RE: [NMusers] Parameter uncertainty
Hi Fanny, Marc I was thinking in the same direction as Marc. If you use MCMC (BAYES method in NONMEM) the algorithm will provide you with samples from the posterior density (posterior = likelihood * prior). From these samples you can then investigate different statistics, for example variance of your parameters. Be caution about convergence of the algorithm, since these algorithms are not guaranteed to sample uncorrelated samples. On the same topic, are there any good comparisons out there comparing the standard covariance matrix approach, bootstrap, profiling and MCMC? /Jacob From: owner-nmus...@globomaxnm.com [mailto:owner-nmus...@globomaxnm.com] On Behalf Of Marc Gastonguay Sent: den 16 februari 2017 13:23 To: Fanny Gallais Cc: Williams, Jason ; nmusers@globomaxnm.com Subject: Re: [NMusers] Parameter uncertainty Dear Fanny, One additional method to obtain the parameter uncertainty, which I don't believe was mentioned, is Bayesian estimation using Markov-Chain Monte Carlo (MCMC) simulation. This method provides a full joint posterior distribution (e.g. uncertainty distribution) of the parameters and any predicted quantities, and is really the gold standard for this type of goal. It is possible to implement this method in NONMEM (with some limitations on the prior distributions), or you could use BUGS or Stan with associated PK model libraries. You can also extract the samples from the posterior distribution and simulate using the methods already described in this thread. Marc On Thu, Feb 16, 2017 at 6:01 AM, Fanny Gallais mailto:gallais.fa...@gmail.com>> wrote: Thank you all for your responses. It is going to be very useful for my work. Best regards, F.G. 2017-02-15 17:35 GMT+01:00 Williams, Jason mailto:jason.willi...@pfizer.com>>: Dear Fanny, Another useful tool you may want to try is using the mrgsolve package available in R, developed by Kyle Baron at Metrum Research Group. I have found mrgsolve to be very efficient for PKPD simulation and sensitivity analysis in R. There is an example of incorporating parameter uncertainty (from $COV step in NONMEM) in Section 9 of the example on Probability of Technical Success (link below). https://github.com/mrgsolve/examples/blob/master/PrTS/pts.pdf Best regards, Jason From: owner-nmus...@globomaxnm.com<mailto:owner-nmus...@globomaxnm.com> [mailto:owner-nmus...@globomaxnm.com<mailto:owner-nmus...@globomaxnm.com>] On Behalf Of Fanny Gallais Sent: Wednesday, February 15, 2017 2:55 AM To: nmusers@globomaxnm.com<mailto:nmusers@globomaxnm.com> Subject: [NMusers] Parameter uncertainty Dear NM users, I would like to perform a simulation (on R) incorporating parameter uncertainty. For now I'm working on a simple PK model. Parameters were estimated with NONMEM. I'm trying to figure out what is the best way to assess parameter uncertainty. I've read about using the standard errors reported by NONMEM and assume a normal distribution. The main problem is this can lead to negative values. Another approach would be a more computational non-parametric method like bootstrap. Do you know other methods to assess parameter uncertainty? Best regards F. Gallais -- Marc R. Gastonguay, Ph.D.<mailto:ma...@metrumrg.com> CEO Metrum Research Group LLC<http://metrumrg.com> 2 Tunxis Rd., Ste 112, Tariffville, CT 06081 USA Tel: +1.860.735.7043 ext. 101, Mobile: +1.860.670.0744, Fax: +1.860.760.6014 Confidentiality Notice: This message is private and may contain confidential and proprietary information. If you have received this message in error, please notify us and remove it from your system and note that you must not copy, distribute or take any action in reliance on it. Any unauthorized use or disclosure of the contents of this message is not permitted and may be unlawful.
