Introduction

Cognitive impairment and dementia are frequent in Parkinson's disease (PD)1 with major consequences for daily functioning, quality of life, care giver burden and health-related costs. Cortical Lewy bodies, diffuse α-synuclein aggregates, Alzheimer-type changes including amyloid plaques and neurofibrillar tangles, and neurochemical changes all contribute to dementia in PD. Of the cortical neurochemical alterations, cholinergic deficits appear to be particularly important, and are associated with cognitive impairment in both postmortem2 and in vivo3 studies.

There is increasing awareness in a variety of conditions that certain drugs may accelerate cognitive decline. For example, drugs with anticholinergic properties are widely used and may cause central nervous system anticholinergic syndromes like delirium and cognitive impairment.4–6 Because of the increased permeability of the blood–brain barrier, age-related pharmacodynamic changes, the risk for polypharmacy and interactions, the ageing brain is particularly sensitive to these toxic effects.7 8

Although the prescription of anticholinergic agents as a treatment for the motor symptoms of PD is now less favoured, it is not at all uncommon that patients with PD receive drugs with anticholinergic properties. Some agents targeting motor and bladder symptoms, and drugs used to treat common neuropsychiatric symptoms including depression9 and psychosis,10 have anticholinergic properties. Comorbid physical conditions are also common in elderly PD patients, and many drugs commonly prescribed for older people such as bronchodilatators, antiarrhytmic drugs, corticosteroids, analgesics, antihistamines or antihypertensives have also anticholinergic properties, many of which are not recognised as anticholinergics by a substantial proportion of physicians.5 7 8

Patients with brain diseases with cholinergic deficits are particularly sensitive to anticholinergic drug-effects. For example, drugs with anticholinergic effects were found to be significantly associated with a higher risk in women to progress from mild cognitive impairment to dementia.11 In another study, antipsychotic drugs were associated with worsening cognitive decline in people with dementia, potentially due to their anticholinergic activity (AA).12 In a direct comparison of two antipsychotic agents with and without AA, the increase in AA measured with radioreceptor assay was found in the group taking the drug with AA, and this increase was associated with increase in anticholinergic side effects and cognitive slowing.13

Given the reduced cholinergic function in PD and its association with cognitive impairment, drugs with anticholinergic activities may contribute to the observed cognitive impairment in people with PD. However, this hypothesis has not been systematically investigated. Therefore, using data from a prospective, longitudinal study, the use and impact of drugs with anticholinergic properties in a community-based sample of patients with PD was studied. Such information is important for the management of patients with PD, as avoiding drugs which accelerate cognitive decline may be a key part of optimal therapy.

Material and methods

Patients: longitudinal community cohort of PD patients

Patients were drawn from a longitudinal prospective community-based prevalence study of 245 patients diagnosed as having PD on 1 January 2003 in Rogaland County, Norway. Three patients were rediagnosed as not having PD during follow-up, and seven had insufficient data about their medication or MMSE score, leaving 235 patients for this study. The recruitment strategy, patient samples, and design of this study have been described elsewhere.14 PD was diagnosed according to published criteria15 by neurologists with special expertise in movement disorders. Patients with dementia, diagnosed according to DSM IIIR as previously described,16 were included if it was clearly established that onset of dementia occurred at least 1 year after onset of PD. Twenty-seven of the total PD population have come to autopsy, and the diagnosis of PD was confirmed neuropathologically in all cases.16 After a comprehensive baseline assessment, patients were followed longitudinally and reassessed after 4 and 8 years, including a full drug history. A total of 146 subjects were available at 4-year and 92 at 8-year follow-up evaluation. Attrition was mainly due to death; only 14 (year 4) and five (year 8) subjects refused to participate or could not be traced (figure 1). Information on the assessment at follow-up has been provided elsewhere.17

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Flow chart showing the distribution of patients and use of drugs with anticholinergic activity at the three assessment points.

