RsNLME 2.0.1 - May, 2024

Issues Corrected

  • VPC replicate issue when reset is used resolved: When performing a VPC run with reset mapped, the replicate count would just cycle between 0 and 1. This issue has been resolved and the replicate value now counts up to nreplicates.

  • Random effect parameters now properly reinitialized after removal and being added back: When removing random effects from a model, then subsequently adding again, there were cases where previous random effect parameters were not reinitialized in the model properly. This issue has been resolved.

  • Previously set random effects values are now retained when new effects are added through randomEffect(): When removing random effects from a model, then subsequently adding again, there were cases when previous values would be reset to default instead of the originally specified value.

  • Models containing A1Strip now recognize the variable as a required column mapping.

  • Results for models with secondary parameters are now correctly supplied when Naive Pooled engine is used.

RsNLME 2.0.0 - November, 2023

What’s New

  • TDL5 performance enhancement: Model translation jobs involving less than 600 ODEs has dropped from 2.5 minutes to 10 seconds or less.

  • Certara.RsNLME and NLME-Engine is now supported on Ubuntu 22.04.

  • License information is now included in the model fit output: The license type (Academic or Commercial), the expiration date, current date and days until the license expires is now present in the output. If the license is no longer valid, a statement to this effect replaces the days until expiration data, and a link is present for contacting Certara Support.

  • MPICH implementation of the Message Passing Interface (MPI) standard on Windows has been replaced by Microsoft MPI.

  • NLME now utilizes a library of correctly rounded elementary functions in double precision. Switching to this platform-independent math library eliminates the observance of differences in results when executing on different platforms.

Certara.RsNLME v2.0.0

Issues Corrected


  • The peakreset statement now resets the peak variable to blank and works for both Cmin/Tmin and Cmax/Tmax. Previously, the peakreset statement only worked when used for Cmax/Tmax tracking, since it reset the peak variable to 0.

  • NLME now uses predefined sequence variables in structural parameters calculation. When some variables were initialized in the sequence block, they were initialized after defining dependent structural parameters, causing the initial compartment value to be an inappropriate value if closed form was used. This issue has been fixed.

  • NLME now properly reads data column definition files. A case was reported where NLME did not read the column definitions properly, causing the observation definition to be doubled. This problem has been fixed.

  • Setting the initial LL too high no longer causes optimization by FOCE algorithms to exit prematurely. For some models, parameters were not fully optimized by FOCE algorithms when the initial LL was too high. This resulted in exiting the optimization prematurely with exit code 1. This issue has been fixed

  • QRPEM now correctly shows zero etas for the subjects with zero observations. Previously, all etas were updated irrespective of the number of observations.


  • Population models with zeroed etas can now be created. It was reported that a simulation for a population model could not be run while setting all eta and sigma terms to 0. Setting all error terms to 0 caused the model to default to individual and, as a result, id was being removed from the mappings. This issue has been resolved.

RsNLME 1.2.0 - January, 2023

What’s New

Certara.RsNLME v1.2.0

  • PRED is now output as a separate column in the file output of vpcmodel(). Previously, PRED was output in the DV column with REPLICATE =-1.

  • LLOQ value is now outputted to predcheck0 (Observations data frame) for vpcmodel() function.

  • For each observed variable per each replicate, predout (PredCheckAll worksheet) now lists simulated values, population predictions if either OutputPRED is requested or prediction correction is enabled, correction factor and corrected simulated value if predication correction is enabled.

  • RsNLME print() model generic displays the values of modelName and workingDir slots.

  • A new wrapper function called hostParams() allows setup of host objects without using the NlmeParallelHost class initializer.

  • A new wrapper function called tableParams() allows defining table output instead of using NlmeTableDef-class/NlmeSimTableDef-class initializers.

  • The syntax for the colMapping() function now allows for unquoted key-value pairs that may be specified via the ellipses argument.

  • A new filesToReturn argument is available for the fitmodel() and sortfit() functions, which allows specification of data tables to return.

  • In estimatesUI(), when overlay=TRUE, the number of lines in the resulting plot now reflects the number of unique dosing values in data.

  • UNC paths are now supported, allowing RsNLME to be utilized with a shared directory inside a network drive.

Certara.Xpose.NLME v1.2.0

  • New xposeNlmeModel() function in Certara.Xpose.NLME creates xpdb with the model object. This function has two arguments: model (object of class NlmePmlModel) and fitmodelOutput (returned value of fitmodel execution).

Certara.RsNLME.ModelBuilder v1.2.0

  • The model working directory now defaults to the current working directory in an RStudio Session, instead of a temp directory, and can be overridden with a user-defined directory.

