#> Warning: package 'knitr' was built under R version 4.3.3
preDose is a free and open source package that automatize the process of external evaluation process using an independent dataset from which the original popPKPD model was developed.
Currently, the user can use the package choose to use the package based on the mapbayr package
You can perform an external evaluation for a single model from :
- a population PKPD model (coded in mrgsolve,
- a data set with concentrations (NONMEM format)
You can install the development version of preDose from GitHub with:
install.packages("devtools")
devtools::install_github("Martin-Umpierrez/preDose")This document presents a case of study to illustrates the external evaluation of two population pharmacokinetic model of Tacrolimus:
-
Han et al. (2011)
Prediction of the tacrolimus population pharmacokinetic parameters according to CYP3A5 genotype and clinical factors using NONMEM in adult kidney transplant recipients.
European Journal of Clinical Pharmacology, 69(1), 53–63. -
Zuo et al. (2013)
Effects of CYP3A4 and CYP3A5 polymorphisms on tacrolimus pharmacokinetics in Chinese adult renal transplant recipients: a population pharmacokinetic analysis. Pharmacogenet Genomics. 2013 May;23(5):251-61. doi: 10.1097/FPC.0b013e32835fcbb6. PMID: 23459029.
The external evaluation was performed using data from an internal study conducted between 2022 and 2024 by Umpierrez et al.
library(preDose)
## basic example codeModels can be defined in two different ways:
- Inline model code, which can be stored and accessed globally within the project.
- External
.cppfiles, which can be read and compiled directly by mrgsolve.
Additionally, the package provides several pre-coded models that can be
accessed using the list_models() function, allowing users to readily
explore and apply available population pharmacokinetic models.
Han_etal_test<-
'$PROB
# One Comparment Model with first order absorption- Ka is FIXED
$GLOBAL
#define CP (CENT/iV)
$CMT @annotated
EV : Extravascular compartment
CENT : Central compartment#two compt model with first order absorption
$PARAM @annotated
CL : 24.13 : Clearance for CYP3A5*3*3
V : 716 : central volume
KA : 4.5 : absorption rate constant
ETA1 : 0 : IIVCl (L/h)
ETA2 : 0 : IIVV (L)
$PARAM @annotated @covariate
POD : 0 : COV POST OPERATIVE DAY
HCT : 0 : COV HCH
WT : 0 : COV WT
CYP3A5 : 0 : Polimorfismo CYP3A5
OCC : -99 : Occasion, shall be passed by dataset imported
$ODE
dxdt_EV = -iKA*EV;
dxdt_CENT = iKA*EV - iCL*CP;
$MAIN
##CYP3A5 effect on Cl##
double HM = 1.186 ; ####Rapid Metabolizer ####
double IM = 1.13 ; ####Intermediate Metabolizer ####
double PM = 1 ; ####Poor Metabolizer####
if(CYP3A5==1) double CL_EFFECT = HM ;
if(CYP3A5==2) CL_EFFECT = IM ;
if(CYP3A5==3) CL_EFFECT = PM ;
double CL_HCT1 = 1.3458 ; ##Effect of HCT< 33
double CL_HCT2 = 1.124 ; ##Effect of HCT >33
if(HCT< 33) double CL_HCT = CL_HCT1 ;
if(HCT >= 33) CL_HCT = CL_HCT2 ;
double CL_POD = - 0.00762 ;
double iCL = CL *exp(ETA(1) + ETA1)* pow(POD, CL_POD) * CL_EFFECT * CL_HCT ;
double iV = V *exp(ETA(2) + ETA2) * exp (0.355*WT/59.025) ;
double iKA = KA ;
$OMEGA @name IIV
0.248
0.237
$SIGMA @name SIGMA @annotated
ADD : 0 : ADD residual error
PROP : 0.16 : Proportional residual error
$TABLE
double IPRED = CENT/iV;
double DV = IPRED * (1 + PROP) ;
$CAPTURE @annotated
CP : Plasma concentration (mass/volume)
iCL : Clearance
iV : :Central Volume
iKA : KA: absorption rate constant
EVID : EVENT ID
DV : PREDICCION
OCC: OCCASION
'Import the data set in NM-TRAN-formatted datasets Ensure that the dataset structure aligns with the required format for proper processing.
data("tacrolimus_pk1_kidney", package = "preDose") # Cargar dataset desde el paquete
head(tacrolimus_pk1_kidney) # Ver primeras filas
#> # A tibble: 6 × 30
#> ID OCC DD AMT TIME POD DV EVID CMT MDV II SS
#> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 1 1 6 3000 168 7 0 1 1 1 12 1
#> 2 1 1 6 0 168 7 9.4 0 2 0 0 0
#> 3 1 2 6.5 3250 264 11 0 1 1 1 12 1
#> 4 1 2 6.5 0 264 11 8.4 0 2 0 0 0
#> 5 1 3 7.5 3750 360 15 0 1 1 1 12 1
#> 6 1 3 7.5 0 360 15 8.4 0 2 0 0 0
#> # ℹ 18 more variables: Creatinine <dbl>, SCR <dbl>, eGFR <dbl>, ClCrea <dbl>,
#> # AGE <dbl>, SEX <dbl>, WT <dbl>, HCT <dbl>, CYP3A5 <dbl>, EXPRESSION <dbl>,
#> # PDN_DOSE <dbl>, PDNXWT <dbl>, Heigth <dbl>, Height..m. <dbl>, BSA <dbl>,
#> # BMIcalc <dbl>, LBW <dbl>, DMELITU <dbl>External model evaluation is performed using the exeval_ppk()
function.
