*5.1. Meta-Analysis*

Our meta-analysis shows that, in more than 30 years, the outcome of ET-DBS improved only modestly. Given that there was no relevant change in the studied DBS parameters, the observed improvement might possibly be attributed to improved patient selection [6], DBS targeting [7,21] and stereotactic planning [22]. Although beneficial, these developments are efficient mostly in the short-term. Provided that tremor outcomes significantly decline over time [23], with habituation of the stimulation settings accounting for more than 10% of the decrease in outcome [24], more efforts should be focused on programming. In line with this, there is evidence for alternating stimulation settings to reduce habituation. However, side-effects remain a grea<sup>t</sup> limiting factor.

From the available literature, there appears to be no substantial relationship between DBS parameters and tremor suppression or side-e ffects. Given that several follow-up reports extend up to 150 months, it raises the concern that tremor outcomes might be a ffected by either disease progression or tremor adaption. Although tremor outcomes di ffered significantly in the four time-groups (follow-up-duration quartiles), no downward trend suggesting declining outcomes was observed. As such, the observed di fferences stem more likely from patient heterogeneity, rather than being the effect of time. In line with these results, we introduced random combinations of stimulation parameters for conducting comprehensive and time-saving parameter searches in individual patients. We show that this novel programming strategy is e ffective in optimizing individual tremor control and resolving side e ffects.

Despite that DBS parameters remain relatively constant over time, the corresponding tremor outcomes show large variations. Although voltage is the only parameter to show a significant correlation with tremor suppression and dysarthria, the low value of the explained variation falls short to explain this observation (significance in this case reflects the large *N* rather than a clinical significance) These findings confirm the inconsistent results from DBS-parameter explorations reported elsewhere [25–27]. Such discrepancies sugges<sup>t</sup> that uniform parameters might not exist, and that programming should particularly address the individual anatomy [7] and tremor characteristics [25].

### The Rationale behind Conventional Stimulation Parameters

Conventional stimulation parameters have originally been extrapolated from structure–e ffect relationships to address the anatomical target and tremor characteristics [10,28]. However, they consistently appear to fall short of the mark, with room for improving clinical outcomes. Despite increasing evidence that better individualizing the DBS dose is key to maximizing symptom control [17,27], conventional parameters have remained the mainstay. However, as long as the precise spatiotemporal coordinates of the neuroanatomical substrate required for clinical benefits remain insu fficiently understood, finding the optimal DBS parameters will remain challenging.

Firstly, the complexity of stimulating neural tissue stems from multiple determinants, e.g., the interaction with di fferent neuronal elements and relative distance to the electrode [29,30], as well as the direction of propagation of the action potentials [31,32]. This is consistent with the di fficulty of modeling electric field predictions [33,34]. Secondly, this spatial component of DBS is further raveled by the ye<sup>t</sup> unknown mechanism of action. Several hypotheses have been proposed [35,36], illustrating the wide-ranging e ffects of DBS. Elucidating the precise e ffects that determine the clinical outcome will be key to refining the electric field predictions to address both the local and global dynamics of the targeted circuitopathies. Lastly, the third obstacle in finding optimal DBS parameters are the temporal adjustments of stimulation. Refining DBS to be delivered only in response to pathological biomarkers (adaptive DBS; aDBS) appears promising for both ameliorating symptoms and reducing side-e ffects [37,38].

Attempts to understand the stimulation parameters have yielded inconsistent results. Most commonly, the e ffect of varying one DBS-parameter is documented, while the remaining parameters are maintained at constant [39]. The limitation of such an approach is that the constant parameters determine the therapeutic window of the parameter of interest [16]. This generates irreproducible results due to inter-patient variability. In this study, we emphasize how insu fficiently understood the interdependence of the stimulation parameters is, i.e., unexpected stimulation parameters can be clinically meaningful only when provided with the right interaction of these three parameters. Given that (1) the stimulation substrate exhibits a highly complex, dynamic and individualized spatiotemporal fabric and that (2) individual stimulation parameters cannot be considered for titration alone, future studies should allow for mutual dynamism to be exercised between the two. Therefore, programming strategies should be as robust as possible.

### *5.2. Experimental DBS Programming*

Increasing understanding of the complexity of the stimulation substrate, as well as that of the interdependence of the stimulation parameters, has increasingly discouraged the use of conventional parameters. However, the infeasibility of clinically exploring the range of DBS parameters, to a greater extent, has precluded this transition. Testing ten random combinations of stimulation parameters results in a thorough, time-saving exploration, which allows for ET-DBS optimization in individual patients. Safety is ensured by gradually transitioning between combinations and having the patient report the emergence of side e ffects, beyond which stimulation is not further increased. Notably, the most optimal parameters are selected by patient intrinsic factors, e.g., anatomopathological substrate [40–42] and lead positioning [16,43]. The fact that significantly lower frequencies have been favored in this assumption-free trial is in line with the observed deleterious e ffects of supra-threshold frequency [14]. It is also tempting to attribute the resolution of side-e ffects in three of our four patients to this. This pilot approach raises thus grea<sup>t</sup> interest by opening the gate towards more individualized, comprehensive, and faster DBS titration.

The foremost limitations of the meta-analysis are publication bias [4] and perhaps overly enthusiastic early reports [44]. Additionally, the analysis of the relationship between DBS parameters and tremor outcomes could have benefited from objective tremor measurements (e.g., accelerometry). Nevertheless, validated tremor scales have been used [45]. Regarding the experiment, one limitation might be using shorter (between 1–2 min) wash-out periods than usual [18] in some patients. However, it has been shown that Vim-DBS for ET provides tremor suppression over seconds [2]. Additionally, the sustained tremor suppression upon follow-up excluded that the therapeutic benefit of the definitive experimental settings was confounded by phase-resetting or carryover e ffects. In addition, it would have been desirable to also provide tremor scores. However, the value of tremor scales would have been limited in this case, given that only one posture of one limb was assessed and tremor scales cannot detect subtle di fferences due to their ordinal character. Ideally, medium-term tremor reports should have been supplemented by accelerometer measurements. Another limitation might be using pre-determined electrode configurations. This was done because the electrode configurations with the largest therapeutic windows had already been determined in previous programming sessions and we wanted to limit patient fatigue to a minimum. However, it would be interesting, in a future study, to explore the role of random stimulation parameters in determining the therapeutic window. Lastly, there is no evidence that, after ten trials of random parameters, convergence is reached. However, we started with a pragmatic approach that could be tested during a regular outpatient visit. In future studies, we will explore whether testing more parameters brings further improvements.
