NONMEM Users Guide Part VI - PREDPP
Preface. to 3rd Edition

NONMEM Users Guide Part VI - PREDPP

Preface. to 3rd Edition

The appearance of this 3rd edition of the PREDPP Guide coincides with the appearance of PREDPP Version III Level 1. The new features in Version III are:

There are new additional PK parameters: absorption lag times, one for each compartment. An absorption lag time for a compartment into which a dose is given is the difference between the time PREDPP actually enters the dose into the compartment and the event time at which the dose is administered. Absorption lag times can be modeled with ’s. Doses with positive lag times are called lagged doses

Section III.F.6

PK and ERROR can be called with every event record, simply by setting an element of IDEF to an appropriate value, and this is now the default. This change can cause the output from Version III to differ from that from Version II. The calling-protocols of Version II only allow more limited sequences of calls to PK and ERROR (without also using CALL data items). These protocols are still implementable. [With Version II ERROR is called during the Simulation Step with exactly the same records as during non-Simulation steps. With Version III ERROR is called during the Simulation Step with every record, regardless of the particular calling-protocol implemented in non-Simulation steps.]

Sections III.H and IV.C

PK can be called with every event record, or in other words, at every event time, and also at any time an additional dose or lagged dose enters the system, simply by setting an element of IDEF to an appropriate value. With such a call PK has access to information in the event record following this time and to information in the dose event record specifying the dose. It also has access to the time in question.

Sections III.B.2, III.H, III.I

There is an improved format for the IDEF array in PK. This format must be used if absorption lag times are used. If absorption lag times are not used, the Version I-II IDEF format can be used.

Section III.G

Bioavailability fractions apply to all types of doses (except continuous steady-state infusions). With Version II they do not apply to infusions. [If both infusions and other kinds of doses are present in a data set, and a bioavailability fraction is used, the output from Version II can differ from that from Version III because with Version II the fraction does not apply to the infusions, whereas with Version III it does.]

Section III.F.2

Multiple infusions specified on a steady-state dose record may have a duration that exceeds the (size of the) interdose interval. With Versions II the duration cannot exceed the interdose interval.

The rate of a steady-state infusion can be modeled.

Sections III.F.5, V.E

PK, ERROR, and INFN have access to a NEWIND variable that functions analogously to the NEWIND variable accessible to a PRED routine.

Sections III.I, VI.A

The algorithm used in ADVAN5 to compute the matrix exponential has been improved and is faster.

10.

The algorithm with which ADVAN6, ADVAN8, and ADVAN9 handle a zero-order bolus dose whose duration depends on an has been greatly improved. However, if a rate parameter, duration parameter, or absorption lag time parameter is modeled with an variable, then the right sides of the differential equations should depend on time only through their dependence on compartment amounts, i.e. time should not appear explicitly in these equations. With ADVAN9, time may appear explicitly in the algebraic equations, as long as the differential equations do not depend on the equillibrium compartment amounts. These restrictions (with respect to modeled rate and duration parameters) are new to Version III. Version II can be used without these restrictions. Hopefully, a future release of PREDPP will also not include them.

11.

The algorithm used in ADVAN10 to compute Michaelis-Menten kinetics has been improved.

12.

Conditional estimation methods are supported. The Laplacian method of estimation, however, is not supported with ADVAN routines ADVAN5-9, and is not supported with any SS routines except SS1.

Sections III.D, III.E

13.

PRED error-recovery is supported.

Sections III.K, IV.F

14.

Data transgeneration during the Simulation Step is supported.

Section III.L.1

15.

Amounts from all compartments (and their -derivatives) are available to the ERROR routine. This allows easier modeling of pharmacodynamics and in general, decreases the need to use PCMT data items.

Section IV.D

This edition of the PREDPP Guide represents a substantial rewriting of the earlier edition. It contains more information than the earlier edition, both about new features and features that have existed in Version II. Nonetheless, an attempt has been made to simplify various discussions. It contains more cross-referencing. An example involving a pharmacodynamic model has been added, and another involving a mixture model. Both NONMEM and NM-TRAN control streams for all examples are shown.

Algorithms and codes for computing the matrix exponential, which are used in certain ADVAN and SS routines implementing general linear compartmental models, were developed by Professor Beresford N. Parlett and Drs. Kwok Choi Ng and Jing Li of the Center for Pure and Applied Mathematics at the University of California at Berkeley.

Codes for computing the solution to a system of differential equations, which are used in certain ADVAN and SS routines implementing general nonlinear compartmental models, as well as codes for other purposes, were developed by IMSL, Inc. in Houston, Texas, and are used under an agreement with IMSL.

A code for computing the solution to a system of differential-algebraic equations, which is used in the ADVAN and SS routines implementing the general nonlinear compartmental model with equilibrium compartments, was developed by Drs. Alan Hindmarsh and Jeffrey F. Painter of the Computing and Mathematics Research Division of Lawrence Livermore Laboratory, and modified by the NONMEM Project Group.

The concepts explained in this document are illustrated in the body of the document with two sets of data: The first is a set of population phenobarbitol data, kindly supplied by Professor Ted Grasela and discussed in more detail in:

Grasela, T. H. and Donn, S. M. (1985). Neonatal population pharmacokinetics of phenobarbitol derived from routine clinical data. Dev. Pharmacol. Ther.: 8, 374-383.

The second is a set of theophylline data from a single subject, which is also used in Guide I, section C, but without using PREDPP. In addition, Appendix III illustrates how PREDPP can be used to specify the same analysis of the full set of population theophylline data as that given in Guide I, section F. In this 3rd addition this illustration includes alternative approaches, using ADVAN’s 2, 6, and 7. NM-TRAN control streams, as well as NONMEM control streams, are listed. The theophylline data is kindly supplied by Profs. Sidney Riegelman and Robert Upton.