Article publication date
May 2025
ARPANSA review date
May 2025
Summary
This systematic review evaluated the literature for evidence of a relationship between radiofrequency electromagnetic field (RF-EMF) exposure and cancer incidence as reported in animal (in vivo) studies. After searching the literature for appropriate articles, 52 studies were found suitable for inclusion in the review. Each article was assessed for risk of bias (RoB) according to the method described by OHAT. Cancer endpoints where a statistically significant result was reported in the literature were progressed to full analysis and a certainty of evidence (CoE) evaluation was conducted according to GRADE. The review reported that the included studies were too heterogeneous to combine in a meta-analysis and the authors elected to base their findings on statistically significant results which occurred in a very small subset of the included studies.
No evidence was found for any cancer outcome in most biological systems including gastrointestinal, kidney, mammary gland, urinary, endocrine, musculoskeletal, reproductive and auditory. The review reported moderate certainty evidence for an increased incidence of lymphoma, liver cancer, and lung and adrenal gland tumours in response to RF-EMF exposure. The review also reported high certainty evidence for an increased incidence of brain cancer (glioma) and heart cancer (malignant schwannoma) in response to RF-EMF exposure.
Published in
Environment International
Link to the study
Effects of radiofrequency electromagnetic field exposure on cancer in laboratory animal studies, a systematic review
Commentary by ARPANSA
The systematic review conducted a high-quality search and identified all the appropriate studies, collating a large evidence base. However, their results placed particular emphasis on just two studies: the National Toxicology Program (NTP) 2018 study and the study by Falcioni et al (2018). Many of the methodological failures of those studies, outlined in reviews by health organizations including ARPANSA, ICNIRP, and the US FDA, were not adequately considered in this review. Specifically, issues such as the lack of blinding in pathological assessments, unexplained differences in cancer incidence between male and female animals, and the longer lifespans of exposed animals resulting in a higher chance for tumours to develop were overlooked. Additionally, the studies failed to account for chance in their statistical analyses, leading to potential false positive results. For example, with 12,800 tests made in the NTP (2018) study, many hundreds would be expected to be significant due to chance alone. These methodological weaknesses, combined with inconsistencies in the findings between the two studies and with other literature, make it difficult to reconcile the weight the authors placed on the conclusions regarding RF-EMFs and cancer risk from these studies.
One of the most important parts of a systematic review is the synthesis of results. This is when all the results from previous studies are combined, allowing the evidence to be assessed as a whole rather than as individual pieces. Unfortunately, the authors of this systematic review elected not to consider the evidence in its totality and instead focused on a small number of positive findings. The review considered RF-EMF to have an effect on cancer if at least one study showed a significant effect (on exposed versus not-exposed) or significant trend (across different exposure levels) regardless of other studies not finding an effect or trend. Further, a trend could be calculated as statistically significant even if it was based on individual non-significant results across different exposure levels. This issue is demonstrated repeatedly in the review, most prominently in the assessment of brain tumours, lymphoma, and heart schwannomas. It is particularly egregious for glioma, with the authors purporting a high certainty of an increased risk of brain tumours. This is despite there being no single significant result across all studies, with the reviews reported outcome based on insignificant results from a single paper (NTP, 2018) that have been interpreted as a significant trend. Furthermore, the very low actual numbers of glioma tumours reported in the NTP (2018) study make this trend very tenuous, as a single tumour in the control group would have resulted in a non-significant trend. The review focused on this trend in their results and ignored the many non-significant outcomes reported by other studies. The approach taken by the authors to consider only positive outcomes fails to consider all of the available evidence, ultimately defeating the purpose of conducting a review.
Another point to consider with the synthesis is the absence of a meta-analysis. The authors reported that the studies were too heterogeneous for a meta-analysis to be performed. Although this rationale could be considered valid, in another recent systematic review on cancer in laboratory animals conducted by Pinto et al. (2023), a meta-analysis was performed. The Pinto et al systematic review found that there was low or inadequate evidence for an association between RF-EMF exposure and the onset of tumours of any type.
The risk of bias (RoB) assessment in the current systematic review aimed to evaluate the methodological rigor and transparency of included studies by considering six key bias domains: selection bias, performance bias, detection bias, attrition bias, selective reporting and other sources of bias. There are several issues with the conduct of the RoB in this review which results in inappropriately generous characterisations of the included studies. Many of the limitations of the included studies, particularly those reporting positive findings for glioma and malignant heart schwannomas in male rats (Falcioni et al., 2018; NTP, 2018a; b), were not adequately considered. The authors did not acknowledge the well-documented flaws with these two studies that have been extensively commented on by ARPANSA and ICNIRP and have consequently under-represented the RoB in the conclusions of these studies. These choices in the RoB assessment are crucial as these two studies overwhelmingly informed the conclusions of the present review.
