Patient and Clinician Impressions of Cognitive Impairment in Parkinson’s Disease

INTRODUCTION

Cognitive decline is a common non-motor feature of Parkinson’s disease (PD). Nearly 30% of patients may develop dementia within 5 years of diagnosis [1], and over 80% within 20 years [2]. Cognitive decline in PD can develop anytime in disease course, and the rate and degree of progression is variable and difficult to predict. The spectrum of cognitive impairment in PD ranges from subjective cognitive decline (PD-SCD) to mild cognitive impairment (PD-MCI) to dementia (PDD). The implications of these diagnoses affect quality of life, caregiver burden, morbidity, and mortality [3, 4].

There is great need for predictors of cognitive decline in PD, some of which are MCI, age, and age at PD diagnosis [5]. SCD, or “cognitive complaint,” in Alzheimer’s disease has been shown to correlate with pathology and predict disease, but few studies have examined SCD in PD. In two small studies following PD patients for two-years, PD-SCD predicted future cognitive impairment [6, 7]. A longer study (mean 7.5 years), found that while the highest rate of conversion to dementia occurred in PD-MCI (50%), the PD-SCD group closely followed (33.3%) [8]. Consistent with these findings, Purri et al. recently found that over 5 years, SCD was associated with higher risk of cognitive impairment and poorer functional abilities [9].

Moreover, patient and clinician impressions of cognitive function are important since objective clinical evaluation is fraught with limitations. Screening measures like the Montreal Cognitive Assessment (MoCA) are not sensitive to early, mild changes or to change over time. Comprehensive neuropsychological testing, while useful for detecting deficiencies in specific cognitive domains and diagnosing cognitive impairment in patients with co-morbidities, can be limited by access to a trained neuropsychologist and can be time-consuming and stressful for the patient. Therefore, clinical impression is often used in practice to determine diagnoses of MCI or PDD. Clinicians may have difficulty reconciling subjective reports from patients and their care partners, however, since apathy or lack of insight may accompany PD and caregiver burden may impact impressions of cognitive status. Therefore, it is important to understand how patient and clinician impressions of cognitive impairment relate to one another and how these impressions correlate with gold-standard diagnostic criteria.

In this brief report, we investigated patient and clinician impressions of cognitive impairment and whether they correlated with objective measures of cognitive impairment over a seven-year period.

METHODS

PD patients and healthy controls (HC) were enrolled in Parkinson’s Progression Markers Initiative (PPMI). The methodology and inclusion/exclusion criteria are available at (http://www.ppmi-info.org/study-design). De novo PD patients with under 2 years of disease duration were recruited. Healthy controls were without neurological disorder or a first degree relative with idiopathic PD and scored 26 or higher on MoCA.

The PPMI database was accessed on November 25, 2019. Clinicians documented whether the participant reported cognitive decline (PD-SCD), the clinician’s impression of the participant’s cognitive status (cognitive categorization), level of confidence in cognitive diagnosis, and whether neuropsychological tests were reviewed in making this determination. Cognitive categorization, neuropsychological assessment scores, and MoCA scores were documented at baseline, 3 years, and 7 years for 388 PD patients. Neuropsychological assessments included Benton Line of Judgement, Letter Number Sequencing, Semantic Fluency, Symbol Digit Modalities, and Hopkins Verbal Learning Test (delayed recall). We employed two gold-standard measures of cognitive status, defining objective test-based cognitive impairment as a) score of less than 26 on MoCA, or b) raw score performance of 1.5 or more standard deviations below mean of the healthy control group on at least two neuropsychological assessments.

Generalized estimated equations (GEE) models were run with either the predictor a) cognitive decline (patient/caregiver) or b) cognitive categorization (clinician), and with dependent variables of cognitive impairment as defined by a) 2 or more neuropsychological tests at least 1.5 standard deviations below the mean, or b) MoCA less than 26. GEE models were controlled for age, sex, and time. Statistical analysis was performed using Statistical Analysis Software (SAS).

