The world is facing a cancer pandemic, with exponential growth occurring in many cancers (International Agency for Research on Cancer (IARC) World Cancer Reports) (1–4). While population aging is contributing to this trend (5), it cannot fully explain the observed rise in certain cancers, suggesting that environmental and/or lifestyle factors are playing a role. One potential factor for consideration is the increasing global background levels of anthropogenic radiofrequency electromagnetic fields (RF-EMF) (6), which coincide with a general increase in cancer incidence rates over the last several decades. While this temporal overlap does not imply causation, it highlights the need for careful investigation into the potential role of RF-EMF exposure among the many environmental and lifestyle factors associated with cancer.

The current perspective held by IARC is that RF-EMF exposures, including exposure to mobile phones, are a group 2B possible carcinogen (IARC, May 2011) (7, 8). This classification was based on the available evidence at the time, which covered both epidemiological and animal experimental studies. While the evidence was deemed credible, bias and confounding could not be completely ruled out (8). A limited understanding of the underlying mechanisms, regarded by some as weak (9), also prevented a higher classification.

Genetic alteration is a well-established trigger for cancer development (10). Genotoxicity is the ability of a physical or chemical agent to induce genetic damage, which may result in genetic mutations (11), and represents a critical pathway to cancer. Therefore, if RF-EMF exposure is linked to genotoxicity, this would provide strong evidence for a plausible mechanism describing how RF-EMF may initiate carcinogenesis in humans and potentially in all living organisms. With the recommended prioritization for an IARC (12) assessment of the carcinogenic potential of radiofrequency exposures, an unbiased synthesis of the evidence on RF-induced genotoxicity is a crucial resource needed for such an investigation.

This review aims to determine whether RF-EMF exposure can damage DNA, thereby potentially contributing to the rising global incidence of cancer.

Radiofrequency electromagnetic fields are defined as non-ionizing electromagnetic frequencies in the range of 3 kilohertz (kHz) to 300 gigahertz (GHz) (13), which sit between extremely low-frequency fields (ELF), i.e., electrical power frequencies, and infrared light in the electromagnetic spectrum. RF-EMF is produced both naturally (as background cosmic radiation, lightning and other atmospheric activity) and from a wide range of man-made sources including radar, radio/TV broadcast antenna, satellite communications, mobile phone base stations, smart meters, smartphones, Bluetooth devices, game consoles, baby monitors, computers, Wi-Fi routers, microwave ovens, radiofrequency implanted devices, diathermy machines and wireless power transmission devices.

The primary source of natural, isotropic microwave background radiation on Earth is the Cosmic Microwave Background (14). This radiation falls to earth in a non-polarized and continuous manner, at very low intensity levels as low as 10⁻¹⁸ W/m² (6) for frequencies used in the mobile phone RF-EMF spectrum. In contrast, man-made RF-EMF exposure is typically polarized and pulsed with intensities reaching up to 10 W/m² for certain frequencies used in wireless communication, as permitted by the International Commission of Non-Ionizing Radiation Protection (ICNIRP) (15) for member of the public exposures. These important characteristics distinguish man-made from natural sources of radiation. Biological life has never before experienced this type of artificial radiation, which may have unique biological effects (16). Radiofrequency waves can penetrate human skin, depending on wave frequency and tissue properties (17). Lower-frequency RF waves penetrate deeper, reaching internal organs, while higher frequencies are more readily absorbed by the skin layers or reflected (18). Wireless communication technologies, such as mobile phones and Wi-Fi routers, rely on RF frequencies that can pass through barriers like walls, windows, and roofs so as to maintain signal coverage.

DNA damage encompasses various forms (see Figure 1), including single-strand breaks (SSBs), double-strand breaks (DSBs), chromosome aberrations, sister chromatid exchanges and the presence of micronuclei (11), all of which can significantly impact cellular integrity.

Approximately 10,000 DNA modifications occur every hour per cell (19), and of these, SSBs are the most frequent types of DNA lesions (75%) (19). SSBs can arise from exposure to free radicals such as reactive oxygen and nitrogen species (ROS/RNS) (20). SSBs may also occur as intermediate products of the DNA repair process or as a result of abortive cellular enzyme activities (20). If they are not promptly and correctly repaired, SSBs can disrupt critical processes such as DNA replication and transcription, which ultimately compromises genome stability (21). DSBs, though less frequent, are particularly severe because they involve breaks in both DNA strands. DSBs pose a significant challenge to repair mechanisms and increase the risk of chromosomal aberrations and micronuclei, which serve as biomarkers for genome instability (22). The accumulation of unrepaired DNA damage, including SSBs, DSBs, and associated abnormalities, has been strongly implicated in the development of cancer (11), aging-related disorders (23), and neurodegenerative diseases (24).

Each type of DNA damage has specific assay methods (e.g., comet assay) to detect and evaluate the damage. The types of DNA damage and their associated assays that are investigated as part of the evidence map are detailed in Supplementary Table 1.¹

DNA damage as measured via an assay represents the net result of three factors: (i) the damage that is induced by exposure to an external agent under test, (ii) the baseline level of endogenous DNA damage that occurs naturally under normal physiological conditions, and (iii) the extent of damage repaired by the cellular DNA repair mechanisms.

It has been well-established that the biological and health effects of microwaves depend on various biological and physical parameters that differ across studies, leading to variations in observed effect outcomes (25, 26).

Hundreds of experimental studies of varying quality have been conducted over many decades investigating whether RF-EMF exposure can damage DNA or result in genetic mutations. Several reviews have also been performed, each using a smaller subset of studies, with mixed results. The balance of evidence determined by these past reviews ranges from significant evidence of genetic damage or interference, Lai (27), to slightly favoring DNA damage, Ruediger (28), to claims that most studies show no significant effects, Vijayalaxmi and Prihoda (29, 30) and more recently, Romeo et al. (31).

