Exploring Asymmetric Fine Motor Impairment Trends in Early Parkinson's Disease via Keystroke Typing

Abstract

Background

The nQiMechPD algorithm transforms natural typing data into a numerical index that characterizes motor impairment in people with Parkinson's Disease (PwPD).

Objectives

Use nQiMechPD to compare asymmetrical progression of PD‐related impairment in dominant (D‐PD) versus non‐dominant side onset (ND‐PD) de‐novo patients.

Methods

Keystroke data were collected from 53 right‐handed participants (15 D‐PD, 13 ND‐PD, 25 controls). We apply linear mixed effects modeling to evaluate participants’ right, left, and both hands nQiMechPD relative change by group.

Results

The 6‐month nQiMechPD trajectories of right (**P = 0.002) and both (*P = 0.043) hands showed a significant difference in nQiMechPD trends between D‐PD and ND‐PD participants. Left side trends were not significantly different between these two groups (P = 0.328).

Conclusions

Significant differences between D‐PD and ND‐PD groups were observed, likely driven by contrasting dominant hand trends. Our findings suggest disease onset side may influence motor impairment progression, medication response, and functional outcomes in PwPD.

In early Parkinson's disease (PD) stages, fine motor function impairment often presents asymmetrically, ¹ leading to classification into dominant‐side onset (D‐PD) or non‐dominant side onset (ND‐PD). ² Asymmetry helps distinguish PD from other motor disorders ³ and might influence factors like first symptom, disease survival, and time to diagnosis. ² , ⁴ Despite neuroimaging evidence showing reduced dopamine uptake ⁵ and increased neurodegeneration ⁶ in the less affected side, the underlying pathophysiological mechanism remains unclear. ³

The literature conflicts on whether premorbid hand dominance influences PD symptom onset side, ⁴ , ⁷ , ⁸ , ⁹ though handedness may affect motor performance and prognosis trajectories.

Right‐handed PwPD and healthy controls (HC) studies showed the unaffected dominant hand in ND‐PD compensated more than in D‐PD during high precision tasks. ⁴ Scharoun et al ⁴ found right‐handed D‐PD increased non‐dominant hand usage, potentially switching hand preference as the disease progresses. Some research suggests D‐PD has a more debilitating effect on hand performance than ND‐PD, ¹⁰ while other studies report faster motor progression for ND‐PD ⁴ , ⁵ or D‐PD. ¹¹ , ¹² Quantifying asymmetrical motor impairment could help clinicians confirm PD diagnosis, refine prognosis, and design personalized rehabilitation plans.

Standard subjective assessments in PD have limitations, highlighting the need for sensitive, objective tools to detect fine motor performance abnormalities and characterize disease facets like motor asymmetry. ¹³ Typing involves complex, bimanual processes and cortical and subcortical brain areas, ¹⁴ , ¹⁵ making finger‐keyboard interactions ideal for evaluating fine motor performance asymmetry. The nQiMechPD score is a digital biomarker assessing fine motor skills through natural typing patterns. ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ Trained against the Unified Parkinson's Disease Rating Scale part III (UPDRS‐III), it generates a 0–1 score, where higher scores indicate more impairment. ²⁰ This work explores an alternative nQiMechPD application, focusing on longitudinal fine motor performance asymmetry trends in early PD and evaluating the influence of disease onset side relative to handedness.

Methods

The initial study design involved a cross‐sectional sample of PwPD and healthy controls. Self‐reported handedness and disease onset data were collected but not used in the study itself. The post‐hoc analysis was performed on this pre‐existing data.

A total of 53 right‐handed participants were included: 15 participants with D‐PD, 13 with ND‐PD, and 25 age‐, sex‐, and education‐matched HCs without parkinsonism or related disorders. All PwPD were de‐novo and initiated dopaminergic therapy following baseline visit. Participants were recruited from seven health institutions in Madrid, Spain, and were eligible if they self‐reported using a laptop/desktop for at least 30 min daily. Exclusion criteria included cognitive impairment (Montreal Cognitive Assessment score ≤25), ²¹ hand deformities, sleep disorders, antipsychotics/sedatives use, or other confounding diagnoses/medications. The UPDRS‐III ²² was administered to all participants.

Ambidextrous patients and those with unclear PD onset sides were excluded from analysis.

