The effect of 8 weeks of treatment with transcranial pulsed electromagnetic fields on hand tremor and inter-hand coherence in persons with Parkinson’s disease

Study design

The present study includes accelerometer data from a subsample of participants of a double-blinded randomized clinical trial [27, 30, 31]. In the trial, 97 participants diagnosed with idiopathic PD (Hoehn & Yahr I-IV) according to the United Kingdom Brain Bank Criteria were randomized to receive either 8 weeks of T-PEMF or placebo treatment in a 1:1 allocation ratio. Inclusion criteria for the trial were an unchanged and optimal medical treatment regarding PD six weeks prior to and during the intervention period; Mini Mental State Examination score > 22; age > 18 years; and cognitive skills enabling certification in the use of the T-PEMF device and to give informed consent. Exclusion criteria were any known neuromuscular or neurological diseases other than PD that might interfere with motor function; psychopathological treatment of other conditions than depression; substance abuse; active medical implants; pregnancy or nursing; current or previous cancer in the brain, head or neck region; leukaemia; autoimmune disease; epilepsy; and open scalp wounds. The clinical trial was registered at clinicaltrials.gov (NCT02125032), was approved by the Regional Scientific Ethical Committees for Southern Denmark (S-20130114) and the Danish Health Authority (CIV-14-01-011780), and was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent prior to participation. For further details of the trial, please refer to Additional file 1 or [27, 30, 31].

Participants

For the present study, trial participants diagnosed with idiopathic PD with at least one hand with pathological PD tremor defined by having a rest or postural tremor intensity larger than mean + 2SD of a healthy reference group were included (cut-off values, rest: 0.300 m/s², postural: 0.783 m/s²). Fifty of the 97 trial participants with PD had pathological PD tremor (hereinafter referred to as tremor). Seven of these had rest tremor only, 14 had postural tremor only, and 29 had both rest and postural tremor. Thus, 36 participants with PD (17 females, mean age (SD) of 66 (8.9) years, having a total of 52 hands with rest tremor) were included in the analysis of rest tremor, and 43 participants with PD (19 females, mean age (SD) of 65 (9.2), having a total of 64 hands with tremor) were included in the analysis of postural tremor. Endpoint data from participants with a post-interventional treatment compliance of less than 80% were excluded from the analyses. The participants with PD were evaluated by the Unified Parkinson’s Disease Rating Scale (UPDRS) for clinical disease severity characterization [32]. The UPDRS was chosen in preference to the MDS-UPDRS since the latter has not been translated to and validated in a Danish version.

Inspired by our previous finding suggesting that mildly affected persons with PD have a larger potential for neural rehabilitation than more severely affected persons with PD [27], we sub-grouped the participants with PD. Thus, the participants with PD were divided into a group with unilateral tremor and a group with bilateral tremor to investigate the effect of T-PEMF treatment on different levels of disease progression from a tremor perspective.

T-PEMF treatment

The T-PEMF treatment has been described in details elsewhere [27, 30, 31]. In short, participants with PD received one 30-min session of home-based T-PEMF or placebo treatment daily for 8 weeks. The T-PEMF device (Re5 NTS Parkinson Treatment System, Re5, Frederiksberg, Denmark) consisted of a pulse generator and a head applicator with 7 electromagnetic coils located as follow: one in the central occipital region, one in the frontal-parietal region (bilateral), and two in the anterior-temporal and posterior-temporal region (bilateral). During T-PEMF treatment, the pulse generator supplied the coils with squared bipolar pulses of ±50 V at 50 Hz and with a pulse duration of 3 ms. The stimulation intensity depends on the distance from the surface of the coil and the radial distance from the centre of the coil. At the periphery of the coil and close to the skull, maximal stimulation was estimated to be 2.5 mV/cm and decreasing with distance [34]. The subjects could not feel this very low stimulation intensity. During placebo treatment, the same treatment duration and treatment device was used but no pulsed electromagnetic fields were generated (sham stimulation). Treatment allocation was encoded to a chip card that was inserted in the pulse generator. The generator interface was identical for the two treatment types. Thus, it was not possible for the participants to see or feel difference between T-PEMF and placebo treatment, and both participants and investigators were blinded to the allocation until after endpoint assessment of the last participant. Participants with PD were optimally medicated at baseline and followed their usual medication scheme throughout the intervention. The REF group did not receive any treatment.

Protocol and accelerometer measurements

Two-dimensional cylindrical accelerometers (Catsys PD, Danish Product Development Ltd., Snekkersten, Denmark) were fixed on the hand dorsum along and between metacarpal bone II and III on each hand. The accelerometers were sensitive to accelerations in the plane orthogonal to the metacarpal bones. Hand tremor was sampled synchronously from both hands at 50 Hz.