Re: [NMusers] Parameter uncertainty
Dear Fanny, One additional method to obtain the parameter uncertainty, which I don't believe was mentioned, is Bayesian estimation using Markov-Chain Monte Carlo (MCMC) simulation. This method provides a full joint posterior distribution (e.g. uncertainty distribution) of the parameters and any predicted quantities, and is really the gold standard for this type of goal. It is possible to implement this method in NONMEM (with some limitations on the prior distributions), or you could use BUGS or Stan with associated PK model libraries. You can also extract the samples from the posterior distribution and simulate using the methods already described in this thread. Marc On Thu, Feb 16, 2017 at 6:01 AM, Fanny Gallais wrote: > Thank you all for your responses. It is going to be very useful for my > work. > > Best regards, > > F.G. > > 2017-02-15 17:35 GMT+01:00 Williams, Jason : > >> Dear Fanny, >> >> >> >> Another useful tool you may want to try is using the mrgsolve package >> available in R, developed by Kyle Baron at Metrum Research Group. I have >> found mrgsolve to be very efficient for PKPD simulation and sensitivity >> analysis in R. There is an example of incorporating parameter uncertainty >> (from $COV step in NONMEM) in Section 9 of the example on Probability of >> Technical Success (link below). >> >> >> >> https://github.com/mrgsolve/examples/blob/master/PrTS/pts.pdf >> >> >> >> Best regards, >> >> >> Jason >> >> >> >> *From:* owner-nmus...@globomaxnm.com [mailto:owner-nmus...@globomaxnm.com] >> *On Behalf Of *Fanny Gallais >> *Sent:* Wednesday, February 15, 2017 2:55 AM >> *To:* nmusers@globomaxnm.com >> *Subject:* [NMusers] Parameter uncertainty >> >> >> >> Dear NM users, >> >> >> >> I would like to perform a simulation (on R) incorporating parameter >> uncertainty. For now I'm working on a simple PK model. Parameters were >> estimated with NONMEM. I'm trying to figure out what is the best way to >> assess parameter uncertainty. I've read about using the standard errors >> reported by NONMEM and assume a normal distribution. The main problem is >> this can lead to negative values. Another approach would be a more >> computational non-parametric method like bootstrap. Do you know other >> methods to assess parameter uncertainty? >> >> >> >> >> >> Best regards >> >> >> >> F. Gallais >> >> >> >> >> >> >> > > -- Marc R. Gastonguay, Ph.D. CEO Metrum Research Group LLC <http://metrumrg.com> 2 Tunxis Rd., Ste 112, Tariffville, CT 06081 USA Tel: +1.860.735.7043 ext. 101, Mobile: +1.860.670.0744, Fax: +1.860.760.6014
Re: [NMusers] Parameter uncertainty
Thank you all for your responses. It is going to be very useful for my work. Best regards, F.G. 2017-02-15 17:35 GMT+01:00 Williams, Jason : > Dear Fanny, > > > > Another useful tool you may want to try is using the mrgsolve package > available in R, developed by Kyle Baron at Metrum Research Group. I have > found mrgsolve to be very efficient for PKPD simulation and sensitivity > analysis in R. There is an example of incorporating parameter uncertainty > (from $COV step in NONMEM) in Section 9 of the example on Probability of > Technical Success (link below). > > > > https://github.com/mrgsolve/examples/blob/master/PrTS/pts.pdf > > > > Best regards, > > > Jason > > > > *From:* owner-nmus...@globomaxnm.com [mailto:owner-nmus...@globomaxnm.com] > *On Behalf Of *Fanny Gallais > *Sent:* Wednesday, February 15, 2017 2:55 AM > *To:* nmusers@globomaxnm.com > *Subject:* [NMusers] Parameter uncertainty > > > > Dear NM users, > > > > I would like to perform a simulation (on R) incorporating parameter > uncertainty. For now I'm working on a simple PK model. Parameters were > estimated with NONMEM. I'm trying to figure out what is the best way to > assess parameter uncertainty. I've read about using the standard errors > reported by NONMEM and assume a normal distribution. The main problem is > this can lead to negative values. Another approach would be a more > computational non-parametric method like bootstrap. Do you know other > methods to assess parameter uncertainty? > > > > > > Best regards > > > > F. Gallais > > > > > > >
RE: [NMusers] Parameter uncertainty
Dear Fanny, Another useful tool you may want to try is using the mrgsolve package available in R, developed by Kyle Baron at Metrum Research Group. I have found mrgsolve to be very efficient for PKPD simulation and sensitivity analysis in R. There is an example of incorporating parameter uncertainty (from $COV step in NONMEM) in Section 9 of the example on Probability of Technical Success (link below). https://github.com/mrgsolve/examples/blob/master/PrTS/pts.pdf Best regards, Jason From: owner-nmus...@globomaxnm.com [mailto:owner-nmus...@globomaxnm.com] On Behalf Of Fanny Gallais Sent: Wednesday, February 15, 2017 2:55 AM To: nmusers@globomaxnm.com Subject: [NMusers] Parameter uncertainty Dear NM users, I would like to perform a simulation (on R) incorporating parameter uncertainty. For now I'm working on a simple PK model. Parameters were estimated with NONMEM. I'm trying to figure out what is the best way to assess parameter uncertainty. I've read about using the standard errors reported by NONMEM and assume a normal distribution. The main problem is this can lead to negative values. Another approach would be a more computational non-parametric method like bootstrap. Do you know other methods to assess parameter uncertainty? Best regards F. Gallais
Re: FW: [NMusers] Parameter uncertainty
One of the tools available for simulations is Metrum R package metrumrg install.packages("metrumrg", repos="http://R-Forge.R-project.org";) Example of applications can be found here: http://www.page-meeting.org/page/page2006/P2006III_11.pdf Since the time it was written (2005-2006), Nonmem enhanced the simulations options, so now you can simulate from the model-estimated uncertainty directly from Nonmem. The R package could be useful if you do it from the bootstrap results. Leonid *From:*owner-nmus...@globomaxnm.com [mailto:owner-nmus...@globomaxnm.com] *On Behalf Of *Fanny Gallais *Sent:* Wednesday, February 15, 2017 5:55 AM *To:* nmusers@globomaxnm.com *Subject:* [NMusers] Parameter uncertainty Dear NM users, I would like to perform a simulation (on R) incorporating parameter uncertainty. For now I'm working on a simple PK model. Parameters were estimated with NONMEM. I'm trying to figure out what is the best way to assess parameter uncertainty. I've read about using the standard errors reported by NONMEM and assume a normal distribution. The main problem is this can lead to negative values. Another approach would be a more computational non-parametric method like bootstrap. Do you know other methods to assess parameter uncertainty? Best regards F. Gallais *NOTICE: *The information contained in this electronic mail message is intended only for the personal and confidential use of the designated recipient(s) named above. This message may be an attorney-client communication, may be protected by the work product doctrine, and may be subject to a protective order. As such, this message is privileged and confidential. If the reader of this message is not the intended recipient or an agent responsible for delivering it to the intended recipient, you are hereby notified that you have received this message in error and that any review, dissemination, distribution, or copying of this message is strictly prohibited. If you have received this communication in error, please notify us immediately by telephone and e-mail and destroy any and all copies of this message in your possession (whether hard copies or electronically stored copies). Thank you. Personal data may be transferred to the United States of America and, if this occurs, it is possible that US governmental authorities may access such personal data. buSp9xeMeKEbrUze
RE: [NMusers] Parameter uncertainty
Dear Fanny and Bill, The sampling importance resampling (SIR) approach [1] to characterize the parameter uncertainty address the aspects pointed out Bill. In my opinion this is currently in general the most widely applicable and accurate method to characterize parameter uncertainty for NLMEM (bootstrap is likely approximately as good for large datasets and balanced designs). The method is implemented in PsN [2] and ready to use together with NONMEM. [1] Dosne A-G, Bergstrand M, Harling K, Karlsson MO. Improving the estimation of parameter uncertainty distributions in nonlinear mixed effects models using sampling importance resampling. J Pharmacokinet Pharmacodyn. 2016 Oct 11. http://link.springer.com/article/10.1007/s10928-016-9487-8 [2] SIR user guide, PsN 4.6.0: http://psn.sourceforge.net/pdfdocs/sir_userguide.pdf Best regards, Martin Bergstrand, Ph.D. Senior Consultant Pharmetheus AB +46(0)709 994 396 martin.bergstr...@pharmetheus.com www.pharmetheus.com +46(0)18 513 328 U-A Science Park, Dag Hammarskjölds v. 52b 752 37 Uppsala, Sweden *This communication is confidential and is only intended for the use of the individual or entity to which it is directed. It may contain information that is privileged and exempt from disclosure under applicable law. If you are not the intended recipient please notify us immediately. Please do not copy it or disclose its contents to any other person.* *From:* owner-nmus...