Non-thermal membrane effects of electromagnetic fields and therapeutic applications in oncology

Abstract The temperature-independent effects of electromagnetic fields (EMF) have been controversial for decades. Here, we critically analyze the available literature on non-thermal effects of radiofrequency (RF) and microwave EMF. We present a literature review of preclinical and clinical data on non-thermal antiproliferative effects of various EMF applications, including conventional RF hyperthermia (HT, cRF-HT). Further, we suggest and evaluate plausible biophysical and electrophysiological models to decipher non-thermal antiproliferative membrane effects. Available preclinical and clinical data provide sufficient evidence for the existence of non-thermal antiproliferative effects of exposure to cRF-HT, and in particular, amplitude modulated (AM)-RF-HT. In our model, transmembrane ion channels function like RF rectifiers and low-pass filters. cRF-HT induces ion fluxes and AM-RF-HT additionally promotes membrane vibrations at specific resonance frequencies, which explains the non-thermal antiproliferative membrane effects via ion disequilibrium (especially of Ca2+) and/or resonances causing membrane depolarization, the opening of certain (especially Ca2+) channels, or even hole formation. AM-RF-HT may be tumor-specific owing to cancer-specific ion channels and because, with increasing malignancy, membrane elasticity parameters may differ from that in normal tissues. Published literature suggests that non-thermal antiproliferative effects of cRF-HT are likely to exist and could present a high potential to improve future treatments in oncology.


Introduction
Conventional radiofrequency (RF) hyperthermia (HT, cRF-HT) or microwave (MW) HT has been evaluated in clinical studies [1] and is used in the treatment of soft tissue sarcoma in combination with neoadjuvant chemotherapy [2], and for the treatment of cervical cancer [3][4][5] and superficial tumors [6][7][8][9], when combined with radiation therapy (RT). Annular phased array (APA) systems, capacitive systems (CS), and local RF or MW applicators are common technologies to achieve a specific absorption rate (SAR) of 10-60 W/kg in the target region and intended temperatures of !42 C [10,11]. Current knowledge about cRF/MW-HT is based on numerous preclinical in vitro and in vivo studies mainly using water bath (WB) HT [12][13][14][15][16][17]. Generally, WB-HT kills tumor cells at temperatures !43 C, but a combination of WB-HT with RT and/or chemotherapy might show efficacy at lower temperatures. Overgaard [14] described a relevant sensitization at 41.5 C [thermal enhancement ratio (TER) of 1.34] if RT and HT are simultaneously applied. For sequential treatments however, nearly no cytotoxic effect was observed for temperatures 41 C [17]. Thus, preclinical studies predict that temperatures above 41 C are required for cRF/MW-HT to enhance the RT effects.
However, these requirements have not been usually achieved in all parts of the target for patients with deepseated tumors (e.g., cervical cancer) in successful randomized studies investigating cRF-HT [4,5,[18][19][20][21]. A conspicuous inconsistency between preclinical and clinical data remains as clinical trials using cRF-HT have demonstrated positive results. Improved reoxygenation and higher heat sensitivity of hypoxic (radioresistant) cells have been used to explain the favorable clinical results. Likewise, extreme whole body HT [22], a non-RF-based clinical counterpart of water bath HT, has resulted in disappointing clinical results even though the basic requirements, i.e., covering the target at temperatures >41.5 C are always fulfilled.
Recently, a couple of clinical studies provided evidences of non-thermal effects in human tumors using AM-RF or RF. The researchers applied clinical systems like mEHT (modulated electro-hyperthermia [23,24], TTF (tumor treating fields) [25] and the TheraBionic device [26,27]. These studies have in common according to the general understanding of hyperthermia that temperature increases are below 40 C and the observed effects are to be regarded as non-thermal. Therefore, we hypothesize that additional non-thermal effects of the electromagnetic fields (EMF) exist and are clinically validated. This hypothesis is controversial as it challenges current textbook knowledge and has a high potential to impact clinical practice. Thus, we summarize the current information regarding the non-thermal effects of EMF.

Methods
We found an unmanageable number of hits in Google Scholar when we were searching for non-thermal biological effects of electromagnetic fields, such as 25,900 hits for the search term "non-thermal biological effects of electromagnetic fields" or even 1,900,000 hits for "electromagnetic field effects on cells". For the hyperthermia community, we assessed three clinical approaches (mEHT, TheraBionic and TTF) of maximum interest. Current studies with these approaches have been carefully analyzed to determine whether they provide evidence of non-thermal effects (Section 3). They were considered if they were either randomized or they used the EMF application alone. Hence, we discarded clinical phase I/II studies if EMF was used in combination with any other potentially effective schedule, as such studies might have a strong selection bias, making the clinical outcomes non-evaluable.
Then we took into account all publications that provided preclinical and methodological information for these clinical approaches. That is a considerable but manageable amount of work (Section 4).
In both sections, for clinical and preclinical studies the thermometry is crucial to differentiate between thermal and non-thermal effects. We evaluated not only the accuracy of the temperature measurements in selected points but also analyzed the setup in order to estimate the spatial temperature distribution. In particular, we had to discuss (and to exclude) unrecognized hot spots that are most frequently brought up in the field in order to refute non-thermal effects. If temperature measurements were missing, we attempted a reliable estimation of the temperature from the applied power and resulting SAR (specific absorption rate) distribution.
It is a frequent objection to the existence of non-thermal effects that no convincing physical explanations have been found so far. Thus, the presentation of biophysical and electrophysiological models, which are able to explain such effects, is a central part of this work (Section 5). We emphasize that such a theoretical framework is novel but is deduced from a still incomplete database. Verification or even establishment of a final theory about membrane effects of EMF was neither intended nor possible.

Clinical studies
Regional cRF-HT can reach intermediate temperatures of 39-43 C in larger areas and is suitable for extended cancer diseases. Typically, total powers of >500 W are applied using the most advanced APA systems. The gained temperatures are nevertheless not sufficient, since T 90 , i.e., the temperature exceeded in 90% of the tumor, is below 40 C [4,18].
In the following, we report on three clinical approaches (mEHT, TTF and TheraBionic), which unequivocally demonstrate in our understanding that non-thermal effects should exist. No or incomplete thermometry was performed in the outlined studies. Clinically experienced users of hyperthermia know, however, that with the low power levels applied (as listed below), temperatures high enough for any relevant thermal effect could never be reached.

