Methods and selection flowchart
By the search and selection process depicted in the flowchart (Figure 2), we started from 330 arriving to 18 analysed RCT papers (4 depression, 4 fatigue in MS, 1 pain, 5 addictions, 4 fibromyalgia).
Overview of the Class 1 tDCS RCT in no-structural diseases
We reported a description of all the non-structural studies organized by pathology. We started with those for which the meta-analysis was executed and we reported a qualitative description for those excluded for the meta-analysis, clarifying the reason for this choice.
For each pathology, we quantified the efficacy of the treatment for Sham (Tables 3, 6, 9 and Forest plot in Figures 3, 6), Real (Tables 4, 7, 10 and Figures 4, 7) and their relationship (Tables 5, 8, 11 and Figure 5, 8), and specified the population enrolment criteria, the outcome measure and stimulation parameters in Tables 12, 13, 14.
Depression
Meta-analysis
One study 30 reported the percentage mean and standard deviation (SD) variation but none of the studies reported the mean change score and the SD, data were calculated as described in statistical method section. The correlation (Pearson’s r) calculated from data was equal to 0.86 in Sham group (Table 3, Figure 3) and to 0.96 in Real group (Table 4, Figure 4).
The Sampaio-Junior et al.’s study 31, reported data about depression measured by Hamilton depression rating scale (HDRS-17) scores while the other three studies 30,32,33 reported data about Montgomery‑Asberg depression rating scale (MADRS). For one study 33 data were extracted from graph.
Pooling the studies, the total number of patients in was 132 the Sham group and 135 in the Real group.
Depression - Sham effect
The pooling effect size was equal to 0.93 (95% CI 0.52-1.33; p<0.001), indicating a large Sham effect. The heterogeneity was considerable (I2=92.4%, p<0.001).
We performed a sensitivity analysis excluding the Loo et al. study (2018), because the data were extracted from the graph. The pooled effect size was equal to 0.72 SDs (95% CI 0.57-0.87; p<0.001), indicating a moderate Sham effect. The heterogeneity was not significant (I2=26.1%, p=0.259).
Another sensitivity analysis, performed decreasing the correlation to 0.5, showed consistent results about the reduction after sham (ES=1.15, 95% CI 0.59-1.71; p<0.001). The heterogeneity was considerable (I2=84.5%, p<0.001). Reperformed the analysis excluding Loo et al. 2018, the pooled effect was reduced to 0.85 (95% CI 0.61 to 1.095; p<0.001) and the heterogeneity inter trials was not significant (I2=0%, p=0.437) (Table 3, Figure 3).
Depression - Real effect
The pooling effect size was equal to 0.99 (95% CI 0.56-1.42; p<0.001), indicating a large effect after tDCS. The heterogeneity was considerable (I2=98.2%, p<0.001).
We performed a sensitivity analysis excluding the Loo et al.’s study 2018, because the data were extracted from the graph.
The pooled effect size was equal to 1.05 (95% CI 0.41-1.70; p=0.001), indicating a large effect after tDCS. The heterogeneity was considerable (I2=98.8%, p<0.001).
The elevate heterogeneity was maybe due to different administration condition of tDCS.
Another sensitivity analysis, performed decreasing the correlation to 0.5, showed consistent results about the reduction after tDCS (ES=1.78, 95% CI 1.13-2.43; p<0.001). The heterogeneity was considerable (I2=82.5%, p=0.001) (Table 4, Figure 4).
Depression - Real vs Sham effect
The results showed that the tDCS had a large effect, the reduction in depression observed after tDCS was on average 1.09 SDs higher than that observed after sham (95% CI=0.63-1.54; p<0.001)(Fig. 1). Heterogeneity was moderate and significant (I2=66.8%, p=0.029).
By the funnel plot no evidence of publication bias emerged.
We performed a sensitivity analysis excluding the Loo et al.’s study 2018, for the reason previously explained. The pooled effect size was equal to 1.28 SDs (95% CI 0.91-1.65; p<0.001), indicating a large effect after tDCS. The heterogeneity was not significant (I2=23.5%, p=0.271).
