Comparing 2 biomarkers using restricted cubic splines - how-to, interpretation, model building

[jroon]

Really interesting graphs!

Might this have to do with the distribution of creatinine at the extremes? See summary:

> summary(dat$creatinine)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
  37.00   67.00   80.00   92.09   99.00  967.00      84

Which, in my mind, then also means we should apply a spline to creatinine?

Hmm I don’t think creatinine needs a spline - because when we fit a spline to that above the line stayed straight. Also this uptick is showing up non-linear on the curve for MyC already. Hmmmm, I wonder could this be a small group of people who have really bad Cr but relatively normal MyC ? You could try plotting a curve with a higher Cr value, 300, 400 etc and see what those curves look like. Or you could actually look in the data and see are there people who experience AMI despite normal MyC - and then look at their creatinine. Maybe there are a handful of people with truly awful renal function and for some reason normal MyC ?

[tomkaier]

Or you could actually look in the data and see are there people who experience AMI despite normal MyC - and then look at their creatinine. Maybe there are a handful of people with truly awful renal function and for some reason normal MyC ?

That’s not it (worst histogram ever…) - all MyC levels for patients with creatinine >180:

You could try plotting a curve with a higher Cr value, 300, 400 etc and see what those curves look like

Even crazier. Lack of datapoints?

[jroon]

That’s not it (worst histogram ever…):

Sorry I’m confused by the histogram - how did you make it?

Even crazier. Lack of datapoints?

This again makes me think there must be at least 1 someone with crazy high creatinine, +ve AMI, and 0 or low troponin ? If there isn’t I am at a loss to explain this feature!

[tomkaier]

Sorry I’m confused by the histogram - how did you make it?

Sorry. The histogram contained only patients with creatinine >180 umol/L. Reviewing the raw data, all patients with AMI and creatinine >180 have a MyC >140 (almost twice 99th centile, and way above LOD (0.4 ng/L))…at least we’re confused together, but this is rather unsettling.