Description of studies

The search strategy rendered 31 promising abstracts. The full text of five theses could not be traced in peer reviewed journals (see list of studies awaiting assessment). Twenty‐eight studies were excluded for various reasons after screening of the full text articles (see characteristics of excluded studies). The following 11 studies were all suitable for inclusion.

Bakken et al. (Bakken 2001) published a RCT in which 15 participants aged 72 to 91 years were randomly assigned to an aerobic exercise group (n = 8) or a wait list control group (n = 7) for eight weeks. The aerobic exercise group trained in one hour sessions for three sessions per week. Duration and intensity increased over the weeks but the subjects' heart rates did not exceed the upper limit of their target heart rate reserve. The main cognitive outcome parameter was a "tracking test" (visual attention). Five participants were available for analysis in each group. Rate‐pressure product (RPP) was used to determine if an aerobic training effect had taken place. Both groups showed slight increases in RPP from pre‐test to post‐test.

Blumenthal et al. (Blumenthal 1989 a; Blumenthal 1989 b) performed a RCT in which 101 men and women aged 60 to 83 years were randomly assigned to either an aerobic exercise training (n = 33), a yoga/flexibility programme (n = 34) or a waiting list control group (n = 34). The aerobic exercise programme consisted of three supervised sessions for 16 weeks. The aerobic training was based on 70% of maximum heart rate achieved on the baseline exercise test. The yoga/flexibility group trained two times a week for 16 weeks and the control group was instructed not to change their physical activity habits and especially not to engage in any aerobic exercise for the study period. Two subjects were lost from the aerobic group and two were lost from the waiting list group. Blumenthal et al. included tests for cognitive speed, verbal, visual and working memory as well as executive functions, cognitive inhibition, visual and auditory attention and motor function in this study. A summary combination of the scores on both 2&7 test (letters and digits) was calculated and standard deviations were pooled and used in analysis. Subjects in the aerobic training group experienced a significant 11.6% increase in their VO₂ max (from 19.4 to 21.4 ml/kg/min), whereas the participants in the yoga/flexibility and control groups experienced a 1 to 2% decrease in VO₂ max (from 18.8 to 18.7 ml/kg/min and 18.5 to 17.9 ml/kg/min, respectively). Blumenthal and colleagues described the results for men and women separately in their article. In this review the data presented in Blumenthal 1989 a reflects the data of the women, Blumenthal 1989 b those of the men.

In the RCT of Emery and Gatz (Emery 1990 a) a total of 48 men and women aged 61 to 86 years were randomly assigned to an exercise programme (n = 15), a social activity group (n = 15) or a control group (n = 18). The exercise programme consisted of stretching exercises, aerobic exercise (at 70% of age‐adjusted maximum heart rate = 220‐age) as well as rhythmic muscle strengthening exercises (e.g. repeatedly standing up and sitting down), three times a week for 12 weeks. Subjects in the social programme participated in non‐physical activities (card games, art projects, political discussion groups, watching films) three times a week for 12 weeks. Because attrition from the social group was comparable to that of the control group, and because attendance for the social group was poor overall (ranged from 10 to 94%), the analyses were conducted with the two groups pooled. One subject was lost from analyses in the exercise group, four from the social group and four from the control group. This study included tests for cognitive speed and auditory attention in its design. Resting heart rate, maximum heart rate and systolic/diastolic blood pressure indicated no significant differences between the groups. Both groups showed a significant time main effect decrease in diastolic blood pressure, other measures indicated no significant effects.

Emery et al. (Emery 1998) published a second RCT of 79 COPD patients aged 66.6 ± 6.5 years. Twenty nine participants were randomly assigned to an exercise/education/stress management group (EXESM), 25 subjects to an education/stress management group (ESM) and 25 participants were assigned to a control group. EXESM participants trained during a 10‐week period. The training included aerobic exercises as well as strength training on Nautilus equipment, educational lectures on topics relevant for COPD patients and a weekly group meeting for stress management and psychological support. In the first five weeks, EXESM participants followed a programme of daily sessions with 45 min of aerobic exercise, lectures on relevant COPD topics and a stress management meeting. Following the initial five weeks, EXESM participants then participated in five more weeks of exercise for three times a week, 60 to 90 min and one hour a week of stress management. ESM consisted of 16 educational sessions and 10 stress management classes in 10 weeks. Controls were instructed not to alter their physical activities during the study. Four subjects were lost from the EXESM group and two subjects were lost from the ESM group. Tests on cognitive speed, visual attention, executive and motor functions were incorporated in this study. Participants in the EXESM condition achieved a significant 16% gain in VO₂ max (from 12.4 to 14.4 ml/kg/min), whereas participants in the other two groups showed no significant changes in VO₂ max (ESM from 12.0 to 11.7 ml/kg/min and controls from 11.1 to 10.8 ml/kg/min).

