The General Accounting Office study compared all cancer patients in the province of Ontario with a sample of the entire US population of such patients (4 types of cancer over a given period of time).

Such gross comparisons are akin to attempting to find a scientific/policy important needle in a hay stack.

In that they aggregate the outcomes of all patients (the relatively rich and poor, those who live in large metropolitan areas to rural-remote areas, etc.) they are not capable of observing any interesting/important effects among such subgroups.