Clinical application of AM-RF-HT at 13.56 MHz
Szasz et al. [28] provided a summary of the clinical evidence for use of AM-RF-HT, also referred to as mEHT. However, only a few clinical reports are suitable for evaluating the effect and significance of AM-RF-HT. Most notably and very instructive is a randomized study applying AM-RF-HT against cervical cancer [23]. The authors used the AM-RF-HT system, which is basically, a CS, and applied a remarkably low total power of 130 W without conducting any temperature measurement. Lee et al. [29,30] conducted a study on recurrent cervical cancer, utilizing AM-RF-HT with even slightly higher power of 150 W and measured temperatures in and around the cervical tumor, recording a mean temperature of 38.5 C (38-40 C). Thus, we estimate similar or even lower tumor temperatures in the above study that are not efficient in the sense of conventional hyperthermia.
Minnaar et al. achieved enhanced radiochemotherapy efficacy by the inclusion of AM-RF-HT, resulting in minimal cost and effort (low resource setting) without increased toxicity. An additional analysis revealed that effective AM-RF-HT could be employed nearly independently of the body mass index with negligible toxicity [24]. Interestingly, FDG-PET revealed that the majority of patients manifested extrapelvic (mostly para-aortic) lymph node disease ($70%) and were included in the study, while the RT and AM-RF-HT fields were confined to the pelvis. In these patients, complete metabolic remission was significantly more frequent than that in the AM-RF-HTgroup (24.1% vs. 5.6%), suggesting an abscopal effect. These clinical observations are consistent with preclinical studies outlined in the next Section 4.
In comparison, other positive randomized trials on cervical cancer and cRF-HT required significantly higher total power either utilizing the elaborate APA technology with a total power of 400-800 W [3,18] or capacitive technology without AM with a total power of 800-1500 W [4,5]. Recently, an extensive update of heating technology for cRF-HT has been provided [11].
Furthermore, three retrospective reports are of interest in the assessment of the potential of AM-RF-HT. Lee et al. [29,30] treated patients with pre-irradiated cervical cancer recurrences either with chemotherapy alone (n ¼ 20, platinum-based with paclitaxel or 5-FU) or chemotherapy plus AM-RF-HT for 60 min three times each week for a total of 36 sessions in 12 weeks (n ¼ 18). The last follow-up demonstrated a significantly improved response in the AM-RF-HTgroup [50% complete response (CR), 10% partial response (PR), 10% stable disease (SD), 30% progressive disease (PD) versus 15% CR, 15% PR, 5% SD, 65% PD, p ¼ 0.0218]. Both groups were well balanced according to age, FIGO stage, pathology, and recurrent lesions.
Fiorentini et al. [31] applied AM-RF-HT monotherapy to 28 patients with recurrent glioblastoma multiforme (GBM) and 22 patients with recurrent anaplastic astrocytoma (AST) who relapsed after standard first-line treatments (surgery, adjuvant radiochemotherapy and TMZ) [32]. The remaining patients in this retrospective observational multicenter cohort-controlled study (83 GBM, 14 AST) received best supportive care (BSC), including dexamethasone and palliative chemotherapy in 28 patients with GBM. Tumor response after 3 months (CR þ PR þ SD) was significantly higher in the AM-RF-HT group, i.e., 72% (AST) and 54% (GBM) vs. 37% (AST) and 19% (GBM) in the BSC group. Even if we consider a selection bias in favor of patients with AM-RF-HT, we must consider 10/22 objective responses (2 CR and 8 PR) in the AM-RF-HT group with AST (41%) and 7/28 objective responses (1 CR and 6 PR) in the AM-RF-HT group with GBM (25%). Apparently, a mean of 22 (range 11-62) AM-RF-HT sessions alone (150 W for 60 min after a step-up power protocol) can bring about relevant tumor regression. Questions remain due to the retrospective nature of this analysis (patient selection bias) as well as accurate measurement of brain tumor progression using brain MRI (true vs. pseudoprogression).
In the third study, Hager et al. [33] applied AM-RF-HT (70-150 W, 60 min) in 80 patients with liver metastases of far advanced stages of colorectal cancer and retrospectively analyzed the results. For all patients, they started with one course of 8 AM-RF-HT biweekly (without chemotherapy) in the liver and found an overall biochemical response (CEA and CA-19-9 decreased below 75% of initial values) in 25% of the patients. A total of 50/80 patients received only AM-RF-HT, whereas 30/80 patients later received chemotherapy (mytomycin þ 5-fluorouracil þ folinic acid) during the followup period. The authors reported remarkable survival times and objective tumor regression (CT-scans, ultrasound) in these patients. Again, AM-RF-HT alone appeared to cause significant tumor regression.
No firm conclusions can be drawn due to the retrospective design of the above studies. However, the clinical data suggest that AM-RF-HT at moderate temperatures <40 C should have non-thermal effects on various tumors.

Clinical applications of RF-EMF in the intermediatefrequency range 100-300 kHz
RF-EMF in the intermediate-frequency range 100-300 kHz, also known as tumor-treating fields (TTF), was first studied in patients with recurrent GBM and then clinically validated as an additional modality in first-line GBM treatment [25,32]. Treatment was delivered through four transducer arrays with nine insulated electrodes placed on the shaved scalp according to the planning software [34]. TTF was started 4-7 weeks after the end of simultaneous radiochemotherapy (e.g., together with the TMZ maintenance therapy) and was run every day for at least 18 h for a maximum duration of 24 months. The median survival was significantly improved from 16 months (TMZ alone) to 21.9 months in the TTF experimental group (p < 0.001). SAR or E-field distributions in the brain were calculated using the finite-element method [35,36]. Ballo et al. [37] used the Sim4Life quasi-electrostatic solver to ascertain E-field or SAR distributions and verified an association of overall survival and TTF E-field strength, achieving a better outcome for mean E-field strength >100 V/m. The SAR varied in the range of 0.3 to 3.2 W/kg that is not sufficient for any anti-proliferative thermal effect. Even though preclinical findings, dose-response relationships, and clinical outcomes of the study were consistent, the results were called into question. We note that the patients from the experimental treatment group are likely to have received superior medical care, as they were treated mostly within tertiary referral centers, which might also improve overall survival [38]. Furthermore, the long-term and nearly permanent wearing of the "electric hood" is associated with severe restrictions for the affected patients, which limits the quality of life of patients. In addition to newly diagnosed and recurrent glioblastoma, various other oncological applications of TTF are being studied, such as brain metastases, lung cancer, malignant mesothelioma, pancreatic cancer, HCC and ovarian cancer (www.novocuretrial.com). Mun et al. [39] referred to TTF as the fourth modality in cancer therapy and further described the potential for clinical applications on the head, thorax and abdomen. An indisputable advantage of TTF is its applicability to large anatomical regions throughout the human body, while a disadvantage is a need for permanent and long-term exposure to TTF.

