Introduction

This document describes the template files available in FACTS. The templates are described as a “class”, with different versions for different types of endpoint.

Phase 1 Dose Escalation Designs

non Model i3+3, mTPI,mTPI-2, BOIN designs

For all these designs the 6dose templates the the same and differ only in the design type used.

There are 6 doses, these are derived from the CRM 11 dose example. The 11 dose eample has doses at dose strengths of: 25, 36.5, 50, 75, 100, 150, 200, 300, 400, 600 and 800. The 6 doses for the 6 dose example are the first and last dose and ever other dose in-between: 25, 50, 100, 200, 400 and 800. This might be the more normal phase 1, dose doubling arrangement, the 11 dose used to illustrate the CRM’s greater flexibility and and ability to target “in-between” doses.

The non-Model methods don’t depend on the underlying dose strengths, they are simply a sequence of available doses, but the strengths were used to derive the dose-toxicity profiles simulated. See the CRM templates for details.

Profile	Dose 25	50	100	200	400	800
037.5 f	0.095	0.269	0.818	0.990	1.000	1.000
050 f	0.125	0.222	0.364	0.533	0.696	0.820
075 s	0.028	0.118	0.380	0.738	0.928	0.983
100 s	0.023	0.047	0.182	0.818	0.999	1.000
150 f	0.059	0.111	0.200	0.333	0.500	0.667
200 f	0.020	0.048	0.110	0.233	0.428	0.648
300 s	0.023	0.029	0.047	0.119	0.500	0.982
400 s	0.000	0.000	0.001	0.017	0.219	0.817
600 f	0.059	0.063	0.072	0.095	0.161	0.389
800 s	0.000	0.000	0.000	0.000	0.010	0.247

CRM designs

2D-CRM designs

Phase 2 and Phase 3 Designs

Single Arm trial

These are examples show how to set up a FACTS design to simulate a simple, fixed, single arm trial.

Two arm group sequential trial

These templates create a group sequential design, comparing a treatment against control. With two interims and alpha / beta spending boundaries to stop early for success or futility. The boundaries were derived outside of FACTS and are manually entered.

We simulate 5 scenarios: - Null, with no difference in response between the treatment arm and control, with the expected value of the nuisance parameter. - Alternate with the treatment difference the trial has been powered for, with the expected value of the nuisance parameter. - Four sensitivity analysis scenarios, where either the treatment effect is +/- 20% than hoped for, or the value of the nuisance parameter is +/- 10% the expected value.

The multiple endpoint template has two endpoints, the first is a continuous outcome and we use the same GS design for stopping criteria and fibnal decision. The second endpoint is a dichotomous safety endpoint where the rate on control is 0.1 and the non-inferiority delta is 0.1. Stopping bounds and final success criteria are in terms of the posterior probability that the difference between the safety rate on the treatment arm and on the control arm is < 0.1. The false positive rate if either efficacy is null, or the safety rate = control+0.1, is controlled at < 0.025. At the chosen overall sample size of 400, the power if the treatment is 1 point better than control and the safety rate the smae as contorl is between 0.85 and 0.9.

The corresponding fixed design can be simulated by simply unchecking the “enable adaptive features” checkbox on the Study > Study Info tab. Of course the fixed trial can use a slightly smaller maximum trial size.

Two arm Bayesian Goldilocks trial

These templates are identical to the tow arm group sequential trial templates, except that they use Bayesian predictive probabilities to determine stopping. The final test is a p-value test and so should be acceptable to regulators, but the decision to stop is decided on the predictive probability of success of the current trial. If the predictive probability of success if the trial were to fully enrol is sufficiently low then the trial is stopped for futility, if the predictive probability of success if the trial were stop enrolling and follow-up the enrolled but as yet incomplete subjects is sufficiently high then the trial stops enrolling in expectation of success after fully following up the current patients.

The thresholds used to decide to stop in this trial are: - to stop for futility if the predictive probability of success at the end of the trial is < 0.05. - to stop enroling in the anticipation of success if the predictive probability of success after following the currently enrolled population is > 0.95.

Two advantages of the Goldilocks design are: - The framing of the stopping rules is more intuitive - A more complex design can include a longitudinal model to impute the final outcomes of the partially complete patients, increasing the information available at the interim.

Four arm, arm dropping trial

These templates create a phase 2 like trial that tests 3 treatment arms against control, with 3 interims at which poorly performing arms can be dropped. Arms are dropped if the probability that they have the maximum response is < 0.1. This can be seen to have good properties, having greater power and smaller expected sample size compared to a fixed trial.

Four response profiles are included: - Lo Dose Only, the low dose has a treatment effect of - Linear - Med dose peak - Null, with no difference in treatment effect

Enrichment Designs

2 Groups

This template defines a design testing a treatment against a control with 2 subgroups - a Biomarker Positive subgroup, and Biomarker Negative subgroup, the assumption is that the treatment is more likely to work in the Positive subgroup.

The trial is a fixed design, there are no interims.

We simulate the Positive subgroup being slightly more prevalent (60:40), and simulate 3 repsonse profiles: - Null, no treatment difference from the control in either subgroup - Bio Positive successful, where the Biomarker Positive subgroup has the
-

FACTS Templates