The systematic review’s evaluation of statistical methods used by the included studies was also inadequate. This is again particularly problematic for the key studies used to support their conclusions. In the NTP (2018) study, sharing of control groups amplified the likelihood of outcomes being found by chance and the lack of corrections for multiple comparisons testing was not adequately reflected as a confounding factor in the RoB assessment. Similarly, the issues with using significant trends in place of significant results were not adequately considered. When claiming that non-significant results can result in significant trends, the trend should be interpreted with caution (Nead et al, 2018). This is particularly true for the NTP study where the shared control groups and lack of multiple comparisons testing compound the issue, biasing toward a positive result. A more reasonable RoB assessment for the same set of studies was presented in the recent systematic review by Pinto et al. (2023). In that systematic review, the authors rated the NTP studies at a higher RoB compared to the current systematic review.
The assessment of the certainty of evidence (CoE) in the review is based on the RoB in conjunction with other factors downgrading the CoE, such as inconsistency, indirectness, imprecision, and publication bias as well as potential factors upgrading the CoE such as a dose response. A consequence of the generous RoB assessment discussed above is that the RoB can’t have a real influence on the CoE, which ultimately brings into question the higher certainties presented in the review. Further problems arise when examining the robustness of their CoE assessment in the other domains. For example, inconsistency is another major issue across the evidence base that is downplayed by the systematic review. This problem first arises in the way the authors have chosen to synthesise their conclusions but is repeated in the CoE evaluation where non-significant results are largely ignored. Even internal inconsistencies within the NTP study are not judged as cause for a downgrade as is the case for glioma. Another example would be schwannomas, which can occur in a few organs (e.g. salivary gland schwannomas, metastatic, trigeminal ganglion) but only occurred as a significant result in the heart. A related issue is their commentary on the variance of animal models when evaluating inconsistency for CoE. If the evidence base used different animal models, the authors stated that this did not warrant a downgrade in the CoE for inconsistency. However, they also listed this as a reason not to do a meta-analysis and these two justifications oppose each other.
The CoE assessment downplays the indirectness of the animal models and exposure magnitudes in relation to human health outcomes. RF-EMF interacts differently with smaller animals. While higher frequency RF-EMF can reach the heart or liver of a rat or mouse, it cannot do so in humans (Basandraia & Dhamia, 2016; ICNIRP, 2020). This reduces the direct applicability of cancer outcomes in animals, particularly the schwannomas reported in the hearts of rats, to humans. Further, much of the RF-EMF exposure employed in the NTP studies was higher than the whole-body exposure limits (e.g., 6 W kg−1) ((ICNIRP, 2020; ARPANSA, 2021). Such high exposures are not encountered by people in the everyday environment and cannot be realistically associated with human cancer studies.
The CoE assessment wrongly upgraded the certainty in the evidence for brain cancer based on a dose-response relationship with RF EMF exposure, as determined by the authors. As mentioned earlier a dose-response is reported only in the NTP (2018) study based on non-significant individual results across different exposure levels. In contrast, the recent Pinto et al (2023) systematic review did not find a dose-response relationship for brain cancer or any other tumour and did not subsequently upgrade the CoE. The ultimate consequence of the author’s systematic failings in both the CoE and RoB assessments is that results are given a higher certainty rating that is not supported by the scientific evidence.
In addition to the above major flaws there are numerous minor issues scattered throughout the document that degrade the overall article quality. These include undocumented deviations from the published protocol, reference to supplementary information that is not provided, mischaracterisation of exposure level comparisons and self-contradictions in the discussion. While the prominent issues discussed at length above have a greater impact on the interpretation of the review, the extent of additional issues complicates this interpretation and introduces significant doubt surrounding the rigour in which the substantive parts of the article have been conducted.
This review is a part of the WHO commissioned systematic review process, the overall aim of which is to assess the possible implications of RF-EMF exposure on human health. ARPANSA has therefore considered this review carefully and thoroughly when evaluating the scientific evidence regarding effects of RF-EMF exposures. In determining the impact of RF-EMF on human health the most significant evidence comes from studies on humans, not animals. This is because human observational studies are higher in the evidence hierarchy compared to animal studies as they provide more direct and relevant information about human health and disease. The WHO commissioned systematic reviews looking at observational studies in humans did not find an association between RF and any cancer (Karipidis et al 2024, Karipidis et al 2025).
In Australia, exposures to RF-EMF are covered by the Australian radiofrequency standard RPS-S1. This Australian standard was published in 2021, based on revised guidelines by ICNIRP (2020), which considered all of the relevant scientific evidence available at the time, including both the NTP and Falcioni studies. As the major conclusions of this review are primarily based on these two studies and there is no synthesis of all the available evidence, it is ARPANSA’s assessment that no new results are presented by the review and therefore no reasons for policy revisions. This systematic review on laboratory animals does not change the assessment of ARPANSA that there is no substantiated evidence of health effects from RF-EMF exposure below the ARPANSA safety limits.
