RE: Modeling accelerated phase of malignancy
Dear Mark,
I think one problem might be the distributioal assumption on NTLAG in that if you have outlying individuals with early/late lag you may need to think of a sensible transformation for the ETA or there might be no obvious one. Do all the subjects go into the exponential phase or might the typical value of NTLAG be greater than your observed time? That aside, something to try would be to have a continuous function driving NTLAG, e.g.:
NATHL = NATH*T**50/(T**50+NTLAG**50)
Play around with the shape parameter value, or even consider estimating it thus allowing a more gradual transition?
Hope this helps,
Joe
Joseph F Standing
MRC Fellow, UCL Institute of Child Health
Antimicrobial Pharmacist, Great Ormond Street Hospital
Tel: +44(0)207 905 2370
Mobile: +44(0)7970 572435
Quoted reply history
________________________________________
From: owner-nmusers_at_globomaxnm.com [owner-nmusers_at_globomaxnm.com] On Behalf Of Mark Sale [msale_at_nuventra.com]
Sent: 29 May 2015 20:52
To: nmusers_at_globomaxnm.com
Subject: [NMusers] Modeling accelerated phase of malignancy
Dear Colleagues,
Im working on a model of a malignancy that, at some point in the course of the disease enters into an accelerated phase. Im using a sort of standard serial compartment model, with a zero order input rate, then first order transit to the next compartment. I think the correct model for natural history is a slow increase in the input rate over time, then, at some point change to an exponential growth. Im having trouble getting NONMEM to do this. The relevant code I have in $DES is:
IF(T.LT.NTLAG) THEN
LGIND = 0
ELSE
LGIND = 1
END IF
NATHL = LGIND*NATH
DADT(3) = (INPUT-A(3)*K)+ A(3)*NATHL
DADT(4) = A(3)*K-A(4)*K
.
.
.
Where NTLAG is an estimated parameter for the lag time between entry into the study and the onset of the accelerated phase, NATH is the natural history term, NATHL is the lagged natural history term, INPUT is the zero order input rate and K is the first order transit constant. FOCE actually works pretty well for this for the THETA term for NTLAG, gives reasonable values. Probably is with the ETA for NTLAG (which is essential since it varies from person to person. With FOCE I get zero gradient for it. BAYES, SAEM, IMP MAP and ITS give reasonable values for OMEGA, but conditional values for ETA are all zero.
What I think is going on is that, unlike an ALAG, there is not event at that point in time, so small changes in ETA (smaller than the integration step size) dont change the predicted value, so no gradient and all ETAs = 0 with EM methods.
Ive tried to figure out a way to do this with an additional compartment for the natural history and havent been able to yet. That, I think would solve the problem, since an event would be inserted at the end of ALAG.
Any ideas on a solution, is there a way to insert an event at an unknown time?
Thanks
Mark
********************************************************************************************************************
This message may contain confidential information. If you are not the intended recipient please inform the
sender that you have received the message in error before deleting it.
Please do not disclose, copy or distribute information in this e-mail or take any action in reliance on its contents:
to do so is strictly prohibited and may be unlawful.
Thank you for your co-operation.
NHSmail is the secure email and directory service available for all NHS staff in England and Scotland
NHSmail is approved for exchanging patient data and other sensitive information with NHSmail and GSi recipients
NHSmail provides an email address for your career in the NHS and can be accessed anywhere
********************************************************************************************************************