Baseline assessment

Clinical assessment at baseline

Cognition was assessed with the Mini-Mental State Examination (MMSE)18 at baseline and at each follow-up assessment. The clinical examination of motor symptoms included disease stage according to the Hoehn and Yahr Scale.19 Depression was rated using a clinical rating scale, the Montgomery–Åsberg Depression Rating Scale (MADRS), by a neurologist after a training procedure.20 Major depression according to DSM IIIR was diagnosed after a clinical interview as previously described.14 Follow-up assessments were completed with the same instruments at 4- and 8-year visits as part of a systematic longitudinal follow-up of the cohort.

Measurement of anticholinergic load

To estimate the use of with AA in our sample, we modulated the procedures recommended by Chew et al.5 In that study, the AA of 107 medications commonly prescribed to older adults was detected using an in vivo radioreceptor assay, grading them in five levels, 0 (no AA); 0/+ (no or minimal AA), + (low AA), ++ (moderate AA) and +++ (high AA). We transformed this graduation in five categorical scores from 0 to 4. For drugs which were not included in this study, AA scores were specified by two of the authors independently from each other using available evidence from the literature (UE and KB). The scores from every agent used by each patient were summed up, and this sum score was considered as the total AA load. This was done for each assessment point. For the 8-year observation period, a total AA load was calculated by adding together the AA load at baseline and the two follow-up assessments.

The scale does not take into account the individual doses of the drugs. Furthermore, there was no detailed information available for the start and stop date of the drugs. Therefore, to estimate the duration of treatment with AA drugs, the number of assessment points with AA drugs was calculated, that is, each patient would have a duration category between 0 (never) and 3 (at each assessment).

Analyses

Statistical analyses were performed using the software program SPSS V.16 (SPSS, Chicago). To examine if patients who received AA drugs differed from those who did not on demographic and clinical variables at baseline, comparisons of normally distributed continuous variables were carried out using a one-way analysis of variance. The Mann–Whitney U test was used for non-normally distributed variables. The χ² test was used for categorical variables. The primary outcome variable was the rate of cognitive decline, that is, the difference between MMSE scores at baseline MMSE and at 8-year follow-up. A secondary confirmative analysis was undertaken, focusing on the baseline cognitive performance. Normality was assessed using the Kolmogorov–Smirnov test. Only patients with follow-up data on MMSE were included.

Multivariate hierarchical linear regression analyses were performed to adjust for potential confounders, with cognition or change in cognition as dependent variable, and AA load or AA duration as predictor. In the first model, age, education and gender were entered, then baseline MMSE and MADRS scores, and finally the AA drug use.

p Values less than 0.05 were considered statistically significant.

Results

Cross-sectional analyses

Of the 235 patients included at baseline, 114 (48.5%) were men, the mean (SD) age was 74.7 (8.4) years, and the duration of PD was 9.2 (5.8) years. The median (interquartile range (IQR)) Hoehn & Yahr Score was 3,2 MADRS 69 and MMSE 27.6

A total of 99 substances were used by the PD patients, and 29 of these drugs had clinically relevant AA. The AA rating of all the used agents is displayed in table 1. At baseline, 102 of the 235 patients (43.4 %) received at least one drug with AA. The most common drugs with AA were antidepressants (n=84), cardiovascular agents (n=80), anxiolytics and sedatives (n=61), and antipsychotics (n=44). Seventy-two subjects (30.6%) were taking more than one AA drug (total range 0–5). A total of 75 subjects (74.5%) took psychotropics: 17 patients (16.7%) received antidepressants only, five (4.9%) an antipsychotic only, 12 (11.8%) an anxiolytic only and 41 (40.2%) a combination of psychotropic and other AA drugs. Cholinesterase inhibitors were used by only three patients during the study period.