Clinical application of AM-RF at 27.12 MHz
Clinical experience with AM-RF at 27.12 MHz, also referred to as the TheraBionic device, is limited to a few case series where the treatment was applied alone in advanced cancer. Barbault et al. [26] described the identification of tumor-specific AM frequencies using a simple RF generator at 27.12 MHz that powers an intrabuccal spoon-shaped applicator. The maximum SAR deposited in the body was estimated to be below 2 W/kg. Therefore, temperature increases in any tissue or tumor are negligible. AM is possible between 0.01 Hz and 150 kHz with a modulation depth of 85%. Variations in the radial pulse amplitude are claimed to indicate a tumor-specific AM frequency, which seems surprising. However, Sharma et al. [40] recently explained that the Ca 2þ channel CACNA1H might be expressed in tumors and small vessels (endothelial and smooth muscle cells) that could affect the pulse amplitude. In a total of 163 patients with various cancers, Barbault et al. discovered 1524 modulation frequencies between 0.1 and 114 kHz using their bio-feedback technique [26]. HCC, pancreatic cancer, ovarian cancer, BC and prostate cancer commonly exhibited frequencies between 1.8 and 10.5 kHz. Typically, more tumor-specific frequencies were found in advanced and progressive tumors. A total of 28 patients with advanced cancer received a low-level AM-RF at 27.12 MHz, which implies a self-administered treatment of 60 min, three times a day for a mean treatment duration of months (range, 1-50.5 months). For 7/28 patients (25%), complete remission (1/28), partial remission (1/28) or stable disease (5/28) were documented according to the RECIST criteria. These findings suggest non-thermal therapeutic effects of AM-RF with tumor-specific AM frequencies.
Costa et al. [27] conducted a phase I/II study in 41 patients with advanced HCC, who had either failed standard therapy or had severely impaired liver function limiting any kind of systemic treatment, and used AM-RF at 27.12 MHz as monotherapy. Of these patients, 4/41 had partial remission and 16/41 stable disease for !12 weeks (49% long-term control), with 4/41 patients becoming long-term survivors (>3 years). These data confirmed the potential of using AM-RF at 27.12 MHz.

Preclinical studies
Numerous preclinical studies have investigated the non-thermal effects of EMF. Investigations mainly dealt with the possible risks of EMF at low levels without a relevant temperature increase. Epidemiological studies [41,42] have focused on EMF either in the extremely low-frequency range (ELF; <3 kHz) or in the RF-range (3 kHz up to 300 GHz). EMF exposure at ELF is ubiquitous and is mainly caused by electric lines and associated grounding systems, while wireless communication applications are the source of RF exposures. Organizations such as the World Health Organization and the International Commission of Non-Ionizing Radiation Protection stated that although epidemiological studies have suggested links between EMF exposure and health effects, they could not provide a reliable interpretation [43][44][45][46]. Forschungsgemeinschaft Funk thoroughly examined the possible non-thermal effects of weak EMF (http://www.fgf.de/ index2.html) from 1992 to 2009 in Germany and published the results in 2009. The REFLEX (Risk Evaluation of Potential Environmental Hazard from Low Energy Electromagnetic Field Exposure Using Sensitive in vitro Methods) study was an elaborate European union (EU) project from 2000 to 2003 involving 12 research groups from seven countries. Standardized exposure systems for EMF magnetic fields [47] and RF exposures at 900 MHz [48] and 1800 MHz [49] were constructed according to development guidelines [50]. Nevertheless, the results were highly controversial and not reproducible, indicating the difficulties in obtaining clear conclusions in this field of research.
In the actual extensive "Handbook of Biological Effects of Electromagnetic Fields" (two volumes with 1150 pages) the technical term "non-thermal" is not even mentioned [51,52] but the issue was discussed as controversial [53,54].
In the previous Section 3, we listed recent clinical studies (mEHT, TTF and TheraBionic) that strongly suggest the existence of non-thermal effects. As a result, we focused on the associated preclinical investigations, which have provided the prerequisites for these clinical applications. It is precisely this recent clinical and preclinical work that is most likely to contribute to the controversial discussion.
The biological endpoints of cells or animal tumors are an important issue. Clearly, the clonogenic assay is the gold standard if radiotherapy dosages or thermal doses are directly transferred to the clinics. However, the question here is whether there are non-thermal effects and which mechanisms are underlying. Then other biological endpoints than clonogenicity are also of interest, e.g., apoptosis, damage-associated molecular pattern (DAMP) and others. In spite of the huge number of literatures on this issue, we are not at the point to formulate a dose concept for non-thermal effects.
4.1. Amplitude-modulated (AM)-RF-HT at 13.56 MHz Figure 1 depicts the experimental setup for evaluating the biological effects of AM-RF-HT on cell suspensions as the preclinical counterpart of mEHT or oncothermia; the findings can easily be compared to the effects of WB-HT [55,56]. In parallel to the clinical application, the RF source generates a carrier frequency of 13.56 MHz with an adjustable AM starting at a given minimum modulation frequency f and then declining with 1/f (pink noise) with a selected modulation index m (LabEHY-200, Oncotherm Kft., Budapest, Hungary) as described in detail in [57]. Temperatures are typically measured in the center of probes of <1 ml (dimensions <1 cm) and in the surrounding distilled water by fluoro-optic sensors. Then the probe temperature represents the maximum temperature and is controlled by the RF power. Temperature deviations in the probe of a few tenths of a degree are possible but are further reduced if the surrounding water is warming up near to the probe temperature. However, the crucial parameters SAR or E-field in the specimen are not always clearly defined in preclinical studies with this applicator (see Figure 1 legend). SAR can be determined by measuring the rate of temperature increase over time ( C/min).
The bioheat-transfer equation predicts SAR of 200 À 300 W/kg to heat a sphere of 1 cm diameter from 37 C to %42 C. The SAR declines below 100 W/kg when the ambient temperature rises above 40 C [58]. That is the typical Figure 1. Design of an experimental setup to expose a cell suspension (green) to AM-RF-HT [55]. A mean E-field emerges between the electrodes of the capacitor C generated by the voltage V C . The capacitor C in series with an inductivity L (with loss resistance R) establishes a resonance circuit with resonance frequency RF [56]. It is desirable that temperature T (set and maintained using a flow system) and SAR (set by V C ) can be selected separately. If the resonance condition f ¼ 2p(LC) 1/2 is met, the system is well matched, i.e., in resonance.
description of an experiment according to Figure 1-not far away from the clinical situation [55]. The formation of selective small-volume overheating in regions of approximately 1 mm extension would require excessive SAR > 10,000 W/kg or even SAR >1,000,000 W/kg for areas of 0.1 mm extension etc. -as outlined in a series of publications about point heating or nanoheating [58][59][60][61][62]. We note that a simple arrangement of a 1 ml cell suspension in a capacitor cannot generate such high SARs that would be required to generate hot spots on a microscopic level.
Another arrangement is applied for in-vivo experiments with mice tumors. Here, an electrode of 1.8 cm diameter is positioned on tumors of a few millimeter size with a neutral electrode on the opposite. As illustrated by [63] SAR between 100 À 400 W/kg are generally required to heat such small tumors to 40 À 42 C in accordance with our own experience (unpublished results).
Obviously, in case of mEHT, the preclinical studies use 10 times higher SARs than clinically attainable resulting in only 3 times higher E-field strengths. Higher electric field strengths can facilitate the detection of non-thermal effects and lead nevertheless to valid results if relevant thermal effects can be excluded.
The data show higher rates of apoptosis for AM-RF-HT (typically 42 C for 30-60 min) in comparison to standard WBHT at the same temperature [64]. The greater effectiveness of AM-RF-HT compared to infrared (IR) HT for nude mice xenografts in vivo [65] or WBHT for cell suspensions (human cancer cell lines) in vitro [66][67][68] has been demonstrated by microscopic detection of dead cells or apoptotic cell death. A 30 min treatment with AM-RF-HT at 42 C was 3.2-fold more effective than IR-HT at 42 C. Notably, 30 min of AM-RF-HT at 38 C was still 2.6-fold more effective. In cell suspensions, the apoptosis-inducing effect of AM-RF-HT at 42 C was similar to that of WBHT at 46 C. Apoptosis was associated with increased caspase activation, calreticulin expression, and intra-/extracellular HSP70 expression. Andocs et al. [69] found a complex dependency of AM-RF-HT efficacy on temperature, i.e., increased efficacy from 39 C to 42 C (45% dead cells) and then a (slight) decrease at 43-44 C (30% dead cells) suggesting that for temperatures <43 C, AM-RF-HT generates considerably more dead cells than WBHT, although the temperature dependency for AM-RF-HT and WBHT might be different.
Researchers consistently found high rates of apoptosiswhich were markedly higher than those expected for traditional heat treatment at 42 C-in comparison to the control groups. Further, several indicators of cellular stress, such as elevated HSP70, calreticulin, HMGB1 and damage-associated molecular pattern (DAMP) signals are associated with AM-RF-HT alone [63,72,75,77]. AM-RF-HT induced significantly higher HSP-levels than WBHT at the same temperature [67]. It is known that DAMP signals can support immunogenic cell death (ICD) in response to chemotherapy, radiotherapy, immunotherapy or immune checkpoint inhibition therapy [78,79]. In fact, for symmetrical tumors (both femoral legs in mice), treatment-related effects were also observed in untreated tumors (compared to the sham controls) [75,80]. In a CT26 murine colorectal cancer model, Tsang et al. [66] found a synergism between intratumoral dendritic cell injection and AM-RF-HT, and a systemic effect (prevention of contralateral tumor seeding). Thus, AM-RF-HT can support local and systemic ICD-mediated tumor killing. The abscopal effects of AM-RF-HT were clinically confirmed [23].
Combination treatment of AM-RF-HT and RT [81][82][83][84] was found to result in greater tumor damage, particularly if AM-RF-HT was immediately-or at least within 4 h-followed by RT in fibrosarcoma allografts (in C3H mice) and various cancer cell lines (9 L, MCF-7, Panc-1). Interestingly, the opposite order (RT followed by AM-RF-HT), which is common for cRF-HT, was less effective.
Interestingly, most available references do not specify the parameters of the AM (f and m) adjusted for the AM-RF-HT. We determined a typical default setup in the clinical device EHY 2030, amounting to f % 60 Hz and m ¼ 0.2, and comparable parameters are assumed for the preclinical devices (own unpublished measurements). Wust et al. [55] detected already for RF-HT (without additional AM) more antiproliferative effects than those in WBHT (80-100% vs. 45-50% clonogenicity in HT-29 and SW480 cells). This is in line with our experiences with cRF-HT. The importance of any type of AM is currently under investigation. It should be noted that different biophysical mechanisms may underlie RF-or AM-RFinduced non-thermal effects (see below).

