May Bradykinesia Features Aid in Distinguishing Parkinson’s Disease,Essential Tremor,And Healthy Elderly Individuals?

2.1 Participants

One hundred and ninety subjects were enrolled in the study, including 44 PD patients (17 F, mean age±standard deviation - SD: 67.84±8.69 years), 69 ET patients (31 F, mean age±SD: 68.42±11.16 years), and 77 HC (44 F, mean age±SD: 66.1±8.14 years) (Table 1). Participants were consecutively recruited at the outpatient clinic of the Department of Human Neurosciences, Sapienza University of Rome. The diagnosis of PD and ET was based on clinical criteria [2, 22]. PD patient with various phenotypes of the disease were enrolled, including patients with an akinetic-rigid form, as well as with tremor dominant phenotype [23, 24]. Pharmacological therapies possibly acting on the central nervous system were discontinued in patients. In detail in PD, dopaminergic therapy discontinuation was performed at least 24 hours before the experimental evaluation [25–28]. In ET patients, therapy discontinuation was obtained by dose reduction in the week prior to evaluation, with drug withdrawal 24 hours earlier for propranolol and benzodiazepines and 48 hours earlier for primidone and topiramate [20, 29–31]. Demographic data and clinical information were collected (Table 1). Cognitive performance was tested in participants by using the Montreal Cognitive Assessment (MoCA) [32]. The study protocol underwent review and gained approval from the local Institutional Review Board, and all participants provided their written informed consent to undergo the experimental procedures.

Participants were video recorded while repeatedly performing 15 s of opening and closing the index finger and the thumb (finger tapping) as fast as possible and with the widest range of motion [16, 33–35]. Subjects performed finger tapping by holding their arm at roughly shoulder height, with the forearm semi-flexed and the wrist in line with the forearm [20, 35]. PD patients performed the finger tapping with the most affected hand [20, 27]. In light of previous investigations indicating no significant impact of handedness on motor performance, finger tapping assessments were performed with the dominant hand by both ET and HC [20, 27]. The camera only captured the hand performing finger taps, disregarding facial and other body movements, assuring that the presence of other neurological indications, such as hypomimia or change in body posture, would not influence the assessment [16, 36]. Finger tapping videos were randomized for participants’ diagnosis, and a blinded clinical evaluation was performed offline based on video recordings. In detail, seven neurologists with expertise in movement disorders independently scored the tapping performance from 0 to 4, according to item 3.4 (finger tapping) of the Movement Disorder Society-sponsored revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) [33, 34]. As per the MDS-UPDRS protocol, raters were requested to clinically assess movement velocity (Vel^C), hesitations/ halts (Rhythm^C), and decrementing amplitude (sequence effect or Seq_Eff^C) while executing the tapping test. A rating of 0 denotes a normal motor performance, whereas a score of 1 indicates slightly impaired movements. Similarly, score 2 represents mild abnormal movements, while score 3 depicts a moderate tapping abnormality. Lastly, score 4 is indicative of a severe movement abnormality [33, 34]. Besides the overall score, we asked evaluators to distinguish, using a binary system (yes or no), if the MDS-UPDRS score assigned to a specific video was due to a) decreased movement speed, and/or b) modified rhythm, and/or c) sequence effect [16].

2.3 Kinematic recordings and analysis

We recorded the kinematics of repetitive finger tapping using a 3-D optoelectronic system (SMART motion system, BTS, Milan, Italy), consisting of three infrared cameras that detected the movement in three-dimensional space of reflective markers placed on the body segments to be analyzed. Specifically, three markers were placed on the hand and two additional markers were placed on the distal phalanx of the index finger and thumb of the dominant hand in the ET and HC groups and on the most affected hand in the PD group. The cameras had a sampling rate of 120 Hz [25, 37–40]. Motion analysis was conducted offline using specialized software (SMART Analyzer, BTS Engineering, Italy). Using an automatic algorithm, this software accurately identified several kinematic parameters, as described elsewhere [25, 37]. In line with the clinical evaluation approach, we narrowed our focus for further analysis on kinematic data to movement velocity (Vel^K), expressed in degrees/s, movement rhythm (Rhythm^K), which is determined by the coefficient of variation computed by the ratio of standard deviation to mean value of inter-tap intervals. Higher values indicate a less regular repetition of movements. We also considered the amplitude decrement (Seq_Eff^K), expressed in degrees per number of movement [25, 37]. It should be noted that the researcher responsible for the kinematic recordings and analyses (LA) was also blinded to the participants’ diagnoses.

