Comparison of the analgesic effects of “superficial” and “deep” repetitive transcranial magnetic stimulation in patients with central neuropathic pain: a randomized sham-controlled multicenter international crossover study

Abstract

We directly compared the analgesic effects of “superficial” and ‘deep” repetitive transcranial magnetic stimulation (rTMS) of the primary motor cortex in patients with central neuropathic pain. Fifty-nine consecutive patients were randomly assigned to active or sham “superficial” (using a figure-of-8 [F8]-coil) or “deep” (using a Hesed [H]-coil) stimulation according to a double-blind crossover design. Each treatment period consisted of 5 daily stimulation sessions and 2 follow-up visits at 1 and 3 weeks after the last stimulation session. The primary outcome was the comparison of the mean change in average pain intensity over the course of the treatment (group × time interaction). Secondary outcomes included neuropathic symptoms (NPSI), pain interference, patient global impression of change (PGIC), anxiety, depression, and catastrophizing. In total, 51 patients participated in at least one session of both treatments. There was a significant interaction between “treatment” and “time” (F = 2.7; P = 0.0024), indicating that both figure-8 (F8-coil) and H-coil active stimulation induced significantly higher analgesic effects than sham stimulation. The analgesic effects of both types of coils had a similar magnitude but were only moderately correlated (r = 0.39, P = 0.02). The effects of F8-coil stimulation appeared earlier, whereas the effects of H-coil stimulation were delayed, but tended to last longer (up to 3 weeks) as regards to several secondary outcomes (PGIC and total NPSI score). In conclusion, “deep” and “superficial” rTMS induced analgesic effects of similar magnitude in patients with central pain, which may involve different mechanisms of action.

1. Introduction

In recent years, repetitive transcranial magnetic stimulation (rTMS) has emerged as a promising noninvasive approach for the treatment of neuropathic pain (NP).^14,20 The more effective stimulation paradigms are those targeting the primary motor cortex (M1) with conventional figure-8 coil (F8-coil) and using relatively high frequency (ie, ≥5 Hz).^5,20 It has recently been confirmed that it is possible to maintain the analgesic effects for a relatively long period (up to 6 months) in these patients.^5,33 One limitation of standard stimulation paradigms with F8-coils (ie, a frequency of 10-20 Hz and a stimulation intensity of 80%-90% of the motor threshold) is that they can only stimulate superficial cortical regions of the brain, located about 1 cm beneath the surface of the skull. Higher intensities would be required to stimulate deeper brain structures with these coils, but this could significantly increase the risk of adverse events. Technological innovations led to the development of deep rTMS as early as 2002.^35,36,39 This technology is based on a new coil design named Hesed (H)-coils, which include windings in multiple planes and induce a slower decay of the electric field with distance in comparison with standard F8-coils. This allows larger brain areas to be stimulated, reaching structures located up to 4 to 5 cm beneath the surface of the skull.^35,36 Another potential advantage of this technique is that it does not require complex neuronavigation system because the stimulation target is predefined by the manufacturer. Deep rTMS is currently widely used in psychiatry for the treatment of various conditions and has FDA clearance for the treatment of major depression and certification for use in Europe for various psychiatric conditions, such as unipolar depression, bipolar depression, negative symptoms of schizophrenia, and post-traumatic stress disorders.⁷ However, regarding chronic pain, only 2 small sham-controlled, randomized single-center trials using H-coils have been conducted^29,37 and only one study compared H-coil with conventional rTMS.³⁷

To further investigate the potential interest of “deep” rTMS for the treatment of neuropathic pain, we conducted a randomized controlled, double-blind crossover study aiming to directly compare the analgesic effects of 5 daily stimulation sessions with H-coil or F8-coil in patients with central neuropathic pain. The primary outcome measure was average daily pain intensity. Secondary outcomes included pain interference, neuropathic pain symptoms, anxiety, depression, and catastrophizing.

2. Methods

The study was approved by local ethics committee and registered in clinical trials.gov: NCT03370107. All participants provided written informed consent at inclusion, and the study conformed to the Declaration of Helsinki and Good Clinical Practice Guidelines.

