I don't really know. I have never done or seen any rigorous testing of it. One problem in testing it is that some of the observations are time-frame specific. More of a problem is the issue that the observations are somewhat subjective and do not easily lend themselves to the mathematical definition needed to test them rigorously.
My own experience is that divergences in uptrends should be taken as a strong warning. I use them as an important part of my market risk management strategy. Details of how I do this are in my videotapes.
Note: Since writing this answer, I have dropped the Advance-Decline line from my investment plan. This basically because I found that I had sufficient warnings from the chart itself. The notorious downward bias on the A-D line was als an influence in this decision.