Association between Subjective Cognitive Complaints and Incident Functional Impairment in Parkinson's Disease

Early identification of subjective cognitive complaints (SCC) in Parkinson's disease (PD) may improve patient care if it predicts cognition‐related functional impairment (CFI).

Cognitive decline is a major complication in Parkinson's disease (PD), with 20% to 40% of patients meeting criteria for mild cognitive impairment (MCI) at the time of diagnosis, 1 and most patients progress to Parkinson's disease dementia (PDD) long-term. 24][5] Predictors of cognitive impairment include older age, older age at disease onset, greater disease severity, postural instability-gait disorder subtype, rapid eye movement sleep behavior disorder (RBD), psychosis, and depression and anxiety. 6dentifying cognitive decline at its earliest stage in Parkinson's disease (PD) may improve patient management, including long-term outcomes, and help in the development of new treatments for PD cognition, including disease-targeting therapies.Diagnostic criteria for both PDD 7 and PD-MCI 8 exist, with the primary differentiating characteristic of cognitive functional impairment being present in the former but not the latter.Both require that the patient experience a decline in cognitive abilities compared with pre-PD abilities.As a result, there is interest in evaluating the value of subjective cognitive complaints (SCC) in PD as an early marker of cognitive change.Some studies have shown that patients with PD with SCC perform significantly worse on objective cognitive measures than those without SCC, [9][10][11][12][13][14][15] whereas others have not. 16,17Only four of the studies reported longitudinal follow-up data to examine the conversion of nondemented patients with PD to PD-MCI or PDD; two of these studies found higher rates of conversion from normal cognition to PD-MCI during a 2-to 2.5-year period in patients with PD with SCC. 10,13One study found higher conversion to dementia in PD patients with SCC compared with those without SCC during a 7.5-year period, 14 and one study found no change in neuropsychological assessments between nondemented patients with PD with SCC and those without SCC at 1-and 2-year follow-up. 17In a recent study, 18 patients with PD with normal cognition, but expressing SCC, reported poorer cognition-specific functional abilities and were more likely to be diagnosed with cognitive impairment and experience global functional ability decline long term, suggesting that SCC may be a sensitive indicator of initial cognitive decline in PD.
Regarding cognition-related functional impairment (CFI), there is controversy about how to define the "minimal" functional impairment allowable for a patient with MCI, 8 as well as what constitutes significant functional impairment to support a diagnosis of dementia. 19How to best assess CFI also varies, with a small number of validated self-or informantreported questionnaires for use in PD, 20,21 as well as several performance-based measures. 22One such measure is the Penn Parkinson's Daily Activities Questionnaire [PDAQ-15], a 15-item self-completed rating scale to assess cognition-related functional abilities in PD. 20,23 The instrument was developed to be sensitive and specific to cognitive deficits and commonly reported CFI in PD.
Studies to date examining SCC have utilized data generated as part of research studies at individual research centers and by asking PD patients about cognition using specific rating scales or questionnaires.However, little is known about unstructured reporting of cognitive complaints or how to query this in large-scale studies.In this study we evaluated cross-sectional associations between unstructured self-report of SCC, characterized into specific cognitive domains, and cognition-related functional abilities, and investigated if baseline SCC is associated with future development of CFI in patients without impairment at baseline.

Standard Protocol Approvals, Registrations, and Patient Consent
This study was approved by the Western Institutional Review Board-Copernicus Group Institutional Review Board (WCG IRB), and written informed consent was obtained from each participant.

Cohort
Data were obtained from Fox Insight (FI), an online longitudinal observational study that collects patientreported outcomes from individuals self-identified as having PD. 24 This study was approved by the WCG IRB, and written informed consent was obtained from each participant.The sample was comprised of persons with PD (PwP) participating in FI who had completed a PD Patient Report of Problems (PD-PROP) assessment.PD-PROP is a series of open-ended questions that ask respondents to report their most bothersome problems and impact on function (up to five). 25,26FI participants were asked to complete the PD-PROP by keyboard entry every 3 months.For baseline analyses, general inclusion criteria were 1 one or more responses to PD-PROP 2 available data on age, sex, years since PD diagnosis (YSD), cognitionrelated functional abilities (PDAQ-15 score), 20 and motorrelated function in activities of daily living (ie, Movement Disorder Society-Unified Parkinson's Disease Rating Scale [MDS-UPDRS] Part II score 27 ).Baseline was defined as the first PD-PROP report for each individual, and baseline PD-PROP and PDAQ-15 assessments were completed within 30 days of each other.For longitudinal analyses at least one follow-up PDAQ-15 assessment was required.