RESULTS

Incidence of SCD was common and increased over time, from 5.7% (6/106) at baseline to 28.3% (102/361) at 3 years to 42.6% (83/195) at 7 years. Patient and clinician perspectives of cognitive impairment differed in 10.0% (66/662) of visits. 78.8% (52/66) of these discrepancies were patient report of SCD but clinician impression of normal cognition. Discrepancy between patient and clinician perspectives grew slightly over time, from 6.6% (7/106) at baseline to 10.2% (37/361) at 3 years and 11.3% (22/195) at 7 years. The correlation between patient and clinician perspectives was positive and statistically significant (overall: φ= 0.75, p < 0.0001; baseline: φ= 0.51, p < 0.0001; 3 years: φ= 0.74, p < 0.0001; 7 years: φ= 0.78, p < 0.0001).

The number of participants deemed cognitively impaired by clinicians grew from 8.5% at baseline to 22.5% at 3 years to 32.3% at 7 years, more than tripling over time (Table 1). Level of confidence amongst clinicians in making these impressions decreased from 87.7% at baseline to only 56.4% confident within a 90–100% range after 7 years. Despite decreasing confidence and increasing impressions of cognitive impairment, however, fewer clinicians reviewed neuropsychological test results at later visits.

An increasing proportion of participants met objective criteria for cognitive impairment. At baseline, 29.6% (115/388) of participants had cognitive impairment. This increased to 35.9% (129/359) at year 3 and 34.2% (65/190) at year 7.

Participants who reported SCD were moderately likely to score as cognitively impaired on either gold-standard test (neuropsychological tests: OR = 2.20 p < 0.0001; MoCA: OR = 2.06, p < 0.0001) (Table 2). Participants with clinician impressions of PD-MCI were also moderately likely to score as cognitively impaired on gold-standard tests (neuropsychological tests: OR = 1.96, p = 00002; MoCA: OR = 2.02, p < 0.0001). The odds of these groups scoring as cognitively impaired on MoCA versus neuropsychiatric tests were roughly equivalent.

Clinician impressions of PDD were strongly associated with scoring as cognitively impaired on gold-standard tests (neuropsychological tests: OR = 7.95, p < 0.0001; MoCA: OR = 10.83, p < 0.0001).

DISCUSSION

Our study is the largest to our knowledge to evaluate the incidence of SCD in PD with long-term follow-up. As expected, we found SCD to be common in de novo PD, and incidence increased over 7 years. SCD was significantly predictive of having cognitive impairment per objective tests, congruent with previous reports that PD patients do not underestimate cognitive deficits [10, 11]. A further consideration is that SCD was often reported by combination of patient and a care partner. Recently, Deck et al. demonstrated moderate agreement between patients and care partners on a questionnaire focused on cognition-based activities of daily function [12]. We also demonstrate reasonable concordance between patient and clinician impressions of cognitive status. Discrepancies between impressions were uncommon and slightly increased over a 7-year period.

Over time, clinicians increasingly deemed patients cognitively impaired, but fewer reviewed neuropsychological test results. These findings highlight that clinicians often lack adequate tools to assist them in determining cognitive status. The underutilization of objective neuropsychological tests requires further exploration; it may suggest that clinicians have difficulty interpreting these tests or do not find them as clinically helpful as they would like.

Patient-reported SCD and clinician-reported MCI were similarly associated with a gold-standard diagnosis of cognitive impairment, suggesting that either should warrant further testing. Participants with clinician impression of PDD were 4–5 times more likely than those with impressions of PD-MCI to score abnormally on gold-standard tests. Clinician impressions of PDD were the strongest correlate of abnormal cognitive scores, showing greater association with increased severity of impairment.

By utilizing widely accepted gold-standard criteria for cognitive impairment in PD, we found that 29.6% of participants were cognitively impaired at baseline, which grew to 34.2% after 7 years. This is roughly consistent with prior studies reporting on cognitive function in PPMI [13, 14]. Since these publications, more participants have completed year 3 and 7 study visits that have been included in our study. Weintraub et al. has shown that within the PPMI, there appears to be a bias towards study drop-outs having more cognitive impairment, so our findings may be underestimated [14]. Another limitation of our study was that we grouped PD-MCI and PDD into a single category of cognitive impairment. We selected this approach since the number of PDD cases in this de novo cohort were expected to be relatively small. Finally, the questionnaires administered to participants and clinicians were brief and lacked information on cognitive function in specific activities of daily living. Further studies are needed to assess more detailed patient and clinician-sourced information about cognitive function over disease course.

CONFLICTS OF INTEREST

The authors have no conflicts of interest to report.