2.2.2. Experimental Paradigm

The experimental paradigm started with a baseline tremor measurement, using the current stimulation parameters. Tremor was assessed in various arm postures and quantified by an in-house developed accelerometer (UMCG, Groningen, The Netherlands). For each posture, the total distance amplitude (TDamp) was calculated over 5 s by a custom-written script in LabVIEW (2014 SP1) and was

used to assess tremor severity. This was done by calculating the second integral of the accelerometer signal. After the baseline measurement, the posture in which tremor was most severe, was used for the adjustment of DBS parameters.

In this posture, 10 random combinations of stimulation parameters (voltage, pulse width and frequency) were tested. These combinations were generated by a custom-written script in MATLAB (version R2014a, MathWorks, MA, USA), for each patient individually. The ranges from which the stimulation parameters were randomly extracted were as follows: voltage 1.5–4 V with intervals of 0.1 V; pulse width 60–240 μs with intervals of 10μs and frequency 60–185 Hz with intervals of 5 Hz. The script selected *at random* one voltage, one pulse width and one frequency, from a pool of 12,844 theoretical combinations. The process was repeated 10 times for every patient. For each experimental combination, a 5s accelerometer recording was performed once the effects of the previous settings disappeared. Patients also indicated whether stimulation felt better in terms of tremor control compared to baseline. If side-effects emerged, stimulation was not further increased (thus attaining a pseudorandom set of parameters). However, the transition between combinations was done systematically to maximize the chance of employing a given set of parameters. Namely, we identified the parameters that needed to be lowered or increased in the subsequent setting. Given that higher current charges are more likely to cause side-effects, the parameters were prioritized as follows: parameter requiring the greatest decrements, followed by eventual lesser decrements, least increments and greatest increments, respectively. Therefore, parameters requiring decreases were adjusted first. If the remaining parameters needed to be increased, the one requiring the smallest increase was adjusted first. If the remaining changes were to cause side-effects, increments would be stopped, and the final combination would be noted. By applying this system, the pseudo character of the random parameters was deemed by the patient safety and not by the clinician's bias.

In case the new stimulation parameters led to improved tremor control and/or less side-effects compared to the baseline settings, patients maintained these parameters. To evaluate whether the new empirical settings retained tremor control, patients were approached by telephone 6–17 weeks later. Patients indicated whether tremor control was similar, better or worse compared to baseline settings and whether side-effects had emerged.

### 2.2.3. Evaluation of the Experimental Settings

The combinations of stimulation parameters that led to the best tremor reduction were identified by the individual ratings of the patients (subjective) and the accelerometer signal (objective). The effect on tremor of the best subjective and objective random stimulation parameters was compared to that of the baseline settings. Medium-term efficacy (i.e., beyond the clinical setting) was determined by contrasting the patient-reported improvements to baseline tremor control. Next, the subjective and objective stimulation parameters were compared to baseline settings to determine whether tremor control was achieved with significantly different parameters or levels of total electrical energy delivered (TEED, [20]).
