QuantaDose Press Releases

The Tumors That Are Rising—and the Exposure Limits That Were Never Designed to Measure Them

Registry Signals, Biological Timing, and the Scientific Limits of a Heating Standard

An RF Safe scientific analysis and research agenda
By John Coates, Founder, RF Safe
August 2026

 

A cancer registry can identify a population signal. It cannot identify the exposure that caused it. A laboratory study can establish biological plausibility. It cannot, by itself, establish population risk. The public-health task is to connect those layers without confusing them.

Executive summary

Several tumor trends deserve more careful public discussion than they usually receive.

In England, a peer-reviewed analysis of national registry data reported that the age-standardized incidence of glioblastoma increased from approximately 2.4 to 5.0 cases per 100,000 people between 1995 and 2015. Annual diagnoses rose from 983 to 2,531, with much of the increase occurring in frontal and temporal tumors and in adults older than 55. A later 2026 reanalysis by two of the same investigators reported a substantial increase in Grade 3 oligodendroglioma. However, the authors’ 2018 peer-reviewed paper had described little change in anaplastic and Grade 2 oligodendroglioma over the same period. That discrepancy requires methodological clarification before the newer oligodendroglioma estimate can be treated as settled evidence.

In the United States, a May 2026 analysis of National Cancer Institute SEER data reported that recorded nonmalignant meningioma incidence increased markedly from 2004 to 2023, thyroid cancer incidence approximately doubled from 2000 to 2023, salivary-gland cancer increased gradually, and glioblastoma trends differed by age. The overall glioblastoma rate changed little after the mid-2000s, while an increase was reported among people aged 15 to 39 through 2019 and among adults aged 75 and older. These are important descriptive observations, but each requires tumor-specific interpretation. Mandatory collection of nonmalignant brain tumors began in 2004, thyroid cancer is particularly sensitive to imaging and diagnostic intensity, and subgroup analyses can reveal real heterogeneity while also increasing the risk of overinterpreting selected trends.

The human epidemiology is not uniform. Some case-control analyses report higher risks among people with the greatest cumulative mobile-phone call time, including analyses above approximately 1,000 lifetime call hours. In contrast, the prospective COSMOS cohort and a 2024 World Health Organization-commissioned systematic review did not find increased risk of glioma, meningioma, acoustic neuroma, pituitary tumors, salivary-gland tumors, or pediatric brain tumors from mobile-phone exposure as studied. Those reassuring findings must be included in any scientifically serious account. So must their limitations: exposure misclassification, changing technology, the difficulty of studying very long latency, and limited information about early-life, multi-source, and continuously changing exposures.

Experimental evidence creates a different and unresolved question. The U.S. National Toxicology Program reported clear evidence of malignant cardiac schwannomas in male rats and some evidence of malignant gliomas after long-term radiofrequency exposure. A 2025 systematic review of animal cancer bioassays, partially funded by the WHO, rated the evidence as high certainty for glioma and malignant heart schwannoma in male rats while emphasizing that quantitative extrapolation to humans is unusually difficult. It remains uncertain which exposure metric, waveform, target tissue, time course, and dose-response model should govern that extrapolation.

In 2026, Ronald Melnick and Joel Moskowitz applied benchmark-dose methods and low-dose risk extrapolation to the animal findings. They estimated whole-body specific absorption rates associated with a one-in-100,000 excess cancer risk at approximately 0.8 to 5 milliwatts per kilogram when risk was normalized per hour of daily exposure, depending on the study and exposure schedule. They estimated 3.3 to 10 milliwatts per kilogram as a range intended to protect male reproductive endpoints. Those values are below the 80 milliwatts-per-kilogram whole-body public limit used by the FCC and ICNIRP. The comparison is consequential, but it is a model-based risk assessment, not a directly observed human threshold. Its conclusions depend on cross-species extrapolation, endpoint selection, uncertainty factors, daily exposure duration, and the assumption used for low-dose response.

The regulatory point is narrower and stronger than claiming that these calculations have already established the correct new limit:

The FCC limit is a thermal compliance threshold. It was not derived as a quantitative lifetime cancer-risk standard, a developmental-neurobiology standard, or a reproductive-toxicity standard.

RF Safe’s S4–Mito–Spin framework proposes a research program for the biological questions that an energy-only standard does not answer:

  • S4: Could time-varying fields alter the kinetics of voltage-sensing ion channels and thereby change the timing, amplitude, or recovery of calcium signals under non-heating conditions?
  • Mito: Could altered calcium timing be amplified by mitochondria into changes in membrane potential, electron flow, redox signaling, energy reserve, quality control, or stress recovery?
  • Spin: Under which molecular conditions can magnetic fields alter spin-correlated radical-pair reactions involving flavins, hemes, iron-sulfur centers, quinones, or related redox intermediates?
  • Persistence: When does a transient change return to baseline, and when is it retained through impaired repair, altered chromatin, mitochondrial dysfunction, clonal selection, or developmental reprogramming?

RF Safe calls a persistent mismatch between environmental electromagnetic timing and endogenous biological regulation bioelectrical dissonance. We use low-fidelity biology to describe the proposed systems state in which timing, repair, classification, and recovery become less precise. These are not clinical diagnoses and should not be presented as facts already visible in a cancer registry. They are operational hypotheses that must be tested with measurable endpoints.

The central public-health question is therefore not whether radiofrequency exposure has been proven to cause every tumor discussed in this article. It has not. The question is whether a safety framework designed around short-term heating is scientifically adequate for lifelong, localized, pulsed, developmentally timed, and multi-source exposure—and whether the evidence base has been designed to detect the kinds of biological changes now becoming experimentally accessible.

RF Safe’s position is that this question remains open, consequential, and testable. That is sufficient reason for transparent surveillance, independent research, exposure-reduction options, modernized compliance testing, restoration of meaningful public-health oversight, and policies that do not force communities to treat compliance with a thermal limit as the end of the biological inquiry.

One young athlete makes the question human

Archie Goodburn was 22 when neurological symptoms initially interpreted as migraine were ultimately recognized as seizures. Imaging revealed three large oligodendrogliomas. The tumors were described as inoperable, and Goodburn later became an advocate for brain-tumor research while continuing to compete internationally. In 2026, he reached a Commonwealth Games final while living with incurable brain cancer.

