In preparation for a subsequent full scale research study, we discuss further the recruitment, adherence and retention, the choice of a control condition, the effects of cognitive and physical activity intervention on cognition and their consistency with age, as well as predictions of future sample sizes based on the pilot data.
Recruitment for trials of training-based interventions
SHARP-P recruitment was based on mailing lists and presentations at health education meetings. SHARP-P randomized 73 out of 343 participants aged 70-85 years who were screened on the telephone for a yield of 21.3%. While this yield exceeds that of some other pilot trials [11, 23], others who have used a greater breadth of approaches such as print advertisements and television features  or who have targeted institutionalized cohorts  have achieved higher yields (e.g., 45%-60%).
Adherence and Retention
Attendance rates were higher in the cognitive training group and the combined group than in the physical activity group, with the highest rates in the cognitive training group. However, this difference may be due to the availability of make-up sessions in the cognitive training condition. Participants were not provided the opportunity to make up missed physical activity visits. Attendance declined only slightly over time except in the physical activity alone arm where the decline tended to be more pronounced. SHARP-P physical activity attendance is similar to that observed in LIFE-P using a tapered contact schedule: 76% during the first 8 weeks (3 sessions per week) and 65% for weeks 9 to 25  (2 sessions per week). These attendance rates are also consistent with what has been reported in a review of numerous randomized controlled trials of physical activity and older adults . The expectation for participants to re-book missed cognitive training sessions may have motivated the participants in the combined group to attend both cognitive training and physical activity sessions. Furthermore, the PACT condition may have offered a more novel and interesting intervention that served to enhance interest and commitment and thus higher levels of attendance. Additionally, the fact that cognitive training and physical activity sessions were offered on the same day may have facilitated attendance at both sessions.
In terms of adhering to the physical activity goals, participants did not achieve the 150 minutes per week of aerobic physical activity goal. This goal was chosen because it represents the level of physical activity that is recommended by the American College of Sports Medicine . However, although reporting of physical activity adherence in older adults is inconsistent across studies , previous research indicates that achieving recommended levels of physical activity in sedentary samples of older adults is quite difficult. In a sample of formerly sedentary older adults, the LIFE-P intervention included a phased contact schedule (3-days/week during the first 6-months) and a behavioral group-counseling module and reported approximately 128 min/wk of physical activity during the first 6 months of the trial . The present study included only 4 months of physical activity, did not include a behavioral group-counseling module, and had fewer center-based contacts than LIFE-P. Future randomized clinical trials investigating the impact of physical activity on cognition should include structured behavioral modules to address the multitude of barriers to physical activity encountered by sedentary older adults to maximize adherence to study protocols .
Choice of a control condition
We adopted, as our control condition, the Healthy Aging Education program of the Lifestyle Interventions and Independence for Elders Pilot (LIFE-P) trial, which had been developed to foster retention in a clinical trial of a physical activity training intervention while not directly affecting its outcomes, markers of mobility and physical function [22, 23]. We found, across four months of follow-up, retention in this comparison condition to be equivalent to those of the training-based interventions, which parallels the experience of the LIFE-P trial.
A substudy, involving 102 participants, added measures of cognitive function to LIFE-P Pilot at baseline and one-year post-randomization to compare a physical activity intervention to the same control condition we used . Similar to our study, no significant relative benefits were found for the physical activity intervention. Also similar, modest improvements in cognitive function were seen in the performance of participants assigned to the Healthy Aging Education program on several measures of cognitive function. These reached statistical significance for a test of verbal learning. While these may reflect learning effects, it is also possible that the Healthy Aging Education program may have affected cognitive function. Increased social engagement and reading, both features of the Healthy Aging Education program, may produce cognitive benefits in older individuals . Furthermore, within the full LIFE-P trial, the Healthy Aging Education program was associated with modest mean improvements in a battery of physical function tests, which although less than for its physical activity intervention, may signal that the control condition was not benign with respect to other markers of health .
Choice of a control condition in behavioral interventions can be controversial from design and ethical perspectives [38, 39], and active control conditions may reduce estimated effect sizes . SHARP-P incorporated the control condition to enhance retention and reduce differences in exposure to study staff among intervention groups, with the aim of providing an appropriate contrast to the mechanisms specific to the physical activity and cognitive training interventions. Even so, exposure times varied among intervention conditions: these are often difficult to balance with behavioral interventions and may contribute to relative intervention effects. Although the SHARP-P control condition may produced some improvements in performance on cognitive tests, one may argue that for behavioral interventions to be adopted as part of clinical practice, they should be demonstrated to provide benefits above whatever effects the more modest and less resource intensive Healthy Aging Education program provided. However, others argue that true placebo controlled trials are necessary in situations where no effective therapies have been established and mechanisms may be non-specific [41, 42].