2.4 Statistical analysis

We analyzed gender distribution and familial history differences among PD, ET, and HC participants using the Chi-square test. Additionally, the Kruskal-Wallis test was performed to investigate potential variations in demographic and clinical characteristics between the three groups, and post-hoc comparisons were conducted using Dunn’s test.

Fleiss’ K was employed to calculate the inter-raters’ consensus in the blinded video evaluation, whereby a value above 0.21 indicated a satisfactory level of agreement between raters in accordance with Landis and Koch’s (1977) criteria [41]. Kruskal-Wallis testing was utilized to compare clinical scores obtained from the blinded video evaluation of the three participant groups (mean of the 7 raters scores), and post-hoc comparisons were carried out with Dunn’s test. The frequencies of observance for each motor abnormality— Vel^C, Rhythm^C, and Seq_Eff^C— were compared between the groups through the Chi-square test. In addition, finger tapping kinematics were compared between PD, ET, and HC using one-way ANOVA, with kinematic variables analyzed in separate ANOVAs. Post-hoc analyses were performed using the Bonferroni test.

The possible correlations between clinical data, video scores, and kinematic parameters were assessed using Spearman’s rank correlation coefficient. To evaluate the overlapping distribution of the kinematic data, clinical evaluation scores were analyzed and presented through stratified density plots. Furthermore, ordinal and multinomial logistic regression analyses were conducted to determine the impact of each kinematic variable on the probability of belonging to a clinical score category (0, 1, 2, or 3-4) and a diagnostic group (PD, ET, or HC). To obtain a comprehensive assessment, a combined kinematic score (CKS) was developed by considering all the kinematic variables. The CKS was calculated by taking the linear combination of the coefficients derived from the ordinal logistic model. The resulting score accurately reflected the participant’s overall kinematic performance. CKS = - 3.1 Vel K / 1000 + 19.3 Rhythm K - 0.9 Seq _ Eff K

To ensure consistency, we normalized the CKS values to a range of 0 to 1 before proceeding, with a higher score reflecting more severe alterations. ROC curves were utilized for visualization and determination of the predictive power of each kinematic variable and the CKS for a particular diagnostic category. The area under the ROC curve (AUC) was used as a measure of discriminating capability for the model. We chose the optimal cut-off point based on the Youden index.

We set the level of significance at 0.05 and performed data analysis using R (version 4.2.2) and STATISTICA® (TIBCO Software Inc., Palo Alto, California, US).

3.1 Demographic and clinical results

The analysis revealed no clear differences in gender ratio and age between PD, ET, and HC, and MoCA scores were statistically comparable among the three groups of participants (p > 0.05). As expected, the age of disease onset was significantly higher in PD than in ET (p = 0.009), while disease duration showed an opposite trend, being longer in ET than in PD (p < 0.001). Furthermore, while 50.73% of ET cases had a positive family history of tremor, no PD patients reported a similar family history of parkinsonism or other neurological conditions (Table 1).

3.2 Video evaluation

Through our blinded finger tapping video evaluation, a fair global agreement was reached amongst raters (Fleiss K = 0.34). In particular, the Fleiss K was 0.3 for the PD group, 0.31 for ET, and 0.27 for HC. Kruskal-Wallis ANOVA indicated that there was a significant variance in the finger tapping scores between the three groups of participants [H(2)=74.28, p < 0.001]. Post-hoc comparisons showed higher scores in PD as opposed to ET and HC (p = 0.012 and < 0.001, respectively), as well as in ET as opposed to HC (p < 0.001) (Fig. 1A). Intriguingly, raters assigned an MDS-UPDRS finger tapping score of≥1 to 149 out of 190 subjects, and details on finger tapping scores for each group can be found in Fig. 1B.