The sensitivity analysis, considering a correlation of 0.5, showed that the results were consistent. The reduction in depression observed after tDCS was significantly higher than that observed after Sham (SMD=0.63, 95% CI=0.39-0.88; p<0.001). The heterogeneity was not significant (I2=0%, p=0.570) (Table 5, Figure 5).
PICO variables’ values for tDCS against depression
The depressed patients populations who benefit by tDCS suffer by moderate to severe symptoms (Table 12). Data suggest the use of MADRS as outcome measure, since it was used by all but one study, which employed HDRS as primary outcome and observed similar results when assessed by MADRS. When defining the most efficacious tDCS intervention parameters, we observed that all studies position the anode on left dorsolateral prefrontal cortex (dlPFC), targeted according to the 10-20 international system of EEG electrodes placement by centering the anode on F3. Concerning the cathode, F8 seems to drive the most stable result, since the two studies using F4 instead obtained the best and worst result. Nevertheless, the results from F8 cathode position are in-between, so that we suggest that there is no advantage in making the position of anode and cathode asymmetrical in the two hemispheres. We therefore suggest the montage of anode on F3, cathode on F4 (Figure 10).
In relation to the current delivered, it varies little between the studies and the higher current surface density (csd) used in two studies, results in the best and the worst efficacy, so the suggested csd can be the average across the four studies, around to 0.072 mA/ cm2.
In relation to the duration, 10 days of treatment seem sufficient, in fact the best result is with 10 days (Sampaio et al.), and Loo and colleagues observed a lower result with 20 days treatment than the 15 days one. We suggest the most used daily session duration of 30 minutes, even though Loo et al. with 20 minutes duration observed a result comparable to 30 minutes one. Therefore, suggested duration is 30 min / day, 10 days.
MS fatigue
Meta-analysis
Five studies 25,34–36 fulfilled the criteria of class 1 studies. Tecchio et al.’s 2015 study considered the same sample of Tecchio et al. 2014 for the S1 target, so only data of the most recent study was considered in the meta-analysis. The group stimulated on hand SM1 in Tecchio et al. 2015 was not considered since it was included in Ferrucci et al.’s study, which involved a larger group (Table 6, 7, 8, Figure 6, 7, 8).
Since all the trials were cross-over, methodological methods indicated by Elbourne et al. 37 were followed. As measure of correlation between the two conditions (Sham and Real), the coefficient indicated by one of the individuated studies (Cancelli et al., r = 0.55) was assumed. The summary statistics used was the standardized mean difference (SMD). We calculate the effect size (ES) and the relative Standard error (SE) as describe in method-summary measures section.
Pooling the studies, the total number of patients was 46 subjects.
Ms Fatigue – Sham effect
The pooled ES indicated a significant small effect of the sham (ES = 0.27, 95% CI 0.11 to 0.42; p=0.001). The Heterogeneity was moderate (I2=69.7%, p=0.037). One possible source of Heterogeneity was the Electrode positions. We reperformed meta-analysis considering only two studies with the same Electrode positions (Tecchio et al 2015 and Cancelli et al. 2018). The results confirmed what seen previously, a significant small effect of Sham (ES=0.18, 95% CI 0.07 to 0.3; p=0.002). The Heterogeneity was not significant (I2=0%, p=0.803) (Table 6, Figure 6).
Ms Fatigue – Real effect
We calculate the Standardize Mean Difference as effect size (ES) and the relative Standard error (SE) as describe in method-summary measures section. The pooled ES indicated a significant large effect of Real (tDCS) (ES= 0.80, 95% CI 0.42 to 1.17; p<0.001). The Heterogeneity among trials was substantial and significant (I2=71%, p=0.032). One possible source of Heterogeneity was the Electrode positions. We reperformed meta-analysis considering only two studies with the same Electrode positions (Tecchio et al 2015 and Cancelli et al. 2018). The results indicated a significant large effect of Real (ES=0.98, 95% CI 0.69 to 1.27; p<0.001). The Heterogeneity was not significant (I2=0%, p=0.503) (Table 7, Figure 7).
Ms Fatigue – Real vs Sham effect
All the studies about fatigue in multiple sclerosis patients were crossover studies. We estimate the mean difference, SMD, between the mean value post Sham and post Real.
