correct value is used. Errors in the estimate of p are inherent due to biological and analytical variability, but their impact can be reduced drastically by expanding the sampling schedule to longer times. Since this is not always possible because of analytical and practical limitations, the use of statistical moments may not be appropriate if there is insufficient sampling to provide accurate and consistent estimates of p.
Routine analytical error in drug concentration data can affect estimates of MAT and MDT in two ways: first, by its effects on the calculation of AUMCf, and AUCS, which should be rather insignificant, and second, by its effects on the estimate of p, which in turn may affect the calculation of AUMC; and AUC; significantly. The first case is illustrated in Table I11 by the values of MDT generated using the theoretical value of p (i.e. 0.1 hr-l) for all extrapolations. The average ( n = 18) estimated MDT values were good estimates of the theoretical values with coefficients of variation ranging from 23 to 32%. The effect of variations in ,6 caused by random error and the subsequent impact on MDTC is also shown in Table 111. The average MDT" values were again good estimates of the theoretical values but the coefficients of variation were quite large, ranging from 47 to 81%. With this sampling schedule -8% of AUC; and 31% of AUMC; were due to extrapolation, and this would not be uncommon for a typical bioavailability study. The large variability in these MDT values could affect the ability to statistically detect small differences in MDT values under such conditions.
Overall, the results of these studies suggest that the statistical moment approach to the analysis of bioavail-(1) S. Riegelman and P. Collier, J. Pharmacokinet. Eiopharm., 8,509 (2) K. Yamaoka, T. Nakagana, and T. Uno, ibid., 6,547 (1978). (3) "CSSL IV Users Guide and Reference Manual," Nielsen Asso-(1980). ciates, 1976.