Table 4.

GRADE summary of evidence table.

Nonpharmacological interventions for people with Parkinson's and cognitive impairment
Population: People with idiopathic Parkinson's disease and cognitive impairment
Setting: Any
Intervention: Nonpharmacological interventions (only cognitive interventions represented by meta-analysis)
Comparison: Control arm (varied comparators: Usual care; physical activity/rehabilitation; cognitive training)
Outcomes Immediately postintervention (up to 2 weeks post end of intervention)	Comparative risks: Standardised mean difference (95% Cis)	No. of participants (studies)	Quality of the evidence (Grade)	Comments
Health-related quality of life PDQ-39, PDQ-8, Qualidem	−0.20 [−0.57, 0.18]	235 (7∗)	⊕⃝⃝⃝ Very lowa,b,c,d	Negative values indicate improved health-related quality of life (signs reversed for Qualidem to match direction of effect).
Function/Activities of daily Living Bayer ADLs, BAFQ, UPDRS Part II, pill questionnaire	0.14 [−0.23, 0.51]	200 (5∗)	⊕⃝⃝⃝ Very lowa,b,c,d	Positive values indicate worsened function.
Wellbeing Relatives stress scale, brief resilience scale, generalised efficacy scale	See comment	105 (2)	⊕⃝⃝⃝ Very low	Outcomes address different aspects of wellbeing rather than a unified construct and so not considered suitable for meta-analysis.

⁣^∗One study is divided into two independent intervention arms, with the control group halved for inclusion in the meta-analysis.

^aDowngraded by one point due to the risk of bias of included studies: risk about randomisation in two trials, risk from crossover (carryover) in one, risk about outcome measurement due to lack of blinding (expected due to nature of interventions) in 5 of the 7 trials included in HrQoL analysis, and 2 of the 4 for function analysis.

^bDowngraded by one point due to inconsistency: moderate heterogeneity. Some can be accounted for by subgroup analysis by intervention type, comparator type and cognitive diagnosis of participants.

^cDowngraded by one point due to imprecision: confidence intervals cross two possible interpretations (benefit and harm).

^dDowngraded by one point due to detection of publication bias on funnel plot where possible and extrapolated when too few studies for funnel plot.