pact (version 0.5.0)

pact: Predictive Analysis of Clinical Trials

Description

The pact package implements a prediction-based approach to the analysis of data from randomized clinical trials (RCT). Based on clincial response and covariate data from a RCT comparing a new experimental treatment E versus a control C, the purpose behind the functions in pact is to develop and internally validate a model that can identify subjects likely to benefit from E rather than C. Currently, 'survival' and 'binary' response types are permitted.

Details

Package: pact Type: Package Version: 0.5.0 Date: 2016-04-14 Author: Dr. Jyothi Subramanian and Dr. Richard Simon Maintainer: Jyothi Subramanian <> License: GPL-3 pact.fit fits a predictive model to data from RCT. Currently, 'survival' and 'binary' response types are supported. Analysis of high dimensional covariate data is supported. If known and available, a limited number of prognostic covariates can also be specified and fixed to remain in the predictive model. An object of class 'pact' is returned. print, summary and predict methods are available for objects of class 'pact'. Additionally, the function pact.cv takes as an input the object returned by pact.fit and computes predictive scores for each subject through k-fold cross-validation. Evaluations of the cross-validated predictions are performed by the function eval.pact.cv. Finally, the function overall.analysis also takes an object of class 'pact' as input and computes some summary statistics for the comparison of treatments E and C.

Examples

### Survival response
set.seed(10)    
data(prostateCancer)     
Y <- prostateCancer[,3:4]
Xf <- prostateCancer[,7:8]  ## Prognostic covariates fixed to always be in the model
Xv <- prostateCancer[,c(5:6,9)]
Treatment <- as.factor(prostateCancer[,2])
p <- pact.fit(Y=Y,Xf=Xf,Xv=Xv,Treatment=Treatment,family="cox",varSelect="univar")
print(p)
overall.analysis(p)
cv <- pact.cv(p, nfold=5)
eval.pact.cv(cv, method="continuous", plot.score=TRUE, perm.test=FALSE, nperm=100)

### Binary response
set.seed(10)
data(EORTC10994)
Y <- as.factor(EORTC10994[,4])
## No prognostic covariates (Xf) specified
Xv <- EORTC10994[,c(2,5:7)]
Treatment <- as.factor(EORTC10994[,3])
p <- pact.fit(Y=Y,Xv=Xv,Treatment=Treatment,family="binomial",varSelect="none")
print(p)
overall.analysis(p)
cv <- pact.cv(p, nfold=5)
eval.pact.cv(cv, method="discrete", g=log(1), perm.test=FALSE, nperm=100)

### High dimensional data, survival response
## Not run: 
# set.seed(10)    
# data(GSE10846)     
# Y <- GSE10846[,1:2]
# Xv <- GSE10846[,-c(1:3)]
# Treatment <- as.factor(GSE10846[,3])
# p <- pact.fit(Y=Y,Xv=Xv,Treatment=Treatment,family="cox",varSelect="lasso",penalty.scaling=2)
# print(p)
# overall.analysis(p)
# cv <- pact.cv(p, nfold=5)
# eval.pact.cv(cv, method="continuous", plot.score=TRUE, perm.test=FALSE)
# ## End(Not run)