README.md

In Arhit-Chakrabarti/logisticLAGO: Learn-As-You-Go study with a single constraint and binary response

logisticLAGO Package

Arhit Chakrabarti

• Description
• Installation
• Usage
  • In an ongoing trial
  • Before starting a trial
    • Single center LAGO design
    • Multi center LAGO design
      • Equal number of centers per stage and equal sample size in each center
      • Unequal number of centers per stage and equal sample size in each center
      • Unequal number of centers per stage and unequal sample size in each center
• Details

Description

The “Learn-As-You-Go” (LAGO) design is motivated by large scale public health intervention trials, where an intervention package is prospectively modified over each stage such that an “optimal” intervention package is rolled out to the participants in the next stage. This R package implements the optimization for the LAGO study with a single constraint, namely the outcome goal constraint and binary response.

Installation

You can install logisticLAGO R package from GitHub with:

# install.packages("devtools")
# install the package without building vignettes
devtools::install_github("Arhit-Chakrabarti/logisticLAGO")

Note: The library devtools needs to be installed before installing the R package from GitHub. Please note that this package includes vignettes which provides further guidelines on how to use the functions in this R package. To install the logisticLAGO package with the vignettes, use the code

# install.packages("devtools")
# install the package building vignettes
devtools::install_github("Arhit-Chakrabarti/logisticLAGO", build_vignettes = TRUE, force = TRUE)

Usage

Once the logisticLAGO library is installed load the library in the R workspace.

library("logisticLAGO")

The basic setup of the LAGO design requires a multi-component, continuous intervention package and a binary response variable which may be for example, how well the participants perform on a test following administration of the intervention package. The intervention package is prospectively changed at every stage based on the cumulative data collected over the stages and an optimal package is rolled out to the participants in the next stage such that the probability of success for the binary response is above a desired threshold, while simultaneously minimizing the implementation costs.

In an ongoing trial

In the case when a trial has already been designed and data from the trial has been collected, the next step of the study is to estimate the optimum intervention package to be rolled out to the participants in the next stage such that the outcome goal of the study is met and costs minimized. To estimate the optimum intervention package, the vector of per unit costs for the components of the intervention package (cost), the vector of minimum (lower) and maximum (upper) values of the components of the intervention package, desired outcome goal (pstar), the estimated (beta) from fitting a logistic regression model to the observed response, which gives the estimated effect of the corresponding intervention package and the intervention package rolled out at the current stage (starting.value) are required. The estimated optimal intervention is given by the function opt_int

opt_int(cost = cost, beta = beta, lower = lower, upper = upper, pstar = pstar, starting.value = starting.value)

Before starting a trial

Single center LAGO design

This package may also be used before starting a trial to get an idea about the optimal intervention package based on an initial intervention package and best guesses about the effects of the components on the response. Initial package and idea about the effects of the components of the intervention package may be obtained from an investigator or from knowledge of prior or concurrent intervention trials. The simplest case is when the trial is designed to be conducted in a single center or location. The number of stages (nstages) in the LAGO design, sample size per stage (sample.size), the unit costs for the intervention package components (cost.vec), the vector of minimum (lower) and maximum (upper) values of the components of the intervention package, desired outcome goal (prob), the best guess intervention effect (beta.true) and the initial value of intervention package (x0) along with the expected variation in rolling out the intervention package (icc) are used to simulate the estimated intervention package over the different stages, the estimated outcome goal and the estimated power of the test of “no-intervention” effect at the end of the study. This is done using the following function:

sc_lago(x0 = x0, lower = lower, upper = upper, nstages = nstages, beta.true = beta.true, sample.size = sample.size, icc = icc, cost.vec = cost.vec, prob = prob)

Multi center LAGO design

Equal number of centers per stage and equal sample size in each center

Another common study design is when the trial is planned to be conducted in a multiple centers or location. Apart from the information as required by the single center design, the number of centers per stage (centers) and sample size per center per stage (sample.size) is used to simulate the optimal intervention package to be rolled out in each of the centers in the next stage such that the goals of the study are met. As before, estimated power for the test of “no-intervention” effect at the end of the study is also provided through simulations. The simplest design in this case, assumes an equal number of centers per stage and equal number of samples per center in each stage. The corresponding function is:

mc_lago(x0 = x0, lower = lower, upper = upper, beta.true = beta.true, nstages = nstages, centers = centers, sample.size = sample.size, cost.vec = cost.vec, prob = prob, icc = icc)

Unequal number of centers per stage and equal sample size in each center

Another common design under multi-center LAGO study is considering different number of centers in each stage. However every center at any stage have equal samples. The corresponding function is:

mc_lago_uc(x0 = x0, lower = lower, upper = upper, beta.true = beta.true, nstages = nstages, centers = centers, sample.size = sample.size, cost.vec = cost.vec, prob = prob, icc = icc, bcc = bcc)

Here the function argument icc denotes the expected variation among subjects within each center at any stage and the argument bcc denotes the expected variation between the centers at any stage while implementing the intervention package. Thus, this function separates the between center and within center variability, while generating the data for simulations.

Unequal number of centers per stage and unequal sample size in each center

The traditional design under multi-center LAGO study is considering different number of centers in each stage as well as different sample sizes in each center of any given stage. This design arises quite naturally, as it is intuitive to have both small number of centers and small sample size at each center at the beginning of the study and progressively increase the number of centers and sample size as the study progresses. The corresponding function is:

mc_lago_uc.us(x0 = x0, lower = lower, upper = upper, beta.true = beta.true, nstages = nstages, centers = centers, sample.size = sample.size, cost.vec = cost.vec, prob = prob, icc = icc, bcc = bcc)

Here, the argument sample.size is a list denoting the sample size in each center of different stages.

Details

For more information on logisticLAGO Package, please access the package documentations or vignettes. Please feel free to contact the author.

Arhit-Chakrabarti/logisticLAGO documentation built on Dec. 17, 2021, 9:43 a.m.