Typing Task

Participants visited a movement disorder unit five times over 6 months for motor tests, clinical evaluations, and typing tasks. Follow‐up visits occurred at weeks 4, 8, 16, and 24. During typing tasks, participants transcribed a unique, randomly selected folk tale passage at each visit to avoid learning effects. Tasks were performed on a Lenovo G50‐70 i3‐4005U with Manjaro Linux, and the nQ software captured keystroke data in the background.

Each alphanumeric character key was labeled as right‐ or left‐hand based on its physical location on the keyboard. While this labeling does not ensure strict left/right key‐hand correspondence for all individuals, it is expected that the left‐side keys are most likely to be tapped using the left hand and the right‐side keys using the right hand for the majority of individuals. Non‐alphanumeric keys were excluded from the analysis. ²⁰ Then, the nQiMechPD algorithm was applied to full keyboard, right‐side and left‐side keystroke streams for each participant on each visit.

Data Analysis

The nQiMechPD algorithm generates a right (R), left (L) and both (B) hands score on each typing task data stream. Impairment trends by study group (D‐PD, ND‐PD and HC) are estimated by fitting a linear regression model to participants' nQiMechPD change from baseline (Δ_nQiMechPD) measured over subsequent study visits. An independent model is computed for R, L and B keyboard sides’ results.

Linear mixed effects models evaluated disease onset side effect on progression trends by keyboard side, with subject ID as a random factor. Linear regression slopes were reported as standardized coefficients (β) and P‐values from z‐scores. ²³ The Mann–Whitney U significance test ²⁴ was used to compare the groups' age (in years), percentage of male participants, years from diagnosis, levodopa equivalent dose (LED: calculated using final dose achieved at the conclusion of the 6‐month follow‐up period), baseline right‐, left‐ (defined as the sum of items broken out by right or left side) and total UPDRS‐III scores, and baseline right‐, left‐, and both‐hands nQiMechPD scores. Results were considered significant when P ≤ 0.05 (two‐tailed). A second set of linear mixed effects models was developed with the relative change in LED ($\mathrm{\Delta }$LED) as a covariate to assess the impact of $\mathrm{\Delta }$LED on the nQiMechPD change trends. A post‐hoc power analysis was completed to identify the effect size and power.

Results

The comparison of the 6‐month nQiMechPD change trends by keyboard‐side and study group showed a larger decrease in nQiMechPD score in ND‐PD (−0.007) compared to D‐PD (−0.002) based on both hands’ performance regression slopes (*P = 0.043). The nQiMechPD scores for dominant (right) hand decreases in the case of ND‐PD (−0.006) but increases in the case of D‐PD (0.002) (**P = 0.002). Left side trends were not significantly different between these two groups (P = 0.328). D‐PD comparison to normative trends (HC) showed no significant difference in any keyboard side, while ND‐PD's trends were significantly different to HC's for right (**P = 0.002), left (**P = 0.001) and both (***P < 0.001) keystroke streams. The effect size of the difference in nQiMechPD scores between the D‐PD and ND‐PD groups was found to be 0.4631 (Cohen's d = 0.4588) with a power of 0.2178. The ΔLED over the follow‐up period was statistically significant only for the both‐hands model (*P = 0.042), suggesting that ΔLED may have a confounding effect on the both‐hands nQiMechPD change trends. After adjusting for ΔLED, the relationship between disease onset side and both‐hands nQiMechPD progression remained significant (*P = 0.032). ΔLED did not have a significant effect on the right‐hand nQiMechPD change trends (P = 0.238), and as the relationship between disease onset side and left‐hand nQiMechPD change trends was not significant to begin with, the relationship between relative change in ΔLED and left‐hand nQiMechPD change trends was not relevant for analysis in this case.

The Table 1 includes detailed results of the statistical comparison between the linear trends observed by study group. It also reports group‐level baseline demographic and clinical variables that discard significant differences between D‐PD and ND‐PD subpopulations on the basis of age (in years), sex, years since Parkinson's diagnosis, LED reached at the end of the observation period, and baseline motor impairment measured by either UPDRS‐III or nQiMechPD. The Figure 1 presents trend point (A) and regression relative to baseline (B) plots of right‐, left‐ and both‐hands nQiMechPD scores measured for each subgroup of participants over the study observation period.