Variables
Demographic data collected for FI include age, biological sex, race/ethnicity, and education.PD clinical variables include YSD, MDS-UPDRS Part II score, and cognitionrelevant psychiatric variables previously curated as problems from PD-PROP (ie, depression, apathy, and fatigue). 28For SCC self-reported on PD-PROP, a combination of human curation, natural language processing, and machine learning translated verbatim participant responses into one of eight categories: cognitive slowing, executive abilities, concentration/attention, memory, language/word finding, mental alertness/awareness, visuospatial abilities, and cognitive impairment not otherwise specified, as previously described 28 (Table S1).The PDAQ-15 is completed every 6 months as part of FI.Scores range from 0 to 60, with higher scores indicating better cognitive functional abilities.CFI was defined as a score ≤49 (representing the upper bound of the 95% confidence interval [CI] of mean scores for the MCI cohort in the validation study, 20 thereby capturing PwP with subtle difficulties on complex functional tasks, which may be present with PD-MCI 8 ).Incident CFI was defined as a PDAQ-15 score >49 at baseline and a score ≤49 at follow-up.MDS-UPDRS Part II scores were divided into quartiles.

Dataset
Data used in the preparation of this article were obtained from the FI database (https://foxinsight-info.michaeljfox.org/insight/explore/insight.jsp) on February 3, 2020.For up-to-date information on the study, visit https://foxinsight-info.michaeljfox.org/insight/explore/insight.jsp.There were originally 25,194 participants at baseline; participants aged>100 (N = 2) and lacking PDAQ-15 data (N = 4032) were excluded, creating a baseline dataset of 21,160 persons.For longitudinal analyses, individuals with functional impairment at baseline (N = 7332), with no follow-up assessments (N = 4525), or with missing motor scores or YSD (N = 200) were excluded; the sample size for longitudinal analyses at each annual time point over the 3-year period was 9103 (baseline), 6637 (year 1), 4404 (year 2), and 1470 (year 3) persons.

Statistics
Associations between baseline characteristics and baseline CFI were explored using χ 2 test or Cochran-Armitage trend test, as appropriate.For subjects with normal cognition-related functional abilities at baseline, Kaplan-Meier survival analyses examined the development of CFI over the study period, stratified by baseline SCC status.Cox proportional hazards models to examine the development of said impairment included baseline age, sex, YSD, MDS-UPDRS Part II score, depressive symptoms, apathy and fatigue, and any SCC, or specific SCC domains, at baseline.

Prevalence of Cognitive Complaints and Functional Impairment at Baseline
At baseline, 31.9%(N = 6750) of PwP reported one or more cognitive complaints among their five most bothersome symptoms.The most commonly reported complaints as classified by the algorithm were memory (13.2%), language/word finding (12.5%), and concentration/attention (9.6%); least common were visuospatial abilities (0.2%) and cognitive impairment not otherwise specified (1.2%).Regarding function, 34.7% (N = 7332) of PwP self-reported some CFI as defined by the PDAQ-15 cutoff score (Fig. S1).This number is consistent with cross-sectional prevalence rates for PD-MCI. 5

Baseline Association between Cognitive Complaints and CFI
Increasing age, male sex, increasing disease duration, worse motor-related function in activities of daily living (MDS-UPDRS Part II score), presence of depression symptoms, apathy, and fatigue as reported problems were all associated with CFI (all P-values <0.001) (Table 1).
A report of any SCC at baseline was associated with CFI; of those with CFI 50% reported any SCC versus 22% without CFI (P-value <0.001) (Table 1).When examining cognitive domain-specific cognitive complaints, the most notable differences were for cognitive slowing, executive abilities, concentration/attention, memory, language/word finding, and cognitive impairment not otherwise specified (all P-values <0.001).

