People with Parkinson's disease psychosis (PDP) have reported cognitive and executive deficits. However, it is unclear whether these deficits are different depending on psychosis severity. Here, we aimed to compare cognitive performance between PDP patients with varying severity of their psychosis symptoms (such as delusions and hallucinations), relative to patients with Parkinson's disease without psychosis (PDnP), using a meta-analytical approach.

Searches were conducted on PubMed, MEDLINE, Web of Science, PsycINFO and SCOPUS. Standard mean differences between PDP and PDnP patients on cognitive and executive function tests were expressed as Hedges’ g effect sizes from eligible studies (K = 23). PDP patients were classified based on the severity of their psychosis symptoms into formed and unformed hallucinations, and hallucinations with/without insights. Separate multi-level meta-analyses were conducted for symptom severity groups of PDP patients, and for different cognitive domains due to studies contributing to multiple cognitive domains. Meta-regressions were conducted to examine the effect of age. Analyses were conducted in R (version 4.2.2).

Hedges’ g effect sizes were negative in all analyses, suggesting that PDP patients with formed hallucinations (n = 317) showed a significantly worse performance than PDnP patients (n = 734) across all domains (global cognition, g=−0.853, p < 0.001; language, g=−0.602, p < 0.001; episodic memory, g=−0.899, p < 0.001; executive functions, g=−0.543, p = 0.007; processing speed, g=−0.698, p < 0.001). There was presence of significant heterogeneity across all the analyses (all p < 0.05), except for language (p = 0.053) and processing speed (p = 0.077). There was also presence of significant publication bias (assessed with Egger's regression test) in all analyses (all p < 0.001), except for global cognition (p = 0.656). PDP patients with unformed hallucinations (n = 135) performed worse relative to PDnP patients on the same domains but these results were not significant. Publication bias was not significant for global cognition for results of unformed hallucinations (p = 0.4054). Age was shown to be a significant moderator for all domains except global cognition (global cognition, b=−0.02, p = 0.13; language, b=−0.042, p = 0.037; episodic memory, b=−0.055, p = 0.001; executive functions, b=−0.098, p < 0.001; processing speed, b=−0.048, p = 0.022). Relative to PDnP patients (n = 322), both PDP patients with/without insights had worse scores on global cognition tests (no insights, n = 37, g=−2.747, p = 0.021; insights, n = 83, g=−0.942, p = 0.019). Due to the low number of studies (k < 7), Egger's test was not applied.

Decreased cognitive performance may underlie presence of impairments in PDP patients. Formed hallucinations and lack of insights are associated with greater cognitive deficits. In addition, older age could result in worse cognitive scores.