Motor Symptom Variability in Parkinson's Disease: Implications for Personalized Trial Outcomes?

Abstract

Background

The Movement Disorders Society Unified Parkinson Disease Rating Scale (MDS‐UPDRS), Part III, is the gold standard for assessing motor symptoms in Parkinson's disease (PD). However, motor symptoms fluctuate significantly from day to day, potentially limiting the sensitivity of this scale for trials with short duration and crossover designs. This study investigated whether day‐to‐day variability in motor symptoms exceeds the minimal clinically important difference (MCID) in the MDS‐UPDRS, Part III.

Methods

Twenty PD participants (Hoehn & Yahr stages 1.5–3) underwent 10 weekly off‐medication assessments by one assessor on the same morning. Several determinants of day‐to‐day variability were explored.

Results

Symptom variability often exceeded the MCID for worsening and improvement. Current mental stress and fatigue did not correlate with worse scores, nor did physical activity and sleep quality in the previous week.

Conclusions

These findings suggest that day‐to‐day symptom variability impacts MDS‐UPDRS scores in smaller and shorter‐duration trials of symptomatic interventions. Continuous monitoring using wearable sensors may offer more accurate and reliable measures for evaluating PD motor symptoms in clinical studies. © 2025 The Author(s). Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

In Parkinson's disease (PD), the Movement Disorders Society Unified Parkinson Disease Rating Scale (MDS‐UPDRS), Part III, is the gold standard for assessing motor symptoms in the research context. The MDS‐UPDRS, Part III, and its minimal clinically important difference (MCID) are used to provide insight into the smallest effect on motor symptoms that clinicians would classify as important for novel symptomatic and disease‐modifying treatments. ¹

However, motor symptoms fluctuate significantly from day to day. ² Although the reported test–retest reliability of the MDS‐UPDRS is high, ³ , ⁴ its limitations have become increasingly apparent with the deployment of wearable sensors that measure motor symptom severity continuously and estimate day‐to‐day variability to exceed 10%. ⁵ , ⁶ , ⁷ High intraindividual symptom variability may thereby limit the accuracy of the current standard, namely by low frequency in‐center visits where a single image of symptoms is captured, especially in small clinical trials with fewer participants and in crossover designs. ⁸ The aim of this study was to investigate whether day‐to‐day motor symptom variability limits the power of such clinical trials on current gold‐standard scales and thereby increases the risk of type II errors. The secondary aim was to investigate the influence of the potential determinants fatigue, physical activity, sleep, and mental stress on motor symptom variability in patients with PD.

Patients and Methods

This study was performed using the baseline data of a randomized, participant‐ and outcome assessor‐blinded multiple crossover trial of a nonpharmacological intervention with potential short‐term symptomatic effects (NCT05214287). ⁹ All participants were recruited through a national PD research database, diagnosed by a movement disorders–specialized neurologist, and had Hoehn & Yahr stages (H&Y) between 1.5 and 3. Individuals with unstable PD medication or active deep brain stimulation were excluded. More detailed information on the inclusion and exclusion criteria is available elsewhere. ⁹

Twenty participants underwent 10 weekly off‐medication measurements on the same moment of the day (200 measurements in total). A wash‐out period of at least 5 days between measurements was taken into account to avoid lingering effects of study‐related interventions. Each measurement day, the MDS‐UPDRS, Part III, was conducted by one trained assessor (J.J.D.). Furthermore, participants answered questions about four possible determinants of symptom variability: physical activity and quality of sleep over the past week, and current fatigue and mental stress levels. Physical activity in the past week was quantified using the International Physical Activity Questionnaire Short‐Form, which provides an activity level in Metabolic Equivalent of Task minutes. ¹⁰ Sleep quality over the past week was self‐assessed on a four‐point ordinal scale (1 very bad to 4 very good). The amount of current perceived stress and fatigue level were scored on a 10‐point Likert scale (allowing half points), in which a lower value indicates higher stress levels and more fatigue.

This report is not a reliability study but focuses on symptom variability within individual participants and how this variability relates to the MCID. Therefore, instead of using Cohen's kappa, we calculated all potential differences (deltas) between the 10 scores within each participant, to assess the entire spectrum of variability for all participants in this study. This resulted in 90 possible combinations of MDS‐UPDRS, Part III, scores (Δ‐UPDRS) per participant. This assumes that there is no order in the MDS‐UPDRS, Part III, assessments, such as introduced by learning effects or carryover effects. To verify this assumption, chronological changes in MDS‐UPDRS, Part III, scores over all 10 consecutive measurements were explored using a regression analysis. Furthermore, to assess whether intraindividual variability was associated with symptom severity, we explored the association between intraindividual variability (based on the Δ‐UPDRS) and mean MDS‐UPDRS, Part III, sum score using a regression plot. Finally, an exploratory regression analysis was performed to evaluate whether the levodopa‐equivalent dose (LED) was associated with symptom variability. ¹¹

