More Repeated Measures Problems!

Hi guys,

I have a similar question to johnstephenbyrne. I am analysing the results of a clinical trial. I have measured the levels of a cytokine before and after (at 3 time points) giving a drug to one group of 20 patients. I want to know if the levels of the cytokine change significantly with time after treatment. One problem is that the measurements are taken at different time points for different patients - but (very broadly) the 1st reading is between 25-50 days after treatment, the 2nd between 50-100 days and the third sometime after 100 days. So the data look a bit like this:

Patient 1
Day 0= 10 pg/ml; Day 30= 20 pg/ml; Day 68= 23 pg/ml; Day 145= 40 pg/ml

Patient 2
Day 0 = 0 pg/ml; Day 22= 5 pg/ml; Day 52= 8 pg/ml; Day 121= 5pg/ml


My thoughts are either:

Linear regression for time after treatment vs change in cytokine levels from baseline.


One-way repeated measures ANOVA

Do either of these sound sensible? For the repeated measures ANOVA, could I categorize time after treatment into three groups, eg 25-50 days, 50-100 days and 100+ days?

Any input would be greatly appreciated!

I would use the One way repeated measures ANOVA and simply bin the time intervals as you have described. This will be the easiest and most straight forward way to do it.
Could a mixed model with patient as the random effect be used ? ...

Would it make any sense to take into account possible serial correlation between measures?

This would NOT be the easiest or most straightforward way of doing it :D