Hi all,
Would you please advise me how to use the "minimum clinically important difference - MCID" to calculate sample size for interventional studies? Are there any web based sample size calculators that you would like to recommend?
Thank you!
From what I can determine, there's no easy way and it would involve the following steps:
1. decide on outcome measure
2. select mean value +- SD/SE of said outcome measure in control group for calculation
3. Determine what % change your available MCID is from the mean selected (for example, lets say your PROM is a 10 point scale, where a change in 0.8 point is the MCID, and the mean value of that scale you have selected for your control group for calculation of sample size is 8, then a 10% difference from the mean would be clinically significant)
4. use that %change as your effect size
5. Use a sample size calculator by setting your alpha and beta error values, along with selected mean and %change/effect size value to determine sample size
Dear Nguyen,
I think you can use G*Power software program for your purpose. Here is a link below that you can freely download the program:
http://www.gpower.hhu.de/en.html
hi Tu N Nguyen,
Whether you were able calculate sample size with MCID? I am also looking for solution. G power does not give the option to do sample size with SD/SE. But as the MCID is calculated with ROC there is no data on SD/SE.
From what I can determine, there's no easy way and it would involve the following steps:
1. decide on outcome measure
2. select mean value +- SD/SE of said outcome measure in control group for calculation
3. Determine what % change your available MCID is from the mean selected (for example, lets say your PROM is a 10 point scale, where a change in 0.8 point is the MCID, and the mean value of that scale you have selected for your control group for calculation of sample size is 8, then a 10% difference from the mean would be clinically significant)
4. use that %change as your effect size
5. Use a sample size calculator by setting your alpha and beta error values, along with selected mean and %change/effect size value to determine sample size
Hi I think this could be the possible way.
I used an alternative method.
In my research I used FMA lower extremity score as primary outcome and wanted to calculate sample size. But majority of the studies reported only 2 point change in mean, which I felt was too small for clinically important change. That is when I posted my question.
I identified the study which mentioned MCID for FMA lower extremity. The reported Mean and SD for the groups which was divided based variable used for MCID calculation. I just used these Mean and SD to calculate my sample size.
- Badges
- Science topic
- Similar topics
- Applied Mathematics
- Calculations