@globomaxnm.com [mailto:owner-nmus...@globomaxnm.com] *On Behalf Of *William Denney *Sent:* Wednesday, February 15, 2017 1:01 PM *To:* Fanny Gallais *Cc:* nmusers@globomaxnm.com *Subject:* Re: [NMusers] Parameter uncertainty Hi Fanny, It is often good practice to fit parameters that must be positive on the log scale (by exponentiating them). That will ensure that when sampling from a normal distribution (and then exponentiating the sample) you will have a positive value. LLP was suggested, but it won't assess correlation between your parameters which is often important when running simulations. Bootstrap is another good alternative as has already been suggested. Thanks, Bill On Feb 15, 2017, at 5:55 AM, Fanny Gallais wrote: Dear NM users, I would like to perform a simulation (on R) incorporating parameter uncertainty. For now I'm working on a simple PK model. Parameters were estimated with NONMEM. I'm trying to figure out what is the best way to assess parameter uncertainty. I've read about using the standard errors reported by NONMEM and assume a normal distribution. The main problem is this can lead to negative values. Another approach would be a more computational non-parametric method like bootstrap. Do you know other methods to assess parameter uncertainty? Best regards F. Gallais
RE: [NMusers] Parameter uncertainty
Hi Fanny, As I understand it, you’re looking for ways to produce predictions according to your model taking into account parameter uncertainty. We’ve recently published on the importance of parameter uncertainty when considering probability of target attainment for antibiotic dosing regimens. (Colin et al. J Antimicrob Chemother (2016) 71 (9): 2502-2508) The online supplement to this paper holds an R-script which you can use to simulate (and calculate PTA, if relevant) taking into account parameter uncertainty. For this, the script uses the variance-covariance matrix that is produced by the $COV step in NONMEM. Of course other techniques which generate a var-cov matrix could be used as input for the script as well. Kind regards, Pieter Colin From: owner-nmus...@globomaxnm.com [mailto:owner-nmus...@globomaxnm.com] On Behalf Of Fanny Gallais Sent: woensdag 15 februari 2017 11:55 To: nmusers@globomaxnm.com Subject: [NMusers] Parameter uncertainty Dear NM users, I would like to perform a simulation (on R) incorporating parameter uncertainty. For now I'm working on a simple PK model. Parameters were estimated with NONMEM. I'm trying to figure out what is the best way to assess parameter uncertainty. I've read about using the standard errors reported by NONMEM and assume a normal distribution. The main problem is this can lead to negative values. Another approach would be a more computational non-parametric method like bootstrap. Do you know other methods to assess parameter uncertainty? Best regards F. Gallais
Re: [NMusers] Parameter uncertainty
Hi Fanny, It is often good practice to fit parameters that must be positive on the log scale (by exponentiating them). That will ensure that when sampling from a normal distribution (and then exponentiating the sample) you will have a positive value. LLP was suggested, but it won't assess correlation between your parameters which is often important when running simulations. Bootstrap is another good alternative as has already been suggested. Thanks, Bill > On Feb 15, 2017, at 5:55 AM, Fanny Gallais wrote: > > Dear NM users, > > I would like to perform a simulation (on R) incorporating parameter > uncertainty. For now I'm working on a simple PK model. Parameters were > estimated with NONMEM. I'm trying to figure out what is the best way to > assess parameter uncertainty. I've read about using the standard errors > reported by NONMEM and assume a normal distribution. The main problem is this > can lead to negative values. Another approach would be a more computational > non-parametric method like bootstrap. Do you know other methods to assess > parameter uncertainty? > > > Best regards > > F. Gallais > > > > >
RE: [NMusers] Parameter uncertainty
Dear Fanny, I would use either bootstrapping or likelihood profiling, both of them are implemented in PsN ('bootstrap' and 'llp'). Kind regards Max Taubert Von: owner-nmus...@globomaxnm.com [owner-nmus...@globomaxnm.com]" im Auftrag von "Fanny Gallais [gallais.fa...@gmail.com] Gesendet: Mittwoch, 15. Februar 2017 11:55 An: nmusers@globomaxnm.com Betreff: [NMusers] Parameter uncertainty Dear NM users, I would like to perform a simulation (on R) incorporating parameter uncertainty. For now I'm working on a simple PK model. Parameters were estimated with NONMEM. I'm trying to figure out what is the best way to assess parameter uncertainty. I've read about using the standard errors reported by NONMEM and assume a normal distribution. The main problem is this can lead to negative values. Another approach would be a more computational non-parametric method like bootstrap. Do you know other methods to assess parameter uncertainty? Best regards F. Gallais