To account for intra‐rater variability, the Δ‐UPDRS was corrected with twice the standard error of the mean (SEM) of the sole outcome assessor to obtain a 95% confidence interval for each of the 90 calculated delta values. To obtain the SEM, the trained assessor performed the complete official MDS‐UPDRS, Part III, exam thrice, at least 1 month apart. ¹² This exam contains four cases, and the standard error per case was calculated and averaged over all four cases. Next, the percentages of MDS‐UPDRS score differences that were lower than the MCID for improvement (−3.25 points), or higher than the test's MCID for worsening (4.63 points), were calculated per participant. ¹ These percentages were averaged across all participants. The delta values show the fluctuation of symptoms in each participant individually. If the delta values exceed the MCID for improvement or worsening, these symptom fluctuations could be interpreted as clinically relevant. To evaluate whether the contextual factors were determinants of the MDS‐UPDRS, Part III, scores, linear mixed effect models were developed per contextual factor, with MDS‐UPDRS, Part III, as outcome, the contextual factor as covariate, and participants as random effects and random intercepts. Statistical analyses were designed with the help of two statisticians and were conducted using R.

Results

Participant characteristics are presented in Table 1. The SEM of the outcome assessor was 1.74. A histogram of all Δ‐UPDRS scores is shown in Figure S1. There was no significant incremental change in MDS‐UPDRS, Part III, measurements over time (β = 0.016, P = 0.88), indicating that repeated measurements did not induce a learning effect or carryover effects. However, higher MDS‐UPDRS, Part III, scores were associated with higher intraindividual variability in MDS‐UPDRS, Part III, scores (Figure S3).

TABLE 1.

Study population characteristics

Number of participants	20
Women (%)	10 (50%)
Age (mean ± SD)	62 ± 5.9
Disease duration in years (mean ± SD)	4.5 ± 2.5
Baseline MDS‐UPDRS‐score (mean ± SD)	44.7 ± 10.4
Levodopa‐equivalent dose (mean ± SD)	451.3 ± 448
Hoehn & Yahr stage (n)
1.5–2	8
2.5	5
3	7

Abbreviations: MDS‐UPDRS, Movement Disorders Society Unified Parkinson Disease Rating Scale; SD, standard deviation.

Table 2 presents the Δ‐UPDRS for worsening and improvement per participant calculated and corrected for intra‐rater variation. On group level, 29.7% of Δ‐UPDRS exceeded the MCID for improvement and 17.4% of Δ‐UPDRS exceeded the MCID for worsening. Average scores for improvement or worsening for these measurements exceeding the MCIDs were −9.3 and 10.7 points, respectively. LED was not associated with symptom variability (P = 0.78).

TABLE 2.

Percentages and average scores of worsening/improvement in MDS‐UPDRS, Part III, scores beyond the MCID‐range per participant

Participant	Levodopa‐equivalent dose	Mean MDS‐UPDRS	Worst and best MDS‐UPDRS	Worse outlier (%)	Worse outlier (average)	Improvement outlier (%)	Improvement outlier (average)
1	925	51	56; 46	37.8	9.7	48.9	−9.2
2	150	29	37; 24	17.8	10.6	40.0	−9.1
3	525	56	60; 51	2.2	9.0	8.9	−8.0
4	790	30	35;26	6.7	9.3	13.3	−8.3
5	1605	50	55; 43	11.1	10.2	26.7	−8.7
6	750	66	76; 61	35.6	12.4	48.9	−11.1
7	850	44	48; 38	4.4	10.0	15.6	−8.4
8	0	38	45; 28	33.3	11.4	51.1	−10.0
9	450	46	49; 38	11.1	10.4	15.6	−9.4
10	0	46	53; 43	4.4	11.5	20.0	−8.4
11	0	53	59; 49	11.1	9.4	22.2	−8.3
12	450	45	50; 37	24.4	10.5	33.3	−9.7
13	855	37	43; 31	8.9	10.3	22.2	−8.7
14	900	43	49; 35	13.3	11.3	33.3	−8.9
15	400	44	51; 39	11.1	10.0	20.0	−8.8
16	0	60	66; 52	22.2	10.8	33.3	−9.7
17	0	43	52; 36	20.0	11.4	33.3	−9.9
18	375	35	45; 25	46.7	13.1	55.6	−12.2
19	0	49	58; 42	26.7	11.8	37.8	−10.5
20	0	31	35; 27	0.0		13.3	−8.0
				17.4	10.7	29.7	−9.3

Abbreviations: MCID, minimal clinically important difference; MDS‐UPDRS, Movement Disorders Society Unified Parkinson Disease Rating Scale.

With regard to contextual factors, increased stress (lower scores) was not significantly associated with higher MDS‐UPDRS, Part III, scores on a 10‐point Likert scale (β = −0.55, P = 0.077). Additionally, fatigue (β = −0.38, P = 0.18), sleep quality (β = 0.38, P = 0.39), and physical activity during the last 7 days (β < −0.01, P = 0.99) were not associated with symptom severity (Figure S2).

