Comparison of dyskinesia profiles after L-DOPA dose challenges with or without dopamine agonist coadministration

Many patients with Parkinson ’ s disease (PD) experiencing L -DOPA-induced dyskinesia (LID) receive adjunct treatment with dopamine agonists, whose functional impact on LID is unknown. We set out to compare temporal and topographic profiles of abnormal involuntary movements (AIMs) after L -DOPA dose challenges including or not the dopamine agonist ropinirole. Twenty-five patients with PD and a history of dyskinesias were sequentially administered either L -DOPA alone (150% of usual morning dose) or an equipotent combination of L -DOPA and ropinirole in random order. Involuntary movements were assessed by two blinded raters prior and every 30 min after drug dosing using the Clinical Dyskinesia Rating Scale (CDRS). A sensor-recording smartphone was secured to the patients ’ abdomen during the test sessions. The two raters ’ CDRS scores were highly reliable and concordant with models of hyperkinesia presence and severity trained on accelerometer data. The dyskinesia time curves differed between treatments as the L -DOPA-ropinirole combination resulted in lower peak severity but longer duration of the AIMs compared with L -DOPA alone. At the peak of the AIMs curve (60 – 120 min), L - DOPA induced a significantly higher total hyperkinesia score, whereas in the end phase (240 – 270 min), both hyperkinesia and dystonia tended to be more severe after the L -DOPA-ropinirole combination (though reaching statistical significance only for the item, arm dystonia). Our results pave the way for the introduction of a combined L -DOPA-ropinirole challenge test in the early clinical evaluation of antidyskinetic treatments. Furthermore, we propose a machine-learning method to predict CDRS hyperkinesia severity using accelerometer data.


Introduction
Most patients with Parkinson's disease (PD) respond well to dopaminergic pharmacotherapies in the first years after diagnosis, but eventually develop motor fluctuations and dyskinesias (Jankovic, 2005). Dyskinesias are unwanted and intrusive movements that negatively affect health-related quality of life. A higher cumulative incidence of dyskinesias occurs in patients of female gender, low body weight, early-onset disease, and longer duration of L-DOPA treatment (Schrag and Quinn, 2000;Hassin-Baer et al., 2011). Dyskinesias encompass a variety of involuntary movements, including chorea, dystonia, ballism, and myoclonus, which variably affect different body parts in idiosyncratic patterns. The three main clinical presentations are defined based on their relationship to the timing of L-DOPA administration and consist of peak-dose dyskinesia, diphasic dyskinesia, and off-period dystonia (Luquin et al., 1992;Fahn, 2000). Dyskinesias are strongly associated with the use of L-DOPA (Fabbrini et al., 2007). Early use of dopamine (DA) agonists delays the onset of dyskinesias (Chondrogiorgi et al., 2014), but this benefit declines when L-DOPA therapy is added in order to achieve a better symptomatic control (Rascol et al., 2006). In a meta-analysis of previous studies, the frequency of L-DOPA-induced dyskinesia (LID) was significantly increased with adjunct DA agonist treatment in comparison to placebo (Stowe et al., 2011). Another exploratory, case-controlled study in PD patients has implicated the use of DA agonists in the development of postural dystonias (Ameghino et al., 2018). The most commonly used DA agonists for oral treatment have predominant or exclusive activity on D2-class DA receptors (Stocchi et al., 2016). For example, ropinirole and pramipexole have a strong stimulatory action on DA receptor types D2 and D3 (Kvernmo et al., 2006). Recent evidence indicates that adjunct treatment with D2/3 agonists has a significant impact on cellular adaptations underlying the development of dyskinesia (Espinosa-Cárdenas et al., 2020;Kim et al., 2023). Furthermore, a recent study in a rat model of LID shows that adding ropinirole to L-DOPA alters the responsiveness to several antidyskinetic treatment principles (Espa et al., 2023). This recent evidence indicates that the impact of DA agonist cotreatment should be taken into consideration in future preclinical and clinical studies of candidate antidyskinetic interventions.
The early phase of clinical drug development for LID often involves acute dose tests with L-DOPA and ratings of abnormal involuntary movements (AIMs) by trained medical personnel (Fox and Brotchie, 2019). Dose challenges are typically achieved using a suprathreshold oral dose of L-DOPA (Politis et al., 2014;Svenningsson et al., 2015). To the best of our knowledge, no previous study has investigated whether adjunct challenge with a DA agonist may alter some or several features of dyskinesia.
The objective of this study was to explore possible differences in AIMs phenomenology in PD patients affected by LID upon challenge test with either L-DOPA alone or a combination of L-DOPA and the D2/3 agonist ropinirole. Patients were examined for up to 5 h following oral drug intake to study time course, body distribution, and representation of dystonic versus fast hyperkinetic components in their dyskinetic motor patterns.