His story is medically and morally important. It communicates what a rate per 100,000 cannot: a rare tumor is not rare to the person whose nervous system, career, family, and future are reorganized around it.

It is equally important to state what his story cannot establish. A single case, however striking, cannot identify its cause. It cannot demonstrate that a phone, a base station, a chemical exposure, an inherited susceptibility, a developmental event, or any other specific factor produced the tumor. Oligodendroglioma is a molecularly defined disease whose biology involves characteristic genetic alterations, and most individual cases cannot be assigned to a known external cause.

The scientifically appropriate use of a case is therefore not to convert chronology into causation. It is to motivate better questions:

  • Is the tumor truly increasing after diagnostic and classification changes are accounted for?
  • Is any increase concentrated by grade, molecular subtype, anatomical site, age, sex, birth cohort, or geography?
  • Which exposures changed during the relevant latency interval?
  • Are those exposures biologically plausible at the target tissue?
  • Do analytical epidemiology, dosimetry, and experiments support the same explanation?

A case makes the question human. A registry determines whether there is a population pattern. An exposure study determines whether risk differs between better- and less-exposed people. Mechanistic research determines whether the association is biologically coherent. Causal inference requires all of those layers.

What the England data show—and what remains unresolved

The documented glioblastoma increase

In 2018, Alasdair Philips, Denis Henshaw, and colleagues published an analysis of 81,135 primary malignant brain tumors recorded in England from 1995 through 2015. They reported that the age-standardized incidence of glioblastoma increased from approximately 2.4 to 5.0 cases per 100,000 people. The annual number of recorded glioblastomas rose from 983 to 2,531.

The increase was not distributed uniformly across all tumor categories. The investigators reported particularly large changes in frontal and temporal glioblastoma and noted that most cases occurred among people older than 55. They also observed a shift from overlapping or unspecified locations toward more specific frontal and temporal coding during part of the interval, which they acknowledged could reflect improved imaging.

The magnitude of the reported increase warrants investigation. It does not, however, identify one environmental cause. During the same period, magnetic-resonance imaging became more accessible and precise, neurosurgical sampling changed, pathology evolved, older adults were investigated more aggressively, cancer registration improved, and diagnostic coding moved through several classification systems. Molecular definitions introduced later have further changed what is now called glioblastoma.

These factors do not automatically explain the entire trend. Nor can they be treated as irrelevant. The correct question is how much of the increase remains after harmonizing diagnosis, morphology, molecular classification, site, age, and registry completeness.

Broader registry comparisons also matter. A 2022 analysis of 18,232 gliomas among Nordic men aged 40 to 59 found only a small annual increase from 1979 through 2016 and concluded that the observed trend was not compatible with several large risk increases predicted from earlier case-control studies. That analysis has its own assumptions about exposure prevalence, latency, and comparability, but it demonstrates why a national trend cannot be interpreted in isolation.

The England signal should therefore be described as real within the data analyzed and unresolved in its explanation—not as proof of a wireless cause and not as something that can be dismissed by quoting one lumped brain-cancer rate from another country.

The oligodendroglioma discrepancy

A 2026 article from the EM Radiation Research Trust, written by Philips and Henshaw, reported that Grade 3 oligodendroglioma in England rose approximately four-fold from 1995 to 2015, with the increase beginning in the late 1990s. The same article stated that lower-grade oligodendroglioma roughly doubled.

That claim should not be repeated without an important qualification. In their 2018 peer-reviewed analysis of the same broad registry period, the authors wrote that there was little change in anaplastic oligodendroglioma and Grade 2 oligodendroglioma. The newer article appears to be a reanalysis, but the public presentation does not yet supply enough methodological detail to reconcile the two conclusions.

Several possibilities could explain the difference:

  • The later analysis may use a different denominator or age standard.
  • It may include a different combination of histology and behavior codes.
  • It may separate grades differently or correct earlier coding assumptions.
  • It may use annual counts rather than age-standardized rates.
  • It may be affected by changes in pathology, tumor grading, or registration.
  • A rare tumor can show a large relative change from a small baseline even when the absolute increase remains modest.

The appropriate conclusion is not that the newer result is false. It is that the four-fold estimate should be treated as a signal awaiting a fully specified, reproducible analysis. A formal publication should provide the exact ICD-O codes, grade rules, denominators, age standard, confidence intervals, joinpoint model, sensitivity analyses, and treatment of molecular reclassification.

That is how RF Safe should handle evidence that supports its concern: not by lowering the standard of proof, but by demanding the analysis required to know whether the signal is robust.

How to read a cancer trend without being misled

Cancer incidence seems simple: count new cases and divide by the population. In practice, a trend is produced by biology, clinical practice, classification, surveillance, and demography at the same time.

Counts, crude rates, and age-standardized rates answer different questions

Annual case counts can rise because a population grows or ages even when an individual’s age-specific risk is unchanged. Crude rates partly adjust for population size but remain sensitive to age structure. Age-standardized rates reweight age-specific rates to a reference population, making comparisons across time and place more meaningful.

For tumors strongly concentrated in older adults, this distinction is essential. A rising count is a burden on families and health systems, but it is not automatically evidence of a rising age-adjusted risk.

Better detection can create a real increase in recorded incidence

If more people receive MRI, ultrasound, or cross-sectional imaging, more tumors will be found. This is especially important for slow-growing or asymptomatic tumors. A tumor that would once have remained undiagnosed can become a registry case after imaging for an unrelated complaint.

That does not make the diagnosis imaginary. It changes the relationship between recorded incidence and underlying disease occurrence.

Classification can move cases between categories

Brain tumors have been repeatedly redefined. Histology once dominated diagnosis. Modern classification increasingly integrates mutations, chromosomal changes, methylation, and other molecular features. A tumor called glioblastoma under an older system may be assigned to a different molecular category today. Oligodendroglioma now requires an IDH mutation and combined deletion of chromosome arms 1p and 19q.

Long time-series analyses must account for this migration between categories. Otherwise, an apparent rise in one subtype can partly reflect a decline in another.

