The NTP study [1, 2] was conducted in three phases. In the 1st and 2nd phase, the authors determined the (maximum) specific absorption rates (SAR) suitable for whole body exposure that do not lead to increased mortality or excessive warming (above 1 °C) in laboratory rodents (Harlan Sprague-Dawley rats and B63F1 mice) after a short time. Based on the results from the 1st and 2nd phase, the whole body SAR values for the chronic exposure of the laboratory rodents were determined in the 3rd phase, the actual main study (detailed information on the three study phases can be found in Annex 1: Study design and results of the first and second study phase and in Annex 2: Study design and results of the NTP main study).

In the main study, the individually kept, freely moving animals (90 animals/group) were subjected to whole body exposure (for 18 h and 20 min per day, 10 min field off/10 min field on) with GSM and CDMA modulated 900 MHz (rats) and 1900 MHz (mice) signals at SAR values of 1.5, 3, and 6 W/kg for rats and 2.5, 5, and 10 W/kg for mice. The SAR averaged over the entire body of the animals was determined using numerical computer simulations and kept as constant as possible throughout the study. A sham-exposed control group for both modulations (GSM, CDMA) was used as comparison group. The following endpoints were investigated at the end of the study after two years: body weight and survival rate of animals, haematological and clinical-chemical variables, and neoplastic and non-neoplastic changes in organs and tissues.

A) Rats

• Survival: Exposed male rats lived significantly longer than the sham-exposed control animals (number of animals surviving to the end of the study: GSM 25, 45*, 50*, 60* or CDMA 25, 43*, 56*, 43 at 0, 1.5, 3, and 6 W/kg)^a . Exposed female rats lived as long as the control animals except for the animals of the 6 W/kg CDMA-exposed group, which survived significantly longer (number of animals surviving to the end of the study: GSM 48, 53, 48, 57 or CDMA 48, 46, 50, 61 at 0, 1.5, 3, and 6 W/kg).

• Heart: In male rats, the incidence of malignant schwannomas of the heart increased with increasing GSM exposure (0, 2, 1, 5 cases at 0, 1.5, 3, and 6 W/kg) and CDMA exposure (0, 2, 3, 6* cases at 0, 1.5, 3, and 6 W/kg). This exposure-response trend was statistically significant for both modulations and for CDMA in the highest exposure group (6 W/kg) compared with sham exposure. In the female rats, there was neither a trend nor a significantly increased incidence for individual exposure levels (GSM: 0, 0, 2, 0 cases and CDMA: 0, 2, 0, 2 cases at 0, 1.5, 3, and 6 W/kg respectively). The incidence of schwannomas in other parts of the body was unremarkable.

  Significantly (*) increased incidences of right ventricular cardiomyopathies occurred at the highest CDMA and the two highest GSM exposure levels in male rats (GSM: 54, 62, 72*, 74* cases and CDMA: 54, 45, 62, 74* cases at 0, 1.5, 3, and 6 W/kg) as well as in female rats (GSM: 4, 9, 14*, 15* cases at 0, 1.5, 3, and 6 W/kg) compared with sham-exposed controls.
  Increases in the incidence of Schwann cell hyperplasias (possible precancerous stage of schwannomas in the heart) in the male (GSM: 0, 1, 0, 2 cases and CDMA: 0, 0, 0, 3 cases at 0, 1.5, 3, and 6 W/kg) and female rats (only for CDMA: 0, 1, 1, 1 cases at 0, 1.5, 3, and 6 W/kg) were observed compared with controls. However, these were not significant.

• Brain: In male rats, increased incidences of malignant gliomas were found in all exposed groups compared with the control group at GSM modulation (0, 3, 3, 2 cases at 0, 1.5, 3, and 6 W/kg). There was also a significant trend with increasing exposure to CDMA modulation (0, 0, 0, 3 cases at 0, 1.5, 3, and 6 W/kg). In female rats, increases in the incidence of malignant gliomas (GSM: 0, 3, 0, 0 cases and CDMA: 0, 0, 0, 1 cases at 0, 1.5, 3, and 6 W/kg) occurred in only a few exposure levels, all of which were not significant.

  Increased incidences of glial cell hyperplasia (possible precancerous stage of malignant glioma) occurred in most exposed groups of both sexes. However, these were not significant.