2.1. Participants

Eligible participants were recruited and treated at 2 university hospitals in France (Boulogne-Billancourt, Créteil) and 1 in Norway (Oslo) by experienced teams from January 2018 to October 2021. Inclusion criteria were as follows: woman or man >18 and <80 years; pain for at least 6 months; continuous pain (at least 4 days per week) of at least moderate intensity (≥4/10 on a 0-10 numerical rating scale for pain on average on the Brief Pain Inventory)¹¹; central neuropathic pain related to stabilized multiple sclerosis, spinal cord lesion, or stroke fulfilling criteria for probable or definite neuropathic pain¹³ and scoring ≥4 of 10 on the DN4 questionnaire⁸; and stable pharmacological treatment for pain at least 1 month before inclusion.

Noninclusion criteria were as follows: any clinically significant or unstable ongoing medical or psychiatric disorder including major depression, history of substance abuse (alcohol, drugs), peripheral neuropathic pain, past treatment with rTMS, contraindications to rTMS (previous severe head trauma or neurosurgical intervention, past or current epilepsy, active brain tumor, intracranial hypertension, or implanted ferromagnetic devices, eg, cardiac pacemaker, neurostimulator, or cochlear implants), other type of pain more severe than neuropathic pain, pending litigation or work accident, any difficulty to fill out questionnaires (because of language or cognitive problems); impossibility to be followed during the time course of the study; participation in another research protocol within 1 month before the inclusion; and subjects protected by law (guardianship or tutelage measure) and pregnancy or breastfeeding (all women of childbearing age were asked to have negative pregnancy test at inclusion and use contraception).

2.2. Study design

Inclusion took place 2 to 4 weeks before randomization. Patients were assessed for inclusion and exclusion criteria and underwent a brain MRI for the determination of the position of the F8-coil stimulation with neuronavigation.

Patients were randomized after a 1-week baseline observation period during which they had to report their average daily pain intensity in a diary. First, they were assigned to one of 2 treatment groups: active F8-coil and H-coil stimulation or sham F8-coil and H-coil stimulation, according to a 2:1 ratio (2 for active and 1 for sham). Then, for each treatment group (active or sham), they were rerandomized for the order of the stimulation sessions (F8-coil or H-coil) according to a crossover design; thus, each patient received successively either active F8-coil stimulation followed by active H-coil or active H-coil followed by active F8-coil, sham F8-coil followed by sham H-coil, or sham H-coil followed by sham F8-coil. This design, which has been successfully used in a previous study,⁴ was chosen to avoid having patients receiving placebo and active stimulation in the same crossover arm, based on a previous study showing the importance of placebo timing in rTMS efficacy for pain relief.² The 2 treatment periods were separated by 5 to 6 weeks. Each treatment consisted of 5 daily stimulation sessions over 5 consecutive workdays (D1, D2, D3, D4, and D5) and was followed by 2 assessment visits 1 week (W1) and 3 weeks (W3) after the last stimulation session of each treatment period. Clinical assessments were performed at each visit, immediately before the stimulation from D1 to D5 and at W1 and W3. Thus, each patient had a total of 15 visits, including 1 screening visit and 7 visits during each treatment period (ie, from D1 to D5 and at W1 and W3).

The randomization was centralized at the Clinical Research Unit of the coordinating center (Hospital Ambroise Paré, APHP). A computer-based algorithm generated random permuted blocks stratified by study center. The result of active or sham randomization was saved on a USB flash drive or as a personalized code directly entered in the stimulator (Oslo) for the F-8 coil stimulation and a magnetic card for the H-coil stimulation, which were personalized for each patient. During stimulation session, the patient's USB drive, personalized code, or magnetic card were connected to the rTMS machine and allowed the active or sham stimulation to be automatically selected without the operators' knowledge. These procedures fully ensured double blinding.

Patients were allowed to keep their concomitant analgesic treatment throughout the study. For patients whose pain had not recovered at least partially (score ≥4/10) 6 weeks after the first treatment period, no further stimulation was conducted.

