2
$\begingroup$

I'm a medical doctor with no statistical skills, and I need some help on what kind of test to run.

I have a set of 20 patients who underwent CT brain perfusion multiple times (twice in most cases, up to 5 times for one case), meaning they had an injection of iodinated contrast medium, and the subsequent density enhancement was measured in blood vessels, as discrete points of density (y) in time (x). One measurement is performed every 2.5s (approximately) for 50 seconds, and the points in time are always the same for every procedure. I need to compare the intra-individual enhancement (or response to the injection), to assess whether the enhancement is statistically no different when the exam is repeated on the same patient under identical conditions (contrast medium, injection protocol, etc.). The only conditions that are not identical are the site of injection (elbow or wrist), the delay between the start of injection and the start of the brain scans (which is done manually), and sometimes the patient has an additional condition that messes with the enhancement profile (of cardiac origin mostly).

To simplify, I need to compare the solid line with the dotted line in each of the boxes below. First separetely, and then determine if the effect is the same over all the patients. I am interested in any kind of analysis that helps, like the difference between ascending slopes, peak density, area under the curves, and any others that I haven't thought of yet.

enter image description here

Thank you so much for reading, I look forward to reading your answers!

$\endgroup$
  • 1
    $\begingroup$ The answer from @thierryboulain provides an overview of some general ways you could proceed. If you could edit your question to be more specific about the aspects of these enhancement/time curves you wish to examine, several of which are noted in that answer, you might be able to get a more specific response. Details about the number of patients, the number of measurements on each patient (both at the same time if there are multiple regions examined, and the number of separate tests) would also help a lot. $\endgroup$ – EdM Dec 29 '19 at 20:37
  • $\begingroup$ Thank you for your comment @EdM. It's hard to know what to examine because my goal is to prove that the procedure is repeatable and any kind of comparison is relevant here. I would like to do an analysis as complete as possible. $\endgroup$ – Julien Vidal Jan 10 at 16:43
  • $\begingroup$ So follow the advice from @thierryboulain in the answer below, applied to all aspects of the procedure for which it is important to document repeatability. $\endgroup$ – EdM Jan 10 at 16:49
3
$\begingroup$

You may have different primary interests: peak density (the highest y), delay from injection to peak density, or overall increase in density (difference between areas under the solid and dotted curves). Whatever the measure of interest you choose, it can be compared between the different experiments using analysis of variance for repeated measurements. An alternative way is to use linear mixed model with patients handled as random slope and intercept variable, and experiments (order number of injections). You would be able to compare increases in density between experiments and also to compare each point of measurement between different experiments.

$\endgroup$
  • $\begingroup$ Welcome to the site, and thanks for providing an answer that covers major general approaches that could be used (+1). In my experience here I found that it can be good to try to get the OP to provide more details about the study, as then a more specific answer can be provided that will be more useful both for the OP and for other people who might come across your answer. I made a comment on the question that shows the type of things I try to ask about when I have the time and inclination. $\endgroup$ – EdM Dec 29 '19 at 20:43
  • $\begingroup$ @thierryboulain Thanks a lot for this answer, it helps a lot already! $\endgroup$ – Julien Vidal Jan 10 at 16:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.