OPEN IN VIEWER

Fig. 1

Blinded video evaluation results. A) Finger tapping blinded video rating scores in healthy controls (HC), patients with Parkinson’s disease (PD), and essential tremor (ET), according to the Movement Disorder Society-sponsored revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS). Horizontal lines denote median values. Asterisks indicate p < 0.05 in the post hoc comparisons. B) Number of subjects (percentage are in brackets) within the HC, ET, and PD groups who obtained specific clinical scores (ranging from 0 to 4) at the blinded video evaluations. C) Percentage of subjects within the HC, ET, and PD groups showing specific movement abnormalities at the blinded video evaluation, including reduced velocity (RV), altered rhythm (AR), and sequence effect (SE) (PD group: RV vs. AR: p = 0.58; AR vs. SE: p = 0.026; RV vs. SE: p = 0.026; ET group: RV vs. AR: p = 0.043; AR vs. SE: p = 0.19; RV vs. SE: p = 0.003; HC group: RV vs. AR: p = 0.23; AR vs. SE: p = 0.43; RV vs. SE: p = 0.42). Note that movement slowness was the most prominent abnormality and it differed between PD and ET, being much lower in HC (RV: PD vs. ET: p = 0.012, PD vs. HC: p = 0.002; ET vs. HC: p < 0.001). Irregular movement rhythm differentiates controls from patients, but not PD from ET (AR: PD vs. ET: p = 0.55, PD vs. HC: p = 0.03; ET vs. HC: p = 0.03). The SE was more prevalent in PD than in ET and HC, but it did not differ between ET and HC (SE: PD vs. ET: p < 0.001, PD vs. HC: p < 0.001; ET vs. HC: p = 0.15).

When considering the specific motor abnormalities responsible for the MDS-UPDRS finger tapping score (Fig. 1C), we observed that in PD patients, 43.8% had reduced Vel^C, 44.8% had altered Rhythm^C, and 67.2% had Seq_Eff^C. In ET patients, 60.9% had reduced Vel^C, 44.9% had altered Rhythm^C, and only 36.6% showed Seq_Eff^C (Fig. 1). In HC, we observed equal prevalence of the three motor abnormalities (reduced Vel^C - 24.48%, altered Rhythm^C- 30.98%, and Seq_Eff^C- 28.75%), with no significant difference between any of them. In summary, movement slowness, which was identified in both PD and ET patients by the raters, proved to be the most prominent abnormality in movement. It was also observed that this abnormality differed between PD and ET, in addition to being much lower in the control group (all ps < 0.01, see Fig. 1 Legend for more details). On the other hand, although an irregular movement rhythm could differentiate controls from patients, it was not able to distinguish between PD and ET (p = 0.55). The sequence effect was found to be more prevalent in individuals with PD than in those with ET and HC (both ps < 0.001). However, there was no significant difference observed between the occurrence of the sequence effect in ET and HC (p = 0.15).

The kinematic parameters of finger tapping in the three groups are shown in Table 2. As expected, the ANOVAs disclosed a significant effect of the factor ‘GROUP’ for Vel^K (F_2,187 = 26.61, p < 0.001), Rhythm^K (F_2,187 = 10.04, p < 0.001), and Seq_Eff^K (F_2,187 = 6.91, p = 0.001). Vel^K was lower in PD as compared to both ET and HC (p = 0.005 and p < 0.001, respectively), as well as in ET as compared to HC (p < 0.001). Rhythm^K was more irregular in PD than in ET and HC (p = 0.005 and p < 0.001, respectively), with no difference between ET and HC (p = 0.48). Finally, the Seq_Eff^K was higher in PD than in ET and HC (p < 0.001 and p = 0.03, respectively), with no difference between ET and HC (p = 0.43).

In summary, kinematic analysis of finger tapping revealed that subjects with PD exhibit bradykinesia, often accompanied by sequence effect an irregular rhythm, whereas the primary motor abnormality observed in subjects with ET and HC was an isolated bradykinesia.

We found a correlation between age and the median clinical score at the blinded video rating in participants (R = 0.23, p = 0.001). The analysis also showed an inverse correlation between video rating global scores and Vel^K values (r=-0.51, p < 0.001). Conversely, a direct relation was observed between video rating global scores and Rhythm^K values (r = 0.51, p < 0.001). This means that the patients who obtained the highest scores at the blinded video evaluation also showed the slower and the more irregular movement at the kinematic analysis. Notably, we did not find any correlation between clinical scores and the Seq_Eff^K (r=-0.06, p = 0.38). Also, no correlations were found between kinematic values and tremor severity in ET (r = 0.12, p = 0.34), nor between other demographic, clinical, and kinematic data.

3.5 Stratified density plots

In Fig. 2, the distribution of kinematic parameter values in the participants is presented, based on the median score (ranging from 0 to 4) given by the 7 raters. Supplementary Figures 1, 2, and 3 depict the distribution of kinematic parameter based on the clinical score given by each rater. The density plots reveal a substantial overlap among the kinematic data curves, which suggests that a specific clinical score is associated with a broad range of kinematic values. Figure 2 also illustrates the distribution of kinematic parameters among the three groups of participants (PD, ET, and HC), displaying the lowest Vel^K values in the PD group (left-hand graph), with substantial overlap observed in both ET and HC groups. Notably, the kinematic data of Rhythm^K and Seq_Eff^K demonstrated even greater overlap across all three groups (middle and right-hand graphs).

