Movatterモバイル変換

NCC:Simulation and analysis of platform trials with non-concurrent controls

NCC package allows users to simulate platform trials andto compare arms using non-concurrent control data.

Design overview

We consider a platform trial evaluating the efficacy of $K$ treatment arms compared to a shared control. Weassume that treatment arms enter the platform trial sequentially. Inparticular, we consider a trial starting with at least one initialtreatment arm, where a new arm is added after every $d=(d_1,…,d_K)$ patients have beenrecruited to the trial (with $d_1=0$ ).

We divide the duration of the trial into $S$ periods, where the periods are the time intervalsbounded by times at which a treatment arm either enters or leaves theplatform.

The below figure illustrates the considered trial design.

Functions

This package contains the following functions:

Data generation

Main functions for datageneration

datasim_bin() simulates data with binary outcomes
datasim_cont() simulates data with continuousoutcomes

Auxiliary functions fordata generation

get_ss_matrix() computes sample sizes per arm andperiod
linear_trend() is the linear time trend function, usedto generate the trend for each patient
sw_trend() is the step-wise time trend function, usedgenerate the trend for each patient
inv_u_trend() is the inverted-u time trend function,used generate the trend for each patient
seasonal_trend() is the seasonal time trend function,used generate the trend for each patient

Data analysis

Treatment-controlcomparisons for binary endpoints

Frequentist approaches

fixmodel_bin() performs analysis using a regressionmodel adjusting for periods
fixmodel_cal_bin() performs analysis using a regressionmodel adjusting for calendar time
poolmodel_bin() performs pooled analysis
sepmodel_bin() performs separate analysis
sepmodel_adj_bin() performs separate analysis adjustingfor periods

Bayesian approaches

MAPprior_bin() performs analysis using the MAP priorapproach
timemachine_bin() performs analysis using the TimeMachine approach

Treatment-controlcomparisons for continuous endpoints

Frequentist approaches

fixmodel_cont() performs analysis using a regressionmodel adjusting for periods
fixmodel_cal_cont() performs analysis using aregression model adjusting for calendar time
gam_cont() performs analysis using generalized additivemodel
mixmodel_cont() performs analysis using a mixed modeladjusting for periods as a random factor
mixmodel_cal_cont() performs analysis using a mixedmodel adjusting for calendar time as a random factor
mixmodel_AR1_cont() performs analysis using a mixedmodel adjusting for periods as a random factor with AR1 correlationstructure
mixmodel_AR1_cal_cont() performs analysis using a mixedmodel adjusting for calendar time as a random factor with AR1correlation structure
piecewise_cont() performs analysis using discontinuouspiecewise polynomials per period
piecewise_cal_cont() performs analysis usingdiscontinuous piecewise polynomials per calendar time
poolmodel_cont() performs pooled analysis
sepmodel_cont() performs separate analysis
sepmodel_adj_cont() performs separate analysisadjusting for periods
splines_cont() performs analysis using regressionsplines with knots placed according to periods
splines_cal_cont() performs analysis using regressionsplines with knots placed according to calendar times

Bayesian approaches

MAPprior_cont() performs analysis using the MAP priorapproach
timemachine_cont() performs analysis using the TimeMachine approach

Running simulations

all_models() is an auxiliary wrapper function toanalyze given dataset (treatment-control comparisons) with multiplemodels
sim_study() is a wrapper function to run a simulationstudy (treatment-control comparisons) for desired scenarios
sim_study_par() is a wrapper function to run asimulation study (treatment-control comparisons) for desired scenariosin parallel

Visualization

plot_trial() visualizes the progress of a simulatedtrial

For a more detailed description of the functions, see the vignettesin the R-package website (https://pavlakrotka.github.io/NCC/).

Scheme of the packagestructure

The below figure illustrates theNCC package functionsby functionality.

Installation

To install the latest version of theNCC package fromGithub, please run the following code:

# install.packages("devtools")devtools::install_github("pavlakrotka/NCC",build_vignettes =TRUE)

Documentation

Documentation of all functions as well as vignettes with furtherdescription and examples can be found at the package website:https://pavlakrotka.github.io/NCC/

References

[1] Bofill Roig, M., Krotka, P., et al. “Onmodel-based time trend adjustments in platform trials withnon-concurrent controls.” BMC medical research methodology 22.1(2022): 1-16.

[2] Lee, K. M., and Wason, J.“Includingnon-concurrent control patients in the analysis of platform trials: isit worth it?.” BMC medical research methodology 20.1 (2020):1-12.

[3] Saville, B. R., Berry, D. A., et al. “TheBayesian Time Machine: Accounting for Temporal Drift in Multi-armPlatform Trials.” Clinical Trials 19.5 (2022): 490-501

Funding

EU-PEARL (EU Patient-cEntricclinicAl tRial pLatforms) project has received funding from theInnovative Medicines Initiative (IMI) 2 Joint Undertaking (JU) undergrant agreement No 853966. This Joint Undertaking receives support fromthe European Union’s Horizon 2020 research and innovation programme andEFPIA and Children’s Tumor Foundation, Global Alliance for TB DrugDevelopment non-profit organisation, Spring works Therapeutics Inc. Thispublication reflects the authors’ views. Neither IMI nor the EuropeanUnion, EFPIA, or any Associated Partners are responsible for any usethat may be made of the information contained herein.

[8]ページ先頭