Radiofrequency Electromagnetic Field Exposure and Apoptosis: A Scoping Review of In Vitro Studies on Mammalian Cells

In the last decades, experimental studies have been carried out to investigate the effects of radiofrequency (RF, 100 kHz–300 GHz) electromagnetic fields (EMF) exposure on the apoptotic process. As evidence-based critical evaluation of RF and apoptosis in vitro is lacking, we performed a scoping literature review with the aim of systematically mapping the research performed in this area and identifying gaps in knowledge. Eligible for inclusion were in vitro studies assessing apoptosis in mammalian cells exposed to RF-EMF, which met basic quality criteria (sham control, at least three independent experiments, appropriate dosimetry analysis and temperature monitoring). We conducted a systematic literature review and charted data in order to overview the main characteristics of included studies. From the 4362 papers retrieved with our search strategy, 121 were pertinent but, among them, only 42 met basic quality criteria. We pooled data with respect to exposure (frequency, exposure level and duration) and biological parameters (cell type, endpoint), and highlighted some qualitative trends with respect to the detection of significant effect of RF-EMF on the apoptotic process. We provided a qualitative picture of the evidence accumulated so far, and highlighted that the quality of experimental methodology still needs to be highly improved.


Rationale
In 2011, the International Agency for Research on Cancer (IARC) classified radiofrequency electromagnetic fields (RF-EMF, 100 kHz-300 GHz) from mobile phones as a possible human carcinogens (2B group) based on the limited evidence from experimental and epidemiological studies [1]. From the literature published since then, and reviewed by international expert panels, the evidence for health effects due to RF-EMF exposure is still inadequate, and needs more accurate investigation [2,3]. Moreover, mechanistic insights of RF-EMF exposure have not been clarified so far for any cellular endpoint. Many hypotheses have been suggested, but none of them has been proven so far [2][3][4][5].
Among the potential cellular mechanisms that are relevant for cancer occurrence, alteration of apoptotic process is of particular interest due to the pivotal role of the regulation of apoptosis in cell homeostasis [6,7] Moreover, abnormalities in cell death regulation, whether they feature insufficient or excessive apoptosis, can be a significant component of other diseases such as autoimmune lymphoproliferative syndrome, AIDS, ischemia and neurodegenerative diseases such as Parkinson's, Alzheimer's, Huntington's diseases and Amyotrophic Lateral Sclerosis [8].
Apoptosis is an important cell death program, highly conserved within multicellular organisms and genetically controlled, which is responsible for the removal of damaged, 2 of 25 dysfunctional or no longer necessary cells to promote homeostasis and survival of organisms [6][7][8]. Two pathways are involved in apoptosis that work synergistically to assure the removal of the defective cells. The intrinsic cell death pathway, or mitochondrial pathway, is activated by the cell itself upon detection of cell damage via a number of intracellular sensors. It is governed by the Bcl-2 family of proteins, which regulate commitment to cell death through the mitochondria and the activation of caspase 9. The extrinsic cell death pathway is activated by the interaction between a cell of the immune system and a damaged cell. Activation of the extrinsic cell death pathway occurs following the binding on the cell surface of "death receptors" such as Tumor Necrosis Factor Receptor (Fas TNFR1), or TNF-related apoptosis inducing ligand (TRAIL) receptors, to their corresponding ligands. These death receptors recruit adaptor molecules such as Fas-associated protein with death domain (FADD) and caspase 8. Once the caspases 8 and 9 are activated by inactive pro-caspases, the two pathways converge with the activation of the executioner caspases (caspases 3, 6 and 7). At this point, a cascade of events initiates that leads to DNA fragmentation from activation of endonucleases, destruction of cytoskeleton and nuclear proteins, crosslink of proteins, the expression of ligands for recognition by phagocytic cells, such as the phosphatidylserine, and the formation of apoptotic bodies. The exposure of phosphatidylserine on the external surface of the plasma membrane allows the phagocyte recognition of the dying cells [8][9][10].
Different methods for the detection of apoptosis and its peculiar hallmarks (which allow the recognition with respect to the necrosis), have been developed over time concurrently with the knowledge of apoptosis phenomena. They mainly rely on morphological and biochemical analysis aimed at identifying features of apoptotic cells such as shrinkage, membrane blebbing and chromatin condensation, DNA fragmentation, detection of caspases, cleaved substrates, regulators and inhibitors, externalization of phosphatidylserine, alteration of mitochondrial membrane potential, release of cytochrome-c, analysis of apoptotic or anti-apoptotic regulator proteins such as Bcl-2-associated X protein (Bax), as well as BH3-interacting domain death agonist (Bid), and BCL2 apoptosis regulator (Bcl 2) [8].
A number of in vitro and in vivo experimental studies have addressed the effects of exposures to RF EMF, at frequencies and signals typical of telecommunications, on the apoptotic process. These studies have been carried out under different conditions and experimental regimens with conflicting results, which have not been systematically reviewed. Several reviews regarding the effect of RF-EMF on mammalian cells included apoptosis but were not specifically focused on it [11][12][13][14][15]. Moreover, in all the reviews cited above, papers were not retrieved by performing a systematic literature search, and inclusion criteria did not include cogent quality parameters. The latter have been demonstrated to greatly affect the results of experimental studies. Indeed, quality is emerging as a critical issue in bioelectromagnetic research in general, since the majority of studies do not comply with quality criteria such as adequate attention to dosimetry, inclusion of sham control, positive control, blind evaluation and temperature control [16,17].
As evidence-based critical evaluation of RF exposures and apoptosis is still lacking with reference to health risk assessment, here we performed a scoping literature review, with the aim of systematically mapping the research performed in this area and identifying gaps in knowledge. We focus on in vitro studies because they can provide essential information about the potential effects of chemicals or physical agents on specific cell properties, and allow a more rapid, cost effective and well-controlled approach to molecular and mechanistic studies than conventional laboratory animal models [16]. Moreover, the preamble to the IARC Monographs on the Identification of Carcinogenic Hazards to Humans has given new emphasis and highlighted the importance of mechanistic studies in corroborating evidence and providing biological plausibility to other types of studies, and the possibility that they could provide strong evidence in case of consistent findings demonstrated across a number of different systems and in different species [1].

