Erratum to: J Pharmacokinet Pharmacodyn (2008) 35(4):401–421 10.1007/s10928-008-9094-4

The following list of corrections applies to the article, “Likelihood based approaches to handling data below the quantification limit using NONMEM VI” by Jae Eun Ahn, Mats O. Karlsson, Adrian Dunne, Thomas M. Ludden. The article appeared in the Journal of Pharmacokinetics and Pharmacodynamics, Volume 35, Number 4, August 2008, pp. 401–421.

Page 402:

  • Method 7 (M7): Replace all BQL observations with 0 and apply extended least squares estimation.

Page 419, Appendix Part I.b, M3 Code:

DUM = (LOQ − IPRED)/(SIG*IPRED)

Page 419, Appendix Part I.b, M4 Code:

DUM = (LOQ − IPRED)/(SIG*IPRED)

DUM0 = (0 − IPRED)/(SIG*IPRED)

Page 419, Appendix Part I.b, $ERROR:

Adding “CALLFL=0” or “(ONLY OBSERVATIONS)” is necessary to prevent division by zero for dosing records.

In Part I, the negatively biased estimates for RUV with all data, M1, and M2 were due to the fact that an incorrect true value (0.025) was used in calculating estimation error. The value should have been 0.0225. The corrected values for RUV assessment are provided in Table 4-7c.

Table 4-7c Part I.a: Corrections to Mean Estimation Error (MEE, %) and Root Mean Squared Estimation Error (RMSE, %) of RUV
Full size table

Estimation in Part I.b was redone with the correct code above but using a different Fortran compiler, g77. Simulated data are the same. The new results are provided in Tables 8c–10c. M2 tended to be less stable than M3 or M4 at high LOQ values (Table 8c) but still gave less biased estimates than M1. M1 was also fitted to the untransformed data but its estimates were still severely biased as in Part I.a and the results are not presented. M3 and M4 gave very similar and the least biased estimates (Table 9c). Overall precision seems comparable (Table 10c).

Table 8c Part I.b: The Proportion of BQL Data and the Success Rate (%) (Extensive Design: 24 samples/individual; untransformed data)
Full size table
Table 9c Part I.b: Mean Estimation Error (%) of Parameter Estimates (Extensive Design, Untransformed Data, Successful Termination Only)
Full size table
Table 10c Part I.b: Root Mean Squared Estimation Error (%) of Parameter Estimates (Extensive Design, Untransformed Data, Successful Termination Only)
Full size table