RE: [NMusers] Parameter uncertainty
Hi Fanny, Likelihood profiles are very useful to asses parameter uncertainty. I am sure you find a tutorial somewhere how they work. A number of software packages automate the process quite a bit. They are usually much more computationally efficient than bootstrap. Warm regards, Douglas Eleveld From: owner-nmus...@globomaxnm.com [mailto:owner-nmus...@globomaxnm.com] On Behalf Of Fanny Gallais Sent: woensdag 15 februari 2017 11:55 To: nmusers@globomaxnm.com Subject: [NMusers] Parameter uncertainty Dear NM users, I would like to perform a simulation (on R) incorporating parameter uncertainty. For now I'm working on a simple PK model. Parameters were estimated with NONMEM. I'm trying to figure out what is the best way to assess parameter uncertainty. I've read about using the standard errors reported by NONMEM and assume a normal distribution. The main problem is this can lead to negative values. Another approach would be a more computational non-parametric method like bootstrap. Do you know other methods to assess parameter uncertainty? Best regards F. Gallais De inhoud van dit bericht is vertrouwelijk en alleen bestemd voor de geadresseerde(n). Anderen dan de geadresseerde(n) mogen geen gebruik maken van dit bericht, het niet openbaar maken of op enige wijze verspreiden of vermenigvuldigen. Het UMCG kan niet aansprakelijk gesteld worden voor een incomplete aankomst of vertraging van dit verzonden bericht. The contents of this message are confidential and only intended for the eyes of the addressee(s). Others than the addressee(s) are not allowed to use this message, to make it public or to distribute or multiply this message in any way. The UMCG cannot be held responsible for incomplete reception or delay of this transferred message.
[NMusers] Parameter uncertainty
Dear NM users, I would like to perform a simulation (on R) incorporating parameter uncertainty. For now I'm working on a simple PK model. Parameters were estimated with NONMEM. I'm trying to figure out what is the best way to assess parameter uncertainty. I've read about using the standard errors reported by NONMEM and assume a normal distribution. The main problem is this can lead to negative values. Another approach would be a more computational non-parametric method like bootstrap. Do you know other methods to assess parameter uncertainty? Best regards F. Gallais
AW: [NMusers] Parameter Uncertainty and Covariate effects
Dear Mats, Thanks a lot for your answer. Taking into account the correlation/covariance of the estimators for THETA(1)/THETA(2) certainly solves the problem for additive covariate effects with normally distributed errors in both parameter and effect (e.g. CL=TVCL+SEX_EFF**SEX). However, if we assume a multiplicative effect (CL=TVCL*SEX_EFF**SEX) we do not have an additive structure, thus it is not clear a priori if the covariance, which measures linear dependence, will take care of the problem properly. The expression can be transformed to make the relationship linear, but then we lose the normal distribution (CL = EXP(LOG(TVCL)+SEX*LOG(SEX_EFF)). In order to check whether accounting for correlation solves the dependency problem, I ran a small MATLAB script. I think it is not possible to attach pictures to mails to this list (correct me if I am wrong), so I will attach the code I used. From the results (which of course are only toy examples, real-life correlations may look different) it seems that at least in some cases, accounting for the correlation is not enough (see code below). But I still think your post gives the answer. If the model were designed in a way to additively combine the THETAs for parameter and effect, accounting for correlation will solve the problem. Essentially this is what mu-modeling forces one to do anyway when using EM methods (In this case, MU_1 = THETA(1) + SEX*THETA(2); CL=EXP(MU_1) ). So if one is consequently using mu-modelling and accounting for the correlation, one should be on the safe side. Kind Regards, Sven Here is the MATLAB code: Assumed 95%CIs for the parameters CL = [18.04 ; 21.96] SE = [0.504 ; 0.896] Assumed (positive/negative) correlation: 0.5 testmat = [1 .05; .05 .01]; U_gt = chol(testmat)'; testmat2 = [1 -.05; -.05 .01]; U_lt = chol(testmat2)'; U_0 = diag(diag(U_lt)); CL0 = 20; SE0 = 0.7; normals = randn(5,2)'; uncert_gt = U_gt*normals; uncert_lt = U_lt*normals; uncert_0 = U_0*normals; maleStats_gt = CL0+uncert_gt(1,:); femaleStats_gt = (CL0+uncert_gt(1,:)).