Pmetrics User Manual - Laboratory of Applied Pharmacokinetics
Contents
1. M 25 External Validation eroe eroe eer vae netur aa Seer aa e eee yo Ree N ad eee YER Fe e ERERR e VE E ERRARE ERERRNN E MR ERRRSSRERENE E ERR MARE EKRTEENV T 26 DOES REO TETOID IIT D ELEC ILDLMET 27 Pmetrics User s Guide 2 Introduction Thank you for your interest in Pmetrics This guide provides instructions and examples to assist users of the Pmetrics R package by the Laboratory of Applied Pharmacokinetics at the University of Southern California Please see our website at www lapk org for more information Here are some tips for using this guide Items that are hyperlinked can be selected to jump rapidly to relevant sections Atthe bottom of every page the text Pmetrics User s Guide can be selected to jump immediately to the table of contents Items in courier font correspond to R commands Disclaimer You the user assume all responsibility for acting on the results obtained from Pmetrics The USC Laboratory of Applied Pharmacokinetics members and consultants to the Laboratory of Applied Pharmacokinetics and the University of Southern California and its employees assume no liability whatsoever Your use of the package constitutes your agreement to this provision System Requirements and Installation Pmetrics is a package for the freely available statistical and programming R software which can be obtained from http www R project org R and Pmetrics run in Windows and Mac Unix e2. 8 SUDFOUTING 4 GETFA 9 Subroutine 5 GETIX ciisicscssecciaciesciicsseccds Hr i 9 SUDPOUTING 62 GETTEAG iiicsssccsdesesisisessadsetesenstscssnaseieeesia sdaeaussveusessicedavcsseGeaassadaase ssceuuvessccadeas caubvesess vobas s cseeeessccesas 9 Subroutine 7 ANALI iicidiccsedds cieeccctiiieadd cocececedaaadssevadsdceceddussasdvadedcecadeetsdsasssedcsseseedeceueducicevedssse0sdssvcccedsdesaeedscsicde 10 How to use R and Pmetrics sisccesccccaccsscscncsvcceacesnaconconsesveasuacevcnoneuceusecdbscccduevbactcctebesvdvoncessarsectvacciscbadseatsbeveines 10 Pmetrics Data ODJECtS iine orte lors kac co IG Eos Cc n REV CURED PU RUE R R ENDE DC TOR d E Ds cR RUDI eaten 12 Making New Pmetrics Objects iiiiieeoioi civis eder cues eue sx vu toU dod sutu epu kt DOE Dali dw SERE DL EVE eK CV EU DR 14 IT 2B UI 15 NPAG T LE 18 Simulator RUNS vsscissdsntenicctsacesdescebicucsscesscoveseveesceovacconctucsesdusdncodscucoucablcdbescdvaclelcdasiocdeudesuaponecpedsedoesacdaaisonbanss 21 Plotting eT Vetta cs carbs M 21 Model DiagioSlits uis snsccarensecandasssasnacacssccessesteencoucossdussacs bU MUR DURAN RN ideo Qr MET d VE IN EP Esni Ear Saksana IMPIIS 25 Internal Validation mE
3. Pmetrics User Manual November 21 2011 Package Version 0 17 An R package for parametric and non parametric modeling and simulation of pharmacokinetic and pharmacodynamic systems IM LAPK Laboratory of Applied Pharmacokinetics Table of Contents Dteniipiieee U 3 Disclaimer eo ceia coc ae epa cv cu eov aeos voa rcov v ap sae cod ek caue aera aoa VE Sae V Peu Fu Gv Na eu a apo Ue Va aae Sap s DESEE UE aasia ia aD EK OUO 3 System Requirements and Installation reiaco oria a vvv edi dE Ry gye qu exa gye rci o aS Fey ax ES day Free ir dr sYP ev I REY 3 What This Manual IS NOW tir eov pones nre suy vED vaa eva cub ue ee cr YR Eo orae ucenic oTo gU Re ep ek euy eua ry eere KURVE ra Ure Ossa Gunn aioi 3 Pretrics Components eee eee Seno ep eno deg eno eene rete van ee ob ae ua eon ue eo ao pare ao va eee unn kk o eae e Neo AY EE UeNa VY ur ua Eua S Sua esenea eass 4 General PI SP 2 ascsdscas concinesvadecacsbostechedeacbancsecnsenecd sdeesvecpndnedardne spade sipodsecaadebceviseusedariebdsvesseaccusscoaesedssudevaats 5 Pmetrics Input Fle Ss ccecnccatencheates inireseta CE p REL I IE dur ir aE EAE E Dee Ea EA ESES 6 BEKER T S E CU MER 6 OPRIIEPBSILIIM 7 SUDFOUTING 12 DIFFEQ em o 7 Subroutine 2 OUTPUT T 8 S bro utine 3 SYMBOL e M
4. 2 P H 1 g w Z i 1 3 Be co h o Mean 0 27 P 0 169 SD 2 34 7 E R squared 0 949 L Inter 0 12 95 Cl 0 193 to 0 433 Slope 1 01 95 C 0 968 to 1 05 s L Predicted Predicted Time Weighted residual error plot PMop data plot PMop data resid T Pmetrics User s Guide 23 oro 600 800 400 900 s00 T T 0 06 0 08 0 04 oro 600 soo 400 900 S00 04 02 oro 600 soo 400 900 s00 Vol plot PMfinal data Ka 2 Tlag oro 600 soo 400 900 S00 AIC BIC 2 x Log likelihood BIC ANC soz spz sez a a oser ozer Qv 800 700 600 500 400 300 200 800 700 600 500 400 300 200 Cycle Cycle Normalized Mean Gamma Lambda SO 00i S6 0 060 oz oz osez 800 700 600 500 400 300 Cycle Cycle Normalized Median Normalized SD 00 060 800 700 600 500 400 800 700 600 500 400 Cycle Cycle plot PMcycle data lt N i S 2 IU n p 0 eo 2 n e co hz GU t G Q Q plot versus N 0 1 for npde Distribution of npde E o J mno o O S E s gt 2 lt S S y 7 2 x S lt gt AD 5 9 A L a o J Z 9 o o coca o E T
5. P 1 etc are the natural logs of the variables Note that all variables are implicitly declared to be REAL 8 in Pmetrics model files Variables cannot start with the letters I through N as these are reserved hence the XKA notation CV 1 is the notation to define the value of covariate 1 as defined in the data csv file It will refer to the values contained in the first column after the C3 entry Of course CV 2 CV 3 etc can be used if these columns are defined in the data Note that any covariate relationship to any parameter may be described as the user wishes by mathematical equations and Fortran code allowing for exploration of complex non linear time dependent and or conditional relationships Pmetrics User s Guide 7 RATEIV 1 is the notation to indicate an infusion of input 1 typically drug 1 The duration of the infusion and total dose is defined in the data csv file Up to 7 inputs are currently allowed These can be used in the model file as RATEIV 1 RATEIV 2 etc The compartments for receiving the inputs of oral bolus doses are defined in subroutine SYMBOL This is a simple example which does not actually require differential equations in Pmetrics as we shall see However it gives the idea of how to specify a model Subroutine 2 OUTPUT In this subroutine users specify the equations that define the outputs of the model As for DIFFEQ all variables unless secondarily defined are referred to as P x w
6. POPDATA APR 11 ID EVID TIME iz ES c S c m z P o d OUT OUTEQ CO C1 C2 c3 COV GH 1 0 0 400 1 GH 0 0 5 0 42 1 0 01 0 1 0 0 GH 0 1 l 0 46 1 0 01 0 1 0 0 GH 0 2 l 2 47 1 0 01 0 1 0 0 GH 4 0 x GH 1 3 5 0 5 150 1 0 01 0 1 0 0 GH 0 5 12 0 55 1 0 01 0 1 0 0 GH 0 24 l 0 52 1 0 01 0 1 0 0 1423 1 0 0 400 1 1423 1 0 1 0 100 2 2 2 1423 0 1 l 99 1 0 01 0 1 0 0 1423 0 2 0 38 1 0 01 0 1 0 0 1423 0 2 l l 1 6 2 0 05 0 2 0 11 0 002 POPDATA APR 11 This is the header for the file and must be in the firstline It identifies the version ZID This field must be preceded by the symbol to confirm that this is the header row It can be numeric or character and identifies each individual All rows must contain an ID and all records from one individual must be contiguous Any subsequent row that begins with will be ignored which is helpful if you want to exclude data from the analysis but preserve the integrity of the original dataset or to add comment lines IDs should be 11 characters or less but may be any alphanumeric combination There can be at most 800 subjects per run EVID This is the event ID field It can be 0 1 or 4 Every row must have an entry 0 observation 1 input e g dose 2 3 are currently unused 4 reset where all compartment values are set to 0 and the time counter is reset to 0 This is useful when a
7. variation at standard concentrations You can use the Pmetrics function PMerrorPoly to choose the best set of coefficients that fit the data from the laboratory Alternatively if you have no information about the assay you can use the Pmetrics function ERRrun to estimate the coefficients from the data see 15 1 3 below Finally you can use a generic set of coefficients We recommend that as a start Co be set to half of the lowest concentration in the dataset and C1 be set to 0 15 C2 and C3 can be 0 15 1 If you choose 1 one set of coefficients for all subjects you will then be presented with 3 additional choices 15 1 1 Choice 1 Gamma is a scalar to capture additional process noise related to the observation including mis specified dosing and observation times In general well designed and executed studies will have data with gamma values approaching 1 Poor quality noisy data will result in gammas of 5 or more Choose this option if you wish to fix the assay error coefficients to values either in the data csv file or as specified in item 16 and to fix gamma to 1 15 1 2 Choice 2 Choose this option if you wish to fix the assay error coefficients to values either in the data csv file or as specified in item 16 but to estimate gamma based on the data This is the usual option Pmetrics User s Guide 16 16 17 18 19 20 21 22 23 24 25 26 15 1 3 Choice 3 Choose this option if y
