Ezra Klein had a post last week about the accuracy of projections about health care, in which he called reports like the recent one from HHS a sort of Rorschach test--conservatives and liberals each see in them what they want to. I think this is true, to a point. The history of health care spending projections is considerably more checkered than either liberals or conservatives acknowledge. Over time, it is true that far more health care programs have busted their budgets than have come in considerably under budget; on the other hand, some cost cutting programs have ended up netting more than expected. But virtually all of the reports written on this question consists of simply cherry picking your examples very carefully in order to generate the answer you want. This would be a perfect job for the CBO, except that members of Congress request those reports, and I doubt either party is willing to risk getting the wrong answer.
But there is one example of "cost underruns" that I keep hearing, which I think should be used cautiously if at all. That's Medicare Part D, which came in substantially beneath projections.
As it happens, I just turned in the first draft of a piece on the status of prescription drug pipelines. You'll have to wait to read my opinions on that question, but investigating the question has given me a better appreciation of just how unique an environment surrounded the enactment of Medicare Part D. There was a broad shift in the market for pharmaceuticals on several fronts: fewer blockbuster drugs were being approved, and more blockbuster drugs were going off patent. Meanwhile, pharmaceutical benefit managers were really cracking down on what drugs went into their formularies.