RF-EMF in the intermediate-frequency range 100-300 kHz
Pairs of insulated wires or electrodes are suitable for generating low-intensity RF-EMF of 100-300 kHz in cell cultures or tissues. In vitro studies on eleven different types of cancer cell lines and in vivo studies on mice tumors showed arrested proliferation, cell destruction, and reduction in tumor growth [86]. A 24 h exposure of EMF with moderate strengths of 100-140 V/m (comparably low SAR of approximately 5-10 W/ kg without considerable temperature elevation) resulted in a significant decrease in growth rate (relative to that of the control) of 10-90% depending on the cell line. This application was labeled as tumor treating fields (TTF). Antiproliferative effects increased with E-field strength and depended on frequency and cell line (e.g., maximum for B16F1 cells at 120 kHz and F-98 cells at 170 kHz). Kirson et al. [86] identified two E-field effects, which are not temperature-dependent. First, the E-field polarizes and orients tubulin dimers and other charged macromolecules in the direction of the E-field during the early phases of mitosis, which interferes with proper chromosome alignment and separation. Second, during cleavage, the E-field is strongly inhomogeneous, exerting an electric force on all intracellular charged and polar macromolecules. Since dividing cells are disturbed and sometimes destroyed, the authors suggested a novel treatment modality for cancer. Another paper extended the investigations to additional cell lines and animal tumor models [87]. The combination of chemotherapy (paclitaxel, doxorubicin, and cyclophosphamide) with TTF resulted in strong sensitization [88]. Silginer et al. [89] supplemented preclinical information for several glioma cell lines and identified 100 kHz as the most effective frequency for glioma. TTF also hampered the invasiveness and migration of glioma cells. Importantly, glioma cells responded to TTF independent of MGMT status and/or TMZ resistance.

AM-RF at 27.12 MHz
Barbault et al. [26] exposed cancer patients to low-level AM-RF (<2 W/kg) with a carrier frequency of 27.12 MHz and selected modulation frequencies from 0.01 Hz to 150 kHz with a high modulation index of 0.85, which does not cause any temperature increase. This system is also referred to as TheraBionic TM (Ettlingen, Germany). The authors briefly described a bio-feedback method (measuring skin electrical resistance, pulse amplitude, and blood pressure) and identified a variety of cancerspecific modulation frequencies (see also [90]). This low-level AM-RF was evaluated in three preclinical studies. Zimmerman et al. [91] utilized cancer-specific modulation frequencies identified in hepatocellular carcinoma (HCC) and breast cancer (BC) cell lines. They described the antiproliferative effects associated with changes in gene expression in two representative HCC cell lines (HepG2 and Huh7) exposed to AM-RF for 21 h in seven days (3 h per day) or longer with HCC-specific AM frequencies. A commonly used tritium incorporation assay determined the growth inhibition that persisted over a period of seven weeks. We highlight some interesting results of this early preclinical study. First, significantly lower growth inhibition was measured using randomly chosen (nonspecific) or BC-specific AM against HCC cells, instead of HCCspecific AM. Second, HCC-specific AM did not affect the proliferation of BC cells (here MCF-7) or normal hepatocytes (THLE-2). Third, within the entire SAR range (from very low levels of 0.05 to 1.0 W/kg), they measured significant antiproliferative effects. Fourth, growth inhibition depended on the exposure time. They did not observe any effect for reduced exposure times such as 1 h per day for seven days or 3 h per day for three days but achieved the same level of effect for longer exposure times such as 6 h per day for seven days. Application of HCC-specific modulation frequencies for more than 21 h resulted in mitotic spindle disruption in more than 60% of HCC cells. Therefore, for both approaches (TTF and TheraBionic), a minimum exposure time was required to achieve a significant antiproliferative effect but this effect did not always increase with prolonged exposure time.
These results were confirmed and explored in two recent studies. Jimenez et al. [92] exposed additional HCC cell lines (HepG2, Hep3B, Huh7, HCCLM3 and MHCC97-L) to the same AM-RF (27.12 MHz with low SAR between 0.03 and 0.4 W/kg) for 3 h daily for seven days. They found antiproliferative effects only for AM-RF with HCC-specific AM. Interestingly, growth inhibition was associated with increased intracellular Ca 2þ resulting from Ca 2þ influx through a special T-type low voltage-gated Ca 2þ -channel (VGCC) CACNA1H [93][94][95], which appeared to be selectively sensitive to HCC-specific modulation frequencies.
In vivo experiments with mice carrying human HCC xenografts confirmed the inhibition of tumor growth by AM-RF exposure with HCC-specific AM.
Sharma et al. [40] described the same pattern for human BC cell lines (T47D, BT-474, SKBR3, MDA231 and MDA-MB-453) with different BC-specific AM. Additionally, after intracranial injection of HER2-positive or triple-negative BC cells into SCID mice, the implanted brain metastases were exposed to AM-RF with the same low SAR (0.03-0.4 W/kg) 3 h each day until the experimental endpoint. For both cell types, they found significant suppression of metastases and tumor growth, resulting in significantly improved brain metastasis-free survival. Again, AM-RF-EMF with BC-specific frequencies in BC cells was associated with increased Ca 2þ influx through the same CACNA1H channel. However, an exact biophysical mechanism has not yet been described.

