BM_R: Bayesian efficacy monitoring for the difference of two...

In IDDI-BE/PhII_Bayes: Bayesian tools for PhII clinical trial design

Description

Comparative Bayesian Efficacy Monitoring Via Predictive Probability (BEMPR) or Bayesian Efficacy Monitoring Via Posterior Probability (BEMPO)
The frequentist properties can be investigated for combinations of
1) Different sample sizes 2) Different true response rates # BEMPO interim futility stopping rule is P(Delta<=Delta_fut)>P_fut BEMPO interim efficacy stopping rule is P(Delta>Delta_eff)>=P_eff BEMPO final efficacy stopping rule is P(Delta>Delta_fin)>=P_fin

BEMPR interim futility stopping rule is PP<P_fut BEMPR interim efficacy stopping rule is PP>=P_eff BEMPR final efficacy stopping rule is P(Delta>Delta_fin)>=P_fin

Usage

BM_R(
  N,
  ar,
  p_exp,
  p_ctrl,
  design_fut,
  design_eff,
  interim,
  Delta_fut = NULL,
  P_fut,
  Delta_eff = NULL,
  P_eff,
  Delta_fin,
  P_fin,
  beta_par_exp,
  beta_par_ctrl,
  nsim,
  distrisize = 10^3,
  PP_nsim = 10^3
)

Arguments

N	sample size to investigate operating characteristics: can be a scalar or a vector
ar	allocation ratio experimental:control arm
p_exp	response rate experimental arm to investigate operating characteristics: can be a scalar or a vector
p_ctrl	response rate control arm to investigate operating characteristics: can be a scalar or a vector
design_fut	choice between "futPR" (predictive probability) and "futPO" (posterior probability)
design_eff	choice between "effPR" (predictive probability) and "effPO" (posterior probability)
interim	Information fraction at interim, e.g. c(0.5,0.75)
Delta_fut	This may be a scalar or a vector. Must only be specified for BEMPO
P_fut	This may be a scalar or a vector
Delta_eff	This may be a scalar or a vector. Must only be specified for BEMPO
P_eff	This may be a scalar or a vector
Delta_fin	This must be a scalar
P_fin	This must be a scalar
beta_par_exp	two shape parameters c(alpha,beta) for prior beta distribution experimental arm
beta_par_ctrl	two shape parameters c(alpha,beta) for prior beta distribution control arm
nsim	number of simulations
distrisize	Size of sampled distributions (the larger, the better)
PP_nsim	Number of simulation to sample from predictive distribution

Value

a list of 3 data.frames: first with design parameters + operating characteristics ($param_simul) ,
one with futility decision rules ($Fut_rules), and one with efficacy decision rules ($Eff_rules)

• param_simul: dataframe with all input parameters + operating characteristics

• param_simul$scenario: combination of vector N and p

• param_simul$N: sample size for which frequentist properties simulated

• param_simul$p: binomial parameter (true proportion) for which frequentist properties simulated

• param_simul$param: string with all input parameters

• param_simul$N_interim: number of interim analyses (final analysis not included)

• param_simul$interim: string with number of patients per interim analysis

• param_simul$pow: proportion of simulations where H0 was rejected

• param_simul$eff_fin: proportion of simulations where H0 was rejected only at the final analysis

• param_simul$eff_stop: proportion of simulations where H0 was rejected at an interim analysis

• param_simul$fut: proportion of simulations where H0 was not rejected

• param_simul$eff_fin: proportion of simulations where decision of not rejecting H0 was at the final analysis

• param_simul$fut_stop: proportion of simulations where trial stopped for futility at an interim analysis

• param_simul$N_avg: average number of patients for simulated datasets

• param_simul$int_nrx: number of patients at x'th interim analysis

• param_simul$effx: proportion of simulations where trial stopped for efficacy at xth interim analysis

• param_simul$futx: proportion of simulations where trial stopped for futility at xth interim analysis

• param_simul$decision rules: for efficacy: if >=x: stop for efficacy; for futility: if <=x: stop for futility

• Fut_rules: dataframe with futility rules for each interim

• Fut_rules$n: number of patients at interim

• Fut_rules$x: number of successes at interim (Futility if observed successes <=x)

• Fut_rules$P_Bayes/Fut_rules$PP: actual posterior probability or PP at cutoff

• Eff_rules: dataframe with efficacy rules for each interim

• Eff_rules$n: number of patients at interim

• Eff_rules$x: number of successes at interim (Efficacy if observed successes >=x)

• Eff_rules$P_Bayes/Fut_rules$PP: actual posterior probability or PP at cutoff

References

Lee JJ, Liu DD.A predictive probability design for phase II cancer clinical trialsClinical Trials 2008; 5: 93–106

Examples

pow<-BM_R(N=c(100),ar=1,p_exp=c(0.475),p_ctrl=c(0.35),interim=c(0.3,0.5,0.75),Delta_fut=NULL,
P_fut=c(0.25,0.5,0.5),Delta_eff=c(0,0,0),P_eff=c(1,0.75,0.75),Delta_fin=0,P_fin=0.75, 
distrisize=10^3, PP_nsim=10^3, design_fut="futPR", design_eff="effPO",nsim=1000)
typeI<-BM_R(N=c(100),ar=1,p_exp=c(0.475),p_ctrl=c(0.475),interim=c(0.3,0.5,0.75),Delta_fut=NULL,
P_fut=c(0.25,0.5,0.5),Delta_eff=c(0,0,0),P_eff=c(1,0.75,0.75),Delta_fin=0,P_fin=0.75, 
distrisize=10^3, PP_nsim=10^3, design_fut="futPR", design_eff="effPO",nsim=1000)

IDDI-BE/PhII_Bayes documentation built on May 19, 2021, 3:04 p.m.