Introduction

Sample size calculations¬†– don’t panic! Before an investigator starts collecting data, they need to decide how many patients they want to recruit. It’s easy to cheat by just keeping going until you get the result you want!

Trial sizes depend on the rate of outcome and the clinically relevant difference. The investigator then has to say what power they want and what type 1 error rate they are willing to accept

The type 1 error is the chance that you find an effect when actually there isn’t really one. Power is the chance of detecting a difference if one truly existed and is usually above 80%.¬†An acceptable type 1 error rate is set at 5% by convention.

Power is worked out by 1 minus the type 2 error. (A type two error is the chance of not finding an effect when actually there is one.)

The power (of a trial) is the chance of detecting a difference between the groups if one truly exists.