Which test to use

#1
Hi. Hope someone can help with my data.

Essentially it’s a clinical trial where we gave patients a dye and checked for it’s presence in the abdomen using a fluorescence camera (intervention) and a plain white light camera (control). We have a binary outcome (present or not) as well as signal intensity (continuous variable) for the intervention data.

I want to see whether the fluorescent dye was superior to the control and what the probability is
Over time that the dye is better than the control (with graphs for these).

Additionally I want to fit a model to see whether other confounders (eg dose of dye, weight, sex etc) have any effect on the difference.

We are dealing with small numbers (40 patients).

Hope someone can help!

Tom
 
#2
My first thought was a logistic regression model, but the fact fluorescence was different for some of the interventions might be tricky. Does it make sense to pool the signal into low and high intensity? Then you could try multinomial logistic regression.

Should also be easy to add confounders into such a model but have to be careful there are not too many for your sample size. How correlated are these anyway, I mean was dose of dye calculated on patient's weight?.

That's my guess anyway: a grown up should be along shortly.