Objective
The aim of this scoping review is to survey the available evidence on the effects of RF-EMF exposures on the apoptotic process in mammalian cells cultured in vitro by mapping how research was conducted, by identifying key characteristics of the experiments and any existing gaps in knowledge. A systematic literature search was performed and the review was restricted to studies that adhere to basic quality criteria defined a priori, and thus characterized by low risk of bias. The scientific question, formulated as a PECO (Population, Exposure, Comparator, Outcome) statement, is outlined in Table 1.

Methods
The scoping review conforms to PRISMA-ScR (Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews) guidelines, provided as Supplementary Material (Table S1: PRISMA-ScR-Checklist) [18,19].

Eligibility Criteria
We have restricted inclusion to peer-reviewed journal articles reporting findings from primary studies and published in English. Meeting abstracts, conference proceedings, and commentaries were excluded, whereas reviews have been used to check for missing articles.
We have included in vitro studies assessing the capability of RF-EMF in the frequency range between 100 kHz and 300 GHz, to affect apoptosis process in mammalian cells (Table 1), with no restrictions on biological model (freshly collected or immortalized cells), cell status (healthy or cancerous), or cell lineage. Regarding the apoptosis outcome, we have included the endpoints reported in the PECO statement (Table 1).
For studies that evaluate apoptosis in relation to both RF exposure alone, and to coexposure to RF fields and other agents, only findings concerning RF exposure alone have been considered, because we want to focus on potential apoptosis effects of RF-EMF themselves.
In order to restrict the analysis to papers characterized by a low risk of bias, we have adopted the following quality-based exclusion criteria [20].
First, we excluded studies that did not provide information to adequately characterize exposure conditions, such as frequency range, signal type, exposure level and duration, as detailed in the PECO statement (Table 1). We excluded studies where dosimetry analysis was not performed at all or was not carried out with adequate methods. For example, estimation of SAR from measurements of the electric field in absence of the sample is not appropriate because the sample significantly perturbs the electric field in the RF range. On the contrary, estimation of SAR from computation of electric field in the sample or by calorimetric measurements are acceptable [20,21].
We included studies that used sham-exposed controls, i.e., a sample placed in an exposure system identical to that used to administer the treatment, except for the emission of RF-EMF [20]. A further quality-based exclusion criteria was the absence of temperature control. As a main confounding factor for RF exposure, temperature inside the samples must be monitored to ascertain absence of heating, or to counteract possible thermal increase. Finally, we also excluded studies that performed less than three independent experiments or did not report the number of independent experiments.