2.3. Figure-8-coil stimulation

Patients were seated in a comfortable reclining chair and kept their hands as relaxed as possible. Magnetic stimulation was applied with a MagProX100 machine (Magventure Tonika Electronik, Farum, Denmark), using a figure-8-shaped (F8) coil oriented at a tangent to the scalp, with the main phase of the induced current in the anterior–posterior direction. All patients were fitted with air plugs during rTMS. Each rTMS session consisted of 30 trains of TMS pulses delivered at 10 Hz for 10 seconds (100 pulses/train) with 20-second intertrain interval, resulting in 3000 pulses per session for a total duration of 15 minutes. The stimulation intensity was set to 80% of the resting motor threshold, which was measured as the minimal intensity of stimulation evoking an electromyographic response ≥50 µV on the first dorsal interosseus muscle of the hand contralateral to the stimulated hemisphere in at least 5 of 10 trials. For this measurement, a C-B60 figure-of-8 coil (MagVenture) was positioned over the hand motor hotspot and oriented perpendicular to the central sulcus (45° away to the hemispheric midline). To perform rTMS, a Cool-B65 A/P figure-of-8 coil (MagVenture) was used. This coil has a symmetrical design with 2 nondifferentiable sides, one for active stimulation and the other for sham stimulation. The side used for the rTMS session was determined by the personalized USB flash drive or code for each patient and connected to the rTMS machine. The sound emitted during stimulation was similar regardless of the side of the coil used. Finally, to further increase treatment blinding for the patient and the operator, the scalp tapping sensation induced by active rTMS was reproduced by applying a low-intensity electrical stimulation on the head with surface electrodes,^3,23 synchronized to the magnetic pulses during both active and placebo stimulation sessions.

All rTMS sessions were performed with a neuronavigation system (Syneika ONE, Syneika, Cesson-Sévigné, France; visor2, ANT Neuro, Hengelo, Netherlands; or TMS Navigator, Localite GmbH, St. Augustin, Germany) integrating T1-weighted magnetic resonance imaging (MRI) (1.5 T) thin slices to allow tridimensional brain reconstruction.^{1,6,17,22,27,34} As in our previous studies, the motor cortical area corresponding to the hand was stimulated, regardless of the location of neuropathic pain. We stimulated the M1 target contralateral to the painful side, or the left hemisphere in case of bilateral pain, as in previous rTMS studies in chronic pain patients.^{4,18,19,24,31–33} In all cases, the figure-of-8 coil was positioned tangentially to the target, with a postero-anterior orientation. Once the location of the stimulation target and the orientation of the coil were determined, these coordinates were included in the neuronavigation system and used for each stimulation session. Then, a robot (TMS-Robot, Axilum Robotics, Strasbourg, France or Smartmove, ANT Neuro) was used in the 2 French centers to ensure proper and reproducible positioning of the coil, optimizing contact between the coil and the scalp, and compensating for any movement of the head during the rTMS session.^16,33 The Syneika ONE neuronavigator combined with the Axilum TMS-Robot was used in Ambroise Paré Hospital, whereas the visor2 and smartmove system from ANT Neuro was used in Henri Mondor Hospital. In Oslo, a flexible arm (MagVenture) was used to position the coil.

2.4. “Deep” H-coil repetitive transcranial magnetic stimulation

H-coil rTMS was administered with the Brainsway H10-coil (Brainsway, Jerusalem, Israel) applied via a helmet placed on the head designed for optimal positioning over the primary motor cortex (M1). The coil contains 14 windings; 3 medial groups conduce current along a postero-anterior axis and 2 other groups return currents in the anterior-posterior direction. Each coil element measures 10 to 13 cm in length. H-coil rTMS was applied through the H-coil connected to a Masgtim Rapid2 stimulator (Magstim, Whitland, United Kingdom). For targeting the hand area, the coil was progressively tilted laterally until reaching the optimal position for obtaining a motor response in the contralateral hand (first interosseus). Akin to conventional rTMS, the resting motor threshold (defined as the lowest stimulation intensity producing motor-evoked potentials [≥50 µV] of in 50% of the trials) was measured at the beginning of the session by stimulating the hand primary motor area on the hemisphere contralateral to the painful side or the left hemisphere in case of bilateral pain. The active and sham coil were installed in the same system. The mode of operation (active or sham stimulation) was switched by the personalized magnetic card for each patient defined during the randomization to ensure that both the patients and the investigators were blinded. The sham coil produces a similar acoustic artefact and scalp sensation as the active coil. It induces a negligible electric field inside the brain because of its nontangential orientation on the scalp, and components cancelling the electric field will ensure that it rapidly reduces the field as a function of distance. Patients were instructed to insert earplugs to minimize adverse effects on hearing. The spatial coordinates of the stimulation were recorded with markings on a cap placed on the subjects' head to ensure placement reproducibility between sessions. Active stimulation sessions with H-coil consisted of 30 consecutive trains of 10-second pulses delivered at a frequency of 10 Hz, at 80% resting motor threshold (RMT), separated by intertrain intervals of 20 seconds, according to a procedure similar to that previously described,²⁹ aiming to mimic exactly the stimulation parameters used for conventional rTMS with F8-coil. The duration of the session was 20 minutes.

