Of the 3,430 candidate studies, 517 were identified as relevant. The relevant studies were classified into 3 categories: in vitro (53%), in vivo (37%), and epidemiological (10%) studies. Overall, the balance of evidence for DNA damage was 59% "effect" versus 41% "no effect."
The results of the in vitro studies leaned slightly toward "no effect" (55%), whereas most of the in vivo (75%) and epidemiological (75%) studies reported statistically significant DNA damage. In higher-quality studies, the balance shifted slightly toward "no effect" (52%) versus "effect" (48%). The simplified environment of in vitro studies may lead to an underestimation of RF-induced genotoxic effects. To fully understand the biological effects of RF exposure, in vitro studies must be interpreted with caution and validated by in vivo research. Human studies showed an almost even split (51% versus 49%) in the evidence. In vivo and epidemiological studies consistently reported statistically significant DNA damage, while in vitro studies tended to show negative results. Similar trends were observed in rats (75% "effect" studies) and other mammals (73%). All studies on plants, worms, birds, and amphibians reported statistically significant DNA damage. Similarly, 71% of insect studies identified statistically significant DNA damage. Approximately 80–100% of non-mammal studies showed statistically significant effects, particularly DNA base damage and oxidative stress. Spindle disruption, a potential mechanism of DNA damage, was found in 100% of mammalian and plant studies. The highest proportion of genotoxic effects (81% of 21 studies) was reported in studies examining extremely low exposure levels (< 0.001 W/kg). Furthermore, the proportion of statistically significant DNA damage findings decreased with increasing intensity. However, at extremely high intensities (> 10 W/kg) exceeding ICNIRP limits, the proportion of statistically significant DNA damage effects increased again (58% of 59 studies). These results suggest that DNA damage does not follow a linear dose–response pattern but rather a U-shaped curve and that non-thermal mechanisms likely play a significant role in RF-induced DNA damage.
In addition, the effects varied in a U-shaped pattern depending on the duration of the study's exposure. The highest proportion of effects occurred in long-term studies (> 3 months), followed by medium-term studies (1–3 months). The fewest effects occurred in short-term studies (1–24 hours), while a higher proportion of effects occurred in acute studies (< 1 hour). This complex interaction confirms Lai and Levitt's (2022) earlier evidence of non-linear response patterns in terms of both intensity and duration. Linear models are not suitable for describing the results in this area. These results challenge overly simplistic conclusions from earlier reviews that found no effects. Elevated concentrations of reactive oxygen species have been linked to DNA damage. In response to cellular stress, cells employ various protective mechanisms, including the upregulation of DNA repair genes, heat shock proteins, and enzymes that mitigate oxidative stress. Gene expression is a sequential, time-consuming process that begins with the activation of signaling pathways and the binding of transcription factors. This is followed by mRNA processing and protein synthesis. These adaptive mechanisms likely explain the observed U-shaped curves for EMF exposure duration. Real-world RF exposure is usually chronic and variable, raising concerns about its cumulative effects. Although repair mechanisms appear to resolve much of the damage in the short term, prolonged or repeated exposure may overwhelm these defense mechanisms, resulting in permanent genomic changes.
A comprehensive analysis examining potential biases, such as authors' affiliations, funding sources, and journals used for publication, revealed that these factors strongly influence reported outcomes. Studies funded by interest groups (e.g. industry, government telecommunications regulators, and the U.S. Air Force) were more likely to conclude that there was "no DNA damage" than studies funded by governments, institutions, or private/public sources.
Industry- or military-funded research primarily consisted of in vitro studies: 30% of the 142 in vitro studies reported statistically significant results compared to 26% of the 31 in vivo studies. Quality filtering had no significant impact on these results (25% of 56 in vitro studies and 25% of 12 in vivo studies). In contrast, 63% of the 465 experimental studies were classified as independent research, including 130 in vitro studies (44.5%) and 162 in vivo studies (55.5%). Overall, study quality had little influence on independent research. Seventy-four percent of all independent studies (n = 292) showed statistically significant DNA damage, compared to 73% of the 62 higher-quality studies.
Independently funded research conducted a much broader range of experiments and primarily reported statistically significant DNA damage. These findings suggest that the more conservative results of higher-quality studies, compared to all studies overall, are likely due to the high proportion of industry-funded researchers conducting higher-quality studies, the larger proportion of in vitro studies with short exposure times, and the exclusion of epidemiological studies.