Study approval
The study was approved by the Swedish Ethical Review Authority (DNR: 2019-04047), further information is provided below (see paragraph "Institutional Review Board Statement").

Patients
We recruited 34 patients in total. Eight patients dropped out of the study due to severe motor impairment through the treatment washout period before the first visit, and one patient dropped out shortly after the initiation of the first visit due to feeling unwell. Twenty-five patients were included in the final analysis. The patients were selected according to the following inclusion criteria: (i) age 40-85 years; (ii) idiopathic PD with clinical records of classical, predominantly hyperkinetic LID (MDS-UPDRS part IV, item 1 ≥ 1) and/or suspected treatment-related dyskinesia/dystonia, as assessed by patient history; (iii) able to perform a withdrawal of PD medications for 12-24 h before each test day, and amantadine during a week, before and during the study.
We excluded patients according to the following criteria: (i) Hoehn & Yahr stage 5; (ii) dementia or neuropsychiatric disorder interfering with the completion of the study; (iii) ongoing treatment with continuous dopaminergic drug infusions or deep-brain stimulation; (iv) pregnancy.

Study design
This was an open-label study comparing two sequential dose challenges with L-DOPA, given alone, or combined with ropinirole according to a randomized crossover design. Immediate-release oral formulations of L-DOPA-benserazide and ropinirole were used. Challenge doses consisted of 150% of the usual L-DOPA morning dose up to a maximum of 250 mg, or a combination of equipotent parts of L-DOPA & ropinirole with the same total L-DOPA equivalent dose (LEDD), calculated according to Schade et al. (2020). The PD medication remained stable for each patient during study participation. PD patients with dyskinesias were recruited, and PD medication was withdrawn before each challenge dose as follows: (i) amantadine one week before the challenge dose; (ii) all PD medications (except for amantadine and L-DOPA) 24 h before the challenge dose; (iii) L-DOPA preparations at 10 p.m. the night before the challenge dose. The patients were asked to eat the same food categories prior to both evaluation days, and they were not allowed to eat breakfast nor drink any beverages except for water prior to the study visit or during the evaluation.
The patients were evaluated before and consecutively every 30 min after the challenge dose for up to 5 h. The evaluation consisted of approximately 2-3 min long video recordings with the patients performing the same predetermined sequence of motor tasks each time according to the Rush filming protocol (Goetz et al., 1994(Goetz et al., , 2008a. The Clinical Dyskinesia Rating Scale (CDRS) (Hagell and Widner, 1999) was used to rate AIMs severity on video recordings by 2 experienced neurologists (P.O and J.T.), blinded as to challenge test condition.
During each video recording, a sensor-recording smartphone was secured to the patient's trunk through a belt fastened around the abdomen. A mobile application recorded accelerometer and gyroscope data, which were then used to train support vector machines predicting presence and severity of trunk hyperkinesia (methods are described in the Supplemental file).
To apply the scale, patients were filmed every 30 min for up to 5 h following drug challenge, while carrying out this sequence of tasks: (i) sitting and describing a picture; (ii) drinking from a cup; (iii) putting on and off a lab coat; (iv) standing up, walking 4.5 m forward, turning by 180 • , walking back, and taking a seat again. The majority of patients required 2-4 min to complete the whole sequence. Videotapes were independently evaluated by raters with experience in movement disorders and dyskinesia assessment (authors P.O. and J.T.) after training on the CDRS method. The highest severity score observed through each video recording period was annotated for each body part (face, neck, trunk, arms, and legs). A highly significant degree of reliability was found between the two raters (intraclass correlation coefficient 0.801 ± 0.01; p < 0.001). The mean CDRS score of the two raters (for each body part and time point) was therefore used in the final statistical analysis. Only 4 patients exhibited AIMs in the face after L-DOPA and 3 patients after the L-DOPA/ropinirole challenge, therefore no meaningful statistical analysis could be performed for this body part.
The drug challenges were overall well-tolerated. Two patients who had not previously been treated with DA agonists experienced nausea at 2 time points (60 and 90 min) after the administration of L-DOPAropinirole, which led to them missing the evaluations on these time points. Patients who were receiving treatment with DA agonists did not experience any adverse effects other than dyskinesia. One patient experienced transient nausea (lasting for approx. 45 min) after the L-DOPA challenge, not leading to a disruption of the study procedures.
Rigidity and bradykinesia were assessed on-site at the same time points, by certified study personnel according to MDS-UPDRS items 3.3 & 3.6 ( Goetz et al., 2008b). The two drug challenges had a similar effect on rigidity and bradykinesia scores (see Supplemental file, Part A).

