Longitudinal motor decline in Dementia with Lewy Bodies, Parkinson Disease Dementia, and Alzheimer’s dementia in a community autopsy cohort

Abstract

INTRODUCTION:

We examined the progression of extrapyramidal symptoms and signs in autopsy-confirmed dementia with Lewy bodies (DLB), Parkinson’s disease dementia (PDD) and Alzheimer’s Disease dementia (AD).

METHODS:

Longitudinal data were obtained from Arizona Study of Aging and Neurodegenerative Disease, with PDD (n=98), AD (n=47) and DLB (n=48) further sub-grouped as with or without parkinsonism (DLB+ and DLB−). Within-group UPDRS-II and UPDRS-III trajectories were analyzed using non-linear mixed effects models.

RESULTS:

In DLB, 65.6% had parkinsonism. Baseline UPDRS-II and III scores (off-stage) were highest (p<0.001) for PDD (mean±SD 14.3±7.8 and 27.4±16.3), followed by DLB+ (6.0±8.8 and 17.2±17.1), DLB− (1.1±1.3 and 3.3±5.5) and AD (3.2±6.1 and 8.2±13.6). Compared to PDD, the DLB+ group had faster UPDRS-III progression over 8-years (Cohen’s-d range 0.98 to 2.79, p<0.001), driven by gait (p<0.001) and limb bradykinesia (p=0.02) subscales.

DISCUSSION:

Motor deficits progress faster in DLB+ than PDD, providing insights about expected changes in motor function.

1. BACKGROUND

Parkinson’s Disease Dementia (PDD) shares several clinical and pathological features [1] with Dementia with Lewy bodies (DLB). Parkinsonism is a core feature of both PDD and DLB [2,3] but the diagnosis is dependent on the temporal emergence of motor symptoms relative to cognitive symptoms, with an arbitrary 1-year cut off [4]. Lower diagnostic accuracy of DLB antemortem has been associated with absence of parkinsonism [5] and higher Braak stages of comorbid Alzheimer’s Disease (AD) neuropathology [6,7]. Studies comparing outcomes in DLB vs. PDD are scarce, and it can be clinically difficult, even for subspecialists, to distinguish the two once dementia has progressed. While significant overlap exists in their presentation and evolution [4,8], studies comparing survival and cognitive profiles have shown inconsistent results [9–12].

Limited information is available about the characteristics and evolution of parkinsonian motor features in DLB [13,14]. Motor features in DLB are considered less likely to be associated with rest tremor [15], and to be less severe with lower response rates to levodopa [4,16]. Other studies have described the value of recognizing prodromal motor changes prior to development of either Parkinson’s Disease or DLB [17] and one prospective study found differences in motor signs and symptoms in prodromal DLB and PD [18]. However, long-term comparative trajectories for progression of motor impairments in DLB and PDD are not well-studied. Characterizing the trajectories of parkinsonian symptoms and signs is foundational for early recognition of DLB, prognostication of clinical progression, and identifying treatment targets for disease modifying therapies in clinical trials. We examined longitudinal progression of motor signs and symptoms in a community-based, clinicopathological study (Arizona Study of Aging and Neurodegenerative Disorders, AZSAND) with autopsy confirmed DLB, PDD and AD dementia. We further evaluated whether the evolution was related to any sub-features of parkinsonism.

2. METHODS

2.1. Selection of participants

All cases were participants of AZSAND [19] and had a final clinicopathological diagnosis of DLB, PDD or AD at data freeze (January 2021). DLB and PDD cases with AD co-pathology were included. Participants were only included in the AD subgroup if they had no other major secondary neuropathological diagnoses including vascular dementia and other neurodegenerative conditions. All participants had annual (when able) neuropsychological/cognitive evaluations and movement exams (by movement disorders subspecialists); when possible, these were performed in the practically-defined off state. For this study, each participant had to have completed at minimum two movement exams with the final exam within 2 years prior to death (Supplemental Figure 1) and only off-state measures were used. Consensus movement and cognitive diagnoses for participants were determined after each visit and after death utilizing data from all AZSAND evaluations and private medical records. Neuropathologic assessment was performed independent of any clinical information [19]. Final clinicopathologic diagnoses were determined by consensus between neuropathologists, movement disorder neurologists, cognitive-behavioral neurologists and neuropsychologists in the program consortium [19,20].

