A good easy to do , quick test for colorectal cancer is something diagnostic labs have been wanting (and trying to make) for many years. Now, DNA-direct has introduced a new genetic test. The test is based on methylation of the vimentin gene.
As a learning excercise for myself (and my lab), I’m going to run through the numbers to try to assess whether this is a good test or not and consequently whether we should look into something similar (like this commercial test-kit).
The Bayesian method of assessment is very general and can be used for any diagnostic test where the following is known:
- Disease incidence – Colorectal cancer in US approx 50 / 100 000 = 0,05 % (numbers from National Cancer Institute).
- Test specificity (true negatives) – approx 86 % (from DNA-direct web site).
- Test sensitivity (true positives) – approx 75 % (from DNA-direct web site).
Now what we want to do is calculate the Bayesian probability of actually having cancer when you have a positive test result (procedure from here). Summary of calculations as follows: i) 75 % of 50 true cancer patients test positive = 37,5, ii) 100-86 = 14 % of 99950 true healthy patients test positive = 13993. Bayesian probability i/ii = 0,27 %.
The test may be cost-effective as a general population screening test since a positive test means your chances of having colon cancer at the time of testing has increased from 1/2000 to 6/2000. In addition, if you choose the right age group (aged over 50), screening using this test may be a good thing. Decisions on population screening however, are made by health officials on a national level and not by individual labs.
Representing an individual lab and seeing that the chances of you having cancer when you receive a positive test result using this test is, – believe it or not -, o,3 %, which means that out of a thousand people tested only 3 of them will be diagnosed correctly, – the only possible conclusion must be:
Worse than a coin flip, – Not a test I would use in my laboratory.