Contribution of Aging to the Severity of Different Motor Signs in Parkinson Disease FREE

Extrapyramidal motor signs, including poverty of movement, bradykinesia, rigidity, stooped posture, and gait impairment, have long been recognized in normal aging.¹ Although these extrapyramidal features resemble those seen in patients with idiopathic Parkinson disease (PD), clinical,^{2 - 4} pathologic,^{5 - 6} biochemical,⁷ and pharmacologic⁸ evidence does not support the view that PD represents an accelerated aging process.^{9 - 10} However, the additional contribution of aging to the severity of different motor signs in patients with PD is not known. This knowledge may have implications for clinical trials of neuroprotective agents in PD,¹¹ such as the choice of primary outcome measure and analytical strategies.

In a previous community-based study,¹² we observed that axial (gait and postural) impairment, but not tremor, rigidity, and bradykinesia, was correlated with age in patients with PD and postulated that axial impairment in PD might result from a combined effect of the disease and the aging process. This hypothesis implies both (1) an independent effect of disease duration and age on axial impairment and (2) an interaction effect of disease duration and age on axial impairment. Recently, we concluded a familial aggregation study of PD,^{13 - 14} which oversampled patients with early-onset (≤50 years) PD from a tertiary referral center. As a result, the sample presented a wide range of disease duration and age, and the mean age at examination of this service-based sample was lower than in our previous community-based sample (62.0 vs 70.8 years).¹² The present study allowed us to examine the contribution of aging to the severity of motor signs in PD in a distinct sample of patients with PD. In addition, we used a different statistical method than in our previous analysis.

Patients with PD were recruited during a 4-year period from the Center for Parkinson Disease and Other Movement Disorders of the Columbia University Medical Center and a neurology clinic that primarily served individuals from the Washington Heights–Inwood community in New York City. Because the study aimed to assess the familial aggregation of PD in families of both early- and late-onset PD probands, patients with early-onset PD (≤50 years) were oversampled. We restricted ascertainment of patients with age at onset older than 50 years to those residing in the New York metropolitan area (to increase the opportunity to examine both patients and family members in their homes) and to those examined at the Center for Parkinson Disease within 5 years of onset (to facilitate discrimination between 2 motor subtypes before the initiation of levodopa). All patients with age at onset 50 years or younger were invited to participate, regardless of their duration of illness when first examined or where they lived.^{13 - 14} The protocol was approved by the Columbia University Medical Center Institutional Review Board.

Criteria for idiopathic PD included 2 of 4 neurologic signs: bradykinesia, rest tremor, rigidity, and postural instability, one of which had to be either bradykinesia or rest tremor.^{15 - 17} Patients with PD were categorized by age at onset of PD (≤50 or >50 years) based on the patient’s recall of the age of the first motor symptom of PD.¹⁸ Duration of PD was defined as the period between the first motor symptom of PD and the evaluation. Neurologic evaluation included rating according to the Unified Parkinson Disease Rating Scale (UPDRS)¹⁹ and Hoehn and Yahr stage²⁰ performed by one of several neurologists who specialized in movement disorders. Patients were not required to temporarily interrupt use of antiparkinsonian medications before rating according to the UPDRS motor examination (part III). Additional assessments have been described in detail elsewhere.^{13 - 14}

The UPDRS motor examination (part III) comprises 27 items, each with ratings that range from 0 (absent or normal) to 4 (most severe impairment).¹⁹ The total UPDRS motor score (range, 0-108) was divided into 6 motor domains (tremor, rigidity, bradykinesia, facial expression, speech, and axial impairment) based on the cardinal clinical manifestations of PD. The number of items in each domain is 7 for tremor, 5 for rigidity, 9 for bradykinesia, 1 for facial expression, 1 for speech, and 4 for axial impairment. In addition, the 6 motor domains were grouped into 2 subscores that represented predominantly dopaminergic (subscore A: tremor, rigidity, bradykinesia, and facial expression) and nondopaminergic (subscore B: speech and axial impairment) deficiency based on levodopa responsiveness.¹²

The contribution of age to each of the 6 motor domains and 2 subscores (dependent or outcome variables) was analyzed using 2 multiple linear regression models.²¹ For each outcome variable, model 1 included disease duration and model 2 included both disease duration and age at examination (independent or predictor variables). Models 1 and 2 adjusted for sex, years of education, levodopa dosage, and use of other antiparkinsonian medications (dopaminergic agonists [yes/no], selegiline hydrochloride [yes/no], amantadine hydrochloride [yes/no], and anticholinergics [yes/no]).