Information Sources and Search Strategy
Our primary information sources were NCBI PubMed and Web of Science (WOS) databases. In both cases, we did not apply restrictions in terms of time coverage, and the most recent search was performed on 12 August 2021. The search strategies developed for both databases are provided as Supplementary Material (Table S2: Search queries). We also checked reference lists of review papers and authors' personal literature databases to retrieve studies that were missed by the web search.

Selection of Sources of Evidence
All bibliographic records have been imported into the reference management software Endnote ® X9, and the appropriate functions have been used to remove duplicates and classify the papers by relevance and inclusion/exclusion status. Two independent reviewers (SR and MRS) have performed a two-phase selection process: first, the papers have been included/excluded for relevance based on the screening of title and abstract; second, the full text of all potentially relevant papers has been retrieved and assessed for compliance with the predefined eligibility criteria. The results of the papers selection process have been graphically displayed in a PRISMA flow-chart, and the papers excluded at the stage of full-text examination have been recorded in a separate table, with indication of at least one motivation for exclusion.

Data Charting Process and Data Items
The same investigators in charge of the papers selection have also extracted the relevant information regarding the experiments, using the forms reported as Supplementary Material 3. More specifically, for each paper we have identified individual experiments, characterized by different exposure conditions (in terms of frequency, signal, exposure level or duration), or different cell models or endpoints. For each experiment, the following data have been extracted and recorded in the form: We also assessed the overall incidence of statistically significant or non-significant effects in the experiments, and the relative incidence of effects with respect to the endpoints and the exposure parameter subgroups.

Selection of Sources of Evidence
The results of the literature search and of the screening process are summarized in the PRISMA flow-chart in Figure 1. The literature search yielded a total of 4649 records, which reduced to 4362 after duplicates removal in Endnote X9. The first round of screening, based on information and terms in the title and abstract, led to the exclusion of 4241 publications, whereas for the remaining 121 records the full text was assessed for eligibility. Among these, 79 papers were excluded because they were not compliant with either basic or quality criteria, whereas the remaining 42 were fully analyzed for data extraction and synthesis. The full references of excluded papers with motivations for exclusion are reported in Table 2.
To the latter aim, we identified several subgroups within the exposure parameters We also assessed the overall incidence of statistically significant or non-signific fects in the experiments, and the relative incidence of effects with respect to the end and the exposure parameter subgroups.

Selection of Sources of Evidence
The results of the literature search and of the screening process are summar the PRISMA flow-chart in Figure 1. The literature search yielded a total of 4649 re which reduced to 4362 after duplicates removal in Endnote X9. The first round of s ing, based on information and terms in the title and abstract, led to the exclusion o publications, whereas for the remaining 121 records the full text was assessed for e ity. Among these, 79 papers were excluded because they were not compliant with basic or quality criteria, whereas the remaining 42 were fully analyzed for data ext and synthesis. The full references of excluded papers with motivations for exclus reported in Table 2.     Figure 2 shows the number of included, excluded and retracted studies, and the motivations for exclusion with relative proportions (when more than one motivation applied, only one of them was counted): the most recurrent motivations were the absence of sham control and the absence of dosimetry or of appropriate dosimetry methods. 2014 [99] No sham control. No dosimetry performed. 2015 [100] No temperature control at 18 W/kg SAR. Figure 2 shows the number of included, excluded and retracted studies, and t tivations for exclusion with relative proportions (when more than one motivation ap only one of them was counted): the most recurrent motivations were the absence o control and the absence of dosimetry or of appropriate dosimetry methods.     Tables 3 and 4. We separated studies into two categories. First, those that did not observe statistically s icant (according to the statistical analysis performed by the authors) alterations o apoptotic process due to RF-EMF exposures in any of the experimental conditions co ered (Table 3). Secondly, those that did report significant alterations in at least one o experimental conditions considered (Table 4).

