Statistical constraints on microvascular measurements using fluorescent erythrocytes.

We show that two separate estimators of microvessel hematocrit (H) using the same fluorescent cell technique are not equally precise. The precision of H (indicated by the coefficient of variation, CV) depends on in vivo labeled cell counts (m), the labeled fraction (p), and for time-averaged H estimates, the mean cell velocity (v). Thus, for length-dependent estimates of H, CV2(HL) = 1/E(m) + CV2(p) and for time-averaged H estimates, CV2(Ht) = 1/E(m) + CV2(p) + CV2(v). Because CV(p) can be made small arbitrarily and independently, the precision of H is principally dependent on the expected value of m. Practical sampling constraints and minimum sampling intervals are identified and used to define strategies to minimize CV(H). We show that Ht is preferable to HL because it is more precise and the useful in vivo sampling range of m and p is more flexible. In addition, Ht allows simultaneous determination of cell flux and mean cell velocity.