Parent drug with metabolites and BQL

[davidq8]

HI, I am working on this project involving building a population Pk model for a parent drug with two metabolites.

Plasma concentrations were taken for all three compounds from the time of dosing until the end of the study (the last plasma sample were taken 7
days later).

The drug was not detected in many samples at the end of the plasma collection points for the parent drug. For many subjects, the drug was not detected in the last 3 plasma samples (48h,72h, and 7 days later).

I am familiar with Beal (2001) methods for imputation of BLOQ. Still, I am unsure if implanting it with the parent drug could give us an accurate estimation since we have many plasma concentrations, either BLQ or not detected (0).

I have attached a sample of the data to show the problem and VPC for the parent drug with handling all BQL and non-detected as missing data.

Can I get help on how to deal with this issue?

Thanks

[Samer_Mouksassi]

I suggest that you try to have a simple model ( only parent) with BLQ before going into a full blown multiple observation with BLQs.

you can have the value of the BQL in your observation and then you tick the BQL box and map the proper information or you can a have a fixed blq

https://onlinehelp.certara.com/phoenix/8.3/index.html#t=topics%2FObserve_statement_for_Gaussian_Residual_models.htm&rhsearch=bql&rhhlterm=bql&rhsyns=%20

Phoenix also can deal with vpc of bql when you have bql

https://onlinehelp.certara.com/phoenix/8.3/index.html#t=topics%2FContinuous_observed_variable_tab_s_options.htm&rhsearch=vpc&rhsyns=%20

[davidq8]

Thanks, Sameer, for your reply.

I am using the following PML codes for modeling a parent drug with a metabolite assuming inversible metabolite formation:

test(){

deriv(A1 = (Aa * Ka)- (Cl2 * (C - C2))- (CLPar2Met * C))

deriv(Aa = - (Aa * Ka))

deriv(A2 = (Cl2 * (C - C2)))

deriv(AMet = (CLPar2Met * C)- (CLMet * C3)- (CL3 * (C3 - C4)))

urinecpt(A0Met = (CLMet * C3))

deriv(A4 = (CL3 * (C3 - C4)))

C = A1 / V

dosepoint(Aa,tlag = Tlag)

C2 = A2 / V2

error(CEpspar = 0.1)

observe(CObsPar = C + CEpspar * sqrt(1 + C^2 * (CMultStdevpar/sigma())^2))

C3 = AMet / VMet

C4 = A4 / V4

error(CEpsmet = 0.1)

observe(CObsMet = C3 + CEpsmet * sqrt(1 + C^2 * (CMultStdevmet/sigma())^2))

stparm(V = tvV * exp(nV))

stparm(Ka = tvKa * exp(nKa))

stparm(V2 = tvV2 * exp(nV2))

stparm(Cl2 = tvCl2 * exp(nCl2))

stparm(VMet = tvVMet * exp(nVMet))

stparm(CLPar2Met = tvCLPar2Met * exp(nCLPar2Met))

stparm(CLMet = tvCLMet * exp(nCLMet))

stparm(V4 = tvV4 * exp(nV4))

stparm(CL3 = tvCL3 * exp(nCL3))

stparm(Tlag = tvTlag * exp(nTlag))

stparm(CMultStdevpar = tvCMultStdevpar)

stparm(CMultStdevmet = tvCMultStdevmet)

fixef(tvV = c(, 11904761.9, ))

fixef(tvKa = c(, 1, ))

fixef(tvV2 = c(, 22874500, ))

fixef(tvCl2 = c(, 2226500, ))

fixef(tvVMet = c(, 879700.9, ))

fixef(tvCLPar2Met = c(, 2226500, ))

fixef(tvCLMet = c(, 62000, ))

fixef(tvV4 = c(, 847500, ))

fixef(tvCL3 = c(, 62000, ))

fixef(tvTlag = c(, 0.9, ))

fixef(tvCMultStdevpar = c(, 1, ))

fixef(tvCMultStdevmet = c(, 1, ))

ranef(diag(nV, nKa, nV2, nCl2, nVMet, nCLMet, nCLPar2Met, nTlag,nCL3, nV4) = c(1, 1, 1, 1, 1, 1, 1, 1, 1 ,1))

The fit is quite acceptable based on GOF plots (attached) but I am wondering if you have any suggestion to improve it .

Thanks