likely to be cost-effective and whether it should be a component of a strategy to prevent or reduce underage drinking. Inevitably, connecting the dots between evidence and policy requires a contextual judgment. How strong does the evidence bearing on effectiveness have to be to justify an intervention of this particular type in light of its likely range of costs? In making these judgments and designing the proposed strategy, the committee has been guided by several general considerations:
In dealing with complex social phenomena, such as underage drinking, comprehensive, multipronged strategies usually work best. As we note above, one of the reasons to embrace a portfolio approach is to capture the synergistic effects of coordinated and reinforcing interventions. Moreover, although any one intervention may produce no effect at all or an effect too small to detect, it might make an important contribution to a multipronged strategy.
It is necessary to distinguish between what is possible and what is likely. This distinction has two parts. One is between efficacy (what can be achieved in an experimental design?) and effectiveness (what can be achieved in the real world?) The second involves implementation. An intervention may be effective in a real-world context when it is carried out in faithful conformity with the recommended protocol, but not otherwise, or the effects may vary widely in relation to the quality of implementation. Whether a particular intervention should be included in a national strategy must depend on a judgment about implementation—how often would it be deployed effectively and with what cumulative effect—and whether that effect is worth the cost.
One must carefully consider the risk that an intervention will produce a harmful effect. Some interventions may have a perverse effect—in the context of underage drinking, perhaps a media campaign or a school-based education program could have a boomerang effect that stimulates alcohol use rather than depressing it. This risk may be especially great if a program with proven effectiveness with a specific group is implemented for another group or is poorly implemented. Moreover, an intervention that is effective overall may have widely varying results for subpopulations, including harmful outcomes for some of them. This possibility raises an important ethical concern in balancing benefits and risks. The committee has been sensitive to any evidence that an intervention presents a risk of harm to any youth subgroup and suggests ways of reducing such risks when they might exist.
Specific evidence of effectiveness for refinements of an intervention known to be effective and for which investments have already been made (e.g., limiting access) is not required. Because it is rarely possible, at this time, to quantify either the anticipated benefits or costs of proposed inter-