BookmarkSubscribeRSS Feed
☑ This topic is solved. Need further help from the community? Please sign in and ask a new question.
JoakimE
Obsidian | Level 7

Hi,

 

Background:

I am working with a clinical trial in which an endpoint of interest is to analyse a parameter in an organ in the body (deliberately vague language here, but hopefully clear enough to understand the statistical considerations). The trial has a parallel group design in which patients receive either active treatment or placebo. The organ in question is subdivided into 14 sections and measurement are taken for each section at baseline and post-baseline (two time points). So each patient has in total 28 observations. There is no a priori logical hierarchy between the sections. The organ is simply "sliced up" in 14 regions/sections.

 

Current analysis suggestion:

My current idea is to analyze this as an extension of ANCOVA, where the baseline value is added to the model as a covariate, but where the repeated measures in each patient (correlations between sections) are modeled. The proposed code is:

 

proc mixed data=indata;

          class patient treatment section;

          model post_baseline_value = treatment section baseline_value / ddfm=kenwardroger2;

          lsmeans treatment / diff cl adjust=tukey ADJDFE=ROW;

          repeated section / subject=patient(treatment) type=UN;

run;

 

An unstructured covariance structure is proposed, since there is no clear organisation/structure/hierarchy of sections. Since the analysis is akin to a "repeated measures ANCOVA", this model will use 14 observations per patient (post-baseline measurements is modeled using the baseline measurements).

 

Questions:

Does this model seem reasonable to the problem at hand? Are there alternative approaches I should consider? I am not entirely sure the repeated statement as defined is suitable in this context.

 

I can provide additional details if needed. Any input would be greatly appreciated!

 

Kind Regards,

JE

1 ACCEPTED SOLUTION

Accepted Solutions
SteveDenham
Jade | Level 19

The only change I can think of for the model is to include an interaction term for treatment by section, but that may only be meaningful if sections are made at approximately the same place within an organ. If that is the case, you may want to consider some of the spatial covariance structures as well as UN.

 

Returning to the interaction term and whether it should be included - is there any reason to suspect that the active group may demonstrate different responses for various sections. If so, then the interaction term is probably needed to get a better idea of differences within the organ.

 

SteveDenham

View solution in original post

2 REPLIES 2
SteveDenham
Jade | Level 19

The only change I can think of for the model is to include an interaction term for treatment by section, but that may only be meaningful if sections are made at approximately the same place within an organ. If that is the case, you may want to consider some of the spatial covariance structures as well as UN.

 

Returning to the interaction term and whether it should be included - is there any reason to suspect that the active group may demonstrate different responses for various sections. If so, then the interaction term is probably needed to get a better idea of differences within the organ.

 

SteveDenham

JoakimE
Obsidian | Level 7

Many thanks Steve! Then I conclude I was not totally barking down the wrong tree. I primarily posted to get a second opinion on the model.

 

I shall consider your thoughtful points on the interaction term as well.

sas-innovate-2024.png

Available on demand!

Missed SAS Innovate Las Vegas? Watch all the action for free! View the keynotes, general sessions and 22 breakouts on demand.

 

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
  • 738 views
  • 0 likes
  • 2 in conversation