8. Akaike Information and Bayesian Information Criteria and root mean squared errors RMSE for observed vs predictions from the population prior distribution and individual posterior distributions By specifying the option plot T observed vs predicted plots for all the models will be generated The option to generate residual plots of prediction errors described next can be specified with the additional switch resid T which is ignored if plot F As an option to plot PMop resid T will generate a residual plot instead of an observed vs predicted plot A residual plot consists of three panels 1 weighted residuals predicted observed vs time 2 weighted residuals vs predictions 3 a histogram of residuals with a superimposed normal curve if the option ref T is specified the Pmetrics User s Guide 25 default The mean of the weighted residuals expected to be 0 is reported along with the probability that it is different from 0 by chance Three tests of normality are reported for the residuals D Agostino Shapiro Wilk and Kolmogorov Smirnof An example is shown in the Plotting section Two more complex and time consuming options are also available the normalized prediction distribution error npde method of Brendel et al and the prediction discrepancy pd method of Mentr and Escolano 4 recently recast as a standardized visual predictive check SVPC by Wang and Zhang Both of these can be computed from the same simul
9. Output files from the simulator can be read into R using the SIMparse command see documentation in R There is a plot method plot PMsim for objects created by SIMparse Plotting There are numerous plotting methods included in Pmetrics to generate standardized but customizable graphical visualizations of Pmetrics data Taking advantage of the class attribute in R a single plot command is used to access all of the appropriate plot methods for each Pmetrics object class To access the R help for these methods you must query each method specifically to get details for p1ot will only give you the parent function PMop NPload plot PMop Plot observed vs population or ITload individual Bayesian posterior makeOP predicted data Optionally you can generate residual plots Pmetrics User s Guide 21 PMfinal NPload plot PMfinal Plot marginal final cycle parameter ITload value distributions makeFinal PMcycle NPload plot PMcycle Plots a panel with the following ITload windows 2 times the log likelihood at makeCycle each cycle gamma lambda at each cycle Akaike Information Criterion at each cyle and Bayesian Schwartz Information Criterion at each cycle the mean parameter values at each cycle normalized to starting values the normalized standard deviation of the population distribution for each parameter at each cycle and the normalized median parameter values at each cycle The
10. VPC External Validation Should you wish to use your population model to test how well it predicts a second population that is separate from that used to build the model i e externally validate your model you may do that in Pmetrics After completing an NPAG run place the same model file and instruction file located in the inputs subdirectory of the NPAG run whose model you are validating along with your new validating data file in your working directory Navigate to the output subdirectory of the NPAG run whose model you are validating and locate the DENxxxx usually DEN0001 file Place this file which contains the non parametric joint parameter value density information i e the population model also into your working directory So there should be four files in your working directory s model for file same as for model building NPAG run found in the inputs subdirectory s instruction file optional but if included it should be the same as from the model building NPAG run found in the inputs subdirectory e DENxxxx file density file from model building NPAG run found in the outputs subdirectory s data csv file new subjects for validation Beginning users Pmetrics User s Guide 26 1 Initiate an NPAG run in Pmetrics as usual without specifying the instruction or density files that you placed in the working directory e g NPrun model mymodel for 2 When answering the questions from NPAG to c
11. exponentiated value of P 4 Pmetrics User s Guide 9 multiplied by the third covariate specified in the data csv file for example an indicator of 1 if capsule 0 if liquid Subroutine 7 ANAL3 If N 1 in SYMBOL users can specify equations which can be used to estimate parameter values without the use of differential equations analytical models speeding up run times 10 000 fold Equations are defined for three basic models 1 one compartment 2 one dosing and one central compartment and 3 one dosing one central and one peripheral compartment The parameters are formulated as KA KE KCP KPC and V which cannot be changed However any secondary equations that define these primary variables are permissible KA DEXP P 1 KE DEXP P 2 CV 1 70 0 25 KCP 0 KPC 0 V DEXP P 5 CV 1 70 This specifies a model with an absorptive compartment elimination from a central compartment scaled allometrically to normalized body weight if this is the value of the first covariate no peripheral compartment by setting KCP and KPC to 0 and volume scaled to normalized body weight All parameters in this example are log transformed so must be exponentiated in the model For users who prefer parameterization in terms of clearances if P 1 is assigned to CL and P 2 is VOL then by the relationship K CL VOL you may specify KE P 1 P 2 All other parameters including fraction absorbed G
12. noted The choices are 1 to 7 corresponding respectively to 2129 5003 10007 20011 40009 and 80021 points If you choose 7 you then have an additional choice to select one or more multiples of 80021 points Enter the maximum number of cycles This can be 0 or greater If you enter an integer greater than 0 the engine will terminate at convergence or the number of cycles you specify whichever comes first Early in model exploration values of 10 to 100 can be useful with larger values later in model development In order to facilitate model comparison however we recommend using the same cycle limit for all early models e g 100 rather than choosing 10 for one and 100 for another If you enter 0 this is the way to test the predictive power of a model on an independent data set and a non uniform prior must be specified see 30 below This means that the engine will only calculate the individual Bayesian posteriors for the new subjects using the population joint density from a previous run as a Bayesian prior Information about convergence criteria Answer 1 or some other keystroke plus Return to acknowledge that you have read it In order to predict concentrations from a non parametric distribution you have the option to use the 1 mean 2 median or 3 mode of the Bayesian posterior distribution for each subject We typically use the median first and then the mean in a separate run and compare the differences Select the time interval t
13. on a subject by subject basis either those in the data csv file already the default values entered into the program or other values 18 For each output equation after you have selected the option in item 17 you will be prompted to supply the required information including the general default values for missing or overridden values in the data csv file 19 After assay error pattern and estimates are specified for all output equations enter the salt fraction of the drug usually 1 Salt fraction is the percentage of administered compound that contains active drug For example the mean salt fraction for theophylline is 0 85 This is not the same as bioavailability which is the fraction of drug absorbed after non parenteral administration e g oral compared to intravenous administration 20 Enter the grid point index This number corresponds to the number of grid support points which will initially fill the model parameter space The larger the number of random parameters to be estimated the more points Pmetrics User s Guide 19 21 22 23 24 25 26 27 28 29 30 31 32 33 are required The program will make a suggestion based on the number of random parameters in the model The more you choose the slower the run will be but results may improve It is reasonable to choose fewer points early in model exploration and increase in later phases or if poor model fits or lack of convergence are