Certara.RsNLME.ModelExecutor v1.2.0

  • R code generation of table and engine parameters for workflows performed in Model Executor is now supported.
  • Real time convergence plots for simple run mode are now available. Previously, convergence plots of –2LL and model parameters were only available after execution was finished.
  • Stratification by categorical covariates is now supported for Bootstrap.
  • A single host can now be specified without wrapping in c().
  • During model execution, the same output that is returned in the R console is now reported in the Progress Text window in Model Executor.

Certara.ModelResults v1.2.0

  • Font customization options are now formal font names. Previously, the choices were typefaces (i.e., serif, sans, mono). Now the choices are font names such as Times New Roman, Arial, Courier New, etc.

Certara.VPCResults v1.2.0

  • The binning methods headtails, maximum, and box are now available.
  • Font customization options are now formal font names. Previously, the choices were typefaces (i.e., serif, sans, mono). Now the choices are font names such as Times New Roman, Arial, Courier New, etc. Customization options for VPC plot legend are now available (DRWN-186)

Issues Corrected

Certara.RsNLME and Certara.NLME8

  • An issue where stepwise covariate searches with criterion set to AIC could give incorrect results during backward elimination has been resolved (PHX-8226). Previously, for a stepwise covariate search run involving a multicore host and AIC, the page would use “+” instead of “-” for certain scenarios during the backward elimination.

  • For shotgun covariate searches, the total number of jobs/scenarios executed is set to 2^n, where n denotes the number of enabled covariate effects (DRWN-447).


  • An issue where using special words like ‘dose’, ‘obs’, ‘covr’ , and ‘mode’ could cause NLME to fail has been resolved (PHX-8024): In the previous version, using such special words as variables in the simulation table or as stratification variables would occasionally cause NLME to fail.

  • The issue of fraction excreted theta, used in urinecpt, being fixed irrespective of other statements has been resolved (PHX-8015).

  • An issue where having an overall number of fixed effects (including frozen and not enabled for the current scenario run) exceed the number of thetas + 100 could cause optimization to stop has been resolved (PHX-8020).

  • For non-time-based models with covariate effects, if the When covr set field contains a covariate that is not the last one defined in the Column Definition Text area, NLME no longer generates an incorrect execution error message that the covariate is not set (PHX-8199).

  • Using a sscol() statement in column definitions no longer creates additional rows in the doses.csv file (PHX-8214).
    The QRPEM engine now supports models in which the same thetas are used in different stparm statements (PHX-8168).

  • In the calculation of steady state, when the DelayInfCpt statement is used and reset statements are present, the reset statements are now handled correctly (PHX-8013).

Files Changed

  • R Package: Certara.RsNLME_1.2.0.tar.gz
  • R Package: Certara.NLME8_1.2.2.tar.gz
  • R Package: Certara.Xpose.NLME_1.2.0.tar.gz
  • R Package: Certara.RsNLME.ModelBuilder_1.2.0.tar.gz
  • R Package: Certara.RsNLME.ModelExecutor_1.2.0.tar.gz
  • R Package: Certara.ModelResults_1.2.0.tar.gz
  • R Package: Certara.VPCResults_1.2.0.tar.gz
  • NLME-Engine: NLME-Engine-23.1.1.exe (Windows)
  • NLME-Engine: (Linux)

RsNLME 1.1.1 - June, 2022

RsNLME 1.1.1 release corrects the issue where the posthoc table was missing if numIterations = 0 for model execution (RSNLME-2). The following files have been updated:

  • R Package: Certara.NLME8_1.2.1.tar.gz
  • NLME-Engine: NLME-Engine-21.11.2.exe (Windows)
  • NLME-Engine: (Linux)

The updated version of Certara.NLME8 package v1.2.1 will be automatically installed when performing the installation for R packages. However, existing users should request the updated NLME-Engine-21.11.2.

Request NLME-Engine-21.11.2

For Windows, after running NLME-Engine-21.11.2.exe you will first be prompted to remove the previous NLME-Engine installation. After removing NLME-Engine-21.11.1 via the installer, running NLME-Engine-21.11.2.exe once again will proceed with installing the updated version.

Linux users may simply extract contents of and replace their existing NLME_Engine folder with the updated version.

Note: While the version of the RsNLME software suite (R packages and NLME-Engine) has been incremented to v1.1.1 for the above hotfix, no changes have been made to the Certara.RSNLME package and the version remains v1.1.0. Only the Certara.NLME8 R package and NLME-Engine versions have been incremented.

RsNLME 1.1.0 - November, 2021

What’s New

NLME-Engine v21.11.1

  • New “type” argument for LL statements enables generation of appropriate summary statistics worksheets for VPC mode (PRN-652): In PML code, the type argument is now available for specifying an observation variable’s type. This argument can be set to cont, cat, count, or event and ensures the proper output worksheets will be generated for VPC mode.

  • Simple and Simulation tables include all five ID columns (PRN-645): NLME supports up to 5 ID columns for the input data. To take this into account, five ID columns are added for all simple and simulation tables.