The results showed that the Real tDCS had not a significant effect compared to Sham, the pooled standardized mean difference of mFIS observed after tDCS was 0.34 standard deviation (sd) lower than the fatigue observed after Sham (95% CI= 0.24 to 0.92; p=0.247) (Fig. 1). Heterogeneity was substantial and significant (I2=71.7%, p=0.029). There were only three small studies and it is difficult to evaluate the funnel plot.
We reperformed meta-analysis considering only two studies with the same Electrode positions (Tecchio et al 2015 and Cancelli et al. 2018). The results showed a marginally significant moderate effect of Real (tDCS) compared to that of Sham (ES=0.61, 95% CI 0.02 to 1.23; p=0.056). The Heterogeneity was not significant (I2=50.6%, p=0.155) (Table 8, Figure 8).
PICO variables for tDCS against MS fatigue
Given that the few Class 1 RCTs involved small patients populations, we deduced the suggested parameters (Table 13) aware of the need of confirmation in larger groups.
The MS patients populations who benefitted by tDCS suffered by severe fatigue symptoms, as quantified by Modified Fatigue Impact Scale (mFIS) > 35, which appears to be a sufficient cut-off, since positive effects of the treatment were observed from this inclusion threshold. All cross-over studies involved small populations, and the results suggest that the inclusion of minimal to moderate clinical severity (Expanded Disability Status Scale, EDSS < 3.5) is opportune at the present stage instead of offering the tDCS treatment to patients in a wider range of disease-related impairment. While it will be relevant to assess the efficacy in presence of increasing disability in larger populations, it is of note that MS is typically accompanied by severe fatigue even in the absence of other disabling symptoms, thus a treatment efficacious against fatigue becomes decisive for the patient's quality of life.
Present data, corroborated by a recent wide-population multi-centre RCT 38, suggest mFIS as proper outcome measure, sufficient with its 21-items to sense the induced variations, with no necessity to collect the longer 40-items Fatigue Impact Scale (FIS).
When defining the most efficacious tDCS intervention parameters, we observe a huge difference in efficacy when the anode targeted the bilateral whole-body primary somatosensory cortex, S1 35,36 instead of the left and right hand sensorimotor regions 25 where Real tDCS had not a significant effect compared to Sham. For the cathode, it appears appropriate to enlarge the surface area to minimize the effects in the reference area selected in occipital cortex to focus the induced current in post-central regions (Figure 10). Another source of difference with Ferrucci et al. study 25 could be the extra-cephalic reference, but this choice to minimize the undesired effects under the cathode is frequently used, with positive results in other cases.
In relation to the current intensity, the lower efficacy in Ferrucci et al. study 25 could be due to the half current superficial density delivered across the two electrodes centred on C3 and C4, each 35 cm2 study. Thus the suggested value is 0.04 mA / cm2.
In relation to the duration, 15 minutes per day for 5 days of treatment seem sufficient.
The result of the present meta-analysis of the Class 1 tDCS RCTs further strengthens the findings of a recent systematic review and meta-analysis focused on non-invasive brain stimulations against fatigue 39, which indicated that short-term and long-term treatment effects were significant for tDCS, whereas real TMS and transcranial random noise stimulation were not superior to sham stimulation. Among the 11 tDCS RCTs, the strongest efficacy emerged targeting the bilateral whole body S1 34–36.
Positive indications of tDCS efficacy against fatigue when targeting primary somatosensory cortex suggest that positive effects can be induced also in dystonia. In fact, although dystonia is a heterogeneous neurological condition which manifests mainly as a movement disorder, from a physiological point of view the core generating mechanism is an abnormal sensorimotor integration 40. Abnormalities can be found at multiple level such as a loss of inhibition, sensory disfunction and alterations in synaptic plasticity at different levels of sensorimotor circuit 4142. Although there are no Class 1 RCTs available to support a consistent use of tDCS in dystonia, there are seminal results of efficacy coming from studies targeting specific dystonic conditions 24. We can speculate about future perspectives in the treatment of dystonia by means of electroceutical interventions aiming at restoring the pathological functional unbalances, capitalizing on the results coming from available and sound RCTs that in counteracting efficaciously fatigue are able to rebalance altered mechanisms similar to those found in dystonia within the sensorimotor circuit.