Short term effects of cognitive and physical training on cognitive performance
Assignment to our cognitive and physical activity interventions did not produce significant relative improvements in our overall composite measure or in the separate components focused on executive function and episodic memory after four months. As designed, the SHARP-P trial targeted the detection of intervention mean relative effects of 20% differences on domain specific tests in marginal comparisons, which was equivalent to an average of 0.64 standard deviations across domains. Thus, the observation of fairly substantial differences was required for statistical significance in this pilot study. For the composite primary outcome, post hoc power calculations from the observed standard errors of differences indicate 80% power was available to detect mean marginal differences of 0.49 standard deviations. Although SHARP-P, as conducted, provided sufficient power to detect the targeted differences, observed differences were smaller than these. Our composite outcome was projected to provide greater statistical power than its individual components if intervention effects were distributed among its components. Composite outcomes allow one to deal efficiently with problems of multiple comparisons . However, if intervention effects vary markedly among individual components, some important differences may not be detected. Our results should be interpreted later in the context of larger trials.
Of great interest, functional magnetic resonance imaging was conducted at the conclusion of the trial on five participants who had been assigned to the Healthy Aging Education condition and six who had been assigned to the PA. Despite the small sample size, the hippocampi of PA participants had significantly greater cerebral blood flows and network connectivity than Healthy Aging Education participants . Thus, despite the lack of an effect on our cognitive measures, four months of physical activity training may have produced beneficial effects on other measures of brain function.
Several recent small trials report stronger intervention effects on cognitive measures. Fabre, et al. found that logical memory, paired associates learning scores and memory quotient were significantly improved after two months of aerobic and/or mental training compared to a control group . Smith, et al. found four months of a multi-component behavioral intervention that included physical exercise, weight loss, and diet was associated with a relative improvement in measures of executive function in overweight/obese individuals with mean age 52 years . Baker, et al. found six months of aerobic exercise training produced significant benefits in measures of executive function in women; these were not evident in men and did not occur after three months of intervention . Liu-Ambrose found significant improvements in executive function after 12 months of resistance training in women with mean age 70 years, but not at 6 months . Albinet, et al. found 12 weeks of aerobic training improved executive function in older sedentary adults compared to stretching . Three other recent pilot trials, however, found no significant intervention effects on cognitive outcomes. Stuss, et al. found three months of a memory, goal management, and psychosocial training did not significantly affect measures of memory . As noted above, Williamson, et al. found no significant differences in executive function and other domains after 12 months of a physical activity intervention compared to the healthy aging control condition adopted by SHARP-P; however, an analysis pooling data across intervention groups found a significant positive correlation between improvements in physical function and cognitive function . Dechamps, et al. found little improvement in global cognitive function after 12 months of an exercise program .
Overall, it appears that the greatest effects may be for interventions that 1) last longer than four months, 2) target executive function, 3) feature a minimal control condition and, possibly, 4) are multi-factorial. The choice of executive function as a primary outcome is supported by meta-analysis . However, combining individual measures into a composite outcome, as we did, can sometimes dilute focused effects . In contrast, examining neuroimaging data following intervention may provide a more sensitive outcome measure than cognitive tasks.
Consistency across age ranges
We investigated if follow-up and intervention effects were attenuated by age in our participants. We found no evidence, across the relatively short time frame of four months, of poorer adherence rates among our older volunteers. Given issues of comorbidities and other threats to intervention participation, one might expect longer trials to face greater issues. Interestingly, however, the Diabetes Prevention Program found that older participants (>65 years of age) were more likely to achieve the physical activity and weight loss goals of the study . The eligibility and recruitment approaches used in SHARP-P appeared to be effective for identifying older individuals who can be adequately adherent to the interventions being offered, at least in the short-term.
Of great potential interest is that we found a trend that physical activity training may convey greater relative short-term benefits on cognitive function among our older volunteers. Assignment to the physical activity intervention appeared to lead to improvements in cognitive function that increased, in a graded fashion, across the age range of 70-85 years. While this comparison was pre-specified, we cannot claim statistical significance due to the many inferences we conducted in our pilot study. All SHARP-P participants self-identified as having cognitive deficits; while this is known to be a weak risk factor for future cognitive decline [48, 49], it is possible that this relationship may signal differences in responses between individuals who experience relatively earlier versus later declines. There is precedence for greater benefits of physical activity training among individuals with lower baseline functioning in other outcomes. For example, Marsh, et al. found baseline lower extremity functioning moderated the influence of two different walking programs on improvements in physical functioning . Whereas those with higher levels of functioning responded more positively to a traditional walking program, the lower functioning participants responded more positively to a novel walking program that included more complex walking tasks, such as stepping over obstacles.
Sample size projections for the full-scale trial
Using benchmark data generated by SHARP-P, we projected that a two-armed four-year trial to detect relative differences in cognitive deficits can be mounted with fewer than 1,000 participants (continuous outcome) or 2,000 participants (categorical outcome). The required sample sizes are dependent on event rates, and thus on the risk profiles of participants. Trials aimed at affecting the endpoints of mild cognitive impairment or dementia typically would be expected to require larger sample sizes because events rates may be lower .
While large enough to meet its objectives, our pilot trial involved a modest sample size and short follow-up. Our participants were predominantly Caucasian and reported relatively high levels of education. We examined only two training-based interventions; it is likely that the most effective programs may include tool-box approaches with multiple options [1, 2]. Similarly, the physical activity program was a relatively short, traditional walking program and did not include anaerobic forms of activity, such as strength training. Evidence suggests that longer physical activity programs (>6 months) that include both aerobic and anaerobic exercise have greater effects than aerobic exercise alone . Additionally, the nature of the Healthy Aging Education control condition may have influenced outcomes.