Estimating and comparing reduction in HIV-1 RNA in clinical trials using methods for interval censored data.

The magnitude of reduction in HIV-1 RNA levels provides an important complement to the end point based on the percentage of patients achieving HIV-1 RNA levels below a threshold value. Analyses and interpretation of this end point, however, is difficult due to the lower limit of quantification. Crude methods of analyzing HIV-1 RNA data provide biased estimates of the HIV-1 RNA reduction. Censored methods that take into account the censoring of HIV-1 RNA measurements by the limit of quantification greatly improve the analysis of HIV-1 RNA reduction end points. It was shown, however, that when there is a high percentage of censoring, those methods can overestimate HIV-1 RNA reduction. We suggest going a step further, considering that HIV-1 RNA reduction is left-censored by the limit of quantification and right-bounded by the HIV-1 RNA levels at baseline. We then suggest using nonparametric and parametric methods introduced for interval-censored data to analyze such data. A convenient feature of the methodology is the ability to easily handle missing HIV-1 RNA data, although some assumptions are required. For instance, the HIV-1 RNA reduction can be estimated using the so-called "missing = failure" scenario. Graphic procedures to check the validity of using parametric methods are described. The methods are discussed and illustrated with data of 2 recent clinical trials. Surprisingly, it was found that the log10 transformation of the HIV-1 RNA reduction was not appropriate in our data.