Pain
Quantitative analysis
A single study 43 on 18 people suffering by back pain either bilaterally of with a monolateral prevalence satisfied the Class 1 criteria. Tables 9, 10, 11 reportsthe pre- and post- Sham and Real treatments values of the outcome measure and the differential effect of Real with respect to Sham.
PICO variables for tDCS against pain
Straudi et al 2018 43, the only study against pain matching the Class 1 requirements, involved a group of people suffering by back pain, without comorbidities, at a mean level of 5.6±1.8 cm of the Visual Analogue Scale (VAS) for pain intensity with maximum 10 cm. The VAS is proposed as a proper outcome measure. In dependence on the prevalence of symptoms, the anode was placed on the primary motor cortex (M1) of the dominant hemisphere if pain was central in the back or bilateral, or on contralateral M1 if pain irradiated to one side. The cathode was positioned on the supraorbital area contralateral to the anode. We formulate a doubt about the choice of centering the electrode on C3/C4, corresponding to the hand representation, instead of a more central position corresponding to lower body representation. For cathodal position, electrode size, current intensity and stimulation duration refer to the Table 14 Pain.
Fibromyalgia
In our query, four studies reported positive results of tDCS intervention against fibromyalgia symptoms. Since applying the PICO model no clear diagnostic criteria exist for fibromyalgia, we excluded this condition from the quantitative analysis and we report the studies in a descriptive manner.
Two studies focused on the primary symptom of this pathology, pain. Riberto et al. 44 applied tDCS over left M1 (2mA, 20 min, 1 session per week, 10 weeks) combined with a rehabilitation program. They observed a reduced impact of pain on the quality of life after Real vs Sham tDCS but not differential effect on pain intensity, depression and anxiety.
To et al. 45 applied tDCS (1.5 mA, 20 min, 2 sessions per week, 4 weeks) comparing the efficacy of two bilateral montages, occipital (C2 dermatomes) and dlPFC (F3 and F4), in reducing fibromyalgia-related pain and fatigue. They observed a significant reduction of pain but not fatigue when targeting the C2 dermatomes region and a significant reduction of both pain and fatigue when targeting bilateral DLPFC.
Other two studies focused on secondary cognitive aspects related to fibromyalgia, as attention and memory performance.
Santos et al. 46 combined tDCS on left DLPFC (1 mA, 20 min, 8 session for 8 consecutive days) with working memory training, obtaining an improvement of immediate memory capacity and verbal fluency. Moreover, Silva et al. 47 combining tDCS with Go/No-Go task, reported an improvement of attention performance as primary outcome together with an improvement of pain threshold considered as secondary outcome.
Overall, these results from Class 1 RCTs integrate with promising evidence of tDCS efficacy in the treatment of pain and cognitive dysfunctions.
Addiction
Five Class1 RCTs used tDCS against addiction behaviors. We review the results without executing the meta-analysis since it was not possible to pool them in a common statistic, either for absence of indication of a primary outcome among multiple variables to quantify the effect, or for inhomogeneity of the outcome (e.g., relapse rate vs. number of smoked cigarettes).
Against alcohol addiction, Klauss et al. 48 targeted right dlPFC by anode (2mA, 13 min, twice per day, 5 days) against left dlPFC and reported a significant reduction in relapse rate up to six months. Den Uyl et al. 49 tested the efficacy of tDCS, anode over left dlPFC and cathode over right dlPFC, in enhancing the effect of a Cognitive Bias Modification (CBM) in patients with alcohol addiction. Despite the results could not isolate the effect of tDCS intervention, tDCS combined with CBM showed no effect on alcohol approach bias, craving or time to relapse and only a trend of significance on reduction in probability of relapse.
Two other studies targeted smoking behavior with similar montage over right or left DLPFC.
