Does mobile phone radiation cause cancer?

Principle findings

Following detailed lab experiments, researchers at the US National Toxicity Program (NTP[1]) have published[2] data which provides evidence of a statistically significant association between high exposure to radio frequency emfs (RF emfs) and malignant schwannoma of the heart in male rats. The evidence revealed thus far is consistent with a causal explanation. Extrapolating to human disease is however unclear; as this depends on the mechanism behind this association. It is hoped that in the final report, NTP will be as specific about this mechanism as they can be. As it is, there are signs in the data that the mechanism may be unusual, so much so that a new challenge could be made to the norms of cancer compensation.

The final NTP report is due before the winter of 2018. As a draft for peer review, the NTP report may not be quoted.

A challenge to the norms of cancer compensation

The courts, and therefore insurers, nearly always manage cancer claims as if the causal mechanism was stochastic[3] in nature. This gives rise to the familiar ‘doubled risk’ test of legal specific causation. However, one way to make sense of the published NTP data is to propose a non-stochastic disease mechanism. If this is correct it could cause some uncertainty. For example, the legal causation tests for non-stochastic cancer have not been firmly established, but in principle should often favour the claimant. This because they may only need to show a material contribution to gain 100% of the damages.

Academic interest in non-stochastic cancer has been growing and several State expert committees[4] have adopted it in public health statements. It is now very clear that not all cancer mechanisms are stochastic. How will the common law respond? Will liability insurers help shape that response?

The insurance industry would need to step carefully if the final NTP report promotes a non-stochastic explanation for their findings. If it can be achieved, a consistent response to cancer claims which allege a non-stochastic mechanism would seem to be advisable.

The NTP experiment went wrong

The US NTP is a world class institution for testing carcinogenicity in rodents. This latest work on radio frequency emfs, as used in 2G and 3G mobile phones, was of a very high quality, but interpretation was hampered by an unforeseeable and as yet unexplained trend in kidney disease in male rats.

Kidney disease is very common in older rats. What was unexpected in this experiment was that the severity of the disease was inversely correlated with the intensity of emf exposure. The kidney disease severity was highest in the control group. The lowest severity was in the group with the highest emf exposure.

Kidney disease has many effects on health, including organ-specific lesions such as cancer and benign tumours, so any disease trends with increasing RF emf exposure could also be explained by trends in kidney disease severity. In effect, some real RF emf-related trends could be confounded by an unrelated trend in kidney disease especially if the emf effect competes with the kidney disease effect.

The problem then is when or if to compare RF emf exposed animals with this particular control group. Might it be better to use historic controls from other lab work instead? Experts are working through the options and will give their final view later this year.

The rats were examined in minute detail after 2 years of RF emf exposure[5].

Are lab experiments relevant to disease and liability in humans?

Lab experiments are qualitatively relevant for human disease and personal injury claims if the mechanism of disease is comparable in both animals and humans. If an association between exposure to hazard and injury is found in rats, as it was in this case, this must be followed up by experiments on injury mechanism if the relevance to human health and litigation is to be decided.

But is there a true association here?

One problem with rat experiments is that standard statistical methods[6] of fact determination are not the same as fact determination at common law. Lab scientists usually adopt statistical techniques which are precautionary in nature and have no natural meaning in the context of a common law judgment. Common law relevant statistical techniques are usually not considered by lab experiment experts simply because their primary aim is to inform public health politics – where a precautionary approach is mandated. If the courts are to decide the facts for themselves, as indeed they should, they need to see the results of using common law compatible statistical techniques[7] for such lab experiments[8].

Common law compatible statistical techniques have now been applied[9] to the data provided by NTP. The interpretation of this work informs the findings presented here. By chance, many of these conclusions are very similar to those offered in the draft NTP report.

What would a good case for cancer causation look like?