Tremor assessment was performed before (week 0) and within one day after the last treatment session (week 8). Measures at week 0 and 8 were performed on the same time of day for each individual to reduce the influence of intra-day fluctuations. The participants with PD were tested in self-reported ON-state. The subject sat on a chair with backrest and no arm support with their feet on the ground. Hand tremor was assessed in two conditions: 1) rest, while the hands were placed with palms down approximately on the middle of the thigh in a position allowing the subject to relax and 2) postural, while the arms, hands and fingers were extended in front of the body at shoulder height, approximate shoulder width between hands, and with palms facing towards the floor (Additional file 2). PD tremor can be related to the level of attention of the patient [9]. Thus, some patients can deliberately suppress the tremor when focusing on it. Furthermore, PD tremor can be affected by stress with more pronounced tremor in a stressed state [8, 9]. We attempted to standardize the attention level of the participants without inducing stress. Thus, we asked the participants to close their eyes during the assessments to avoid distraction from the environment. Furthermore, we asked the participants to vocally perform a serial subtraction task to focus their attention on the subtraction task and not the tremor task. The use of a serial subtraction task was inspired by Lee et al. [8]. However, we modified the subtraction task and asked the participants to vocally count down from 100 in steps of two (instead of 7 or 8) in a self-paced manner (instead of as quickly as possible) to avoid stress. Finally, before each 30 s assessment, it was emphasized, that the subject should sit as calm and relaxed as possible while counting. The assessments were performed in calm surroundings and two 30-s trials of each condition were performed in the following order: rest, postural, rest, postural. Approximately 30 s pauses separated the assessments.

Data analysis

Data analyses were performed in MATLAB (Mathworks Inc., USA) using a custom-made script. The script was validated towards conventional tremor analysis software (Catsys PD software, Snekkersten, Denmark). Tremor intensity related measures and inter-hand coherence measures were calculated.

Statistics

Statistics were performed in SAS 9.4. To investigate potential treatment effects we used linear mixed models for the analysis of the intensity related measures (the log transform of tremor intensity and peak power were used for analyses to meet normal distribution). The analysis of inter-hand coherence related measures was performed by Wilcoxon statistics, as transformation of data did not induce a normal distribution. Age was not correlated to any of the effect measures among the PD group at baseline (Pearson correlation for tremor intensity related measures; Spearman correlation for coherence related measures) and was thus not included as covariate in the models. The level of significance was set to 0.05.

with unstructured covariance. Week entailed week 0 and week 8, and group entailed baseline adjusted treatment groups (T-PEMF and placebo subjects were all considered placebo subjects at week 0). In case of significant week × group interaction, pairwise comparison within group across weeks and between groups within week were performed and a 2-level Bonferroni correction was applied to adjust for multiple comparisons (corrected significance level 0.025). To determine if there was a difference in response to active and placebo treatment on coherence related measures, we used a Wilcoxon Ranked Sum test performed on the differences between week 0 and week 8. The effects of treatment on coherence related measures were evaluated both for the whole PD group and for the uni- and bilateral tremor subgroups.

Intergroup differences between PD and REF on intensity related measures were evaluated by the model dependent outcome = group, where group entailed PD and REF and measures from both hands from one subject regarded as repeated measures. To determine differences between participants with PD and REF on the inter-hand coherence measures we used a Wilcoxon Ranked Sum test. Intergroup differences in UPDRS measures were determined by t-test.

Normal distributed variables are presented by means and standard deviations (SD). Non-normal distributed variables are presented by medians and the inter quartile ranges (IQR) indicating where the middle 50% of the data lie.

Participants

Of the 36 and 43 participants with PD having rest and postural tremor respectively, three participants with both rest and postural tremor had missing data at week 8 due to withdrawal (n = 1) and exclusion from the analyses due to lag of compliance and change of medication (n = 2) (all receiving T-PEMF treatment). In addition, one participant with postural tremor only did not show up for endpoint assessment due to personal reasons (placebo). Thus, rest data from 33 subjects (19 T-PEMF and 14 placebo) and postural data from 39 subjects (20 T-PEMF and 10 placebo) were available at week 8. The treatment compliance of the participants completing the intervention was on average 98%.