The variation of the outcome variable explained by each of the 2 linear regression models was calculated using the adjusted coefficient of multiple determination (adjusted R²). The coefficient of multiple determination measures the proportion of total variation in the outcome variable associated with the use of a set of predictor variables. Unlike the unadjusted R², which can only become larger when an additional variable is included in the model, the adjusted R² adjusts for the number of predictor variables and can actually become smaller when an additional variable is included in the model.²¹ The change in adjusted R² obtained by adding age at examination to the models (model 2 adjusted R² minus model 1 adjusted R²) was used to gauge the contribution of aging to each motor domain and subscore of the UPDRS. We used the bootstrapping procedure to calculate approximate 95% confidence intervals (CIs) for the adjusted R² and change in adjusted R² (1000 bootstrap samples, random X sampling). The bootstrapping procedure involves sampling with replacement from the observed sample data a large number of times and calculating the estimate of interest from the bootstrap sample each time. Based on the distribution of the bootstrap sample estimates, 95% CIs were calculated using the bias-corrected percentile method.²²

For each outcome variable, we also investigated whether there was an interaction between disease duration and age at examination. To avoid multicollinearity, the 2 variables were centered before being multiplied; the mean disease duration and age at examination were subtracted from each patient’s disease duration and age at examination, respectively. Disease duration, age at examination, and the product term were included in linear regression models that adjusted for sex, years of education, levodopa dosage, and use of other antiparkinsonian medications.

The assumptions of constant variance and normal distribution of the error terms in the linear regression models were checked by examining plots of the residuals vs fitted values and normal probability plots of the residuals. Nonconstancy of error variance was observed for subscore B, tremor, and axial impairment, which was remedied by log transformation of these outcome variables. We report the nontransformed models to ensure comparability of model fitting across outcome variables, but results were similar when using the transformed models for subscore B, tremor, and axial impairment. The possibility of multicollinearity in the linear regression models was assessed by the variance inflation factor.²¹ No serious multicollinearity was observed in the reported models. Because of the large number of outcome variables in this study, we corrected for multiple comparisons and increase in type I error rate by using a significance level of .005 (conventional significance level of .05 divided by 10).

Of 536 patients recruited, 85 were excluded from this analysis: 47 because they missed 1 or more items of the UPDRS, 18 because they had undergone fetal tissue transplantation, 17 because they had undergone deep brain stimulation, and 3 because they had missing medication information. Patients who underwent fetal tissue transplantation and deep brain stimulation were excluded because these therapeutic modalities would add other sources of influence on the outcome measures that would need to be adjusted for in the analysis. Excluded patients had lower age at onset of PD (mean ± SD, 48.2 ± 14.4 vs 54.9 ± 13.4 years; P<.001) and longer disease duration (mean ± SD, 13.8 ± 7.1 vs 7.2 ± 5.9 years; P<.001) than the 451 included patients. No significant differences were observed for age at examination, years of education, sex, and ethnicity.

Demographic and clinical characteristics of the 451 PD patients are given in Table 1. Mean age at onset of PD was 54.9 years (SD, 13.4 years). Two hundred one patients (44.6%) had early-onset PD (age at onset ≤50 years) and 87 (19.3%) had age at examination of 50 years or younger. Mean age at examination was 62.0 years (SD, 12.6 years; median, 62.0 years; range, 18-93 years), and mean disease duration was 7.2 years (SD, 5.9 years; median, 5.6 years; range, 0.1-41.6 years). Age at examination and disease duration were not significantly correlated (Pearson correlation r, 0.07; P = .15).

Both disease duration and age at examination were significant independent predictors of the total UPDRS motor score and subscore B when included in the same model (model 2, Table 2). For subscore A, only age was a significant predictor in model 2. The additional variation of the outcome variable explained by including age in the models was higher for subscore B (14.3%; 95% CI, 9.9%-20.4%) than subscore A (4.7%; 95% CI, 2.0%-9.1%) (Table 2). The bootstrapping procedure was also used to calculate an approximate 95% CI for the difference between subscore B and subscore A adjusted R² changes (difference, 14.3% − 4.7% = 9.6%). This interval did not include zero (95% CI, 6.1%-14.5%).

Both disease duration and age at examination were significant independent predictors of bradykinesia, speech, and axial impairment when included in the same model (model 2, Table 3). For rigidity and facial expression, only age was a significant predictor in model 2, whereas for tremor, neither disease duration nor age was a significant predictor. The additional variation of the outcome variable explained by including age in the models was highest for axial impairment (13.6%; 95% CI, 9.4%-19.6%) (Table 3).

When we investigated the interaction of disease duration and age at examination, the interaction term was significant at the .005 level for both subscore A (β = .136, P = .003) and subscore B (β = .204, P<.001). Among the 6 motor domains, the interaction term was significant at the .005 level for bradykinesia (β = .125, P = .005), facial expression (β = .152, P<.001), speech (β = .126, P = .004), and axial impairment (β = .206, P<.001) and was not significant for tremor (β = .060, P = .21) and rigidity (β = .091, P = .05).