Model approaches to describe non-thermal antiproliferative effects
Biophysical mechanisms were not presented to the best of our knowledge until now to explain the controversial nonthermal effects, in particular, in the clinically relevant RF range from 1 to 150 MHz. This lack of theory building is a common argument against the existence of non-thermal effects. In fact, works that refute the existence of non-thermal effects are predominantly of a theoretical nature [43,44,46,[96][97][98][99][100]. How else can you deal with something that does not exist? Predominantly in past experiments, higher frequencies (microwaves up to terahertz) were investigated with regard to their non-thermal resonance effects [101,102]. Therefore, theoretical considerations mentioned above have attempted to refute such resonances in macromolecules, e.g., DNA or enzymes [44,[97][98][99]103]. These high frequencies (>300 MHz) are, however, of minor importance for clinical hyperthermia because of their limited penetration depth.
In the following section, we focus on the membrane effects of EMF in the lower RF range (<100 MHz).

Nanoheating
The term nanoheating (formerly known as point heating) refers to the deposition of a very high SAR in a microscopic volume, e.g., a protein cluster (nm to mm size) in the cell membrane or the cytoplasm. This theoretically fascinating therapeutic approach occupied many researchers in a series of publications, beginning at the time of World War II [58][59][60][61][62]. The underlying problem can be solved by an analytical solution of the partial differential equation for the temperature distribution, which ultimately represents a diffusion equation with a transport term (perfusion) known as the bioheat transfer equation. In the cited works it was consistently shown that a circumscribed SAR increase in regions <1 mm must be unrealistically high in order to generate a thermal effect (e.g., SAR > 10,000 W/kg for a 1 mm sphere). On the other hand, with realistic SAR increases, only minimal temperature increases of fractions of 1 C are expected. In conclusion, non-thermal antiproliferative effects, at least in part, must exist to explain preclinical and clinical data.
Nanoheating is specially investigated after the administration of nanoparticles in the target and exposure to an alternating magnetic H-field [104][105][106]. Analysis of available ferrofluids and tolerated H-fields (frequency and amplitude) shows that effective heating of lesions <1 cm is difficult. Thus, nanoheating, without nanoparticles, appears even more difficult.
Based on current technology, typical incident E-field intensities under clinical conditions are in the range of E ¼ 50-350 V/m corresponding to SAR ¼ 1-60 W/kg. Uncomplicated and cost-effective devices achieve a low SAR of 5-15 W/kg (E ¼ 100-170 V/m). These E-fields are negligible compared with the E-fields between the inner and outer surfaces of the lipid bilayer of E > 10 6 V/m associated with the resting potential of approximately 30-100 mV at membranes of 4-5 nm thickness. However, the distribution of electrical constants such as the relative permittivity e r and conductivity r (in S/m) is inhomogeneous in living tissues and might generate higher E-fields caused by electrical boundaries compared with the incident E-field [107]. The field distribution is determined by solving Maxwell's equations, which is certainly a challenge at the microscopic level. In particular, Maxwell's 4 th equation causes elevations of E at the electrical boundaries.
Eibert et al. [108] investigated this issue and performed an electromagnetic and thermal analysis of an exposed structure depicted in Figure 2. They showed that with such arrangements, voltages of a few millivolts arise on and in the membrane, which is in the percentage range of the resting membrane potential. In Figure 2, the electrical constants e r and r are taken for the frequency range of !70 MHz. Figure 3 schematically shows the frequency-dependent behavior of e r in living tissues [109,110], indicating the range of cRF-HT with e r $80 (above 70 MHz). The so-called d-dispersion formed by proteins, protein-bound water and organelles [110] leads to this well-known plateau representative of APA technology . However, the d-dispersion also causes a 10-fold higher e r of $1000 at 10-30 MHz (of AM-RF), while the aand b-dispersions result in a 1000-fold higher e r > 100,000 at intermediate frequencies (100-300 kHz). Such a high e r might create a large E of 500 V/m (corresponding to local SAR of 120 W/kg) in certain microscopic areas (as shown in Figure 2) if the AM-RF-HT-technique at 13.56 MHz is applied (or other CS around 10 MHz). This fact might partially compensate for the technological advantage of the APA technique at RF around 100 MHz.
In the intermediate-frequency range, local E-field peaks along the membranes can be large ($100,000 V/m) in addition to the described specific mechanisms during cell division. This might be a further explanation for the effectiveness of RF-EMF in the intermediate-frequency range.
Papp et al. [111] investigated the power absorption in transmembrane protein clusters (rafts) with an assumed diameter of 1 mm and enhanced permittivity and conductivity in comparison to the adjacent membrane layer. They created a 3D model of the raft embedded into the membrane and simulated the E-field distribution using CST EM Studio (Computer Simulation Technology software, Darmstadt, Germany). Assuming realistic excitations of 100 V/m E-field elevations up to 300 V/m (SAR of 40 W/kg) occurred at the Figure 2. Three-dimensional configuration analyzed by Eibert et al. [108] using sophisticated numerical methods. They found for this special arrangement that the high electrical jump between the extracellular fluid (e r ¼ 70, r ¼ 1.2 S/m, blue) and fatty or connective deposits (e r ¼ 3, r ¼ 0.002 S/m, yellow) cause high E-fields of some 1000 V/m (and SAR of 500 W/kg) at a lipid bilayer (membrane) between. Voltages as high as millivolts can be generated in these membrane sections. This analysis is valid for RF > 70 MHz (see dispersion curve in Figure 3). rafts and are probably of minor effect. We note that the authors [111] assumed a much higher excitation of 50,000 V/ m and resulting SAR peaks that are clinically unrealistic.
In summary, current theoretical approaches do not indicate that energy transfer by EMF to biological macromolecules can disturb their function or even cause chemical conversions.

