Do mobile phones cause brain cancer?

The evidence now available says no. Brain tumour incidence has remained flat through thirty years in which phone use became universal — the pattern harm would have to somehow evade. Prospective studies that record use rather than asking patients to remember it, including the quarter-million-person COSMOS cohort reported in 2024, find no increased risk even among the heaviest long-term users, and a WHO-commissioned systematic review of 63 studies reached the same conclusion the same year. The residual positive signals come from case-control studies with documented recall bias. Certainty about 40-year latencies awaits 40 years of data, but every line of strong evidence is currently null.

Then why did the WHO cancer agency call it 'possibly carcinogenic'?

IARC's 2011 Group 2B classification is a hazard label meaning limited evidence that could not be entirely dismissed at the time — a category whose historical members include pickled vegetables. It rested mainly on the case-control studies whose recall-bias problems are now well characterised, and it predates the prospective cohorts and the 2024 systematic review. IARC classifications are also rarely revisited quickly. Citing 'WHO says possibly carcinogenic' against 'a WHO-commissioned review found no link' is quoting a 2011 evidence snapshot against the 2024 evidence — the same organisation's process, thirteen years of data apart.

Is 5G more dangerous than 4G?

No property of 5G points that way. Most 5G traffic runs on frequencies comparable to 4G; the new millimetre-wave bands are absorbed in the outermost millimetre of skin, making them less penetrating than their predecessors; and denser antenna networks cause phones — always the dominant personal exposure — to transmit at lower power. Compliance measurements around operating 5G sites find exposures at small fractions of international limits. The claims that circulated in 2020, up to and including a causal role in a viral pandemic, were not scientific hypotheses that failed testing; they were never candidates, and their main documented harm was arson against infrastructure and the technicians maintaining it.

I get headaches near Wi-Fi — is electromagnetic sensitivity real?

The symptoms are real; the attributed cause does not survive testing. In dozens of double-blind provocation experiments, people identifying as electrosensitive could not detect whether a field was on at better than coin-flip rates, and their symptoms tracked belief about exposure rather than exposure. This is nocebo — genuine physiology triggered by expectation, the mirror image of placebo — and it can be experimentally induced by alarming media. That reframing points to what helps: approaches used for persistent physical symptom syndromes, including CBT-based therapy, have trial support, while shielding products entrench the attribution and can even raise real exposure by forcing phones to transmit harder. Taking the suffering seriously and the attribution sceptically is not a contradiction; it is the evidence.

Should children limit phone and Wi-Fi exposure?

On cancer grounds, the data are reassuring — the MOBI-Kids international study found no association between wireless phone use and childhood brain tumours, and no agency sets child-specific RF limits below ICNIRP. The strong, boring, evidence-based reasons to manage children's phones lie elsewhere entirely: sleep displacement, attention, sedentary time and the contested territory of social media and mental health. A family worried about radiofrequency but relaxed about a phone in the bedroom at midnight has the risk ordering inverted. For anyone wanting exposure reduction anyway, physics makes it free: distance and speakerphone cut head exposure by orders of magnitude more than any product sold for the purpose.

Does the flat brain tumour trend really rule harm out?

It rules out anything sizeable. Incidence analyses have estimated that the relative risks reported by the alarming case-control studies would have produced clearly visible rises in national registries — rises that did not occur; modest increases seen in some registries for specific subtypes are largely attributable to imaging and diagnostic changes and predate mobile saturation. What trend data cannot exclude is a very small effect, a very rare susceptible subgroup, or latencies beyond the observation window — which is why cohort follow-up continues and why honest communication says 'no detected risk despite great statistical power' rather than 'proven absolutely safe'. In risk communication those sentences must be kept distinct; in practical decision-making they point the same way.

Home › Topics › EMF and 5G

EMF and 5G

GMJ News knowledge hub · last reviewed August 2026 · Georgian Medical Journal

Summary Statistics Glossary GMJ news FAQ Documents Organizations Research

Radiofrequency electromagnetic fields from mobile phones, Wi-Fi and 5G networks are among the most feared and least dangerous exposures in modern life: the radiation is non-ionising, incapable of breaking chemical bonds, and decades of accumulating evidence — including a 2024 WHO-commissioned systematic review covering 63 studies — show no association between mobile phone use and brain cancer, while brain tumour incidence has stayed flat through thirty years of explosive phone adoption. The much-cited IARC “possibly carcinogenic” (Group 2B) classification from 2011 is a weak-evidence category, not a finding of harm. Electromagnetic hypersensitivity deserves a careful answer of its own: the suffering is real, but blinded provocation studies consistently show sufferers cannot detect fields at better than chance — the symptoms are genuine and the attributed cause is not (WHO: electromagnetic fields and public health).