The RCT of Fabre et al. (Fabre 2002) presented data from 32 men and women aged 60 to 76 years. Participants were randomly assigned to an aerobic exercise programme (n = 8), a mental training programme (n = 8), a combined aerobic/mental programme (n = 8) or a control group (n = 8); no one was lost to follow‐up. Physical training consisted of two supervised exercise sessions per week for two months: walking and running to maintain target heart rate (target heart rate corresponded to the ventilatory threshold). Subjects in the mental training programme met once a week for two months and Israel's method (Israel 1987) was used. The controls met as many times as the other groups for leisure activities such as singing. Fabre and colleagues included tests for verbal and visual memory, perception and executive functions in their study. The physical training resulted in an average significant increase in VO₂ max of 12% (from 1350 to 1630 ml/min) and 11% (from 1510 to 1625 ml/min) in the aerobic training group and the combined aerobic/mental group, respectively. The VO₂ max scores of the participants in the other two groups were unchanged compared to initial values (mental training group from 1060 to 999 ml/min and controls from 1256 to 1265 ml/min).

Hassmén and Koivula (Hassmén 1997 a; Hassmén 1997 b) presented an RCT in which 40 men and women aged 55 to 75 years were randomly assigned to an exercise group (n = 20) or a control group (n = 20), no one was lost to follow‐up. Participants in the exercise group walked three times a week during three months with intensities according to "Rating of Percieved Exertion" (RPE, Borg scale) of 9 (very light), 11 (fairly light) and 13 (somewhat hard). The performed exercise gave rise to heart rate responses of 95 to 125 beats/min. Subjects in the control group were given home assignments to be performed three times per week. Cognitive speed, verbal memory, perception and auditory attention were tested in the participants. Heart rate data and data on RPE showed no significant differences between the groups over time. The results for men and women were described separately in this article. In this review the data presented in Hassmén 1997 a reflects the data of the women, Hassmén 1997 b those of the men.

A total of 174 participants, aged 60 to 75 years, were included in the RCT of Kramer et al. (Kramer 2001). After randomisation and testing, 124 individuals remained in analyses; 58 (13 men) in the aerobic walking group and 66 (20 men) in the stretching and toning programme. Participants in the aerobic programme walked briskly in three supervised sessions per week for six months. After warming up they spent 40 minutes on brisk walking (gradually beginning at 10 to 15 minutes up to 40 minutes) and finished with a cooling down. Initial exercise was performed at 50 to 55% of VO₂ max and increased to 65 to 70% of VO₂ max. The subjects in the stretching and toning programme met three times a week for six months. The programme emphasized stretches for all the large muscle groups of the upper and lower extremities. Each stretch was held for 20 to 30 seconds and repeated 5 to 10 times. Each session was preceded and followed by 10 minutes of warm‐up and cooling down. Cognitive speed, verbal and visual memory, perception, executive and motor functions as well as cognitive inhibition, visual and auditory attention were assessed with various cognitive tests. Mean results of the subtests of the pursuit rotor task, Rey's auditory verbal learning test, spatial attention and visual search task were summed and divided by the number of tasks. Standard deviations of these subtests were pooled. The physical training resulted in improvements of 5.1% on VO₂ max measures (from 21.5 to 22.6 ml/kg/min). The toning group showed a 2.8% decrease in VO₂ max scores (from 21.8 to 21.2 ml/kg/min).

Madden et al (Madden 1989) published an RCT of 85 men and women 60 to 83 years of age. All participants were randomly assigned to either an aerobic exercise group (n = 28), a yoga group (n = 30) or a control group (n = 27). The subjects in the aerobic exercise group trained three times per week for 16 weeks and all exercise was performed in target (training) heart range (70% of maximum during initial exercise test). Subjects in the yoga group participated in supervised sessions twice a week during 16 weeks. The controls were instructed not to change their physical activity habits for the length of the study. A total of six subjects were lost from analyses; three from the aerobic exercise group, two from the yoga group and one from the control condition. All mean error rates for the word comparison task (executive functions) and the letter search task (visual attention) were recalculated into summary scores. Standard deviations of these two tasks were pooled. Aerobic capacity remained constant for the yoga and control groups between pre‐ and post‐test (respectively from 18.8 to 18.6 ml/kg/min and from 19.1 to 18.6 ml/kg/min), whereas the aerobic exercise group showed a significant 11% increase in VO₂ max (from 19.7 to 21.9 ml/kg/min).

Moul et al. (Moul 1995) presented data of an RCT in which 30 men and women aged 65 to 72 years of age were randomly assigned to a walking condition (n = 10), weight training (n = 10) or control condition (n = 10) for 16 weeks. No one was lost to follow‐up. Participants in the walking condition walked five times per week for 30 minutes at 60% of "Heart Rate Reserve" (HRR, as determined by treadmill testing). Walking duration was increased two minutes per week until they reached 40 minutes and HRR values were adjusted after eight weeks of training to 65% of HRR. Weight training consisted of five sessions per week of upper and lower body exercises on alternate days of the week. Abdominal crunches and back extension were performed each session. The weight training group employed a "Daily Adjusted Progressive Resistive Exercise Program" (DAPRE) using weights. Subjects in the control condition met five times per week for mild stretching exercises which challenged the cardiovascular or muscular systems minimally. All participants were kept in analyses. The Ross Information Processing Assessment was used to evaluate changes in cognitive function. Post‐test data revealed that the subjects in the walking condition significantly increased their VO₂ max by an average of 16% (from 22.4 to 26.6 ml/kg/min), whereas there were no significant changes in VO₂ max for the other two groups (weight training group from 21.4 to 20.4 ml/kg/min and controls from 20.9 to 19.3 ml/kg/min).

The RCT of Panton et al. (Panton 1990) included data on 49 retired university professionals (23 males and 26 females), aged 70 to 79 years. The participants were randomly assigned to a walk/jog group (n =17), a strength group (n = 20) or a control condition (n = 12) for 26 weeks. A total of eight participants were lost to follow‐up; it is unclear from which group these subjects were lost. Subjects in the walk/jog condition participated in three exercise sessions per week for the duration of the study. Initially, participants started walking/jogging for 20 minutes at 50% of their maximal heart rate reserve (HRRmax). The duration was increased by 5 min every two weeks until the subjects walked for 40 minutes. Training intensity was gradually increased until subjects could walk at 60 to 70% of their HRRmax. During the 14th week of training, exercise intensity was further increased by alternating fast walk/moderate walk or fast walk/slow jog intervals. Five participants increased their training intensity by increasing the slope of the treadmill. By the 26th week of training, all subjects performed at 85% of HRRmax for 35 to 45 min. Participants in the strength group had workout sessions for three times a week for 30 min per session. The workouts consisted of one set of 10 variable resistance Nautilus exercises (leg, arm and torso muscles). During the first 13 weeks, subjects used light to moderate weights and performed 8 to 12 repetitions for each exercise. During the last 13 weeks, resistance was increased substantially and subjects were encouraged to train to volitional muscular fatigue. When subjects could complete 12 or more repetitions, the resistance was increased. The controls were asked not to change their lifestyle over the six month duration of the study. Tests for cognitive speed were performed to analyse cognitive function. Aerobic capacity significantly improved by 20.4% (from 22.5 to 27.1 ml/kg/min) for subjects in the walk/jog group; participants in both the strength as well as the control group showed no significant changes in VO₂ max (from 22.5 to 23.3 and 22.2 to 22.0, respectively).

Whitehurst (Whitehurst 1991) published an RCT on 14 females aged 61 to 73 years. The subjects were randomly assigned to an exercise programme (n = 7) or a control condition (n = 7) for eight weeks. No one was lost to follow‐up. The exercise condition consisted of three supervised sessions per week. The subjects cycled for 8 to 10 minutes the first week to provide acclimation. Thereafter, 3 to 5 minutes was added to subsequent sessions so that by week 4 all subjects were cycling for 35 to 40 minutes at their target heart rate (70 to 80% of predicted maximum by submaximal bicycle test). The controls were instructed not to engage in any form of vigorous physical activity during the course of the study. Choice reaction times were tested for evaluation of cognitive function. The subjects in the exercise group significantly increased their VO₂ max values by an average 16% (from 25.4 to 29.7 ml/kg/min), whereas the subjects in the control group increased their VO₂ max by a (nonsignificant) 2% (from 24.7 to 25.4 ml/kg/min).

Risk of bias in included studies

Results of the quality assessment can be found in the characteristics of included studies table and the additional table (Table 3).

The overall methodological quality score of the included studies ranged from 23 to 39 (minimum possible score of 14 points, maximum possible score of 48 points; lower scores denote a better methodological quality). In most of the trials the blinding procedures of the outcome assessors, treatment providers, and participants frequently scored "no, because blinding is not feasible".
Cohen's kappa (K) was calculated as a measure of inter‐observer reliability after the initial screening by two reviewers (MA and GA) and reached 0.81, almost perfect according to Landis 1977.

Effects of interventions

Analyses on the effects of aerobic exercise on cognitive function compared to any other intervention rendered significant positive effects of aerobic exercise on cognitive speed (SMD random effects 0.26, 95% CI 0.04 to 0.48, P = 0.02) and visual attention (SMD random effects 0.26, 95% CI 0.02 to 0.49, P = 0.03). Analyses on the effects of aerobic exercise on cognitive function compared to no intervention (controls or waiting list groups) yielded significant positive effects of aerobic exercise on auditory attention (WMD random effects 0.52, 95% CI 0.13 to 0.91, P < 0.01) and motor function (WMD random effects 1.17, 95% CI 0.19 to 2.15, P = 0.02).

However, in the comparisons of exercise versus any intervention, nine of the 11 cognitive functions yielded no significant effects. Nine of the 11 cognitive functions showed no significant effects for exercise versus no intervention.

Analyses on the effects of aerobic exercise on cognition compared to a strength programme rendered no significant differences. However, these results should be interpreted with care since they were based on one study (Moul 1995). Analyses on the comparisons of aerobic exercise versus a social/mental programme included three studies (Emery 1990 a; Emery 1998; Fabre 2002). These comparisons were are not shown separately (data of the three were already pooled in the analysis of effects of aerobic exercise versus any intervention).