Biophysical processes on transmembrane ion channels
Preclinical and clinical experience with AM-RF-HT at 13.56 MHz and AM-RF at 27.12 MHz suggest that non-thermal mechanisms exist, which stress and finally damage the exposed cells. Suitable AM might enable or at least enhance these processes. The non-thermal effects of AM-RF could arise from electrical disturbances at the membranes. Therefore, we examined the ion channels more closely. We validated the non-thermal effects of RF at 13.56 MHz without AM in a typical preclinical experiment using LabEHY-200 [55]. For a better understanding, we developed a model for ion channels based on textbook knowledge [112][113][114][115]. Figure 4 shows an example of a Ca 2þ channel, which was proven to be particularly important for non-thermal effects. This model is also extrapolatable to potassium (K þ ), sodium (Na þ ), chloride (Cl -) or hydrogen (H þ ) ions. The basic elements of a typical ion channel are the entrance (here outer) pore, which has a mechanism for opening and closing (gating) with some selectivity for the particular ion (here Ca 2þ ), a cavity to collect the ions, and the selectivity filter as the most important and highly specific part for the particular ion. The ion gains the dehydration energy by a specific chemical reaction with carbonyl oxygen atoms lining the selectivity filter, stalls in the filter, and can only exit by exploiting electrostatic repulsive forces if the next ion enters from the cavity [112].
We assume that every channel act like a diode because ions can only be directed along the electrochemical gradient. Our example ( Figure 4) shows a huge Ca 2þ gradient between the extracellular concentration of 1.3 mmol/l and intracellular concentration of only 100 nmol/L (>10,000-fold). Then, the equilibrium (Nernst) potential for Ca 2þ is U Ca % 250 mV, which is far from the typical membrane potential of -60 mV. This electrochemical gradient is enormous, causing a strong tendency for calcium influx and intracellular calcium overload if the calcium channel is open. Instead, in the case of potassium, the K þ equilibrium (Nernst) potential of U K ¼ -95 mV is below the membrane potential of 60 mV. Therefore, K þ ions flow from intracellular to extracellular to bring the membrane potential nearer to the K þ equilibrium potential. Thus,  Model of a calcium channel to estimate R Ca and G Ca that we used as a template for ion channels [112]. The ion channels act like diodes with forward direction along the electrochemical gradient (for calcium from extra-to intracellular). The selectivity filter is magnified for clarification. There is a large electrochemical gradient toward intracellular due to the difference between the membrane potential of -60 mV and the Ca 2þ equilibrium potential of 250 mV. High concentration gradients >10,000 predispose cells to excessive calcium influx. Thus, disequilibrium of intra-/extracellular calcium concentrations is more probable for calcium than for other electrolytes. a potassium channel acts like a diode with the forward direction from intra-to extracellular.
The resistances of ion channels can be estimated using basic electrical principles (e.g., R Ca ¼ 323 GX or R K ¼ 4.3 GX, see [55]) and the surrounding membrane contributes a capacitance C. Figure 5 (right) shows the equivalent circuit diagram, which is basically a diode in series with a low-pass filter RC (with resistance R and capacitance C). Thus, the voltage associated with the RF E-field along the channel (i.e., perpendicular to the membrane), is rectified and smoothed. In the case of a pure sinusoidal RF wave (E ¼ 200 V/m assumed), a direct voltage of $1 mV is emerging (see dotted line). Additional AM with a single modulation frequency, after demodulation, yields the solid sinus curve overlaying this direct voltage. Direct voltage and superimposed alternating voltage might exert different effects on the channel (see below). In certain regions with electrical inhomogeneities, E-fields can be higher (see the previous section) and the generated direct voltage can approach the millivolts range under special conditions ( Figure 2) [108].
While the ion flux through a single channel caused by some microvolts to millivolts might be considered ineffective, a mean channel density of 1 channel/mm 2 (i.e., 2000 channels in the membrane that point in the direction of the E-field) and an exposure time of 30-60 min (1800-3600 s), can enhance the stress on the cell by a factor >10 6 . This contrasts with the fact that the channels are often (but not always) closed.
Even such a small voltage offset of 1 mV along >1000 ion channels for >1000 s will cause a relevant disequilibrium of ions in the affected cells, which can be much higher than the complete ion content for sodium, chloride, and protons and >100-fold larger than the intracellular calcium content. Table 1 (1 st column) shows the relative intracellular ion increase resulting from Ohm's law. However, it is supposed that mostly, fast chemical reactions control the flux through the ion channel (facilitated diffusion), which considerably accelerates the exchange (2 nd column). A DC voltage of 1 mV (or the associated E-field of 200 V/m) causes a drift velocity v X of the ion X. Then, we estimate the maximum flux through the selectivity filter dividing v X by the hydration diameter d X , if we assume that the ions pass the channel (and, in particular, the filter) in a single file with maximum rapidity. Wust et al. [55] provided further details of the models, basic equations, and calculations. Table 1 summarizes the results and suggests that increased intracellular Ca 2þ most likely explains non-thermal damage to cells. Indeed, researchers reported increased intracellular calcium in association with non-thermal antiproliferative effects of AM-RF-HT [68], AM-RF at 27.12 MHz [40,92], and RF-EMF in the intermediate-frequency range [116]. Researchers identified VGCC of the T-type CACNA1H (in HCC or BC) or L-type CACNA1C (glioblastoma) as sensitive to EMF used with AM-RF at 27.12 MHz or RF-EMF in the intermediate-frequency range and causal for intracellular Ca 2þ  .g., potassium) showing the internal pore, cavity and selectivity filter. This model is an archetype for other ion channels. Right: Equivalent circuit diagram of an ion channel according to the ion channel model of Figure 4. After rectification of the amplitude-modulated (AM) carrier radiofrequency (RF) ($10 MHz) a low-pass filter (specified by RC) generates a mean DC voltage (1 mV for SAR of 25 W/kg) with a superposed sinus oscillating with the modulation frequency in the kHz range (red solid line). For RF of 10 MHz the condition RC ) 10 À7 s holds. Table 1. Estimated fluxes through ion channels caused by 1 mV either derived applying Ohm's law or assuming maximal facilitated diffusion (passage in file) and their impact on ion imbalances assuming 1000 channels and 1000 s exposure time resulting in an enhancement factor 10 6 . The highest imbalance is expected for Ca 2+ (bold).
Ohm's law  Table 1 (bottom  line). Interestingly, T-type VGCCs require only a moderate depolarization of 10-20 mV for activation/opening [118] and are probably particularly sensitive to influences on the cell membrane. An overload of intracellular calcium and protons and/ or a decline in the intracellular pH will certainly harm the cell. Additionally, intracellular overload of NaCl may cause cell edema/swelling, leading to cell death, and intracellular loss of K þ may promote apoptosis [119]. Ion channels provide the key to understanding the non-thermal effects, even though the exact biochemical mechanisms are unknown.
Abnormalities of ion channels in tumor cell lines have been described in numerous publications [119][120][121][122][123][124][125][126]. Ashrafuzzaman's textbook (Chapter 8) provides a summary [127]. In tumor cells, with increasing aggressiveness, certain channel families are often overexpressed, e.g., transient receptor potential (TRP) channels. They are responsible for calcium transport, proton transport (pH regulation), and water exchange (aquaporins). T-type calcium channels are broadly expressed in various malignancies [93]. Differences in ion channels (channelomes) in tumors and normal tissues can contribute to an increase in the therapeutic quotient. Both the duration and sequence play a decisive role in the non-thermal effects and may differ significantly from cRF-HT.