Association between Baseline Cognitive Complaint and Incident CFI
Median PDAQ-15 scores did not change significantly over the 3-year period of observation (median score = 53 at each time point).The median PDAQ-15 score at baseline was 53 and over time was 54 (year 1), 53 (year 2), and 53 (year 3).

A.
B.

C. D.
FIG.In the model including any SCC reported at baseline as a variable, baseline clinical or demographic variables associated with incident CFI were older age and higher MDS-UPDRS Part II score (highest quartile vs. all other quartiles); sex and YSD were not.
Kaplan-Meier curves to 3 years from baseline for any cognitive complaint show that an estimated 45% of participants with any cognitive complaint reported at baseline had developed new-onset CFI by year 3, compared with 31% of participants without any cognitive complaint reported at baseline (Fig. 1).When baseline complaints were classified as specific domains, the highest rates of incident CFI at year 3 were for individuals who reported cognitive impairment not otherwise specified (56%), executive abilities (52%), memory (50%), and cognitive slowing (50%) at baseline compared with those who did not report those complaints (Fig. 2).In addition, regardless of the presence of cognitive complaint at baseline, older age was associated with a higher rate of incident CFI (Fig. S2).

Discussion
Using a novel patient-reported outcome measure, PD-PROP that allows reporting of PD-related problems in participants' own words, this study shows that SCCs are common in Parkinson's disease, with one-third of participants reporting a cognitive complaint among their five most bothersome symptoms.The most frequent categories of complaints, as separated by the PD-PROP algorithm, were memory, language/ wording finding, and concentration/attention difficulties.In addition, these complaints correlated with selfreported CFI at baseline and were associated with the development of future incident CFI.In particular, nearly half of participants reporting a cognitive complaint as a most bothersome symptom at baseline exhibited newonset CFI over the 3-year follow-up period.
This study differs from previous ones examining SCC in several ways.Previous studies typically have been conducted at single research centers, have had a relatively small sample size, and have specifically asked participants about their cognition, which precludes unprompted reporting and may affect the results (eg, overreporting of cognitive complaints).PD-PROP is accessible to PD patients broadly, has a very large sample size, and asks PwP to generate their own list of problems without the imposing interaction of a clinician interview, although it is possible that obtaining input from both a patient and an informant 29,30 or objective testing of cognitive functional abilities 31 may be more accurate than only patient self-report, given reported deficits in self-awareness of cognitive functional abilities in PD, particularly as cognition declines. 32Our cohort included a wide range of participants in terms of sex, age, and disease duration.Thus, PD-PROP is practical, generalizable, open ended, and unbiased.
Regarding the assessment of daily function as it pertains to cognition, there is controversy in the PD field about how best to assess this.Brief self-completed rating scales offer the advantage of being able to be administered remotely and on a large scale, making them ideal for inclusion in research studies such as PD-PROP and FI.The measure used here, the PDAQ-15, has been validated for use in PD, with good sensitivity and specificity for a diagnosis of a cognitive disorder. 20,23Assessment of report on cognitive decline and assessment of cognitive functional abilities are two of the three components required to make a diagnosis of normal cognition, PD-MCI, 8 or PD dementia. 7ne-third of participants reported some level of CFI at baseline, based on the PDAQ-15 cutoff score applied.Correlates of this impairment were increasing age, male sex, increasing disease duration, and worse motor function, all of which have been shown to be predictors of dementia in PD. 33 Additionally, at baseline any SCC was associated with CFI, as were most individual SCCs, with the most notable associations for cognitive slowing, executive abilities, memory, and cognitive complaint not otherwise specified.The SCC not otherwise specified category included self-report of "brain fog," "confusion," "mental sharpness," "thinking less clear," "less coherent," and "losing things," a category that may also include cognitive symptoms related to PD medications or complications (eg, nonmotor fluctuations).The baseline results suggest that multiple types of cognitive complaints correlate with cognition-related function in PD.
