Sample size calculation determines the number of participants a trial needs to have a good chance (power) of detecting a pre-specified treatment effect on the primary endpoint, while limiting the risk of a false positive (alpha).
Free sandbox · No credit card · 21 CFR Part 11 aligned
Per arm
93
+15% dropout
110
Total
220
In short
Try it
Adjust the control and treatment response rates, alpha, power and dropout. For other endpoint types, work with a statistician.
You need, per arm
108
91 completers + dropout margin
Total to enroll (2 arms)
216
Two-proportion z-test, equal allocation. A rough planning estimate, not a substitute for a biostatistician on your protocol.
Explained
The most common mistake is an optimistic effect size. Early-phase results often overestimate effects, and a trial powered for an effect that is twice the real one has much less than its nominal power. Basing the effect on the smallest clinically meaningful difference, often informed by the MCID, is more robust.
Other frequent problems include using a variability estimate from a different population, forgetting dropout, and powering for the primary endpoint while planning confirmatory claims on secondary endpoints that are underpowered. The sample size calculator handles the common two-proportion case; complex designs such as cluster, adaptive or Bayesian trials need a statistician.
The probability that a trial will detect a treatment effect of the specified size if it truly exists. 80% or 90% is typical.
The probability of concluding there is an effect when there is none (a false positive). 0.05 two-sided is typical.
Participants who leave reduce the analysable sample. Inflating enrolment keeps the planned power.
A statistician, using assumptions agreed with the clinical team and documented in the protocol.
Build the study your calculation supports. Free sandbox.