Table 3 in Appendix C summarizes the literature review by describing the evidence at each link in the analytic framework (Figure 1 in Appendix C). The first column lists the numbered link and the quality of the evidence in parentheses (only for link 1). We did not systematically review the evidence at links 2 and 3. The second column, labeled "critical path," reveals where emphasis was placed in evaluating the evidence. The primary questions in this review are listed in column 3. The fourth column lists a summary of findings, and the final column presents the supporting evidence.

The purpose of this review has been to determine whether adults counseled in primary care settings improve and/or maintain physical activity behavior. Several recent trials on efficacy of counseling for physical activity in primary care of good and fair quality demonstrated modest or no improvements in physical activity behavior. Most studies had at least one of the following limitations: provided limited details on the counseling intervention, had fair follow-up rates, excluded nonresponders from the analysis, studied selected provider populations, reported differences in physical activity levels at baseline between intervention and control groups, and/or reported uncertain or low provider compliance. In several trials it was difficult to assess whether patients had actually received a physical activity behavioral intervention.

Assessing a patient for sedentary behavior in these trials often took a few minutes (or more). More work is needed to develop a brief tool to correctly identify sedentary adults. The accuracy of current assessment systems and methods to measure physical activity levels is uncertain. Self-report physical activity tools are only moderately valid and reliable in identifying sedentary patients (42). Only one trial in this review attempted to collect self-reported physical activity and also a measure of fitness (Vo₂ max) (14). Fitness measures are considered more objective than self-report measures and are correlated with physical activity levels. New technologies are needed to directly measure physical activity intensity, duration, and frequency over time (43).

More research is needed regarding the use of office systems (e.g., electronic reminder systems, chart stickers, incentives for completed assessments) to improve assessment compliance (44, 45). The PACE methodology has recently been implemented using an interactive (Web-based or hard drive) communication technology for adults and adolescents (46). Just prior to the visit with a clinician, the patient completes the computerized assessment program in the waiting room. The patient then takes the printed action plan in to the visit with a clinician.

Changes in the health care system are needed to support effective physical activity interventions (45). Providers currently have limited options for reimbursement. Federal and state payers of Medicaid could set up reimbursement measures for ongoing physical activity counseling (45). Currently, clinicians are not reimbursed directly for reinforcing an active lifestyle.

We still do not have definitive evidence about whether specific physical activity-promoting behavioral strategies used in primary care settings affect long-term patient behavior (25). Because most of the cited studies in this review followed patients for less than two years, we know very little about patients' maintenance of activity. Similarly, most physical activity interventions in primary care have focused on changing sedentary behavior to active behavior. Many people move between being sedentary and being active at different times in their lives. That is, this progression is dynamic not static (47). In designing a study to address an information gap on maintenance, the investigators will need to follow interventions for several years and study which strategies best maintain physical activity (43).

Only one trial in this review reported adverse effects.(14) We need large prospective studies that report types and intensities of exercise, and injuries for more than two years. The trials should clearly describe both the intervention patients and protocol and the usual care patients and protocol. These trials document why patients drop out of studies. Understanding the potential harms and revising future interventions to reduce them may improve patient compliance. It is likely that some of the nonresponders in trials experienced an adverse effect and stopped exercising.

Because sedentary behavior is one of the most important risk factors for premature mortality and morbidity, a concerted community- and clinic-based effort will be needed to improve physical activity levels among Americans. With the methodological limitations described earlier, we found it difficult to assess the efficacy or effectiveness of the interventions. The evidence is inconclusive. To identify effective behavior change strategies we need trials with: more robust study designs; settings with typical baseline physical activity counseling rates and typical activity levels among patients; quick, reliable, and valid assessment tools; clearly defined counseling interventions; longer follow-up time periods (2-10 years); methods to ensure good provider compliance and complete reporting of it; and methods to reduce harms and good documentation of adverse effects. Finally, we need to design delivery and financial systems to translate effective strategies into practice.