Jokha Al Qassabi1, Adam Darwich1, Amin Rostami1,2
1Centre for Applied Pharmacokinetic Research, University of Manchester, United Kingdom 2 Certara UK Ltd., Sheffield, United Kingdom
Introduction/Objectives: Renal impairment (RI) leads to physiological changes that can affect the pharmacokinetics (PK) of drugs [1]. These changes, besides the reduced glomerular filtration rate (GFR), include reductions in the activity of hepatic drug metabolising enzymes and plasma protein binding (and hence fraction unbound (fu))[1]. The latter can be due to alteration in proteins levels (mainly human serum albumin (HSA) and α1-acid glycoprotein (AAG))[2]. The main objectives of this study were to develop (1) a model that can predict the (fu) based on known changes in the levels of (HSA and AAG) in various degrees of RI for 23 drugs and, (2) a model that can predict the clearance (CL) using the (fu) model for 14 drugs in different stages of RI. The overall aim of this study was to examine the importance of (fu) in determining the changes in the (PK) for various drugs at varying degree of RI, when accounting for (GFR), (fu), and metabolic enzyme activity (CYP enzymes). This attempt hopefully will help in understanding the role of protein binding and more importantly, it might help to explain some of the variability seen with drug exposure in RI population.
Method: A list of renal impairment pharmacokinetics studies was collected, where fu of a drug was reported as a measured value. Total drugs that were found with a measured fu were 23 drugs. Beside this, information on protein binding, pka values, compound type, fraction eliminated or fe, metabolism via CYP enzymes (fm if reported), total clearance and renal clearance if reported, bioavailability, route of administration, dose of the drug given, the study group and the number of patients was also collected. The same list of drugs was used for the prediction of total clearance, renal clearance as well as non-renal clearance of the drugs at different stages of renal impairment. Nine drugs were excluded from this list as there was either no information on fe, or renal clearance, or no information on their metabolic clearance or the fraction metabolised at varying degree of RI. 4 models were developed using data from the list of dedicated renal impairment studies.
The equation used for calculating the HSA for a given GFR is as follow(Model1):
Equation 1
y=0.00005x^2-0.0004x+3.9875
Where y represents the HSA in g/dL, and x is the glomerular filtration rate in mL/min.1.73m2.
The equation used for calculating the AAG for a given GFR is as follow (Model2):
Equation 2
y=-0.0031×2 + 0.3239x + 20.143
Where y represents the AAG in µmol/ L, and x is the glomerular filtration rate in mL/min.1.73m2.
Model3:
Equation 3
fui=1/(1+((1-fu)*[P]i)/([P]*fu))
Where, fui is the individual fraction unbound in the renal impairment patient, fu is the fraction unbound in healthy individual, [P]i is the concentration of either HSA or AAG in renal impaired patient and [P] is the concentration of either HSA or AAG in healthy individuals.
Model 4:
Equation 4: CLr Predictions:
CLrmild = (CLrHV*((GFRmild*fumild) / (GFRHV*fuHV)))
CLrmoderate =(CLrHV*((GFRmoderate*fumoderate)/ (GFRHV*fuHV)))
CLrSevere = (CLrHV*((GFRsevere*fuSevere) / (GFRHV*fuHV)))
Equation 5: CLnr Predictions:
CLnrPred(mild)= (((ClnrHV*fmCYPn*fraction of CYPn activity in mild) *(fumild/fuHV)
CLnrPred(moderate)= (((ClnrHV*fmCYPn*fraction of CYPn activity in moderate) *(fumoderate/fuHV)
CLnrPred(Severe)= (((ClnrHV*fmCYPn*fraction of CYPn activity in severe) *(fusevere/fuHV)
Equation 6: Clt Prediction:
Clt mild = Clr mild +Clnr mild
Clt moderate = Clr moderate +Clnr moderate
Clt Severe = Clr severe +Clnr severe
Results: The results showed that the (fu) model performed well when compared with those reported within clinical studies as the predicted values of 23 drugs were within 2-fold of observations as well as the (CL) model estimation. Predictions of (fu) (AFE=0.96, AAFE=1.27) and (CL) (AFE=0.65, AAFE=1.78) based on HSA seems to be better than predictions of (fu) (AFE=0.69, AAFE=1.48) and (CL) (AFE=0.53, AAFE=1.98) based on AAG.
Conclusion: The statistical models proposed in this study for fu and CL seems to be predictive of the clinical data. In the absence of measured protein levels this model can be of use. Clearance and plasma protein binding predictions showed superior performance for HSA substrate drugs as compared to AAG, as more data was available to inform the HSA binding model
References:
[1] Rowland Yeo, K. et al. (2011) ‘Modeling and predicting drug pharmacokinetics in patients with renal impairment’, Expert Review of Clinical Pharmacology, 4(2), pp. 261–274. doi: 10.1586/ecp.10.143
[2] Benet, L. Z. and Hoener, B. A. (2002) ‘Changes in plasma protein binding have little clinical relevance’, Clinical Pharmacology and Therapeutics, 71(3), pp. 115–121. doi: 10.1067/mcp.2002.121829.
Reference: PAGE 28 (2019) Abstr 9176 [www.page-meeting.org/?abstract=9176]
Poster: Drug/Disease Modelling - Other Topics