Frequency-dependent effects on cells
Frequency-dependent effects caused by particular modulation frequencies on an RF carrier were described by two research groups more than 40 years ago [128,129]. Both described a calcium release in brain tissue caused by AM-RF of 147 MHz with an ELF of 10-20 Hz at low intensities of $1 W/kg; in a control without AM, they found no effect. Later, researchers described comparable effects in brain tissue exposed to ELF directly without a carrier [130] and noted its dependence on temperature [131]. These results are controversial [96,100,[132][133][134]. We note that in some refuting studies pulsed microwaves were used, which is clearly not the same as AM-RF (see below).
An analysis of the experimental data reveals a non-thermal gain in effectiveness through AM of the RF with modulation frequencies hertz to 20 kHz (audio frequencies). Preclinical and clinical experience with the AM-RF-HT indicates effectiveness for lower AM frequencies in the range of 100 Hz to a few kilohertz considering the pink-noise approach, which typically starts at a minimum frequency of 100 Hz. Instead, experiences and data using AM-RF at 27.12 MHz predominantly indicate frequencies >1 kHz (up to 20 kHz). Increased efficacy with duration and number of applications is always observed.
For a reasonable explanation of the resonance frequencies generated by AM-RF-HT on cells (Hz-20 kHz), we refer to our model of an ion channel. Because the dehydrated ions are trapped in the selectivity filter, every periodic force is transmitted to the adjacent membrane. As depicted in Figure 6, the membranes consist of vibratory sheets or tiles, which are isotropic or at least six-fold symmetric. Such membrane Figure 6. Schematics of a single section of the cell membrane that is excited by the ion channel to vibrate. This membrane section perpendicular to the incident wave behaves like a sheet a Â b with supported edges. Such a vibratory structure can be excited to natural bending oscillations in the frequency range from hertz to kilohertz depending on the dimensions a, b and the stiffness of the membrane described by the Young's modulus Y (see text). Excitable tiles with larger extensions can be composed of these sections with an increased number of exciting ion channels. The Eigen frequency declines with increasing dimensions a and b, given in mm.
portions are characterized by only two elastic parameters [135,136]. The first parameter is the volume compression modulus K V , which amounts to 10 7 -10 8 Pa. These consistently high values of the compression modules for lipid bilayers and cell membranes reflect the incompressibility characterizing both solids and liquids [135]. The second parameter describes the fluidity (elasticity) of membranes and may vary considerably (see below). Poissons ratio r is a suitable parameter to describe fluidity. For r ¼ 1/3, the stiffness is high and describes a solid. Instead for r ) 1 = 2 , the membrane increasingly takes on the properties of a liquid. The corresponding elasticity parameters are either Young's modulus Y (also called stiffness) or shear modulus m (also called torsion). The basic relationships between elasticity parameters are derived according to [136 §5]: Both decline (and finally approach zero) in comparison to K V as r goes toward 1 = 2 .
Assuming the Young's modulus Y of a membrane section in Figure 6, the resonant frequencies for the bending vibration of this sheet with supported edges can be calculated accordingly [136, § 25]: with Y in Pa, a and b in mm. For n ¼ m ¼ 1, we obtain the fundamental oscillation and for n, m > 1 high-frequency harmonics.
For Y ¼ 500 Pa and a ¼ b ¼ 10 mm, we calculated a fundamental resonant frequency f ¼ 53 Hz assuming a vibration of the whole membrane face. If only parts of the membrane are excited, for example, a ¼ b ¼ 1 mm, we calculate f ¼ 2.7 kHz. Depending on the stiffness Y [Pa] of the membrane and the extension a, b [mm] of the oscillating sheet, we predict the resonant frequencies in the hertz and kilohertz ranges that match the modulation frequencies used in the previous sections.
The mechanical properties of tumor cells and normal cells may strongly differ [137,138]. Generally, tumor cells are more deformable and elastic in correlation with higher adhesion, migration, invasiveness, and metastasis formation. Table 2 shows the upper lines for a typical K V ' 10 7 Pa that theoretically, the calculated Y of a cell membrane (using Equation 1) decreases from very high Y ¼ 10 7 Pa (for r ¼ 1/3 indicating a behavior like a solid) down to Y ¼ 600 Pa, if r ¼ 0.49999 describes a fluid-like behavior. The measured Y in Table 2 demonstrates that we may expect Y a few thousand Pa for normal cells and only a few hundred Pa for the corresponding tumor cells [139][140][141][142]. Generally, the stiffness decreases with increasing aggressiveness and invasiveness of tumor cells.
In summary, the membrane stiffness Y of aggressive tumor cells ranges from 200 to 800 Pa, leading to resonant frequencies between 10 Hz and 10 kHz (see above), while normal cells do not resonate or have less resonance in this frequency window. Furthermore, because of the typical microenvironment, tumor cells are isolated and surrounded by extracellular fluid.
Such single cells are more susceptible to an incident AM-RF and show the theoretically predicted resonance behavior. In comparison, normal cells are connected to one another in a tissue and show a different-presumably significantly lower resonance behavior-at higher frequencies. The different mechanical properties of tumor cells and tissues can increase the therapeutic ratio of AM-RF if an appropriate modulation frequency spectrum is selected; data suggest that lower frequencies <1 kHz are more effective.
We hypothesize that a large number of ion channels demodulate the AM-RF and transfer these low-frequency oscillations (in time to the AM frequency) mechanically to the membrane. Then, parts of the cell membranes or the entire cells oscillate in resonance with considerable displacements. In extreme cases of mechanical stress, the membrane might rupture, creating a hole. Such a long-dated water channel between the extracellular and intracellular space will cause ion fluxes along the concentration gradient that predominantly increases the Ca 2þ load in the cell, but might also lead to other ion imbalances (Na þ , Cl -, K þ and pH). Less pronounced membrane vibrations at lower SAR levels might cause depolarization of the membrane potential or directly open ion channels by distortion of the voltage sensor or gate. Likely, T-type VGCCs that have already been identified open at low depolarization (of only 10 À 20 mV) with consequential calcium influx into the cell (see Table 1). Due to these oscillations, we expect a variety of membrane changes, especially on ion channels, when we examine different levels of SAR and different frequencies. Further systematic investigations with different parameters of AM-RF-HT are therefore worthwhile. In particular, we expect a strong interdependence of membrane properties on temperature and future experiments should examine these interactions. Relevant changes in the membrane properties, e.g., decreasing stiffness Y, are expected even at low temperature increases to 38 or 39 C.