Reasons for the large discrepancies in past reviews include:

Given the current incompleteness of available systematic reviews, inconsistencies in reporting and potential biases, an evidence map investigating the potential genotoxic effects of RF-EMF is both timely and essential.

The primary question for this scoping review and evidence map is: What evidence exists regarding the genotoxic potential of anthropogenic RF-EMF exposures? A systematic mapping and synthesis of the following factors will help address this question:

Further questions aim to explain the reasons for conflicting outcomes and inconsistencies found in previous reviews:

Populations: all organisms and cell types, microorganisms or free DNA used in in vitro, in vivo and epidemiological studies.

Exposures: Anthropogenic radiofrequencies (3 kHz to 300 GHz) from real-world wireless transmitters and signal generators (or other methods used to simulate real-world device signals in a laboratory setting).

Comparators: Comparison populations exposed to lower levels, sham exposure, or no exposure (control).

Outcomes: DNA damage as depicted in Figure 1 as well as related potential biological mechanisms such as spindle disturbances, free radical production/oxidative stress, heat shock protein expression and apoptosis.

A quality assessment was performed on studies using specific quality criteria. The assessment adopted and extended the recommended quality attributes used by Vijayalaxmi and Prihoda (30). The specific quality attributes used for the selection of “higher quality” studies covered; waveform specified, exposure duration described, exposure intensity details provided, frequency of signal qualified, dosimetry calculated or measured, blinding/coding used, sham control used, and statistical methods described. A positive control was not judged to be critically important for quality determination, which is a deviation from Vijayalaxmi and Prihoda quality review protocol. While positive controls are desirable for evaluating an assay's sensitivity by comparing effects against a known genotoxic agent, they are not essential for determining cause and effect using experimental logic (34). Studies that incorporated all of these criteria were classified as higher quality (n = 130). For more details on quality assessment procedures refer to Supplementary Data Sheet 1.

The main evidence for genotoxicity, potential biological mechanisms and funding influences are summarized within an evidence map using tables, graphs, a flow chart, heat maps, and network diagrams. The systematic map database containing the selected articles, including bibliographic information and extracted data, is summarized below, with full details provided in Supplementary Data Sheet 4.

In all graphs and tables, the main analysis presents the proportion of statistically significant damage effects (displayed in orange) vs. the proportion of no significant damage (displayed in gray). All trends in studies were classified as non-significant effects, and studies showing trends or protective effects were grouped under the “no effect” category, indicating no significant DNA damage.

The PRISMA 2020 flow diagram (Figure 2) provides an overview of the search and screening process. An initial search was conducted on 12th April 2023, starting with the EMF-Portal database. A total of 3,430 candidate records were identified from all literature databases combined. After removing duplicates (n = 1674) and non-relevant or incomplete articles (n = 1,226), a total of 530 articles remained. Specific details for article exclusions are provided in Supplementary Data Sheet 4 (spreadsheet tab labeled “Review”, columns B:C).

Five articles reported on both in vitro and in vivo studies and one article covered an epidemiological and in vitro study (bringing the total number of studies to review n = 536). Of these, nineteen studies (6% of in vitro and 1% of in vivo) reported only “possible” DNA damage and were thus excluded from the main balance-of-evidence mapping process (leaving n = 517 studies for review).

Approximately 80–100% of the non-mammalian studies demonstrated statistically significant effects, particularly DNA base damage and oxidative stress (Table 1). Spindle disturbances, a potential mechanism for DNA damage, were found in 100% of studies for both mammals and plants. Although smaller in number (so conclusions are less certain), studies with plants, insects, worms, and birds have high proportions of outcomes showing RF-induced DNA damage and oxidative stress. Given the limited number of studies on amphibians, worms and snails, robust conclusions are not possible.

In vitro studies primarily focus on cellular damage across various organisms. Similarly, in vivo and certain epidemiological studies also examine specific cell types. When statistically significant DNA damage is observed in > 50% of studies for particular cell types, they are deemed more sensitive to RF-EMF exposures.

Genotoxicity findings for different cell types revealed varying sensitivities to RF exposure (see Table 2). Reproductive cells were highly sensitive, with statistically significant DNA damage effects in 80% of 20 ovarian, 80% of 30 testicular, and 74% of 27 spermatozoa studies. Insect larvae (75% of 4 studies) and embryos (55% of 20 studies) also showed differing levels of susceptibility. Normal brain cells were found to be very sensitive (76% of 54 studies), whereas neoplastic brain cells exhibited lower sensitivity (33% of 12 studies). A moderate proportion of liver and lung studies (56% of 16 studies) found damage, as well as studies of buccal (oral) mucosa cells (65% of 23 studies). This raises potential concerns due to mobile phone technology changes, which relocated the main antenna from the top of the phone to the bottom, nearer to the mouth (36).

Blood cells showed greater tolerance to RF-induced genotoxicity, particularly lymphocytes (47% of 95 studies) and leukocytes (30% of 20 studies), which were analyzed separately. However, this tolerance may be limited because longer exposures, particularly in real-world settings (e.g., epidemiological studies), showed statistically significant DNA damage across all blood cell types. On the other hand, 67% of 21 erythrocyte studies found evidence of genotoxicity. Some studies investigating genotoxicity on human blood (37–39) also demonstrate the existence of individual sensitivities and responses to radiofrequency exposures, which should not be confused with electromagnetic hypersensitivity (EHS). Some studies pooled their results, washing out potentially sensitive individual data (40–42).

Eye and skin tissues were studied less frequently, with only 33% of 12 and 15 studies, respectively, showing statistically significant DNA damage findings. Similarly, bone marrow (30% of 10 studies) showed less damage. Primary cells were more likely to present DNA damage than cell lines. These findings reveal the variability in cell sensitivity.

