We appreciate Dr. Hahn's thoughtful consideration of the potential distinctions between and within USPSTF recommendation grade letters and the role of certainty in arriving at the recommendation grade.

His suggestion that the USPSTF never assign any grade higher than B to a recommendation based on the chain-of-evidence approach (in fact, his proposal would create a new B-minus category for any positive recommendation made on the basis of indirect evidence) implies that only evidence from randomized trials merits a judgment that "certainty of net benefit is high." In at least some cases, randomized trials of preventive services are not necessary to ascertain high net benefit. Support for smoking cessation counseling in primary care is, for example, an A recommendation with a high certainty of net benefit. It is based on the chain-of-evidence approach: counseling results in smoking cessation, smoking cessation prevents lung cancer, and prevention of lung cancer increases longevity.

The USPSTF also recognizes that the randomized-trial-as-evidence standard is unattainable for many clinical preventive services. It is not feasible, for example, to conduct a randomized trial of the effect of HIV screening in pregnant women on childhood morbidity and mortality for ethical reasons and because the sample size and cost of such a trial would be prohibitive. The USPSTF rated its recommendations for universal HIV screening in pregnancy as an A recommendation by using the chain-of-evidence approach: Screening detects HIV infection in the mother, treatment of the mother during pregnancy prevents HIV transmission to the newborn, and prevention of newborn HIV infection prevents HIV morbidity and mortality.

The USPSTF is not indifferent to Dr. Hahn's concern about the use of surrogate outcomes instead of patient-oriented outcomes as the desirable basis for recommending interventions, including preventive interventions. If randomized trials are feasible and necessary to draw a conclusion that has high certainty, they should be conducted, and conclusions about the use of the intervention should be delayed pending the results.

The USPSTF knows that much of its intended audience is unaware of the distinction between grade A and B recommendations—both are recommended for use in routine practice. As a potential guide to policymakers, those who design quality measurements, or even clinicians who may use the distinction between A and B recommendations for prioritization purposes, the USPSTF will continue to use both and reserve the grade A recommendation for services with the most ironclad evidence, and therefore the highest certainty, about the impact of the service on health outcomes.