*(SE0+uncert_gt(2,:)); maleStats_gt_X = (CL0*SE0+uncert_gt(1,:))./(SE0+uncert_gt(2,:)); %This is just a posteriori change of reference, to see what would really happen, I think that one would need to estimate both ways with a real dataset maleStats_lt = CL0+uncert_lt(1,:); femaleStats_lt = (CL0+uncert_lt(1,:)).*(SE0+uncert_lt(2,:)); maleStats_lt_X = (CL0*SE0+uncert_lt(1,:))./(SE0+uncert_lt(2,:)); maleStats_0 = CL0+uncert_0(1,:); femaleStats_0 = (CL0+uncert_0(1,:)).*(SE0+uncert_0(2,:)); maleStats_0_X = (CL0*SE0+uncert_0(1,:))./(SE0+uncert_0(2,:)); subplot(4,1,1); ksdensity(maleStats_gt); title('correlation>0'); hold on; ksdensity(femaleStats_gt); ksdensity(maleStats_gt_X); hold off; legend('male','female','male with female as reference') subplot(4,1,2); ksdensity(maleStats_lt); title('correlation<0'); hold on; ksdensity(femaleStats_lt); ksdensity(maleStats_lt_X); hold off; legend('male','female','male with female as reference') subplot(4,1,3); ksdensity(maleStats_0); title('correlation=0'); hold on; ksdensity(femaleStats_0); ksdensity(maleStats_0_X); hold off; legend('male','female','male with female as reference') subplot(4,1,4); ksdensity(femaleStats_gt); title('Female only'); hold on; ksdensity(femaleStats_lt); ksdensity(femaleStats_0); hold off; legend('>0','<0','=0') SVEN STODTMANN, PHD Pharmacometrician AbbVie Deutschland GmbH & Co KG Clinical Pharmacology and Pharmacometrics Knollstrasse 50 67065 Ludwigshafen am Rhein, Germany OFFICE +49 621-589-1940 EMAIL sven.stodtm...@abbvie.com abbvie.com Please note that any views or opinions presented in this email are solely those of the author and do not necessarily represent those of the Company. -Ursprüngliche Nachricht- Von: Mats Karlsson [mailto:mats.karls...@farmbio.uu.se] Gesendet: Tuesday, January 12, 2016 6:02 AM An: Stodtmann, Sven Cc: nmusers@globomaxnm.com Betreff: Re: [NMusers] Parameter Uncertainty and Covariate effects Dear Sven If you don't assume the covariance between THETA(1) and THETA(2) to be zero but use the estimated covariance value, you do let the data speak. A problem in this respect is that publications never give such values even if it of course is possible. With online access to model code and output (as with the DDMoRe reposito
Re: [NMusers] Parameter Uncertainty and Covariate effects
Dear Sven If you don't assume the covariance between THETA(1) and THETA(2) to be zero but use the estimated covariance value, you do let the data speak. A problem in this respect is that publications never give such values even if it of course is possible. With online access to model code and output (as with the DDMoRe repository (repository.ddmore.eu)) it will be more likely to find the information. Best regards, Mats Skickat från min iPhone > 11 jan 2016 kl. 14:28 skrev Stodtmann, Sven : > > Dear All, > > In order to account for uncertainty in estimated parameters when running a > simulation, a natural approach would be running multiple simulations for > different parameter vectors which are drawn from the (theoretical, > asymptotic) distribution of the estimator (i.e. normal with mean THETA and > covariance according to the NONMEMs $COR output for the THETAs). > This approach may in some cases (particularly, when there are a lot of > covariate effects estimated) lead to very broad parameter distributions, even > assigning some quite high probability of unphysiological values if one didn’t > have good quality data, strong priors or a very careful parametrization of > the model (e.g. transforming/bounding parameters, which requires/introduces > prior knowledge as well). > > Another problem connected with parameter uncertainty on covariate effects is > the following. Say we model > TVCL = THETA(1) > SEX_EFF = THETA(2) > CL = TVCL * SEX_EFF**SEX, (Eq. > 1) > where male is coded as SEX=0, female as SEX=1. > In this case, when using the above mentioned technique to account for > parameter uncertainty, the female population will have a more variable > (uncertain) PK, not just different one. If we phrase the problem differently, > using > CL = TVCL * SEX_EFF**(1-SEX) ,(Eq. 2) > The conclusion would be the other way around (i.e. male PK is more uncertain). > > One approach to deal with the second problem could be this: > In order to remove this (usually unjustified) assumption (the female > population having a less certain PK compared to the male), one could try to > model the same covariate effect as follows: > TVCL = THETA(1) > SQRT_SEX_EFF = THETA(2) > CL = TVCL * SQRT_SEX_EFF**SEX / SQRT_SEX_EFF**(1-SEX) > In this case TVCL would already include “half” of the effect (on the log > scale; the “new” TVCL would be TVCL*SQRT(SEX_EFF) in terms of the parameters > used in Eq.1). > With this approach, both sub-populations, male and female get “some part” of > the uncertainty effect. > Of course it would be even nicer to let the data decide which sub-population > gets how much uncertainty exactly instead of evenly splitting it. > > How do you deal with uncertainty in the estimates of covariate effects when > it comes to simulations/predictions? > > Kind Regards, > > SVEN STODTMANN, PHD > Pharmacometrician > > AbbVie Deutschland GmbH & Co KG > Clinical Pharmacology and Pharmacometrics > Knollstrasse 50 > 67065 Ludwigshafen am Rhein, Germany > OFFICE +49 621-589-1940 > EMAIL sven.stodtm...@abbvie.com > > abbvie.com > > > > > Sitz der Gesellschaft: Wiesbaden - Registergericht: AG Wiesbaden HRA 9790 > Persönlich haftende Gesellschafterin: AbbVie Komplementär GmbH > Sitz der persönlich haftenden Gesellschafterin: Wiesbaden - Registergericht: > AG Wiesbaden HRB 26371 > Geschäftsführer: Dr. Patrick Horber, Thomas Scheidmeir, Dr. Stefan Simianer, > William J. Chase > Vorsitzende des Aufsichtsrats: Dr. Azita Saleki-Gerhardt > > This communication may contain information that is proprietary, confidential, > or exempt from disclosure. If you are not the intended recipient, please note > that any other dissemination, distribution, use or copying of this > communication is strictly prohibited. Anyone who receives this message in > error should notify the sender immediately by telephone or by return e-mail > and delete it from his or her computer. > > Diese Kommunikation kann Informationen enthalten, die geheim, vertraulich > oder hinsichtlich der Offenlegung beschränkt sind. Wenn Sie nicht der > beabsichtigte Empfänger sind, nehmen Sie bitte zur Kenntnis, dass jede > Weitergabe, Verteilung, Verwendung oder Vervielfältigung dieser. > Kommunikation strikt untersagt ist. Jeder, der diese Nachricht fehlerhaft > erhält, sollte den Sender unverzüglich telefonisch oder durch Rücksendung der > E-Mail benachrichtigen und diese von seinem oder ihrem Computer löschen.
[NMusers] Parameter Uncertainty and Covariate effects
Dear All, In order to account for uncertainty in estimated parameters when running a simulation, a natural approach would be running multiple simulations for different parameter vectors which are drawn from the (theoretical, asymptotic) distribution of the estimator (i.e. normal with mean THETA and covariance according to the NONMEMs $COR output for the THETAs). This approach may in some cases (particularly, when there are a lot of covariate effects estimated) lead to very broad parameter distributions, even assigning some quite high probability of unphysiological values if one didn’t have good quality data, strong priors or a very careful parametrization of the model (e.g. transforming/bounding parameters, which requires/introduces prior knowledge as well). Another problem connected with parameter uncertainty on covariate effects is the following. Say we model TVCL = THETA(1) SEX_EFF = THETA(2) CL = TVCL * SEX_EFF**SEX, (Eq. 1) where male is coded as SEX=0, female as SEX=1. In this case, when using the above mentioned technique to account for parameter uncertainty, the female population will have a more variable (uncertain) PK, not just different one. If we phrase the problem differently, using CL = TVCL * SEX_EFF**(1-SEX) ,(Eq. 2) The conclusion would be the other way around (i.e. male PK is more uncertain). One approach to deal with the second problem could be this: In order to remove this (usually unjustified) assumption (the female population having a less certain PK compared to the male), one could try to model the same covariate effect as follows: TVCL = THETA(1) SQRT_SEX_EFF = THETA(2) CL = TVCL * SQRT_SEX_EFF**SEX / SQRT_SEX_EFF**(1-SEX) In this case TVCL would already include “half” of the effect (on the log scale; the “new” TVCL would be TVCL*SQRT(SEX_EFF) in terms of the parameters used in Eq.1). With this approach, both sub-populations, male and female get “some part” of the uncertainty effect. Of course it would be even nicer to let the data decide which sub-population gets how much uncertainty exactly instead of evenly splitting it. How do you deal with uncertainty in the estimates of covariate effects when it comes to simulations/predictions? Kind Regards, SVEN STODTMANN, PHD Pharmacometrician AbbVie Deutschland GmbH & Co KG Clinical Pharmacology and Pharmacometrics Knollstrasse 50 67065 Ludwigshafen am Rhein, Germany OFFICE +49 621-589-1940 EMAIL sven.stodtm...