2.2. Data extraction

For each participant, the following data were extracted: Age at enrollment, sex, ethnicity, education, global deterioration scale (GDS) [21], Mini Mental State Exam (MMSE) [22] at first visit and prior to death, Apolipoprotein-E (APOE) Ɛ4 genotype, diagnosis at first and last movement and cognitive evaluation, and age at death. In conjunction with neuropathologic evidence, PDD was clinically diagnosed based on previously outlined criteria [23] and DLB was diagnosed using McKeith criteria [2,7]. DLB was divided into two groups based on the presence or absence of parkinsonism at final exam (DLB+ and DLB−). Total scores for Unified Parkinson’s Disease Rating Scale (UPDRS)-II and UPDRS-III [24,25] in the practically defined-off stage (if on dopaminergic medications) were extracted. Individual subscales of interest for UPDRS-III were grouped as: Body bradykinesia, Limb bradykinesia, Gait, Postural instability, Limb rest tremor, Neck and Limb rigidity [26]. Where applicable, if more than one limb was involved, the average score of all limbs was used. Finger tapping, hand movements, pronation-supination and leg agility were grouped together as Limb bradykinesia (for asymmetric patients, left and right were averaged). Limb rest tremor was defined as mean of rest tremor in all limbs. Gait, postural instability, neck rigidity and body bradykinesia were standalone subscales. Face, lip, and neck tremor were averaged separately but not included in model as a subscale of interest. Frequency of presence vs. absence of subscales were also extracted separately. From the pathological diagnosis, we determined Braak staging [27] for participants who had concomitant AD pathology and Unified LB stage [28] for participants with Lewy body pathology. The tissue processing and staining methodology have been described previously [19,29].

2.3. Statistical Analysis:

Demographic data was analyzed using Chi-square or Kruskal Wallis, with statistical significance set a priori at p < 0.05. To assess between group differences in motor trajectories (defined by UPDRS-II and UPDRS-III) and longitudinal change, mixed, linear, and non-linear quadratic, effects models with random intercept and random slope were employed. The mixed effects modelling approach accounted for missing data by differential weighing such that cases with more data points were given higher weight. Supplemental Figure 2 and Supplemental Table 1 outlines the fixed effect and random terms in our initial model. The most non-significant terms were removed by backward-elimination [30]. Sensitivity analyses were performed by adjusting for baseline MMSE in the original models; these were removed in the final model to increase power as they did not change model estimates. Model predicted mean scores were then generated for the same baseline scores (UPDRS-II of 5 and UPDRS-III of 10) for a 76-year-old male with 15 years of education (mean age and years of education for all participants with most participants in the PDD and DLB groups being male). Estimation of change in UPDRS standardized effect sizes were calculated at 2-year intervals from baseline movement visit using Cohen’s d-measure. The mixed model derived above was applied to individual subscale scores (0–4) to evaluate which subscales potentially contributed most to between group differences. The variance explained (R²) by mixed-effects model, by fixed effects only or by fixed terms only including group effect were calculated using Nakagawa and Schielzeth’s method [31]. Modeling was performed with the use of the PROC MIXED procedure in SAS software, version 9.3 (SAS Institute, Cary, NC).

3. RESULTS

There were 257 cases with a final clinicopathologic diagnosis of DLB, PDD or AD without any major secondary co-pathology. Data included were from 193 participants: DLB (n=48), PDD (n=98) (±AD co-pathology for each) or AD without co-pathology (n=47), after exclusion of participants who did not have sufficient movement exams. In sum, over 1485 participant-years of longitudinal data were included in the analyses.

3.1. Baseline Characteristics:

At enrollment, participants with DLB and PDD diagnosis were predominantly male; female sex was more common in the AD group. Table 1 outlines the demographic characteristics, initial cognitive and movement diagnoses, and final pathologic staging of participants in the four groups.

Table 1:

Demographic, clinical features, and pathologic diagnosis for all groups.