In this large sample of PD patients with a wide range of disease duration and age, age was a significant predictor of rigidity, bradykinesia, facial expression, speech, and axial impairment. However, the change in adjusted R² was highest for axial impairment, and age explained the variability of a subscore of the UPDRS that comprised speech and axial impairment (subscore B) to a greater extent than a subscore that comprised tremor, rigidity, bradykinesia, and facial expression (subscore A).

The assumption that nondopaminergic lesions play a predominant role in speech and axial impairment in PD is based on clinical studies that show that these motor signs, unlike tremor, rigidity, and bradykinesia, are relatively refractory to levodopa therapy, especially in the middle and late stages of the disease.^{23 - 26} This is also supported by electrophysiologic and quantitative investigations of the influence of dopaminergic medication on postural responses and balance impairment^{27 - 28} and gait parameters^{29 - 30} in PD patients.

Our findings are consistent with the proposal that older PD patients have increased motor disability because of more widespread subcortical involvement, including nondopaminergic structures.^{31 - 33} While many cross-sectional and longitudinal studies have demonstrated that older age is associated with more rapid progression of PD,^{34 - 41} clinical and pharmacologic evidence suggests that the effect of older age on increasing disability is more prominent for non–levodopa-responsive axial motor impairment.^{32 - 33 ,41 - 44}

In one previous cross-sectional investigation,³⁷ no significant interaction between the effects of disease duration and age on progression of PD was observed, leading to the conclusion that disease duration and age influenced the natural history of PD in an additive fashion; however, clinical severity was quantified with the limbs bradykinesia scores from the UPDRS only. In the present study, we observed an interaction effect of disease duration and age on bradykinesia, facial expression, speech, and axial impairment. However, the interaction was of highest magnitude and most significant for axial motor impairment (β = .206, P<.001). In another study, we observed that a group of PD patients with older age and high severity had a significantly increased risk of dementia compared with a group with younger age and low severity, but the groups with older age and low severity and younger age and high severity did not.⁴⁵ Overall, these findings suggest that late-stage non–levodopa-responsive motor and cognitive manifestations of PD may result from the combined effect of the disease and the aging process on nondopaminergic subcortical structures, such as the locus ceruleus, pedunculopontine nucleus, and nucleus basalis of Meynert.

The main strength of our study is the wide range of age at onset and age at examination of the PD patients. Although we reported the results using age at examination as a predictor variable in the linear regression models, we repeated the analysis using age at onset of PD instead. In general, findings were similar, but the models that used age at onset accounted for a lower variation of the outcome variables. One limitation of this study is its cross-sectional rather than longitudinal design. In addition, patients were taking antiparkinsonian medications when they were evaluated using the UPDRS. However, detailed information on medication use, including levodopa dosage, was available for the analysis and used in the linear regression models. Still, one might argue that the effect of dopaminergic medication in reducing the severity of levodopa-responsive motor signs to a greater extent could obscure the effect of age on tremor, rigidity, and bradykinesia. We addressed this possibility by performing additional analyses stratified by the median levodopa dosage. Among patients taking less than 300 mg of levodopa or not taking levodopa, we did not observe a higher contribution of age to levodopa-responsive motor signs.

In conclusion, aging contributed most to the severity of axial impairment among the motor signs of PD, and a highly significant interaction between disease duration and age was observed for axial impairment. This finding has implications for therapeutic trials of agents that may affect PD progression,¹¹ such as consideration of axial impairment as a primary outcome measure and investigation of an interaction between the effects of neuroprotective drug treatment and age.

Correspondence: Gilberto Levy, MD, Department of Biostatistics, Columbia University, 722 W 168th St, Sixth Floor, New York, NY 10032 (GL227@columbia.edu).

Accepted for Publication: May 25, 2004.

Author Contributions:Study concept and design: Levy. Acquisition of data: Levy, Louis, Cote, Perez, Mejia-Santana, Andrews, Harris, Waters, Ford, Frucht, Fahn, and Marder. Analysis and interpretation of data: Levy. Drafting of the manuscript: Levy. Critical revision of the manuscript for important intellectual content: Louis, Cote, Perez, Mejia-Santana, Andrews, Harris, Waters, Ford, Frucht, Fahn, and Marder. Statistical analysis: Levy. Obtained funding: Marder. Administrative, technical, and material support: Cote, Perez, Mejia-Santana, Andrews, Harris, Waters, Ford, Frucht, and Fahn.

Funding/Support: This study was supported by federal grants NS36630, RR00645, AG07232, NS39422, and NS29993 and the Parkinson’s Disease Foundation, New York, NY.

Acknowledgment: We appreciate the assistance of Paul Greene, MD, and Susan Bressman, MD, in the recruitment of individuals for participation. We thank Bruce Levin, PhD, for statistical advice.