Hi, I wonder could you help me regarding a model for tumour growth inhibition. I am trying to use an indirect model as detailed in a paper by Laurent Salphati - http://dmd.aspetjournals.org/content/38/9/1436.full.pdf+htm -but I am not having much success and would really appreciate some help identifying where I have gone wrong. Many thanks in advance. the model I have is test(){ deriv(Aa = - Ka * Aa) deriv(A1 = Ka * Aa - Cl * C) deriv(E = (kng * E) - (K * E)) dosepoint(Aa) K=Kmax * C/KC50 + C C = A1 / V sequence{E = kng / K} error(EEps = 1) observe(EObs = E + EEps) stparm(kng = tvkng) stparm(K = tvK) stparm(Kmax = tvKmax) stparm(KC50 = tvKC50) fixef(Ka(freeze) = c(, 0.427, )) fixef(V(freeze) = c(, 439, )) fixef(Cl(freeze) = c(, 135.2, )) fixef(tvkng = c(, 25, )) fixef(tvK = c(, 0.33, )) fixef(tvKmax = c(, 0.76, )) fixef(tvKC50 = c(, 1050, )) } I have a data sheet with time and tumour volume, and a dose sheet with dose for 14days of dosing. “The error I get is as follows Could not merge sort times Some data that is expected to be numeric is not Please check your source data” I think this is becasue Aa rate appears in the dose sheet and I am not sure what to fill in here as the dose regimen is oral QD for 14 days. Any help most welcome of if you have a model I could use which does this already. kind regards Anne
Anne, please can you attach your PHXPROJ, or at least the dose and data sheets as this error message seems to suggest a set up issue rather than necessarily a model error. Simon.
Thank you Simon, file attached. Anne [file name=2011PKPD141.phxproj size=284607]Certara | Drug Development Solutions
2011PKPD141.phxproj (278 KB)
Anne, there was indeed a conflict in the data, I think it didn’t like the multiple ‘null’ times in the raw data sheet, I’ve deleted your internal dosing sheet and added a full time column to your original data file so the model now can execute. [file name=2011PKPD141_PKPD.phxproj size=756582]Certara | Drug Development Solutions 
however there is still some issues with your internal estimates and/or model, unfortunately I have to complete something else today but at least that gets you a bit nearer; do you have other dose levels and/or conc data available? Simon.
Thanks Simon, I’ll have a look at what you have done and will get back to you later with concs and other dose levels. I have a few meetings to attend to in the next few hours. kind regards Anne
Dear Anne I see 2 major problems one in the codee, one in the data set. 1: In the code, you use C before it has been defined. You need to put that code at the beginning as follws: C = A1 / V K=Kmax * C/KC50 + C The reason is that K=Kmax is a mathematical equation that uses C and therefore C must be defined before. 2: In the data set, you put on the same row a dose and an observation. YOu cannot do that. Phoenix will not pay attention to the observation or to the dose. You need separate rows for observation and doses. Just duplicate all your data set and on one duplciate clear all the dosing ans the other all the responses. Your initial estimate also leads to values of the response that are far away from what we observe. Adjust your initial estimates. Best Serge
The model for E is not right either you have in your model either an exponential decrease or increase and this does not match with the initial conditions that assume steady state of an indirect response model. in sequence you just need to put sequence {E=E0} and define E0 as a structural parameter stparm(E0= tvE0) fixef(tvE0 = c(, 100, )) Now, your data suggest more an exponential increase. Therefore choose K and kng and other parameters that you have drug induced increase in E. BEST SERGE
tHE DRUG INCREASES E based on the data. Therefore the effect of the drug which is related to K should show the K part positive, not negative deriv(E = (-kng * E) + (K * E)) would make it. best Serge
you have also a scaling problem. your drug levels are in the range of 1000000 to 4000000. This is no sense when I am looking at the ec50 you are using. Apparenlty there is a factor million here to take care of. Also apparenlty you want to use indirect response model and nto exponential growth. Am I right. If you are using indirect response mdoel, then the model should be C = A1 / V kng=Kmax * C/KC50 + C deriv(Aa = - Ka * Aa) deriv(A1 = Ka * Aa - Cl * C) deriv(E = kng - K * E) In this model I made, the production of E increases with C which would lead to E increasing with C. Anyway, first of all check the units of your dosing as C in millions will not make anything good in the fitting process. best Serge
Hi Simon and Serge Thank you for your input. serge, the units for dosing is in ng/Kg. I have incorporated your suggestions into the model and have attached the file. As I mentioned previously I was looking to write a model in line with models in the literature according to Salphati et al and Yamasaki et al. This is to allow me to look at tumour growth inhbition in relation to conc and biomarker modulation and define whether effect on tumour growth is off-target and deter,iine a stasis concentration. In the file attached you will see that the model based on the Yamasaki paper is fairly resonable Model_829_30mpk_Yamasaki Model_829_100mpk_Yamasaki However I would appreciate some advise on the model that based is on the Salphati paper. The Salphati model would be my preferred model. 829 PK_indirect_30mpk_Salphati Attached are both papers combined in a single pdf file for reference (Yamasaki P1269, Salphati P1438)
sorry forgot to add files Anne [file name=TGI_model_development.phxproj size=1141099]Certara | Drug Development Solutions
TGI_model_development.phxproj (1.09 MB)
papers combined in single pdf for reference having trouble with sending multiple files Regards Anne [file name=Yamasaki_and_Salphati_papers.pdf size=730869]Certara | Drug Development Solutions
Yamasaki_and_Salphati_papers.pdf (714 KB)
Hi Anne, simply use “Edit post” after submitting and you can keep attaching multiple files to the same post. Simon.
Hi I wonder whether someone could look at my indirect model for Tumour growth inhibition. just to say that I am not very experienced in PK/PD modelling. I have used PK parameters to simulate the conc. I have fitted the dose groups individually. I would like to calculate the conc required for stasis and use the PK/PD parameters to simulate the ED50. Fitting the data individually gives me 3 very different sets of parameters (except for KC50s- I don’t know how to fit the data simultanously. Can you please let me know how to do to do this and am I right in thinking that a simulatanous fit would aid the in the calc of the Cstasis and and the ED50? Your help would be most appreciated. Anne My phxproj file is attached for reference. The model is test(){ deriv(Aa = - Ka * Aa) deriv(A1 = Ka * Aa - Cl * C) deriv(E = kng * E - (K * E)) dosepoint(Aa) C = A1 / V K = (Kmax * C)/(KC50 + C) sequence{E = E0} error(EEps = 1) observe(EObs = E*(1+EEps)) stparm(E0=tvE0) stparm(kng = tvkng) stparm(Kmax = tvKmax) stparm(KC50 = tvKC50) fixef(Ka(freeze) = c(, 0.427, )) fixef(V(freeze) = c(, 439, )) fixef(Cl(freeze) = c(, 135.2, )) fixef(tvE0(freeze) = c(, 100, )) fixef(tvkng = c(, 4.8, )) fixef(tvKmax = c(, 5.25, )) fixef(tvKC50 = c(, 200, )) } where E = tumour volume kng is net growth constant for tumour K is rate constant describing tumour reduction effects of compound [file name=TGI_model_development_Salphati_2013_04_23.phxproj size=959584]Certara | Drug Development Solutions