As expected, the group with unilateral postural tremor was less clinically affected by the disease than the group with bilateral postural tremor as they had a significantly lower UPDRS Total score (unilateral = 41 ± 12, bilateral = 51 ± 17, p = 0.0361) and a tendency of lower UPDRS Motor score (unilateral = 24 ± 8, bilateral = 29 ± 11, p = 0.0590). Thus, the subgroup with bilateral tremor represented a clinically more severely affected group than the subgroup with unilateral tremor. The subgroups with uni- and bilateral rest tremor consisted of 19 participants (10 T-PEMF, 9 placebo) and 13 participant (9 T-PEMF, 5 placebo) respectively. The subgroups with uni- and bilateral postural tremor consisted of 20 participants (8 T-PEMF, 12 placebo) and 19 participants (12 T-PEMF, 7 placebo) respectively.

The adverse events of the treatments have been reported elsewhere. They were benign, mild and transient and the frequency did not differ between treatment groups [31].

PD vs REF

The PD group had a much larger median tremor intensity (rest: PD 1330% of REF, p < 0.0001, Fig. 1A; postural: PD 445% of REF, p < 0.0001, Fig. 1D) and median peak power (rest: PD 2336% of REF, p < 0.0001, Fig. 1B; postural: PD 770% of REF, p < 0.0001, Fig. 1E) than the REF group in both conditions. In addition, the PD group had slightly lower estimated mean peak frequency at rest (PD 85% of REF, p < 0.0001, Fig. 1C) but a slightly higher peak frequency than the REF group at the postural condition (PD 107% of REF, p = 0.0421, Fig. 1F). We found a higher median coherence (rest: PD 384% of REF, p < 0.0001, Fig. 2A; postural: PD 185% of REF, p = 0.0704, Fig. 2D), median peak coherence (rest: PD 165% of REF, p < 0.0001, Fig. 2B; postural: PD 130% of REF, p = 0.0271, Fig. 2E), and median frequency of peak coherence (postural: PD 114% of REF, p = 0.0339, Fig. 2F) in the PD group compared to REF in both conditions, except for the frequency of peak coherence at rest where no between group difference was found (p = 0.7784, Fig. 2C). When dividing into the uni- and bilateral tremor subgroups, both subgroups had significantly higher median coherence (unilateral: 0.233 ≈ 384% of REF, p < 0.0001; bilateral; 0.277 ≈ 456% of REF, p = 0.0010) and median peak coherence (unilateral: 0.620 ≈ 178% of REF, p < 0.0001; bilateral; 0.555 ≈ 159% of REF, p = 0.0011) than REF in the rest condition. However, in the postural condition only the bilateral tremor group differed from REF (coherence: 0.416 ≈ 251% of REF, p = 0.0042; peak coherence: 0.666 ≈ 46%, p = 0.0097).

Effect of the T-PEMF treatment

The statistical analysis showed no statistical difference between the effect of T-PEMF and placebo treatment on tremor intensity, peak power or peak frequency. However, main effects of time across treatment groups were found. Thus, the tremor intensity tended to decrease from week 0 to week 8 in both treatment groups in the resting (p = 0.0604, model estimated change in median for each group: T-PEMF -22%, placebo − 23%, Fig. 1A) and in the postural condition (p = 0.0585, model estimated change in median for each group: T-PEMF -3%, placebo − 19%, Fig. 1D), but no difference of improvement between groups was found. Likewise, peak power decreased significantly from week 0 to week 8 across groups (rest: p = 0.0453, model estimated change in median for each group: T-PEMF -32%, placebo − 22%, Fig. 1B; postural: p = 0.0128, model estimated change in median for each group: T-PEMF -7%, placebo − 28%, Fig. 1E), but no difference of improvement between groups was found. The peak frequency did not change (Fig. 1C & F).

With all participants included, no statistical differences in coherence, peak coherence or frequency of peak coherence were found between T-PEMF and placebo treatment (Fig. 2). However, when rerunning the analysis after sub-grouping into unilateral and bilateral tremor groups, a statistically significant treatment effect was found for the PD group with unilateral tremor in the postural condition (p = 0.0339). Thus, coherence was reduced in the T-PEMF group with unilateral postural tremor (Δcoherence from week 0 to 8, median (IQR): − 0.10 (− 0.190 to − 0.0019)) while the corresponding coherence values for the placebo group was not (Δcoherence from week 0 to 8, median (IQR): + 0.068 (− 0.052 to + 0.147)).

