BookmarkSubscribeRSS Feed
🔒 This topic is solved and locked. Need further help from the community? Please sign in and ask a new question.
Jay4
Calcite | Level 5

I have 5 timepoints (week 0, 2, 6, 12, 26) and the change from Baseline (BL) at week 12 is the variable interested. To work out the sample size for a future trial I would like to estimate the SD from a data set (N=400).  Snce the original data are highly skewed the change from BL was log transformed (taking the ratio of visit xx/visit 0). I ran the following sas code to get the SE for each visit and then thought multiply by sqrt of n at that visit would give me the required SD?

PROC GLM  Data=powa2  ;
Class sub prot visit ;
Model logratio = base visit visit*base / solution ;
LSMeans visit/  cl alpha=0.05;
Run;
quit;

 Can I use the SE for log transformed means for visit=12?

Another way I am thinking is to run the model is to run model by visit and use the RMSE as the standard deviation for sample size calculation. 

Which way is better or is there any other way I should get the SDs for sample size cal.

1 ACCEPTED SOLUTION

Accepted Solutions
SteveDenham
Jade | Level 19

Please read Frank Harrell's excellent comments regarding analysis of change from baseline data (which boils down to - DON'T analyze change from baseline)  located here:

 

http://biostat.mc.vanderbilt.edu/wiki/Main/MeasureChange

 

Now, there is an additional problem.  The standard error on a log transformed estimate cannot be easily back-transformed to the original scale.  You would need to apply the delta method to get a reasonable approximation.

 

Please consider what is presented in that link before continuing.  Once you have, then it is far more likely that you can estimate sample size through simulation than implementing a closed form solution that doesn't correctly model the data.

 

Steve Denham

View solution in original post

2 REPLIES 2
SteveDenham
Jade | Level 19

Please read Frank Harrell's excellent comments regarding analysis of change from baseline data (which boils down to - DON'T analyze change from baseline)  located here:

 

http://biostat.mc.vanderbilt.edu/wiki/Main/MeasureChange

 

Now, there is an additional problem.  The standard error on a log transformed estimate cannot be easily back-transformed to the original scale.  You would need to apply the delta method to get a reasonable approximation.

 

Please consider what is presented in that link before continuing.  Once you have, then it is far more likely that you can estimate sample size through simulation than implementing a closed form solution that doesn't correctly model the data.

 

Steve Denham

Jay4
Calcite | Level 5

Dear Steve, Many thanks and very helpful advise & thanks for the link too. I am surprised to see that FDA is considering change from BL as an endpoint for most neurology clinical trials. The change in cognition from BL measured as a score arising from a questionnaire at BL & at a specified Follow up visit (e.g 24 weks). When the chg from BL is not normal it is log transformed.  I was just tryingto use of those datasets to calculate the SD to work out a sample size for a future trial. But I understand that it is not advisable.

So I ran the model as the response (not change) which is also unfortunately lognormal and keeping BL as a covariate.

I am still stuck as the SE is given in log. Perhaps I will workout the sample sizes considering the log values of the response rather than the orginal scale.

Once again many thanks for your swift response and valuable advise

Best regards

Jay4

sas-innovate-2024.png

Join us for SAS Innovate April 16-19 at the Aria in Las Vegas. Bring the team and save big with our group pricing for a limited time only.

Pre-conference courses and tutorials are filling up fast and are always a sellout. Register today to reserve your seat.

 

Register now!

What is ANOVA?

ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 2 replies
  • 3814 views
  • 1 like
  • 2 in conversation