2.5. Primary outcome

The primary outcome measure was the comparison of the mean change from baseline over the course of the treatment (group × time interaction) in average pain intensity extracted from the diary (scored on a 0-10 NRS, with 0 = no pain and 10 = worst pain imaginable) between active stimulation with F8-coil, active H-coil stimulation, and sham stimulation. Baseline average pain intensity corresponded to the average of pain intensities measured the first week before the first stimulation then just before the first stimulation session (day 1). Pain intensity was further recorded immediately before each stimulation session at days 2, 3, 4, and 5 (D2-D5), then at weeks 1 (W1) and 3 (W3) after the last stimulation of each treatment period.

2.6. Secondary outcomes

Secondary judgment criteria were the comparison of the efficacy of active rTMS, active H-coil stimulation, and sham stimulation between the baseline period and 1 and 3 weeks after the end of each stimulation period. These included the following outcomes:

• (1) The neuropathic pain symptom inventory (NPSI).⁹ This questionnaire quantifies the mean intensity of 10 neuropathic symptoms and their combination into 5 distinct dimensions (burning, deep pain, paroxysmal pain, evoked pain, and paresthesia/dysesthesia) during the last 24 hours on a 11-point (0-10) numerical scale.

• (2) The proportion of responders, expressed as the proportion of patients achieving at least 30% and 50% pain relief as compared with baseline values, based on the average pain intensity from the Brief Pain Inventory (BPI).¹¹

• (3) The mean pain interference score calculated from the 7 interference items of the BPI rated from 0 (does not interfere) to 10 (complete interference) to measure the impact of pain on general activity, mood, walking ability, normal work, relations with other people, sleep, and enjoyment of life.¹¹

• (4) The Hospital Anxiety and Depression Scale (HADS)⁴⁰ including 14 items scored as anxiety and depression scores.

• (5) The Pain Catastrophizing Scale (PCS).³⁸ The PCS consists of 13 items describing the thoughts and feelings that individuals may experience when in pain (range 0-52).

• (6) The patient global impression of change (PGIC) including 7 items to evaluate the subjective improvement or deterioration (from very much improved to very much deteriorated).

Furthermore, a blinding questionnaire was used at W3 to ask the patients who completed the protocol what treatment they thought they had received and the reasons for this response.

2.7. Safety

The occurrence of any adverse event was recorded over the entire study period, as well as any cause of treatment discontinuation. The severity of adverse events was rated as mild, moderate, or severe. Severe adverse events were those necessitating close monitoring, medical intervention, or hospitalization, with serious health impairment leading to prolonged disability or death.

2.8. Statistical analyses

The sample size calculation was based on the unique study of H-coil in neuropathic pain that was available when we designed the present study.²⁹ This study that randomized 23 patients was positive on the primary outcome and showed 13% to 54% efficacy of active H-coil vs sham depending on the time visit, with a median improvement of 30%. In our previous study using the same experimental design and comparing rTMS to direct current electrical stimulation (tDCS) and to sham stimulation,⁴ we found a 26% difference in efficacy between rTMS and sham. We thus calculated that we would need 50 patients to complete the protocol for the study to have a power of 90% to detect (2-tailed test with an alpha risk of 0.05) a mean difference in pain intensity of 26% between sham and active treatments. Assuming an estimated dropout rate of 20%, we estimated that 60 patients would have to be enrolled.