Regardless of disease mechanism[10], a case for RF emf carcinogenesis would be supported:

1. By finding a higher number of malignant growths in exposed rats, especially if there was a positive trend with increasing exposure.
2. By finding that benign lesions[11] were also more common at the particular site of any excess malignancy and again, especially if their number correlated with exposure.
3. Male and female rats have different cancer susceptibilities and different susceptibilities to benign lesion so sometimes if the same cancers and benign lesions were found in both sexes the strength of the case would be increased.
4. GSM and CDMA have different duty rates but these differences may be mechanistically irrelevant. If what they have in common is relevant then the case for causation is strengthened if the effect is seen in both types of radiation.

The rats in this NTP research were extensively tested for evidence of genetic damage. The tests used are generally sensitive but not specific[12]. A positive test would suggest the possibility of genetic damage, especially if the rate of positive results corresponded with exposure intensity. Again, a common law compatible test of such trends would be useful in advising the courts. This statistical test work has been done.

Common law compatible tests of NTP data gives the following findings

The most significant and robust finding was a raised rate of malignant Schwannoma[13] at the heart. This was found most clearly in male rats. The numbers correlated with RF emf intensity (both for CDMA and GSM) and correlated with the number of benign lesions at roughly the same anatomical location.

Benign lesions at this site are much less common in females but a similar dose response curve was obtained. A small excess of malignant Schwannoma at the heart was also seen in female rats.

Body wide, Schwannomas were not in excess. This suggests the findings at the heart are not attributable more generally to the type of cell involved but to a location-specific factor at the heart.

There was no apparent effect of kidney disease on Schwannoma. This increases the confidence that the findings are real.

Evidence of genetic damage was at best patchy, and did not correlate with exposure.

A great many other lesions and cancers were recorded but none were at significant excess. As expected, several trends in disease were dominated by the kidney disease effect.

Injury causation?

The pattern of results is probably real and is consistent with a causal link between high RF emf exposure and male heart Schwannoma. Being seen in both GSM and CDMA the mechanism would depend on something these signal formats have in common, rather than something which differentiates them from each other.

Lack of genetic damage and lack of increase in Schwannoma at other body sites/organs suggests a non-genotoxic mechanism. Lack of other kinds of tumour suggests and organ-specific effect of RF emfs with males being more vulnerable than females.

Causation concerns would be raised if further research establishes the mechanism of heart Schwannoma development in male rats and it could be shown that the same mechanism would apply in humans. The theory would also have to account for why there was no excess Schwannoma at other sites in the rat and why there was probably no genetic damage in the tested tissues.

A self-consistent hypothesis would be that the RF emfs caused localised heating at certain parts of the heart. Such ‘hot spots’ would be expected where fluid flow was slow (e.g. cartilage or, nerve) and where there are RF emf absorption peaks. Such a disease mechanism would probably be non-stochastic. If so, degree of heating would be mechanistically important, and, there may be a temperature below which the risk is not increased.

It would be possible to observe localised heating at specific organs, but this would be a difficult experiment.

Perhaps the most useful work for now would be to repeat the NTP experiment but without the severe kidney disease.

Legal causation

Recent UK precedent suggests that the ground work for non-stochastic cancer claims is already in place.

A non-stochastic injury is one which is caused by the degree of an effect as opposed to the existence of an opportunity for an effect. For example, a brief exposure to heat, even if 1,000 °C will cause no injury. The pulse of additional heat is carried away by blood flow, conduction, re-radiation etc. No damage is done. However, if the rate of heat input exceeds the rate of heat loss then the tissue will warm up. Injury occurs when the temperature exceeds a threshold determined by chemistry e.g. protein is thermally denatured. Bigger injury occurs when bigger doses of super-threshold exposure to hazard are experienced. It is natural to think of the net effect of several super-threshold doses of heat as being cumulative and to apply the term divisible.