Parameters	PDD (n= 98)	DLB + (n = 33)	DLB − (n = 15)	AD (n=47)	p-value
Age	73.4 (8.1)	77.8 (6.1)*	81.1 (6.4)	78.7 (7.4)	<0.001
Males (%)	69 (70.4%)	24 (72.7%)	12 (80.0%)	15 (31.9%)‡	<0.001
Education, years (SD)	15.4 (2.6)	14.9 (2.6)	17.7 (4.5)	14.8 (2.4)	0.083
MMSE, mean (SD)	24.3 (6.0)	20.9 (8.8)	23.6 (6.3)	24.1 (7.4)	0.115
Global Deterioration Scale, mean (SD)	3.2 (1.2)	3.7 (1.5)	3.5 (1.7)	3.4 (1.5)	0.566
MMSE prior to death, mean (SD)	19.4 (7.0)	14.0 (8.3)*	17.1 (7.8)	17.7 (8.0)	0.011
APOE Ɛ4 allele					0.010
None	70 (72.9%)	19 (57.6%)	11 (73.3%)	19 (43.2%)
Heterozygous	26 (27.1%)	13 (39.4%)	3 (20.0%)	22 (50.0%)
Homozygous	0 (0.0%)	1 (3.0%)	1 (6.7%)	3 (6.8%)
Age at death (SD)	80.6 (6.2)	83.6 (6.7)*	88.9 (8.0)†	87.3 (7.7)	<0.001
Age at dementia diagnosis, y	76.2 (7.3)	75.3 (8.4)	81.9 (8.6)	79.4 (10.2)	0.013
Dementia duration	4.5 (3.0)	8.3 (4.0)	7.0 (5.1)	7.7 (4.4)	<0.001
Cognitive diagnosis at first evaluation
Normal	10 (25.6%)	6 (23.1%)	3 (33.3%)	11 (42.3%)	0.049
MCI	16 (41.0%)	5 (19.2%)	0 (0.0%)	4 (15.4%)
Dementia	13 (33.3%)	15 (57.7%)	6 (66.7%)	11 (42.3%)
Pathologic diagnosis and staging
Secondary neuropathologic diagnosis of AD	48 (49.0%)	32 (97.0%)*	13 (86.7%)	N/A	<0.001
Braak Stage 0 – III	45 (45.9%)	5 (15.2)	2 (13.3%)	5 (10.6%)	<0.001
Braak Stage IV – VI	52 (53.0%)	28 (84.8%)	13 (86.7%)	42 (89.4%)
Unified LB stage IIA	4 (4.1%)	0	0†	0‡	<0.001
Unified LB stage IIB	2 (2.0%)	0	0†	0‡
Unified LB stage III	30 (30.6%)	5 (15.2%)	6 (40%)†	0‡
Unified LB stage IV	62 (63.3%)	28 (84.8%)	9 (60%)	0‡
Movement diagnosis
Parkinsonism at first movement visit	95 (96.9%)	17 (51.5%)*	2 (13.3%)†	8 (17.0%)	<0.001
Parkinsonism at final movement visit	100%	100%	0†	14 (29.8%)‡	<0.001
Number of movement visits (median, range)	4 (2, 16)	3 (2, 9)	5 (2, 12)	4 (2, 13)	0.321
Baseline UPDRS II total score (mean, SD)	14.3 (7.8)	6.0 (8.8)*	1.1 (1.3)†	3.2 (6.1)	<0.001
Baseline UPDRS III total score (mean, SD)	27.4 (16.3)	17.2 (17.1)*	3.3 (5.5)†	8.2 (13.6)	<0.001
Medications
Treatment with dopaminergic medication (Levodopa/Carbidopa or Dopamine agonists)	86 (87.8%)	19 (57.6%)*	0 (0%)†	1 (2.1%)	<0.001

Symbols (*,†,‡) represent post-hoc paired comparisons of p<0.05.

^*= PDD vs. DLB+

^†= DLB+ vs. DLB−; and

^‡= DLB− vs. AD.

Abbreviations: DLB = Dementia with Lewy bodies, PDD= Parkinson’s Disease Dementia, AD = Alzheimer’s Disease, DLB+ = DLB with Parkinsonism, DLB− = DLB without parkinsonism. MMSE = Mini Mental Status Exam, UPDRS = Unified Parkinsons Disease Rating Scale.

