Unhealthy alcohol use (the spectrum of risky use through dependence) is common in HIV-infected persons, yet it can interfere with HIV medication adherence, may lower CD4 cell count, and can cause hepatic injury. Carbohydrate-deficient transferrin (CDT), often measured as %CDT, can detect heavy drinking but whether it does in people with HIV is not well established. We evaluated the operating characteristics of %CDT in HIV-infected adults using cross-sectional data from 300 HIV-infected adults with current or past alcohol problems. Past 30-day alcohol consumption was determined using the Timeline Followback (TLFB), a validated structured recall questionnaire, as the reference standard. Sensitivity and specificity of %CDT (at manufacturer's cut-off point of 2.6%) for detecting both "at-risk" (≥4 drinks in a day or >7 drinks per week for women, ≥5 drinks in a day or >14 per week for men) and "heavy" drinking (≥4 drinks in a day for women, ≥5 drinks in a day for men on at least seven days) were calculated. Receiver operating characteristic (ROC) curves were estimated to summarize the diagnostic ability of %CDT for distinguishing "at risk" and "heavy" levels of drinking. Exploratory analyses that stratified by gender and viral hepatitis infection were performed. Of 300 subjects, 103 reported current consumption at "at-risk" amounts, and 47 reported "heavy" amounts. For "at-risk" drinking, sensitivity of %CDT was 28% (95% confidence interval (CI) 19%, 37%), specificity 90% (95% CI 86%, 94%); area under the ROC curve (AUC) was 0.59. For "heavy" drinking, sensitivity was 36% (95% CI 22%, 50%), specificity 88% (95% CI 84%, 92%); AUC was 0.60. Sensitivity appeared lower among women and those with viral hepatitis; specificity was similar across subgroups. Among HIV-infected adults, %CDT testing yielded good specificity, but poor sensitivity for detecting "at-risk" and "heavy" alcohol consumption, limiting its clinical utility for detecting unhealthy alcohol use in this population.