The statistical analyses were conducted in all randomized patients who received at least one stimulation, corresponding to a modified intention-to-treat (ITT) population subjects. All longitudinal continuous efficacy measures, that is, the primary outcome (changes in average pain intensity), and all secondary efficacy variables (ie, scores for the NPSI, BPI, HADS, and PCS) were analyzed with a repeated-measures analysis of variance with a mixed-effects model. The model included the following as explaining factors: treatment (F8-coil stimulation, H-coil stimulation, and sham stimulation), time, the interaction of treatment with time, and the sequence of stimulation (F8-coil–Hcoil or Hcoil–F8-coil). The patient was the random effect. If a significant interaction between treatment and time was observed, pairwise comparisons were made between H-coil, F8-coil, and sham stimulation at the various time points, by t tests calculated from the model. The baseline observation carried forward method was used for missing data (<5% of the sample). Pearson correlation test with Bonferroni correction for multiple comparisons was used to assess the correlation between the analgesic effects of each treatment and HADS, BPI interference, PCS, DN4, and NPSI scores at baseline. Fisher exact test was used to compare proportions. In all cases, P values <0.05 was considered significant.

3. Results

3.1. Patients

In total, 131 patients with central neuropathic pain were screened for participation in this trial; 72 patients were excluded because of contraindications, psychiatric or somatic comorbidities, refusal, and litigation (see Fig. 1 for Consort flowchart). The 59 patients meeting the inclusion criteria were randomly assigned to the active stimulation (n = 39) or sham stimulation (n = 20) groups, but 2 patients (1 in the sham group and 1 in the active group) withdrew from the trial before any stimulation was performed and were not included in the modified ITT population. Of the 57 patients who received the first rTMS treatment, 6 withdrew before the second rTMS treatment. Thus, 51 patients participated in at least 1 session of both treatments.

Figure 1.

Study flowchart.

Sociodemographic and clinical characteristics, including sex, pain intensity, neuropathic symptoms, symptoms of anxiety/depression, and catastrophizing, did not differ between groups (Table 1). In addition, average pain intensity was similar before both treatment sessions (6.2 ± 1.5 and 6.1 ± 1.7).

Table 1.

Demographic and clinical characteristics of the patients at baseline.

	Active stimulation	Sham stimulation
	n = 59	n = 20
Age (y ± SD)	54.9 ± 11.8	52.6 ± 9.8
Sex, women, n (%)	34 (58)	11 (55)
Pain condition, n (%)
Central poststroke pain	12 (30)	4 (20)
Spinal cord injury	25 (62.5)	16 (80)
Multiple sclerosis	2 (7.5)	0
Maximal pain area, n (%)
Lower limbs	21 (53)	11 (55)
Upper limbs	6 (15)	5 (25)
Hemibody (without the face)	8 (20)	2 (10)
Other (eg, face, neck, and trunk)	4 (10)	2 (10)
Average pain intensity ± SD	6.2 ± 1.6	6.2 ± 1.4
Pain duration (y ± SD)	7.3 ± 7.0	8.5 ± 6.5
BPI interference score ± SD	26.7 ± 12.0	29.6+ ± 11.3
NPSI total score ± SD	35.1 ± 12.1	34.2 ± 18.9
Anxiety score (HADS) ± SD	7.4 ± 3.6	7.0 ± 4.1
Depression score (HADS) ± SD	7.3 ± 5.2	7.4 ± 4.7
Catastrophizing score (PCS) ± SD	20.9 ± 12.2	19.1 ± 11.6
Concomitant analgesic treatments, n (%)
None	27 (69)	15 (75)
Antidepressants	4 (10)	2 (10)
Antiepileptics	5 (13)	2 (10)
Opioids	2 (5)	1 (5)
Others	4 (10)	2 (10)

BPI, brief pain inventory; HADS, hospital anxiety and depression scale; NPSI, neuropathic pain symptom inventory; PCS, pain catastrophizing scale.

3.2. Comparison of the efficacy of figure-8-coil, H-coil repetitive transcranial magnetic stimulation and sham repetitive transcranial magnetic stimulation

Because there was no difference between the analgesic effects of either sham F8-Coil and sham H-Coil on average pain intensity or proportion of responders (data not shown), both sham stimulations were grouped together to form a unique sham group. The comparison of the changes in pain intensity between F8-coil, H-coil, and sham stimulation showed a significant effect of “time” (F = 12.6; P < 0.001) with a significant interaction between “treatment” and “time” (F = 2.7; P = 0.0024), indicating that both F8-coil and H-coil induced significantly higher analgesic effects than sham stimulation over the course of the treatment. Pairwise comparisons showed that F8-coil had significantly higher efficacy (P = 0.018) as compared with sham stimulation at D5 (ie, after the fourth stimulation session) and that H-coil stimulation induced significantly greater effects (P = 0.019) than sham stimulation at W1 (ie, 1 week after the fifth stimulation session) (Fig. 2A).

Figure 2.

(A) Effects at the different time points of active F8-coil rTMS (blue), H coil rTMS (red), and sham rTMS (green) on the primary outcome (average pain intensity, expressed as a mean difference from baseline in NRS [numerical rating scale]). *P < 0.05 for the comparisons with sham stimulation. (B) Correlation (r = 0.39, P = 0.02) between the overall effects of F-8 coil rTMS and H-coil rTMS on average pain intensity from D2 to W3 (expressed as a mean difference from baseline). rTMS, repetitive transcranial magnetic stimulation.

There was a moderate correlation (r = 0.39, P = 0.02) between the mean analgesic effects of F8-coil and H-coil over the course of the treatment (Fig. 2B).

We observed no correlation between baseline clinical characteristics (ie, sex, location of the lesion [in the spinal cord or brain], and pain area [lower limbs, upper limbs, or hemibody]), average pain intensity at baseline, pain duration, or total scores for DN4, NPSI, HADS, BPI interference, and PCS at baseline and the effects of F8-coil, H-coil, or sham stimulation (data not shown).

3.3. Comparison of the effects of treatments on secondary outcome measures

The proportion of responders with at least 50% reduction of pain intensity at D5 was significantly higher (P < 0.001) in the F8-coil stimulation group (%) in comparison with sham groups (Fig. 3A), whereas the proportion of responders was significantly higher (P < 0.001) in the H-coil group at W1 (Fig. 3A).

Figure 3.

(A) Percentage (%) of responders (50% pain relief) to F8-coil rTMS (blue columns), H-coil rTMS (red columns), and sham-rTMS (green columns) from D2 (after the first rTMS session) to W3 (3 weeks after the last session). ***P < 0.001 for the comparisons with sham stimulation. (B) Comparisons of Patient Global Impression of Change (PGIC), expressed as percentages of patients much or very much improved, at D5 (last stimulation session), W1 (1 week after the last session), and W3 (3 weeks after the last session) for F8-coil rTMS (blue columns), H-coil rTMS (red columns), and sham-rTMS (green columns). *P < 0.05; **P < 0.01 for the comparisons with sham stimulation.

The proportion of patients reporting an improvement (ie, much or very much improved according to the PGIC) was significantly higher in comparison with sham stimulation at D5 in patients treated with F8-coil stimulation and at W1 and W3 for patients treated with H-coil stimulation (Fig. 3B).

Regarding other secondary outcomes (Table 2), H-coil, but not F8-coil, stimulation significantly improved the total NPSI score as compared with sham stimulation at W1 and W3, but the changes in the NPSI subscores (symptoms or dimensions) were not different after active or sham treatments. There was no difference between F8-coil, H-coil, and sham stimulation on anxiety or depression scores, pain catastrophizing, and BPI interference (Table 2).

Table 2.

Changes from baseline to week 1 and week 3 after sham, F8-coil, and H-coil stimulation, in neuropathic pain symptom inventory total score and subscores (burning pain, deep pain, paroxysmal pain, evoked pain, and paresthesia/dysesthesia), pain interference assessed with the brief pain inventory, anxiety and depression assessed with the HAD, and catastrophizing assessed with the pain catastrophizing scale.

