http://books.nap.edu/html/health_behavior/

"...The committee noted the vast array of interventions at various levels, with varying endpoints on different populations, with different methodologies. Each committee member brought to the table their own perspective about what would be most effective, but the data were inadequate to convince any of the experts of a best approach to shaping and maintaining behavior change. A critical obstacle to answering definitively the question of what works best is the difficulty of generalizing the findings of current studies. Many factors contribute to the problem: outcome measures among the studies differ, populations studied differ, and methodologies differ. For example, an intervention may be exceptionally effective on a highly motivated population but fail for the general public. Measurement of a behavioral outcome such as self-reported tobacco use is difficult to compare with an outcome measure such as change in sales of tobacco. Another obstacle is that there are no rigorous evaluations of interventions. Evaluations may assess short-term changes, but long-term effectiveness should also be assessed..."