Dr. Thompson notes that an abbreviated version of the AUDIT that includes 5 of the 10 questions gives nearly the same results as the complete instrument. A three-item version of the AUDIT (the AUDIT-C), including the alcohol consumption questions, has also been developed by Bush and colleagues (1). It is interesting that the consumption questions are in large part what distinguish the AUDIT from the other self-report screening instruments we evaluated. Although beyond the scope of this letter, a reanalysis of our data focusing on an abbreviated version, including the consumption questions, may help validate these other findings.

Dr. Saitz argues that the accuracy findings may be a function of the methods we used. Methodologic standards for conducting and reporting studies of diagnostic test performance guided our design and analysis (2). We used a well-validated diagnostic interview (the Alcohol Use Disorder and Associated Disabilities Interview Schedule), completed by all study participants (absence of work-up bias), to determine Diagnostic and Statistical Manual of Mental Disorders, fourth edition, diagnoses of alcohol abuse and dependence (both in the past year and over the participant's lifetime). An analysis using lifetime disorders as the criteria was consistent with findings reported in our paper for current (past year) disorders. Although at least one other study has raised concerns about asking consumption questions before administering the CAGE questionnaire, the findings for the CAGE questionnaire and for the Self-Administered Alcoholism Screening Test (SAAST) were similar in our study (in addition, the SAAST, which includes three of the CAGE questions, was completed before the consumption questions were asked). It is true that adopting a cutoff of one "yes" response to the CAGE questions, rather than two, would increase the sensitivity of this instrument. However, the overall accuracy of the CAGE questionnaire in several patient subgroups was poor, and lowering the cutoff would decrease specificity in addition to positive predictive value.

Dr. Saitz also comments that the AUDIT may be too long to memorize and incorporate as part of the medical interview. Again, an abbreviated version might be helpful. Furthermore, the AUDIT can easily become part of the periodic health examination if a computerized information system is used for administration and scoring. A greater challenge seems to be demonstrating the utility of the AUDIT (or any screening approach) in guiding the management of patients with alcohol use problems.