A joinpoint is a statistical observation, not an historical cause

Joinpoint regression identifies years when the slope of a trend changes. A change near the introduction of a technology can generate a hypothesis, but temporal alignment alone cannot establish causation. Many exposures, behaviors, diagnostic practices, and population characteristics change together.

The inference becomes stronger when a predicted latency pattern, dose-response relationship, anatomical distribution, age or birth-cohort pattern, and mechanistic pathway all agree. Without those elements, a calendar-year coincidence remains ecological evidence.

Subgroup trends can reveal biology—and can multiply false leads

An overall rate can conceal opposing trends by age, sex, tumor site, or molecular subtype. Stratification is therefore necessary. But examining many subgroups and highlighting only those that rise creates a multiple-comparison problem.

A credible analysis should pre-specify hypotheses where possible, report all strata, provide confidence intervals, test interactions, and replicate findings in independent registries.

These principles do not weaken cancer surveillance. They make it capable of distinguishing a real etiological signal from a changing measurement system.

What the U.S. SEER data say

The National Cancer Institute’s Surveillance, Epidemiology, and End Results program is one of the most important cancer-surveillance systems in the world. SEER 21 covers a large and diverse portion of the U.S. population and provides age-adjusted incidence by tumor type, age, sex, race, stage, and calendar year.

Joel Moskowitz of the University of California, Berkeley has used SEER 21 through 2023 to examine tumor categories relevant to localized exposure of the head and neck. The resulting trends deserve attention, but they should be presented with the relevant surveillance caveats.

Nonmalignant meningioma: a large recorded increase with an ascertainment problem at the baseline

The SEER-based analysis reported an age-adjusted nonmalignant meningioma incidence of 6.59 per 100,000 in 2004 and 12.18 in 2023. It also reported continued increases among people aged 15 to 39 and 40 to 64 during later segments of the series.

Meningioma arises from the meninges, the membranes surrounding the brain and spinal cord. Most meningiomas are histologically nonmalignant, but that classification does not mean clinically inconsequential. Depending on location and growth, a meningioma can compress brain tissue, disturb vision or hearing, provoke seizures, produce neurological deficits, require surgery or radiation, and recur.

It is therefore inappropriate to exclude nonmalignant tumors when discussing the total neurological burden of primary brain and central-nervous-system tumors.

It is also inappropriate to describe the 2004-to-2023 change as a clean biological doubling without qualification. U.S. registries began mandatory collection of nonmalignant brain and CNS tumors for diagnoses beginning January 1, 2004. Early years were affected by implementation, case-finding, and completeness. Increased use of MRI and incidental detection also raises recorded meningioma incidence.

Peer-reviewed analyses have found that the steepest rise occurred in the first years after mandatory reporting and that the rate of increase later slowed or stabilized in some age groups. The more informative question is not whether the 2004 and 2023 endpoints differ. It is whether well-ascertained, age-specific and site-specific rates continue to increase after the reporting system matured—and, if so, which risk factors explain the residual trend.

Thyroid cancer: approximately twice the 2000 rate, but not a simple exposure marker

The SEER-based analysis reported thyroid cancer increasing from 7.65 per 100,000 in 2000 to 15.35 in 2023. It also identified a new increase from 2021 through 2023, concentrated primarily among adults aged 40 to 64.

The thyroid is located superficially in the lower anterior neck, and its position makes localized exposure a legitimate dosimetric question for some phone-use configurations. That anatomical fact does not make wireless exposure the default explanation for the incidence trend.

Thyroid cancer is one of the clearest examples of diagnostic intensity changing recorded incidence. Ultrasound, CT, MRI, carotid imaging, and fine-needle biopsy detect small papillary cancers that may never have produced symptoms. Changes in clinical thresholds and efforts to reduce overdiagnosis have altered the trend in recent years.

At the same time, overdiagnosis is not necessarily the complete explanation. NCI-supported research has reported increases in some advanced-stage papillary cancers and mortality during earlier decades. A serious etiological analysis must separate small localized tumors from larger, regional, distant, aggressive, and fatal disease.

The next scientific step is therefore not to place the thyroid trend beside a drawing of a phone and declare causation. It is to analyze stage, tumor size, histology, age, sex, calendar period, diagnostic intensity, side or lobe where available, and individualized exposure.

Salivary-gland cancer: a gradual rise in an anatomically relevant site

The same SEER analysis reported salivary-gland cancer increasing by approximately 0.73 percent per year from 2000 through 2023. The parotid glands lie anterior and inferior to the ears and can receive localized radiofrequency energy during handset use.

This makes salivary tumors important in mobile-phone epidemiology, and they were included in the 2024 WHO-commissioned systematic review. That review concluded with moderate certainty that mobile-phone exposure as studied likely does not increase adult salivary-gland tumor risk.

The registry trend and the epidemiological conclusion are not logically incompatible. Incidence can rise for reasons unrelated to phones. A cohort or case-control analysis can also miss a small effect because of exposure error, insufficient latency, rare outcomes, or changing technology. The scientific response is better exposure reconstruction and tumor-specific research, not selective acceptance of whichever result supports a preferred conclusion.

Glioblastoma: overall stability can coexist with divergent age trends

The SEER-based analysis reported an overall glioblastoma rate of 3.02 per 100,000 in 2000 and 3.15 in 2023. After an early increase, the overall trend from approximately 2004 through 2023 was essentially flat.

That overall finding is important and should not be hidden.

The analysis also reported a 1.12 percent annual increase among people aged 15 to 39 from 2000 through 2019 and a 0.60 percent annual increase among adults aged 75 and older, while rates in some middle-aged groups declined slightly. These subgroup findings may reflect differences in tumor biology, classification, diagnostic intensity, cohort exposures, or random and modeling variation.

A young-adult increase in a highly lethal tumor deserves replication and investigation. It does not justify saying that U.S. glioblastoma is broadly surging. The scientifically precise statement is that the aggregate rate has been comparatively stable while selected age strata show different trends.

Children and young adults: burden is not the same as an increasing trend

Brain and other CNS cancers are the leading cause of childhood cancer death in the United States. Among adolescents and young adults, primary brain and CNS tumors are a major source of morbidity and the second leading cause of cancer-related death; among those aged 15 to 24, they are the leading cause.