Of the 517 included studies that were confirmed to investigate actual DNA damage, the majority were in vitro studies, comprising 53% (n = 272) of all studies. In comparison, only 37% (n = 193) were in vivo studies, and a smaller 10% (n = 52) were observational (epidemiology) studies. There were five papers that investigated both in vitro and in vivo outcomes. The overall balance of evidence for DNA damage was 59% Effects compared to 41% No Effects (Figure 4). Outcomes for in vitro studies were slightly weighted toward No Effect (55%), whereas the majority of both the in vivo studies (75%) and the epidemiological studies (75%) reported statistically significant DNA damage. For higher quality studies, the balance of evidence for overall DNA damage shifted in favor of No Effects (52%) vs. Effects (48%).

These results reveal substantial variability in the potential for RF-EMF to induce genetic damage. In vitro studies, which dominate the evidence base, predominantly indicate no significant DNA damage. In contrast, an identical proportion of in vivo and epidemiological studies primarily indicate genotoxic effects. These opposing outcomes from studies of live organisms vs. isolated cells highlight the limitations of in vitro models for replicating the complexity of living organisms, particularly animals (44).

Animals are comprised of interconnected systems, including the nervous, endocrine, and immune systems, that can all influence cellular functions and responses (45). These systems mediate intricate feedback mechanisms, hormonal signaling, and immune responses (46). Furthermore, in vivo models include physiological processes and interactions such as blood flow, metabolic activity, and tissue-level interactions that can influence RF absorption and its biological effects (47, 48).

In vitro systems, however, are devoid of these regulatory influences, interacting systems and processes, providing a limited and potentially misleading perspective on how cells respond to RF exposure in a living organism. Additionally, the simplified environment of in vitro studies may lead to an underestimation of RF-induced genotoxic effects. To fully understand the biological impact of RF exposures, in vitro studies must be interpreted cautiously and validated through in vivo research.

Epidemiological research on the other hand, reflects real-world exposure conditions and indicates the consequences of long-term exposures. However, as the research map shows (Figure 4), epidemiological research is less common, potentially due to complexity, costs, and time requirements. Evaluating in vivo data and epidemiological data together results in a holistic view and resilient confirmation. Therefore, more weight should be given to these study types when assessing the health risks of RF exposure.

Statistically significant DNA damage was prevalent in studies investigating DNA breaks (58% of 283 studies), DNA base damage (86% of 42 studies), chromosome aberrations (63% of 100 studies), and micronuclei induction (54% of 140 studies). DNA base damage was found in the majority of studies for all study types: 92% of 12 in vitro studies, 80% of 20 in vivo studies, and 100% of 4 epidemiological studies (Figure 5C), and DNA breaks/fragmentations were found in the majority of in vivo studies (Figure 5A). Conversely, the balance of evidence was weighted toward no significant effects for studies investigating mutations (60% of 45 studies) and sister chromatid exchange (88% of 25 studies).

When only the higher-quality studies were selected from the above DNA damage subtypes, results typically showed < 50% of studies reporting damage, except for DNA base damage, where 75% of the 12 higher quality studies produced statistically significant findings. The largest shift in the balance of evidence across most DNA damage subtypes occurred for in vitro experiments, where, for higher-quality studies, the proportion showing no effects increased markedly. In contrast, the balance of evidence for in vivo studies remained largely unchanged (see Figures 5A–L, gray regions of each pie chart).

Some biological changes observed suggest DNA damage, but with no clear one-to-one correspondence between the occurrence of these biological changes and DNA damage. The evidence for these possible indicators was not included in the main systematic mapping detailed above, but instead, is presented in Figures 6A–D for completeness. Studies investigating these indicators predominantly found DNA conformational changes (93% of 46 studies) and cell apoptosis (60% of 111 studies). Selecting only the higher-quality studies had minimal impact on the balance of evidence for DNA conformational change, i.e., findings increased (100% of eight studies), but for apoptosis studies, findings decreased (50% of 26 studies). DNA conformational changes showed a consistently higher proportion of studies showing DNA alteration effects for both in vitro and in vivo higher quality studies, whereas for apoptosis, statistically significant findings decreased for both in vitro and in vivo studies.

Additive, synergistic, antagonistic, and potentiative effects can be observed and is dependent on the sequence of exposure (prior, follow, or simultaneously) and the initial functional state of the exposed biological system [(49), p. 915].

Studies were conducted to investigate adaptive or protective cell responses to radiofrequency exposures in conjunction with known genotoxic agents. This was contrasted with studies showing synergistic effects. The presence of some agents (such as gamma rays) in both categories underscores the pivotal influence of the timing and duration of an RF exposure in determining whether the cellular response is likely to be protective or synergistic (damaging) (Supplementary Table 11).

In some cases, the published data indicates that exposures to RF-EMF can produce beneficial protective effects against other genotoxic agents, suggesting the induction of an adaptive response at both the cellular and organism level (50). Potential biological mechanisms underlying these protective effects include the activation of signaling pathways, upregulation of specific gene expression and protein synthesis (e.g., HSPs, antioxidant enzymes, cell repair), enhanced activity of the DNA repair system, and a reduction in free radical levels (51).

Although the pool of studies investigating combinative and synergistic effects was relatively small, it reveals how RF exposures can have potential therapeutic benefits (cancer treatments) or can enhance the harm caused by ionizing radiation and chemical agents, depending on exposure conditions. Unfortunately, current international RF Guidelines do not consider harm caused by synergistic interactions of RF with other genotoxic agents (52).

