Limited research done by us in South Africa shows that none of the current tests available (including CAD) reliably differentiate between safe colour deficients and truly unsafe ones. We have found that some colour deficient persons are unreliable in critical colour perception based on testing in full simulator facilities, but there was no correlation with Ishihara, Farnsworth CAD or Signal testing.
The problem will remain: How do we identify the risky vs the safe pilot? Should a colour deficient pilot be tested in a standardised and calibrated simulator by means of a set of set tests. Should this testing be left to the instructor rather than the doctor, or should it be assessed by partnering the instructor and doctor? How many colour related mistakes are made by those with normal colour perception during similar testing?