Prediction of CD4 count from CD4 percentage experience from three laboratories

Abstract

CD4 counts have been used to monitor progression of disease in HIV infection as criteria for initiation of therapy, and to stratify and follow patients in clinical trials. Recently, the Centers for Disease Control and Prevention (CDC) has made CD4 counts part of the classification of HIV disease. Because a CD4 percentage may be the only laboratory information available, this study was initiated to determine whether the correlation between CD4 percentages and CD4 counts is sufficiently high to enable these measures to be substituted for each other. One thousand consecutive CD4 measurements from the University of Washington (UW) were used to create a model that was tested using datasets of 1000 CD4 measurements each from Maryland Medical Laboratories (MML) and Rush-Presbyterian-St Luke's Medical Center (Rush). The patients were not selected for age, sex, risk group or treatment. All patients from MML and Rush were known to be HIV-positive, while the HIV status of all UW patients was unknown. The model predicted that a patient with a CD4 percentage > or = 14% would have a CD4 count > or = 200 x 10(6)/l(if CD4 percentage of 14% was used, 9% of patients would have a CD4 count > or = 200 x 10(6)/l), and a patient with a CD4 percentage > or = 27% would have a CD4 count > or = 500 x 10(6)/l(if CD4 percentage of 27% was used, 17% of patients would have a CD4 count > or = 500 x 10(6)/l). These CD4 percentage correlations may be useful when a white blood cell and lymphocyte count are not available to calculate the CD4 count.