Those facts establish urgency. They do not establish that the burden is caused by wireless exposure or that every pediatric brain-tumor category is increasing. Incidence, mortality, survival, and relative rank among causes of cancer death are different measures.

Children nevertheless require special attention in exposure science because development changes anatomy, tissue conductivity, calcium signaling, synaptic organization, endocrine regulation, sleep, repair, and the number of remaining years in which a long-latency outcome can emerge. A standard claimed to protect the public should explicitly evaluate these life-stage differences rather than assume that an adult thermal model fully represents them.

Anatomical proximity is a hypothesis, not dosimetric proof

During a voice call, a handset can create localized radiofrequency exposure in tissues near the device. Depending on position, frequency, antenna configuration, network conditions, adaptive power control, use of speaker mode or a headset, and tissue properties, comparatively exposed structures can include superficial regions of the brain, meninges, vestibular nerve, parotid and other salivary tissues, skin, eye, and portions of the neck.

This is why glioma, meningioma, acoustic neuroma, and salivary-gland tumors have been repeatedly examined in phone epidemiology.

But “near the phone” is not an exposure measurement. Individual absorbed dose varies substantially. A modern phone changes output power. Different frequencies deposit energy differently. The highest local absorption need not coincide with the site at which a tumor is later detected. People change devices, networks, sides of use, and behavior over decades. Texting and data use move the device away from the classic call position, while earbuds, wearables, and ambient infrastructure change the exposure mixture.

The strongest studies should therefore integrate:

  • operator or device records rather than memory alone;
  • model-specific power and frequency information;
  • laterality and habitual device position;
  • near-field computational dosimetry;
  • use of speaker mode, wired headsets, Bluetooth, and cordless phones;
  • cumulative call time and latency;
  • childhood and prenatal exposure;
  • occupational and residential sources;
  • tumor location at high anatomical resolution;
  • and molecular tumor subtype.

Anatomical plausibility tells us where to look. It does not tell us what we will find.

The human evidence is contested—and must be presented whole

The high-use case-control signal

A 2020 updated meta-analysis led by Moon and Choi reviewed 46 case-control studies. The pooled analysis did not find an increased risk for regular mobile-phone use under all exposure definitions. However, a subgroup with more than approximately 1,000 cumulative hours of call time showed a statistically significant increase in tumor risk. Public summaries described the magnitude as roughly 60 percent.

This finding matters because it proposes an exposure-response threshold in cumulative use. Approximately 1,000 hours can be reached by about 17 minutes of calling per day over ten years.

Case-control studies are also vulnerable to systematic error. People with a brain tumor may remember past use differently from controls. Participation can differ between cases and controls. The side of preferred phone use can be recalled after the person already knows the tumor location. Rapid changes in technology complicate conversion of call time into absorbed dose. These biases can either create, obscure, or distort an association.

The high-use result is therefore evidence, not a final estimate of causal risk.

The prospective COSMOS result

COSMOS follows more than 250,000 mobile-phone users. Participants reported detailed phone use and were linked to cancer registries. In results published in 2024 after a median follow-up exceeding seven years, the ten percent with the greatest lifetime call time did not have higher incidence of glioma, meningioma, or acoustic neuroma than lighter users.

Prospective design reduces the recall bias that affects case-control studies. The large cohort and registry linkage are major strengths. Limitations remain. Median follow-up after enrollment is modest for tumors that may take decades to become clinically detectable, even though many participants had used phones for years before entry. Self-reported use and changing behavior create exposure misclassification. Very heavy early-life use, current multi-source exposure, and rare molecular subtypes remain difficult to evaluate.

COSMOS is not evidence that should be ignored because it is reassuring. It is one of the strongest human datasets available and weighs against a large near-term increase in the tumors studied under the exposure conditions captured.

The WHO-commissioned human systematic review

The 2024 systematic review led by Karipidis evaluated 63 human observational reports selected from more than 5,000 records. It concluded that there was moderate-certainty evidence that mobile-phone exposure likely does not increase the risk of adult glioma, meningioma, acoustic neuroma, pituitary tumors, salivary-gland tumors, or pediatric brain tumors.

Critics have challenged study selection, exposure classification, treatment of latency, weighting of cohort and case-control designs, and potential conflicts of interest. Those critiques deserve evaluation on their methods, not on the identity of the conclusion. Conversely, it is not scientifically defensible to describe the literature as though the review does not exist.

The most accurate synthesis is that the present human evidence does not establish a population-wide increase in brain or head tumors from mobile-phone use, while important uncertainty remains for very long latency, early-life exposure, changing waveforms, highly exposed subgroups, and rare or molecularly defined outcomes.

IARC’s 2011 classification of radiofrequency electromagnetic fields as Group 2B, possibly carcinogenic to humans, remains part of this history. Group 2B identifies limited evidence and uncertainty; it is neither a finding of no hazard nor proof of causation.

The animal evidence changed the scientific landscape

The National Toxicology Program

The U.S. National Toxicology Program conducted large, long-term studies in rats and mice exposed to 900-megahertz and 1,900-megahertz radiofrequency fields, respectively, modulated to resemble GSM or CDMA signals. The exposures were whole-body, began before birth for rats, and continued for up to two years.

NTP concluded that there was clear evidence of malignant schwannoma of the heart in exposed male rats. It reported some evidence of malignant glioma in the brain of male rats. Other findings differed by sex, species, modulation, and endpoint.

The studies do not reproduce ordinary human phone use. Rodents received whole-body exposure for long daily periods, and the highest exposures exceeded typical environmental levels. Thermal management, survival differences, dosimetry, multiple comparisons, and the relevance of rare rodent tumors have all been debated.

Yet the results cannot be dismissed on the ground that non-ionizing radiation is incapable of carcinogenic activity unless it directly breaks chemical bonds like an X-ray. A carcinogen can act indirectly through signaling, oxidative chemistry, endocrine effects, immune regulation, repair, or promotion. The animal findings establish a hazard signal requiring mechanistic and translational explanation.

The 2025 animal systematic review

A 2025 systematic review led by Mevissen examined 52 animal studies, including 20 chronic bioassays. Using a GRADE/OHAT framework, the authors rated evidence as high certainty for increased glioma and malignant heart schwannoma in male rats. They assigned moderate certainty to several other tumor outcomes. A corrigendum was published in 2026, so readers should consult the corrected record.

The review also stated a limitation that should guide policy interpretation: extrapolating RF cancer bioassays to humans is particularly complex. The appropriate exposure metric may be localized or whole-body absorption, field strength, peak intensity, cumulative exposure, modulation, or another variable. A monotonic dose-response may not apply to every mechanism, and specific absorption rate may not be the only biologically relevant metric.

High certainty that an animal association exists is not the same as high certainty about the magnitude of human risk. It is, however, a stronger basis for precaution and research than “no animal evidence.”

What the 2026 risk assessment does—and does not—establish

Ronald Melnick, who helped design the NTP program, and Joel Moskowitz published a 2026 risk assessment using the NTP and Ramazzini Institute animal data.

For cancer, the authors used benchmark-dose modeling and linear low-dose extrapolation to estimate an exposure corresponding to an additional lifetime cancer risk of one in 100,000. They reported SAR estimates of approximately 0.8 to 5 milliwatts per kilogram when risk was normalized per hour of daily exposure. Depending on assumed daily exposure duration, they calculated that the current 80-milliwatt-per-kilogram whole-body public limit is approximately 15 to 900 times higher than their cancer-risk estimates.

For male reproductive toxicity, they applied uncertainty factors to animal endpoints and reported protective estimates of approximately 3.3 to 10 milliwatts per kilogram, eight to 24 times below the current whole-body limit.

These comparisons expose a genuine policy question: RF limits were not derived using the quantitative risk-assessment procedures commonly applied to carcinogens and reproductive toxicants.

They do not establish a universally accepted cancer threshold. The resulting values depend on several contestable choices:

  • whether the animal tumors are causally attributable to RF exposure;
  • which study and tumor endpoint should anchor the model;
  • whether whole-body animal SAR maps appropriately to localized and whole-body human exposure;
  • whether low-dose risk is linear;
  • how daily duration and lifetime exposure should be represented;
  • which uncertainty factors are appropriate;
  • and whether SAR adequately captures waveform-dependent effects.

The correct description is therefore:

Melnick and Moskowitz provide a health-risk-based alternative to the existing thermal framework. Their estimates are scientifically consequential and should be independently reproduced, stress-tested under alternative models, and compared with human and mechanistic data.

That is more defensible than saying the study has already proved that every current limit is exactly 900 times too high.

What the FCC limit actually measures

The FCC adopted its present radiofrequency exposure limits in 1996, drawing on standards developed by ANSI/IEEE and the National Council on Radiation Protection and Measurements. For the general public, the whole-body specific absorption rate limit is 0.08 watts per kilogram. Localized limits are higher, including 1.6 watts per kilogram averaged over one gram of tissue for many portable-device evaluations.

The historical evidence base centered on acute behavioral disruption and heating in animals. A whole-body SAR near 4 watts per kilogram was treated as a threshold for adverse thermal effects; uncertainty factors were then applied to derive lower occupational and public limits.

This history supports a precise criticism:

  • The limit is designed to prevent excessive energy absorption and established thermal injury.
  • It is not a dose-response standard derived from lifetime human cancer incidence.
  • It is not based on developmental calcium signaling, mitochondrial recovery, redox timing, radical-pair chemistry, fertility, sleep architecture, or long-term neurological endpoints.
  • Compliance therefore demonstrates compliance with the specified exposure metric and averaging rules. It does not prove the absence of every biological effect.

That does not make thermal protection obsolete. Heating is real and must be controlled. The problem is treating one necessary protection as a complete biological safety assessment.

In 2021, the U.S. Court of Appeals for the D.C. Circuit remanded the FCC’s 2019 decision to retain the 1996 limits. The court held that the agency had not provided a reasoned explanation for its conclusion regarding harmful effects unrelated to cancer. It specifically addressed long-term exposure, children, pulsation or modulation, and changes in wireless technology. The court did not rule that RF exposure causes cancer or other disease. It ruled that the agency’s reasoning was inadequate on important non-cancer questions.

That distinction is legally and scientifically important. A remand is not a toxicological verdict. It is a requirement for reasoned agency analysis.

Why time structure matters even when average energy is unchanged

Specific absorption rate describes the rate at which RF energy is absorbed per unit mass. It is indispensable for thermal dosimetry. It does not fully describe a time-varying signal.

Two exposures can have the same time-averaged SAR and differ in:

  • carrier frequency;
  • pulse duration;
  • repetition rate;
  • duty cycle;
  • peak-to-average ratio;
  • modulation;
  • polarization;
  • intermittency;
  • exposure phase relative to sleep or development;
  • and the interval available for biological recovery.

Whether those differences matter depends on the receiver. A molecule or cellular circuit that integrates only total energy will respond primarily to average absorption. A nonlinear oscillator, voltage-sensitive gate, spin-correlated reaction, adaptation circuit, or calcium-dependent transcriptional system may respond differently to timing.

This is not proof that ordinary wireless pulses cause cancer. It is a reason why equal-average-power comparisons are experimentally necessary.

S4–Mito–Spin: a formal mechanistic hypothesis

RF Safe’s S4–Mito–Spin framework is intended to organize testable pathways, not to replace experimental evidence with a slogan. Each branch begins with established biology and then identifies an electromagnetic question that remains to be resolved.

S4: voltage sensing and calcium-code fidelity

Voltage-gated sodium, calcium, and potassium channels contain specialized voltage-sensing domains. Their S4 transmembrane segments carry regularly spaced positive charges. Movement of these gating charges in the membrane electric field helps couple voltage changes to channel opening and closing.

This is established ion-channel biophysics.

The RF Safe hypothesis is that some time-varying external fields, under specific geometries and tissue conditions, may change channel-gating probabilities or kinetics without producing substantial heating. The most informative endpoint would not be a nonspecific increase in total calcium. It would be a change in the calcium code:

  • pulse frequency;
  • amplitude;
  • rise and decay time;
  • spatial localization;
  • phase relation to other oscillators;
  • baseline recovery;
  • and cell-to-cell synchronization.

Calcium controls secretion, contraction, metabolism, transcription, proliferation, differentiation, migration, apoptosis, and mitochondrial activity. A small change can be amplified if it occurs at the right phase of a sensitive circuit. The relevant prediction is therefore state dependent: the same field may produce different responses in cells with different channel expression, membrane potential, differentiation state, or prior activity.

This proposed route must be tested with patch clamp, optical voltage reporters, compartment-specific calcium imaging, matched-temperature controls, pharmacology, channel knockouts, S4 mutations, and waveform comparisons at equal average power.

It should not be presented as established that Wi-Fi, DECT, GSM, or 5G envelopes add timing noise to S4 sensors. That is the hypothesis.

Mito: amplification through calcium, membrane potential, and redox state

Mitochondria are electrically active organelles. Their inner membrane maintains a large electrochemical potential generated by respiratory electron transport. That potential powers ATP synthesis, supports protein import and metabolite exchange, and drives calcium uptake through the mitochondrial calcium uniporter.

Calcium can stimulate mitochondrial metabolism when delivered in appropriately timed and localized signals. Excessive or persistent calcium can promote oxidative stress, membrane-potential collapse, permeability transition, fragmentation, and cell death. Mitochondria also shape the calcium signal by buffering and returning ions, while producing redox signals that regulate channels, transcription, and stress responses.

This creates a bidirectional loop:

calcium timing influences mitochondrial state, and mitochondrial state influences calcium timing.

The 2026 Cell study by Kim and colleagues adds a remarkable experimental clue. The researchers engineered an electromagnetic-field-inducible gene switch and used a CRISPR screen to identify cytochrome b5 type B, or CYB5B, as an essential mediator likely acting as a sensor. Activation depended on rhythmic calcium oscillations rather than generic calcium influx. The reported stimulation used a defined 60-hertz, 2-millitesla magnetic field in an engineered system; an erratum later corrected a supplementary-image issue.

The study establishes that a defined electromagnetic input can be coupled through specific cellular machinery to a patterned calcium response and transcriptional output. It does not establish that ordinary environmental RF uses CYB5B, that the CYB5B heme operates through a radical pair, or that the response produces cancer. Those are separate questions.

The RF Safe Mito branch asks whether electromagnetic perturbations that alter calcium or redox timing can change:

  • mitochondrial membrane-potential stability;
  • respiratory reserve;
  • ATP-to-demand matching;
  • superoxide and hydrogen-peroxide dynamics;
  • fusion, fission, and mitophagy;
  • DNA-repair support;
  • apoptosis thresholds;
  • and recovery after the field ends.

This branch provides an experimentally tractable bridge from an acute signal to persistence. If every change returns rapidly and completely to baseline, long-term risk becomes less plausible. If repeated exposures create cumulative recovery delay, altered quality control, or stable transcriptional changes, the persistence hypothesis gains support.

Spin: magnetic control of reaction probability

Some chemical reactions generate pairs of radicals whose unpaired electron spins are correlated. The pair can interconvert between singlet and triplet configurations. Magnetic fields can alter this spin evolution through Zeeman and hyperfine interactions, thereby changing reaction lifetimes or product yields under suitable molecular conditions.

Radical-pair chemistry is not speculative in the abstract. It is a well-developed field of spin chemistry and is central to leading models of biological magnetoreception.

Two 2026 experiments extended its relevance. Burd and colleagues demonstrated radiofrequency magnetic-resonance control of spin-correlated radical-pair dynamics in a live transgenic animal. Meng and colleagues showed radio-wave control of photogenerated spin-correlated radical pairs in flavoproteins, including cryptochrome and engineered LOV proteins.

These experiments establish that radiofrequency fields can control selected spin-correlated biochemical reactions in living or biologically compatible systems. They do not show that ambient telecommunications exposure perturbs human tumor suppression. The engineered proteins, optical excitation, resonance conditions, field strengths, frequencies, and readouts must all be considered.

The RF Safe Spin branch asks whether naturally occurring mitochondrial or cellular radical intermediates involving flavins, hemes, iron-sulfur clusters, quinones, oxygen, or associated partners have the lifetimes and coupling needed for field sensitivity under realistic exposure conditions.

Specific predictions include changes in:

  • radical-pair product ratios;
  • redox reaction lifetime;
  • superoxide versus hydrogen-peroxide production;
  • electron-transfer efficiency;
  • oxygen consumption;
  • and downstream calcium or transcriptional timing.

The required experiments include electron-paramagnetic-resonance measurements, magnetic-isotope substitution, static-field orientation controls, frequency sweeps around predicted resonances, oxygen and light dependence, engineered loss and rescue of the candidate redox center, and direct temporal ordering from spin chemistry to redox to calcium to phenotype.

Persistence: the difference between a perturbation and a disease-relevant process

Cells experience countless transient changes without becoming diseased. A mechanism relevant to cancer must explain persistence.

Persistence can arise through several routes:

  • repeated exposure faster than complete recovery;
  • stable chromatin or transcriptional remodeling;
  • impaired mitochondrial quality control;
  • accumulation of mitochondrial DNA defects;
  • altered stem-cell state;
  • chronic inflammation or tissue remodeling;
  • failure of immune surveillance;
  • selection and expansion of a pre-existing mutant clone;
  • or disruption during a developmental window when a transient signal becomes a lasting structural decision.

Long-lived cells and low-turnover tissues may retain consequences longer, but longevity alone is not a persistence mechanism. Neurons, glia, Schwann-lineage cells, endocrine cells, germ cells, and stem-cell compartments differ in channel expression, mitochondrial density, redox metabolism, repair, turnover, and tissue context. Claims that a particular tumor is predicted by a simple “high S4 times high mitochondria” score remain hypotheses until those variables are quantitatively mapped.

The persistence gate is what converts a mechanistic possibility into a research program relevant to cancer.

Biological fidelity, bioelectrical dissonance, and recovery debt

RF Safe uses biological fidelity to describe the precision with which cells encode, transmit, interpret, repair, and terminate regulatory signals.

High fidelity does not mean biological stillness. Healthy systems fluctuate. They respond to stress, adapt, and sometimes enter noisy states. Fidelity refers to whether those fluctuations remain appropriately timed, spatially constrained, and recoverable.

