FACTS 6.4.0 Release Notes

April 23, 2021

FACTS 6.4.0 is now available for official release. Please contact us regarding any questions.

The key features of this release are:

In detail the new features in FACTS 6.4.0 are:

  1. Three new dose response models have been added across all FACTS (Staged) Core designs. These new options will appear in the model selection dropdown on the Dose Response tab. The new models are as follows:

    1. The Simple Hierarchical model – a model in which the mean responses for each of the arms in the design are drawn from a normal distribution, whose mean and variance are estimated by FACTS. The control arm can be included in the hierarchical model, or modeled separately, in which case it has its own prior mean and variance. The control arm cannot be included in this model for Time-to-Event designs.

    2. The Simple Linear model – a linear model which assumes that the mean responses for each of the arms in the design are linear functions of the associated arm strength. In particular, the arm with the largest mean response is guaranteed to be either the largest dose or the smallest arm in this model. Note that the “2-Parameter Logistic” model in FACTS (Staged) Core Dichotomous designs has been replaced by the “Simple Linear model”. FACTS (Staged) Core Dichotomous designs making use of the 2-Parameter Logistic model will be automatically migrated to the Simple Linear model.

    3. The Simple Hierarchical Linear model – a model which uses a linear model as a base dose-response structure but allows deviations from linearity in a manner similar to the Hierarchical Logistic dose response model. Given appropriate priors, if the data and prior distributions are consistent with linearity, the hierarchical variance parameter will be estimated to be small and the model fit will be essentially linear, but if the data is non-linear the variance parameter will be large allowing a significantly non-linear model fit.

  2. In FACTS (Staged) Core Dichotomous and TTE designs, Posterior Probability QOIs with alternative parametrizations can be set when creating a new QOI. This can be achieved by selecting the appropriate option in the “Compare” dropdown of the QOI dialog. The options are as follows:

    1. For FACTS (Staged) Core Dichotomous designs, Posterior Probability QOIs comparing the log-odds ratio of the response rate for each arm against that of a given arm can now be created. Previously, only the response rates could be compared.
    2. For FACTS (Staged) Core TTE designs, Posterior Probability QOIs comparing the hazard rates of the response for each arm against that of a given arm can now be created. Previously, only the hazard ratios (HR) could be compared.

  3. Enterprise FACTS licensees will now be able to access and run FACTS Core and Enrichment Design (ED) analysis models from R via an R wrapper, the output of which is an MCMC file pertaining to the model. This can be used to simulate trials that require posterior quantities that FACTS does not include (e.g., probability that a dose has a treatment effect in a certain range) or simulate trials that make decisions that FACTS does not include (e.g., sample size re-assessment).

  4. Enterprise FACTS licensees will also now be able to run FACTS in command-line mode on Linux via a separate executable: FACTS Linux File Loader Lite (FLFLL). Mono 6.8.0+ is a pre-requisite for running FLFLL. Executing a valid FACTS design with FLFLL will generate the same results output as its Windows GUI counterpart; in particular, it will generate the simulations, summary, weeks and patients files. FLFLL can be used to automate the simulation of multiple (potentially related) FACTS designs and, more generally, can be used as a key component of a more complex trial design simulation pipeline.

The following features were also implemented in FACTS 6.4.0: