Table 5 Outcomes and risk of bias of clinical trials on human Parkinson’s disease

Studies that specifically measured executive function (n = 4)

McKee et al. 2013 [29]

• Tango improved disease severity (UPDRS-III) and spatial cognition/mental imagery (Brooks Spatial Task) more than education group, maintained gains 10–12 weeks post-intervention

• Detection bias: study was underpowered (n = 23 tango, n = 8 education) to evaluate some main effects within groups, so main effect of time was evaluated

Cruise et al. 2011 [26]

• Exercise improved verbal fluency and spatial working memory on Cambridge Neuropsychological Test Automated Battery

• Selection bias: the control group received usual care, no control for the effect of social interaction with exercise

• Exercise was of “possible benefit” on semantic fluency and mood

• Information bias: the variable intensity level of the intervention could have affected outcomes

• Exercise did not benefit spatial or pattern recognition, quality of life, had no negative impact

Ridgel et al. 2011 [28]

• Time to complete Trail Making Test A & B (tests executive function) decreased after passive cycling

• Selection bias: no control

• Information bias: the same test pattern was used pre- and post-intervention, although practice effects were attempted to be controlled through pre-test training with the task

• Performance improved on Trail Making Test B following passive cycling

Tanaka et al. 2009 [27]

• Exercise improved executive function for “Categories Completed” (i.e., capacity for abstraction) and “Preservative Errors” (i.e., mental flexibility) on the Wisconsin Card Sorting Task

• Selection bias: small sample size, no long-term follow-up, not purely randomized

•Information bias: no mention of medication administration; only one participant in the group had a heart rate monitor, so the intensity was targeted towards the group and not the individual

• No interactions for confounding variables: concentrated attention, trait or state anxiety, depression

Studies that measured unspecified aspects of cognition (n = 4)

Dos Santos Mendes et al. 2012 [25]

• PD showed no deficit in learning or retention on 7/10 games, deficits related to cognitive demands of tasks

• Selection bias: the baseline physical fitness of the subjects was not compared

• PD had worse performance than healthy individuals on 5 tests

• Performance bias: no PD controls not performing intervention, no control for enjoyment or motivation

• PD could transfer learning to an untrained motor task at follow-up

Pompeu et al. 2012 [24]

• Both groups improved UPDRS-II, MoCA and balance, no additional advantage from Wii Fit^TM

• Information bias: the baseline physical fitness of the subjects was not compared, so potential for differences between groups

• Improved scores on Wii Fit^TM games, maintained at follow-up

• No differences in outcomes between groups pre- to post-intervention or in follow-up

Müller et al. 2010 [23]

• Reaction time increased after rest and decreased after exercise, movement time decreased after exercise

• Selection bias: no PD control group, no healthy controls

• Information bias: one-day washout period (24 hours) may not have been long enough; pilot trial

• Number of correct answers decreased after rest

• Tapping rate increased after exercise

• Detection bias: unclear how reactivity was measured

• Peg insertion interval time decreased after exercise (complex movement sequences, visual and spatial cognition, sorting and planning)

Baatile et al. 2000 [22]

• Improved UPDRS score (only total score significant)

• Selection bias: limited sample size, no control group; pilot trial

• Improved PDQ-39 score, most improved in cognition component

• Information bias: exercise intensity not standardized