Statistical analysis
Statistical analyses were performed using IBM SPSS Statistics for Windows (Version 28.0, Armonk, NY: IBM Corp) and GraphPad Prism (version 9.5.1). Interrater reliability was evaluated using the Intraclass correlation coefficient (ICC). Two-way mixed effect ANOVA was first applied to search for differences in dyskinesia time course between the two challenge tests. This test was chosen because it provides valuable information on the interaction between time and treatment that is not readily available using non-parametric tests. Next, the related-samples Wilcoxon signed-rank test was used to examine differences in AIM scores at specific phases of the dyskinesia time curve. Comparisons of hyperkinesia and dystonia CDRS scores were carried out using the Kruskal-Wallis test followed by Dunn's multiple comparison test. The significance level was set at p < 0.05. Statistical data related to each dataset are shown in the corresponding Figure legends.

Patient characteristics
The first patient was included in October 2019 and the last patient completed the study in February 2021. Of the 25 patients included in the final analysis, 2 were on L-DOPA monotherapy, and 23 were receiving a combination of PD medications most often including DA agonists (14/25 pramipexole, 7/25 ropinirole, 1/25 rotigotine). Other adjunctive medications were entacapone (11/25); monoamine oxidase B (MAO B) inhibitors (10/25 rasagiline, 4/25 safinamide, 1/25 selegiline), and amantadine (12/25). The majority of patients (22/25) reported motor fluctuations at study inclusion, whereas 3/25 patients had only dyskinesias during the study duration (two of them, however, had experienced motor fluctuations previously). Patients' demographic and clinical data are summarized in Table 1.

Temporal profiles of the AIMs
The time course of total CDRS scores (sum of all body parts) was compared between the two challenge tests separately for hyperkinesia and dystonia (Fig. 1). This analysis revealed a significant difference in the temporal profile of hyperkinesia and dystonia scores between the two challenge conditions (p < 0.05 for time-treatment interaction) ( Fig. 1A and B). An inspection of the CDRS time curves for both hyperkinesia and dystonia showed a distinct peak of dyskinesia severity at 60-120 min followed by a sharp decline at 180-210 min upon L-DOPA (LD) administration ( Fig. 1A and B). In contrast, the CDRS curve after L-DOPA-ropinirole coadministration (LD + R) showed a blunter peak and a slower decline, with some degree of dyskinesias present through the duration of the test for patients who completed the 5-h observation period ( Fig. 1A and B).
We next examined differences between the two challenges tests on the sum of the CDRS points at 60-120 min (Peak phase, Fig. 1A', B') and 240-270 min (End phase, Fig. 1A'', B''; the 300 min data were not included in the end-phase computation because many patients did not complete the recording session through this last time point).
Peak-phase hyperkinesia scores were significantly higher after the challenge with LD compared with LD + R (Fig. 1A', median CDRS values 18 on LD vs. 13 on LD + R, p = 0.026). Although a higher grade of endphase dystonias was observed after LD + R, the difference vs. LD did not quite reach statistical significance (p = 0.06) (Fig. 1B'').