	Sham stimulation			F8-coil stimulation			H-coil stimulation
	Baseline	Week 1	Week 3	Baseline	Week 1	Week 3	Baseline	Week 1	Week 3
NPSI total score	34.2 ± 18.9	31.5 ± 15.1	30.3 ± 19.6	33.9 ± 13.2	27.0 ± 16.4	29.3 ± 16.6	32.5 ± 15.4	23.9 ± 13.2*	24.4 ± 12.9*
Burning pain	4.5 ± 3.6	4.1 ± 3.6	4.2 ± 3.8	4.2 ± 3.5	3.2 ± 2.9	3.6 ± 3.1	3.9 ± 3.0	3.3 ± 3.1	3.5 ± 3.2
Deep pain	4.3 ± 2.5	3.8 ± 2.9	3.9 ± 2.5	3.9 ± 3.1	3.0 ± 2.7	3.5 ± 3.1	3.7 ± 3.0	3.2 ± 3.1	3.2 ± 3.0
Paroxysmal pain	3.2 ± 3.2	2.9 ± 3.5	3.0 ± 3.4	3.3 ± 3.3	2.2 ± 3.2	2.8 ± 3.5	2.7 ± 3.1	2.1 ± 2.9	2.4 ± 2.9
Evoked pain	3.8 ± 3.0	3.2 ± 2.9	3.2 ± 2.9	3.2 ± 2.4	2.7 ± 2.6	2.9 ± 2.8	3.1 ± 2.2	2.4 ± 2.3	2.7 ± 2.4
Paresthesia/dysesthesia	3.9 ± 2.7	3.3 ± 3.2	3.0 ± 3.1	3.7 ± 2.5	3.1 ± 2.3	3.4 ± 2.3	3.7 ± 2.4	3.3 ± 2.6	3.2 ± 2.5
Pain interference (BPI)	29.6 ± 11.3	30.7 ± 14.6	29.9 ± 15.5	29.4 ± 11.9	25.7 ± 13.6	26.9 ± 15.3	28.9 ± 12.1	26.5 ± 15.3	26.3 ± 14.3
Anxiety (HADS)	7.0 ± 4.1	6.0 ± 4.7	5.9 ± 4.0	7.2 ± 3.5	6.5 ± 3.6	6.4 ± 3.7	7.3 ± 3.9	5.7 ± 3.3	7.0 ± 4.0
Depression (HADS)	7.4 ± 4.7	7.6 ± 4.7	7.7 ± 4.5	7.4 ± 5.2	7.6 ± 4.9	7.7 ± 4.8	8.1 ± 4.6	7.3 ± 5.2	7.6 ± 5.2
Catastrophizing (PCS)	19.1 ± 11.6	18.9 ± 13.1	18.6 ± 13.8	20.9 ± 11.8	19.2 ± 11.5	18.1 ± 11.7	19.8 ± 11.0	18.8 ± 12.6	18.1 ± 12.0

Only the change in NPSI total score measured at week 1 and week 3 after H-coil stimulation was significantly different from sham stimulation (*P < 0.05).

BPI, brief pain inventory; HADS, hospital anxiety depression scale; NPSI, neuropathic pain symptom inventory; PCS, pain catastrophizing scale.

3.4. Blinding

Almost half of the patients (47%) did not know whether they received active or sham treatment. The treatment was wrongly identified by 27% and correctly identified by 25% of the patients.

3.5. Safety

The most frequently reported adverse effects were pain at the stimulation site (26% of the patients) and headache during or transiently after the stimulation (35% of the patients). These adverse events were slightly, but not significantly, more frequent after active stimulation than sham stimulation (55% vs 45%) and after H-coil (52%) than F8-coil (41%). They were judged mild to moderate, except in 5 patients who reported severe pain at stimulation site during some of the stimulation sessions (2 after active H-coil stimulation, 2 after H-coil stimulation, and 1 after sham stimulation), but they did not discontinue the protocol. Other common adverse effects (≥10% of cases) included increase in bodily pain and fatigue. No death was reported. A serious adverse event occurred in 1 patient in the active stimulation group (discovery of a rectal cancer) and was considered to be unrelated to the treatment by investigators.

4. Discussion

In this multicenter, international, crossover, randomized, sham-controlled double-blind study, we directly compared the effects of conventional “superficial” (F8-coil) and “deep” (H-coil) rTMS in patients with central neuropathic pain. Our results showed that both types of stimulation induced analgesic effects of similar magnitude, albeit with a different time course. The effects of F8-coil stimulation appeared earlier (D5), whereas the effects of H-coil stimulation were slightly delayed (W1) but tended to last longer (W3) as regards to several secondary outcomes.

Only 2 single-center studies had assessed the effects of H-coil rTMS in patients with chronic pain.^29,37 These studies were conducted in small cohorts of patients (respectively, 23 and 18 patients) and only one directly compared F8-coil and H-coil stimulation.³⁷ As was the case in our study, these studies used 5 daily stimulation sessions according to a crossover design. However, the study of Shimizu et al.³⁷ found only very short-lasting analgesic effects (ie, up to 1 hour after the stimulation) in patients treated with H-coil, but not F8-coil stimulation, whereas both F8-coil and H-coil stimulation induced analgesic effects in our study. This discrepancy could be because of several reasons including in the study by Shimizu et al.,³⁷ a lower number of patients, a complex study design (6 subgroups), and a heterogenous cohort of patients (with both patients with peripheral and central neuropathic pain). In addition, and probably more importantly, the stimulation parameters used by Shimizu et al.³⁷—a frequency 5 Hz with a total of 500 pulses per session—were much lower than those used in most previous positive rTMS studies in patients with chronic pain.^5,20