Discussion and prospects for oncologic applications
Clinical and preclinical data suggest the existence of antiproliferative non-thermal effects (Sections 3 and 4) [143][144][145][146][147] applying the mobile-specific frequencies around 900 MHz, 1800 MHz or 1950 MHz. The investigators found several biological effects, in particular, DNA damages that they attributed to the functionality of DNA as a fractal, i.e. broadband, antenna [148]. Yakymenko et al. [149,150] identified oxygen species overproduction reactive to mobile-specific EMF as a major correlate, and described several possible mechanisms in their reviews. Panagopoulos et al. [151] elaborated the concept that free ions on both sides of the cell membrane can oscillate and might cause membrane disorders. Interestingly, most of these studies confirmed our educated guesses that modulated and/or pulsed fields are more effective than non-modulated sinusoidal MW, e.g., [143,144,146]. These researchers confined themselves to the risks of mobile phones and did not discuss possible therapeutic applications. Some of their results concerning modulation/pulsation are nevertheless of interest for therapeutic applications. Note that these frequencies have only minor applicability for therapeutic purposes because of their limited penetration depth.
Our overview is intended to show that non-thermal effects exist in the RF range of 10 À 30 MHz that can be exploited for therapeutic purposes as a further development of conventional hyperthermia. Microscopic hot spots, which are often cited as an explanation for alleged non-thermal effects, were discussed in detail and could largely be excluded both in the clinical studies discussed and in the preclinical experiments (Section 5.1). Instead, we presented possible mechanisms at ion channels for non-thermal effects of sinusoidal RF (Section 5.2). The data suggest that additional AM with selected (tumor-specific) modulation frequencies might enhance the antiproliferative effects of sinusoidal RF (Sections 3.3 and 4.3). We presented membrane resonances as a possible explanation for such frequency-dependent behavior and could correlate the observed modulation frequencies (hertz to kilohertz) with elastic membrane properties that are characteristic for tumor cells (Section 5.3).
These novel model approaches give some insight into possible mechanisms of non-thermal effects, but do not claim to be validated models that can quantitatively describe the experiments. The exact mechanisms and the dependencies of non-thermal effects are still unresolved and require much more data that cover broader ranges of SAR, temperatures and modulation frequency spectra as treatment parameters. Until now, the data regarding non-thermal effects for therapeutic purposes are not nearly comparable to the scope of the data on WB-HT and cRF-HT.
In particular, there is a lack of comparative measurements of RF ± AM in cells and experimental tumors in order to quantitatively record the influence of AM and enable refinement of the theoretical approaches. Interestingly, in the preclinical investigations with the mEHT, the data on the AM were incompletely documented and apparently, no systematic variation was performed (Section 4.1). Such comparative experiments are urgently needed.
However, the available data together indicate that the feasibility, tolerance and effectiveness of optimized AM-RF-HT might be distinctly superior to cRF-HT. AM-RF at 27.12 MHz (TheraBionic) refers to some frequency-dependent non-thermal effects at very low SAR (<0.4-2 W/kg) but is of limited clinical effectiveness and requires long exposition times.
The AM-RF-HT at 13.56 MHz promises higher effectiveness with a few sessions of one hour per week [23]. The existing data on AM-RF-HT suggest that low AM frequencies less than kilohertz might be more effective. At these low frequencies, the resonance of whole cells might be explained by forced oscillations of a variety of ion channels.
In preclinical studies, application of AM-RF-HT at 13.56 MHz (mEHT) alone at moderate clinically feasible SAR levels (<20-30 W/kg) and temperatures (<41-42 C) might achieve an estimated cell kill of 30-50% in the target region, which is not achieved by RF-HT at temperatures 42 C (see [69]). Therefore, AM-RF-HT monotherapy at 13.56 MHz is a debatable option. For example, if we integrate 2-3 AM-RF-HT at 13.56 MHz sessions per week up to a total of 20 sessions into an adjuvant scheme, this procedure might be comparable to 20 Â 2 Gy with respect to cytotoxicity, taking into account preclinical data. Preclinical results confirmed in vivo that a series of AM-RF-HT monotherapy sessions with 48 h intervals (2-3 times a week) might be effective [63]. This is a clear deviation from the rules of cRF-HT, which requires a combination of RT and/or chemotherapy.
Available RF heating technology is suitable for exposing anatomical regions such as the brain, liver, lung, and major parts of the abdomen or pelvis. However, we require high SAR [W/kg] for temperatures >42 C [temperature increase (DT) ¼ 4.5 C from 37.5 C to 42 C] in a target with perfusion w [ml/ 100 g/min] as given by the following easy relationship [152]: SAR ½W=kg ¼ ðDT=1:4Þ Â w ½ml=100 g=min Only for w < 10 ml/100 g/min are SAR of %30 W/kg sufficient, which are achievable using the elaborated APA technique. The less expensive capacitive technology might achieve 10-15 W/kg in deep anatomical regions. Therefore, only in low-perfused tumors or tissues with w < 5 ml/100 g/ min capacitive systems can gain temperatures >42 C.
Considering the non-thermal effects outlined in this review, we expect relevant cytotoxicity at moderate SAR and lower temperatures <42 C significantly reducing the high demands for cRF-HT listed in quality assurance reports [10]. There is evidence that a pure RF has non-thermal antiproliferative effects [55], whereby an RF of 10 MHz might have advantages over 100 MHz (see Section 5.1). Further, preclinical and clinical studies strongly suggest that a suitable AM can increase RF effectiveness. However, we do not know exactly the optimal modulation frequencies, SAR, and temperatures for clinical applications. Therefore, further systematic preclinical studies are required to complete the fragmentary data on non-thermal effects.
For cRF-HT, only tumors with low perfusion <10 ml/100 g/ min are accessible for SAR ¼ 30 W/kg to achieve 42-43 C; even tumors with perfusion up to 40 ml/100 g/min would be accessible for the same SAR if temperatures of 38 C would be considered sufficient to ensure non-thermal antiproliferative effects in conjunction with a slight temperature increase (see the equation above). Here, we regard temperature as a co-variable. Consequently, the possible applications in oncology would increase considerably. These applications include AM-RF-HT monotherapy but especially in combination with other therapy modalities. If the existence of non-thermal antiproliferative effects can be further substantiated, we believe a new field opens up for oncology.

Author contributions
P.W. conceptualization of the project, review of the literature, development of the theory and performance of calculations, writing and approval of the manuscript. U.S. editing and approval of the manuscript. P.G. funding acquisition, the conceptualization of the project, methodology, writing and approval of the manuscript. All authors have read and agreed to the published version of the manuscript.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
We acknowledge support from the Berliner Krebsgesellschaft e.V., grant number GHFF202006, www.berliner-krebsgesellschaft.de, and from the German Research Foundation (DFG) and the Open Access Publication Fund of Charit e -Universit€ atsmedizin Berlin.