Topographic analysis of the AIMs
To appreciate the contribution of different body parts to the CDRS total score, hyperkinesia and dystonia were plotted separately for neck, trunk, arms, and legs for each challenge test ( Fig. 2A and B). In both test conditions, the extremities contributed a larger proportion of CDRS scores relative to axial regions (neck and trunk). The difference relative to neck and trunk scores reached significance for leg hyperkinesia and     (caption on next page) dystonia in the LD challenge test ( Fig. 2A), and for both leg and arm dystonia scores in the LD + R test (Fig. 2B). Dystonia scores tended to be higher in the legs than in the arms, and this difference reached significance after LD administration, whereas LD + R resulted in more comparable dystonia scores in the arms and legs (Fig. 2B). We next carried out a temporal analysis of each topographic AIMs subtype. The time course of leg CDRS scores ( Fig. 3A and B) followed the same pattern as for the total CDRS scores. Thus, compared to LD + R, the LD challenge resulted in more severe leg hyperkinesia (Fig. 3A') as well as dystonia (Fig. 3B') during peak-phase (p < 0.05). In the end phase ( Fig. 3A'', B''), there was a clear trend for LD + R to yield larger dystonia scores compared with LD, but the difference did not reach statistical significance ( Fig. 3B'').
The time course of arm CDRS scores is illustrated in Fig. 4 A, B. The time course of arm hyperkinesia differed significantly between the two challenge tests (Fig. 4A, p < 0.05 for time-treatment interaction). In the peak phase, arm hyperkinesia was generally more severe after the challenge dose with LD compared to LD + R, but the difference did not quite reach statistical significance (p = 0.051) (Fig. 4A'). Arm dystonia scores were comparable between the LD and LD + R challenge in peak phase (Fig. 4B, B') and significantly larger after LD + R challenge in end phase ( Fig. 4B'', p < 0.05 vs LD).
The results of CDRS analysis for neck and trunk are shown in Fig. 5. In these body segments, the temporal profile of dyskinesia differed significantly between LD and LD + R for trunk hyperkinesia, trunk dystonia, and neck hyperkinesia (Fig. 5 A-C; p < 0.05 for time-treatment interactions). For these items, peak CDRS scores tended to be higher after challenge with LD ( Fig. 5A'-C'). Although these CDRS items were often not detected in end-phase, there was a consistent trend for the highest scores to occur after the LD + R challenge (Fig. 5A''-C''). Neck dystonia scores were generally low and overall comparable between the LD and LD + R challenges in peak phase (Fig. 5 D, D'). Also in this case, the highest scores detected in end phase occurred after the administration of LD + R (Fig. 5D'').

A machine-learning approach to predict CDRS hyperkinesia scores
Data from the single body-worn accelerometer was used to train kernel support vector machines to predict the presence and severity of trunk hyperkinesia. We focused this analysis on the trunk hyperkinesia scores based on the sensor's location on the body (which would most likely preclude an accurate prediction of CRDS scores in other body segments). We utilized the accelerometer data generated during each video recording where the patients completed the series of predefined motor tasks, a process of approximately 2-3 min. We did not perform any analysis on dystonia as trunk dystonia observed in most patients consisted of mild fixed dystonic postures that would be difficult to detect using accelerometer/gyroscope data. The model-predicted hyperkinesia scores showed an excellent correlation with the raters' CDRS values (mean Pearson correlation coefficient 0.829; p-value <0.001) (additional data in Supplemental file, part B). Moreover, the algorithmgenerated scores faithfully reproduced the time courses of actual CDRS values (Fig. 6).