The baseline characteristics of patients receiving drugs with AA at baseline and those without drugs are shown in table 2. Patients taking AA agents had significantly lower cognition (median 25) and higher depression scores9 than those not taking AA agents (28 and 6, respectively), whereas age, duration of PD, or severity of motor symptoms did not differ significantly. Total AA load correlated significantly with MMSE (Spearman r=−0.205, p=0.002) and MADRS (Spearman r=0.321, p<0.001). Using linear regression analysis, after adjusting for age and gender, MADRS (standardised β 0.291, p<0.001) but not MMSE score remained significantly associated with the AA load in a linear regression analysis.

Longitudinal analyses

Seventy-six of the 92 patients seen at 8-year follow-up completed an MMSE. The characteristics of those completing MMSE at follow-up and those who did not are shown in table 3. The characteristics at follow-up of those who had used AA or not at some point during the study are shown in table 4 . The median (IQR) change in MMSE over the full 8-year period in subjects who had used AA drugs at some point (n=56) was 6.5 points (IQR 15.3), whereas those who had not used AA drugs at any assessment (n=20) declined by only 1 point (IQR 6.25). This difference was significant (Mann–Whitney test z=−2.5, p=0.025). The correlation between total AA load during the 8 years (ie, sum of AA load at all three assessments) and the change in MMSE score was significant (Spearman r=0.32, p=0.004). In the linear regression analysis, after adjustment for gender, education, age and baseline Hoehn & Yahr stage scores at baseline, the association between AA load and MMSE decline was significant and remained significant after including baseline MMSE and MADRS scores (β 0.162, SE 0.077, standardised β=0.229, p=0.040; total model F=3.1, p=0.010). Similarly, there was a significant association between duration of use of AA drugs and cognitive decline after adjustment for age, education, gender, baseline Hoehn and Yahr stage, and baseline MMSE and MADRS scores in a multivariate linear regression analysis (β=1.8, SE 0.9, standardised β 0.231, p=0.032, total model F=4.9, p<0.001).

Discussion

We report a significant association between drugs with AA and the rate of cognitive decline in a large and community-based cohort of patients with PD followed prospectively for 8 years. The rate of decline was 6.5 times higher in the group using such drugs compared with those never taking drugs. Our finding of an association between cognition and estimates of load and duration of AA drug use after adjustment for potential confounders adds robustness to the findings and is consistent with studies in the general population. Emphasising the importance of the finding to the population of people with PD, at baseline, more than 40% used at least one drug with AA.

The additional finding of an association between AA drug use and depression at baseline is in line with previous studies reporting an association between cholinergic deficit in Lewy body diseases with depression.3 However, although this association remained significant after adjustment for age and gender, it is possible that it is entirely due to antidepressant drugs being given to patients with more severe depression. Similarly, an uneven distribution of other psychopathologies, some of which may be associated with cognitive impairment in PD, and frequently treated with drugs with AA effects, might also have influenced the findings. A recent trial with nortriptyline, a tricyclic antidepressant with AA, demonstrated superior efficacy over placebo, with no worsening of cognition.25 Our findings, however, suggest marked long-term cognitive effects of drugs with AA which may offset the positive effect on mood. Clinical judgement is required to solve this dilemma.

There are several potential mechanisms for a more rapid decline in PD patients taking AA drugs. First, cholinergic activity is reduced in PD.2 Cholinergic activity enhances attentional processes and other cognitive functions,6 and drugs with anticholinergic effects induce impairments in various cognitive domains including memory22 and language.21 Stimulation of cholinergic circuits seem to increase cerebral perfusion, and thus cholinergic blockade may cause neurovascular dysregulation and worsen cognition.26