TABLE 1.

Summary of baseline demographics and disease characteristics by group (D‐PD, ND‐PD, and HC).

	Mean (SD)			P‐value
	D‐PD (n = 15)	ND‐PD (n = 13)	HC (n = 25)	(D‐PD, ND‐PD)
				(D‐PD, HC)
				(ND‐PD, HC)
Age (years)a	63.4 (9.8)	56.1 (12.7)	60.3 (10.9)	P = 0.088
				P = 0.433
				P = 0.224
Sex (% male)a	47%	54%	48%	P = 0.931
				P = 0.997
				P = 0.943
Years from diagnosisa	1.7 (1.3)	1.3 (0.95)	‐	P = 0.369
				‐
				‐
Medication (ΔLED)a	290.7 (140.0)	347.9 (105.1)	‐	P = 0.473
				‐
				‐
UPDRS‐IIIa (T=Baseline)	19.6 (7.3)	21.0 (6.2)	1.6 (1.6)	P = 0.615
				P = 0.000
				P = 0.000
UPDRS‐III‐RIGHTa (T=Baseline)	5.7 (2.0)	2.4 (1.7)	0.1 (0.4)	*** P < 0.001
				*** P < 0.001
				*** P < 0.001
UPDRS‐III‐LEFTa (T=Baseline)	2.3 (1.4)	5.4 (1.8)	0.2 (0.4)	*** P < 0.001
				*** P < 0.001
				*** P < 0.001
nQiMechPD‐BOTHa (T=Baseline)	0.077 (0.047)	0.102 (0.061)	0.039 (0.019)	P = 0.197
				*** P < 0.001
				*** P < 0.001
nQiMechPD‐RIGHTa (T=Baseline)	0.056 (0.033)	0.100 (0.058)	0.037 (0.018)	* P < 0.05
				P = 0.105
				*** P < 0.001
nQiMechPD‐LEFTa (T=Baseline)	0.086 (0.077)	0.121 (0.092)	0.044 (0.027)	P = 0.231
				** P < 0.01
				** P < 0.01
${\mathrm{\beta }}_{{\mathrm{\Delta }}_{\mathit{\text{nQiMechPD}}-\mathit{\text{RIGHT}}}\left(T\right)}$ b	0.002 (0.002)	−0.006 (0.002)	−0.001 (0.001)	** P = 0.002
				P = 0.070
				** P = 0.002
${\mathrm{\beta }}_{{\mathrm{\Delta }}_{\mathit{\text{nQiMechPD}}-\mathit{\text{LEFT}}}\left(T\right)}$ b	−0.006 (0.002)	−0.009 (0.002)	−0.002 (0.001)	P = 0.328
				P = 0.107
				** P = 0.001
${\mathrm{\beta }}_{{\mathrm{\Delta }}_{\mathit{\text{nQiMechPD}}-\mathit{\text{BOTH}}}\left(T\right)}$ b	−0.002 (0.002)	−0.007 (0.002)	−0.001 (0.001)	* P = 0.043
				P = 0.764
				*** P < 0.001

Note: Mean and standard deviation are reported for age (in years), years from diagnosis, UPDRS‐III score, LED, baseline nQiMechPD‐BOTH, as well as the percentage of male participants. The P‐values reported are from the Mann Whitney U significance test (a) and the Linear Mixed Models (b). A P‐value with no asterisk indicates no statistical significance. Bold indicates significant value (P < 0.05).

^* P < 0.05;

^** P < 0.01;

^*** P < 0.001.

Figure 1.

(A) presents the group mean and standard deviation of the nQiMechPD score measured on each hand (right, left, both) over time. (B) presents regressed line fitted to the relative change from baseline in the nQiMechPD score measured on each hand (right, left, both) over time. Shaded areas represent the 95% confidence interval for each regression line.

