In Reply.—
The letter by Kirch and colleagues has raised questions about the statistical procedures used in our analysis of haloperidol levels and clinical response.1 We believe that some of the points Kirch et al raise may not be valid; other legitimate concerns they raise have been addressed in a subsequent report from our group.2First, Kirch et al question the use of the linear regression procedure to estimate improvement in psychopathology scores. The rationale and justification for our statistical procedures and the use of linear regression have been presented in greater detail in our subsequent article,2 which analyzed a slightly enlarged sample. Briefly, linear regressions were fitted to the patients' BPRS psychosis factor scores because scatter plots of these psychopathology scores by the day of treatment did not support the assumption of curvilinear trends for their response to haloperidol during the first 31/2 weeks of drug treatment. Only one of 34 patients of