Of 125 cases with GDS scores, 17.6% had no cognitive impairment; 30.9% scored in the mild cognitive impairment (MCI) range; and 26.5% had scores with mild dementia, 19.1% with moderate dementia and 5.9% with severe dementia. There were no differences in baseline MMSE between groups; MMSE prior to death was lower for the DLB+ group (p=0.01). At enrollment, consensus cognitive syndromic diagnosis with complete neuropsychological examination (n=100) showed 30% were cognitively normal, 25% were deemed MCI and 45% with dementia. More participants with DLB or AD tended to have a diagnosis of dementia at enrollment (p=0.05); participants with a final diagnosis of PDD were more likely to have been enrolled at the MCI stage (Table 1). Participants with AD were more likely to be heterozygous carriers of APOE Ɛ4 allele followed by DLB+; then PDD and DLB− groups (Table 1, p=0.01).

At their first movement visit participants with PDD, by definition, had a diagnosis of parkinsonism. For the DLB+ cases (n=33), 51.5% had evidence of parkinsonism at first visit and the rest developed parkinsonism during the follow up period. Of the total DLB cohort (n=48), 48.5% were diagnosed as DLB antemortem. PDD participants were more likely to be treated with dopaminergic agents in their lifetime (87.8%) compared to (p<0.001) DLB+ (57.6%), DLB− (0%) and AD (2.1%).

3.2. Longitudinal Motor Evaluation

Baseline UPDRS-II total scores were highest for PDD (14.3 ± 7.8, mean ± SD), followed by DLB+ (6.0 ± 8.8), AD (3.2 ± 6.1) and DLB− (1.1 ± 1.3). UPDRS-II scores at final movement exam were highest in the PDD group (22.5 ± 8.3), followed by DLB+ (16.9 ± 10.2), while AD and DLB− were similar (6.9 ± 8.6 and 8.0 ±7.5 respectively) (Table 1, Figure 1). UPDRS-III scores (mean ± SD) at baseline movement exam were highest for PDD (27.4±16.3), followed by DLB+ (17.2±17.1) and then AD (8.2±13.6) and DLB− (3.3±5.5). Final UPDRS-III scores (mean ± SD) were 46.9 ± 17.6 (PDD), 38.8 ± 22.5 (DLB+), 16.4 ± 17.9 (AD) and 7.9 ± 7.4 (DLB−).

Figure 1:

UPDRS-II total scores during off stage examination for each participant and their trajectories for groups: PDD (A), DLB+ (B), DLB− (C) and AD (D). Each point represents one movement exam visit and corresponding score. Blue lines depict each participant. Red line represents mixed model fit line based on fixed effects. Shaded area (funnel) surrounding mixed model fit line represents 95% confidence interval for the mixed model fitted curves. DLB = Dementia with Lewy bodies, PDD= Parkinsons’s Disease Dementia, AD = Alzheimer’s Disease; DLB+ = DLB with parkinsonism and DLB− = DLB without parkinsonism.

Figures 1 and 2 depict spaghetti plots showing the severity of UPDRS-II and UPDRS-III total scores, respectively, for each participant over time from first movement assessment; along with the best fitting non-linear mixed model regression lines surrounded by a funnel plot representing 95% confidence interval for each as overlay. The DLB+ group has steeper slopes compared to other groups for UPDRS-III. Non-linear modelling with predicted mean UPDRS-II and UPDRS-III scores for each group for a given baseline score, mean age of 76 years, and male sex is shown in Figures 3A–3D. Our modelling demonstrates that rate of change in UPDRS-II is similar between PDD and DLB+ at year 2, 4 and 6 post-first movement diagnosis, with baseline UPDRS-II score ~5 or 10 (Figure 3A–3B, Table 2). Similarly, the rate of change in UPDRS-II scores between DLB− and AD remained similar throughout and were lower than groups with parkinsonism.

Figure 2:

UPDRS-III total scores during off stage examination for each participant and their trajectories for groups: PDD (A), DLB+ (B), DLB− (C) and AD (D). Each point represents one movement exam visit and corresponding score. Blue lines depict each participant. Red line represents mixed model fit line. Shaded area (funnel) surrounding red fit line represents 95% confidence interval for the mixed model fitted curves. DLB = Dementia with Lewy bodies, PDD= Parkinsons’s Disease Dementia, AD = Alzheimer’s Disease; DLB+ = DLB with parkinsonism and DLB− = DLB without parkinsonism.

