RE: bug list

Leonid: Here is the the bug list of 7.1.0 that were fixed in nm 7.1.2. The particular bug you mentioned is item 5, and has been fixed in nm 7.1.2. Several bugs have been found in nm 7.1.2, which have appeared in the last couple of months. As they appear, we have corrected them for the next major NONMEM release (8.0) scheduled in January. Some of these are: 1) With very large problems of more than to 180 estimated parameters (thetas, omegas, and sigmas), the eigenvalues list with two sets of column labels. 2) When the number of records in a subject exceeds 250, a "stack overflow" in the intel version of NONMEM may occur. 3) On occasion after an analysis with SAEM with a very complex problem, estimation of objective function with IMP or IMPMAP results in ever increasing objective function values without stabilization, even though the SAEM result is reasonable. The usual adjustment of options in nm 7.1.2 fails to correct the problem. In NONMEM 8.0, some internal scaling parameters have been adjusted. Also, the user can further adjust these scaling parameters. 4) For certain estimation problems, ADVAN 5 and ADVAN7 provide inaccurate prediction values, which are sensitive to the initial thetas. The workaround is to use ADVAN6 or ADVAN9. ******************************************************* NONMEM 7.1.0 bugs that were fixed in NONMEM 7.1.2 ________________________________________________________________________ NONMEM 1. There is a bug in NONMEM VI 1.x & 2.0 and NONMEM 7.1.0 that results in the spurious error message below, or a similar message, when certain band symmetric matrices are defined in $OMEGA. PROGRAM TERMINATED BY OBJ OMEGA ESTIMATED TO BE SINGULAR MESSAGE ISSUED FROM ESTIMATION STEP AT INITIAL OBJ. FUNCTION EVALUATION Workaround: None is available. 2. When single subject data were analyzed, NONMEM VI would impose the use of the SLOW gradient feature, without this option appearing explicitly on the $ESTIMATION record. In NONMEM 7.1.0, the SLOW option is not imposed unless it appears on the $ESTIMATION record. Small differences in results may occur for some single subject analyses. Workaround: Include the SLOW option on the $ESTIMATION record. 3. In NONMEM 7.1.0, whenever MAXEVAL=0 appears on the $ESTIMATION record, the table number in the raw output files is 0. Workaround: Use MAXEVAL=1 on the $ESTIMATION record if it is important to correct the table number. 4. If the SAEM or MCMC method is the first method used in a problem and the INTERACTION option is not specified (even with homoscedastic residual error), the model initialization will not be complete and will result in a failed analysis. Workaround: Always specify INTERACTION for the new methods. This option does not impact the analysis time. 5. In NONMEM 7.1.0, if a problem is run, completed and saved in an msf file and then later the msf file is used to implement the Covariance Step without an Estimation Step, the results would not be consistent with those obtained when both Estimation and Covariance Steps were implemented in the same control stream. Workaround: Add an Estimation Step to the control stream using the msf file to implement the Covariance Step, and specify the same method that was used when the msf was created. E.g., $MSFI myfile.msf $EST METHOD=0 MAXEVAL=0 $COV or $MSFI myfile.msf $EST METHOD=0 MAXEVAL=1 $COV 6. For METHOD=MCMC the inverse covariance matrix of the estimates is not calculated. The inverse covariance matrix associated with the MCMC method that is displayed in the <root>.coi file should be ignored. 7. When the Simulation, Estimation and Covariance steps are performed in the same problem and there are multiple subproblems then the gradients for the second and subsequent subproblems are slightly affected by the failure to initialize a variable in the covariance step. 8. The #METH tag may appear in the output for the Simulation step. 9. A segmentation fault may occur when using Intel Fortran on Linux. Workaround: Change the value of NO in SIZES to be at least as large as the maximum number of records for any given individual. _________________________________________________________ PREDPP 1. There is a PREDPP bug in pr/PRED.for (NONMEM VI 2.0) and pr/PRED.f90 (NONMEM 7.1.0) that affects the behavior of the steady-state (SS) data item when: 1) SS has a value of 2 and 2) the record with SS=2 precedes all other records with a non-zero SS data item in an individual record or after a reset record, EVID=3 or 4. The bug causes the steady state dose to be applied twice. Workaround: 1) If SS=2 is the first record of the individual record, then a work around is to specify the AMT value on the affected record to be one-half the intended dose. Or, the $PK block could be modified instead. Suppose there is only one steady state dose record having SS=2, and the dose is for compartment 1. Add: $PK F1=1 IF (SS.EQ.2) F1=.5 2) If SS=2 is used in a subsequent record (with a change in TIME) and it is not preceded by a record with a non-zero value for SS then there is no workaround. This situation occurs when a transient dose precedes the steady state dose. Suggestion: The behavior of the steady-state data item, in the absence of the bug, is relatively complex. A user planning to incorporate the SS data item into a data set should review Section V.F in NONMEM Users Guide VI, PREDPP, and Chapter 6, Section 8.2.3 of Guide V, Introductory Guide. ___________________________________________________________ NMTRAN 1. In some cases the $COVARIANCE record may be incorrectly parsed resulting in an NM-TRAN error message. Workarounds: 1) Place the $COVARIANCE record prior to the $ESTIMATION record in the control stream. Or 2) Include the option COND or UNCOND on the $COVARIANCE record. Robert J. Bauer, Ph.D. Vice President, Pharmacometrics ICON Development Solutions Tel: (215) 616-6428 Mob: (925) 286-0769 Email: [email protected] Web: www.icondevsolutions.com