Cumulative injury, such as NIHL, is an established theory and is not stochastic in causation; the insult is direct, action and reaction. In principle, with powerful detection methods, the injury could be detected immediately. The body can cope well with certain levels of noise but is damaged at higher levels. There is a practical threshold. Once passed, the severity of effect depends on the degree of excess exposure to hazard. NIHL is non-stochastic, divisible and cumulative.

Cancer is not divisible. Yet the argument is that some kinds of cancer are non-stochastic. What this means is that a key step in the disease mechanism was harmless until a threshold was exceeded. It doesn’t matter that all the other steps were stochastic, one non-stochastic step determines the non-stochastic nature.

Among the key questions for non-stochastic causation theory for cancer relate to threshold, de minimis and contribution.

Does the common law recognise the concept of threshold?

In Bussey v Anglia, it was stated that a threshold test would apply to the question of foreseeability in case of a disease which was given to be stochastic in nature. The problem with this is that for a stochastic disease mechanism, all exposures give rise to foreseeable risk of injury, even if legally speaking, they are not actually causal and are too small to count as breach of duty. A threshold test of foreseeability is therefore meaningless for stochastic injury mechanisms, but the court adopted this approach non-the-less. Since threshold would often be relevant in judgements about foreseeability in non-stochastic disease and since the court has indicated an unjustifiable preference for threshold tests in Bussey it seems likely that the concept of thresholds in non-stochastic injury would be accepted without much difficulty. Such tests offer a pragmatic method of relating facts to legal principles even if they are illogical.

Does the common law still recognise the principle of de minimis?

In Sienkiewicz v Grief, the Supreme Court decided that for an indivisible injury: The only circumstances in which a court will be able to conclude that wrongful exposure of a mesothelioma victim to asbestos dust did not materially increase the victim’s risk of contracting the disease will be where that exposure was insignificant compared to the exposure from other sources. This confirms, even in case of mesothelioma, that the Court respects the concept of de minimis, even if it has no method for arriving at a scientific statement[14] of de minimis.

For a non-stochastic disease, de minimis is a pivotal decision to make. The question this time is whether all hazard exposures below the threshold are de minimis if their effect is reversible, or can they accumulate to the point where the threshold is exceeded? The answer depends on the rate of resolution of the effect of exposure to hazard and the rate of acquisition of the next exposure to hazard.

Does the common law recognise divisibility of contribution in indivisible injury?

In Heneghan v Manchester Dry Docks, despite finding that ‘lung cancer and mesothelioma are legally indistinguishable’, the court decided that damages for lung cancer should be in proportion to exposure to hazard rather than the joint and several liability which applies in mesothelioma claims. In Heneghan there was in total a doubled risk of lung cancer due to asbestos but the contribution to this made by any one of the defendants was much less than could be assigned a doubled risk. In so finding, a mathematical test of contribution to an indivisible injury is an accepted principle. Bailey v MOD clarified the disease mechanism test of contribution in indivisible injury.

In effect, the law is already prepared for non-stochastic cancer in respect of its understanding of threshold, de minimis and divisibility. Changing the court’s view of these concepts would be quite a challenge.

One key challenge remains:

Would the common law hold as harmless any exposure to hazard which arrived after the threshold had been exceeded?

Given that cancer is indivisible once in progress, further exposures should not count as causal (but could be held to accelerate the progress of disease). Regardless of the order of events, a non-stochastic cancer cannot be said to exist until after the threshold condition has been exceeded. Where cancer is manifest (as it must be in a valid claim) the threshold must have already been exceeded. It would seem to me that the question of causation entirely rests on how the threshold has been exceeded.

Proof of causation?

To show causation, the claimant must show that the rate of effect of exposure to hazard exceeds the rate of resolution of that effect and, does so to the point where an injury or mechanistic threshold is exceeded. If the effect is heating, the tissue must actually warm up and must do so to the point where the injury can be caused.