Discussion

In this work, we evaluated fine motor impairment asymmetry in 28 de‐novo PwPD over 6 months of dopaminergic treatment and 25 matched controls, focusing on disease onset side impact. The nQiMechPD algorithm assessed fine motor impairment progression in right, left, and both hands. Results show significant differences in overall progression trends between D‐PD and ND‐PD groups, driven by diverging trends in the dominant hand (right). Combined with previous work, this suggests PD onset side may influence medication response and fine motor impairment progression rate in early PD. ¹⁶

The differences between the both hand trend slopes may suggest poorer response to medication in the D‐PD group (ie, β closer to 0). Moreover, the level of response showed by this group is not significantly different from the healthy control group. Clinically, this could indicate a worse clinical prognosis for D‐PD, consistent with previously published literature. ¹¹ , ¹² , ²⁵ Heinrichs‐Graham et al. used magnetoencephalography to evaluate the relationship between symptom asymmetry and the laterality of neural activity, also concluding that the ND‐PD group presented healthier oscillatory neural activity patterns than the D‐PD group.

Evaluating single‐hand fine motor performance trends in response to medication, ND‐PD exhibit a significant positive response in both the right and left hand, while D‐PD show a less pronounced response in the left hand and even a negative response (ie, worsening) in the right hand. This asymmetric progression could explain behavioral changes in D‐PD driven by unbalanced impact of the disease in activities of daily living compared to ND‐PD. In the literature, Sharoun et al. reported a switch in hand preference in right‐handed D‐PD, and Huang et al suggested ND‐PD may show better quality of life in early stages of the disease. ²⁵

These findings suggest that the relationship between side of onset and handedness might play a role in understanding disease progression and medication response in PwPD. The nQiMechPD algorithm is well‐suited to support research and improve PD care, providing detailed insights into asymmetrical fine motor progression.

The post hoc analysis we report here was designed as an experimental endpoint leveraging data collected from a study conducted to evaluate typing‐derived biomarkers as predictors of drug response in early PD. In other words, our raw data are not from a de novo study originally designed to assess disease progression compared against most affected body side. Therefore, in this naturalistic study design, dopaminergic treatment parameters should be considered a relevant and potential confounding factor when interpreting our results. Extending from this situation, additional limitations include the limited sample sizes for PwPD, and lacking data on left‐handed or ambidextrous individuals.

Future work could expand on these results with a larger, more diverse sample and use the Edinburgh Handedness Inventory ²⁶ for classification. As typing styles may vary, future research could refine the analysis by obtaining self‐reported typing styles.

In conclusion, this work proposes a novel approach for evaluating upper limb motor function asymmetry in PD using digital biomarkers. The nQiMechPD typing tool aids in fine motor impairment characterization in PD and potentially assesses asymmetry, important for diagnosis and disease progression. While this study does not provide definitive conclusions on prognosis or treatment approaches, it highlights digital biomarkers’ potential in complementing clinical standards for PD diagnosis and assessment. Further research is needed to explore clinical applications of digital biomarkers in PD and their potential to improve patient care and facilitate drug discovery.

Author Roles

(1) Research project: A. Conception, B. Organization, C. Execution; (2) Statistical Analysis: A. Design, B. Execution, C. Review and Critique; (3) Manuscript Preparation: A. Writing of the first draft, B. Review and Critique.

A.A.H.: 1A, 1B, 1C, 2A, 2B, 2C, 3A, 3B.

M.M.: 2A, 2C, 3B.

I.M‐C.: 2C, 3B.

E.K.: 3B.

R.M.: 2C, 3B.

T.A‐G.: 1A, 1B, 1C, 2A, 2B, 2C, 3A, 3B.

Disclosures

Ethical Compliance Statement: All research study protocols were approved by the Massachusetts Institute of Technology (no. 1412006804), HM Puerta del Sur University Hospital, Spain (no. 15. 05.796‐GHM), 12 de Octubre University Hospital, Spain (no. CEIC:14/090), and Clínico San Carlos University Hospital, Spain (no.14/136‐E). All study participants provided written informed consent prior to enrollment. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this work is consistent with those guidelines.

Funding Sources and Conflicts of Interest: This study was conducted with research funds from Michael J. Fox Foundation for Parkinson's Research Grant 10,860. This research was also supported by nQ Medical Inc. A.A.H., I.M‐C., and T.A‐G. are employees at nQ Medical Inc, a company that has licensed the neuroQWERTY technology, and received a regular salary while contributing to the work. E.K. is a former employee at nQ Medical Inc. and received a regular salary while contributing to the work. R.M. reports consulting fees from nQ Medical Inc.

Financial Disclosures for the Previous 12 Months: The authors declare that there are no additional disclosures to report.