Confirming the clinical importance of reports on SCC, its presence as a most bothersome symptom at baseline was associated with future development of CFI over a relatively short period of time in those unaffected at baseline, showing it may be a simple, sensitive indicator of future clinically relevant cognitive decline.This association was true for any category of SCC, with close to half (45%) developing CFI over 3 years.Further, there were strong effects (based on HR) for complaints categorized as cognitive impairment not otherwise specified, executive abilities, memory, and cognitive slowing, with at least 50% of PwP having these complaints at baseline developing incident CFI over 3 years.These findings highlight the clinical significance of a broad range of cognitive deficits for future cognitive decline, consistent with research showing that impairments on cognitive testing across multiple domains predict future development of dementia. 34ased on these findings, we recommend that clinicians treating PD patients with new-onset SCC should prioritize routine screening of global cognitive abilities and be more mindful of comorbid medical conditions and use of medications known to adversely affect cognition.
There are study limitations to note, as follows.6][37][38] Those enrolling in PD-PROP are not diverse from a racial or ethnic standpoint and are highly educated on average, and those participating had to be familiar with computers and Internet access, so the results are not generalizable with regard to those factors.There was substantial decrease in sample size from baseline to year 3, although there were still nearly 10,000 participants who completed the year 3 visit.The PDAQ-15 cutoff point utilized to indicate at least subtle CFI has not been validated specifically for this purpose.Regarding the latter point, at baseline 50% of participants reporting CFI did not have a cognitive complaint.Explanations for this are that participants were asked to list their five most bothersome problems (complaints), and cognition may not have been one of those top five in spite of reporting mild CFI.Second, related CFI is different conceptually from a cognitive complaint.Finally, the PDAQ-15 cutoff utilized here would have some false positives, so participants without a cognitive complaint may still score as having CFI on the PDAQ-15.However, we do want to note that there was still a strong association between cognitive complaint and CFI at baseline, so the overall point still holds.A lack of a cognitive complaint does not mean that patients are not going to develop incident CFI, just that they appear less likely to do so.It has been shown that there are many demographic, motor, and nonmotor (eg, RBD, depression, anxiety, and psychosis) and neurobiological predictors of cognitive impairment in PD, but we did not have assessments for many of these features that we could enter as covariates in our models.Finally, these analyses were not able to examine if cognitive reserve is protective against development of incident CFI.
Cognitive impairment continues to be a major complication in PD that progresses notwithstanding treatment.Cognitive changes can be present even at disease onset and are variable across domains.Identifying cognitive decline at its onset provides valuable information to patients, family members, and clinicians.Future clinical trials aiming to prevent development of significant cognitive decline may target patients at the stage of having an SCC but no functional impairment.Our findings suggest that PD patients commonly identify a range of cognitive complaints and that the reported complaints are associated with concurrent and future development of CFI.Future research needs to determine if such complaints map to cognitive testing and if they predict conversion to a formal diagnosis of MCI or dementia.Regarding the ease in collecting such data serially and on a large scale, compared with formal screening, consideration should be given to making patient reporting of problems a routine part of clinical research and care.

FIG. 1 .
FIG. 1. Kaplan-Meier curve for incident CFI (cognition-related functional impairment) based on reporting of any SCC (subjective cognitive complaint) status at baseline.[Color figure can be viewed at wileyonlinelibrary.com]

2 .
FIG. 2. Kaplan-Meier curve for incident CFI (cognition-related functional impairment) based on reporting of specific SCC (subjective cognitive complaint) status at baseline.(A) Cognitive impairment not otherwise specified.(B) Executive abilities.(C) Memory.(D) Cognitive slowing.[Color figure can be viewed at wileyonlinelibrary.com]

TABLE 1
Participant characteristics at baseline

TABLE 2
Cox models of association between baseline cognitive complaint and incident cognition-related functional impairment Note: Each model includes age, sex, YSD, MDS-UPDRS Part II score, depressive symptoms, apathy, and fatigue as covariates.Abbreviations: HR, hazard ratio; YSD, years since PD diagnosis; MDS-UPDRS, Movement Disorder Society-Unified Parkinson's Disease Rating Scale; NOS, not otherwise specified.