tenable. Therefore, monitoring only VPA plasma levels could be misleading. The serial 9-h time-course (7 a.m. 0 10-1-4 p.m.) of free versus plasma VPA levels was investigated in 24 patients. Limits for diurnal fluctuations were (mean +/- SD): Group A (n = 14); 49.0 mg/kg; multiple equal dosing: Free VPA (micrograms/ml) 6.06 (1.55 to 12.62 (4.89), plasma VPA (micrograms/ml) 52.9 (11.6) to 84.2 (21.3), percent free VPA 11.5 (1.8) to 14.9 (2.5). Group B (n = 10); 30.6 mg/kg; b.i.d.: 12.0 (2.4) mg/kg a.m. and 18.7 (3.3) mg/kg p.m.: Free VPA 5.53 (1.04) to 9.92 (1.51), plasma VPA 52.0 (7.3) to 79.2 (9.6), percent free VPA 10.7 (1.6) to 12.5 (0.8). Reducing the dosage by 19 mg/kg (A to B) decreased VPA plasma levels by 6.9% nd free VPA levels by 28.9%. For B: y = 8.15 + 4.03 x; n = 10, r = 0.954, Sy.x = 1.38, when x = steady-state (7 a.m.) free VPA concentration and y = VPA mg/kg/day. The findings suggest that multiple dosing is unnecessary. Similar plasma levels with far less diurnal fluctuations of free levels are achievable by a smaller drug dose with approximately two-thirds of total daily dose being administered in the evening and one-third in the morning.