Discussion
Dyskinesias are a dose-limiting complication of DA replacement therapy in PD. They are associated with discomfort, stigma, a higher risk of disability, motor and non-motor fluctuations (Cenci et al., 2020). Experimental models of LID in rodents and non-human primates have generated a large body of concordant information about neurochemical, cellular, and circuit dysfunctions at the basis of this movement disorder (Bastide et al., 2015;Cenci and Crossman, 2018). Against this background, the pursuit of novel therapeutic approaches for LID is expected to attract continued interest on the part of both experimental and clinical investigators. Yet, in spite of the many clinical trials so far carried out in this area, very few medications have obtained regulatory approval for the treatment of LID . A key lesson learnt from the failed trials is that the endpoints used to evaluate antidyskinetic efficacy should be as similar as possible between preclinical and clinical studies (Fox and Brotchie, 2019).
Dyskinesias are a complex movement disorder encompassing both fast hyperkinetic motions and dystonic components in varying topographic patterns. Dystonia consists of either slow twisting motions or abnormal postures that can appear as an isolated feature or coexist with choreiform movements as a part of peak-dose LID (Luquin et al., 1992;Marconi et al., 1994;Fahn, 2000). Choreiform or dystonic forms of dyskinesia may not be equally distributed across body segments (Luquin et al., 1992;Fahn, 2000), and may reflect distinct network dysfunctions (Marconi et al., 1994;Skovgård et al., 2022). Recent studies in parkinsonian rodent models have highlighted that dystonic and hyperkinetic forms of LID may differentially rely on the stimulation of D2 vs D1 receptors, and that the D2 receptor mediates predominantly dystonic components (Andreoli et al., 2021). In addition, the D3 receptor has been implicated in LID pathogenesis, (Bordet et al., 1997;Guigoni et al., 2005;Visanji et al., 2009;Kumar et al., 2009;Solís et al., 2017), although its importance appears to depend on the animal model and pharmacological treatments used (reviewed in Lanza and Bishop, 2021). These and other considerations indicate that the relative engagement of different types of DA receptors can vary considerably between dyskinetic states.
To the best of our knowledge, the impact of D2/3 agonist cotreatment on the functional and phenomenological features of LID has not previously been investigated in human patients. As a first step to address this question, we have here examined the effects of ropinirole coadministration in the setting of acute L-DOPA challenge tests, a paradigm that is frequently applied to evaluate candidate antidyskinetic drugs. We chose ropinirole over the related D2/3 agonist pramipexole because, (i) it has a faster elimination half-life (T 1/2 6 h for ropinirole compared with 8-12 h for pramipexole, reviewed in Cenci et al. (2011)), making it more suitable for a 5-h acute-dosing test; (ii) experimental studies have shown that ropinirole cotreatment alters cellular and pharmacological features of LID (Espa et al., 2023), whereas equivalent studies have not yet been performed using pramipexole.
The CDRS was used to evaluate dyskinesia because it is well-aligned with the methods used to assess LID in laboratory animals, enabling differentiation between hyperkinetic and dystonic components, and evaluating each affected body part using separate scores (Hagell and Widner, 1999). CDRS proved to be a very accurate tool, with high reliability between raters and suitability to train an automatic prediction of hyperkinesia scores utilizing data from a body-worn accelerometer. Importantly, a high correspondence between model-derived and actual CDRS scores was found even at short registration periods of 2-3 min. Further testing of the model's applicability should be performed in a less controlled setting (e.g., during the patients' usual daily activities at home) and over a larger time span. Moreover, further work is needed to develop algorithms that can predict and assess the slow dystonic components of LID, which may be difficult to accurately differentiate from hypokinetic features based on accelerometer data.
Taking advantage of the CDRS methods, we here present a detailed topographic analysis of dyskinesia patterns, which is usually missing in the pharmacological literature on LID and may serve as a reference for future clinical and preclinical studies. For both types of drug challenges, the CDRS scores were overall higher in the extremities compared to axial regions (neck and trunk), a finding that was consistent in both peak and end phase. This result is somewhat expected given that arms and legs were rated on both sides of the body and the sums from both sides were used for statistical comparisons. The fact that CDRS scores from arms and legs were larger than those from neck and trunk probably facilitated the detection of treatment-related differences in the extremities. An important finding of this study is that LD resulted in more severe peak-phase dyskinesia (particularly evident in total and leg hyperkinesia scores), whereas LD + R yielded a blunter peak and a more prolonged end phase. The observed differences in dyskinesia time course may reflect the pharmacokinetic properties of the two drug challenges because the elimination half-life of L-DOPA and ropinirole are quite different (1.5-2 and 6 h, respectively) (Deleu et al., 2002). In addition, a relatively larger stimulation of D2-class receptors after the LD + R challenge would be expected to result in a more prolonged motor effect due to the slower desensitization rate of the D2-compared to the D1 receptor type (Asin et al., 1995). Regarding AIMs topography, both hyperkinesia and dystonia scores were larger in the legs than in the arms after LD, while LD + R resulted in a more comparable severity of hyperkinesia and dystonia in arms and legs. This was probably due to the more pronounced end-phase arm dystonia seen after LD + R administration. The more severe end-phase arm dystonia after LD + R administration is likely to depend on a specific effect of ropinirole. Indeed, based on pharmacokinetic considerations, the effect of a single LD dose is expected to subside after 210 min, therefore the end-phase scores after LD + R challenge can be assumed to rely on the D2/3 stimulatory action of ropinirole. Although neck and trunk CDRS scores were overall low and more variably expressed than those in the extremities, it is worth noting that the highest values recorded for these items in end-phase occurred after LD-R administration.