14. the instructions coming from the keyboard or a file If this is the first time choose 0 for keyboard If you have a previously saved instruction file that you want to use choose 1 However typically if you have an instruction file that you want to use you will specify this in the R script with TTrun instr filename for automated analysis 7 1 If you selected keyboard input you will answer the following questions otherwise you will be prompted for the name of your instruction file and once loaded you will verify your previously supplied answers to the following questions What data input format will you use The standard format is the matrix block csv file so the answer should be 1 Working copy files are an older format The csv file is actually converted to these wrk files one file per subject prior to an IT2B run However some function will be lost in the Pmetrics package by using wrk input directly without a csv file such as the ability to plot raw subject data via the plot PMmatrix function Enter the name of your csv file now Enter the total number of unique subjects defined by ID in the csv file How many of the total number do you want to analyze Enter 1 if you want to analyze all of them 0 if you want to analyze a subset 11 1 If you entered 0 you will then choose 1 to include specific subjects or 2 to exclude specific subjects 11 2 Enter the inclusion or exclusion subject numbers not IDs in order
15. using a combination of numbers hyphens and commas For example 1 3 5 7 10 Press return and then enter 0 to conclude entry The program will then open the csv file and read the number of output equations reporting each subject as it is read This can take some time if it is a very large dataset Enter the initial boundary values for the random parameters in the model in the form min max followed by return The estimated mean for each parameter value distribution during the first iteration will be the median of the range specified in 13 You now have the option to specify the standard deviation for the parameter value distribution which by default is half of the range in 13 Choose 1 to accept this the usual answer or 0 to change it to something else expressed as a multiple of the range In IT2B the standard deviation SD of the observation obs is modeled by a polynomial equation with up to four terms Co Ci obs Cz obs Cs obs You will specify the coefficients Co C1 C2 and C3 You can now choose 1 if every subject has the same coefficients or 0 to use a unique set of coefficients for each subject The first case is the more usual when all samples from all subjects are analyzed in the same lab If samples are analyzed in different labs and you have the assay data from each lab then you would enter 0 This information should ideally come from the analytic lab in the form of inter run standard deviations or coefficients of
16. AG the load functions can be used to load the above Rdata files after a successful run The report functions are automatically run at the end of a successful run and these will generate an HTML page with summaries of the run as well as the Rdata files and other objects The default browser will be automatically launched for viewing of the HTML report page Within Pmetrics there are also functions to manipulate data csv files and process and plot extracted data Manipulate data csv files PMreadMatrix PMcheckMatrix PMwriteMatrix PMmatrixRelTime PMwrk2csv Process data makeAUC makeCov makeCycle makeFinal makeOP makeErrorPoly Pmetrics User s Guide 4 e Plot data plot PMcov plot PMcycle plot PMfinal plot PMmatrix plot PMop plot PMsim Model selection and diagnostics PMcompare plot PMop with residual option PMdiag Again all functions have extensive help files and examples which can be examined in R by using the help command or command syntax General Workflow The general Pmetrics workflow for IT2B and NPAG is shown in the diagram below Data csv file Model for file aie Preparation program Instruction file Ww Engine program Output R is used to specify the working directory containing the data csv and model for files and possibly an instruction file if IT2B or NPAG have been run previously Through the batch file generated by R the preparation program is compiled and executed If th
17. ETFA initial conditions GETIX and lag times GETTLAG are also permissible with analytical models As for differential equation models any covariate relationship to any parameter may be described as the user wishes by mathematical equations and Fortran code allowing for exploration of complex non linear time dependent and or conditional relationships How to use R and Pmetrics The first thing to do is to ensure that appropriate Pmetrics model for and data csv files are in the working directory R can be used to help prepare the data csv file by importing and manipulating spreadsheets e g read csv The Pmetrics function PMcheckMatrix can be used to check an R dataframe that is to be saved as a Pmetrics data csv file for errors The function PMwriteMatrix can be used to write the R data object in the correct format for use by IT2B NPAG or the Simulator If this is a first time run the R commands to run IT2B or NPAG are as follows Recall that the character is a comment character library Pmetrics setwd working directory NPrun for NPAG dor ITrun for IT2B The first line will load the Pmetrics library of functions The second line sets the working directory to the specified path The third line generates the batch file to run NPAG or IT2B and saves it to the working directory If this is the very first time ever to run NPAG or IT2B the user will be first prompted in R to specify the For
18. T T 2 4 0 1 2 3 2 4 0 1 2 3 Sample quantiles npde npde o o o 9 6 E e J apo apa 2 Lass S N o0 e E a 2 o ER o 843 o 9 856 e eo i 5 o 8 27 T s 8 o 7 B wo a o 888 d a E o o o ge Q 918 8 8 a 9 S L 6 o 9 5 amp g 8 ES S A amp 98 o of 9 2 eol Bp gt 571808 8 8 8 7 o d o Q o E oo o o 8 Daw 8 88 L 8 3 T 9 o e ri Q o o 8 o o 9 656 8 o 8 8 d oo o o e l oo 00 aw o oo o o 9 SJ i l 120 125 130 135 140 145 4 6 8 10 120 125 130 135 140 145 Time Predicted Time plot PMdiag data 20 Output T T 120 125 130 135 140 Time h plot PMsim data 1 Model Diagnostics Internal Validation Several tools are available in Pmetrics to assist with model selection The simplest methods are using PMcompare 1 and plot PMop via the plot command for a PMop object made by makeOP or by using NPload or ITload after a successful run All these functions are carefully documented within R and accessible using the command or help command syntax To compare models with PMcompare simply enter a list of two or more PMetrics data objects These should be of the NPAG or IT2B class made either by using NP1oad ITload or NPparse ITparse Although it is possible to compare models of mixed classes the validity of this is dubious The return object will be a data frame with summaries of each model and key metrics such as log likelihood final cycle
19. all populations with analytic solutions or days for large populations with complex differential equations At the end of a successful run the results will be Pmetrics User s Guide 20 automatically parsed and saved to the output directory Your default browser will launch with a summary of the run Simulator Runs The simulator is run from within R No batch file is created or terminal window opened However the actual simulator is a Fortran executable compiled and run in an OS shell It is documented with an example within R You can access this by using the help SIMrun or SIMrun commands from R In order to complete a simulator run you must include a data csv file and a model file in the working directory The structure of these files is identical to those used by IT2B and NPAG The data csv contains the template dosing and observation history as well as any covariates Observation values the OUT column for EVID 0 events can be any number they will be replaced with the simulated values You can have any number of subject records within a data csv file each with its own covariates if applicable Each subject will cause the simulator to run one time generating as many simulated profiles as you specify from each template subject This is controlled from the SIMrun command with the nsub and nsim arguments The first specifies how many subjects are in the data csv file corresponding to the number of simulator runs The second specifies h
20. ary values for the random parameters in the model in the form min max followed by return 16 Select how you would like to model assay observation error for each output equation You have four choices In all four the standard deviation SD of the observation obs is modeled by a polynomial equation with up to four terms Co Ci obs C2 obs Cs obs You will specify the coefficients Co Ci C2 and C3 This information should ideally come from the analytic lab in the form of inter run standard deviations or coefficients of variation at standard concentrations You can use the Pmetrics function PMerrorPoly to choose the best set of coefficients that fit the data from the laboratory Alternatively if you have no information about the assay you can use the Pmetrics function ERRrun to estimate the coefficients from the data Finally you can use a generic set of coefficients We recommend that as a start Co be set to half of the lowest concentration in the dataset and C1 be set to 0 15 C2 and C3 can be 0 16 1 Error model 1 SD Choose this option if you have already run IT2B and multiplied your assay error polynomial by gamma see next option for a description of gamma 16 2 Error model 2 SD gamma Gamma is a scalar to capture additional process noise related to the observation including mis specified dosing and observation times In general well designed and executed studies will have data with gamma values approach
