Extreme values of CWRESI in NONMEM 7.4.1?

Dear all, I am reporting about a suspicious behaviour I am observing with CWRESI in NONMEM 7.4.1. I have a patent/metabolite model which fits the data decently in terms of VPCs and other diagnostics, but the CWRESI have some EXTREME values (e.g. some are less than -15 and more than +40). After checking for outliers or mistakes in the dataset and finding no obvious reason for this, I realised something funny was happening. For all those patients with extreme CWRESI values, all the CWRESI were exactly the same as WRES. I noticed that this happened for other subjects as well, without extreme values. In fact it was observed in a number of consecutive patients in the dataset, then the CWRESI and WRES were different once again for a number of IDs, then different again. I pasted below the beginning of my output table (I hope it won't get horribly mangled in the email). You can see for ID 1002 things are fine, but for the next subject, all CWRESI and WRES are 100% the same, which I find very odd. Going on in the table, one then encounters the crazy values (e.g. +40) in those patients where CWRESI=WRES. After thinking that I had messed up something in the code (although I have a fairly standard $ERR chunk), I reran the same models with NM 7.3 and no extreme CWRESI were to be seen and in no patient CWRESI=WRES. I tried ifortran and gfortran, same results. Not sure if this is a bug or something has changed in the implementation of CWRESI in the new version? Anything that needs to be added into the code? Did anyone else experience something similar? Thanks for sharing your experience, Paolo ID TIME Y DV PRED WRES IPRED IWRES CWRESI 1002 0 1.11E-07 0 1.11E-07 0 1.11E-07 -3.9E-06 0 1002 24 0.003244 0 0.000414 0 0.003244 -0.11491 0 1002 47.17 0.004529 0.0234 0.000633 0.80448 0.004529 0.66837 0.7499 1002 47.17 0.001072 0.00782 8.13E-05 0.57553 0.001072 0.8618 0.76168 1002 47.27 0.004351 0 0.000601 0 0.004351 -0.15409 0 1002 47.27 0.004351 0 0.000601 0 0.004351 -0.15409 0 1002 48.27 0.94518 0.967 4.5118 -0.95261 0.94518 0.10129 -1.1938 1002 48.27 0.077732 0.0657 0.33279 0.53181 0.077732 -0.40398 0.55325 1002 49.27 2.7804 2.58 6.0306 -0.4145 2.7804 -0.3187 -0.798 1002 49.27 0.35996 0.339 0.65388 0.57726 0.35996 -0.15724 0.86232 1002 51.25 3.1544 3.09 3.9118 0.63922 3.1544 -0.09023 -0.068 1002 51.25 0.67321 0.814 0.64312 1.0183 0.67321 0.56538 1.0395 1002 53.25 2.2343 2.48 2.0147 1.1994 2.2343 0.48601 0.69346 1002 53.25 0.61916 0.821 0.37813 1.2318 0.61916 0.88123 0.60795 1002 55.23 1.3392 1.56 0.91675 1.0769 1.3392 0.72666 0.70831 1002 55.23 0.41051 0.485 0.16968 1.1534 0.41051 0.49014 -0.11759 1003 0 1.17E-07 0 1.17E-07 0 1.17E-07 -4.1E-06 0 1003 23.92 2.19E-05 0 0.000106 0 2.19E-05 -0.00078 0 1003 46.58 4.65E-05 0.0234 0.000208 0.82167 4.65E-05 0.82766 0.82167 1003 46.58 8.66E-06 0.00782 2.55E-05 0.56411 8.66E-06 0.99888 0.56411 1003 46.63 4.52E-05 0 0.000202 0 4.52E-05 -0.0016 0 1003 46.63 4.52E-05 0 0.000202 0 4.52E-05 -0.0016 0 1003 47.68 1.1432 1.09 5.096 -0.99314 1.1432 -0.20478 -0.99314 1003 47.68 0.065831 0.053 0.39584 0.43182 0.065831 -0.50191 0.43182 1003 48.67 5.38 6.44 6.427 0.93526 5.38 0.87171 0.93526 1003 48.67 0.57361 0.907 0.7257 0.60985 0.57361 1.571 0.60985 1003 50.65 3.6587 3.71 4.0526 -0.30631 3.6587 0.061999 -0.30631 1003 50.65 0.65408 0.662 0.68469 -0.03219 0.65408 0.032746 -0.03219 1003 52.65 1.5856 1.82 2.0167 -0.16771 1.5856 0.65208 -0.16771 1003 52.65 0.33653 0.402 0.38292 0.080183 0.33653 0.5252 0.080183 1003 54.65 0.58013 0.594 0.86702 -0.41124 0.58013 0.10342 -0.41124 1003 54.65 0.12566 0.12 0.16026 -0.07836 0.12566 -0.12016 -0.07836 1004 0 1.69E-07 0 1.69E-07 0 1.69E-07 -6E-06 0 1004 23.83 0.000114 0 0.000285 0 0.000114 -0.00404 0 1004 46.33 0.000242 0.0234 0.000576 0.8068 0.000242 0.82074 0.8068 1004 46.33 3.76E-05 0.00782 7.31E-05 0.57412 3.76E-05 0.99518 0.57412 1004 46.4 0.000233 0 0.000555 0 0.000233 -0.00824 0 1004 46.4 0.000233 0 0.000555 0 0.000233 -0.00824 0 1004 47.38 4.3404 4.59 3.1408 0.54238 4.3404 0.25445 0.54238 1004 47.38 0.34788 0.295 0.20459 -0.38164 0.34788 -0.41035 -0.38164 1004 48.38 5.5978 5.57 4.1733 0.22522 5.5978 -0.02199 0.22522 1004 48.38 0.69068 0.669 0.4033 0.58087 0.69068 -0.08485 0.58087 1004 50.38 3.1389 3.62 2.4749 0.42327 3.1389 0.67784 0.42327 1004 50.38 0.59966 0.721 0.3565 0.7287 0.59966 0.54698 0.7287 1004 52.38 1.4082 1.55 1.1339 -0.29857 1.4082 0.44389 -0.29857 1004 52.38 0.30079 0.434 0.18172 1.7245 0.30079 1.1949 1.7245 1004 54.37 0.54719 0.555 0.45719 -0.3361 0.54719 0.061603 -0.3361 1004 54.37 0.11383 0.114 0.071932 0.039771 0.11383 0.003873 0.039771 -- ------------------------------------------------ Paolo Denti, PhD Pharmacometrics Group Division of Clinical Pharmacology Department of Medicine University of Cape Town K45 Old Main Building Groote Schuur Hospital Observatory, Cape Town 7925 South Africa phone: +27 21 404 7719 fax: +27 21 448 1989 email: [email protected]<mailto:[email protected]> ------------------------------------------------ Disclaimer - University of Cape Town This e-mail is subject to UCT policies and e-mail disclaimer published on our website at http://www.uct.ac.za/about/policies/emaildisclaimer/ or obtainable from +27 21 650 9111. If this e-mail is not related to the business of UCT, it is sent by the sender in an individual capacity. Please report security incidents or abuse via [email protected]

