BiostatisticsGlossary Term
ICH E9(R1) Section 3.5
The statistical determination of the number of participants required for a clinical trial to have adequate statistical power to detect a clinically meaningful treatment effect, accounting for expected variability, desired significance level, and anticipated dropout rates.
Sample size calculation represents a critical element of trial design that balances scientific rigor with practical and ethical considerations. An underpowered study risks failing to detect true treatment effects, wasting resources and potentially exposing participants to research procedures without adequate scientific return. An overpowered study enrolls more participants than necessary, exposing individuals to research risks beyond what is required for valid conclusions and consuming resources that could support other research.
The fundamental inputs to sample size calculation include the desired statistical power, typically 80% or 90%, meaning the probability of detecting a true treatment effect if one exists. The significance level, commonly 0.05, defines the acceptable probability of falsely concluding that an effect exists when it does not. The expected treatment effect, often based on prior studies or clinical judgment, specifies the magnitude of difference the trial aims to detect. The expected variability in outcomes influences the precision with which effects can be estimated. Additional factors such as anticipated dropout rates, multiple testing adjustments, and stratification affect the final sample size requirement.
Sample size calculations must be documented in the protocol and statistical analysis plan, including the assumptions underlying the calculation and their justification. Sensitivity analyses exploring how sample size requirements change under different assumptions help assess robustness. During trial conduct, the actual observed variability and event rates are monitored against assumptions, as departures may necessitate sample size re-estimation to maintain adequate power. Any modifications to planned sample size must be scientifically justified and implemented through protocol amendments.
Continuous endpoint
"Assuming a mean difference of 5 points on the symptom scale, standard deviation of 12 points, 90% power, and two-sided alpha of 0.05, the sample size calculation indicated that 122 participants per group were required, increased to 145 per group to account for 15% anticipated dropout."
Event-driven design
"For the time-to-event primary endpoint, sample size calculation determined that 350 events were required to detect a hazard ratio of 0.75 with 80% power, translating to approximately 700 participants based on expected event rates."
A range of values calculated from study data that is expected to contain the true treatment effect with a specified probability, typically 95%, providing information about both the estimated effect size and the precision of that estimate.
A statistical analysis strategy that includes all randomized participants in the groups to which they were originally assigned, regardless of whether they completed the study treatment or adhered to the protocol.
A planned statistical analysis conducted before all participants have completed the study, typically to evaluate accumulating data for evidence of efficacy, futility, or safety concerns that might warrant early termination of the trial.
The probability of obtaining results at least as extreme as those observed in the study, assuming that the null hypothesis of no treatment effect is true.
A statistical analysis that includes only participants who completed the study according to protocol requirements, without major protocol violations, adequate treatment exposure, and complete outcome assessments.
ICH E9(R1) Section 3.5
The statistical determination of the number of participants required for a clinical trial to have adequate statistical power to detect a clinically meaningful treatment effect, accounting for expected variability, desired significance level, and anticipated dropout rates.
Sample size calculation represents a critical element of trial design that balances scientific rigor with practical and ethical considerations. An underpowered study risks failing to detect true treatment effects, wasting resources and potentially exposing participants to research procedures without adequate scientific return. An overpowered study enrolls more participants than necessary, exposing individuals to research risks beyond what is required for valid conclusions and consuming resources that could support other research.
The fundamental inputs to sample size calculation include the desired statistical power, typically 80% or 90%, meaning the probability of detecting a true treatment effect if one exists. The significance level, commonly 0.05, defines the acceptable probability of falsely concluding that an effect exists when it does not. The expected treatment effect, often based on prior studies or clinical judgment, specifies the magnitude of difference the trial aims to detect. The expected variability in outcomes influences the precision with which effects can be estimated. Additional factors such as anticipated dropout rates, multiple testing adjustments, and stratification affect the final sample size requirement.
Sample size calculations must be documented in the protocol and statistical analysis plan, including the assumptions underlying the calculation and their justification. Sensitivity analyses exploring how sample size requirements change under different assumptions help assess robustness. During trial conduct, the actual observed variability and event rates are monitored against assumptions, as departures may necessitate sample size re-estimation to maintain adequate power. Any modifications to planned sample size must be scientifically justified and implemented through protocol amendments.
Continuous endpoint
"Assuming a mean difference of 5 points on the symptom scale, standard deviation of 12 points, 90% power, and two-sided alpha of 0.05, the sample size calculation indicated that 122 participants per group were required, increased to 145 per group to account for 15% anticipated dropout."
Event-driven design
"For the time-to-event primary endpoint, sample size calculation determined that 350 events were required to detect a hazard ratio of 0.75 with 80% power, translating to approximately 700 participants based on expected event rates."
A range of values calculated from study data that is expected to contain the true treatment effect with a specified probability, typically 95%, providing information about both the estimated effect size and the precision of that estimate.
A statistical analysis strategy that includes all randomized participants in the groups to which they were originally assigned, regardless of whether they completed the study treatment or adhered to the protocol.
A planned statistical analysis conducted before all participants have completed the study, typically to evaluate accumulating data for evidence of efficacy, futility, or safety concerns that might warrant early termination of the trial.
The probability of obtaining results at least as extreme as those observed in the study, assuming that the null hypothesis of no treatment effect is true.
A statistical analysis that includes only participants who completed the study according to protocol requirements, without major protocol violations, adequate treatment exposure, and complete outcome assessments.