21. ation The basic idea is that each subject in the population serves as a template for a simulation of 1000 further profiles using the population structural model and parameter values joint probability distribution i e together the population model The simulated profiles are compared to the observed data and the npde and pd are generated The command to generate a PMdiag object is PMdiag which is documented in R The same model file and data csv file used in the NPAG or IT2B run must be in the working directory prior to executing the command Because of the extensive simulations involved execution of this command can be slow if the population is large the model complex the time horizon long and or the number of observations to be simulated per profile is large There are print and plot methods for PMdiag objects print PMdiag and plot PMdiag both of which are also documented within R An example of a PMdiag plot is shown in the Plotting section Note that simulation from a population model can be a fickle thing which may lead to errors when trying to execute this command Parameter value distributions in linear space run the risk of simulating extreme or even inappropriately negative parameter values which can in turn lead to simulated observations far beyond anything corresponding to possible reality In Pmetrics the method used to simulate from a prior NPAG non parametric distribution is the split method described above in the Simulator s
22. atrix for each random parameter distribution Matrix of boundaries for random parameter values For NPAG this is specified by the user prior to the run for IT2B itis calculated as a user specified multiple of the SD for the parameter value distribution cycle class PMcycle list names Vector of names of the random parameters l Matrix of cycle number and 2 Log likelihood at each cycle gamlam A matrix of cycle number and gamma or lambda at each cycle see item 16 under NPAG Runs below for a discussion of gamma and lambda mean A matrix of cycle number and the mean of each random parameter at each cycle normalized to initial mean s A matrix of cycle number and the standard deviation of each random parameter at each cycle normalized to initial standard deviation median A matrix of cycle number and the median of each random parameter at each cycle normalized to initial standard deviation Pmetrics User s Guide 13 aic A matrix of cycle number and Akaike Information Criterion at each cycle bic A matrix of cycle number and Bayesian Schwartz Information Criterion at each cycle post class data frame Subject identification NPAG only Bayesian posterior predictions for each output equation based on mean median and mode as specified by the user and with frequency also specified by the user in the run instructions see NPAG Runs below items 23 and 24 mex eM eame 00000 NPdata ITdata Raw data
23. cess the data such as the following plot final plot cycle plot opSpop1 plot opSpost1 plot op popl resid T plot opSpost1 resid T Of course the full power of R can be used in scripts to analyze data but these simple statements serve as examples We suggest that the R script for a particular project be saved into a folder called Rscript in the working directory Folders are not be moved by the batch file Within the script number runs sequentially and use comments liberally to distinguish runs as shown below library Pmetrics Run 1 Ka Kel V all subjects setwd working directory NPrun setwd output directory 1 NPload 1 Run 2 Ka Kel V allometric scaling all subjects setwd working directory NPrun model model2 for instr file assumes an appropriate instruction file from a earlier run possibly edited setwd output directory 2 NPload 2 Remember in R that the command example function will provide examples for the specified function Most Pmetrics functions have examples Pmetrics User s Guide 11 Pmetrics Data Objects After a successful IT2B or NPAG run an R datafile is saved in the output subdirectory of the newly created time and date stamped folder in the working directory After IT2B this file is called IT2Bout Rdata and after NPAG it is called NPAGout Rdata As mentioned in the previous section these data files can be load
24. default is to omit the first 10 of cycles as a burn in from the plots PMcov makeCov E L PMcov Plots the relationship between any two columns of a PMcov object el plot PMmatrix PMreadMatrix plot PMmatrix Plots raw time observation data from a data csv file read by the PMreadMatrix command with a variety of options including joining observations with line segments including doses overlaying plots for all subjects or separating them including individual posterior predictions post objects as described above color coding according to groups and more PMsim SIMparse eplot PMslm Plots simulated time concentration profiles overlaid as individual curves or summarized by customizable quantiles e g 5th 25th 50th 75th and 95th percentiles Inclusion of observations in a population can be used to return a visual and numerical predictive check Pmetrics User s Guide 22 PMdiag PMdiag plot PMdiag Plots an npde qqnorm npde histogram npde vs time npde vs prediction and standardized visual predictive check to visualize results of simulation based internal model diagnostics accessed with the PMdiag command Examples of Pmetrics plots 120 125 130 135 140 Time h plot PMmatrix data D Agostino P 0 267 S e Shapiro Wilk P 0 00 Kolmogorov Smirnoff HR 0 S i A lt
25. ection Division of the covariance matrix over all the multi variate distributions in the multi modal multi variate distribution mitigates the problem of extreme values When using an IT2B parametric unimodal multi variate distribution it is likely that extreme values will be simulated Mitigating techniques include transformation of the model into log space or switching to an NPAG prior There is no command in Pmetrics to automatically generate the simulations necessary for a Visual Predictive Check VPC in contrast to the methods described above VPCs are cumbersome when models include covariates or have heterogeneous dosing sampling regimens among subjects in the population It is nonetheless possible to obtain a VPC and numerical predictive check NPC using the plot PMsim command via plot on a PMsim object made with SIMparse If an observed vs predicted PMop object made with makeOP is passed to plot PMsim with the obspred argument the observed values will be overlaid upon simulated profiles if possible and an NPC will be returned in addition to the plot The NPC is simply a binomial test for the percentage of observations less than the quantiles specified by the probs argument 0 05 0 25 0 5 0 75 0 95 by default The simulations for the VPC must be done manually using SIMrun and extracted with SIMparse priorto plotting them It is up to the user to decide if the study population and model is homogeneous enough to justify a
26. ed by ensuring that the correct output folder is set as the working directory and then using the Pmetrics commands ITload or NPload Both commands load their respective Rdata files into R making the contained objects available for plotting and other analysis Objects loaded by ITload and NPload op class PMop data frame pop1 post1 Population and posterior predictions for each output equation i e 1 2 sabe O sobe Observation 000 pred Prediction based on median of population or p pop posterior parameter value distributions block Dosing block usually 1 unless data file contains EVID 4 dose reset events in which case each such reset within a given ID will increment the dosing block by 1 for that ID obsSD Calculated standard deviation error of the observation based on the assay error polynomial final class PMfinal list popPoints NPAG only Data frame of the final cycle joint population density of grid points with column names equal to the name of each random parameter plus prob for the associated probability of that point Pmetrics User s Guide 12 popMean The final cycle mean for each random parameter distribution popCor The final cycle correlation matrix for each random parameter distribution popMedian The final cycle median for each random parameter distribution gridpts NPAG only The initial number of support points ab l d popCov The final cycle covariance m
