Sample size calculation in a 3 group parallel clinical trial

#1
Hello everyone.

I'm having trouble with biostatistics in general but particularly with finding out how to calculate what sample size i need in setting up a clinical trial with the 3 groups.

The set up is as follows:

3 groups ( placebo, low dose, high dose)

Minimal relevant change in end point: 0.3 ng/ml (pre and post measurements)

SD 0.6

alpha: 0.05

Power 80%

I'm working in SPSS (but I have access to STATA, Sigmaplot, PRISM if necessary)

Any help would be very much appreciated.

Kind regards

TK
 
Last edited:
#3
I don't know exctly what you mean by SRS but I'm guessing something to do with random sampling?

The participants all suffer the same condition and are block randomised into treatment the groups (1:1:1). I hope that answers the question.
 
#4
oh, I was asking if they were randomly assigned or if you were pulling from say different groups.

If they are randomly assigned, you dont have to adjust for intraclass correlation, so your power analysis is much simpler.
 
#5
oh, I was asking if they were randomly assigned or if you were pulling from say different groups.

If they are randomly assigned, you dont have to adjust for intraclass correlation, so your power analysis is much simpler.
That was fortunate.

Simpler works...I'm all for it. Especially when it comes to biostats.
 
#6
well lets walk through this, I have a feeling this is a homework problem :).

So, you want to detect a change of .3 when you had a standard deviation of .6

That means you need a sample for your groups that can detect an effect size of .5

So now you have your beta level- .8
your alpha level- .05
and your effect size to detect -.5

You should have all the information you need to do the final step :).
 
#7
I'm still not getting it (as stated I'm struggling with biostats in general as I haven't used biostats I learned years ago....well for years).

I think I need (in brief) to see the intermediate steps to follow and then an explanation on how to use them to do the final calculations of total n (or n in groups) as I would surely need to explain my power calculations in a funding application e.g. when that time arrives.

I hope you can help.
 

rogojel

TS Contributor
#8
hi,
this would be a anova sample size calculation which is generally performed with some stats SW package. All the necessary inputs are already collected in the previous discussion with the42up - you just need to use your stat package, fill in the data and habe a go at it.

regards
 
#9
Ok....I think I might get it to some extent now.

The change we want to detect is 0.3 which is the minimal difference between the mean of the groups we want to be able to detect. We normalize this change by the SD and then arrive at the effect size (delta) for the groups of 0.5 =(0.3/0.6) Correct?

I found this little supplementary programme package to STATA (which I normally never use cause biostats are hard enough without coding issues) called fpower (example of use here http://www.ats.ucla.edu/stat/stata/dae/fpower.htm).
And heureka(below)....I've got useful output that I can mess around with to design my study. Granted I don't know the inner workings of the program or theory behind it but if it is good enough for STATA I don't have a problem with it.

If I'm incorrect in any of the above or anybody has pros/cons about the fpower programme please let me know.

Otherwise thank you for your time and help. It was much appreciated.


. fpower, a(3) delta(0.5) alpha(0.05)
(1 observation deleted)

a = 3 b = 1 c = 1 r = 1 rho = 0 delta = .5

nobs power
2 .0581418
3 .0681818
4 .0784195
5 .0888281
6 .0994166
7 .110184
8 .1211226
9 .1322216
10 .1434687
12 .1663546
14 .1896742
16 .2133234
18 .2372028
20 .261219
25 .3213024
30 .380531
35 .437967
40 .4928984
45 .544813
50 .5933703
55 .6383743
60 .6797466
75 .7825794
100 .8930816
125 .9507151
150 .9784294