CBL-514 Sample Size: Power, Precision and Rare Events
A participant count needs a purpose
The 2025 CBL-514 publication reports 76 randomized participants.[1] That count alone does not describe the strength of every possible inference. A sample-size calculation concerns a selected endpoint, assumed effect, variability and statistical design.
FDA’s E9 guidance treats sample-size planning as part of trial design.[2] When reading a CBL-514 report, locate the calculation and compare its assumptions with the actual enrollment and available outcomes. Do not calculate a new βobserved powerβ percentage from the result and treat it as a replacement for uncertainty estimates.
An original rare-event example
Hypothetical, not CBL-514 safety data: if an event independently affects 1% of exposed people, the probability of observing no events in 50 people is 0.99 raised to the 50th power, approximately 61%. Zero observed events in such a group would therefore be quite compatible with a 1% underlying event probability.
The example assumes independent observations, equal risk and complete observation. Real safety studies may not satisfy those assumptions. Its purpose is narrower: to show why βnone observedβ is not the same as βcannot occur.β
Follow-up also matters. Fifty people observed for a week and fifty observed for a year offer different opportunities to detect delayed events, even though the headline participant count matches.
What larger studies can and cannot fix
More informative observations generally improve precision, but additional participants do not automatically repair a biased measurement procedure, an unsuitable comparator or selective reporting. The quality of data and the relevance of the question remain essential.
For subgroup claims, record how many people actually contributed to that subgroup. A total enrollment figure can conceal very small cells after participants are divided by baseline severity, age or other characteristics.
Our suggested evidence note includes the planned size, actual randomized size, endpoint-specific count, follow-up duration and stated purpose of the calculation. That record supports a more useful conclusion than calling a CBL-514 trial simply βlargeβ or βsmall.β
Frequently asked questions
Does statistical significance show that a study was large enough for safety?
No. An efficacy test and the ability to detect uncommon or delayed harms are different questions.
Can I infer a subgroup result from the total sample size?
No. You need the subgroup's own participant count, analysis and uncertainty.
Continue your CBL-514 research
Compare the original study evidence with the documentation for the specific material you are reviewing.
