Sample size estimation

Planning a study
A crucial step in study planning is determining the required sample size. In a well-planned study, the sample size is calculated with a predefined likelihood of showing an effect of the primary study objective given that this effect exists (power).

Setting the objective
Precise formulation of the primary objective is imperative in order to calculate the required sample size. Does the planned study aim to demonstrate superiority or non-inferiority, or is the aim to estimate a population statistic with a certain degree of precision?

Specifying desired power
When the primary objective has been defined, the researcher is responsible for defining the probability of detecting an effect given that it exists (power). Often, this value is chosen arbitrarily. By using a commonly used value of 90% power, for example, the researcher accepts a 10% risk of not detecting an effect given that it exists. In reality, the chosen power represents a delicate balance among financial, practical and ethical trade-offs.

Evaluating the effects of unknowns
Besides the study aim and power, which are determined by the researcher, two other factors influence how large a sample must be in order to achieve the desired power:

  • How large are the differences between study arms, or in what range are the estimated effects expected to be?
  • How much variability in the data is expected?

These two factors are mostly unknown before the study. It can therefore be useful to evaluate the required sample size for a range of assumptions in order to assess their sensitivity.

Statistical support from DKF
If we assist you in your study design, we will guide you through the points discussed above. Instead of simply presenting you with a number when asked to calculate the sample size, we will illustrate the effect of various assumptions about the unknowns on the required sample size (see figure below).