Methods

Study design

We conducted a 1-year RCT that primarily examined the effects of GE on physical functioning and falls in seniors living in retirement villages.16 Based on previous work,81319 we hypothesized that 6 months of exercise training would be sufficient to have beneficial effects on cognitive performance and mood. Thus, we measured cognitive performance and mood at baseline and 6 months later.

The study was approved by the University of New South Wales Human Research Ethics Committee, and all participants gave written informed consent.

Participants

The sample was a subset of men and women who participated in our previous RCT and has been detailed elsewhere (fig 1).16 Briefly, this subset sample comprised 154 participants (19 men (12.3%), 135 women (87.7%); mean age 79.6 (SD 6.3) years) who were drawn, after randomisation, from four intermediate care and four self-care retirement village sites (ie, 8 of the 20 villages from the RCT) in Sydney, Australia. The age range of the participants was 62 to 95 years. Participants were excluded from the study if they had a neurological, cardiovascular or musculoskeletal problem that precluded safe participation in an exercise programme; had a Mini Mental Status Examination score (MMSE)20 of <20; were in the hospital or away from the village at the time of recruitment; or were already attending exercise classes of equivalent intensity to the study intervention.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Number of participants in each experimental group.

Descriptor variables

Medical conditions were ascertained by a medical examination and self-report. Medications were recorded directly from the labels of participant’s medication containers. Education was assessed as number of years in secondary (high) school plus number of years in university or college if applicable.

General functional status was assessed by the 6 Minute Walk Test (6-MWT).21 Participants walked as quickly as they could on a pre-determined level course for 6 minutes. The total distance walked was accurately measured in metres with a measuring wheel.

We used a strain gauge to measure maximum isometric strength of three muscle groups in both legs. Knee extension and knee flexion strength were measured in the seated position with the angles of the hip and knee set at 90°. Ankle dorsiflexion strength was assessed with the angle of the knee set at 110°. Isometric strength was measured in kilograms, and the composite isometric strength score was calculated by summing the values from both legs.

Hours of physical activity participation in the previous 7 days was determined from a structured interview that included specific questions on the frequency and duration of planned physical activity.

Global cognitive state was assessed using the MMSE.20 General intellectual ability was estimated using a short form of the Wechsler Adult Intelligence Scale-Revised (WAIS-R).22 The short form contains four of the WAIS-R subtests: similarities, arithmetic, picture completion, and digit symbol. The individual subtest age-scale sores were combined to form the estimated IQ (EIQ). A description of the four subtests follows.

Cognitive performance

Fluid intelligence

The similarities, arithmetic, picture completion, and digit symbol subtest from the WAIS-R comprised the tests for the fluid intelligence domain.

The similarities subtest assesses verbal concept formation. It requires the participant to explain commonalities between word pairs. The arithmetic subtest is a test of oral arithmetic and has a working memory component that becomes more important with increasing age.23

The picture completion subtest involves visuoperceptual organisation and reasoning abilities, and requires the participant to identify what is missing from small incomplete pictures. People who have difficulty verbalising a response may indicate an answer by pointing.

The digit symbol subtest requires the copying within a 90-second time limit of symbols that are paired with numbers; it is a test of psychomotor performance.23

Executive functioning

In addition to assessing fluid intelligence, three other instruments were used to assess specific aspects of executive functioning: the Trail Making Test (part B) (TMT-B), the Stroop Neuropsychological Screening Colour–Word Test (Stroop-CW)and the Controlled Oral Word Association Test (COWAT). We used the TMT-B23 to assess visual conceptual and visuomotor tracking. This part of the test requires the participant to connect a series of circles alternating between letters and numbers (eg, 1-A-2-B-3-C). An attenuated form of part B was used because many seniors found this test daunting in our pilot testing. Time in seconds to complete the test was used for statistical analysis.

The Stroop-CW23 assesses mental flexibility. Participants were shown a page with words printed in incongruent coloured inks (eg, the word “BLUE” printed in red ink) and were asked to name the ink colour in which the words were printed (while ignoring the word itself). This test is based on the findings that it takes longer to say the ink colour than to read itself. This has been interpreted as the result of response conflict or the failure of response inhibition. The time in seconds to read the 112 words was used for statistical analysis.

The COWAT24 examines generativity and requires the participant to produce as many words as possible that begin with a given letter of the alphabet (F, A, S). There is 1 minute allowed for each of the three letters. The score is the sum of all acceptable words produced in the three trials.