This function automates the full evaluation workflow, including:
- Calculation of individual parameters
- Model updating based on individual estimates
- Simulation of predicted concentrations
- Comparison between observed and predicted data
A step-by-step description of the external evaluation procedure
(i.e., individual parameter estimation, model updating, and
simulation)
is provided in the corresponding vignette.
Please refer to the vignette for a detailed and reproducible explanation of each step.
res.1 <- exeval_ppk(model = Han_etal_test,
drug_name = "Tacrolimus",
model_name = "Han_etal_2011",
tool = "mapbayr",
data = subset(tacrolimus_pk1_kidney, ID <11),
evaluation_type= "Progressive",
assessment = 'Bayesian_forecasting') # Cargar dataset desde el paquete
print(res.1)
#> ===================================
#> Data summary
#> argument value
#> 1 Num IDs 10
#> 2 Observations 30
#> 3 Max Num Occasion 6
#> ===================================
#>
#> ===================================
#> Evaluation information
#> argument value
#> 5 Drug Name Tacrolimus
#> 6 Model Name Han_etal_2011
#> 7 Evaluation Progressive
#> 8 Assessment Bayesian_forecasting
#> ===================================
#>
#> ===================================
#> Evaluation metrics
#> # A tibble: 5 × 8
#> OCC rBIAS rBIAS_lower rBIAS_upper MAIPE rRMSE IF20 IF30
#> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 2 14.5 -7.96 37.0 31.5 38.5 23.1 69.2
#> 2 3 5.86 -27.5 39.2 40.5 44.6 10 20
#> 3 4 17.0 -8.49 42.4 35.1 43.9 38.5 53.8
#> 4 5 -8.59 -21.0 3.87 16.7 18.6 70 90
#> 5 6 -8.72 -28.5 11.1 23.8 27.7 50 60
#> ===================================Several plot types are available to visualize model performance using
the metrics_plot() function. ##### 4.1) Bias BarPlot
plot1 = metrics_plot(res.1,
type = "bias_barplot")
print(plot1)
##### 4.2)
Bias boxplot
plot2 = metrics_plot(res.1,
type = "bias_boxplot")
print(plot2)
plot3 = metrics_plot(res.1,
type = 'error_plot')
print(plot3)
source("inst/model_examples/ZuoX_etal_2013.R")
res.2 <- exeval_ppk(model = ZuoX_etalfull_noCYP3A4,
drug_name = "Tacrolimus",
model_name = "Zuo_etal",
tool = "mapbayr",
data = subset(tacrolimus_pk1_kidney, ID <11),
evaluation_type= "Progressive",
assessment = 'Bayesian_forecasting') # Cargar dataset desde el paquete
print(res.2)
#> ===================================
#> Data summary
#> argument value
#> 1 Num IDs 10
#> 2 Observations 30
#> 3 Max Num Occasion 6
#> ===================================
#>
#> ===================================
#> Evaluation information
#> argument value
#> 5 Drug Name Tacrolimus
#> 6 Model Name Zuo_etal
#> 7 Evaluation Progressive
#> 8 Assessment Bayesian_forecasting
#> ===================================
#>
#> ===================================
#> Evaluation metrics
#> # A tibble: 5 × 8
#> OCC rBIAS rBIAS_lower rBIAS_upper MAIPE rRMSE IF20 IF30
#> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 2 102. 67.4 136. 102. 116. 0 7.69
#> 2 3 74.6 13.0 136. 80.9 111. 20 40
#> 3 4 41.9 -2.32 86.2 58.5 81.9 23.1 38.5
#> 4 5 20.7 -4.55 45.9 29.6 39.4 50 60
#> 5 6 1.53 -19.5 22.6 25.9 28.0 40 70
#> ===================================- Use
combine_metrics()to generate a summary of all evaluation metrics across the tested models. - Visualize and compare model performance using the
plot_combined()function.
###### Generate a summary of metrics for all tested models
model_list <- list(list(model_name="Han_etal", metrics_list=res.1$metrics),
list(model_name="Zuo_etal", metrics_list=res.2$metrics))
#### Use combine_metrics() function with the summary created
combined_results<- combine_metrics(model_list)
#### Make the Plot!
plot_comparrison <- plot_combined(combined_results,
'bias_barplot')
print(plot_comparrison)
5.2.2) Select models according to a specific evaluation metric and threshold using select_best_models() function
The select_best_models() function selects the best models from a
dataframe of combined metrics based on a specified ranking metric.
It requires a dataframe containing model evaluation metrics and the name
of the metric to use for ranking.
Optionally, you can specify a particular occasion to focus on and the
number of top models to select.
Best_fit <- select_best_models(combined_results, metric = "rBIAS",
top_n = 1)
print(Best_fit)
#> # A tibble: 5 × 9
#> OCC rBIAS rBIAS_lower rBIAS_upper MAIPE rRMSE IF20 IF30 Model
#> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <chr>
#> 1 2 14.5 -7.96 37.0 31.5 38.5 23.1 69.2 Han_etal
#> 2 3 5.86 -27.5 39.2 40.5 44.6 10 20 Han_etal
#> 3 4 17.0 -8.49 42.4 35.1 43.9 38.5 53.8 Han_etal
#> 4 5 -8.59 -21.0 3.87 16.7 18.6 70 90 Han_etal
#> 5 6 -8.72 -28.5 11.1 23.8 27.7 50 60 Han_etal