Hi,

I am trying to analyze a data set. The question we are looking at is whether the rate of a successful outcome and the risk of a complication change with pass number when trying to remove a clot from a vessel.

The data is organized by patient number (rows), and outcome of each pass (column). There are a few ways we are analyzing this - continuous (what % of the clot is removed) and categorical [any change in clot? (Y/N); complete removal? (Y/N); complication (Y/N)] per pass.

Because the treatment ends when the clot is removed, there are different numbers of patients for each pass (i.e., 100% of the patients got 1 pass, 80% 2 passes, 60% 3 passes, etc., etc.). We expect that as the pass number increases, the risk of a complication increases, and the likelihood of successful revascularization decreases.

For the continuous variable (% clot removed), should I be using repeated measures ANOVA or Friedman's test? Do I need to look at the normality of the distribution of each condition (pass number) to decide which?

According to the research I've done on the internet, the correct test for looking at the categorical variables is Cochran's Q. I had never heard of this test before, so I wanted to check that this seems correct.

Alternatively, I could analyze, say, complication rate as a 4 x 2 contingency table (number of passes needed for success 1, 2, 3 or 4+ vs. complication (y/n)). This is trickier for likelihood of success by pass number. Would this be more appropriate, at least for complications?

Thank you for your time!

Sincerely,
Jennifer