Visual, verbal and working memory

We examined three different aspects of memory: visual, verbal and working. The Visual paired associates subtest of the Wechsler Memory Scale-Revised (WMS-R)25 requires the participant to learn the colour associated with each of six abstract line drawings, and is a measure of visual learning and memory. The verbal paired associates subtest of the WMS-R measures verbal memory and requires the participant to learn eight word pairs, four of which reflect easy associations and four of which are more difficult. The WAIS-R digit span subtest22 requires the participant to repeat a series of orally presented digits, either in the order given or in exact reverse order. Digit span forward primarily involves rote learning and sequential processing, whereas digit span backward requires transformation of the stimulus before responding, and is regarded as a measure of working memory. The WAIS-R arithmetic subtest was also included when assessing exercise effects on the working memory domain (and on fluid intelligence) in the statistical analyses.

Mood

Assessments of mood were made using the Geriatric Depression Scale (GDS) and the Positive and Negative Affect Schedule (PANAS). The GDS2627 was designed specifically for assessing depressed mood in older people. It has been found to be highly internally consistent and stable over a 1-month interval.28 The PANAS29 is a questionnaire that takes ⩽5 minutes to complete. The scale contains 10 negative and 10 positive items, which describe different feelings and emotions. The participant is required to indicate the extent to which these feelings have been experienced during the past few weeks. The measures used for statistical analysis were the total number scored on each scale and were labelled PANAS-N (negative) and PANAS-P (positive).

Experimental groups

Participants were allocated to groups according to their place of residence as dictated by the cluster randomisation. Of the 154 participants, 82 were randomised to the GE group, 34 to the FR group and 38 to the NEC group (fig 1).

The proportions of men in the three groups were similar: GE (13.4%), FR (8.4%) and NEC (13.2%). Participants of the intervention groups (GE and FR) were required to participate in 1-hour exercise classes twice a week for a total period of six months16 (a requirement of 52 classes in total). Classes were generally offered twice a week, with a third class offered occasionally to enable participants with conflicting schedules to maximise their participation in the exercise programme. The classes were held in a common room at each village and were led by experienced instructors trained to provide the same programme at each site. The exercise instructors were not blinded and different exercise instructors were allocated to the GE and FR programmes. The NEC group did not take part in any group activity.

Group-based exercise programme

The GE classes comprised a 5–15 minute warm-up period, a 40 minute conditioning period and a 10 minute cool-down (relaxation) period. The warm-up initially included slow paced walking, which was increased to a moderate speed in the first 10 weeks of the programme. The duration of the walking warm-up also increased from 5 to 15 minutes in the initial 10 weeks, and was maintained at 15 minutes for the remainder of the programme.

The GE conditioning period contained specific resistance training exercises, balance (both static and dynamic) training exercises, and activities for challenging hand–eye and foot–eye co-ordination and flexibility.16 Dynamic balance training exercises included dancing patterns, directional walking changes and speeds, and complicated routines of whole-body movements. There were also walking pattern exercises consisted of large strides, heel–toe walking, narrow-based and wide-based walking, and sidestepping.

In the GE cool-down period, a variety of techniques was used including muscle relaxation, controlled breathing and guided imagery.

Flexibility and relaxation programme

The FR group was designed to assess the effects of a group activity involving minimal-intensity exercise but which did not seem to be a sham activity to the participants. Participants remained mostly seated during the FR classes. Each FR session consisted of gentle bending and rotation of the joints, trunk and neck and controlled rhythmical breathing.16

No-exercise control group

Participants of the NEC group did not take part in any group activity. An NEC group was necessary to control for the possible benefits of social interaction on cognitive function in seniors.

Statistical analysis

For statistical analysis, χ² tests for cross-tabulation tables and one-way ANOVAs with Scheffé post-hoc comparisons were used to compare the prevalence of health and lifestyle measures, baseline cognitive performance, baseline 6-MWT and composite isometric strength of the legs of patients in the GE, FR and NEC groups.

Separate repeated-measures multiple analyses of variance (MANOVA) models assessed whether exercise had beneficial effects on the cognitive performance and mood domains: fluid intelligence, memory, working memory, executive functioning, and mood. The cognitive performance and mood variables were treated as the within-participants factors and group allocation was treated as the between-participants factor. Group by time effects were assessed for statistical significance. Univariate analyses are also presented for all variables regardless of whether the overall MANOVA group by time effect was significant to show the individual variable probability values and how they would have contributed to the multivariate association. These complementary analyses were performed using forced entry multiple linear regression analysis with robust standard errors to adjust for the clusters (ie, the fact that the unit of randomisation was the village and not the individual and the likelihood that there would be more similarities between residents within a village (cluster) would be more similar than between residents in different villages). In these models, baseline scores, age and experimental group were included as independent variables. All between-group comparisons were made on untransformed variables (as the distributions were near normal without significant skewing) using an intention-to-treat basis to minimise re-test bias and provide a more realistic indication of the generalisability and effectiveness of the intervention.30

Paired t test and Pearson correlation analyses were performed within each group to assess associations between baseline mood measure scores and degree of improvement at the 6-month re-test. Pearson correlation coefficients were also calculated to assess whether any significant changes in the cognitive measures were associated with changes in mood over the trial within the GE group. The data were analysed using STATA V.7 and SPSS for Windows statistical software.

