We conducted a study to compare the cost-effectiveness of adding the serum carbohydrate deficient transferrin (CDT) test to a questionnaire when screening individuals for at-risk drinking amounts in the primary care setting. We constructed a decision tree of three strategies for detecting at-risk drinking amounts in primary care: 1) questionnaire alone, using an abbreviated version of the Alcohol Use Disorders Identification Test (AUDIT-C); 2) questionnaire followed by CDT if questionnaire negative; and 3) no formal screening (case-finding only). We estimated costs based on Medicare reimbursement for physician services and the medical literature. We determined gender-specific screening test sensitivity and specificity and prevalence of at-risk drinking amounts from the literature. We then calculated the cost per case of at-risk drinking amounts detected in 2006 dollars. Finally, we performed sensitivity analyses on all parameters. The ICER for the strategy of questionnaire alone compared with no screening was $341 per additional case of at-risk drinking amounts detected. Compared to questionnaire alone, the ICER of the combined questionnaire/CDT strategy was $4067 per additional case detected. The cost-effectiveness of the questionnaire/CDT strategy was sensitive to the prevalence of disease in the population screened but still cost $1929/case at the upper limit of prevalence tested (30%). Analyses of screening tool accuracy revealed that the ICER for a strategy with CDT is sensitive to the sensitivity and specificity of both CDT and the questionnaire in both male and female populations. When the sensitivity of the questionnaire in the male population was lower than 60% (baseline specificity of 82%), the ICER for the questionnaire/CDT strategy was less than $600 / case detected and was more cost-effective than questionnaire alone. The questionnaire/CDT strategy was also sensitive to the cost of CDT (ICER range $2500-$9600) Adding CDT to questionnaire screening is expensive in typical primary care conditions. There may be a role for adding CDT when prevalence is especially high or if the sensitivity of questionnaire is especially low. In the future, we plan to study if downstream improved health status and costs saved by detecting more at-risk drinking amounts justifies the added initial cost.