Figure 3:

Non-Linear mixed model for all groups: PDD (green), DLB+ (blue), DLB− (brown), AD (black) – with predicted mean UPDRS-II (A and B) trajectories shown based on male, age at baseline of 76 years old (mean age for participants), and estimated baseline UPDRS-II, ranging from 3–6 strata (A) and 11–14.5 strata (B). Predicted mean UPDRS-III trajectories (solid lines) based on male, age at baseline 76 years old (mean age for participants), and estimated baseline UPDRS-III, ranging from 9–17 strata (C) and 16–24 strata (D). Shaded areas (funnels between dashed lines) represent 95% confidence intervals around predicted mean point-estimate for each group’s predicted trajectory.

Table 2:

UPDRS-II and UPDRS-III summary at each time point since first movement exam and comparison between groups

Time (years)	PDD	DLB+	DLB−	AD	Cohen’s d (p-value)
Model Predicted mean UPDRS II (95% Confidence Interval)*					PDD vs. DLB+	DLB + vs. DLB−	DLB− vs. AD
2	10.2 (8.7, 11.7)	8.8 (6.9, 10.6)	5.8 (2.9, 8.7)	6.6 (4.8, 8.3)	0.30 (0.215)	0.64 (0.084)	0.17 (0.625)
4	13.2 (11.2, 15.1)	14.2 (11.4, 16.9)	7.6 (3.5, 11.8)	8 (5.7, 10.3)	0.20 (0.551)	1.47 (0.010)	0.07 (0.876)
6	16.5 (13.9, 19.1)	20.5 (16.6, 24.4)	10.7 (5.1, 16.4)	9.6 (6.5, 12.6)	0.20 (0.551)	1.68 (0.005)	0.19 (0.719)
8	20.2 (16.9, 23.6)	27.7 (22.3, 33.2)	15.1 (7.7, 22.4)	11.3 (7.4, 15.2)	1.44 (0.020)	1.71 (0.007)	0.50 (0.370)
Time (years)	PDD	DLB+	DLB−	AD	Cohen’s d (p-value)
Model Predicted mean UPDRS III (95% Confidence Interval)**					PDD vs. DLB+	DLB+ vs. DLB−	DLB− vs. AD
2	15.5 (12.4, 18.7)	25.0 (21.1, 28.8)	9.1 (3.1, 15.1)	13.7 (10.1, 17.4)	0.94 (<.001)	1.65 (<.001)	0.51 (0.177)
4	19.5 (15.6, 23.3)	33.3 (27.9, 38.6)	10.3 (2.4, 18.2)	15.9 (11.4, 20.5)	1.42 (<.001)	2.82 (<.001)	0.58 (0.215)
6	24.3 (20.0, 28.7)	41.3 (34.8, 47.9)	12.0 (2.9, 21.2)	17.6 (12.4, 22.8)	1.67 (<.001)	3.00 (<.001)	0.55 (0.287)
8	30.0 (25.3, 34.8)	49.2 (41.2, 57.2)	14.3 (4.1, 24.4)	18.8 (13.2, 24.4)	2.57 (<.001)	3.38 (<.001)	0.42 (0.439)

Abbreviations: DLB = Dementia with Lewy bodies, PDD= Parkinsons’s Disease Dementia, AD = Alzheimer’s Disease. DLB− (DLB without Parkinsonism), DLB+ (DLB with parkinsonism)

^*: The estimated means were based on male patient assuming baseline age as of 76 years, baseline UPDRSII score as of 5, education years as 15

^**: The estimated means were based on male patient assuming baseline age as of 76 years, baseline UPDRSIII score as of 10, education years as 15

