Statistical analysis: Comparing effects of drug doses on blood glucose over time

#1
Hi all,

New user and new to biostats so please go easy on me!

I am working on a project where I am determining the effects of drug doses on blood glucose levels over time, after a single dose. Technically, there are two measured variables for each "dose" 1.) blood glucose levels and 2.) time. However, my impression is that a 2-way ANOVA is inappropriate for statistical analysis because this analysis will not take into account the order of the time points. So if I want to answer the question: Is there a significant difference between blood glucose levels for each of these doses at this single time point? what would be the most appropriate test?

Shying away from some sort of complicated nonlinear regression of the dose response over time, I was thinking of comparing blood glucose levels for 5 different doses at a single time point and thought I could possibly use ANOVA, followed by Tukey's?

Any help or advice (even if it involves making fun of my noobery) is welcome!

Thanks!
 
#2
Well if you're treating five sets of patients at t=0 is baseline glucose (fasting or otherwise) and then applying the dose, waiting whatever time and measuring again you can use Tukey's between each stratum afterwards and ANOVA across stratum. However blood glucose has a LOT of variance so you'll need a decent sample size. If your patients are diabetic it's even more difficult.

But ANOVA/Tukey's could work and paired T-tests. I've never done analysis on Blood Glucose but have read a few studies back when I was getting my masters and I remember using one way ANOVA in a similar setting. It's not a STRONG study method IMO because blood glucose is best studied continuously with repeated measures but one step at a time works I guess.

There could be more efficient methods but I don't know them. I could make a few assumptions to limit confounders/misclassification due to excursion but if you're more familiar with that then apply it appropriately.

Good luck!

Oh anyone else feel free to comment; I'd love to learn more about blood glucose analysis.
 
#3
Thanks Lowpro!

These are preclinical studies in mice, so for a preliminary analysis I think I will go with the ANOVA/Tukey's for now (anyone see anything wrong with this?!)

You're so right about variability in blood glucose. FYI, I settled on a study design where animals were fasted for 5 hours before dosing and this really seemed to help with the variability. What I am really interested in is if drug tx results in "hypoglycemia" which I am considering below 100 mg/dL blood glucose. Right now, I have a n=4/dose, with glucose being measured at 15min, 30min, 1.5hr and 4hr postdose with water ad libitum and food unavailable until after the final measurement.

Thanks again. I am happy to discuss further if anyone is interested.
 
#4
Correction: I ended up using ANOVA as we said before, then I used Dunnet's to compare each dose to a vehicle dosed group at each time point.