  • New column ‘LLOQ’ is added to the output of predcheck output (PRN-726): To facilitate the computations of VPC data in R packages, the LLOQ column is added to the predcheck0.csv output file if the model has BQL observation data specified.

Certara.RsNLME v1.1.0

  • Additional steady state dosing can now be specified using the optional “SSOffset” argument of “addSteady” (PRN-317): This argument produces the “ssoffcol” statement in cols1.txt.

  • Columns mapped as covariates in the model object are now checked for validity (PRN-629): If data is loaded to a model object and covariate model terms are mapped to data columns, an automatic check is performed regarding data column reliability. A Warning/Error is given in cases where there is inappropriate data.

  • Initial Estimates Shiny application now uses ggplot2 (PRN-537): The estimatesUI initial estimates Shiny application in the Certara.RsNLME package has been updated to use the ggplot2 plotting library and to support plot faceting.

Command line support for the following has been added to the RsNLME package:

  • Distributed delay (PRN-618)

  • Supply class initializer arguments for execution params directly inside execution functions using ellipses (PRN-709)

  • Data mapping can now be performed without column mapping inside built-in model functions by supplying data argument and setting columnMap = FALSE (PRN-717).

Support for new syntax in mmdl:

  • Dosing cycles (PRN-633)

  • Covariate levels and labels (PRN-805)

Certara.Xpose.NLME v1.1.0

  • New covariate model diagnostic functions have been added to the ‘Certara.Xpose.NLME’ package (PRN-541): New covariate plot functions include support for ETAs, parameters, and residuals plotted against a continuous or categorical covariate.

RsNLME.ModelBuilder v1.1.0

  • Additional functionality has been added to the RsNLME.ModelBuilder Shiny application:

  • Administration-absorption parameters for distributed delay (PRN-773)

  • Dosing cycle specification (PRN-785)

  • Covariate labels (PRN-805)

Certara.RsNLME.ModelExecutor v1.1.0

  • Additional functionality has been added to the RsNLME.ModelExecutor Shiny application:

  • Plot convergence data after model execution (PRN-580)

  • Return estimation arguments and table statements to mmdl in Pirana (PRN-704)

  • Add sort column selections for model execution (PRN-423)

Certara.ModelResults v1.1.0

  • Addition of new Shiny application, Model Results (PRN-620): The ‘Certara.ModelResults’ Shiny application in installed as an R package and used to generate and report model diagnostics plots and tables. The application includes functionality to generate corresponding R code from the xpose, ggplot2, and flextable libraries, given operations performed in the GUI.

Certara.VPCResults v1.1.0

  • Addition of new Shiny application, VPC Results (PRN-691): The ‘Certara.VPCResults’ Shiny application is installed as an R package and used to parameterize, plot, and report Visual Predictive Checks (VPC). The application includes functionality to generate corresponding R code from the tidyvpc and ggplot2 libraries, given operations performed in the GUI.

Misc Documentation Updates

Additional information for Certara R packages that are accessed through Pirana is now available as pkgdown websites (PRN-838):

Issues Corrected

NLME Engine

  • All interoccasion omegas have standard errors in the output (PRN-623): Previously, the output only contained the standard errors (SEs) for omegas related to the first occasion. Now, SEs are included for all interoccasion omegas.

  • An issue that prevented using more than five observation variables has been resolved (PRN-545): Previously, it was impossible to run a model with more than five observation variables using the QRPEM engine. This issue is fixed in RsNLME 1.1.

  • An issue involving the failure of parsing when more than one stratification variable is specified has been resolved (PRN-614).


  • Non-time-based models now work correctly (PRN-300): Previously, in a non-time-based model, if the name of a covariate in the model was the same as the corresponding one in the input dataset, the covariate would not be listed as a covariate in the database.

  • For models involving reset, the original dataset appended in the generated data is now correct (PRN-303): Previously, the generated data showed data for all occasions as being exactly the same as the ones corresponding to the first occasion.

  • Individual plot legends now properly display the line type and color (PRN-307).


  • Issue where column named “X_ID” in the data used as subject ID was incorrectly used for subjects accounting has been resolved (PRN-507).

  • A reported issue where executing a large NLME job on a grid led to the master node for the grid to run out of memory has been resolved (PRN-538).

  • NLME now accepts negative estimates as initial estimates for the model executed (PRN-543/CS00211446).


  • The Sort Columns option now works as expected and is now available in the Individual modeling mode (PRN-542).

Known Issues

  • The NLME Engine installer must be run as an administrator or the installation will fail (RSNLME-346): If an error is encountered during installation, check that the person installing has administrator privileges.

  • NLME engine returns a negative value for the standard deviation of residual error variable (PRN-540): If such a situation occurs, then the absolute value of the standard deviation should be considered as the final estimate.