Lecture 7 Sample size calculation

Question 1

Why are sample size calculations necessary?

Answer 1

• Necessary for grant providers
• Necessary for medical ethical committees
• Necessary for the ‘consort statement’

Question 2

What do you need for sample size calculations?

Answer 2

• Expected effect (clinical relevant effect)
• Standard deviation of the outcome variable
• Power en significance

Question 3

What is the formula for continuous variables

Answer 3

Formula: ….

N = Sample size (per group)
Z(1-alpha/2) = (1- alpha/2) percentile of the standard normal distribution
Z(1-beta) = (1- beta) percentile of the standard normal distribution
= standard deviation of the outcome
r = ratio of the numbers in the groups
v = difference between the groups

Question 4

Different formulas for sample size calculations

Answer 4

1. Conservative adjustment formula
2. Liberal adjustment formula
3. Mostly used/ accepted formula

• conservative formula is the better formula
• the other two will give you a non-significant value

Question 5

Dichotomous outcomes

Answer 5

Less possibilities to change paramaters

- standard deviation directly depends on the clinical relevant effect (which is not the case in continuous outcomes)

Question 6

Types of errors

Answer 6

Alfa error (almost always 5%)
Beta error: Power = 1 - beta (accepted to use values higher or equally to 80% generally)

Question 7

What are sample size calculations based on?

Answer 7

Based on testing theory

• This is a silly thing, it’s not as good as people think
• Sample size calculations are therefore based on silly things

Question 8

What do sample size calculations highly depend on?

Answer 8

They depend on arbitrary assumptions

- You can change them easily therefore you can use any number you need

Question 9

Analysing of RCT data

What do you distinguish between?

Answer 9

Distinguish between:
- one follow up measurement
( comparing change over time (between baseline and follow-up) between the two groups
- more than one follow- up measurement