Semiparametric modeling of labeled-cell kinetics, with application to isotope labeling of erythrocytes.

We propose a stochastic model for the kinetics of cells that have been tagged with a chemical label. The proposed model consists of two components: a parametrically specified distribution for the time to incorporation of the label into the cells and a nonparametric survival function reflecting the survival time of the label-cell combination. The target quantity of this modeling approach is the fraction of labeled cells among all cells, viewed as a function of time. Longitudinal measurements of this labeled-cell fraction are available from a recent experiment with folate-labeled red blood cells. The proposed semiparametric model is fitted to these data and some of the implications are explored. The proposed method also includes bootstrap-based inference.