In contrast, UPDRS-III scores increased rapidly and were higher at 2, 4, 6 and 8 years from first movement exam (p<0.001) in DLB+ compared to PDD with similar baseline scores (Figure 3C–3D). DLB− and AD had similar rates of change for UPDRS-III scores throughout. For any given initial baseline UPDRS-III score (~10 or 15), participants with DLB+ progress more rapidly in their motor examination severity than PDD, DLB− and AD. For DLB+ vs. PDD comparison (time*group) the estimate is 3.009 (p=0.01) in the model without MMSE while the estimate for model with MMSE is 2.786 (p=0.02) (Supplemental Table 1). As expected, DLB+ had a higher rate of increase in UPDRS-II and III scores compared to group of DLB− (Table 2). Table 2 displays Cohen’s d effect size estimates for UPDRS-II and UPDRS-III changes at years 2, 4, 6, and 8. The effect sizes continue to increase over time despite increasing variance in the data over time. Although, overall, non-linear in nature, the average annualized increase in UPDRS-II scores were as follows: 1.67 points (PDD), 3.15 points (DLB+), 1.55 points (DLB−) and 0.82 points (AD). For UPDRS-III, the annualized rates increased by 2.42 (PDD), 4.03 (DLB+), 0.87 (DLB−), and 0.85 (AD) points. As a measure of goodness of fit, the R-square (R²) is 91.2%, indicating our model accounted for 91.2% of total UPDRS-III score variation over time when both fixed and random effects were considered. It is reduced to 64.1% for fixed effects (Supplemental Table 2).

Amongst the individual subscale scores at final exam, gait was abnormal in 98.5% of PDD, 90.3% of DLB+, 60% of DLB− group and 56.8% of AD. UPDRS-III subscale scores (mean±SD) were 2.7±1.1 (PDD), 2.2±1.4 (DLB+), 1.0±1.1 (DLB−) and 1.0±1.1 (AD). Frequency (29.4% vs. 29.0%) and severity (mean±SD = 0.4±0.7 and 0.4 ± 0.7) of limb rest tremor was similar for PDD and DLB+. Rigidity was most frequent in the PDD group (88.2%), followed by DLB+ (71%), AD (20%), and none (0%) in the DLB− group. Limb bradykinesia was present in all PDD participants, 96.8% of DLB+, 20% of DLB− and 61.4% of AD cohort. Abnormal postural stability was more common in PDD (95.6%) and DLB+ (87.1%) compared to DLB− (73.3%) and AD (72.7%), with higher severity scores reported in PDD and DLB+ (Supplemental Table 3). Over 8 years, the trajectories for limb bradykinesia and gait were worse for DLB+ than PDD (Figure 4, Cohen’s d range 0.90–2.20 and 0.63–1.98 over 8 years, p<0.001 and p=0.024, respectively).

Figure 4:

Individual subscale predicted mean scores, surrounded by 95% confidence interval funnel plots, with non-linear mixed modelling as previously described for limb bradykinesia (A) and Gait (B) for cohorts of AD (black), DLB− (DLB without Parkinsonism; brown), DLB+ (DLB + parkinsonism; blue) and PDD (green). Shaded areas (funnels between dashed lines) represent 95% confidence intervals around predicted mean point-estimate for each group’s predicted trajectory (solid lines). DLB = Dementia with Lewy bodies, PDD= Parkinsons’s Disease Dementia, AD = Alzheimer’s Disease

3.3. Neuropathology:

Almost all cases with a pathologic diagnosis of DLB and 49% with PDD met the pathologic criteria for AD at autopsy (Table 1). Amongst those also meeting neuropathologic criteria for AD, participants with DLB were more likely to have higher Braak Stages IV-VI (84.8% in DLB+; 86.7% in DLB−) compared to PDD (53%) (Table 1). Participants with DLB+ were more likely to have Unified Lewy body stage-IV (84.8%) compared with PDD (63.3%) and DLB− (60%) (p<0.001). On the other hand, cases with PDD and DLB− had higher frequency of stage-III diagnosis (30.6% and 40%, respectively) compared to DLB+ (15.2%) (p<0.001).

4. DISCUSSION

In this community-based, longitudinal, clinicopathological study, motor disability progression (UPDRS-II and III scores) were significantly different between PDD, DLB+, DLB− and AD dementia without co-pathology. Adjusting for demographics; for any given baseline UPDRS-II score, the DLB+ group showed faster rates of progression in UPDRS-III (motor signs) compared to PDD, while DLB− progressed at rates similar to AD. These differences in UPDRS-III score progression were driven by differential progression in gait and limb bradykinesia sub-scores. In comparison, UPDRS-II scores progressed at similar rates over 8-year trajectories post-baseline evaluation in both PDD and DLB+. Our study provides insights into changes in motor function over time in DLB and can inform better clinical prognostication, as well as improved design of clinical trial outcomes in DLB.

