Getting started with clinDR
In clinDR: Simulation and Analysis Tools for Clinical Dose Response Modeling

Introduction

This introductory document includes

Installation instructions highlighting the one-time use of compileStanModels()
A brief summary of the primary clinDR functions
Instructions to access the dose response meta-data used to support the Emax model and construct prior distributions for its parameters
A brief example

A detailed example using clinDR is in the vignette "Example of Bayesian Emax dose response modeling".

Installation

The clinDR package can be installed from the CRAN package library. It does not include any compiled code so there are seldom problems with installing it and checking the installation using library(clinDR) code.

The Bayesian computing in clinDR is performed using STAN thru the rstan R package, which must be installed before using most of the functions in clinDR. Most clinDR installation problems occur when rstan has not been successfully installed. To confirm that rstan is executing correctly, we recommend executing the simple example included in the documentation for the rstan function sampling.

Once rstan execution has been confirmed, one additional step is required before the Bayesian calculations in clinDR can be used. Before the first use of clinDR, the clinDR function compileStanModels() must be executed to compile the Emax model code. This preliminary step only needs to be executed one time. It typically requires 1-3 minutes to execute. It greatly accelerates the execution of the model fiting codes.

If a new version of rstan is installed, it is strongly recommended to re-execute the compileStanModels() command. If clinDR is re-installed or a new version is installed, the compileStanModels() function must be executed again before using clinDR.

Primary clinDR functions

The primary clinDR function is fitEmaxB, which fits the Bayesian Emax model. The prior distribution input to it is constructed by the function emaxPrior.control, and the MCMC specifications are constructed by mcmc.control. Many of the input settings have default values specified. There are numerous generic functions specialized to handle the fitEmaxB output such as print, plot, coef, predict, etc. The fitEmaxB function can be applied to continuous(normal) or binary data, and individual patient-level data or data aggregated to the dose group level. A corresponding fitEmax function computes maximum likelihood estimates with a step down sequence to the best fitting simpler dose response functions like linear, log-linear, and the exponential functions when the Emax model fits do not converge.

Simulation studies can be simulated without any programming using the corresponding emaxsimB and emaxsim functions. These functions also include many supporting functions to summarize their output. The documentation of these functions includes methods that allow specialized output to be created with minimal additional coding.

Dose response meta-data

Clinical dose response data from more than 200 compounds were evaluated using graphical methods and hierarchical modeling to support the use of the Emax function for dose response modeling. Combined with compound-specific information for a future dose response study, it is also used to create a moderately informative prior distribution for some of the model parameters. The data includes all available dose response studies displaying a non-null trend from one large sponsor since 1990 for both approved and subsequentially terminated compounds. The data includes publicly available data for all FDA approved compounds between 2009-2019. It also includes some older data from biological compounds. All of the data are aggregated to the dose level.

The data can be accessed from the clinDR package using data(metaData).
Details of the data format are available using the usual help functions. Some references for the different data sources are

Thomas, Sweeney, and Somayaji (2014)

Thomas and Roy (2016)

Wu, Banerjee, Jin, Menon, Martin, and Heatherington(2017)

Example

library(clinDR)
data("metaData")
exdat<-metaData[metaData$taid==1,]

prior<-emaxPrior.control(epmu=0,epsca=4,difTargetmu=0,difTargetsca=4,dTarget=20,
                p50=(2+5)/2,
                sigmalow=0.01,sigmaup=3)

mcmc<-mcmc.control(chains=3)

### estimate of within dose group SD
msSat<-sum((exdat$sampsize-1)*(exdat$sd)^2)/(sum(exdat$sampsize)-length(exdat$sampsize))

fitout<-fitEmaxB(exdat$rslt,exdat$dose,prior,modType=4,prot=exdat$protid,
                count=exdat$sampsize,msSat=msSat,mcmc=mcmc)
plot(fitout)