Measurable indicators could include:

  • variance in calcium-pulse interval and amplitude;
  • delayed return of membrane voltage to baseline;
  • loss of phase synchronization among cells;
  • mitochondrial membrane-potential instability;
  • reduced respiratory reserve;
  • altered redox pulse timing;
  • increased DNA-repair latency or error;
  • incomplete restoration of chromatin state;
  • abnormal fusion, fission, or mitophagy;
  • and reduced discrimination between repair, senescence, apoptosis, and proliferation.

Bioelectrical dissonance is RF Safe’s term for a persistent mismatch between externally imposed electromagnetic timing and endogenous bioelectrical regulation.

Recovery debt describes the cumulative state in which a new perturbation arrives before the preceding response has fully resolved. It can be quantified experimentally as progressively slower or incomplete return to baseline across repeated exposure cycles.

Low-fidelity biology is the proposed systems outcome: biological processes continue, but with less precise timing, classification, error correction, and recovery.

These concepts are not established medical diagnoses. A cancer registry does not measure calcium jitter, mitochondrial recovery, or radical-pair yield. It would therefore be inaccurate to say that rising tumor lines are “exactly what low-fidelity biology looks like.” The registry lines are population observations that the hypothesis might help explain only if the intervening steps are demonstrated.

The framework becomes scientifically useful when it makes discriminating predictions:

  • Equal-average-power waveforms should produce different effects if timing matters.
  • Effects should depend on receptor abundance, cellular state, and exposure phase.
  • Calcium, voltage, redox, or spin changes should precede transcriptional and phenotypic changes.
  • Removing the proposed receiver should abolish the effect, and restoring it should rescue the effect.
  • Adequate recovery intervals should reduce cumulative changes if recovery debt is central.
  • A persistent phenotype should correlate with failure to return to baseline, not merely with one acute molecular fluctuation.
  • Tissue susceptibility should be predictable from measured receiver, metabolic, repair, and persistence variables—not assigned retrospectively after a tumor appears.

This is the standard required to move from an integrative hypothesis to a causal model.

What a decisive research program would measure

The cancer question cannot be resolved by repeating short, underpowered studies with incompletely characterized exposure. Nor can it be resolved by comparing national phone-subscription curves with national tumor curves.

A modern program should integrate six levels.

1. Exposure physics

Every experiment should report carrier frequency, modulation, pulse structure, repetition rate, duty cycle, polarization, peak and average fields, near- or far-field geometry, harmonics, temperature, induced current, SAR where applicable, static magnetic background, and sham performance.

Waveforms should be compared at equal average absorbed power. Continuous-wave controls should be included when pulsed or modulated fields are tested.

2. Immediate transduction

Measurements should include membrane voltage, voltage-gated channel kinetics, cytosolic and organelle-specific calcium, CYB5B redox state, flavin and heme chemistry, radical intermediates, and electron-transfer dynamics.

The sequence of events should be resolved at millisecond-to-minute timescales.

3. Mitochondrial amplification and recovery

Studies should measure mitochondrial membrane potential, ATP, oxygen consumption, respiratory reserve, NADH and FAD redox state, compartment-specific reactive species, permeability transition, fusion, fission, mitophagy, and time to recovery.

Repeated-exposure experiments should test whether recovery slows or remains complete.

4. Genome maintenance and cell fate

Measurements should include oxidative DNA lesions, double-strand breaks, replication stress, repair kinetics, chromosome instability, mutational signatures, epigenetic persistence, apoptosis, senescence, immune signaling, stem-cell state, and clonal expansion.

The key outcome is not a single stress marker. It is whether the exposure produces a persistent, reproducible change relevant to tumor initiation or promotion.

5. Tissue and organism context

Experiments should compare development, adulthood, and aging; males and females; genetically susceptible and typical backgrounds; exposed and shielded tissues; and conditions of normal versus reduced recovery.

Target tissues should be chosen prospectively from measured receiver and metabolic characteristics. Blinding, randomization, adequate sample size, preregistration, and independent replication are essential.

6. Exposure-informed epidemiology

Prospective cohorts should integrate operator data, device telemetry that preserves privacy, network and device models, occupational sources, residential measurements, laterality, developmental exposure, sleep-time exposure, and tumor molecular profiling.

Registry analyses should use consistent molecular categories and should publish all pre-specified strata, not only rising ones.

Only this integrated program can determine whether the mechanistic signals are too small or transient to affect human disease, or whether current epidemiology has been measuring the wrong exposure variables and disease groupings.

A formal RF Safe policy position

Scientific uncertainty does not require political passivity. It requires policies proportionate to the uncertainty, potential severity, feasibility of exposure reduction, and distribution of risk.

Modernize the exposure standard

The FCC should undertake a transparent reassessment that distinguishes thermal compliance from long-term health-risk assessment. It should evaluate modulation, intermittency, cumulative exposure, children, pregnancy, sleep, multiple simultaneous sources, real device positions, and the possibility of nonthermal interaction mechanisms.

The assessment should include scientists with expertise in toxicology, epidemiology, oncology, developmental biology, electrophysiology, mitochondrial biology, spin chemistry, dosimetry, and risk assessment. Conflicts of interest should be fully disclosed, and minority scientific opinions should be documented rather than erased through consensus wording.

Restore a strong public-health research function

Public Law 90-602, the Radiation Control for Health and Safety Act of 1968, established federal authority to protect the public from hazardous and unnecessary electronic-product radiation. Its provisions now reside in sections 531 through 542 of the Federal Food, Drug, and Cosmetic Act.

RF Safe calls for reinvigorated implementation: sustained federal research, product surveillance, performance standards where evidence supports them, transparent reporting, and interagency responsibility that does not leave health evaluation subordinate to spectrum and communications policy.

Returning a central RF health-research and risk-assessment role to the Environmental Protection Agency is a policy proposal, not a description of current law. Its purpose would be to place environmental exposure assessment within an agency experienced in chronic risk, susceptible populations, uncertainty factors, and cumulative exposure.

Reconsider Section 704

Section 704 of the Telecommunications Act, codified at 47 U.S.C. 332(c)(7)(B)(iv), prevents state and local governments from regulating the placement, construction, or modification of personal wireless facilities based on the environmental effects of RF emissions when the facilities comply with FCC rules.

