Develop analysis plan on Medical data - trend analysis and model

#1
First, let me say thank you for any help you can give. I have been searching for nearly two days for a solution to my problem and continue to be confused. I am sure the answer has been right in front of me, but it has been a couple years since I had to perform statistical analysis in SAS. I am using SAS for the analysis of the data described below.

I have the following data:
Medication status (binary - yes/no)
Age (quantitative)
Race (categorial)
Smoking (binary)
Site (binary - where tooth extraction took place)
Site status (binary - case/control - each individual can be their own case and control)
Medical measurement at baseline (quantitative)
Medical measurement as 3 months (quantitative)
Medical measurement at 6 months (quantitative)

Main hypothesis:
Will there be a less increase in medical measurement at the site in patients who are taken medication versus those that do not?

When I first received this data, I did some recoding to make the data easier to analyze in SAS. Ever since I have been stumped on where to go. I know that I want to answer the following question: is there a significant change over time for medical measurement? To begin my analysis.

I cannot seem to figure out how to answer this question properly. I have gone back and forth on whether to use proc glm or proc corr or even proc freq.

My plan is, is that if I find there to be significant change over time in medical measurement then I would keep the medical measurement stratified. If this be the case, I would then like to model if medical measurement is affected by medication status, site, age, race and/or smoking. Medication status would be my intercept and the remaining variables I would add to the model individually to see if a significant relationship is present. For those found to be significant I would include them in my final model.

Example of the model equations I would test for significance:
medical measurement = medication status + age
medical measurement = medication status + race
medical measurement = medication status + site
medical measurement = medication status + smoking

I am trying to figure out if for my outcome variable I should take the difference between medical measurement taken at the case site versus the medical measurement taken at the control site within each individual at baseline, then at 3 months and at 6 months - creating 3 different outcome variables for each individual.

As you can probably tell, I have several different thoughts on how to analyze this data, but have problems executing them. I cannot remember the correct statistical tests to test any of my possibly analysis plans.

Again, I appreciate any help you can give. Thank you.

Best Regards,
J
 
#2
Hi J,
Since your main dependent variable (medical measurement) is continuous, you usually think of regression and/or ANOVA. And since in your case the dependent variable is measured more than once (3 times), it leads you to repeated measures (RM) ANOVA.

To start, you may want to run a one-way RM ANOVA using proc glm. Then move to a two-way RM ANOVA by including the medication status variable as your independent variable - this will address your hypothesis. Then, include any control variables you want in your final model.

Hope that helps.