OPEN IN VIEWER

Fig. 2

Stratified density plots. Upper part: Density plots were used to evaluate the overlapping in the distribution of the kinematic data based on the median score (ranging from 0 to 4) given by the 7 raters. Note the marked overlapping between data curves, indicating that a given clinical score reflected a wide range of kinematic values in participants. Velocity is expressed as degrees/s (Vel^K). The coefficient of variation (CV), computed by the standard deviation/mean value of the inter-tap intervals, expresses movement rhythm (with higher CV values representing a lower regularity of repetitive movements) (Rhythm^K). Amplitude slope is expressed in (degrees/s)/n.mov (Seq_Eff^K). Lower part: distribution of kinematic parameters among the three groups of participants (HC: healthy controls, ET: essential tremor, PD: Parkinson’s disease). Note that the PD group had the lowest Vel^K values (left-hand graph). However, Vel^K greatly overlapped in ET and HC. The overlapping of kinematic data between the three groups was even greater for the Rhythm^K and Seq_Eff^K (middle and right-hand graph).

OPEN IN VIEWER

Fig. 3

Ordinal and multinomial logistic regression analysis results. A) The figure depicts the probability of belonging to a specific clinical score category (0, 1, 2, or 3-4) based on individual kinematic parameters, i.e., movement velocity (Vel^K), coefficient of variation (Rhythm^K), and amplitude slope (Seq_Eff^K), and on the combined kinematic score (CKS). Note that subjects with high CKS, i.e., greater than 0.6, had a high probability of being classified as bradykinetic, i.e., with Movement Disorder Society-sponsored revision of the Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) clinical scores higher than 2. The opposite occurred for subjects with low CKS score (<than 0.2), which had a high probability of being classified as normal (MDS-UPDRS clinical scores = 0). Intermediate CKS values, however, correspond to a low probability of being correctly classified as bradykinetic or not bradykinetic. B) The figure depicts the probability of belonging to a specific diagnostic group (HC, healthy controls; ET, essential tremor; PD, Parkinson’s disease) based on individual kinematic parameters, and on the CKS. Note that subjects with CKS values greater than 0.7 had a high probability of belonging to the PD group, and a very low probability of belonging to the HC group. The opposite was observed for subjects with low CKS score values (lower than 0.2), which had a high probability of being normal subjects and a very low probability of belonging to the PD group. At intermediate CKS values corresponded comparable probabilities of belonging to the PD, ET, and HC groups.

The results of the ordinal and multinomial logistic regression analyses are presented in Fig. 3, which depicts the probability of being assigned to a specific clinical score category (0, 1, 2, or 3-4, Fig. 3A) or diagnostic group (PD, ET, and HC, Fig. 3B) based on individual kinematic parameters and CKS values. Notably, individuals with high CKS scores (>0.6) were more likely to be classified as clinically bradykinetic, with MDS-UPDRS clinical scores exceeding 2 (Fig. 3A). Conversely, those with low CKS scores (<0.2) had a high probability of being clinically classified as normal, with MDS-UPDRS clinical scores of 0. Intermediate CKS values were associated with a low probability of proper bradykinetic classification (Fig. 3A). Similarly, subjects with CKS values greater than 0.7 had a high probability of being in the PD group and a very low probability of being in the HC group (Fig. 3B). Conversely, individuals with low CKS scores (less than 0.2) had a high probability of being normal and a very low probability of being in the PD group (Fig. 3B). Again, intermediate CKS values were associated with equal probabilities of belonging to the PD, ET, and HC groups, reflecting more limited possibilities for proper diagnostic classification. ROC curves (Fig. 4) demonstrated that CKS cutoffs of 0.7 and 0.2 were highly specific for differentiating PD from HC [sensitivity: 0.136 (0.052, 0.274), specificity: 0.986 (0.951, 0.998), and sensitivity: 0.247 (0.156, 0.358), specificity: 0.938 (0.877, 0.975), respectively]. It is worth noting that the area under the curve (AUC) of the Vel^K parameter was comparable to that of the CKS, highlighting the crucial role of this kinematic parameter in assessing bradykinesia.

OPEN IN VIEWER

Fig. 4

Receiver operating characteristic (ROC) curves. ROC curves were used to graphically represent and to determine the capacity of the kinematic variables and of the combined kinematic score (CKS) to evaluate the clinical score category and the diagnostic group in participants. The value of area under the ROC curve (AUC) was considered to measure for how well the model can discriminate between subjects.