27. er coefficients IT2B Runs When IT2B is launched by the R generated batch script without an instruction file the user must answer the following questions prompted by the program There is an opportunity at the end to review and correct answers so making a mistake is not a reason necessarily to abort Answers to the following questions can be saved in an instruction file which can be used for future runs Instruction files are simply text files with specific entries which can be modified directly by advanced users Less advanced users can initiate an IT2B run from Pmetrics without specifying an instruction file even if one exists and then load it as described in item 7 below You will then have the opportunity to see previous responses to each question and modify them if desired Note that IT2B and NPAG instruction files are NOT interchangeable 1 Arethe files in the current directory The answer should always be 1 2 Doananalysis or examine results from prior run Almost always 1 3 Awarning about using the correct current model format Press 1 or some other key and then return 4 Enterthe name ofthe Fortran model file 5 For each of the parameters in the model file specify whether it is to be random estimated or fixed not estimated Pmetrics User s Guide 15 10 11 12 13 14 15 What are the ordinary differential equation solver tolerances Accept the default value by choosing 1 unless advanced Are
28. eyboard input you will answer the following questions otherwise you will be prompted for the name of your instruction file and once loaded you will verify your previously supplied answers to the following questions 10 What data input format will you use The standard format is the matrix block csv file so the answer should be 1 Working copy files are an older format The csv file is actually converted to these wrk files one file per subject prior to an NPAG run However some function will be lost in the Pmetrics package by using wrk input directly without a csv file such as the ability to plot raw subject data via the plot PMmatrix function 11 Enter the name of your csv file now Pmetrics User s Guide 18 12 Enter the total number of unique subjects defined by ID in the csv file 13 How many of the total number do you want to analyze Enter 1 if you want to analyze all of them 0 if you want to analyze a subset 13 1 If you entered 0 you will then choose 1 to include specific subjects or 2 to exclude specific subjects 13 2 Enter the inclusion or exclusion subject numbers not IDs in order using a combination of numbers hyphens and commas For example 1 3 5 7 10 Press return and then enter 0 to conclude entry 14 The program will then open the csv file and read the number of output equations reporting each subject as it is read This can take some time if it is a very large dataset 15 Enter the bound
29. fies gastric pH for example Subroutine 5 GETIX In this subroutine users can specify the initial conditions of a given compartment By default all compartments are set to X 0 The syntax is as follows In the first case the initial condition for compartment 1 becomes another variable P 4 to fit in the model given the observed data In the second case the initial condition for compartment 2 becomes the value of covariate 4 multiplied by the current estimate of P 3 during each iteration This is useful when a subject has been taking a drug as an outpatient and comes in to the lab for PK sampling with measurement of a concentration immediately prior to a witnessed dose which is in turn followed by more sampling In this case CV 4 or any other covariate can be set to the initial measured concentration and if P 3 is the volume of compartment 2 the initial condition amount in compartment 2 will now be set to the measured concentration of drug multiplied by the estimated volume for each iteration until convergence Subroutine 6 GETTLAG In this subroutine users can specify a variable to be fitted as the lag time for a given input By default all lag times are set to 0 The syntax is as follows TLAG 1 P 4 TLAG 2 DEXP P 4 CV 3 In the first case the lag time for input 1 becomes another variable P 4 to fit in the model given the observed data In the second case the lag time for input 2 becomes the
30. for each function in R will provide further details on the arguments defaults and output of each command Pmetrics User s Guide 14 makeAUC Make a data frame of class PMauc containing subject ID makeAUC and AUC from a variety of inputs including objects of PMop PMsim or a suitable data frame makeCov Generate a data frame of class PMcov with subject makeCov specific covariates extracted from the data csv file mean median NPAG IT2B or mode NPAG of Bayesian posterior parameter value distributions and AUC from specified start to end times This object can be plotted and used to test for covariates which are significantly associated with model parameters makeCycle Create a PMcycle object described in the previous section makeCycle Create a PMfinal object described in the previous section Create a PMop object described in the previous section makeErrorPoly This function plots first second and third order makeErrorPoly polynomial functions fitted to pairs of observations and associated standard deviations for a given output assay In this way the standard deviation associated with any observation may be calculated and used to appropriately weight that observation in the model building process Observations are weighted by the reciprocal of the variance or squared standard deviation Output of the function is a plot of the measured observations and fitted polynomial curves and a list with the first second and third ord
31. he ordinary differential equation solver to fit the model to the data 2 Zero This implies that the model is defined as an analytic equation or equations in OUTPUT 3 1 which tells Pmetrics to use analytic equations as defined in the subroutine ANAL3 In this case any differential equations in DIFFEQ are ignored As an example N 2 which corresponds to the number of differential equations in DIFFEQ above Again N cannot be greater than 20 This statement must start in column 8 The third section of SYMBOL is for defining the random and fixed variable names in the model It is headed by a statement declaring the number of variables followed by the variable names Pmetrics User s Guide 8 NP 3 PSYM 1 Ka PSYM 2 Kel PSYM 3 Vol Note the use of single quotes Variable names must be no greater than 11 characters and not contain any spaces These statements must also begin in column 8 The number of parameters must be between 2 and 32 with at most 30 random or 20 fixed Subroutine 4 GETFA In this subroutine users can specify which parameter if any is to be fitted as the bioavailability FA of a given input By default all inputs have an FA 1 The syntax is as follows Now the FA for input 1 is assigned to variable 4 Of course secondary expressions are also permissible FA 3 DEXP P 7 CV 3 Here FA for input 3 is assigned to the product of e and CV 3 CV 3 might be a covariate that speci
32. here x is the number of the variable The variables are assigned in subroutine SYMBOL An example of an output equation follows Y 1 X 2 P 3 Y 1 is the Fortran notation for Output 1 Alternatively one could specify more meaningful parameter names and secondary equations V DEXP P 3 CV 1 Y 1 X 2 V Since V does not start with the letters I through N it is an acceptable variable name CV 1 is the same as defined before the value of covariate 1 at the appropriate time as specified in the data csv file There can be a maximum of 6 outputs They are referred to as Y 1 Y 2 etc Subroutine 3 SYMBOL In this subroutine there are up to three areas which need attention from the user In the first section the user can define compartments that will receive bolus doses By default these compartments correspond to the input number so that compartment 1 will receive input 1 etc This can be reassigned using the syntax NBCOMP 1 2 which assigns compartment 2 to receive any bolus doses from input 1 A dose is defined as a bolus ina data csv file by setting duration to 0 Any value greater than 0 implies an infusion which serves as an input as defined in the DIFFEQ subroutine In the second section of SYMBOL the user specifies the number of compartments in the model This can be one of three kinds 1 A positive integer equal to the number of XP equations in DIFFEQ This will instruct Pmetrics to use t
33. ing 1 Poor quality noisy data will result in gammas of 5 or more If you choose this option you then can specify the starting value of gamma Good values are 1 for high quality data 3 for medium and 5 or 10 for poor quality 16 3 Error model 3 SD lamda Lamda is an alternative additive model to capture process noise rather than the multiplicative gamma model Good starting values for lambda are 1 times Co for good quality data 3 times Co for medium and 5 or 10 times Co for poor quality Note that Co should generally not be 0 as it represents machine noise e g HPLC or mass spectrometer that is always present 16 4 Error model 4 SD gamma This model is rarely used and is equivalent to specifying a model with Co only i e a constant error regardless of concentration 17 Once you select the assay error model for each output equation you are then offered four more options on which assay error polynomial coefficients to use Choices 1 and 2 are the most commonly used 17 1 Choice 1 The default Use coefficients in the subject record Co Ci C2 C3 in the data csv file and if missing use the default values to be entered in the program item 18 below 17 2 Choice 2 Use the default values to be entered in the program for all subjects regardless of what is in the data csv file 17 3 Choice 3 To multiply data csv values and default entered values by a fixed gamma and use them 17 4 Choice 0 Specify coefficients