Figure 7 summarizes research and genotoxicity findings by frequency band. The most studied bands include 900–999, 1800–1899, 1900–1999, and 2400–2499 MHz, which align with common frequencies used by mobile phones, base stations, microwave ovens, and Wi-Fi. The balance of evidence favors DNA damage effects for the 900, 1800 and 2450 MHz frequency bands. However, results for 1900 MHz, are equivocal.

Additional studies covering MHz and GHz ranges, corresponding to FM broadcasts, magnetic resonance imaging (MRI), radar, and satellite communications, are listed in Supplementary Figure 3. However, many wireless communication frequencies, especially those used by 5G new radio, remain under-researched. Consequently, novel frequencies and modulation schemes are being deployed without adequate testing for genotoxicity or broader health implications.

The data was investigated for the effects of exposure duration on study outcomes. The proportions of studies showing statistically significant DNA damage effects were graphed across a wide range of exposure durations (1 min to 1 year) for all exposure brackets containing five or more studies.

Figure 8A suggests a crude U-shaped dose response relationship, with over half of the studies in each time bracket showing effects for both short (less than half an hour) and longer exposures (more than 2 days), while less than half of the studies using exposure durations between half an hour and 2 days showed effects.

Both in vitro and in vivo studies showed the least amount of DNA damage in studies in the 16–24 h time bracket where < 30% of studies showed DNA damage. More than 50% of in vitro and in vivo studies showed statistically significant DNA damage for exposures < 15 min and >96 h. The range of exposure time where >50% in vivo studies showed damage is also lot broader than for in vitro studies (see Supplementary Figures 17a, 18a).

The exposure duration outcomes were then investigated according to specific DNA damage type. Figure 8B illustrates the percentage of studies reporting effects, each suggesting a bi- or tri- phasic response curve for most DNA damage types. This provides evidence for potential cellular adaptive responses, particularly in the 1–2 h and 16–24 h time bands, where a greater proportion of studies reported no effects. Figure 8C focuses on the existing studies using longer-term exposures (2 days or more), showing high proportions of statistically significant effects for DNA breaks, chromosome aberrations and micronuclei induction.

To simplify the analysis, exposure durations were categorized into four groups: acute (167 studies), short-term (243 studies), medium-term (108 studies), and long-term (70 studies) (see Figure 9C). Most experiments (70%) focused on short-term and acute exposures, with medium- and long-term exposures being less studied (30%). In vitro studies primarily focused on acute and short-term exposures, while in vivo studies encompassed all exposure durations, and epidemiological studies concentrated on long-term exposures. DNA damage was reported in 63% of acute, 43% of short-term, 69% of medium-term, and 76% of long-term studies (see Supplementary Table 12 for further details). These exposure duration categories were used for analyzing duration in combination with other factors, as described below.

The data was investigated for the effects of exposure intensity on study outcomes by plotting the proportion of studies showing effects for a range of exposure intensity windows (extremely low to extremely high). Figure 9A illustrates that the strength of evidence for DNA damage is moderated by intensity of exposure in a non-linear manner. Most studies have focused on near-field, low- to medium intensity exposure levels (>0.01–2.0 W/kg), typical of mobile phone use, where half or more of these studies found DNA damage effects. Fewer studies have examined higher exposure intensities (e.g., occupational settings) or extremely low intensities, which are relevant to far-field sources such as base stations, smart meters, radar, and Wi-Fi routers.

Studies investigating extremely low exposure levels (< 0.001 W/kg) reported the highest proportion of genotoxic effects (81% of 21 studies). Moreover, the proportion of statistically significant DNA damage findings decreased as intensity increased. However, at extremely high intensities (>10 W/kg), which exceed ICNIRP limits, the proportion of statistically significant DNA damage effects rose again (58% of 59 studies) (see Supplementary Table 13). These results suggest that DNA damage does not follow a linear dose-response pattern, and also suggesting that non-thermal mechanisms are likely to play a significant role in RF-induced DNA damage. Subsequently, exposure intensity was further explored by categorizing exposure relative to ICNIRP occupational limits. Surprisingly, studies at or below ICNIRP limits showed a slightly greater proportion of statistically significant genotoxic effects (58% of 439 studies) than those above the limit (54% of 92 studies) (see Supplementary Table 13).

Exposure intensity and duration were further explored to determine their combined effect on genotoxicity findings. Figure 9B shows the breakdown of Figure 9A into exposure duration categories where there are five or more studies investigating that exposure-duration combination.

The proportion of studies showing effects varied by exposure duration and intensity, following a pattern similar to Figure 9A. For studies with acute (< 1 h) and short (1–24 h) exposure durations, a high proportion showed effects at extremely low (< 0.001 W/kg) and very low (0.001–0.01 W/kg) intensities, but this proportion decreased as intensity increased. These studies reached a minimum at very high intensities (4–10 W/kg), with fewer studies showing effects, followed by an increased proportion of effects at extremely high intensities (>10 W/kg). For medium-duration studies (1 day−3 months), the proportion of studies showing effects reached an initial low point at high intensities (2–4 W/kg) but remained above 50%. At extremely high intensities (>10 W/kg), only one in five medium-duration studies showed effects. In contrast, long-duration studies (>3 months) exhibited a different trend: the proportion showing effects was lowest at medium intensities (0.5–2 W/kg) but increased at higher intensities. Notably, long-duration studies consistently showed effects in over 65% of cases across all intensity levels.

