Prognosis of early cognitive deficits in PD
It cannot be assumed that cognitive dysfunction in early PD is in all cases a prodrome of PD-D. While some early deficits may progress inexorably, and thus represent the early stages of the dementing process, other deficits may have an alternative aetiology and more favourable prognosis. Until recently, longitudinal studies measuring cognitive function in PD have reported rather disparate findings, with a range of neuropsychological deficits including executive deficits [19-21], impaired verbal fluency [21, 22], visuospatial problems , and memory and language dysfunction [20, 23] reported to predict the development of PD-D. The findings of these studies, however, have been limited by two main factors. First, the nature of the cohorts was suboptimal: they included patients at varying disease stages and were mostly hospital-based, and thus not necessarily representative of PD in the general population. Secondly, their findings were highly dependent on the particular selection of tests employed in each study. The lack of a standardized approach to neuropsychological testing inevitably introduces bias and makes any comparison between studies difficult.
The first of these problems has been addressed in recent years through the establishment of a number of prospective studies in population-based cohorts of newly diagnosed patients, enabling a better description of the pattern and temporal evolution of cognitive dysfunction in PD. To date, most of these studies remain in their early stages with limited longitudinal data [24-27]. However, the CamPaIGN study , the first and longest running of these incident population-based cohort studies, has recently reported its 10-year follow-up data . This study has followed up a population-representative cohort of 142 patients with idiopathic PD in the county of Cambridgeshire, UK, with detailed clinical and neuropsychological assessments at approximately 2-yearly intervals. Analysis at multiple time points has shown that the two most significant early neuropsychological predictors of later dementia, independent of age and other potential confounding factors, are poor performance on tests of semantic fluency and pentagon copying [28-30]. A semantic fluency score of less than 20 words in 90 seconds at diagnosis was associated with a hazard ratio (HR) of 3.1 (p =
0.005) for dementia at 10 years (Cox survival analysis), while impaired pentagon copying at diagnosis was associated with a HR for dementia of 2.6 (p = 0.001). Phonemic fluency, however, was not associated with increased dementia risk, and this dissociation between semantic and phonemic fluency implicates the temporal lobe-based semantic system in early dementia, rather than the frontal lobe- based search and retrieval system which is common to fluency tasks in general. Pentagon copying is a mixture of visuospatial and constructional ability, widely accepted to depend on the integrity of the parietal lobe . There was no association between ‘frontostriatally based’ executive tasks and later dementia, and in fact there was no significant deterioration in performance on executive tasks over time. On the basis of these findings, as well as the finding of dissociable relationships between certain genetic variants and different cognitive functions as discussed later, we have previously suggested that cognitive impairments in early PD can be segregated into two main types. Executive deficits, primarily due to dysfunction in dopaminergic frontostriatal networks, are likely to fluctuate with disease course and medication, and are not clearly associated with dementia. In contrast, more posterior cortically based deficits represent the early stages of a dementing process which is proposed to be non-dopaminergic and likely to be related to cortical Lewy body deposition, subcortical cholinergic cell loss, and possibly to Alzheimer’s type changes .
Other authors have provided further support for the idea that there are dissociable cognitive syndromes in PD. Pagonabarraga et al.  used the PD-Cognitive Rating Scale, which characterizes ‘subcortical’ and ‘cortical’ cognitive deficits, in subgroups of PD patients with no cognitive impairment, MCI, and dementia. They found evidence for progressive frontal-subcortical impairment as the disease evolves, but the transition from MCI to dementia was marked by the addition of deficits with a cortical basis. Muslimovic et al.  conducted a large meta-analysis, including 901 PD patients without dementia, to evaluate changes in multiple cognitive domains over time. Over a mean follow-up of 29 months, there was a statistically significant decline in global cognitive ability, visuospatial function, and memory, but not in executive function, in keeping with the idea that posterior cortical deficits represent a progressive dementing process while executive impairments do not.
The second problem affecting the majority of longitudinal studies of cognition in PD is the lack of a standardized approach to neuropsychological testing and defining cognitive impairment. The MDS diagnostic criteria for PD-MCI were proposed to alleviate this problem . However, their prognostic utility remains to be determined. In one of only two longitudinal studies so far using the PD-MCI criteria, Pedersen et al.  studied a population-based cohort of 182 patients with incident PD and reported that of the 20.3% meeting PD-MCI criteria at baseline, 27% had developed dementia by 3-year follow-up, compared with only 0.7% of patients without MCI. There were similar rates of conversion to dementia among those meeting PD-MCI criteria at 1 year from baseline. However, a significant proportion with PD-MCI reverted to normal cognition at 3-year follow-up (21.6% of those with PD-MCI at baseline and 19.4% of those with PD-MCI at 1-year follow-up). In a second study, Broeders et al.  studied a consecutive clinic-based sample of 123 patients with newly diagnosed PD, following up 73 of these to 5 years. With the caveat that there was significant attrition of more cognitively impaired patients, they found that of the 35% diagnosed with PD-MCI at baseline 26% had developed dementia at 5 years. While the majority of those developing dementia at 5 years had had a previous diagnosis of PD-MCI, 18% had been cognitively normal at the previous assessment. In keeping with the findings of Pedersen et al., some individuals with PD-MCI subsequently reverted to normal cognition, although the numbers were lower (9% between baseline and year 1, and 6% between baseline and year 5), possibly due to their more detailed neuropsychological assessment with higher specificity.
While the numbers in these studies have not been sufficient to formally calculate positive and negative predictive values, these data support the idea that PD-MCI as a single entity is not a useful construct for predicting later dementia. It is critical that future studies explore subtypes of cognitive impairment to establish which domains are the most informative in terms of predicting dementia. The PD-MCI criteria do allow for such subtyping (level II criteria), but only if a full neuropsychological battery is used with at least two tests in each of five cognitive domains . Level II PD-MCI criteria were used by Broeders et al. , but the numbers were too small to explore associations between MCI subtypes and progression to dementia. The most common subtype of MCI at baseline was ‘multidomain’ (65.1%), which is not unexpected given that the proposed cognitive syndromes in PD are not mutually exclusive. However, this may limit future attempts to examine the prognostic value of PD-MCI subgroups, and deficits on individual cognitive tests may remain the most useful prognostic variables.