34. is is the first run the user will answer several questions about the run supplying necessary information This information can be saved as an instruction file for future runs The batch file will then compile and execute the engine file which will generate several output files upon completion Finally the batch file will call the R script to generate the summary report and several data objects including the IT2Bout Rdata or NPAGout Rdata files which can be loaded into R subsequently using ITload or NPload On subsequent runs the user may specify the name of the instruction file as an argument to ITrun or NPrun This will result in fully automated execution with no further input from the user required AIl input files data model and instruction are text files which can be edited directly by advanced users Pmetrics User s Guide 5 Pmetrics Input Files Data csv Files Pmetrics accepts input as a spreadsheet matrix format It is designed for input of multiple records in a concise way Files are in comma separated values csv format Examples of programs that can save csv files are any text editor e g TextEdit on Mac Notepad on Windows or spreadsheet program e g Excel Click on hyperlinked items to see an explanation IMPORTANT The order capitalization and names of the header and the first 12 columns are fixed All entries must be numeric with the exception of ID and for non required placeholder entries
35. lable coefficients these cells may be left blank or filled with as a placeholder Any column after the assay error coefficients is assumed to be a covariate one column per covariate for Files iles for Pmetrics are Fortran text files The current version is TSMULTI The safest thing to do is to edit old s of the model files to make new versions Fortran is very fussy when it comes to spacing one column off can result in a failed run There a re seven subroutines that can be modified by the user all by writing simple Fortran code Model files have comment lines to indicate where user code should go Subroutine 1 DIFFEQ In this subroutine users specify the differential equations that define the model There can be a maximum of 20 such equations All parameters unless secondarily defined are referred to as P x where x is the number of the variable The parameters are assigned in subroutine SYMBOL An example of differential equations follows XP 1 P 1 X 1 XP 2 P 1 X 1 P 2 X 2 XP is the Fortran notation for dx dt P 1 is parameter 1 and X 1 is the amount in compartment 1 Alternatively one could specify more meaningful parameter names and secondary equations as shown below XKA DEXP P 1 XKEL DEXP P 2 CV 1 70 0 25 XP 1 XKA X 1 XP 2 RATEIV 1 XKA X 1 XKEL X 2 DEXP is the Fortran code for double precision eJ and in this example
36. lerances Accept the default value by choosing 1 unless advanced 7 NPAG creates a temporary instruction file in case something happens so that instructions entered to date can be recovered Accept the default by choosing 1 You can save with a meaningful filename later If it already exists you will be asked if you wish to overwrite the file which is usually the thing to do 8 If you have previously run IT2B you can automatically import the suggested parameter ranges Choose 1 to do this and 0 to run NPAG without a previous IT2B run 8 1 If you choose option 1 you must specify the name of a FROMxxxx file typically FROM0001 that you can find in an output directory after a successful IT2B run Note that the program will then assume that you are using the same wrk files that IT2B made from your data csv file It is strongly recommended to override this and specify a data csv file At this point you will jump to 10 below and continue on however your answers will often be supplied by the instructions contained in the FROMxxxx file and you need merely confirm them 9 Are the instructions coming from the keyboard or a file If this is the first time choose 0 for keyboard If you have a previously saved instruction file that you want to use choose 1 However typically if you have an instruction file that you want to use you will specify this in the R script with NPrun instr filename for automated analysis 9 1 Ifyou selected k
37. ling and Monte Carlo simulation study of the pharmacokinetics and antituberculosis pharmacodynamics of rifampin in lungs Antimicrob Agents Chemother 2009 53 7 2974 2981 2 D Agostino R Transformation to Normality of the Null Distribution of G 1 Biometrika 1970 57 3 679 681 3 Brendel K Comets E Laffont C Mentr F Evaluation of different tests based on observations for external model evaluation of population analyses J Pharmacokinet Pharmacodynam 2010 37 1 49 65 4 Mentr F Escolano S Prediction discrepancies for the evaluation of nonlinear mixed effects models J Pharmacokinet Pharmacodyn 2006 33 3 345 367 5 Wang DD Zhang S Standardized Visual Predictive Check Versus Visual Predictive Check for Model Evaluation Clin Pharmacol 2011 Pmetrics User s Guide 27
38. mpted for the name of a file that contains the prior density This will be DEN0001 unless you have changed its name It will be found in the output directory of a prior NPAG run The model used to generate the DENO0001 file must be exactly the same as the current model including parameter boundaries However this option is useful to specify a non uniform density for two reasons The first is to test the predictive power of a model on a new set of subjects Do this in combination with setting the number of cycles to 0 see 21 above The second use for a non uniform prior is to continue a previous run For example if you only set the number of cycles to 50 to get a rough idea of model fit you may continue where you left off by specifying the DEN0001 file from the 50 cycle run and continuing with as many additional cycles as you specify in item 21 So in the example if you specify 100 cycles in 21 the total number of cycles will be 50 100 lt 150 Enter 1 if you wish to save all the instructions in an instruction file If you do this you can specify this instruction file in Pmetrics by using the NPrun instr yourfile option and including yourfile in the working directory with the model for file and the data csv file Some output will print to the terminal window which contains information that you can ignore while running NPAG from Pmetrics Press 1 followed by return to begin the NPAG analysis The NPAG run can complete in seconds for sm
39. n individual has multiple sampling episodes that are widely spaced in time with no new information gathered TIME This is the elapsed time in decimal hours since the first event It is not currently clock time e g 21 30 although this is planned Every row must have an entry and within a given ID rows must be sorted chronologically earliest to latest DUR This is the duration of an infusion in hours If EVID 1 there must be an entry otherwise it is ignored For a bolus i e an oral dose set the value equal to 0 Pmetrics User s Guide 6 DOSE INPUT OUT OUTEQ CO C1 COV Model Model f version This is the dose amount If EVID 1 there must be an entry otherwise it is ignored This defines which input i e drug the DOSE corresponds to Inputs are defined in the model file This is the observation or output value If EVID 0 there must be an entry if missing this must be coded as 99 It will be ignored for any other EVID and therefore can be There can be at most 150 observations for a given subject This is the output equation number that corresponds to the OUT value Output equations are defined in the model file C2 C3 These are the coefficients for the assay error polynomial for that observation Each subject may have up to one set of coefficients per output equation If more than one set is detected for a given subject and output equation the last set will be used If there are no avai