@abbvie.com abbvie.com Sitz der Gesellschaft: Wiesbaden - Registergericht: AG Wiesbaden HRA 9790 Persönlich haftende Gesellschafterin: AbbVie Komplementär GmbH Sitz der persönlich haftenden Gesellschafterin: Wiesbaden - Registergericht: AG Wiesbaden HRB 26371 Geschäftsführer: Dr. Patrick Horber, Thomas Scheidmeir, Dr. Stefan Simianer, William J. Chase Vorsitzende des Aufsichtsrats: Dr. Azita Saleki-Gerhardt This communication may contain information that is proprietary, confidential, or exempt from disclosure. If you are not the intended recipient, please note that any other dissemination, distribution, use or copying of this communication is strictly prohibited. Anyone who receives this message in error should notify the sender immediately by telephone or by return e-mail and delete it from his or her computer. Diese Kommunikation kann Informationen enthalten, die geheim, vertraulich oder hinsichtlich der Offenlegung beschränkt sind. Wenn Sie nicht der beabsichtigte Empfänger sind, nehmen Sie bitte zur Kenntnis, dass jede Weitergabe, Verteilung, Verwendung oder Vervielfältigung dieser. Kommunikation strikt untersagt ist. Jeder, der diese Nachricht fehlerhaft erhält, sollte den Sender unverzüglich telefonisch oder durch Rücksendung der E-Mail benachrichtigen und diese von seinem oder ihrem Computer löschen.
[NMusers] Parameter uncertainty in simulations
Dear all, I would appreciate to learn your experience and tips on how to a) estimate parameter uncertainty, and b) sample from such uncertainty in simulations. References would also be much appreciated. In the archives there is mention of parameter uncertainty, but I could not find this subject discussed directly. Hopefully, we may start an informative thread. Methods I found mentioned for estimating parameter uncertainty are either taking the covariance matrix from NONMEM, or obtaining the - not necessarily normally distributed - covariance structure from a non-parametric bootstrap. Estimation and simulation can be quick, but a bootstrap of 1000 replicates or more often is not done that quickly. It may be quicker to use the NONMEM covariance matrix. When should one not do this, or how could one tell that actually there may be problems with using the reported precision of parameters in NONMEM, and the true uncertainty is much better estimated via the bootstrap? When taking the parameter precision from NONMEM's covariance matrix, should one log transform parameters and estimate random subject level and residual error parameters as thetas, added to fixed etas and sigmas? (See also: http://www.cognigencorp.com/nonmem/current/2008-July/1060.html) I have seen an example for a PKPD model where the concentration effect relationship is modelled as a linear relationship with estimation of slope and inter-patient variability in slope. When parameters are not transformed, NONMEM reports a precision of 48% in the population slope and 99% in the variance, however these both drop to 7% after the transforms. Are these typical of the reductions in biased estimation of parameter uncertainty we seek, or may such large changes prompt you to further examine the model and / or trigger you to run the bootstrap? Thank you for your thoughts and responses. Kind regards, Mendel Mendel Jansen Director Modeling and Simulation Clinical Pharmacology and Translational Medicine Eisai Limited London UK e-mail mendel_jan...@eisai.net Eisai Limited Registered in England No. 2242511 Registered Address: 3, Shortlands, London W6 8EE, UK Please note our new telephone no. is +44 (0) 845 676 1400 and the fax is +44 (0) 845 676 1401. THIS EMAIL AND ANY ATTACHED FILES ARE CONFIDENTIAL. If you are not the intended recipient you are notified that any disclosure, reproduction,copying, distribution, or action taken in reliance on the contents of this information is strictly prohibited. If you have received this transmission in error please notify the sender immediately and then delete this email. Email transmission cannot be guaranteed to be secure or error free as information could be intercepted, corrupted, lost, destroyed,arrive late or incomplete, or contain viruses. The company/sender accepts no liability for any changes made to this email during transmission or any damage caused by any virus transmitted by this email. Any views or opinions expressed in this email are solely those of the author and do not necessarily represent those of the company. The company/sender accepts no liability for the content of this email, or for the consequences of any actions taken on the basis of the information provided, unless that information is subsequently confirmed in writing.