In certain populations, the biological alcohol marker carbohydrate-deficient transferrin (CDT) is known to have a high diagnostic accuracy. The aim of this study was to compare the diagnostic accuracy of CDT, gamma-glutamyltransferase (gamma-GT), and mean cell volume (MCV) in a general practice population; more specifically, to ascertain whether CDT is a better tool than gamma-GT and MCV for (early) recognition of excessive alcohol use. To represent the general practice situation as realistically as possible, three different drinking patterns are defined: irregular excessive, regular excessive, and very excessive. From a sample of 524 men from seven general practices, sensitivity, specificity, and predictive values of the three markers for the three drinking patterns were compared, and receiver-operating characteristic analysis was used to compare differences between the markers. The results indicate that drinking patterns do influence the (difference in) diagnostic accuracy. CDT has a higher diagnostic accuracy for all three drinking patterns than gamma-GT and higher predictive values for hazardous [(ir)regular excessive] drinking patterns than MCV. However, receiver-operating characteristic analyses failed to demonstrate a significant difference between these patterns. It is concluded that the performance of all tests is too low to be useful for screening procedures in a general population; however, some tests may be useful for case finding. CDT seems to be the best alcohol marker available, although the difference between CDT and MCV is small.