Characteristics of Sources of Evidence
A total of 255 experiments were extracted from the 42 papers analyzed. Each ex ment was identified on the basis of either one of the exposure parameters (frequency  Tables 3 and 4. We have separated studies into two categories. First, those that did not observe statistically significant (according to the statistical analysis performed by the authors) alterations of the apoptotic process due to RF-EMF exposures in any of the experimental conditions considered (Table 3). Secondly, those that did report significant alterations in at least one of the experimental conditions considered (Table 4).          (Table S3: Data extracted from experiments).

Results and Critical Appraisal of the Source of Evidence
We surveyed the data extracted from included studies with respect to relevant parameters, namely the cell origin (human vs. animal, primary vs. cell lines), the endpoints analyzed, the frequency, exposure level and exposure duration subgroups, the reporting of statistically significant effects.

Results and Critical Appraisal of the Source of Evidence
We surveyed the data extracted from included studies with respect to relevant parameters, namely the cell origin (human vs. animal, primary vs. cell lines), the endpoints analyzed, the frequency, exposure level and exposure duration subgroups, the reporting of statistically significant effects.
As shown in Figure 4, human cells were used as biological model more than animal cells (72.5% vs. 27.5%), whereas in both cases immortalized cell lines (81.6% human and 85.7% animal) were mainly used with respect to primary cells (18.4% human, 14.3% animal). The percentage of endpoints analyzed were as reported in Figure 5, where the vast majority (44.3%) of experiments assessed apoptosis by analyzing the phosphatidylserine externalization, followed by apoptosis signaling (18.8%), caspase activity (12.6%), and DNA fragmentation (10.6%). The remaining 13.7% accounted for the other considered apoptosis endpoints, namely alteration of mitochondrial membrane potential (4.7%), morphological hallmarks (3.9%), PARP cleavage (3.9%), membrane integrity (0.8%), and expression of cytochrome-c (0.4%).   Figure 6 shows the percentage of experiments belonging to the three subgroups (as defined in Table 1) within the main exposure parameters (frequency, exposure level and