Effect of T-PEMF on inter-hand coherence

A major new finding was that eight weeks of treatment with T-PEMF plausibly reduced coherence relative to placebo treatment among the participants with unilateral postural tremor. Thus, T-PEMF seemed to lower the common input to movement-patterns of the limbs within the PD frequency range (3–8 Hz). Interestingly, the group having unilateral postural tremor was less clinically affected by the disease evaluated by UPDRS Total and Motor scores. This pattern of the T-PEMF treatment positively affecting a less affected group distinctively from placebo treatment was also found in the analysis of the influence of T-PEMF treatment on functional rate of force development in the clinical trial. Here, the least functionally impaired PD group benefitted from the T-PEMF treatment relative to placebo treatment by increasing their functional rate of force development, while the most functionally impaired PD group did not [27]. A recent study on repetitive transcranial magnetic stimulation in PD rats showed that functional dopaminergic neurons in substantia nigra are required to induce motor plasticity [37]. Thus, a certain level of neural functionality may need to be present to gain positive effects of neurostimulation. This emphasizes the importance of early initiation of daily treatment with T-PEMF in PD.

A possible neural explanation for the positive effect of T-PEMF in the unilateral postural tremor group (least affected) could be an increased inhibition of the inter-hemispheric crosstalk [11, 17].

Extending the treatment period could potentially induce a larger effect, and could maybe induce effects in more affected persons with PD as well. However, this requires further investigations to determine.

Effect of treatments on intensity related measures

Both T-PEMF and placebo treatment reduced peak power and tended to reduce tremor intensity in both the postural and resting condition. However, the T-PEMF group did not differ statistically from the placebo group. We attempted to induce a calm environment during the assessments, as stress is known to amplify tremor in PD [8, 9]. However, we cannot exclude the possibility that the participants experienced less stress post treatment due to the familiarity of the test situation. Furthermore, previous studies have reported significant placebo effects on PD tremor. A recent study of the effect of placebo on resting tremor in persons with PD showed a major reduction of tremor amplitude 30 min after a subcutaneous injection of saline, which they were told was apomorphine, a dopamine agonist. This effect was found in approximately half of the subjects classified as placebo responders who benefitted equally from a saline and an apomorphine injection [38]. Thus, using tremor intensity as an effect measure in clinical trials is challenging, as it is highly sensitive to other factors such as stress, level of attention, and placebo effects and thereby may mask a potential treatment effect. Thus, we cannot exclude the possibility of an effect of 8-weeks of T-PEMF treatment on tremor intensity and peak power, but we can conclude that the group effect of the T-PEMF treatment was not statistically different from the effect of the placebo treatment.

Inter-hand coherence PD vs. REF

Our results showed that inter-hand coherence and peak coherence were larger in the PD group than in the REF group. At first, this seems to be in contrast to a recent finding by Morrison et al. of no difference of peak coherence of postural hand tremor accelerations between persons with PD having bilateral tremor and healthy peers [39]. However, this discrepancy may be caused by analytic differences, as our analyses were specified to the parkinsonian tremor band (4–7 Hz extended by 1 Hz to 3–8 Hz) to focus the analysis. The study by Morrison et al. included tremor frequencies up to 40 Hz and thus included frequencies of both the parkinsonian and physiological tremor bands. Morrison et al. found that 75% of the accumulated proportional tremor power in the healthy group extended up to 10–11 Hz while it was found below 5–6 Hz in the PD group [39]. In addition, they found the mean frequency of peak coherence for the healthy group to be 8.9 Hz. Thus, when focussing on tremor in the 3–8 Hz band in our study, we by-passed most of the physiological tremor band and may not have included the frequencies of largest coherence in the REF group. However, the peak coherence values found by Morrison et al. (2012) were lower for both the PD and healthy peers (both group means around 0.2), than the values of the present study. The findings of higher peak coherence in the present study during both rest and postural conditions may be attributed to the attention distracting counting task. Performing attention distracting subtraction tasks has been shown to increase intra-limb inter-muscular coherence at rest in persons with PD [40], and to decrease the complexity (i.e. lower the sample entropy) of an EMG signal while performing an isometric contraction in healthy subjects [41]. In addition, inter-hemispheric connectivity has been shown to increase with the addition of a counting task to a unimanual tapping task, though this increase was not significant during a bimanual tapping task [42]. Together, this indicates that performing a task while distracting attention gives rise to a more synchronized motor unit activation that could plausibly occur bilaterally and thus affect inter-hand coherence.

In addition, it is an interesting finding that coherence was higher in the PD group with unilateral rest tremor than in healthy peers, which suggests a more similar frequency content of hand movements between hands during the resting condition in the PD group with unilateral tremor. This indicates that the lag of relaxation and/or control of movement are bilaterally affected despite the fact that only unilateral tremor was present.

Methodologic considerations

In participants with substantial tremor, there is a risk of mechanical transmission of tremor from leg to hand and even from one hand to the other especially during the resting condition. However, in the present randomized study we expect both treatment groups to be equally affected.