We stimulated 20 healthy volunteers (10 men), mean age 25.9 ± 4.95, as part of a study investigating a common framework of action for NiBS. The subjects were interviewed prior to experimentation regarding their state of health. We applied established exclusion criteria for NiBS (Brunoni et al., 2012) and added evidence for photophobia and photosensitive epilepsy. Written informed consent was obtained from all individual participants included in the study. The subjects received financial compensation for their participation. All procedures were performed in accordance with the ethical standards of the Ethics Committee of the Charité – Universitätsmedizin Berlin (“Ethikkommission der Charité – Universitätsmedizin Berlin”) and with the 1964 Declaration of Helsinki and its later amendments. This study adheres to the principles of good scientific practice of the Charité – Universitätsmedizin Berlin (“Grundsätze der Charité zur Sicherung guter wissenschaftlicher Praxis”).

Retinofugal alternating current stimulation was applied via a multi-channel low-voltage stimulation device certified for clinical use, which delivered weak oscillatory current sinus-pulses over four individual periorbital electrodes, respectively (NextWave, Eyetronic, Germany). The four superficial active stimulating electrodes (Grass SAFELEAD^TM gold electrodes, Astro-Med, Inc., RI, United States) were contained in foam-padded stimulation goggles and bilaterally made skin contact via small felt buffers (0.35 cm²) superior and inferior to the eye. The return electrode (rectangular electrode, 30 × 30 mm polished stainless steel) was fastened on the back of the neck at the midline.

Alternating current was applied at 10 Hz, as ACS has shown robust effects at this frequency (Kanai et al., 2008; Helfrich et al., 2014; Vossen et al., 2015) and gold standard PS typically also employs 10 Hz stimulation (Photic driving) (Walker et al., 1944). Stimulation amplitude was set to 120% phosphene threshold (resulting in 351.69 μA (SD 63.95) peak-to-peak amplitude). The phosphene threshold was determined employing an ascending method of limits (Herrick, 1967) provided by the NextWave software. rACS was delivered in 30 s blocks followed by 30 s pauses over 10 min. The subjects were told to keep their eyes open and focus a fixed point on a white surface in 1 m distance for the duration of the experiment.

To assess the safety parameters of stimulation we additionally calculated the effective amplitude. The effective amplitude of the applied current is defined as the time normed integral of the signal, which simplifies to its mean value for discrete signals as is the case here, since the used stimulator receives an equidistant sampled discrete function as input. In the simplest case of a pure sine wave this simplifies to the following formula:

In the case of more complex stimuli such as noise + sine wave or signals with an additional amplitude modulation, the use of peak-to-peak “a(max)” values to describe the resulting electrical power of an electric current stimulation would be misleading.

Regarding charge, we decided to refrain from more complex line integral calculations, and instead used the following simple formula:

This was done to ensure straightforward comparability of resulting values. It also adds to the rigidity of our safety considerations by rather over-than underestimating the injected charge.

Photic Stimulation was applied via two 3 × 5 cm multi-color white LEDs contained in the stimulation goggles, which work via red, green and blue LEDs mixing their emissions to form white light. To be able to compare stimulation intensities with rACS, sinusoidal pulses of white light were applied at an intensity of 120% light threshold and with a frequency of 10 Hz. This threshold was also determined by an ascending method of limits and resulted in an average luminous intensity of 1.24 cd (±0.44) for stimulation.

The stimulation was also delivered in 30 s blocks followed by 30 s pauses over 10 min, and the subjects received the same instructions as for rACS.

The ultra-high resolution head and neck model (MIDA: Multimodal Imaging-Based Detailed Anatomical Model) available through the IT’IS Foundation was used in this study (Iacono et al., 2015). The nifti (.nii) color masks from the MIDA model were first processed in MATLAB to re-create segmentation masks based on intensity values. These masks were then imported into Simpleware (Synopsys Ltd., CA, United States) and any errors in continuity and anatomical details were manually corrected for Datta et al. (2012). Masks with similar electrical conductivities were then merged to a single compartment barring the regions of interest (eye structures) in order to perform individual current flow analysis through them. For instance, mandible, teeth, vertebra, skull dipole, skull inner table, skull outer table, hyoid bone were combined with the skull mask but eye retina, choroid, and sclera were treated as individual masks.