Results and Critical Appraisal of the Source of Evidence
We surveyed the data extracted from included studies with respect to relevant parameters, namely the cell origin (human vs. animal, primary vs. cell lines), the endpoints analyzed, the frequency, exposure level and exposure duration subgroups, the reporting of statistically significant effects.
As shown in Figure 4, human cells were used as biological model more than animal cells (72.5% vs. 27.5%), whereas in both cases immortalized cell lines (81.6% human and 85.7% animal) were mainly used with respect to primary cells (18.4% human, 14.3% animal). The percentage of endpoints analyzed were as reported in Figure 5, where the vast majority (44.3%) of experiments assessed apoptosis by analyzing the phosphatidylserine externalization, followed by apoptosis signaling (18.8%), caspase activity (12.6%), and DNA fragmentation (10.6%). The remaining 13.7% accounted for the other considered apoptosis endpoints, namely alteration of mitochondrial membrane potential (4.7%), morphological hallmarks (3.9%), PARP cleavage (3.9%), membrane integrity (0.8%), and expression of cytochrome-c (0.4%).   Figure 6 shows the percentage of experiments belonging to the three subgroups (as defined in Table 1) within the main exposure parameters (frequency, exposure level and exposure duration). The majority of the experiments (94.5%) were performed by applying   Table 1) within the main exposure parameters (frequency, exposure level and exposure duration). The majority of the experiments (94.5%) were performed by applying EMF in the F2 subgroup (10 MHz to ≤6 GHz), only 5.5% of them applied EMF in the F3 (>6 to ≤300 GHz) subgroup, and none of them employed frequencies below 10 MHz (F1). The distribution within the exposure level subgroups was definitely more uniform, with 38.4% of experiments performed at SAR < 1 W/kg or S ab < 20 W/m 2 or S inc < 10 W/m 2 (EL1), 29% at 1 W/kg < SAR ≤ 2 W/kg S ab = 20 W/m 2 or S inc = 10 W/m 2 (EL2), and 32.5% SAR > 2 W/kg or S ab >20 W/m 2 or S inc >10 W/m 2 (EL3). In most of the experiments (65.1%) exposure duration was long (ED2, >1 h to ≤24 h), whereas in 21.2% and 13.7% they were acute (ED1, ≤1 h) and chronic (ED3, >24 h), respectively.
Int. J. Mol. Sci. 2022, 23, x FOR PEER REVIEW (>6 to ≤300 GHz) subgroup, and none of them employed frequencies below 10 MH The distribution within the exposure level subgroups was definitely more uniform 38.4% of experiments performed at SAR < 1 W/kg or Sab < 20 W/m 2 or Sinc < 10 W/m 29% at 1 W/kg < SAR ≤ 2 W/kg Sab = 20 W/m 2 or Sinc = 10 W/m 2 (EL2), and 32.5% S W/kg or Sab>20 W/m 2 or Sinc>10 W/m 2 (EL3). In most of the experiments (65.1%) ex duration was long (ED2, >1 h to ≤24 h), whereas in 21.2% and 13.7% they were acut ≤1 h) and chronic (ED3, >24 h), respectively. In the majority of experiments (84.7%), no statistically significant effects on t lyzed endpoints were found; only the 15.3% reported statistically significant effec ure 7a). The percentage of experiments reporting effects is presented in Figure 7 respect to exposure parameters. The highest incidence occurred in the F3 subgrou 13 out of 14 experiments (belonging to two different studies) reporting significant Regarding the exposure level and duration subgroups, the highest incidence was ob for above limits (EL3, 30.1%) and acute exposures (ED1, 42.6%). The incidence of cant and non-significant effects with respect to the analyzed endpoints is shown in 7c. The highest incidence of significant effects was found in terms of MMP modifi (6 out of 12 experiments), followed by caspases activation (7 out of 32 experimen the observation of morphological hallmarks (2 out of 10 experiments). For the PE tosis signaling and DNA fragmentation endpoints, the incidence of significant ranged from 10.6 to 14.3% (12 out of 113 experiments for PE; 7 out of 48 for ap signaling; 3 out of 27 experiments for DNA fragmentation). Only two experime sessed the membrane integrity endpoint and did not find significant effects. Only periment assessed expression of cytochrome-c and found a significant alteration.
Moreover, the incidence of statistically significant effects in experiments per with human or animal cells was 13.5% and 20%, respectively. Table 3 reports an overview of the studies that did not observe significant alte of the apoptotic process due to RF-EMF exposures. In 17 out of 31 studies, only on tosis endpoint was analyzed; in 20 out of 31 studies the analysis was not perfor blind; and in 7 out of 31 studies positive control was not included. The experim which significant alterations of apoptosis endpoints were found belonged to twelv In the majority of experiments (84.7%), no statistically significant effects on the analyzed endpoints were found; only the 15.3% reported statistically significant effects (Figure 7a). The percentage of experiments reporting effects is presented in Figure 7b with respect to exposure parameters. The highest incidence occurred in the F3 subgroup, with 13 out of 14 experiments (belonging to two different studies) reporting significant effects. Regarding the exposure level and duration subgroups, the highest incidence was obtained for above limits (EL3, 30.1%) and acute exposures (ED1, 42.6%). The incidence of significant and non-significant effects with respect to the analyzed endpoints is shown in Figure 7c. The highest incidence of significant effects was found in terms of MMP modifications (6 out of 12 experiments), followed by caspases activation (7 out of 32 experiments) and the observation of morphological hallmarks (2 out of 10 experiments). For the PE, apoptosis signaling and DNA fragmentation endpoints, the incidence of significant effects ranged from 10.6 to 14.3% (12 out of 113 experiments for PE; 7 out of 48 for apoptosis signaling; 3 out of 27 experiments for DNA fragmentation). Only two experiments assessed the membrane integrity endpoint and did not find significant effects. Only one experiment assessed expression of cytochrome-c and found a significant alteration.  Table 1; Number above each bar refer to the absolute number of experiments where significant effects were observed); (c) percentage of experiments reporting statistically significant effect or no effect with respect to the endpoints analyzed (numbers on the bars indicate the absolute numbers of experiments performed).