Altogether, the pattern of effects varied in a U-shape pattern across study duration, with the highest proportion of effects for long (>3 months) and then medium (< 1 day−3 months) duration studies, fewer effects for short duration studies (1–24 h) and a greater proportion of effects again for acute (< 1 h) durations. Figure 9C depicts this relationship more clearly, showing that for all exposure intensities, the proportion of studies showing effects was at a minimum for short exposure durations (1–24 h). Therefore, two U-shaped dose-response patterns emerged, one for intensity and one for duration, interacting with one another to give an overall non-linear and non-monotonic dose response pattern. This complex interaction confirms earlier indications by Lai and Levitt 2022 (43) of non-linear response patterns for both intensity and duration. It reveals that intensity alone is not the only important factor determining outcomes, and that the duration of exposure is a crucial, moderating factor. Moreover, linear models are not appropriate for describing results in this field.

These findings challenge the oversimplified ‘no effects' conclusions of some past reviews, which may have failed to account for the complex interplay between exposure intensity and duration. They also give direction for future research, to further explore these non-linear relationships.

Studies were investigated for the effects of different telecommunication signal modulations (see Supplementary Table 15). A greater proportion of studies using real mobile phone signals found DNA damage, except for GSM-Basic and CDMA communication protocols, where the small number of studies limits the ability to draw reliable conclusions. Studies using simulated signals were less consistent, with 50% or less showing evidence of genotoxicity across most modulation protocols. One exception, GSM-Talk, showed statistically significant DNA damage for 67% of 15 studies using simulated signals as well as 91% of 11 studies using a real signal. UMTS and Wi-Fi signals exhibited strong genotoxicity evidence from real signals (88% and 75% of 8 and 12 studies, respectively), but less evidence for effects (17% and 33% of 29 and 5 studies, respectively) for simulated signals. Overall, these results demonstrate that signal modulation can impact study outcomes and highlight the potential limitations of using simulated signals to evaluate genotoxic risks.

Pulsed and continuous waves both showed a greater proportion of DNA damage for in vivo and epidemiological studies and a lower proportion for in vitro studies. Continuous waves produced a greater proportion of DNA damage than pulsed waves for in vitro experiments (but still < 50% of 116 studies). When all studies are combined, both pulsed and continuous wave studies tend toward greater DNA damage. Multiple exposures, which involve separate repeated exposures on a single day or over several days, were found to yield more positive findings than a single exposure.

Comparisons of exposure patterns i.e., continuous, intermittent (e.g., 5 min on, 10 min off), and variable (dynamically changing intensity, such as during mobile phone use) reveal a progressive increase in the proportion of DNA damage findings: 55% of 369 studies for continuous, 62% of 60 studies for intermittent, and 76% of 96 studies for variable exposures.

Further effects of signal characteristics on outcomes are presented in Supplementary Tables 15–17. Across all exposure characteristics, in vivo and epidemiological studies consistently report a higher proportion of DNA damage than in vitro studies.

Altogether, the results reveal that the relationship between exposure intensity and DNA damage varies non-linearly with exposure duration. The evidence is heavily weighted by short-term, in vitro studies, where adaptive responses appear to be occurring. Signal type, frequency, and exposure patterns showed significant influences on study outcomes. Commonly used bands like 900–999, 1800–1899 and 2400–2499 MHz, used by mobile phones, smart devices, and Wi-Fi showed greater proportions of effects. 5G bands, on the other hand, are under-researched. Pulsed waves produced a greater proportion of effects for in vivo and epidemiology studies. Real-world devices consistently showed a greater proportion of effects than signal generators.

Altogether, the exposure characteristics more likely to produce DNA damage effects are those that describe authentic exposure conditions. These findings underscore the importance of future studies to incorporate diverse frequencies, realistic exposure patterns, and emerging technologies.

A review of studies investigating potential biological mechanisms (Figures 10A–F) revealed that free radical production or oxidative stress was the most frequently studied endpoint, with most studies showing effects (83% of 118 studies). Heat shock proteins were found to be expressed in just under half of the studies (46% of 28 studies), and spindle disturbances were found in all the relevant studies (100% of 10 studies). For higher-quality studies, the balance of evidence remained weighted toward effects for free radical production/oxidative stress (68% of 31 studies) and tilted toward effects for heat shock protein expression (57% of seven studies) and remained unchanged for spindle disturbances (100% of two studies).

The proportion of studies showing effects for potential DNA damage mechanisms was compared for various exposure time intervals. Only those time intervals with five or more studies (for effects or no effects combined) were analyzed.

When mechanisms were investigated across exposure time intervals, a bi- or tri- phasic response pattern was suggested (see Figure 11A), where free radical production/oxidative stress effects were dominant in studies using 30 min to 2-h exposures, and free radical production/oxidative stress and apoptosis were dominant in studies using exposure durations longer than 2 days. These results follow the same patterns of effects for DNA damage types shown in Figure 11 (also see Supplementary Figures 21–27 for more details). The possible relationships between the underlying mechanisms for DNA damage and the type of DNA damage were subsequently explored by overlaying the proportion of statistically significant evidence for each across the various exposure durations.

For example, studies on free radical production/oxidative stress and DNA breaks show similar patterns of results across the time intervals (Figure 11B), suggesting that free radical production/oxidative stress may be causally related to DNA breaks. Similarly, Figure 11 suggests strong correspondences between (C) apoptosis and DNA breaks/fragmentations, and correspondences between (D) free radical production/oxidative stress and DNA base damage.

Potential connections were also observed between (E) free radical production/oxidative stress and chromosome aberrations, with a time lag between the former and the latter; however, more evidence is needed to strengthen this conjecture. Possible further associations were also observed between spindle disturbances and chromosome aberrations, chromosome aberrations and micronuclei induction, and between free radical production/oxidative stress and both chromosome aberrations and micronuclei induction (see Supplementary Figures 23–26 for full details). The correspondences between the patterns of evidence for mechanisms and DNA damage effects in specific time bands support the finding from Figure 9 above, as well as existing theory suggesting that effects from EMF-RF exposures are stronger for certain time windows (53, 54).