The stimulation electrodes were created as CAD files mimicking the exact physical geometry and dimensions of the electrodes used in the experiments. The electrodes were positioned interactively within the image data simulating the electrode montage used for rACS (see Figure 1C). The adaptive meshes derived from the segmentation masks were then imported into COMSOL Multiphysics (Burlington, MA, United States) for finite element computation. The final model comprised >10 million elements with >15 million degrees of freedom.

The representative isotropic average electrical conductivities assigned to the different tissue compartments and the electrode materials (in S/m) are listed in Table 1.

The Laplace equation was solved and current densities corresponding to 350 μA total current were applied at the anode or active electrode (s). Ground was applied at the return electrode and all other external surfaces were treated as insulated. The linear iterative solver of conjugate gradients with a relative tolerance of 1E-6 was used.

Surface as well as cross-sectional EF magnitude maps on the gray matter, retina, and the optic nerve were obtained. For the scalp, the surface current density magnitude plot was obtained.

The questionnaire we employed is based on the one proposed by Brunoni et al. (2011) and investigated the presence of headaches, difficulties in concentrating, acute mood changes, visual perceptual changes, fatigue and discomforting sensations tingling, itching and/or burning under the electrodes during and after rACS, as well as PS. The item “Difficulties in concentrating” was defined in accordance with Montgomery and Asberg (1979), while the item “Fatigue” was defined in accordance with Chaudhuri and Behan (2004).

We modified the questionnaire by adding a description of phosphenes. Furthermore, to assess the overall tolerability, we defined the broad category of “Pain” as a summary of all discomforting sensations mentioned above and added a Numeric Rating Scale (NRS-11, 11 stages from 0 to 10, 10 being the strongest imaginable pain and 0 the absence of pain) (Farrar et al., 2001) as a more in-depth and reliable measurement (Downie et al., 1978). We discarded the four-point intensity rating for the other categories to avoid a “halo effect” bias (Streiner and Norman, 2008). We assumed that the foreign body feeling reported for physiologically similar transcorneal electrical stimulation (TCES) came from the electrode lying directly on the cornea (Gekeler et al., 2006) and therefore decided not to include it.

Three months after stimulation, the subjects received a second questionnaire to identify late and longer lasting after-effects.

As the data is not normally distributed and equal variance of residuals cannot be assumed, the severity of pain was analyzed in Wilcoxon Signed Ranks Tests for paired samples. The nominally scaled side effects were analyzed in Fishers Exact Tests, as expected values in several of the cells of a contingency table are below the recommended threshold for a classical Chi-Squared Test (Larntz, 1978). P-values of <0.05 were considered significant. All analyses were performed using IBM SPSS Statistics, Version 19.0.0.1 (IBM, United States).

An average 10 Hz phosphene threshold at 290.50 μV (SD 45.36), impedances at 12.05 kΩ (SD 2.89), and an average amplitude of 351.69 μA (SD 63.95) were noted. Calculated from peak-to-peak amplitude, the current density at electrode level amounted to a mean 1.00 mA/cm² (SD 0.28), and the charge density to 0.60 C/cm² (SD 0.11). As sine waves pulses were employed, we additionally calculated the effective amplitude, resulting in a mean 248.68 μA (SD 47.0). Using effective amplitude, current density amounted to a mean 0.71 mA/cm² (SD 0.13), and the charge density to 0.42 C/cm² (SD 0.08). RACS was found to be well within safety limits and the findings comparable to other similar stimulation methods (see Table 2). Regarding stimulation amplitude, rACS (0.35 mA) was comparable to most TCES and tES montages (ranging from 0.08 to 1.2 mA). Electrosleep and the maximum intensity stimulation employed by Gekeler et al. (2006) were found to employ higher amplitudes (3–25 mA). The stimulated area (0.35 cm²) is smaller than most tES montages (16–35 cm²), comparable only to Electrosleep and TCES (0.35–1.25 cm²). Regarding stimulation frequencies, rACS was compared to studies using similar frequencies (10–20 Hz), with the exceptions of Electrosleep, which is set at higher frequencies (100 Hz) as well as the non-oscillating tDCS and Gekeler’s TCES. The calculations following these observations place the charge density of rACS just above the TCES of Ma et al. (2014) and far below the safety limit published by Liebetanz et al. (2009). This is consistent for charge per phase and charge density per phase. Regarding current density, rACS (1 mA/cm²) ranks below Ma (1.2 mA/cm²), well below the maximum intensity employed by Gekeler (8.57 mA/cm²) and far below the safety limit proposed by McCreery (25 mA/cm²). These findings are even more pronounced when using effective amplitude.

