How do I calculate statistically by SPSS the cutoff value of a biomarker to differentiate between 2 groups if we had the confidence interval?
My question: I have marker called MMP3 which was significantly more in active rheumatoid arthritis (RA) disease than Inactive. I was asking if the means [ 95%CI ] of MMP3 in active was > than Inctive (means [ 95% CI). Can we find the cutt of value to differentiate between active and inactive RA by MMP3 By the following method: [Sum the lower edge of 95%CI in the active disease+ The upper edge of 95% CI of its value in inactive disease] then dividied by 2 = cut off value. Or is this method not correct?
You can't. At least, not with the information you are trying to use.
The first thing to remember is that only a perfect biomarker will discriminate perfectly between two groups. In the real world, you have to choose your cutoff point to maximise one of three (yes three!) aims
- the optimum sensitivity (best yield) of 'cases' detected
You won't be able to detect them all, probably, without an unacceptable false-positive rate, so you will need to balance the risks associated with failing to detect a case and the workload involved in the false positives.
- the optimum specificity
In this case, you base your detection around the maximum workload in terms of false positives that your system can handle. Useful for situations like type II diabetes where non-detection does not pose an immediate threat.
- total prediction accuracy
In this case, you wish to maximise the number of cases correctly classified.
I advise you to a) read up a little on the literature on sensitivity/specificity positive and negative predictive values and b) get hold of some software that will do this sort of thing for you. I use Stata, but there are other packages. Indeed, for all I know, you can purchase an add-on for SPSS for the cost of a holiday in Barbados that will do it. ;–)}
Dear Ronan :Regarding my question: I have marker called MMP3 which was significantly more in active rheumatoid arthritis (RA) disease than Inactive. I was asking if the means [ 95%CI ] of MMP3 in active was > than Inctive (means [ 95% CI). Can we find the cutt of value to differentiate Between active and inactive RA by MMP3 By the following method: [Sum the lower edge of 95%CI in the active disease+ The upper edge of 95% CI of its value in inactive disease] then dividied by 2 = cut off value. Or is this method not correct?
I must recommend an excellent paper by Xinhua Liu which discusses the strategies you can use to set a cut point. It also has an excellent bibliography. It would be my first choice in reading up on the area.
Liu, X. (2012). Classification accuracy and cut point selection. Statistics in Medicine, 31(23), 2676–2686. doi:10.1002/sim.4509
http://onlinelibrary.wiley.com/doi/10.1002/sim.4509/abstract
If a study is already published and we don't have access to the raw data to evaluate ROC, and all we are given is that Group A had mean 0.54+/-.06 and Group B had a mean of 0.46+/-.05 and groups were different (p
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS