Patients were planned to receive 10 consecutive sessions of the infiltration and stimulation series with 48–72 h between each session following recommendations for long-lasting after-effects (30, 46–48). Before each session, the attending physician decided in a context-sensitive approach if a continuation of infiltration and thus stimulation series was indicated. If applicable, first tDCS (anodal/cathodal/sham, depending on study group) and afterwards local infiltration was performed. After completion of series, patients were followed-up for 6 months.

The primary outcome parameter was relative pain intensity reduction after completion of therapy series (typically after 2 weeks of treatment) measured in NRS score as a numeric value between 0 and 10. The secondary outcome parameter include absolute and relative pain reduction measured in NRS score after completion of tDCS stimulation vs. initial NRS score measured before stimulation and time until patients need additional regional-anaesthesiological interventions. The number of required regional-anaesthesiological interventions to achieve sufficient pain reduction and the analysis of adverse reaction (skin redness, headache, concentration, other) were other secondary outcome parameter.

We further evaluated patients' conditions, the used blockade technique, the response rate and the effect of sole tDCS stimulation. In addition to that, analyses of side-effects, drop-outs, self-assessment and a post-hoc sample size calculation was conducted.

For pain assessment, two assessment tools were used: the pain assessment protocol for infiltration series and the German Pain Questionnaire [daily report form, version 2007 (49)]. The assessment protocol was the basis of the evaluation of the primary outcome parameter and contained the NRS on an 11-point Likert scale (0–10) for static (at rest) and dynamic (maximum pain in stress) pain before and after tDCS. The German Pain Questionnaire was used both during stimulation and infiltration series and for the Follow-Up. Its core questions are derived from the grading of chronic pain status (50). They consist of four questions on a 11-point Likert scale (0–10): average pain in the last week, maximum pain in the last week, mental distress and impairment in daily activities. Furthermore, the German Pain Questionnaire consists of a question regarding the endurance of pain (1 = not applicable, I have no pain, 2 = I can tolerate it well, 3 = I can just tolerate it, 4 = I can tolerate it badly). For the pain assessment, patients were asked to name the most predominant painful side.

Before each local anesthetic infiltration series, a 20 min tDCS stimulation was performed using the NeuroConn DC Stimulator^® with saline soaked, square sponge electrodes (surface 25cm²) similar to Nitsche and Paulus et al. (22, 47, 48) (see Supplementary Figure 1). Following the protocol of Morosoli et al., we placed the anode electrode over the primary motor cortex (M1) contralateral to the most predominant painful side, and the cathode electrode over the contralateral supraorbital area (51). The primary motor cortex is located in the Brodman location 4 (52, 53). Electrode position of C3,4 correlates with Brodman location 4, which is located in the precentral gyms, shoulder to wrist area, caudal to middle frontal gyrus (54). We determined the C3 or C4 placement using the recommendation of Jasper (55).

The patients were divided in three subgroups with either anodal, cathodal or sham stimulation. The same stimulation setting was used in every subgroup. Due to the triple-blinding study design, electrode placement was identical in either anodal or cathodal stimulation group with inverse current flow, depending on study allocation (e.g., anodal stimulation: anode=anode and cathode=cathode whereas cathodal stimulation anode=cathode and cathode=anode).

In the active groups (anodal and cathodal), stimulation started with an initialization phase with increasing current over 30 s. Afterwards, tDCS with 2 mA was applied for 20 min, following a phase with decreasing current over another 30 s. In the sham group, patients received increasing and immediately decreasing current at the beginning and similar application at the end of the stimulation for the purpose of blinding.

This study was performed in a triple-blinded setting. The patient, the tDCS applying physician and the person performing statistics were blinded to group allocation. Blinding of tDCS was performed using the study mode of the NeuroConn DC Stimulator^®. A block-randomization was used with blocks of six to ensure comparable group sizes in case of early stop of the study. Randomization was performed using a computer-generated random list for patient allocation with four blocks with size of six provided by study statistician. After inclusion, patients were treated following the randomization list. For allocation concealment, block sizes and study randomization list were prepared blinded to study physicians.

To explore the intrinsic effect of the stimulation setting, we compared tDCS patients with patients without tDCS obtained in a historic cohort [NCT03066037, report in (10)] with the same infiltration techniques applied in the same outpatient clinic.

Patients were followed-up at 1, 3, and 6 months after completion of the combined stimulation and infiltration series. Follow-Up was performed via telephone calls using the German Pain Questionnaire [daily report form, version 2007 (49)]. Furthermore, patients' records at the outpatient clinic were screened whether and when study patients received a new infiltration series outside the study after completion of stimulation and infiltration series.

All statistical analyses were performed using SPSS 29. Descriptive data was summarized using mean and standard deviation or median and range depending on scale level and distribution. Analysis of immediate tDCS effect was performed using data from each session. For analyses of statistical significance, NRS-scores were explored using the exact Wilcoxon-signed-rank-test for paired data. To analyse independent groups of ordinal variables, Mann-Whitney-test or Kruskal-Wallis-test was applied, as appropriate. Distribution of continuous data was examined with graphical exploration and Kolmogorov-Smirnov-test. For binary data, Fishers exact test was applied. To describe odds between groups, Mantel-Haenszel estimation was performed. All statistical significance tests used a two-sided alpha level of < 5% and were intended as exploratory in this pilot trial. Similar to previous studies, we defined responders as patients with a pain reduction measured in NRS of at least 50% (9, 10, 16). Patients treated per protocol were included into analysis (n = 13). The study was a priori planned to explore a clinical meaningful difference in pain reduction measured in NRS (0.33 vs. 0.5 pain, SD ±0.1) with a power of 0.8 resulting in a number of 24 patients to be randomized. In 2019, the study was temporarily on hold due to explore unexpected high rates of drop outs (4 out of 17), however, restart of this trial was not feasible due to ongoing restriction to perform studies during COVID-19 pandemic and thus terminated.

Included patients suffered from either trigeminal neuralgia or persistent idiopathic facial pain (PIFP) (basic characteristics, Table 1).

Most patients received a blockade at the sphenopalatine ganglion (SPG) n = 11 (84.6%) as main infiltration site. Ganglionic local opioid analgesia (GLOA) infiltration [n = 1 (7.7%)] and infiltrations at the N. occ. major [n = 1 (7.7%)] were seldom reported as main infiltration site. For SPG blockade local anesthetics (2–3 ml bupivacaine 0.25%) was applied via infra-zygomatic injection. For GLOA infiltration, lipophilic opioids and local anesthetics (5 ml 0.5% bupivacaine and 0.03 mg buprenorphine) were injected close to paravertebral cervical ganglions. In patients with mononeuropathic pain patterns, nerve blocks with local anesthetic (ropivacaine 0.2%, 3 ml) were used. Included patients received 2–10 infiltrations and stimulations in a series with median 9 infiltrations and stimulations (IQR 6–10). All 13 per-protocol treated patients received 20 min of tDCS before each infiltration in the series.

The NRS score before infiltration and stimulation series for per-protocol patients were at median 7 (IQR5.00–9.50). There was no significant difference between dropouts and non-dropouts (p = 0.249). Throughout series, there was a significant overall decrease of NRS scores in treated patients (before median 7 (IQR 5.00–9.50), at the end of series median 3 (IQR 1.00–4.00), p < 0.001; see Figure 2A).

The NRS scores before intervention were comparable between stimulation groups (p = 0.181). NRS scores decreased in all three study groups throughout series. The anodal stimulation group started with lowest NRS scores before series 5.50 (median, IQR 4.50–7.00) and achieved lowest NRS scores after series 1.00 (median, IQR 0.75–4.00), resulting in the highest relative NRS reduction throughout series of 73.33% (median, IQR 50.00-87.50%). The cathodal stimulation group had higher NRS scores before series 9.00 [median, (IQR 8.00–9.00)] and after series 4.00 (median, IQR 3.00–4.00) and, thus, lowest relative NRS reduction 50.00% (median, IQR 44.44–50.00%). The NRS scores in the sham group before series were 7.50 (median, IQR 4.75–9.50), after series 2.50 (median, IQR 1.25–4.50) and a consequent relative NRS reduction of 68.75% (median, IQR 37.05–78.75%). The relative NRS reduction throughout series did not differ significantly between the study groups (p = 0.532; see Figure 2B).

76.9% of patients had a 50% NRS reduction throughout series. Response of 50% NRS reduction was highest in anodal group (83.3%), followed by sham group (75.0%) and cathodal group (66.7%). Differences in response rate were non-significant (p = 0.850). Odds ratio in the anodal group for 50% NRS reduction was 1.667 (CI 0.074–37.728) and in the cathodal group 0.667 (CI 0.025–18.059) compared with sham group.

The time to the event of 50% NRS reduction was at Median 3 sessions (IQR 2–3). The time to event was shortest in the cathodal group (median 2, IQR 2–2), followed by sham (median 3, IQR 2.25–3) and anodal group (median 3, IQR 2.75–3.50). Differences in time to event were not significant between stimulation groups (p = 0.644).

With a response equalling 50% NRS reduction achieved in 75% of patients in the sham group and an odds ratio of 1.667 in the anodal group, the number needed to treat with anodal stimulation as verum treatment equals NNT = 12.

All three subgroups reported an overall NRS reduction throughout series. Course of reported median NRS scores are indicated in Figure 3.

Overall, 10 patients suffered from trigeminal neuralgia whereas 3 patients had a diagnosis of PIFP. Differences in relative reduction of pain intensity over time were not significant (p = 0.469, see Supplementary Figure 2 and Supplementary Table 1).

Overall, there was a significant immediate effect of tDCS before performance of local anesthesia on pain intensity: NRS scores were reduced in median 1 point (IQR 0–2; p < 0.001). Nevertheless, no difference between anodal, cathodal or sham stimulation was noted (p = 0.482).

Based on the relative NRS reduction throughout series comparing anodal treatment and sham, a sample size calculation for a confirmatory study was performed. A sample size of 93 patients is necessary to achieve level of significance in a confirmatory study. Using a conservative estimation including the drop-out rate of this study with 23.5%, altogether 120 patients would necessary to be randomized.

Of included per-protocol treated patients, 76.9% (n = 10) reported some kind of side-effects due to tDCS. A prickling (53.8%, n = 7) or burning sensation (53.8%, n = 7) was the most common side-effect. Mild local pain (15.4%, n = 2) and skin redness (7.7%, n = 1) was reported seldom. No patient reported increased headache due to stimulation. There was no severe side-effect or reported drop-out due to side-effects.

Four patients dropped out of the study. One reported psycho-social distress not related to the study as the reason for dropout. One patient reported non-sufficient effect of the infiltration and two did not report a reason. Notably, most patients dropped out of the cathodal (n = 3) group followed by the sham group (n = 1). No patient dropped out of the anodal group.

The differences in relative NRS reduction between non-stimulation group (n = 83, Median 44.44%, IQR 0.00–70.00%) and stimulation group (n = 13, Median 66.66%, IQR 47.22–80.00%) were not significant (p = 0.054). A higher proportion of responders with 50% NRS reduction was noted in group with stimulation setting (n = 10, 76.9%) than in non-stimulation group (n = 41, 49.4%, p = 0.079; see Figure 4). This results in an OR of 3.41 for stimulation vs. non-stimulation setting and a subsequent NNT by stimulation setting of 3.63.

There was no statistically significant difference in tDCS study groups regarding maximum pain measured in NRS at 1 month follow-up time point (p = 0.645), three-month follow-up time point (p = 0.626) and 6 month follow-up time point (p = 0.835). There was as well no statistically significant difference in tDCS study groups regarding average pain measured in NRS at 1 month follow-up time point (p = 0.404), 3 month follow-up time point (p = 0.618) and 6 month follow-up time point (p = 0.632).