Key messages

SETTLED: the physics sets the terms

Radiofrequency fields from phones, Wi-Fi and 5G are non-ionising: photon energies millions of times too low to break chemical bonds or damage DNA directly, which forecloses the mechanism by which ionising radiation causes cancer. The only established biological effect at environmental intensities is tissue heating, and international exposure limits (ICNIRP) are set fifty-fold below heating thresholds. Any claimed harm therefore requires a novel non-thermal mechanism — and after decades of searching, none has been reproducibly demonstrated. This does not make harm impossible; it makes the prior probability low and the burden of evidence correspondingly high.

SETTLED: the population data are reassuring at scale

Mobile phone use went from zero to near-universal within a generation — a massive natural experiment. If phones caused brain tumours at meaningful rates, incidence should have risen; registry data across countries show age-adjusted brain tumour rates essentially flat over thirty years. Large cohorts — the Danish subscriber cohort, the UK Million Women Study, the prospective COSMOS study reported in 2024 — find no association, and a WHO-commissioned systematic review published in 2024, covering 63 studies from 22 countries, concluded mobile phone use is not linked to brain cancer, including with long-term and heavy use.

CONTEXT MOST COVERAGE OMITS: what IARC 2B meant

The 2011 IARC classification of radiofrequency fields as Group 2B — 'possibly carcinogenic' — is the most miscited datum in the debate. Group 2B is the weak-evidence category, historically shared by pickled vegetables and aloe vera extract; it rested largely on case-control studies (Interphone and Hardell) whose self-reported phone-use data carry documented recall bias, producing implausible artefacts such as apparent protective effects of moderate use. The prospective evidence and incidence trends accumulated since 2011 point the other way, and the 2024 WHO-commissioned review is the systematic expression of that shift.

SETTLED IN AN UNCOMFORTABLE DIRECTION: electromagnetic hypersensitivity

People reporting EHS suffer genuinely — headaches, fatigue, cognitive complaints, sometimes disabling. But across dozens of double-blind provocation studies, sufferers cannot detect the presence of fields at better than chance, and symptoms follow the belief that a field is present rather than the field itself: the nocebo mechanism, demonstrated repeatedly including with sham 5G exposure. WHO's position is precise — the symptoms are real, and no scientific basis links them to EMF exposure. The clinically useful consequence: treat the suffering (which responds to CBT-based approaches in trials) rather than the attributed cause, and recognise that media scare coverage measurably generates the syndrome.

UNSUPPORTED: the 5G-specific claims

5G changed frequencies and antenna architecture, not the physics: most 5G runs on frequencies similar to 4G, and millimetre-wave 5G penetrates skin less than a millimetre — more superficial, not more dangerous. Claims that 5G suppresses immunity, spread COVID-19 (a virus, transmitted between people including in countries without any 5G), or constitutes an untested exposure are unsupported; measured public exposures near 5G infrastructure run at small fractions of ICNIRP limits, typically far below those from the phone against one's own head. The 2020 arson attacks on telecom masts remain a case study in what unrebutted misinformation does.

GENUINELY OPEN, NARROWLY: what honest uncertainty remains

Residual questions exist and should be named rather than rounded to zero: very-long-latency effects beyond 30 years cannot yet be fully excluded; childhood exposure has shorter observational history (though MOBI-Kids found no association); and a minority of laboratory reports of non-thermal biological effects — the NTP rodent study at whole-body exposures above human limits being the most prominent — remain contested and unreplicated at human-relevant conditions. Continued monitoring through cohorts such as COSMOS is the proportionate response. None of this uncertainty is of a size that justifies precautionary limits below ICNIRP, and no health agency recommends them.

Key statistics

63 studies

from 22 countries in the 2024 WHO-commissioned systematic review — no link between mobile phones and brain cancer

Karipidis et al. 2024

Flat

age-adjusted brain tumour incidence across three decades of explosive mobile phone adoption

Cancer registries

50-fold

safety margin below established thermal-effect thresholds built into ICNIRP public exposure limits

ICNIRP 2020

Group 2B

the 2011 IARC category for RF fields — the weak-evidence tier, based on recall-bias-prone case-control data

IARC

Chance level

accuracy with which self-identified EHS sufferers detect real vs sham fields in double-blind provocation studies

Systematic reviews

<1 mm

skin penetration depth of millimetre-wave 5G — more superficial than lower frequencies, not more penetrating

Biophysics

EMF and 5G — where the disagreement actually lies

Source: Bars show strength of supporting evidence. High values indicate well-supported propositions; low values indicate claims tested and not supported.

Glossary of key terms

Ionising versus non-ionising radiation
PhysicsThe word 'radiation' does the scare's heavy lifting, so the distinction is worth stating exactly. Ionising radiation (X-rays, gamma rays) carries photon energies above roughly 10 electronvolts — enough to strip electrons and break DNA, the initiating event of radiation carcinogenesis. Radiofrequency photons carry around a millionth of that; no accumulation of them ionises anything, any more than many dim lamps make a laser. Their only established effect on tissue at high intensity is dielectric heating — the microwave-oven mechanism — and exposure limits are set far below where measurable heating begins. Sunlight, by contrast, includes UV that borders the ionising range: the radiation people fear least is the one that demonstrably causes cancer.

Interphone, Hardell and recall bias
Evidence/HistoryThe case-control studies underlying the 2011 IARC 2B classification asked people with and without brain tumours to recall their historical phone use. Tumour patients searching for explanations systematically over-recall exposure — and Interphone's own oddities betray the problem: most use categories showed apparently reduced tumour risk, an artefact no one believes, with elevation only in the highest self-reported decile, driven by implausible reported usage. The Swedish Hardell studies reporting stronger effects have faced sustained methodological criticism. Prospective designs, which record use before diagnosis, eliminate this bias — and uniformly find nothing. The RF story is a textbook case of study design determining conclusions.

The COSMOS study and the 2024 WHO review
EvidenceCOSMOS is the prospective answer to the recall-bias problem: about 250,000 adults across six European countries enrolled with operator-recorded and baseline-reported phone use, followed for over a decade; its 2024 tumour analysis found no increased risk of glioma, meningioma or acoustic neuroma even in the highest cumulative-use groups. The same year, a WHO-commissioned systematic review led by Karipidis synthesised 63 human observational studies and concluded mobile phone use is not associated with brain cancer or other head and neck cancers, including long-term users. Together they represent the field's mature verdict, and the basis on which WHO's ongoing RF health risk monograph proceeds.

The NTP rodent study — the strongest opposing card
ToxicologyThe US National Toxicology Program exposed rats and mice to whole-body RF for nine hours daily across their lifespans at intensities up to four times the human whole-body limit, reporting 'clear evidence' of heart schwannomas in male rats — with no effect in females or mice, exposed animals living longer than controls, and whole-body dosing that no phone user approaches (human exposure is local to the head at far lower whole-body average). A companion Italian study (Ramazzini) reported partially consistent findings at lower intensities. These studies are why honest accounts do not claim zero laboratory signal; the combination of internal inconsistencies, exposure irrelevance and null human data is why no agency changed its risk conclusion on their basis.

Electromagnetic hypersensitivity and the nocebo mechanism
ClinicalEHS describes attribution of real symptoms — headache, fatigue, tinnitus, cognitive fog — to EMF exposure. The provocation literature is unusually decisive: in double-blind conditions sufferers neither detect fields nor develop symptoms in response to real versus sham exposure, while symptoms reliably appear when they believe a field is on. Experiments have induced symptoms in healthy volunteers simply by showing alarming documentaries before sham exposure — nocebo generation in real time. Clinically this reframes rather than dismisses: the suffering is genuine, its driver is expectation, effective approaches resemble those for other persistent physical symptom syndromes, and 'shielding' products (some of which measurably increase exposure by blocking handshake signals, causing phones to boost power) reinforce the attribution that sustains the condition.

5G, millimetre waves and small cells
TechnologyThree facts dissolve most 5G-specific fear. First, the majority of 5G deployment uses frequencies below 6 GHz, overlapping or adjacent to 4G bands used for years. Second, the novel millimetre-wave bands (24 GHz upward) penetrate tissue less than lower frequencies — energy is absorbed within the top millimetre of skin — so 'higher frequency' means more superficial, not more invasive. Third, denser small-cell networks lower individual exposure: phones transmit at power inversely related to link quality, so a nearby antenna makes the device by the user's head — always the dominant exposure source — transmit less. Measurement campaigns around live 5G infrastructure consistently find public exposure at small percentages of ICNIRP limits.