In dungtsa/ClinicalTrialInterimAnalysis: Sample size for clinial trial with interim analysis report

========================================================

library(knitr)
library(rmarkdown)
options(scipen=3)

options(width=60)
opts_chunk$set(comment = "", warning = FALSE, message = FALSE, echo = TRUE, tidy = FALSE, size="small",fig.height=5)

#tmp99<-get.result.MultiArm()

         power <- somevalues()$pow 
         typeI <- somevalues()$typI 
         cutoffStage1 <- somevalues()$cut1 
         cutoffStage2 <- somevalues()$cut2
         PETH0 <- somevalues()$PETH0  
         PETH1 <- somevalues()$PETH1

Statistical Plan

Sample size justification: A total of r input$n1+input$n2 patients will be enrolled in the first r input$time.accrual months with an estimated accrual of r ceiling((input$n1+input$n2)/input$time.accrual) subjects per month. All the patients will be followed up for at least r input$time.followup months. We hypothesize a median r input$endpointLong (r input$endpointShort) of r input$medH1 months (H1: alternative hypothesis) in the treatment group compared to r input$medH0 months (H0: null hypothesis) in the historic control group (Hazard ratio (HR)=r input$medH0/r input$medH1=r round(input$medH0/input$medH1,2)). r input$endpointShort is defined as r input$endpointDescription. An interim futility analysis will be performed after the first r round(input$time.accrual/2) months (r input$n1 patients enrolled). A statistic, W, utilizes the expected events (Event.E) under the null hypothesis to compare to the observed events (Event.O) and defines the stopping boundary where W = (Event.O-Event.E)/sqrt(Event.E). If W is higher than r cutoffStage1 (i.e., the number of observed events is higher than expected, meaning a shorter r input$endpointShort), the treatment will be considered ineffective and the trial will be stopped. Otherwise, additional r input$n2 patients will be enrolled in the second r round(input$time.accrual/2) months. At end of the study, if W is less than r cutoffStage2 (i.e., the number of events is lower than expected, meaning a longer r input$endpointShort), the treatment will be considered promising. Simulation analysis show that the design has r power*100% power to detect the effect size of HR=r round(input$medH0/input$medH1,2) controlled at one-sided r typeI*100% type I error. Probability of early termination (PET) is r PETH0*100% if the true median r input$endpointShort is r input$medH0 months under the null hypothesis with an expected sample size of r round( PETH0*input$n1+(input$n1+input$n2)*(1 - PETH0)) (r PETH0 * r input$n1 + r input$n1+input$n2 * r 1- PETH0). When the true median r input$endpointShort is r input$medH1 months (H1), PET is r PETH1*100 % and the estimated sample size is r round( PETH1*input$n1+(input$n1+input$n2)*(1- PETH1)) (r PETH1 * r input$n1 + r input$n1+input$n2 * r 1- PETH1).

Data Analysis

Kaplan-Meier curves of estimated r input$endpointShort will be generated. One-sided log-rank test will be used to test if the treatment group yields a longer r input$endpointShort compared to the historical control (median r input$endpointShort of r input$medH0 months). Median r input$endpointShort times with 95% confidence intervals will also be determined. As defined in the protocol, the survival analysis will be based on the intent-to-treat population, which will include all eligible patients enrolled in this study. This will be also performed using multivariate Cox regression model by incorporating variables of interest as covariates in the model. Data will be summarized overall using descriptive statistics. For example, continuous data will be summarized with number of patients (n), mean, median, minimum, maximum, standard deviation, coefficient of variation, and geometric mean (where applicable). Categorical data will be summarized using frequency counts and percentages.

dungtsa/ClinicalTrialInterimAnalysis documentation built on Dec. 20, 2021, 2:15 a.m.