README.md
In eniclas/FAIRsimulator: Simulation Platform for the FAIR Trial

Background

This is a package with functions to implement the FAIR study simulator.

Status

The package is under development but it does run and offers a bit of flexibility. The documentation is a bit sparse right now. The best source of information is the vignette.

This is an overview of what the package is currently able to do:

Simulate studies with two or more age groups, e.g. 0-6, 6-12 an 12-18 months.
Subjects are reqruited into each of the age groups according to a specifiable recruitment rate.
The subjects are randomized to an intervention according to a set of probabilities.
The treatments are regarded as independent within age groups, which effectivey means that each treatment is age group specific.
HAZ "observations" are simulated from the FREM model (assuming India in the COHORTS study as the population) at age group specific time points.
Subjects may drop-out of the study according to a specifiable dropout rate.
Once the subjects reach the end of the age group age range, the accumulated HAZ observations are analysed using an lme (linear-mixed-effects) model, with birt weight, maternal age, maternal height, sex and sanitation facilities as covariates.
The estimated size of the treatment effects are used to update the randomization probabilities according to the posterior probability of each treatment being the best one.
The updated probabilities can be modified to meet specifiable constraints on minimum assigned probabilities, e.g. it can be specified that SoC (standard of care) should always associated with a probability of 0.25, regardless of its posterior probability of being the best.
After modifying the probabilities, the treatments are subject to a futility analysis, basically throwing away treatments that are not performing well enough according to some critera. The default criteria is to assign a 0 probability to any treatment that have an expected group size of less than or equal to 5 subjects. The remaining probabilities are adjusted accordingly.
Once subjects group out of their current age group, they are moved to the next higher age group and randomized to a treatment according to the updated randomization probabilities.
Various aspects of a simulated study can be visualised:
The number of subjects remaining in each age group over time.
The collected HAZ data over age.
The treatment effects over time.
The randomization probabilities over time.