In addition, there is convincing evidence linking muscarinic receptor activity to the processing of amyloid precursor protein in animal models, with stimulation of muscarinic receptors increasing non-amyloidogenic processing.27 This might explain the findings that drugs with anticholinergic effects were associated with more severe amyloid plaques,27 and conversely that cholinergic drugs may reduce cortical amyloid load.28 There is also evidence that activitation of muscarinic receptors may prevent tau phosphorylation,24 which may explain the findings of increased neurofibrillar tangles after anticholinergic treatment in PD27 and neuroleptic treatment in dementia with Lewy bodies.29 There is also preliminary evidence linking muscarinic receptor activation to the processing of α-synuclein,30 the key protein in the pathogenesis of PD and the main substrate underlying cognitive decline in PD.16 In addition, cholinergic deficits have been reported to be associated with reduced neurogenesis.23

There are methodological limitations that need to be considered when interpreting these findings. First, due to the long test–retest interval, there was a high attrition due to death, and some of the survivors could not provide adequate scores on tests such as MMSE and MADRS. Since dementia is associated with mortality in PD, this may have introduced a selection bias. In addition, the long interval increased the likelihood that changes in drug treatment relevant to the study would go undetected. Second, since this was a naturalistic study, and patients were not randomised to AA treatment, the groups taking AA drugs may have differed from those not taking such drugs. As discussed above, it is possible that patients who were already on a steeper trajectory of cognitive decline, due to coexisting psychopathology or dementia, were more likely to be prescribed AA drugs. However, there were no differences in the motor symptoms or age at baseline. Depression was, however, more severe in those with AA at baseline. This might have influenced the finding, since depression has been found in some studies to be associated with increased cognitive decline.31 However, we have not found such an association in our longitudinal studies.17 This aspect is further complicated by antidepressants being one of the most commonly used AA drugs in this study. However, it is highly unlikely, for ethical reasons, that randomised trials with drugs with anticholinergic effects will be performed to clarify these issues.

Since a relationship between cognitive impairment and drugs with anticholinergic properties has been discussed for many years based on findings in non-PD cohorts, clinicians might also have reduced the anticholinergic load in patients with emerging cognitive decline, which would have attenuated any relationship between cognitive impairment and AA drugs. Finally, MMSE is not a very sensitive measure of cognitive impairment in PD.32 This strengthens the finding of the observed association between AA and cognitive impairment, and an even stronger association might have been observed if more sensitive cognitive measurements had been applied.

Strengths of the study include the large and community-based cohort and the long, prospective follow-up period. Standardised clinical assessments were performed, and the diagnosis of PD was made by movement-disorder specialists, according to operationalised criteria. Patients were continually followed by a study neurologist during the observation period, which increases the accuracy of a clinical diagnosis of PD. In addition, a subset of patients have come to autopsy with confirmation of the diagnosis in all cases.16

A further strength is the detailed assessment of the anticholinergic load, taking into account both the duration and total dosage, and including not only the well-known drugs such as tricyclic antidepressants and neuroleptics, but also other drugs associated with such effects. There is no universally accepted method yet to assess and quantify the anticholinergic burden. Older measures of anticholinergic medication exposure were developed based on clinical experience and knowledge of the pharmacological properties of the medications drawn from the literature.33 34 Serum radioreceptor assay quantifying the drug induced muscarinic blockade has been used to detect serum AA.35 36 However, serum AA may reflect primarily the peripheral endogenous and exogenous AA and may be less representative for the anticholinergic effects on brain and cognition.37 We used a recently published classification that takes into account the dose-dependent anticholinergic properties of the drugs, since higher doses than the clinically relevant doses may be required for some drugs to induce anticholinergic effects.5 We slightly modified the scale by also including drugs which were not included previously, after independent assessment by two of the authors. A limitation with this method was that dosage and duration of treatment are not taken into account.

In conclusion, we found that drugs with anticholinergic properties are commonly administered to patients with PD, and that the number and treatment treatment duration of drugs with AA seem to be associated with more rapid cognitive impairment. These findings have implications for the management of PD patients, since avoiding such drugs may avoid increasing progression of cognitive decline, and highlight the need for continued physician education to ensure avoidance of inadvertent prescription of such drugs.