Detection of alcohol abuse in neurological patients: variables of clinical relevance to the accuracy of the %CDT-TIA and CDTect methods.

OBJECTIVE

Alcohol-related neurological diseases are encountered frequently. Early diagnosis is essential, because minimal intervention effectively reduces hazardous alcohol consumption and may prevent permanent neurological damage. Carbohydrate-deficient transferrin (CDT) is a valuable tool for the identification of alcohol abuse, but for unselected patient populations, reduced test accuracy has been reported. Recently, factors other than alcohol use have been shown to influence CDT levels. Our aim was to identify clinically relevant factors that might reduce test accuracy.

METHODS

We included 397 neurological patients consecutively hospitalized for seizures, ischemic stroke, or sciatica and 87 patients who attended routine outpatient controls for epilepsy. Blood samples were analyzed for CDT by using two commercially available tests, %CDT-TIA and CDTect. All patients underwent a semistructured clinical interview that included a record of the reported ethanol consumption during the last 8 days, and all completed the Alcohol Use Disorders Identification Test (AUDIT). Current medication, medical history, and demographic information also were obtained.

RESULTS

Both tests were elevated in female antiepileptic drug users, compared with others who reported no recent ethanol intake. A higher number of false-positive cases was seen for CDTect than for %CDT. Various combinations of CDT and gamma-glutamyltransferase improved sensitivity, but at the cost of reduced specificity. Variables that predicted the variation of CDT included antiepileptic drug use, sex, body mass index, and smoking. Total transferrin levels were reduced significantly in postmenopausal women, whereas a falling trend was seen for CDTect. Transferrin alterations caused a higher number of false-positive results for CDTect than for %CDT. The area under the receiver operating characteristics curve for women was higher for CDTect than for %CDT, and for %CDT, the area under the receiver operating characteristics curve was higher for men than for women.

CONCLUSIONS

The accuracy of CDT for detection of alcohol abuse in neurological patients was generally low, particularly for women. Combination variables of CDT and gamma-glutamyltransferase did not increase test accuracy. Variables that were associated with higher CDT levels included female sex, antiepileptic drug use, transferrin alterations, and possibly low body mass index. When factors known to cause poor accuracy in particular patient groups are appreciated, CDT may be a good adjunct to the clinical examination.