Mondino et al. 50 delivered tDCS (2mA, 20 min, twice per day, 5 days, anode F4) focusing on self-reported smoking intake and brain activity via fMRI. Despite no effect of tDCS on smoking intake, they found a significant reduction in smoking craving and increased brain reactivity of the right posterior cingulate area to smoking cues. Conversely, Falcone et al. 51 (anode F3, 1-2mA, 20 min, 3 times per week, 1 week) observed no effect of tDCS on abstinence or number of days of abstinence or change in daily smoking rates nor on latency to smoke or number of cigarettes smoked.
There are indications of efficacy as demonstrated by Batista et al. 52 about cocaine addiction. By targeting right DLPFC (2mA, 20 min, 5 days, 3 weeks) they found an improvement in craving scores, anxiety and overall perception of quality of life with significant effects even after weeks.
We note that looking for tDCS effects against addition, the complication emerges about the interaction with a wide range of systemic and multi-organ consequences as the effect of the abused substance.
Estimation of Sham effect
A further contribute of this quantitative review is the estimation of Sham effect in trials planned to assess the efficacy of tDCS.
According to Figure 9, Sham effect resulted clearly dependent on pathology/symptom. After excluding pain (Sham effect was estimable only in one paper and with high level of imprecision), it is evident (test for subgroup differences: I2=94%, Chi2(1)=16.75, p<0.001) that Sham effect was markedly different in fatigue and depression. We estimated that Sham effect size in fatigue is 0.27 (a small effect) and 0.72 in depression (a quite large effect).
In future RCTs aiming at demonstrating an effect of a Real stimulation, researchers could consider these as “first guess” of the control Sham condition. Table 15 presents different scenarios. On the basis of our estimation of Sham effect on fatigue (standardized Effect Size, sES=0.27), in order to reach a 90% probability (power) of recognizing as statistically significant (at two-sided alpha level set at 0.05) an increase of Real stimulation up to a medium effect size (sES=0.50), a sample size of almost 800 patients will need for a two-arm parallel design. Such number decreases to about 200 patients in case of a one-sample design, thus without the recruitment of patients to be assigned to control Sham treatment. A further decrease of sample size could be obtained with a cross-over design (n=122), for which the correlation between pre-post Sham and pre-post Real changes was assumed equal to r=0.7. Such correlation would imply that about 50% of changes’ variance observed after Real stimulation could be accounted for by changes observed after Sham stimulation. We also indicated the sample size estimate in the case of the large Real stimulation effect resulting from the present meta-analysis (sES = 0.98), which corresponds to much smaller population dimensions.
Following this logic, Table 15 reports the appropriate sample size for different effect size of Real stimulation for depression. To be noted that, if it is considered clinically relevant an effect size of 0.8 for Real stimulation on depression (that corresponds to a large effect size according to Cohen’s convention), a huge sample size should be recruited for a parallel design (about 7500 patients). This was due to the high estimated sham-effect for this condition. In other terms, a future trial should face the probable large effect of sham stimulation and, thus, Real stimulation should be able to induce an even higher benefit on depression scales. Nevertheless, the observed effects in depression were huge after Real stimulation, with the meta-analysis estimation of ES=1.79. In addition, more efficient design, such cross-over, are recommended, provided that a good correlation between changes (within-subjects dependency) could be assumed.
Aware of the small populations involved in this sample sizes’ estimate, we pave the way for assessing the Sham effect for future studies, where the assessment of Real efficacy will get rid of the need of employing wide efforts and long times for patients and experimenters in the Sham assessment. We note that in depression the Sham effect was much higher than in fatigue. A strong placebo effect in Depression was also found in pharmacological treatment RCTs, where a high efficacy of the antidepressant medication was almost as strong as placebo's 53. The Authors of that wide review pose the provocative statement suggesting as preferable the placebo to the pharmacological treatments to avoid the documented severe side effects. In multiple sclerosis, a condition where therapeutic drugs contribute to fatigue generation 54,55, the pharmacological treatments of the fatigue symptom are poorly effective 56, thus stimulating the use of alternative substances, which also reach poor efficacy (ES=0.04, 57), and similarly to depression, the ameliorations to placebo can be similar to that to effective therapy 38,58. These findings support the development of electroceutical solutions, almost free of side effects, which produce as presently reviewed relevant efficacies, open to personalization to further enhance the amelioration for individual patients 59.