40. ness Choose 1 each time to indicate correct entries or 0 to change them Enter 1 if you wish to save all the instructions in an instruction file If you do this you can specify this instruction file in Pmetrics by using the ITrun instr filename option and including filename in the working directory with the model for file and the data csv file The program will cycle through your subject records again to extract all relevant information This can take some time if the population is large or individual records are long Specify the nature of each covariate in the data csv file Enter 1 if it is to be considered constant between measurements e g gender or 2 if values should be extrapolated between observations e g creatinine clearance If you chose unique assay error coefficients for each subject in 15 above you will now specify whether you wish to use coefficients found in the data csv file choice 1 the general coefficients specified in 16 above choice 2 or a different set that you enter manually now choice 0 in the form Co C1 C2 Cs Some output will print to the terminal window which contains information that you can ignore while running IT2B from Pmetrics Press 1 followed by return to begin the IT2B analysis Pmetrics User s Guide 17 27 The IT2B run can complete in seconds for small populations with analytic solutions or days for large populations with complex differential equations At the end of a
41. ntax JTrun ITparse ITload ITreport ERRrun NPrun NPparse NPload NPreport SIMrun SIMparse For IT2B and NPAG the run functions generate batch files which when executed launch the software programs to do the analysis ERRrun is a special implementation of IT2B designed to estimate the assay error polynomial coefficients from the data when they cannot be calculated from assay validation data using makeErrorPoly supplied by the analytical laboratory The batch files contain all the information necessary to complete a run tidy the output into a date time stamped directory with meaningful subdirectories extract the information generate a report and a saved RData file of parsed output which can be quickly and easily loaded into R On Mac Unix systems the batch file will automatically launch in a Terminal window On Windows systems the batch file must be launched manually In both cases the execution of the program to do the actual model parameter estimation is independent of R so that the user is free to use R for other purposes For the Simulator the run function will execute the program directly within R For all programs the parse functions will extract the primary output from the program into meaningful R data objects For IT2B and NPAG this is done automatically at the end of a successful run and the objects are saved in the output subdirectory as IT2Bout Rdata or NPAGout Rdata respectively For IT2B and NP
42. nvironments In order to run Pmetrics a Fortran compiler is required The freely available gfortran compiler is available for Mac and Windows users For Mac users Link to http gcc gnu org wiki GFortranBinaries click the MacOS link under Binaries available for gfortran at the top of the page Select the appropriate source to download based on your system For Windows users Linkto http tdm gcc tdragon net download and select the installer download link at the very top Once you launch the downloaded installer accept all the defaults EXCEPT BE SURE to choose fortran under the components dialog window JUST BEFORE YOU CLICK INSTALL Do this by selecting the next to gcc at the top of the list and then scrolling down to check the fortran box If you don t do this you will not get gfortran by default A text editor that can link to R is useful for saving scripts Both the Windows and Mac versions of R have rudimentary text editors that are stable and reliable Numerous other free and paid editors can also do the job and these can be located by searching the internet We prefer Rstudio which is available for Mac and Windows and is free We have also used the freely available Komodo with the SciViews K add on although it is somewhat difficult to set up proper communication with R Once working however it is an intuitive and easy yet powerful scripting environment Email us if you need help at contact lapk org Pmetrics is cu
43. o generate predicted concentrations Additionally there will also be predictions made at the time of each observation In general for most models predictions every 12 minutes provide sufficient granularity Smaller values can result in very large files for big populations Select the default MIC to be used for AUC MIC ratio reporting This should generally be set to the default of 1 by choosing 1 You can always extract the AUC later and divide it by any MIC you choose Enter the value in hours that you want for calculations of AUCs from predicted concentration profiles The default is 24 hours which you can accept by entering 1 or 0 if you want to specify a different interval You are now offered the opportunity to check all of your entries for correctness Choose 1 each time to indicate correct entries or 0 to change them The program will cycle through your subject records again to extract all relevant information This can take some time if the population is large or individual records are long Specify the nature of each covariate in the data csv file Enter 1 if it is to be considered constant between measurements e g gender or 2 if values should be extrapolated between observations e g creatinine clearance For the prior density choose 1 if it is to be uniform This means that the initial grid points will be evenly distributed with equal probability within the boundaries specified in 15 above If you choose 0 you will be pro
44. on PK modeling program It is generally used to estimate parameter ranges to pass to NPAG It will estimate values for population model parameters under the assumption that the underlying distributions of those values are normal or transformed to normal e NPAG is the Non parametric Adaptive Grid software It will create a non parametric population model consisting of discrete support points each with a set of estimates for all parameters in the model plus an associated probability weight of that set of estimates There can be at most one point for each subject in the study population There is no need for any assumption about the underlying distribution of model parameter values The simulator is a semi parametric Monte Carlo simulation software program that can use the output of IT2B or NPAG to build randomly generated response profiles e g time concentration curves for a given population model parameter estimates and data input Simulation from a non parametric joint density model i e NPAG output is possible with each point serving as the mean of a multivariate normal distribution weighted according to the weight of the point The covariance matrix of the entire set of support points is divided equally among the points for the purposes of simulation Pmetrics has groups of R functions named logically to run each of these programs and to extract the output These are extensively documented within R by using the help command or command sy
45. ontrol the run indicate that the instructions will come from a previous run and provide the name of your instruction file Edit the loaded information to reflect the name of the new validating data file the number of new subjects to change the maximum cycle number to 0 to specify a non uniform prior and the file name of the density file e g DEN0001 See NPAG Runs above for more information on this Zero cycles will simply use the population model to calculate the posterior distributions for each new subject without updating the population model by running any additional iterations of NPAG Complete the NPAG run as usual 4 Load the results with NP1oad and plot etc as usual Advanced users 1 Directly edit the instruction file to include the name of the validating data file the number of subjects and to change the cycle number MAXCYC to 0 Zero cycles will simply use the population model to calculate the posterior distributions for each new subject without updating the population model by running any additional iterations of NPAG 2 Initiate an NPAG run in Pmetrics as usual but with an additional argument to specify the density file which will serve as a non uniform prior e g NPrun model mymodel for instr mynewintstr inx denfile DENO0001 Complete the NPAG run as usual 4 Load the results with NP1oad and plot etc as usual References 1 Goutelle S Bourguignon L Maire PH et al Population mode
46. ou wish to estimate the assay error coefficients based on your data for use in future runs Although you can access this option by using either ITrun or ERRrun in R the instruction files that you save and the generated output files will be different Therefore we recommend that if you intend to choose this option use ERRrun in R which will generate an ASS0001 file that contains the estimates for Co Ci C2 and C3 You can then include this file in the working directory along with a model for file and a data csv file to do an IT2B run supplying the file name in 4416 1 below 15 2 If you choose 0 unique coefficients for each subject you will be presented with two choices 15 2 1 Choice 1 Fix gamma to 1 See the discussion above in 15 1 1 15 2 2 Choice 2 Estimate gamma based on the data 15 2 3 You now need to specify where to obtain the values for Co Ci C2 C3 either from the data csv file and from the entry in 16 Choice 1 or on an individual basis during the IT2B run1 Choice 0 Enter the values for Co Ci C2 C3 that will be used for all patients who do not have values associated with them in the data csv file 16 1 You have the option of entering a file name that contains the output of a previous estimation generated by choosing 3 in 15 1 3 above Usually this file will be called ASS0001 and it must be in the working directory After assay error pattern and estimates are specified for all output eq