Summary of Evidence
One of the main concerns regarding health effects of RF-EMF is that prolonged exposures to weak field levels may cause long-term effects. In spite of the high number of studies published on this matter, the evidence accumulated so far is inconclusive and controversial.
The majority of studies regarding biological effects of RF-EMF are based on an in vitro study design, because this approach is rapid, cost effective, allows exposures to be performed under strictly controlled electromagnetic and environmental conditions and provides insight into mechanistic interactions [1]. Among the biological outcomes that can be of interest for the mechanistic assessment of long-term effects, apoptosis has been increasingly considered over the last twenty years. The gained information has not been comprehensively reviewed and does not allow for a complete picture of investigation carried out so far, of the possible effects on this critical cellular process and of the existing gaps in knowledge.
Manna and Gosh reviewed the effects of RF-EMF exposure in cultured mammalian cells on several biological outcomes, including apoptosis. The authors concluded that RF-  Table 1; Number above each bar refer to the absolute number of experiments where significant effects were observed); (c) percentage of experiments reporting statistically significant effect or no effect with respect to the endpoints analyzed (numbers on the bars indicate the absolute numbers of experiments performed).
Moreover, the incidence of statistically significant effects in experiments performed with human or animal cells was 13.5% and 20%, respectively. Table 3 reports an overview of the studies that did not observe significant alterations of the apoptotic process due to RF-EMF exposures. In 17 out of 31 studies, only one apoptosis endpoint was analyzed; in 20 out of 31 studies the analysis was not performed in blind; and in 7 out of 31 studies positive control was not included. The experiments in which significant alterations of apoptosis endpoints were found belonged to twelve studies, which are overviewed in Table 4. In 5 out of 11 studies, only one apoptosis endpoint was investigated; in 7 out of 11 studies the analysis was not performed in blind; and in 10 out of 11 studies, positive control was lacking.