Heating at a particular tissue/organ where the injury occurs would be the norm, though remote effects are not impossible. Experiments would be needed to show that it was heating, and not the mechanism of heating that caused the problem. For such experiments it would be necessary to ensure that localised inflammation was not being caused by the intervention.

Is this a lower hurdle?

Yes. The Claimant does not need to show a doubled risk, he need only show a meaningful contribution to the size of a causative effect at below the threshold. Any source of heat will raise the temperature. If the degree of heating can be distinguished from general thermal fluctuations then it could be regarded as significant.

A testable definition of de minimis is needed.

Two different agents may have the same effect in respect of approach to threshold. If the effect is cumulative then the contribution could be apportioned. If not cumulative then the one which caused the super threshold effect should be identified.

Health Significance?

While there is no reason to suppose that humans would have the same vulnerability at the heart, it is perhaps useful to note that there are around 50 human cases of heart Schwannoma in the UK each year. At such low rates, statistics are not reliable, but there has been no suggestion of any decadal trend.

RF emf emission levels in normal use are between 10 and 1,000 times lower than the immersion rates used in this experiment. Further, most of the emission is not absorbed by the body. It would be remarkable if any part of the human anatomy was exposed at the rates used in this experiment, even if the phone was enclosed within a body cavity. Instantaneous power absorption in humans is usually less than 100^th of the level used in these experiments and average daily absorption is even lower.

Hot spot absorption rates of 1.5 W/kg are permitted in the USA. By exposing the whole animal to this rate or higher, the experimenters ensured that every potential hotspot was overheated[15]. This could not occur under conditions of permitted use for mobile devices and routers.

Perhaps the real significance is that RF emf could be carcinogenic and that if it is, it is unlikely to be a result of direct damage to DNA.

If in humans the real concern is brain cancer then it may be useful to know that there are[16] around 4,000 aggressive brain tumours diagnosed in the UK each year.

 

Claims significance?

The rationale for personal injury claims is fortified by this research. However, to prove generic causation, the claimant would need to show that Schwannoma can be caused in a way which also happens when a person is exposed to RF emfs.

For specific causation there would need to be a relationship between demonstrable use of RF emf devices and the location of the cancer. It need not be at the heart.

Heating would be one proposed mechanism. If this is the true mechanism, then, as it would probably be non-stochastic, any degree of heating above de minimis but below the damage threshold could count as making a contribution and, for an indivisible outcome, any material contribution could attract 100% of the liability. Heating, or some other non-stochastic mechanism, would therefore be an attractive generic causation theory for the claimant. De minimis would need to be defined in the context of competing rates of heating and heat dissipation.

Specific causation would then depend on showing the location of the injury was associated with heating which was significant e.g. the effect was outside the normal fluctuation in body temperature at that site.

Glial cell carcinoma is not showing any trends in humans[17].

Most epidemiology[18] finds no association between phone use and glial cell carcinoma.

Insurance

Exclusions for emf-related injury were once normal and interest in these may now gain renewed momentum.

Assuming defences are offered, valid personal injury exposure would probably be very small. The significant risk potential would be if an undefended non-stochastic mechanism holds sway.

Cancer stories can have an effect on share price. Sometimes such effects are covered by D&O type insurance.

Key points

After ensuring a common law test of fact was applied, it is probable that GSM and CDMA RF emfs caused an excess of malignant Schwannoma in male rats and likely that the same effect was seen in female rats.

A causal explanation is consistent with the evidence but the absence of genetic damage and the specificity of the site of the malignancy suggest the possibility of a non-stochastic disease mechanism. It would be helpful if NTP could comment on this in their final approved report.

Non-stochastic cancer claims would be a new challenge for the common law. Many supportive precedents are already in place.

There is a growing list of non-stochastic cancers which would often fail a doubled risk specific causation test but which could generate successful claims if the test was one of material contribution.