47. ow many profiles are to be generated from each nsub Simulation from a non parametric prior distribution from NPAG can be done in one of two ways The first is simply to take the mean standard deviation and covariance matrix of the distribution and perform a standard Monte Carlo simulation The second way is what we call semi parametric and was devised by Goutelle et al In this method the non parametric support points in the population model each a vector of one value for each parameter in the model and the associated probability of that set of parameter values serve as the mean of one multi variate normal distribution in a multi modal multi variate joint distribution The weight of each multi variate distribution is equal to the probability of the point The overall population covariance matrix is divided by the number of support points and applied to each distribution for sampling Output from the simulator will be controlled by further arguments to SIMrun If makecsv is not missing a csv file with the simulated profiles will be created with the name as specified by makecsv otherwise there will be no csv file created If outname is not missing the simulated values and parameters will be saved in a txt file whose name is that specified by outname otherwise the filename will be simout In either case integers 1 to n n a nsub will be appended to outname or simout e g simout1 txt simout2 txt
48. rrently distributed as a source file To install Pmetrics from the R console use the command install packages file choose and navigate when prompted to the folder in which you placed the Pmetrics tar gz source file Pmetrics will need the following packages for some functions chron rgl and RZHTML However you do not have to install these if you do not already have them in your R library They should automatically be downloaded and installed the first time you use a Pmetrics function that requires them but if something goes awry such as no internet connection or busy server you can do this manually What This Manual Is Not We assume that the user has familiarity with population modeling and R and thus this manual is not a tutorial for basic concepts and techniques in either domain We have tried to make the R code simple regular and well documented A very good free online resource for learning the basics of R can be found at http Pmetrics User s Guide 3 www statmethods net index html We recognize that initial use of a new software package can be complex so please feel free to contact us at any time contact lapk org This manual is also not intended to be a theoretical treatise on the algorithms used in IT2B or NPAG For that the user is directed to our website at www lapk org Pmetrics Components There are three main software programs that Pmetrics controls JT2B is the ITerative 2 stage Bayesian parametric populati
49. successful run the results will be automatically parsed and saved to the output directory Your default browser will launch with a summary of the run NPAG Runs When NPAG is launched by the R generated batch script without an instruction file the user must answer the following questions prompted by the program There is an opportunity at the end to review and correct answers so making a mistake is not a reason necessarily to abort Answers to the following questions can be saved in an instruction file which can be used for future runs Instruction files are simply text files with specific entries which can be modified directly by advanced users Less advanced users can initiate an NPAG run from Pmetrics without specifying an instruction file even if one exists and then load it as described in item 7 below You will then have the opportunity to see previous responses to each question and modify them if desired Note that IT2B and NPAG instruction files are NOT interchangeable Are the files in the current directory The answer should always be 1 Do an analysis or examine results from prior run Almost always 1 A warning about using the correct current model format Press 1 or some other key and then return Enter the name of the Fortran model file Uh qc E op For each of the parameters in the model file specify whether it is to be random estimated or fixed not estimated 6 What are the ordinary differential equation solver to
50. tran compiler in use Pmetrics User s Guide 10 NOTE On Mac systems the batch file will be automatically launched in a Terminal window On Windows systems the batch file must be launched manually by double clicking the np run bat or it run bat file in the working directory ITrun and NPrun both return the full path of the output directory to the clipboard Therefore the next command to enter in the script will change the working directory to this output directory setwd output directory paste the directory name between the double quotation marks This cannot be executed until the completion of the run when the subdirectory tree is constructed by the batch file as described in R documentation for NPrun or ITrun and the run output is sorted to the appropriate folders This subdirectory tree will be contained in the working directory A folder with a date and time stamped name will be created with subdirectories for input output and other files Now the output of IT2B or NPAG needs to be loaded into R so the next command does this NPload dor ITload Details of these commands and what is loaded are described in the R documentation NPload or ITload and in the following section An integer can be included within the parentheses to be appended to the names of loaded R objects allowing for comparison between runs e g NP1oad 1 Finally at this point other Pmetrics commands can be added to the script to pro
51. uations enter the salt fraction of the drug usually 1 Salt fraction is the percentage of administered compound that contains active drug For example the mean salt fraction for theophylline is 0 85 This is not the same as bioavailability which is the fraction of drug absorbed after non parenteral administration e g oral compared to intravenous administration Enter the convergence criterion When the difference between log likelihoods of successive iterations is less than or equal to this criterion IT2B will converge and terminate The default is 0 001 which is the typical response Enter the maximum number of cycles This can be 1 to 41000 and IT2B will terminate at convergence or the number of cycles you specify here whichever comes first Early in model exploration values of 10 to 100 can be useful with larger values such as 1000 later in model development In order to facilitate model comparison however we recommend using the same cycle limit for all early models e g 100 rather than choosing 10 for one and 100 for another IT2B can pass parameter ranges to NPAG via a FROMxxxx usually FROM0001 file The default ranges to be used in NPAG are 5 times the final IT2B cycle standard deviation above and below the final cycle mean If your parameters are normally distributed 5 is a typical number For log normally distributed parameters 3 is a better choice You are now offered the opportunity to check all of your entries for correct
52. used to make the above objects Please use NPparse or ITparse in R for discussion ofthe data contained in these objects Since R is an object oriented language to access the observations in a PMop object for example use the following syntax opSpost1Sobs Note that you can place an integer within the parentheses of the loading functions e g NP1oad 1 which will suffix all the above objects with that integer e g op 1 final 1 NPdata 1 This allows several models to be loaded into R simultaneously each with a unique suffix and which can be compared with the PMcompare command see Model Diagnostics below Making New Pmetrics Objects Once you have loaded the raw NPdata or ITdata or processed op final cycle post data objects described above with NPload or ITload should you wish to remake the processed objects with parameters other than the defaults you can easily do so with the make family of commands For example the default for PMop objects is to use the prediction based on the median of the population or posterior distribution If you wish to use the mean of the distribution remake the PMop object using makeOP If you wish to see all the cycle information in a PMcycle object not omitting the first 10 of cycles by default remake it using makeCycle For all of the following commands the data input is either NPdata or ITdata with additional function arguments specific to each command Accessing the help
Download Pdf Manuals
Related Search