Our results are consistent with prior studies showing lower baseline of severity of parkinsonism, in DLB compared to PDD [13,14]. This was reflected in a higher likelihood of treatment with dopaminergic medication in the PDD group, which may be related to a greater frequency of treatment response, longer course of motor symptoms or adverse reaction to therapy, although these were not directly measured in our study. Progression of UPDRS-III off-scores is a commonly used end point and surrogate marker for disease progression [32] in clinical trials [33,34]. In PD participants during off-states, UPDRS-III progression primarily drives changes in total UPDRS score (51%) with a predicted annualized increase of 4.02 points/year in UPDRS-III [35]. In our study, the PDD group progressed comparatively slower, with an annualized increase of ~2.4 points/year. We delineated an annualized UPDRS-III mean progression rate in DLB+ of ~4.03 points/year, consistent with findings from a prospective longitudinal study, showing an annualized rate of increase in 6.4 points [36] evaluated over 6 months. DLB has a shorter latency to onset of falls compared to PDD [37], an accelerated time to wheelchair dependence, residential care placement, and overall worse survival [38]. This may be explained by the faster UPDRS-III progression, particularly limb bradykinesia and gait impairment. A study of prodromal DLB found parkinsonian gait 3 years prior to diagnosis and motor slowing 5 years prior to conversion to dementia in ~50% of participants with DLB [39]. Future studies are required to identify and predict progression of motor subtypes of prodromal DLB patients who may benefit from early intervention [40]. Our findings of rate of progression in DLB+, therefore, may have important implications for consideration of symptomatic treatment [41], counselling patients (when they present with parkinsonism), designing end points in clinical trials [34], and monitoring response to disease modifying therapies in the future [36]. Future investigations should not only focus on identifying therapies for motor disability in DLB, but also define changes that are clinically meaningful in clinical trial cohorts of DLB.

In our modelling, gait and limb bradykinesia were drivers of the differences in increase in UPDRS-III between DLB+ and PDD. In cross-sectional studies, both PDD and DLB more commonly exhibited gait impairment, postural instability compared to PD alone [13]. Poor response to dopaminergic medications in DLB+ has been reported in the literature [14], and noted by clinicians [2,4,16]. The underlying reason for this is unknown but relatively rapid progression of non-dopamine responsive motor features may contribute. Our findings of gait related changes driving motor progression in DLB+ suggests differential weight of some extrapyramidal features. This should be considered carefully when determining important primary outcomes in future clinical trials [42] and composite scores [43]. Unlike some studies [44], severity of resting tremor score or frequency was similar between PDD and DLB+ in this study. Further work is needed to understand whether relative burdens of nigrostriatal changes and cortical changes in DLB and PDD at various disease stages may be responsible for these differences. Both DLB and PDD cases in our groups had higher frequency of neocortical Lewy body stage, consistent with the literature [45]. Development of antemortem biomarkers that correlate with topographic distribution, pathological burden and pattern of seeding of α-synuclein in-vivo might help improve understanding of mechanisms of disease progression and differential clinical expressions [28]. Data driven methods have identified different subtypes within PD ranging from mild motor impairment to malignant subtypes [46] with variable rates of motor decline. While our data does not allow for subtyping of PDD participants, it is possible that some PDD participants with a more malignant phenotype and greater overlap with DLB cannot be parsed out in this study. Baseline measures of global cognitive deterioration (MMSE) were comparable between PDD and DLB participants, although we did not assess rates of cognitive decline. Our alternate models accounted for baseline MMSE scores, but this did not change our results appreciably (Supplemental Table 1, 4, 5). Associations between increasing motor severity and cognitive decline and/or non-motor features in PDD and DLB [47,48] have been investigated previously. Future studies are warranted to assess whether motor, cognitive and neuropsychiatric symptom trajectories are comparable in this cohort; and whether relationships with burden and nature of neuropathology exist. Further, questions remain about whether clinical prognostication can be improved by delineating distinct profiles of motor, cognitive, and neuropsychiatric components, and their trajectories.

