To the Editor.
—In their recent article, Dr Henrard and colleagues1 show that human immunodeficiency virus (HIV) RNA levels, p24 antigenemia, and CD4 lymphocyte percentage independently predicted the risk of acquired immunodeficiency syndrome (AIDS), using Cox proportional hazards models with time-updated covariates. We wonder whether this conclusion is sensitive to different methods of fitting the model. In particular, we and others have found that the logarithm of the CD4 cell percentage provides a better fit to the model than the untransformed count. In the model fitted by Henrard et al, the relative difference in the risk of AIDS between patients with a CD4 lymphocyte percentage of 30% and 25% is assumed to be the same as that between patients with CD4 cell percentage of 10% and 5%. In comparison, if the logarithm of the CD4 cell percentage was used, the relative difference in the risk of AIDS between people
Mocroft A, Sabin CA, Phillips AN. Percentage of CD4 Lymphocytes and Risk of AIDS. JAMA. 1995;274(23):1836. doi:10.1001/jama.1995.03530230022016