Statistical power is the probability with which a statistical test correctly detects a genuinely existing effect – that is, rejects the null hypothesis (H₀) when it is in fact false. Power = 1 − β, where β denotes the probability of a Type II error (a false negative finding, i.e. failure to detect a true effect).
In clinical research, a power of at least 80% (β ≤ 0.20) is usually targeted, and for pivotal registration trials often 90% (β ≤ 0.10). Power depends on four factors: the significance level (α), the expected effect size (e.g. difference in means), the variability of the measured variable (standard deviation) and the sample size (n). Higher power requires – for the same α and the same variability – a larger sample size.
Sample size calculation is a central task in study design and must be documented in the study protocol. Insufficient power leads to so-called underpowered studies, which fail to reach significance despite a true effect being present. For CROs, correct sample size calculation – taking dropout rates and adjustments for multiple testing into account – is a critical biostatistical task. Regulatory reference: ICH E9.