HELP: If you would like help in completing the sample size calculation click help off. Help Off
Type
Binary Yes/No Type Variable (e.g. response vs no response)
Design
Parallel Group
Objective
Non-inferiority
1) Determine what you expect the outcome percentages will be in the two groups: boxes (a) and (b)
For example a trial is being designed to compare two treatments with an outcome of Response to Treatment. The usual assumption of non-inferiority is that the two groups would have the same response rate, in this case put the same value in (a) and (b). If this is not the expectation put the values you expect to observe in (a) and (b). E.g. if you expect the test product to give an outcome of 55% and the comparator to be 53% then put 55 in (a) and 5 in (b)
Expected % in Group A
(a)
Expected % in Comparator Group
(b)
2. Determine the Non-Inferiority Criteria: box (c)
This is the pre-specified amount that the test product can be 'worse' than the comparator and still claim non-inferiority. The maximum value this can take is the difference between the comparator and placebo but in reality this should be a clinically meaningful amount. E.g. previous trials have shown that the comparator had a response rate (outcome variable) of 53% and the placebo showed the response rate of 33%, hence an absolute difference of 20%. If we allowed the non-inferiority margin to be greater than this 20% then we would not be demonstrating the test product was effective, if we set it at 20% it still may just be as "effective" as placebo so hence a margin perhaps in the region of 10% would be a sensible choice.
Non-Inferiority Criteria
(c)
3. Determine the Power Level: box (d)
In simple terms power is the chance of the trial demonstrating a significant difference if the assumed values in box (a) and (b) are correct. If the assumed values (a) and (b) are incorrect then the power will not be what is stated. Most trials are conducted with 80% or 90% power. For 80% power put 80 in box (d)
Desired Power Level
(d)
4. Determine the Withdrawal/Non-evaluable Rate: box (e).
The sample size calculator will initially determine how many subjects are required to complete the trial (and appear in the final analysis). A percentage of patients will usually not complete the trial and therefore will not contribute to the final analysis. If you expect that 10% of your subjects may do this then put 10 in box (e)
Withdrawal/Non-evaluable rate
(e)
NOTE: If we get a large number of users and/or shares then we will add more sample size calculators to the website and make these free to ALL users.