At 1 month follow-up, there was a significant difference in mental distress between subgroups with cathodal stimulation group having the highest rating in impairment [anodal median 3 (IQR 0–8); sham median 0 (IQR 0–0); cathodal median 8 (IQR 5–8); p < 0.05]. In both, impairment in daily activities [anodal 0 (IQR 0–5); sham median 1 (IQR 0.25–1.75), cathodal median 5 (IQR 2–5); p = 0.231] and endurance of pain [anodal median 2 (IQR 1.5–2.5); sham median 2 (IQR 1.25–2), cathodal median 3 (IQR 2–3); p = 0.154] cathodal stimulation apparently showed worse scores. At both, 3 and 6 months follow-up, we observed no statistical differences in impairment in daily activities, mental distress and endurance of pain.

In total, three patients received additional local infiltration series outside the study after completion of stimulation and infiltration series. The patients with additional series were equally distributed among the study groups (anodal n = 1, sham n = 1 and cathodal n = 1) and the time to the series (median 12, IQR 11–12) did not differ between subgroups (p = 0.368).

To determine patient blinding, patients were asked in a self-assessment to which group they belong. Group assignment and self-assessment to active or sham stimulation did not show any statistical association (p = 0.429).

Although providing insights into tDCS for chronic neuropathic pain patients, this pilot study has some limitations. Especially the small sample size in pilot studies limit generalizability of results. Furthermore, individual pain perception differs between patients and other outcome measurements like change in medication would be an interesting variable to include in future protocols (70). In addition to that, the inclusion and exclusion criteria did not cover all possible confounders (e.g., the effects of pain medication). To cover such unmeasured confounders, the study was performed in a randomized-controlled design with included long term data on patients. Another limitation were differences in the baseline NRS among the different study groups. Though statistically not significant, we used relative instead of absolute NRS reduction and a definition of 50% NRS reduction as a clinically relevant response to address these differences. Regarding suitable outcome measurements, resting pain, maximum pain in exertion as well as frequency of pain attacks and change in pain medication should be assessed.

As blinding is a critical issue in tDCS, we observed sufficient blinding provided by the study mode of the NeuroConn DC-Stimulator^®. There was no connection noted between the self-assessment of patients and the actual study group. Thus, our results support the use of an increasing and decreasing current at the beginning and end of a sham stimulation to imitate effects of tDCS. This goes with the results of the study by Gandiga et al. who described this sham procedure. In their pooled data analysis including several studies over 3 years with 170 stimulation sessions, there was no difference in the incidence of side-effects between tDCS and sham groups (71). This was contrary to results from a study evaluating differences between sham and active tDCS with 131 patients receiving either type of stimulation; a statistically higher rate of sensory side effects was noted in the active tDCS group (72). Based on our finding, blinding in our tDCS population was sufficient.

Future studies could try to lower the barriers to tDCS application with supervised stimulation at home (18). A different approaches is to combine non-invasive brain stimulation with other non-invasive approaches such as neurofeedback to further enhance possible beneficial effects (73). Not only feedback to the patient but also to the tDCS applying in real-time tool could be a new research direction. Thus ongoing brain activity could provide the base for closed-loop technology allowing the delivery of tDCS specific to an individual's internal state (18). Furthermore, future studies should address the limitations reported in our study including small sample size and unmeasured confounders. A sample size calculation for such future studies, revealed the necessity of 120 patients with refractory neuropathic cranial pain to confirm our findings. This is challenging, since infiltration series are only used in patients refractory to standard treatment and thus performed relatively rarely. In 4 years, a study in an university affiliated pain outpatient clinic reported only 74 patients receiving a ganglionic opioid analgesia (GLOA) at the superior cervical ganglion as an infiltration series (16). These numbers are supported by our previous study performed as well in a university affiliated pain outpatient clinic with 83 patients receiving infiltration series in six-and-a-half years (10). Hence, a confirmatory study could only be performed in multicentre study design.