>I'd look first at confounding variables like 33 years of improved methods to detect cancer earlier: the rise of MRI, screening for chemical signatures of cancer (like PSA), etc.

Early detection is one thing, but is cancer generally hard to detect by the time it's fatal?