Sounds like generalized damage to white blood cells they're detecting. It's my understanding that "cancers" of a sort kind of exist in pockets in most everyone - they're just not the sort that get aggressive and kill people, because those mutant pockets just don't break enough of the rules of good cell conduct yet to count as a notable risk.
My big issue with the methodology is that when anyone has already detectable active cancer, they usually are on chemo, or too sick to stop the progress... both of which will cause generalized damage to the body's defenses. If they can reliably distinguish the kinds of damage though, that would be a nice development.
Even as it is stated, sounds useful to help distinguish some symptoms from cancer perhaps - but it seems this could also correlate with radiation damage or other generalized damage too. Cool study all the same - perhaps may help lead to cheaper or more automated screening at some level.