Discussion

Clinically relevant symptom variability has been described in several studies, especially in recent studies that deploy outcome measures with continuous measurement capabilities such as smartphones and other wearable sensors. ² , ⁵ , ⁶ This is the first study to investigate the impact of day‐to‐day symptom variability in participants with PD on the gold‐standard MDS‐UPDRS motor scale. Due to the increasing evidence of symptom variability in patients with PD, and our findings that day‐to‐day symptom variability may exceed the MCID threshold, we believe that relying only on the MDS‐UPDRS, Part III, score to assess motor symptoms in small clinical trials, especially in crossover trials such as multiple n‐of‐1 trials, may not be sufficient to reliably estimate the symptomatic or disease‐modifying properties of the intervention. This is in accordance with high variability as observed in several studies that deploy wearables for continuous monitoring of symptoms, ² , ⁵ , ⁶ and in a study estimating the reliability of the MDS‐UPDRS, Part III, by leveraging yearly assessments of a large prospective observational study. ⁷ The importance of such studies is highlighted by results from conventional test–retest reliability studies of the MDS‐UPDRS, Part III, which did not exhibit such variability across two timepoints. ⁴

No contextual factors were significantly associated with the MDS‐UPDRS, Part III, scores, which is not in line with previous studies. For example, several studies demonstrate the detrimental effects of mental stress on symptom severity, especially on tremor, dyskinesia, bradykinesia, and (freezing of) gait. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ This might be due to the insensitivity of the stress scale that we used, which indeed demonstrated limited granularity (Supplementary Materials), or the fact that not acute stress but stress of the last hours to days is of highest influence. Furthermore, we could not demonstrate associations between motor symptom severity and physical activity, sleep quality, or fatigue. For physical activity and sleep, this could be due to recall bias, as both were retrospective estimates of the last 7 days, contrary to stress and fatigue, which were acute ratings on the morning of the MDS‐UPDRS, Part III, assessment. Furthermore, variation in fatigue levels on a 10‐point Likert scale was limited. Various studies show a relation between fatigue and disease duration or severity on the longer term, ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² and 75% of people with PD report worsening of motor symptoms when fatigued. ²³ Finally, evidence of the influence of sleep on motor symptoms is conflicting. Half the people with PD experience better motor performance after sleeping, also known as “sleep benefit.” ¹⁷ , ²⁴ , ²⁵ , ²⁶ However, discrepancies between objective and subjective (as used here) could have hampered this analysis. ²⁷ Finally, although the positive long‐term effects of exercise on motor symptoms are undisputed, ¹⁷ , ²⁸ , ²⁹ , ³⁰ short‐term (subacute) effects on motor symptoms are less studied. In a previous study, we reported a dose–response relation between participant‐reported physical activity levels and symptom severity (for both improvement and worsening), suggesting subacute effects on symptoms. ¹⁷ However, we note that the MDS‐UPDRS symptom variability, observed under standardized conditions and corrected for measurement errors and determinants, is of limited value to the clinic setting.

Strengths of this study include the high‐frequent off‐medication measurement, the study conducted within a time frame during which no significant disease progression is expected, the measurements being performed by one assessor, and the correction for intra‐rater variability. This study is limited by the relatively small sample size, although 200 measurements were performed in this multiple crossover design. Still, the extent of symptom variability differed significantly between individuals, limiting our ability to analyze predictors of such variability. Moreover, we did not include individuals with severe or late‐stage PD, and the 10‐point measures for sleep quality, stress, and fatigue were not gold‐standard scales to test such parameters. This study demonstrates that day‐to‐day symptom variability in PD can impact interpretation of a gold‐standard rating scale and that mental stress is among the determinants of this variability. Researchers in the field should be aware of this when initiating and selecting statistical analysis plans, especially in the case of small or crossover (including n‐of‐1) studies. The fact that variability correlated with mean MDS‐UPDRS, Part III, scores suggests that this issue would be most pronounced in studies that primarily include people with advanced PD. For larger trials with traditional (frequentist) statistical methods with between‐group comparisons, such symptom variability is unlikely to present significant challenges. Future studies could focus on what domain in the MDS‐UPDRS causes the variability and, more importantly, should work toward validating and advancing wearable sensors as co‐primary outcome measures for symptomatic or disease‐modifying trials in PD. These sensors enable monitoring of symptoms over prolonged periods, offering a more continuous and reliable measure of motor symptom severity and variability. Of note, several sensors are being developed for nonmotor symptoms of PD, which comprise an important part of the PD symptom spectrum but were not part of the current analysis. ³¹ Analyzing severity trends across the spectrum of symptoms allows for more accurate evaluation of the longitudinal impact of interventions on symptoms, at the same time increasing the power of small‐sized studies and allowing for more personalized treatments.

Author Roles

(1) Research project: A. Conception, B. Organization, C. Execution; (2) Manuscript preparation: A. Writing of the first draft, B. Review and critique.

J.J.D.: 1A, 1B, 1C, 2A, 2B

M.H.: 1C, 2A, 2B

M.M.: 1A, 1B, 2B

B.P.: 1B, 1C, 2B

Full financial disclosures of all authors for the preceeding 12 months

The authors report no other financial disclosures.

Supporting information

Data S1. Supporting Information.

Data Availability Statement

The data that support the findings of this study are openly available in The Michael J. Fox Foundation Data Repository at https://zenodo.org/communities/mjff/records?q=&l=list&p=1&s=10&sort=newest.