Web Application for simulating operating characteristics of the Bayesian CRM

Division of Translational Research & Applied Statistics, University of Virginia; nwages@virginia.edu


1. Enter an assumed set of true DLT probabilities, separated by commas. Note: The length of this set should be equal to the number of possible study dose levels.

2. Enter the target DLT probability that defines the MTD for the study.

3. Enter the cohort size required before the next model-based update. Cohort size may be 1, 2, or 3 patients.

4. Enter the maximum sample size for the study. This number should be a multiple of the cohort size entered above.

5. Enter the total number of patients treated on any dose required to stop the trial. At any point in the trial, if the recommendation is to assign the next cohort to a dose that already has the entered number of patients treated on the dose, the study is stopped and the recommended dose is declared the MTD. If the entered number is larger than the maximum sample size, each trial will accrue to the maximum sample size.

6. Enter the number of simulations. A minimum of 1000 is recommended.

7. Enter the index of the starting dose level. Note: Index of lowest dose level is always 1. If the design allows for 'minus' dose levels (i.e. -2, -1, etc.), then the index of the starting dose should account for these lower levels (i.e. if -1 dose level allowed, starting dose is 2.)


8. Set the seed of the random number generator.

9. Specify the confidence level for safety stopping rule at the lowest study dose level.

        

          

            This application simulates operating characteristics for the Bayesian continual reassessment method [1] with the following specifications.
          

          

            1.
            Skipping Restriction:
            The trial is not allowed to skip dose levels when escalating.
          

          

            2.
            Skeleton:
            For the specified target DLT rate and total number of dose levels, the skeleton of power model d^exp(a) is generated according to Lee and Cheung (2009) [2] using a prior MTD located at the median dose level and a spacing measure of delta=0.05.
          

          

            3.
            Prior:
            The prior distribution on the parameter
            a
            is a mean zero normal distribution with the least informative prior variance [3].
          

          

            4.
            Safety Stopping Rule:
            Stop the trial for safety if the lower limit of an Agresti-Coull binomial confidence interval [4] for the lowest study dose level exceeds the target DLT rate
          

          References:
          

            [1] O'Quigley J, Pepe M, Fisher L (1990). Continual reassessment method: a practical design for phase I clinical trials in cancer, 
            Biometrics;
            46
            (1): 33-48.
          

          

            [2] Lee and Cheung (2009). Model calibration in the continual reassessment method, 
            Clinical Trials;
            6
            (3): 227-238.
          

          

            [3] Lee and Cheung (2011). Calibration of prior variance in the bayesian continual reassessment method, 
            Statistics in Medicine;
            30
            (17): 2081-2089.
          

          

            [3] Agresti A, Coull BA (1998). Approximate is better than 'exact' for interval estimation of binomial proportions, 
            American Statistician;
            52
            : 119-126.
          

        

        
      

      

        
Web Application for implementation of the Bayesian CRM

        

          Division of Translational Research & Applied Statistics, University of Virginia;
          nwages@virginia.edu
        

        

        

          Design / Protocol Information
        

        

          1.Enter the index of the starting dose level.
          Note:
          Index of
          lowest
          dose level is always 1. If the design allows for
          'minus'
          
                           dose levels (i.e. -2, -1, etc.), then the index of the starting dose should account for these lower levels (i.e. if -1 dose level allowed, starting dose is 2.)
        

        

          
          
        

        
1. Enter the target DLT rate probability that defines the MTD for the study.

        

          
          
        

        

        

          Observed Trial Data (do not count 'replaced' patients)
        

        

          2. Enter number of observed DLTs at each dose level. If none have been observed or a dose level has not yet been tried, enter '0'.
          Note:
          The length of this set should be equal to the number of possible study dose levels.
        

        

          
          
        

        

        

          3. Enter the number of patients evaluated for DLT at each dose level. If a dose level has not yet been tried, enter '0'.
          Note:
          The length of this set should be equal to the number of possible study dose levels.
        

        

          
          
        

        

        
4. Enter the most recent dose level administered in the study.

        

          
          
        

        

        
5. Specify the confidence level for safety stopping rule at the lowest study dose level..

        

          
          
        

        

          
        

        

        

          

            This application computes a recommended dose level for the next patient in a phase I trial according to the Bayesian continual reassessment method [1] with the following specifications.
          

          

            1.
            Skipping Restriction:
            The trial is not allowed to skip dose levels when escalating.
          

          

            2.
            Skeleton:
            For the specified target DLT rate and total number of dose levels, the  skeleton of power model d^exp(a) is generated according to Lee and Cheung (2009) [2] using a prior MTD located at the median dose level and a spacing measure of delta=0.05.
          

          

            3.
            Prior:
            The prior distribution on the parameter
            a
            is a mean zero normal distribution with the least informative prior variance [3].
          

          

            4.
            Safety Stopping Rule:
            Stop the trial for safety if the lower limit of an Agresti-Coull binomial confidence interval [4] for the lowest study dose level exceeds the target DLT rate
          

          References:
          

            [1] O'Quigley J, Pepe M, Fisher L (1990). Continual reassessment method: a practical design for phase I clinical trials in cancer, 
            Biometrics;
            46
            (1): 33-48.
          

          

            [2] Lee and Cheung (2009). Model calibration in the continual reassessment method, 
            Clinical Trials;
            6
            (3): 227-238.
          

          

            [3] Lee and Cheung (2011). Calibration of prior variance in the bayesian continual reassessment method, 
            Statistics in Medicine;
            30
            (17): 2081-2089.
          

          

            [4] Agresti A, Coull BA (1998). Approximate is better than 'exact' for interval estimation of binomial proportions, 
            American Statistician;
            52
            : 119-126.
          

        

        
      

      

        
Web Application for generating safety stopping bounds at the lowest study dose level based on Agresti-Coull binomial confidence interval estimation.

        

          Division of Translational Research & Applied Statistics, University of Virginia;
          nwages@virginia.edu
        

        

        
1. Specify the confidence level for safety stopping rule at the lowest study dose level.

        

          
          
        

        
2. Enter the target DLT rate probability that defines the MTD for the study.

        

          
          
        

        
3. Enter the maximum sample size for the study.

        

          
          
        

        

          
        

        

        

          

            This application computes the safety stopping bounds for the lowest study dose level based on Agresti-Coull binomial confidence interval estimation [1].
          

          References:
          

            [1] Agresti A, Coull BA (1998). Approximate is better than 'exact' for interval estimation of binomial proportions, 
            American Statistician;
            52
            : 119-126.