2 replies to this topic

[Lance Wollenberg]

Attached is a really nice piece of work on modeling and simulation of adverse events with erlotinib using a continuous Markov chain. Is it possible to implement this approach in Phoenix? If so, how is this accomplished? I searched the forum here and hadn't found anything, however if this has been discussed, please let me know.

 

Thanks!

 

Lance

[Lance Wollenberg]

Attached is a really nice piece of work on modeling and simulation of adverse events with erlotinib using a continuous Markov chain. Is it possible to implement this approach in Phoenix? If so, how is this accomplished? I searched the forum here and hadn't found anything, however if this has been discussed, please let me know.

 

Thanks!

 

Lance

Attached Files

•  Markov Erlotinib Model.pdf   760.57KB   697 downloads

[smouksassi1]

Hi Lance,

PML has all the infrastructure to enable you to fit discrete or continuous markov multistate models. 

from the publication:

DADT(3) = KBB1*A(4)-KFF1*A(3)                       ; No rash
DADT(4) = KFF1*A(3)+KBB2*A(5)-KBB1*A(4)-KFF2*A(4)     ; Mild rash
DADT(5) = KFF2*A(4)+KBB3*A(6)-KBB2*A(5)-KFF3*A(5)     ; Moderate rash
DADT(6) = KFF3*A(5)-KBB3*A(6)                       ; Severe rash

these are just ODE with parameters controlling the probabilities to 
transition from one state to another 
some models assume a hierarchy where you cannot pass 
from mild to severe ( you just missed to observe it ) 
and talk about a latent kind of state.
other multistate models have what we all absorbent 
state where if you attain it your are kicked out e.g. Death State.

Edited by smouksassi1, 22 June 2018 - 06:23 PM.