The EF distributed by rACS is strongest at the eye level, with the highest current density estimates at the retina. Further areas of elevated current densities are optic nerve and cortex (Figures 1A,B).

The calculated maximum current density at the retina amounted to a maximum of 1.24 A/m², while optic nerve (0.33 A/m²) and cortex (0.13 A/m²) were both subjected to less current flow. Regarding the EF, we estimated a maximum of 2.6 V/m in the optic nerve, followed by 1.99 V/m for the retina and 0.47 V/m for the cortex. Finally, current density at skin level underneath the active electrode amounted to a maximum induced value of 14.79 A/m² (Figure 1C), with the EF estimated at 31.80 V/m. It should be noted that due to edge effects, the observed values are higher than the current density toward the middle of the electrode which is simply the current injected over the contact area. For a detailed view, see Table 3.

Table 4 summarizes the adverse events in the 20 rACS and PS sessions in healthy participants. None of the subjects requested the stimulation to be terminated or required medical attention. In their subjective reports, rACS associated adverse events were predominant during stimulation, and PS associated adverse events were predominant following stimulation. More explicitly, a tingling sensation occurred in 70% of the subjects during but not after rACS. An itching sensation under the electrodes was reported by 30% of the subjects during rACS and 25% after rACS. A burning sensation was felt by 30% of the participants during but not after rACS. Fatigue occurred during, as well as after, stimulation in 35 and 20% of the rACS and PS group, respectively. Headaches were reported only by PS participants during stimulation (15%). After stimulation, it was reported by 5% for both PS, as well as rACS participants. Difficulties in concentrating were reported by 10% of the participants after PS, but not after rACS. There were no cases of acute mood changes, nausea and visual perceptual changes or lasting adverse events 3 months after stimulation.

Forty-percent of the subjects reported pain (Figure 2A) during rACS (mean intensity 2.5, SD 1.73) and 20% during PS (2.75, SD 0.83). While none of the participants reported pain after rACS, this was the case for 10% after PS (1.5, SD 0.5).

In the statistical analyses, rACS and PS showed no significant effect of stimulation type (rACS versus PS) on pain intensity as assessed by Wilcoxon Signed Ranks Tests (Figure 2A), fatigue, headache and difficulties in concentrating as assessed by Fisher’s Exact Tests (Figure 2B) during as well as after stimulation. PS and rACS significantly differed regarding skin sensations of tingling, itching and burning (P < 0.05, Fisher’s Exact Tests), which all occurred exclusively in rACS. For a more detailed view, see Tables 5, 6. The full dataset behind this comparison is available as Supplementary Material.

To be efficacious and safe, a stimulation system must stimulate neural tissue without damaging tissue or electrode. Tissue damage is caused when excitable tissue is overstimulated and electrode damage ensues as metal oxidation occurs (Peterchev et al., 2012). Current density and charge density have been proposed as predictors for such damage (Bikson et al., 2016).

No fatal or serious adverse events (Wester et al., 2008) were observed for rACS. The most notable adverse events in the present study were tingling, burning, itching and fatigue. The hazard rate for these adverse events is to be considered “very common” (>1/10 cases). This is comparable to results from other forms of tES (Brunoni et al., 2011), suggesting for tDCS that the type of adverse event is mild and their frequency of occurrence is “common.” Direct experimental evidence shows significantly more cutaneous adverse events, but significantly less concentration deficits after stimulation for rACS as compared to PS (Table 5).

As the modeling results showed high maximum current densities and EF estimates at skin level, the presence of cutaneous adverse events during and after rACS comes as no surprise. Comparing rACS and PS regarding the summary category of pain, we have to note the complete lack of cutaneous sensations in PS and that multiple aversive sensations may be clustered and perceived in sum total as painful (Tuckett, 1982).