Dear Sarawasti

Presentation of sample size calculations in papers is one of the fields in which I see more nonsense than just about any other, so it's good that you are asking advice.

Firstly, depending on what exactly you mean you really don't need to look for a 'similar study'. What you DO need (ideally) is

i) A reasonable idea about what a clinically meaningful minimum change on the measure you are planning to use will look like OR some other of quantifying the possible importance of any change. In a worst case scenario, this reduces to standardised effect sizes 'small' 'medium' or 'large'.

ii) IF (as is likely) you do not have the resources to power your trial to the smallest clinically meaningful effect (or small standardised effect sizes) your study can only have adequate power to detect a bigger effect. In this case your study does risk type 2 error if the true effect is real, potentially important but not as large as the study is powered for

iii) In this case, power calculations can tell you what your study CAN achieve but it is only really directly informative if you have a very strong hypothesis about the actual effect of your intervention. While other studies might give a clue, if they were that robust you wouldn't need a study... Other studies of psychosocial interventions on these outcomes in this patient group might give a clue as to what is achievable 

I think it is likely that what you really need to do here is to undertake a pilot where the sample size calculation is used to inform you of the limits (you'll find a significant result only if the intervention has a strong effect). 

As for your other question: "I want to know that whether is it ok to divide the group into hypnosis and control group " - it is certainly OK but I would strongly urge you consider some form of 'attention' control as well since there is simply no way a 2 group design can tell you whether the effect you are looking at is specific to hypnosis or just a product of the interactions involved.

Peter 

Dear Peter,

Thank you very much.

Hi Saraswati

Peter's advice is excellent. If you decided to power up for a medium effect, then you would need 64 patients in each study arm for a 2-arm trial with 80% power and alpha=0.05. If as Peter suggested, you go for a 3-arm trial, then you would need 69 patients per study arm to allow for (say) two pairwise comparisons and with alpha set to 0.04 to adjust for multiple comparisons. Although choosing a medium effect size might seem a bit arbitrary, a difference of 10 on a 100-point scale (eg the SF-36) has often been found to be of clinical importance. Further, a 100-point scale is likely to have an SD of around 20. This gives an effect size of 0.5 (or medium), which is what we have powered for.

Adrian

Dear Adrian,

Wow! Thank you so much for more clarification. I got idea. I will use EORTC QLQ c3o to measure HRQoL. 

Sincerely

Saras

hi in such case you use non random sampling