dear all forum members, greetings! I want to predict multiple dose data on basis of a single dose data. I am using Winnonlin software for the prediction of bid data from single dose data. It is reported that drug is following non linear kinetics and i guess i cant use non compartmental analysis. so kindly let me know the exact method . thanks & regards stats06
Dear stats06! [quote]It is reported that drug is following non linear kinetics and i guess i cant use non compartmental analysis.[/quote]Correct (see the User’s Guide). [quote]so kindly let me know the exact method.[/quote]I’m not sure what you mean by exact. Even if you model your single dose data you have no means to predict steady state in case of nonlinear PK (the superposition principle is not applicable).
dear helmut , We conducted a pilot bio study on 20 volunteers for 400 mg test product (OD) and 200 mg (BID) innovator product . The be limits were satisfied . My aim now is to predict the multiple dose data and steady state data. in the literature it is given that the drug follows a two compartmental model & is dose dependent. i read the user guide of winnonlin, the steps i am following for my data are: 1.) I fit the plasma data in two compartmental model (model 11). 2.) Then select the option of simulate data . 3.) choose the dosing scheme. 4.) fill the model parameters. Now how to fill these model parameters when we have no knowledge about them ..? The option of winnonlin generated bound is also inactive when i follow the above steps… In case i want to use the NPS option i follow the below mentioned steps but i am not too sure that i am going in the right direction 1.)i have the plasma data of the 2 formulations for 20 subjects( 2 way crossover) 2.) select tools > NPS> variable dosing > 200 mg at 0 hr> repeat every 12 hrs. 3) time output (0-500) hrs 4.)calculate. since i have no past experience in doing simulation so i am a bit confused 
thanks & regards stats06
Hi Stats06 (not related to The Prisoner - Wikipedia are you?) In modelling and simulation you first have to FIT your data compartmentally, then use those final estimates for simulation. There examples of this in the product literature and also in text books e.g Gabrielsson and Weiner’s book; Certara | Drug Development Solutions has many worked examples. If you haven’t done any modelling before I would suggest you get some training or work with a collaborator in you non-linear drug to check your assumptions, but a basic intro would be; 1.) Fit the individual plasma data to feasible models, i.e. does the two compartmental model (model 11) have a lower AIC and better visual inspection of fits than the simpler 1 com model. That it is not over parametised etc. 2.) Regarding initial estimates in FITTING mode these are starting points that if you are lucky WinNonlin maybe able to find itself other wise you will have to perform curve stripping or use the literature values of your drug. 3.) Once you have a fitted model then you can use those model parameters for simulation As a reminder about Simulation set up for Classical models in Phoenix you can take a look at my first attempt at an instructional YouTube video for Phoenix; http://www.youtube.com/watch?v=rH6hCcOeWmw it’s just 3 minutes long but if it proves useful we will record some more when time allows. 4) Once you have fitted your individual data at your dose level see how it compares to literature results - do they give you an idea about how it scales between doses to adjust your simulation - do you know the mechanism/cause of the non-linearity ? 5) Use simulation to try different study designs to ensure that future studies have a good chance of meeting study objectives e.g. review VIF of parameters of interest. As Helmut already indicated NPS is not an appropriate technique in this instance. Simon.
dear Simon, greetings! nope i’m not 
(not related to The Prisoner - Wikipedia are you?) curve stripping or the literature reported values are the two options left for me… since i’m not lucky enough … it has been reported that the drug follows a two compartmental model… but visually analyzing the graphs i feel that i should go for one compartmental model.. so i’ll try it 2mrw … and m a going in the right direction for NPS in case the drug was having linear kinetics ? [In case i want to use the NPS option i follow the below mentioned steps but i am not too sure that i am going in the right direction 1.)i have the plasma data of the 2 formulations for 20 subjects( 2 way crossover) 2.) select tools > NPS> variable dosing > 200 mg at 0 hr> repeat every 12 hrs. 3) time output (0-500) hrs 4.)calculate] thanks for your reply and the 3 min video helped me a lot and helmut thanks 2 u as well thanks & regards stats06
Hi regarding your NPS for 500 hours, for such a long period you may need to increase you number of datapoints since the default will give you only 1 point every 30 mins so you could well miss Cmax. Remember… # NPS can only be as ‘good’ as your source profile, (use the actual times, personally I perform NPS from individual data and then summarise after) # the output should be sampled as least as frequently as the smallest time interval of your original profile, (can take a while if you have lots of subjects) # if you are using raw data then the Log box should be checked for the extrapolations. # and of course Linear kinetics are assumed Simon.
hi simon, greetings! thanks again for your reply (u r a saviour I did NPS as you said for individual subjects and then decided my steady state. I am attaching the raw data file of the pilot study three-treatment, three-period, six-sequence, oral crossover. My results were 6 days (approx) for for R and 8 days for T1 & 7 days for T2 . But according to my seniors these results can not be correct as per their knowledge of the drug is there something i did wrong ? thanks & regards stats06
No attachment. actually the Phoenix project would be much more helpful than the raw data file. As you already mentioned you suspect non-linear kinetics I suspect your supervisors’ thinking is correct that the results from NPS do not look right. We have consultants in India who can work more closely with you under a Non-Disclosure Agreement etc and then knowing the compound etc we may be able to suggest a better direction for you to take. Simon.
sorry forgot to attach the file . i cant attach the whole phoenix project .( as there are some rules here ) attaching the excel file for you and please let me know how can i contact your consultants in India . thanks stats06 [file name=raw_data_file.zip size=47409]/extranet/media/kunena/attachments/legacy/files/raw_data_file.zip[/file]raw_data_file.zip (46.3 KB)