Figure 11B and Supplementary Figure 30 illustrates that a strong correspondence exists between observing free radicals/oxidative stress and finding DNA damage. Evidence of oxidative DNA base damage showed the highest proportion of statistically significant findings among all forms of DNA damage investigated (Figure 5C). Two critical markers for oxidative stress and carcinogenesis (55), 8-oxo-7,8-dihydro-2′-deoxyguanosine (8-oxo-dG) and 8-hydroxy-2′-deoxyguanosine (8-OHdG), were the most frequently measured biomarkers of oxidative DNA damage. Both of these oxidative DNA damage biomarkers are seen as risk factors for many diseases, including neurodegenerative disorders (56) and cancer (57). Failure to promptly remove 8-oxo-dG can result in a base transversion point mutation, where G:C is converted to T:A during DNA replication (56).

Reactive oxygen species (ROS) are free radicals commonly observed as an endpoint in numerous RF exposure studies (58), where they are often associated with oxidative stress. Elevated markers of oxidative stress have been found in individuals with neurological disorders such as Alzheimer's disease and Parkinson's disease, as well as diabetes, cardiovascular diseases, and cancer (59). Small increases in ROS have also been observed to provide therapeutic effects; however, these effects are only evident within narrow windows of exposure intensity (60) and duration.

There are several known and potential pathways identified by which RF-EMF exposure can lead to a free radical imbalance in cells (also see Figure 12):

Reactive oxygen species are also implicated in the activation of cellular signaling pathways, including the regulation of the main pathways of apoptosis, particularly the mitochondrial-mediated pathway (69) (see Figure 12). Increased levels of reactive oxygen species have been associated with DNA damage and linked to environmental stress, with anthropogenic radiofrequency exposures being a contributor (70).

Cells employ various protective mechanisms when faced with cellular stress, including upregulating DNA repair genes, heat shock proteins, and enzymes that mitigate oxidative stress (71, 72). Gene expression is a sequential process requiring time, starting with signaling pathway activation and transcription factor binding, followed by mRNA processing and protein synthesis (73). The time-dependent activation of cellular stress responses may explain the U-shaped or bi-phasic dose response to RF-EMF exposure observed across increasing exposure durations (Figure 11). A recent study investigating free radical production has confirmed a bi-phasic cellular response to RF-EMF exposure (74).

Acute (very short) exposures to extremely high RF-EMF intensities often result in statistically significant DNA damage, possibly due to the inability of repair mechanisms to respond rapidly. In contrast, longer exposures to high intensities may lead to the activation of cellular defenses to repair damage and attempt to mitigate harm. Furthermore, prolonged or cumulative exposures could lead to accumulated DNA damage, genomic instability, and cellular dysfunction (75). Depending on the extent of the damage, cells may activate autophagy as a protective mechanism to remove damaged components and maintain cellular homeostasis (76, 77). While this may be effective in the short term, if damage persists or overwhelms repair mechanisms, cells may undergo apoptosis to remove severely compromised cells. However, should this survival process fail, there is a risk of malignant transformation (75).

Real-world RF exposures are typically chronic and variable, raising concerns about cumulative effects (78). While repair mechanisms appear to address much of the damage in the short term, prolonged or repeated exposures may overwhelm these defenses, leading to lasting genomic alterations. Given a large proportion of public exposures now occur cradle-to-grave and are often non-consensual, there is a need to address the potential long-term health risks of persistent RF exposure.

DNA conformational changes are a potential marker for DNA damage. Various methods have been employed in studies to identify changes in DNA conformation, including Raman spectroscopy, electron microscopy observations, circular dichroism, and dynamic light scattering techniques, as well as UV–vis spectroscopy. However, the most common method applied utilized Anomalous Viscosity Time Dependence (AVTD), which provides insights into structural dynamics and molecular interactions within biological systems. In particular, this method demonstrates how DNA-protein complexes (e.g., chromatin) affect the physical characteristics of the DNA, such as viscosity in a solution over time (79).

Chromatin is a DNA-protein complex found in the nucleus of eukaryotic cells. The structure of chromatin can undergo dynamic changes in response to various cellular processes, including DNA repair, transcription, and replication. Proteins such as histones and non-histone chromatin-associated proteins play crucial roles in organizing and regulating chromatin structure (80).

Results demonstrated that RF exposures can cause DNA conformational changes, with a high level of confidence (93% of studies), (see Supplementary Table 9). While the implications of such a change are not fully understood, one possible scenario is the situation where DNA damage becomes inaccessible to repair proteins or repair function is impeded by repair enzyme conformational changes (81).

While AVTD and similar methods provide valuable insights into DNA conformational changes, they cannot easily determine the underlying cause or whether it relates to DNA damage. Integrating AVTD with other experimental approaches, such as a comet assay, can help elucidate whether the observed changes are directly related to DNA damage or the result of normal cellular regulatory processes.

Micronuclei, chromosome aberrations, and spindle disturbances are all interconnected indicators of genomic instability and cellular stress (82). The mitotic spindle apparatus is responsible for chromosome segregation during cell division. Spindle disturbances can lead to improper chromosome alignment and segregation, leading to lagging chromosomes or fragments that can become encapsulated and form micronuclei (82).

A positive correlation between micronuclei and particular chromosomal aberrations in human in vitro studies has previously been identified, specifically for acentric fragments and dicentric chromosomes (83). Furthermore, micronuclei formation may be induced by chromosomal breakage or inhibition of the spindles during cell division (84).

An examination of the graphs plotting the percentage of studies reporting micronuclei induction, chromosome aberrations and spindle disturbances over exposure time, when overlayed on one another, shows a close relationship validating statements made by past researchers (see Supplementary Figures 23, 25).