Summary of Evidence
One of the main concerns regarding health effects of RF-EMF is that prolonged exposures to weak field levels may cause long-term effects. In spite of the high number of studies published on this matter, the evidence accumulated so far is inconclusive and controversial.
The majority of studies regarding biological effects of RF-EMF are based on an in vitro study design, because this approach is rapid, cost effective, allows exposures to be performed under strictly controlled electromagnetic and environmental conditions and provides insight into mechanistic interactions [1]. Among the biological outcomes that can be of interest for the mechanistic assessment of long-term effects, apoptosis has been increasingly considered over the last twenty years. The gained information has not been comprehensively reviewed and does not allow for a complete picture of investigation carried out so far, of the possible effects on this critical cellular process and of the existing gaps in knowledge.
Manna and Gosh reviewed the effects of RF-EMF exposure in cultured mammalian cells on several biological outcomes, including apoptosis. The authors concluded that RF-EMF exposure might affect the apoptotic process in vitro, with results depending on the type of modulation, intermittent mode of exposure and cell model [15]. Halgamuge and co-workers performed an extensive meta-analysis of data from in vitro studies published between 1990 and 2015, and investigating effects of weak RF-EMF from mobile phones. The analysis revealed a lack of uniform responses in any of the investigated outcomes, which included but was not specifically focused on apoptosis, and also highlighted some recurrent patterns of evidence which depended on cell and signal types [14]. Moreover, in the abovementioned reviews, the study inclusion criteria did not take into account the aspects of quality of experimental methods, which have been widely demonstrated to affect the results [16,17]. In [16], co-authored by two authors of this review, apoptosis was considered together with proliferation to detect possible statistical associations between RF-EMF exposures and cellular response. Cellular response after exposure to RF-EMF was significantly associated to cell lines rather than to primary cells, but not to other experimental parameters.
In this scoping review, for the first time in our knowledge, we specifically addressed apoptosis outcome in studies evaluating the effects of RF-EMF exposures on mammalian cells in vitro. We conducted a systematic literature review, included studies on the basis of quality criteria defined a priori, and provided an overall picture of what has been published so far.
The systematic literature search yielded a total of 121 relevant papers, but only 34.5% of them met the inclusion criteria. The majority of retrieved papers failed to comply with quality criteria for good bioelectromagnetic experiments [16,17,143]. The main motivations for exclusion were the absence of sham controls and the lack of dosimetry analysis, or of appropriate methods for dosimetry analysis. Lack of compliance with these two criteria indicates that experimental conditions were not identical across study groups, and that there was a low confidence in exposure characterization. Quality of experimental methods has become an issue in bioelectromagnetic research, with the majority of published papers presenting flaws on either electromagnetic or biological requirements, or both. Basic quality criteria for in vitro experiments on RF-EMF exposures include the presence of sham control, dosimetry analysis conducted with standardized methods, temperature control, blind analysis and positive control. The absence of some or all of these requirements has been shown to be highly associated with the detection of effect [16,17]. In this review, we considered three out of five basic requirements as inclusion criteria (sham, dosimetry analysis, temperature control). The motivation of this choice was to include papers with a low risk of bias, but we did not consider the absence of blind analysis and of positive control among the exclusion criteria in order to be more inclusive towards the final analysis of relevant papers. The latter criteria were applied for a deeper characterization of the quality of studies. Indeed, in more than half of the included studies analysis of data was not blinded, and in almost half of the studies positive control was not included in the study design.
It is important mentioning that in more than half of the included studies apoptosis was evaluated by assaying a single endpoint. Since apoptosis occurs via a complex signaling cascade that is tightly regulated at multiple points, and since it presents many features in common with necrosis, it is crucial to perform two or more distinct assays, based on different principles, to confirm that cell death occurred via apoptosis [8]. As an example, the detection of phosphatidylserine externalization requires the use of specific dyes (Annexin-V binds to phosphatidylserine on the plasma membrane, while Propidium Iodide only enters necrotic cells) allowing discrimination of apoptotic from necrotic cells, because an increase in membrane permeability is also a feature of necrotic cells. For the same reason, each test must be associated with others based on different apoptotic features.
We have overviewed the main characteristics and outcome of the included studies. The majority of them did not find significant alterations of the apoptotic process due to RF-EMF exposure. Looking at the experiments extracted from the studies, when a statistically significant effect was observed it mainly occurred at frequencies above 6 GHz, and for acute (≤ 1h) exposure durations. However, since the number of studies reporting effects is very small, and the considered experimental conditions are highly heterogeneous, further investigations are needed, together with replication studies, to confirm or confute these results. Moreover, even though the included studies met the basic quality criteria, most of them still presented flaws in the experimental methods (lack of blind analysis and/or positive control, assessment of single endpoints). It can be stated that, to be of value, future studies that investigate the effect of RF-EMF in mammalian cells should aim to be of high methodological quality and be sufficiently powered by performing an adequate number of experiments.

Limitations
The bibliographic search was conducted only on two databases (PubMed and WOS). Even though these two databases may contain the vast majority of studies within the field, it is possible that potentially relevant studies might not be indexed there. Misclassification of studies based on keywords, title or abstract might have also affected the sensitivity of the search strategies. We conducted hand searches to minimize the number of articles missed, but this still may not have captured all eligible articles.

Conclusions
This scoping review sought to systematically map the research regarding the effects of RF-EMF on apoptosis in mammalian cells, and to identify any existing gaps in knowledge within health risk assessment of RF-EMF exposures. This will definitely facilitate to gain reliable information on the effects of RF exposure on the apoptotic process when in a next step, a quantitative analysis of the papers included in this scoping review will be carried out by mainly addressing questions on the direction of the effect (induction or suppression of apoptosis), effect size, possible dose-response relationship, possible association of the effect size with the quality score of the experiments, and possible major capability of certain exposure parameter ranges to exert an effect. The major gap in knowledge from the qualitative analysis conducted here is the lack of a systematic approach based on quality of the experimental methodologies adopted in the studies retrieved and analyzed in this scoping review. Therefore, the evidence here presented is a further confirmation that, in spite of the large amount of relevant papers available in the literature, a huge effort still needs to be made in bioelectromagnetic research towards the improvement of experimental quality, which is crucial to guarantee the reliability, robustness and reproducibility of results.