14.3 Sample size re-estimation methods

Sample size re-estimation methods are crucial in adaptive clinical trials. They allow researchers to adjust participant numbers mid-study based on interim data, ensuring adequate statistical power while maintaining trial integrity.

These methods come in two flavors: blinded and unblinded. Blinded reviews use pooled data to estimate parameters, while unblinded reviews analyze treatment groups separately. Each approach has its pros and cons, impacting trial design and execution.

Blinded and Unblinded Sample Size Review

Sample Size Re-estimation Methods

Top images from around the web for Sample Size Re-estimation Methods

• 

  Applied Psychometrics: Sample Size and Sample Power Considerations in Factor Analysis (EFA, CFA ... View original

  Is this image relevant?

• 

  SMARTAR: an R package for designing and analyzing Sequential Multiple Assignment Randomized ... View original

  Is this image relevant?

• 

  Applied Psychometrics: Sample Size and Sample Power Considerations in Factor Analysis (EFA, CFA ... View original

  Is this image relevant?

• 

  SMARTAR: an R package for designing and analyzing Sequential Multiple Assignment Randomized ... View original

  Is this image relevant?

1 of 2

Top images from around the web for Sample Size Re-estimation Methods

• 

  Applied Psychometrics: Sample Size and Sample Power Considerations in Factor Analysis (EFA, CFA ... View original

  Is this image relevant?

• 

  SMARTAR: an R package for designing and analyzing Sequential Multiple Assignment Randomized ... View original

  Is this image relevant?

• 

  Applied Psychometrics: Sample Size and Sample Power Considerations in Factor Analysis (EFA, CFA ... View original

  Is this image relevant?

• 

  SMARTAR: an R package for designing and analyzing Sequential Multiple Assignment Randomized ... View original

  Is this image relevant?

1 of 2

• Sample size re-estimation involves adjusting the sample size during an ongoing clinical trial based on interim data analysis
• Can be performed in a blinded or unblinded manner depending on whether treatment group information is used
• Aims to ensure the study has adequate power to detect a clinically meaningful treatment effect
• Helps to address uncertainties in initial sample size calculations and adapt to changes in variability or effect size estimates

Blinded Sample Size Review

• Blinded sample size review is conducted without knowledge of treatment group assignments
• Utilizes pooled data from all treatment groups to estimate nuisance parameters (variability, response rates)
• Preserves the integrity of the trial by maintaining blinding and minimizing operational bias
• Requires careful consideration of the timing and frequency of interim analyses to avoid inflating Type I error

Unblinded Sample Size Review

• Unblinded sample size review involves analyzing data by treatment group at an interim analysis
• Provides more accurate estimates of treatment effect size and variability compared to blinded methods
• Requires strict control of Type I error rate through appropriate statistical methods (alpha spending functions, group sequential designs)
• May introduce operational bias and impact trial integrity if not properly managed

Internal Pilot Study Approach

Internal Pilot Study Design

• Internal pilot study approach involves using a portion of the total sample size as a "pilot" phase
• Data from the internal pilot is used to re-estimate the sample size for the remainder of the trial
• Allows for a more accurate assessment of nuisance parameters and effect size estimates
• Requires pre-specification of the internal pilot study design, including the timing and criteria for sample size adjustment

Effect Size Estimation in Internal Pilot Studies

• Effect size estimation in internal pilot studies is based on the observed treatment difference and variability
• Utilizes statistical methods to account for the uncertainty in effect size estimates from the pilot phase (adjusted confidence intervals, conditional power calculations)
• Helps to ensure the final sample size provides adequate power to detect the true treatment effect
• Requires careful consideration of the potential impact on Type I and Type II error rates

Conditional and Predictive Power

Conditional Power Calculations

• Conditional power is the probability of rejecting the null hypothesis at the end of the trial, given the observed data at an interim analysis
• Calculated based on the observed treatment effect, variability, and the remaining sample size
• Helps to assess the futility or promising nature of the trial and inform decisions on early stopping or sample size adjustment
• Requires specification of the true treatment effect and variability for the remainder of the trial (usually assumed to be the same as observed)

Predictive Power Calculations

• Predictive power is the average conditional power over the posterior distribution of the true treatment effect, given the observed data at an interim analysis
• Accounts for the uncertainty in the true treatment effect by integrating over its posterior distribution (based on prior information and observed data)
• Provides a more comprehensive assessment of the trial's prospects compared to conditional power
• Can be used to guide decisions on early stopping, sample size adjustment, or trial continuation based on the probability of success