Heat shock proteins (HSPs) are chaperones that protect cell macromolecules. They play a crucial role in both DNA damage repair and the cellular response to oxidative stress (71, 72). Experimental evidence suggests that RF exposure can induce HSP expression, which protects cells, but may also paradoxically aid cancer cell survival by inhibiting apoptosis. RF exposure has been shown to upregulate HSPs in human endothelial cells, potentially contributing to tumor progression (85).

Only 27 studies investigated the relationship between RF-EMF exposure and the expression of heat shock proteins, which included HSP27, HSP70 and HSP90. Notably, apart from extremely high exposure levels, an inverse relationship appears to exist between exposure intensity and detected HSP levels (see Supplementary Table 20). The evidence follows a similar pattern to the exposure-intensity and DNA damage relationship (see Supplementary Figures 20A, B and Supplementary Table 13). However, the small number of studies at some exposure levels prevent reliable conclusions from being drawn.

Apoptosis is a normal, tightly regulated process used to eliminate superfluous cells and damaged cells, including cells with DNA damage, to reduce the risk of carcinogenesis (69). The apoptotic process can be initiated by three different signaling pathways (86):

Caspase enzymes are crucial mediators of apoptosis, orchestrating the orderly dismantling of cellular components during programmed cell death. Their activation is a hallmark of apoptosis (87), with several studies in this evidence map using caspase assays for the detection of apoptotic events.

Analysis of the results reveals a strong correspondence between DNA damage and apoptosis, as evidenced in (Figure 11C), where numerous studies report both outcomes following RF-EMF exposure. This close relationship raises an important question: in these studies, was apoptosis triggered by excessive DNA damage from RF-EMF, resulting in controlled cell death or was the detected DNA damage the outcome of apoptosis? Both scenarios are plausible. Apoptosis is a cellular mechanism initiated when accumulated damage, including DNA damage, exceeds the cell's capacity to repair and recover (86). The apoptotic biochemical process can include DNA cleavage by activating endogenous endonucleases (88), leading to DNA fragmentation. However, many studies investigating apoptosis did not provide sufficient information to verify the principal initiation mechanism.

The factors above showing effects on study outcomes were collectively examined for any experimental parameters that may explain the lower proportion of DNA damage effects found in higher-quality studies compared to all studies. It was found that higher-quality studies have:

These extra parameters listed above may contribute to the lower balance of evidence for effects observed in higher-quality studies, just as much as the established quality criteria (30).

Real-world wireless devices are expected to comply with public exposure limits recommended by ICNIRP (15). However, calculating dosimetry in studies utilizing these devices is complex due to the variable nature of the emitted signals (89). Many studies were found to lack adequate information on dosimetry or the methods used for its calculation or measurement.

Some researchers have criticized studies employing real-world devices for “poor dosimetry,” using this argument to question study quality and downplay findings (30, 90). However, minimal differences were observed in the balance of evidence when comparing so called studies with “poor dosimetry ” (which typically represent exposures under real-life conditions) to those with “sufficient dosimetry.” Furthermore, considering these unmodified real-world devices operate within, and often well below ICNIRP's public exposure limits, the critique of poor dosimetry becomes less pertinent when evaluating the validity of current public safety standards.

The charts presented in Figure 13 display the balance of evidence by parameter, grouped by primary funding source. (Note that studies may have multiple funding sources including institutional, industry, or government. Therefore, a filter was applied to focus on one funding source for each graph; Note that military and telecom regulator funding was explicitly excluded from government funding).

Research funded by vested interests (e.g., industry, government telecom regulators, and to a lesser extent, military—see section “Classification of potential vested interests”; Supplementary Data Sheet 1) shows a greater proportion of studies concluding “no DNA damage” (as illustrated by the amount of gray color) compared to studies funded by governments, institutions, or private/public sources (not shown). Notably, studies funded by the USAF resemble industry-funded research more than typical military-funded studies (see Supplementary Data Sheet 5, Supplementary Table 25 and Supplementary Figure 35 for full details).

Funding sources appear to have determined various experimental parameters, corresponding to research priorities and potential biases. Studies with potential conflicts of interest (COI), such as those funded by industry, military, or government telecom regulators, organizations who collectively have vested interests in RF technology, exhibit the following research trends:

Some of the above study design decisions are more likely to produce null results (as discussed in sections above).

In contrast, independent research i.e., research not funded by industry, telecom regulator or the military, demonstrates broader coverage of study parameters, characterized by:

Industry, Telecom regulator and USAF-funded studies all have a similar balance of evidence profiles (Supplementary Table 25). In these cases, most studies (>80%) show no significant DNA damage. However, subtle differences are seen between them at the experimental level.

The impact of funding on study outcomes is further compounded by the lack of transparency and disclosure, with 173 studies (~33% of studies) missing a formal funding statement.

The designs and results of research studies that received funding from industry, telecom regulator or military were compared with research studies classified as independent. The two research groups (potential vested interest vs. independent) favored different study types and showed opposing proportions of effects in many cases (refer to Supplementary Table 27). Research funded by, or affiliated with industry or the military, primarily comprised in vitro studies, with 30% of 142 in vitro and 26% of 31 in vivo studies reporting statistically significant findings. Quality filtering did not significantly affect these results (25% of 56 in vitro studies and 25% of 12 in vivo studies respectively).

In contrast, 63% of the 465 experimental studies were classified as independent research, with a mix of 130 (44.5%) in vitro and 162 (55.5%) in vivo studies. Independent in vitro studies, regardless of quality, showed more null results than in vivo studies. Overall, study quality had little impact on independent research, with 74% of all independent studies (n = 292) showing statistically significant DNA damage compared to 73% of 62 higher-quality studies.