Co-pathology with AD was common in those with DLB, with a greater propensity for higher Braak stages, consistent with literature [6]. The effect of this on motor trajectories alone needs further delineation similar to cognitive decline, which is affected by dual pathology [9,48]. The DLB− was like AD in terms of motor disability and progression, as expected, and may be easily diagnosed as AD antemortem, if other phenotypes are not recognized [17]. Both DLB+ and DLB− exhibited dual pathology at autopsy, but clinical expression of parkinsonism appears to drive motor prognosis. No differences in trajectories were noted between PDD participants with and without AD co-pathology. This suggests that duration of parkinsonism, timing of clinical expression and other unknown factors, in addition to LBD and AD pathology distribution, may affect motor trajectory. However, a separate future study with sufficient power for DLB+ with and without AD, compared to PDD with and without AD would address this question comprehensively. Assessing motor characteristics in AD studies may also be of value to identify patients with mixed AD/LBD pathology who would be expected to decline faster, cognitively and globally, and could inform better design, stratification, and analyses of trial results.

Our study has multiple strengths. Foremost, the study included a large sample of longitudinal data that spanned sufficient duration (>5–6 years mean per group) to manifest substantial neuropathologic changes and motor progression. Data was from a very well-characterized community-based clinical research cohort that underwent prospective and standardized clinical, neuropathological assessments and consensus diagnosis. At enrollment, the vast majority of participants fell within a clinical spectrum that ranged from cognitively unimpaired, to MCI and mild dementia; and there was sufficient and broad range in the outcome measures (UPDRS II- and -III), with considerable room for progression from baseline and no ceiling effects, allowing for assessment of longitudinal changes across a spectrum of disease severities. The mixed effects modeling allowed for assessment of linear and non-linear trajectories, and interaction of multiple salient independent variables while adjusting for demographics and baseline differences. Sensitivity analyses showed that 89% of the variance in UPDRS-III scores were accounted for by model variables and 21% of post-baseline UPDRS-III score variance was accounted for by group-related differences, even though expected motor fluctuations in these populations likely resulted in at least moderate variability. Despite a long study duration whereby fewer participants had data in the latter years of follow-up, expanding the 95% confidence intervals around mean point-estimates, the effect sizes were statistically robust through year 8 of post-baseline follow-up.

Our study also has several limitations. We had smaller number of participants in the DLB− group (n=15) which may impact precision and external validity for this group. However, the mean age was comparable to the other groups, and they were followed until death, for ~8 years on average without developing parkinsonism, thus supporting their classification as non-parkinsonian DLB phenotype. While subjects were examined in practically-defined off-state during their movement examination, it does not completely washout any motor benefit the subject may have from dopaminergic medication. Our multivariate and sensitivity analyses adjusted for several covariates and any baseline differences in severity of motor symptoms, however, limitations exist in our ability to covary for “disease severity”. Global disease severity and functional decline may be related to neuropsychiatric symptoms, medical co-morbidities or other factors which were not accounted for in our study. Future analyses that control for other measures of disease severity, such as duration of motor symptoms or falls, dementia rating scales (CDR) as captured in the current uniform data set (UDS) are required to validate our results further. The debate of linear vs. non-linear decline in motor signs also remains, highlighting the need for future research in this field[35,49,50]. Finally, our data is from a single center with limited ethnocultural, racial, socio-economic and educational diversity [19]; not representative of diversity in the U.S. population. Future studies are also needed to understand interactions with LBD stage/density and differential contribution of co-pathologies (vascular pathology).

CONCLUSION

Refining disease progression definitions and quantifying clinical measures in LBD is an area of need. Our study provides preliminary evidence that supports differences in motor trajectories in clinicopathologically defined PDD and DLB with and without parkinsonism. Utilizing and building tools to personalize such modeling (for example, developing webtools for predicted motor trajectories that are individualized to patient characteristics) may allow clinicians to better prognosticate, counsel, anticipate and inform progression while proactively managing symptoms (prevent falls). In the research setting, such approaches may inform decisions about study design, stratification, and detection of early signals of therapeutic effect.

Highlights.

• Dementia with Lewy bodies has faster motor progression than Parkinson’s Disease Dementia

• Linear and non-linear mixed modelling analysis of longitudinal data was utilized

• Findings have implications for clinical prognostication and trial design