_______

 

[1] https://ntp.niehs.nih.gov/

[2] Peer review draft. Scheduled Peer Review Date: March 26 to 28, 2018. NTP Technical Report on the Toxicology and Carcinogenesis Studies in Hsd:Sprague Dawley SD Rats Exposed to Whole-Body Radio Frequency Radiation at a Frequency (900 MHz) and Modulations (GSM and CDMA) used by Cell Phones.

[3] A stochastic process is one where the key steps are random. For cancer to form, a living cell must acquire several pro-carcinogenic changes. In a stochastic process, changes to that cell are acquired at random. In a non-stochastic process the change is as a result of a threshold being exceeded by an effectively cumulative mechanism.

[4] A leading light in this regard is, The Health Council of the Netherlands. https://www.gezondheidsraad.nl/en/home . See for example hexavalent chromium : https://www.gezondheidsraad.nl/sites/default/files/201613_hexavalent_chromium_compounds.pdf

[5] Emf exposure was effectively uniform over the whole body, lasted for 9 hours each day over an 18 hour period with 6 continuous hours a day free from exposure. The exposure was cycled to permit close control of the thermal environment. Power levels were adjusted as the animals gained weight from infancy. Immision levels were at 1.5 W/kg, 3.0 W/kg and 6.0 W/kg. [At 9 W/kg some animals were warmed by around 1.0 °C.] All surviving animals were killed at 104 weeks and all animals were examined in detail.

[6] For example, a one-sided t-test with a p value of 0.05 is hard to reconcile with fact determination in the context of the common law. As a result, expert witnesses often unwittingly provide testimony that would be more suited to a public health debate. This should become apparent at cross-examination.

[7] The essential question when comparing the number of cancers in control rats with the number of cancers in exposed rats is “is it more likely than not that two numbers are different”. If they are probably different then there is a fact in need of an explanation. If they are not probably different then the conclusion must be that nothing happened.

[8] Details are available to Radar service subscribers. Re: Liability (Oxford) Ltd provides a liability emerging risks information service (Radar) to insurers trading in the UK.

[9] Re: Liability (Oxford) Ltd explained the findings at the latest Radar quarterly meeting. (June 2018).

[10] The key question at common law is whether the mechanism is stochastic or non-stochastic. A stochastic mechanism usually implies a doubled risk test for specific causation in humans. Cancer is usually regarded as stochastic because the key DNA damage events are individually complete at inception and occur at random.

[11] Benign lesions already exhibit some of the qualities that are needed for malignancy. DNA damage in a benign lesion is thought to have a higher probability of developing into malignancy.

[12] This is the norm in cancer research. Researchers are looking for any DNA damage. If found, the actual damage they observe is very often mechanistically uninterpretable. If not found, then this is good evidence that the hazard is not genotoxic.

[13] A Schwannoma is a cancer which develops in Schwann cells; a subset of glial cells. These cells are associated with nerve tissue and are found in all parts of the body.

[14] In the absence of a scientifically testable definition of de minimis it is quite difficult for expert witnesses to justify any advice to the court. Very often the expert will unwittingly adopt a public health based approach to his/her thinking. This should be evident at cross-examination.

[15] There were no localised temperature measurements; these would be useful in further experiments.

[16] http://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/brain-other-cns-and-intracranial-tumours/incidence#heading-Four

[17] SCENIHR (2015). Potential health effects of exposure to electromagnetic fields (EMF).

The following summary points were made:

Overall, the epidemiological studies on mobile phone RF EMF exposure do not show an increased risk of brain tumours…

The results of cohort and incidence time trend studies do not support an increased risk for glioma while the possibility of an association with acoustic neuroma remains open.

At that time: animal studies are considered to provide strong evidence for the absence of an effect.

Note that acoustic neuroma is a Schwannoma at the acoustic nerve in the brain.

[18] The Radar service has recorded and analysed the findings of such research since 2002. There is one result which could be used to support an association. By all reasonable tests, the one result is an outlier.