Can you better describe the study design? In particular being on drug x. So they were on x for varying time periods or not and when was outcome collected for x and y? Example data would help!
Hi everyone,
I have a prospective cohort of 90 subjects who were previously on drug X (duration of drug X = "varx"), now on drug Y and followed for 24 weeks. Continuous dependent laboratory variables were checked at week 0 and at week 24.
1) What type of analysis can I run to evaluate how "varx" affected variables at week 0?
I'll be running a paired analysis (likely nonparametric) to evaluate changes from week 0 to week 24 while on drug Y. However,
2) What kind of analysis can I run to evaluate how "varx" correlate with changes while on drug Y?
Should I create a categorical variable for duration of therapy on drug X based off quartiles? Linear regression analysis? I'm not sure what to do.
Thanks in advance!
Can you better describe the study design? In particular being on drug x. So they were on x for varying time periods or not and when was outcome collected for x and y? Example data would help!
Stop cowardice, ban guns!
Prospective cohort followed for 24 weeks; labs performed at week 0 and week 24. The cohort was switched form drug X (with variable duration) to drug Y at week 0. Repeat labs done at week 24.
Example Data:
Subj | Time on drug X | Week 0 GFR | Week 24 GFR
1 44 months 113.5 113.5
2 80 121.1 103.8
3 0 76.24 80.2
4 32 155.1 152.1
5 60 96.2 97.4
For question 1, I thought using linear regression analysis with the dependent variable being the baseline laboratory values and independent variable being the duration on drug X would suffice.
For question 2, I though using linear regression analysis with the dependent variable being the delta values between the two time periods would suffice.
Any thoughts?
But you don't seem to have a baseline variable for before X was initiated, do you?
Stop cowardice, ban guns!
What can you do with it then, since subject could have GFR = 110 then on X 20 months then had GFR = 110. So its effect is completely unknown. So you want to predict effect of Y while controlling for baseline value of GFR and prior weeks of X. That seems more reasonable then try to model varx!
This disconnect is where you are confusing me.
Stop cowardice, ban guns!
The question is whether longer duration of X affects week 0 variables or possibly has an effect on how drug Y affects at week 24.
The first part seems simple with linear regression. The second question does seem a bit unreasonable...I was thinking of recalculating a new value as a delta value from week 24 and week 0 and using linear regression again. Does this seem somewhat reasonable?
blocksteady,
You keep top-rock'in around my issue, you need to freeze and take it to the floor to provide more substantive info. How can you predict the effect of X's course on GFR if you only have GFR at time "0" which appears to be after the application of X. You need to have the value of X before the initiation of X. Unless you know that, but didn't listed it up above.
Stop cowardice, ban guns!
blocksteady (08-29-2017)
Tweet |