This retrospective analysis was designed to evaluate the inactivation index (I3) method used to adjust the isoniazid dose during long-term administration in a pediatric population. Before starting on antituberculosis therapy, sixty-one children received one 10 mg.kg-1 isoniazid test-dose (D). The isoniazid and acetyl isoniazid concentrations were measured by high-performance liquid chromatography on a plasma sample collected 3 hours (C3h) after administration. The patients were separated into slow and fast acetylator groups according to the metabolic ratio. The dose adjustment method using the I3 is based on the assumption that there is a linear correlation between C3h and D [C3h = (I3 x D) - 0.6] in which the slope is I3 and the Y intercept is equal to -0.6 mg.l-1. I3 was determined from a single plasma concentration determination and used to calculate the dose recommended to obtain a desired C3h equal to 1.5 micrograms.ml-1: recommended dose (mg.kg-1) = (1.5 + 0.6)/I3.I3 was significantly higher in the slow acetylator group (0.55 +/- 0.16) than in the fast one (0.26 +/- 0.13), which leads us to recommend a significantly lower dose in the slow acetylator group (4.2 +/- 1.5 mg.kg-1) than in the fast one (10.3 +/- 4.6 mg.kg-1). The data obtained in a subgroup of 21 patients who had at least three consecutive determinations of C3h after different dosages allowed us to verify that there was a linear correlation between C3h and the dose. The mean slope of the correlation lines in that subgroup was 0.61 +/- 0.25 and the 95% confidence interval of the estimated Y-intercept include the theoretical value of -0.60, which shows that our data are consistent with those previously reported in adults. The percentage of patients with a C3h plasma concentration within the expected range (1.5 +/- 0.5 micrograms.ml-1) was significantly higher (69%) in those whose dose was derived from the calculation than in the others (25%). Within each acetylator group, the range of the recommended dose varied widely, and these results emphasize the usefulness of individual dose adjustment based on the inactivation index method.