Results

Baseline descriptors, cognitive performance and mood

The three groups did not differ in education levels, global cognitive state or general intellectual ability (table 1).

For most measures of cognitive performance and mood, the baseline scores were similar for the three groups. However, the digit symbol, visual paired associates, GDS and PANAS-N scores were significantly different. In most cases, the FR group performed worse than the GE group (table 1).

The proportions of participants reporting common medical conditions and medication use were similar across all groups, with the exception of poor hearing, which was more common in the GE group than in the other two experimental groups (table 2; p = 0.01). The mean number of hours spent in planned walks or exercise activities at baseline for the GE, FR and NEC groups was similar: 2.7 (SD 2.9) hours/week, 2.4 (2.2) hours/week and 2.6 (3.1) hours/week respectively, (p = 0.85).

Class attendance and participant availability for re-test

Of the 154 participants, 126 were available for re-test. Participants who withdrew from the study performed significantly worse on several of the baseline cognitive tests.

Of the 82 GE participants, 66 (80%) attended the classes and were available for the 6-month re-test. Of the remaining 16 participants, 1 had died, 2 were away at the time of the re-test, 4 were ill and 9 withdrew. The median number of classes attended for those who participated in the GE programme was 30 (interquartile range (IQR) 18 to 38). The range was 3 to 51 classes with 45 participants (68%) attending ⩾25 classes.

Of the 34 FR participants, 26 (76%) attended the classes and were available for re-test. Of the remaining eight participants, one was away at the time of the re-test, two were ill and five withdrew. The mean number of classes for the FR participants was 21.5 (IQR 4.75 to 32.25). The range was 2 to 62 classes (one participant attended an additional 10 classes) with 12 participants (46%) attending ⩾25 classes. There was no difference in attendance between the GE and FR groups (Mann–Whitney U test 661.5, p = 0.09).

Of the 38 NEC participants, 34 were available for re-test (89%). Of the remaining four participants, one had died, two were away at the time of re-test and one withdrew from the study. None of the NEC participants who returned for the re-test had started a structured exercise programme during the trial period.

Cognitive performance and mood

The MANOVA analyses found significant group by time effects for fluid intelligence variables (F_8,242 = 3.49, p<0.001) that indicated improvements in the test measures in the GE group, but little or no change in the FR and NEC groups. MANOVA for the remaining dependent variables found no significant group by time effects: memory (F_8,234 = 1.21, p = 0.30), working memory (F_4,244 = 1.81, p = 0.13), executive functioning (F_6,224 = 0.05, p = 0.53) and mood (F_6,238 = 0.75, p = 0.61). Means scores at baseline and at the 6 month re-test and p values for univariate analyses adjusting for baseline scores baseline age and participant cluster e-assessed at 6 months are shown in table 3.

Significant improvements in the PANAS-P scale occurred in both activity groups (GE: paired t₆₄ = 2.06, p<0.05; FR: paired t₂₃ = 2.37, p<0.05), but not within the NEC group (paired t₃₃ = 0.20, p = 0.84). Significant associations were also foundbetween initial scores on the Geriatric Depression Scale and degree of improvement on that scale at the end of the trial for both activity groups (r = 0.45, p<0.01 and r = 0.49, p<0.01 for the GE and FR, respectively). In contrast, no such association was evident in the NEC group (r = 0.09, p = 0.63).

Correlates of improved fluid intelligence within GE

Within the GE group, improvements in the similarities and arithmetic tests (the two fluid intelligence subtests in which significant improvements were evident) were not significantly associated with changes in the measures of mood (table 4).