Considerations regarding study subjects and study design
The majority of patients included in the present study (23/25) were receiving treatment with a combination of L-DOPA and a DA agonist, and most of them received additional drugs, that is, MAO-B inhibitors (rasagiline, selegiline, safinamide, 15/25 patients), the COMT inhibitor entacapone (11/25 patients), and amantadine (12/25 patients). This reflects the current scenario of PD pharmacotherapy, where most patients in mid to advanced disease stage receive a combination of different antiparkinsonian drugs (Jagadeesan et al., 2017). Incidentally, these treatment combinations are likely to be encountered also among the patients recruited to current and future clinical trials of antidyskinetic interventions. While determining the impact of each specific drug treatment on LID would require dedicated studies, the clinical evaluations performed before and during the execution of this study did not reveal any overt difference between these drugs combinations. Moreover, a rigorous wash-out protocol was applied to minimize possible effects of these adjuvant treatments on the study outcomes. Thus, COMT and MAO-B inhibitors were washed out for 24 h prior to the drug challenge tests. Due to its longer elimination half-life of 16-17 h (Rascol et al., 2021), amantadine was washed out for 7 days. These wash-out periods are equivalent or longer than those applied in other studies using L-DOPA challenge tests (Kleedorfer et al., 1991;Svenningsson et al., 2015;Herring et al., 2017;Fox and Brotchie, 2019;Corvol et al., 2019). Regarding the COMT and MAO-B inhibitors, based on the known pharmacological properties of these drugs (Youdim and Bakhle, 2006;Habet, 2022), it is unlikely that they can impact on the relationship between D1-and D2 receptor stimulation achieved by L-DOPA. On the other hand, this relationship is most likely affected by ropinirole because of its very high binding affinity for D2/3 receptors (Kvernmo et al., 2006). A strong D2/3 stimulatory action is likely to have significant mechanistic and therapeutic implications because D2 and D3 receptors have quite different cellular distributions and signaling properties compared to D1-class receptors (Beaulieu and Gainetdinov, 2011). Accordingly, our recent studies in the rat model of LID have revealed different patterns of striatal neuroplasticity upon treatment with ropinirole and L-DOPA compared to L-DOPA alone, including a marked inhibition of angiogenic responses (Espa et al., 2023;Elabi et al., 2023) that are known to be D1 receptor-dependent (Lindgren et al., 2009). Importantly, the addition of ropinirole to L-DOPA altered the response to known antidyskinetic drug principles (Espa et al., 2023).
The side effects experienced by a few study participants do not differ from those reported in routine L-DOPA challenge tests, as carried out to evaluate antidyskinetic compounds in various studies. Nausea is not an uncommon side effect when performing a routine L-DOPA challenge test (and other forms of challenge tests, e.g., apomorphine challenge) (Saranza and Lang, 2021). This could possibly be prevented/suppressed by allowing the patients to eat a light breakfast before the study evaluations, which was though prohibited to ensure a reliable and consistent gastrointestinal absorption phase.
The main limitations of the study protocol stem from the fact that many patients were not willing to perform a medication washout prior to study visits, or they experienced severe worsening of dyskinesias after withdrawing amantadine prior to the study, and therefore chose to drop out. As a matter of fact, many patients experienced a severe off state at about 3 h after the challenge dose, and some of them could not complete the registrations through all following time points (three patients after LD and four patients after LD + R). Most patients were already on treatment with a combination of antiparkinsonian agents (including DAs) and had a PD mean duration of >10 years. Enrolling patients with shorter disease duration may have resulted in a smaller dropout rate (as they would tolerate drug withdrawal easier).
The main strengths of the study are its novelty, its prospective nature, the utilization of a dyskinesia scale that distinguishes between hyperkinetic and dystonic AIMs and its use by two experienced blinded raters with a high grade of inter-rater reliability. Furthermore, all patients included completed the medicine withdrawal according to the protocol and every patient completed both challenge tests within two weeks, ensuring a stable PD state and medication.