Higher-quality studies tended to report more conservative results (toward no significant effects), consistent with prior RF-EMF reviews by Wood et al. (95), Karipidis et al. (96), Vijayalaxmi and Prohida (30), and Simko et al. (97). However, these reviews did not assess the combined influence of potential vested interests (e.g., funding or author affiliations) and study quality on outcomes. The significant differences between research group findings (potential vested interests vs. independent) raise serious concerns about the reliability and validity of past review data (30, 95, 96), rendering their effect size comparisons as a potentially unreliable measure, particularly as funding source influence was not considered in their analysis. This finding aligns with previous related research that reported no significant effects of mobile phone use on brain tumors often had industry affiliations either through funding or influence on study design (93, 94). Together, these findings suggest that funding and author affiliation biases study outcomes more than study quality.

Studies funded by industry groups, specific military organizations (particularly the U.S. Air Force), or telecommunications regulators were skewed toward reporting null findings. This is likely due to the experimental methodologies employed by these studies, with a disproportionately large number being conducted as (i) in vitro experiments with (ii) short exposure durations (iii) with simple signal modulations from (iv) signal generators, where the likelihood of observing statistically significant DNA damage is much lower than when using real-world devices emitting signals carrying voice or data over longer exposures. These design decisions appear to bias results toward null outcomes. Supplementary Table 27 shows how studies funded by organizations with vested interests are biased toward in vitro experiments, resulting in a lower proportion of studies finding effects.

On the other hand, independent funded research has conducted a much broader range of experiments and primarily reported statistically significant DNA damage. These findings suggest that, compared to all studies overall, the more conservative results of higher-quality studies are likely due to the high proportion of higher-quality studies being performed by researchers who have received funding from vested interests, the greater prevalence of in vitro studies with short exposures from signal generators, and the exclusion of epidemiological studies due to dosimetry limitations.

The bibliographic network charts in Supplementary Figures 54–56 illustrate the connections between researchers and research groups in this area.

Papers concluding that no DNA damage effects have been observed are primarily published in three journals: Radiation Research, Bioelectromagnetics, and the International Journal of Molecular Sciences. Papers concluding that DNA damage effects have been found appear in some of the same journals but are less concentrated. Overall, independent research has been published across a wide array of journals. In contrast, research associated with industry or the military through funding or employment, has to a larger degree been confined to a small group of journals (see Supplementary Table 29); e.g., Radiation Research (98). The clustering of studies funded by vested interests in select journals can distort the perceived balance of evidence, potentially misleading policymakers, radiation safety practitioners, and the public.

All the above demonstrate how methodological preferences and potential conflicts of interest inherently bias the results of studies (and reviews), reducing the robustness of the overall evidence base and thereby creating uncertainty.

This pattern of bias has been documented in other lucrative industries, such as pharmaceuticals (99), agriculture, and chemical manufacturing (100). Such biases must be acknowledged and accounted for to ensure the integrity and reliability of published findings, particularly when conducting future systematic reviews.

Many issues were identified in both experimental and epidemiological studies, separate from those required for “quality” assessment. Some of the more important issues are summarized below.

The purity of the signal-free sham or control environments could not be confirmed for most studies, because background measurements were either not taken or not reported. Verification of the test environment's EMF integrity and identification of potential stray fields acting as confounding factors was therefore not possible.

Many studies were found to be missing important methodological details. The absence of transparency in methods makes it difficult to evaluate study quality or to replicate studies. Such studies were classified as lower quality studies but were still included in the evaluation of DNA damage outcomes.

Some studies only presented pooled data, which can hide potentially “sensitive responders”; e.g., the blood of some individuals showed more damage from exposure from others (40–42), similar response have been seen in millimeter wavelength research with animals (109, 110).

Several epidemiological studies investigating buccal mucosa micronuclei from long-term mobile phone usage found no statistically significant difference between the left and right cheek. The authors did not consider nonlinear responses or the penetration capability of the RF-EMF from mobile phones. A more relevant comparison is between heavy and light mobile phone usage.

More specific concerns identified in studies are noted in the Systematic Evidence Database (see Supplementary Data Sheet 4, “Final Study List (Experimental)”: column CH and “Final Study List (Epidemiological)”: column CF).

An analysis of study parameters reveals significant variability across studies, which differ in exposure time, intensity, frequencies, modulations, assay methods, DNA assessment timings, equipment, organism and cell types, among others. Assays also use varying stains, preparation techniques, and timings, which can impact sensitivity. Even studies claiming to replicate others often have subtle differences, such as variations in animal species and comet assay parameters as described below.

The evidence map and synthesis presented here have identified several important research gaps. Firstly, there were limited studies investigating the genotoxic effects of RF on non-mammalian species, including insects, birds, and trees. There is also a complete absence of studies on reptiles, and within mammals, there is scant research beyond human and rodent studies.

Most human experimental studies utilized conditions that did not reflect real-life exposure scenarios, typically involving in vitro experiments with short, single, continuous exposures to a simulated signal involving one specific frequency. In contrast, real-world exposures are characterized by chronic, simultaneous exposures to multiple signals overlayed with a variety of modulation patterns and intensities. Such conditions were not used in typical laboratory studies.

There were also few studies using higher frequencies, such as 5G technologies, despite their growing prevalence. The research is therefore lagging behind technological advancements, leaving potential risks largely unexplored.

Inconsistencies observed in past research stem from the complex interplay of numerous variables. For example, both therapeutic and potentially harmful effects have been identified, with this duality possibly tied to the combined effects of intensity and duration of exposures. To clarify these complex relationships, future studies should methodically address gaps in key experimental dimensions, such as exploring long durations and extremely low to very low exposure intensities (see Figure 9B). Additionally, few studies have investigated the combined effects of RF-EMF with other environmental stressors, which could amplify genotoxic outcomes in diverse settings and need further investigation (111).