Discussion

We found that a group-based exercise programme that significantly reduced falls and maintained physical function in seniors16 also significantly improved cognitive performance of fluid intelligence compared with FR or NEC. These novel results suggest that participation in combined (ie, both resistance training and balance training exercises) group-based exercise programme can modify age-related declines in certain domains of cognitive function, even in seniors living in intermediate care and self-care retirement facilities. To our knowledge, this study is the first to show that a general group-exercise programme, not specifically designed to improve cardiovascular fitness (ie, training at a percentage of targeted heart rate), can significantly benefit cognitive performance, specifically fluid intelligence, in seniors without dementia. Although our general group-based exercise programme did not significantly improve cognitive performance in executive functioning, our results still extend the observation that exercise training has specific benefits for executive functioning among high-functioning community-dwelling seniors58 as recent evidence suggests construct similarity in executive functioning and fluid intelligence.32

In recent years, there has been strong interest in physical activity as a primary behavioural prevention strategy against cognitive decline. Physical activity has been widely promoted as a strategy for healthy ageing as it can reduce the incidence of cancer, diabetes and heart disease.33 More recently, accumulating evidence suggests that the benefits of physical activity extend beyond the periphery to the central nervous system.34 To date, most prospective, intervention studies of exercise and cognition have focused on aerobic training both in animals34353637 and humans.81038 Findings from both animal and human studies of aerobic training suggest that this type of physical activity may promote cognitive health through the enhancement of both brain structure and function.81035363738 However, it has been suggested that other types of exercise training, such as resistance training, may also benefit cognition.5

Resistance training may benefit cognition through mechanisms involving insulin-like growth factor (IGF)-1 and homocysteine. Studies have reported that resistance training increased levels of IGF-11339 and reduced serum homocysteine.40 IGF-1 promotes neuronal growth, survival and differentiation, and improves cognitive performance.41 Increased homocysteine levels are associated with impaired cognitive performance,42 AD,43 and cerebral white matter lesions.44 In a 2-year prospective study, elevated homocysteine and increases in homocysteine were associated with decreased neuropsychological functioning in otherwise cognitively intact seniors.45

Results from animal studies suggest that balance or agility training exercises also benefit cognition. For example, Black et al37 found that rats exposed to non-aerobic motor skills (eg, obstacle courses) had a larger number of synapses in the cerebellum than did rats exposed to extensive physical exercise (ie, running on activity wheel) and rats that were inactive. The results of our study support the hypothesis that other types of exercise training, such as resistance and balance training exercises, may benefit cognitive performance, even in seniors.

Although our general group-based exercise programme significantly improved cognitive performance of fluid intelligence, it had no significant benefits onmemory or executive functioning. A possible contributing factor is the relatively short duration of our exercise training, given that improvements in memory were seen by Williams and Lord15 in their 12-month exercise study of community-dwelling women. A second possible factor is the population studied; although Colcombe et al8 saw significant improvement in cognitive performance of exercise functioning after 6 months of aerobic training, their study cohort were high-functioning community-dwelling seniors aged 58–77 years. Compared with their healthier counterparts, our study participants may have a lower capacity for improvement. Finally, we may have not detected a benefit of the GE programme on executive functioning, despite demonstrating a significant beneficial effect on fluid intelligence, because of differences in the characteristics of the neuropsychological tests used to assess performance in these two domains of cognition. All three neuropsychological tests of executive functioning were time-dependent, although this was not true for those used to assess fluid intelligence. Thus, processing speed may have been a significant factor in determining cognitive performance of executive functioning but not of fluid intelligence, as defined in our study.

Our results suggest that improvement in fluid intelligence in the GE group was not mediated by improved mood. We note that most participants scored 0 (no depression) to 5 on the scale at baseline and only five participants (3.2%) scored within the clinical range of depression. Thus, it was likely difficult to detect a meaningful change in this outcome measure. However, there were significant improvements in the PANAS-P scale within both the GE and FR groups. In addition, baseline GDS scores were significantly associated with a degree of improvement at re-test in these two groups. These results suggest that increased peer socialisation may be helpful in improving psychological well-being in seniors. Previous studies have found that group-based exercise may improve mood.1415

We acknowledge that the study has certain limitations. The cluster-randomised design had some advantages in that randomising sites rather than residents within each village prevented possible confounding effects such as controls wishing to join the exercises.46 However, the clustering and the selection of a subgroup for this study resulted in some baseline differences among the three experimental groups; the FR group had significantly worse cognitive performance and mood. The cluster randomisation also made blinding of research staff impractical. Therefore, notwithstanding the quantitative test measures and standardised participant instructions, it is possible that the improvement seen in fluid intelligence at re-test may have resulted from assessor bias.

In conclusion, a general group-based exercise programme composed of resistance training and balance training exercises significantly improved cognitive performance of fluid intelligence in seniors residing in retirement villages compared with a group-based programme of flexibility exercises and with a no-exercise control group. Furthermore, both group-based exercise programmes (ie, GE and FR) was beneficial for certain aspects of mood within the 6-month intervention period.