Reviewed in the WSJ, Geoffrey Kabat's new book talks about the unreliability of epidemiological studies. Quoting Kabat:
The highly charged climate surrounding environmental health risks can create powerful pressure for scientists to conform and to fall into line with a particular position.
He analyzes several suspect claims, including the claim that secondhand smoke causes cancer, and "he finds more bias than biology".
This reminds me of a New York Times article from 2007 that takes a similarly critical look at epidemiology:
The goal of the endeavor is to tell those of us who are otherwise in fine health how to remain healthy longer. But this advice comes with the expectation that any prescription given -- whether diet or drug or a change in lifestyle -- will indeed prevent disease rather than be the agent of our disability or untimely death. With that presumption, how unambiguous does the evidence have to be before any advice is offered?
The overall theme is that solid results from large lifestyle or diet studies are few and far between. For example:
In January 2001, the British epidemiologists George Davey Smith and Shah Ebrahim, co-editors of The International Journal of Epidemiology, discussed this issue in an editorial titled "Epidemiology -- Is It Time to Call It a Day?" They noted that those few times that a randomized trial had been financed to test a hypothesis supported by results from these large observational studies, the hypothesis either failed the test or, at the very least, the test failed to confirm the hypothesis: antioxidants like vitamins E and C and beta carotene did not prevent heart disease, nor did eating copious fiber protect against colon cancer.
And this recent Wired article talks about the "placebo crisis":
Half of all drugs that fail in late-stage trials drop out of the pipeline due to their inability to beat sugar pills. [...] It's not that the old meds are getting weaker, drug developers say. It's as if the placebo effect is somehow getting stronger.
What to do? I'm not a doctor, of course, but one approach seems promising to me -- "individualized medicine". Large-scale studies look at the median performance of drugs, treatments, diets, etc., but individual responses vary greatly. From the SFI bulletin (PDF, sorry, skip to page 16):
We need to look at how various clinical markers change over time for each patient, and based on that -- possibly including genomic data as well -- be able to make a specific, individualized diagnostic at real-time. That will allow us to find out when a patient is developing a disease and treat it before it becomes serious.