Paolo: I reviewed my NONMEM 7.4 testing results on CWRESI, and they provide the same values in NONMEM 7.3 as well as NONMEM 7.4. In case you ran the problem in parallel, because NONMEM 7.4 now executes the diagnostic evaluation in parallel as well, I tested under parallel execution conditions, and the CWRESI values are also the same between NONMEM 7.4 (parallel) and NONMEM 7.3. Incidentally, as a diagnostic that is expected to be normally distributed when the model is correct, CWRES is preferred even when INTERACTION is used in the $EST step. I would be happy to run your problem and see if I can detect something more subtle that is involved in your case. Robert J. Bauer, Ph.D. Senior Director Pharmacometrics R&D ICON Early Phase 820 W. Diamond Avenue Suite 100 Gaithersburg, MD 20878 Office: (215) 616-6428 Mobile: (925) 286-0769 [email protected]<mailto:[email protected]> http://www.iconplc.com/

Hello all: After working with Paolo Denti and Sven Stodtman, it has been determined that the occasional spurious CWRES and CWRESI values occurs during parallelization of the diagnostic evaluation step. The best work-around is to set the PARAFILE=OFF option on the first $TABLE record: $TABLE … PARAFILE=OFF This will allow $EST and $COV step to continue to be parallelized. Single CPU runs will of course not have this issue. Robert J. Bauer, Ph.D. Senior Director Pharmacometrics R&D ICON Early Phase 820 W. Diamond Avenue Suite 100 Gaithersburg, MD 20878 Office: (215) 616-6428 Mobile: (925) 286-0769 [email protected]<mailto:[email protected]> http://www.iconplc.com/