The other study²⁹ evaluated the efficacy of active H-coil stimulation vs sham in patients with diabetic painful neuropathies with a follow-up of 3 weeks. The authors reported significant analgesic effects from D5 to W3. Our observed effects of H-coil stimulation were consistent with these findings, but we observed delayed effects on average pain intensity (the primary outcome) at W1 in our patients, which were not sustained at W3. These differences between the results of the 2 studies could be because of variations in the population of patients (central vs peripheral neuropathic pain) or differences related to the stimulation target. In the study by Onesti et al.,²⁹ the stimulation target was the foot area, which location is deeper in M1. Thus, it is possible that our stimulation target was not optimal and it would be interesting to directly compare the effects of H-coil stimulation of the hand or foot M1 areas in future studies. However, it is worth noting that secondary outcomes, including PGIC and improvement of the NPSI total score, remained significant at W3 after H-coil stimulation in our study.

Interestingly, the differences in the time course of the effects of H-coil and F-8 coil rTMS in our study may reflect distinct mechanisms of action. Consistent with this hypothesis, we found only a moderate correlation between the analgesic effects of H-coil and F8-coil stimulation, suggesting that their mechanisms of action are only partly common. The mechanisms of the analgesic effects of rTMS in neuropathic pain are still largely elusive.^16,25 They may involve endogenous opioids,¹² glutamatergic systems,¹⁰ changes in cortical excitability,^21,24 or in descending pain modulation.^25,26 However, these potential mechanisms only concern F8-coil stimulation and it is unknown whether similar mechanisms are also involved in the effects of H-coil. Onesti et al.²⁹ reported that H-coil stimulation induced an inhibition of the RIII nociceptive reflex, suggesting that the analgesic effects could be related to the activation of descending inhibitory controls. In contrast, we previously found (although in healthy subjects) that F8-coil stimulation does not induce significant changes in the RIII reflex.²⁸ Consequently, one might propose that the deeper stimulation induced by H-coil might act on descending modulatory systems primarily organized in the brainstem, whereas F8-coil affects other (more superficial) neural circuits. It is also possible that differences in the properties and geometry of the electric fields generated by the 2 types of coils may have led to differential activation in the various cortical layers of M1, which are differentially connected to pain modulatory pathways in the brain, as shown in a recent animal study.¹⁵ However, it is essential to interpret these arguments suggesting differential mechanisms for the 2 modes of stimulation with caution because they are only indirect evidence. The fact that the analgesic effects induced by the 2 stimulation modes were only moderately correlated suggests that the responder profiles to superficial or deep rTMS may also be distinct. We were not able to identify differences in the clinical profiles of responders in our patients based on baseline characteristics, but our study was not powered for such analyses. Further studies with larger sample sizes are needed to investigate the mechanisms of action of F8-coil and H-coil rTMS and identify potential clinical responders' profiles.

Our study has several strengths. One was the quality of the blinding that was based on the use of a double-sided active/placebo F8-coil with active and placebo sides with identical appearance, akin to a recent study on neuropathic pain.⁵ The type of stimulation was determined directly by the information supplied to the rTMS machine by the patient's flash drive or personalized code, ensuring no intervention by the operator. The quality of the blinding was further confirmed by a questionnaire at the end of the second treatment period, in which about 3 quarters of our patients were unable to recognize whether they received active or sham stimulation or guessed wrongly. Another strength was the use of a robot driven by MRI-guided neuro-navigation in 2 centers to ensure accurate targeting and maintenance of the F8-coil on the target throughout the rTMS sessions. However, our study had limitations, particularly the short-term treatment period (5 days). Hence, it was not possible based on our results to confirm whether treatment with H-coil would be clinically useful for chronic pain management, which requires long-term maintenance of analgesic effects with sufficiently long intervals between stimulation sessions.^5,30 Although the trend for longer lasting effects of H-coil stimulation is promising, further studies are needed to confirm its ability to maintain analgesic effects over an extended period.

Conflict of interest statement

The authors have no conflicts of interest to declare.