Multi-Arm Multi-Stage (MAMS)


Statistical software

Some existing statistical software resources to support the implementation of MAMS (and platform trials) have been reviewed and functionalities summarised 1.  Here, we summarise some of the statistical software resources to support the design, monitoring, or analysis of MAMS trials.


1. Meyer et al. Systematic review of available software for multi-arm multi-stage and platform clinical trial design. Trials. 2021;22(1):183. 

Open-access software

MAMS (R package)
Supports implementation of MAMS design based on group-sequential methods for normally distributed, binary, ordinal, and time-to-event primary outcomes (see details). It accommodates:
  • a number of statistical stopping boundaries; 
  • treatment selection rules (e.g., selecting all promising or only best performing);
  • recalculation of stopping boundaries for dealing with unplanned changes (e.g., mistimed interim analysis) or after sample size re-estimation;
  • simulation of operating characteristics. 
rpact (R package)
Incorporates the design and analysis of MAMS trials with binary, continuous, and time-to-event outcomes based on combination test and conditional error functions (see details in this book especially chapter on “Applications and Case studies”). Decision rules/boundaries can automatically be recalculated during trial monitoring. Statistical inference is based on stagewise ordering approach to obtain adjusted treatment effect estimates, confidence intervals and p-values. Comprehensively covers step-by-step practical examples in the design and analysis of MAMS trials and other adaptive trials.

asd (R package)
Supports the simulation of a MAMS design that addresses phase 2 and 3 objectives simultaneously in a single trial (seamless 2/3 design) accommodating: 
  • treatment selection to be based on an adaptation outcome that is different from the primary outcome used in phase 3;
  • a number of treatment selection rules. 
See underlying methods for details.

multiarm (R package)
Supports the design of drop-the-loser type of MAMS designs with continuous and binary outcomes.

HECT (Shiny app)
Supports the simulation of operating characteristics of MAMS trials with continuous or binary outcomes using Bayesian methods and offers options for early stopping for efficacy or futility as well as response-adaptive randomisation (RAR).

MASSR (R package)
Performs sample size re-estimation based on the conditional power approach for two-stage flexible MAMS designs with normal and binary outcomes. Allows three flexible MAMS designs: flexible group sequential design, adaptive design based on inverse normal combination function and Fisher's combination function. See underlying methods for details.

BayesMAMS (R package)
Calculates Bayesian sample sizes for multi-arm trials with several study treatments that are compared to a common control at multiple stages or single stage. See underlying methods for details.

BAR (Shiny app)
Supports the implementation of multi-arm Bayesian adaptive randomisation allowing for early stopping for futility or efficacy using posterior probability with or without a control (see BAR).

ComPAS (Shiny app)
Supports the simulation of a Bayesian adaptive design that offer opportunities for the selection of drug combinations by allowing dropping of futile one, graduating effective ones, or adding new combinations during the trial. See ComPAS and published underlying statistical methods.

Closed source software

nstage (Stata program)
Supports the design of MAMS trials with time-to-event outcomes and allows for the adaptation outcome to differ from the primary outcome. Program has been updated to accommodate a number of binding and non-binding stopping boundaries.  See underlying methods for details.

nstagebin, nstagebinopt (Stata program)
Supports the design of MAMS designs with binary outcomes analysed using differences in proportions  and allows for the adaptation outcome to differ from the primary outcome. See underlying methods for details. 

DESMA:  des_mams, sim_mams, des_dtl, and sim_dtl (Stata program and subroutines)
Offers the same functionalities as the MAMS R package described above.

A Fixed and adaptive clinical trial simulator that supports the design, comparison, and simulation of both fixed and adaptive trial designs including MAMS trials using Bayesian methods. It also accommodates additional adaptations such as response-adaptive randomisation (RAR)

Commercial software

East: Multiarm (module)
Supports the design and simulation of MAMS designs using frequentist methods as well as statistical inference based on a stagewise ordering approach to obtain bias-adjusted treatment estimates, confidence intervals, and p-values.  See some underlying methods for details. 
East Alloy
The module implements simulations for a multi-arm two-stage design for normal outcomes using Bayesian methods. Early stopping or dropping of ineffective treatments is based on posterior probabilities. 

ADDPLAN: MC (module)
See brochure for details on functionalities.

Tutorial resources