RF Safe supports repeal or substantial reform of this provision. At minimum, federal law should not transform compliance with a thermal limit into a prohibition on local consideration of health evidence, siting alternatives, setbacks, schools, bedrooms, cumulative exposure, or rapidly evolving science.

Any reform must also preserve reliable communication, emergency access, nondiscrimination, and workable infrastructure planning. The goal is not arbitrary local obstruction. It is democratic participation and health protection under a standard capable of evolving with evidence.

Build a Clean Ether Act

RF Safe’s proposed Clean Ether Act should be developed as a comprehensive policy framework rather than a slogan. Its components should include:

  • health-based and periodically reviewed exposure standards;
  • independent premarket and postmarket testing;
  • waveform and peak-exposure disclosure;
  • child- and pregnancy-specific evaluation;
  • practical wired and low-exposure alternatives in schools, healthcare, workplaces, and housing;
  • consumer right-to-know information that communicates both compliance and uncertainty;
  • incentives for lower-power, distance-aware, and light-based communication technologies where suitable;
  • national exposure mapping and cancer surveillance;
  • protected funding for replication and long-latency research;
  • and a clear process for updating limits when evidence changes.

The end point is not electromagnetic silence. It is an electromagnetic environment designed with biological compatibility as an engineering objective.

Practical exposure reduction without panic

Individuals should not be made solely responsible for a society-wide exposure question. Still, low-cost choices can reduce localized exposure while the science develops.

  • Use speaker mode or a wired headset for longer calls.
  • Keep an active phone away from the body when practical rather than pressed against the head or carried continuously against the skin.
  • Prefer texting or brief calls when that meets the need.
  • Avoid sleeping with an active phone under a pillow or directly beside the head.
  • Place routers and continuously transmitting cordless-phone bases away from beds, nurseries, and locations occupied for long periods.
  • Use wired Ethernet where it is convenient, particularly for stationary work and entertainment.
  • Remember that poor signal can make a phone increase its transmit power; distance remains useful.
  • Do not compromise emergency communication or accessibility. Exposure reduction should support daily life, not produce isolation or fear.

These measures do not prove that an exposure is dangerous. They apply the basic physics that field intensity and absorbed energy generally decrease with distance and reduced transmission time.

What RF Safe is ultimately trying to communicate

The strongest case for reform does not depend on pretending that every tumor trend has one established cause.

It rests on five propositions.

First, cancer surveillance should examine specific tumors, ages, anatomical sites, molecular subtypes, and birth cohorts. A lumped category can conceal important changes.

Second, the present human epidemiology is mixed. Positive high-use case-control results coexist with reassuring prospective and systematic-review findings. Uncertainty remains concentrated in long latency, changing exposure, early-life use, and rare outcomes.

Third, long-term animal studies have produced cancer findings that no scientifically responsible safety review can ignore. Translating those findings to humans is difficult, but difficulty is a research obligation—not a justification for permanent inaction.

Fourth, the existing U.S. limit was designed around prevention of acute thermal effects. It was not derived to quantify lifetime cancer, developmental, reproductive, or biological-timing risk.

Fifth, modern biology provides testable mechanisms by which fields could matter without acting like ionizing radiation. Voltage sensors, patterned calcium, mitochondrial amplification, redox chemistry, and spin-correlated reactions are real biological systems. Whether ordinary environmental exposures perturb them enough, long enough, and in the correct tissues to contribute to cancer remains the decisive question.

RF Safe calls the loss of control-system precision low-fidelity biology. The phrase is valuable only if it increases scientific precision. It must never become a way to label any disease trend as proof of the framework.

The disciplined version of the claim is this:

If time-structured electromagnetic exposure can reproducibly alter voltage, calcium, redox, or spin-dependent chemistry under non-heating conditions; if those changes outrun recovery; if they impair genome maintenance, tissue regulation, or tumor suppression; and if exposure-informed epidemiology finds the predicted patterns, then biological fidelity becomes a causal bridge between environmental fields and disease risk.

Every “if” in that sentence can be tested.

That is what makes the present moment scientifically exciting. We no longer have to choose between an implausibly simple claim that “RF causes tumor X” and an equally simplistic claim that “nothing below heating can matter.” We can identify receivers, measure timing, follow amplification, quantify recovery, manipulate the pathway, and connect the result to tissue-specific surveillance.

The public deserves that research before another generation completes a lifetime of exposure under a standard that never asked the question.

Conclusion

Archie Goodburn’s diagnosis is a human tragedy and a reason to invest in treatment and research. It is not proof of an exposure cause.

The England glioblastoma trend is substantial in the published analysis and requires explanation. The newer oligodendroglioma claim requires methodological reconciliation with the authors’ earlier paper. U.S. meningioma, thyroid, salivary-gland, and age-specific glioblastoma trends contain genuine surveillance signals, but they are shaped by reporting, imaging, classification, age, and subgroup structure.

The human radiofrequency epidemiology remains contested. The animal evidence is more concerning than it was when the FCC adopted its limits. The 2026 risk assessment shows how different the result can be when investigators begin with a lifetime cancer-risk objective instead of an acute heating threshold, while also illustrating how strongly the answer depends on modeling assumptions.

S4–Mito–Spin offers RF Safe a formal way to ask what a heat-only model leaves unresolved. The S4 branch asks whether electrical gating and calcium-code fidelity can be perturbed. The Mito branch asks whether mitochondria amplify those changes and fail to recover. The Spin branch asks whether field-sensitive radical chemistry changes reaction probabilities. The persistence gate asks whether an acute perturbation becomes biological history.

None of those mechanisms can be read directly from a cancer graph. Together, they define the experiments needed to determine whether the graph and the exposure share a cause.

The policy conclusion does not require waiting for perfect certainty. Exposure standards should be capable of evaluating the endpoints that modern science identifies. Communities should not lose their voice because a facility complies with a thermal rule. Federal agencies should maintain an active, independent program of research and surveillance. Safer design, lower exposure, wired options, transparent testing, and protection of children are compatible with technological progress.

The limit is a heating number. The biological question is a lifetime question.

It is time for safety science to measure both energy and time—and to protect not only whether tissue remains cool, but whether living regulation remains precise, resilient, and capable of recovery.

Selected sources and further reading