Concluding remarks
The present study shows the feasibility of using a ropinirole-DOPA challenge test for mechanistic or therapeutically oriented studies on LID. The combined drug challenge was overall well tolerated and yielded a distinctive dyskinesia curve. Some specific items (end-phase arm dystonia) were more strongly represented upon LD-R compared to LD administration, likely reflecting a specific effect of ropinirole in this drug challenge setting. Revealing differences between the drug challenges on other dyskinesia subscores (in particular, trunk and neck CDRS) would have required a larger number of patients. Considering the frequent use of adjunct DA agonist treatment in patients with motor complications and the concerns recently expressed about the use of high doses of L-DOPA in LID research (Chaudhuri et al., 2019), a drug challenge test combining a lower dose of L-DOPA with an immediate-release DA agonist appears to be better aligned with the clinical reality, possibly implying a larger chance to identify effective interventions in clinical trials. Studies including larger PD populations across different disease stages and using more chronic administration regimens may reveal further differences in the phenomenology of dyskinesias and response to different classes of antidyskinetic compounds. Objective, device-aided monitoring could also contribute to better monitoring and understanding of the dyskinesia patterns during different disease stages and treatment combinations. In this regard, the present study is contributing an accurate machine-learning method for a moment-to-moment prediction of hyperkinesia severity based on accelerometer-gyroscope data collected from the affected body part.

Funding
The study was supported by the following research grants to M. Angela Cenci, the strategic research area MultiPark and Lund University (funding program: SRA emerging research topics

Institutional Review Board Statement
The study was approved by the Swedish Ethical Review Authority (DNR: 2019-04047). The study was conducted according to Good Clinical Practice rules (GCP) and to the Declaration of Helsinki. Special care was taken to ensure that all participants understood that their participation was voluntary and could be ended at any time without any consequences.

Informed consent statement
Informed consent was obtained from all subjects involved in the study.

Declaration of competing interest
None.

Data availability
Data will be made available on request.