• Adrenal medulla: In male rats, incidences of benign pheochromocytomas (GSM: 10, 23*, 25*, 14 cases at 0, 1.5, 3, and 6 W/kg) as well as benign, malignant, and complex pheochromocytomas (GSM: 11, 24*, 28*, 14 cases at 0, 1.5, 3, and 6 W/kg) were significantly increased but only at GSM modulation. In female rats, the incidence of benign pheochromocytomas (GSM) was predominantly only slightly elevated (GSM: 1, 3, 3, 2 cases and CDMA: 1, 7, 3, 4 cases at 0, 1.5, 3, and 6 W/kg) as well as benign, malignant, and complex pheochromocytomas (CDMA: 1, 9*, 5, 4 at 0, 1.5, 3, and 6 W/kg).
• Kidney: In male rats, the severity of chronic progressive nephropathy in all exposed groups (GSM and CDMA) was lower than in the sham-exposed control group at equal incidence rates. For a number of possible secondary side effects of chronic progressive nephropathy, significantly reduced incidence rates were observed in all exposed groups of male rats (GSM and CDMA).
• Other organs: Occasionally significantly increased incidences of unclear significance: for example, for benign or malignant granular cell tumours, adenomas and carcinomas of the prostate gland, adenomas of the pituitary gland, and pancreatic islet cell adenoma or carcinoma (GSM) as well as for adenomas of the pituitary gland and hepatocellular adenomas or carcinomas of the liver (CDMA).

B) Mice

• Survival: The survival probability in most groups of male and female mice did not differ significantly from the sham-exposed control group.
• Heart, brain, and adrenal medulla: In contrast to rats, malignant schwannomas of the heart, malignant gliomas, and diseases of the adrenal medulla in mice showed neither a trend with increasing exposure nor significant increased incidences in the individual exposure levels compared with sham-exposed controls.
• Other organs: Occasionally (significantly) increased incidences or trends of unclear significance: Liver tumours, skin tumours, and lung tumours in male mice; malignant lymphomas in female mice.

A detailed list of further results from the main study can be found in Annex 2: Study design and results of the NTP main study.

-----
^a) significant results are marked with *.

[1] 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; NTP TR 595, November 2018

[2] NTP TECHNICAL REPORT ON THE TOXICOLOGY AND CARCINOGENESIS STUDIES IN B6C3F1/N MICE EXPOSED TO WHOLE-BODY RADIO FREQUENCY RADIATION AT A FREQUENCY (1,900 MHz) AND MODULATIONS (GSM AND CDMA) USED BY CELL PHONES; NTP TR 596, November 2018

The final evaluation by the 19-member NTP team took place after a three-day peer review process of a pre-publication in March 2018. Scientists as well as members of the public participated and discussed the results.

According to their specific toxicological classification criteria, NTP sees

• clear evidence for the occurrence of tumours in the heart (malignant schwannomas) (GSM and CDMA modulation) in male rats
• Some evidence for the occurrence of brain tumours (gliomas) (GSM and CDMA modulation) and diseases of the adrenal gland (GSM modulation) in male rats
• equivocal evidence for individual neoplastic and non-neoplastic changes in male and female rats and mice.

The Federal Office for Radiation Protection sees indications in the extensive data but no proof that the tumours in the hearts of male rats could be a consequence of the whole body exposure under the toxicological test conditions of (partly extremely) high and long-lasting exposure. This is due to a number of methodological weaknesses and inconsistencies in the study results as well as particularities in toxicological study design, which are described below:

• Gender specificity

  A significantly increased incidence of malignant schwannomas in the heart was found only in male rats. This result is therefore not consistent with the results in female rats or mice within the NTP study. However, they are consistent with another animal study recently published by Falcioni et al. [3]. However, gender differences in tumour incidence are often observed in NTP carcinogenicity studies. Possibly study-dependent, tumours mostly occur more frequently in male rats [4].

• Low number of preneoplastic lesions

  The number of possible preneoplastic lesions in both exposed and non-exposed control animals was smaller than the tumour incidence in almost all groups. The reverse case would have been expected from the current models of carcinogenesis. However, it is unclear whether the hyperplasias observed (Schwann cell and glial cell hyperplasia) are actually preneoplastic lesions of the malignant schwannomas or gliomas, which is why NTP authors also refer to them as "putative/potentially" preneoplastic lesions.

• Differences in the survival rates

  In the male rats, the control animals had a significantly lower life expectancy than the exposed animals (28 % survival rate in control animals versus 50–68 % in GSM- and 48–62 % in CDMA -exposed groups). The control animals showed no malignant schwannomas of the heart or malignant gliomas compared with the exposed animals. Although this is within the incidence range of historical controls (malignant schwannomas, 0–2 % incidence; malignant gliomas, 0–4 % incidence), it cannot be excluded that the animals died before developing a tumour. This may have led to an under-representation of late-onset heart schwannomas in control animals and thus to an overestimation of the associated risk. The low life expectancy of the control animals might have had an even greater effect on the malignant gliomas, which usually only occur in older animals. The differences in the survival rates were taken into account in the statistical evaluation through appropriate corrections. It is unclear whether the assumptions on which the corrections are based apply to the yet underexplored schwannomas of the heart.

• Exposure conditions and dosimetry

  Exposing freely moving animals while simultaneously ensuring meaningful dosimetry is a great challenge. However, the researchers were able to keep the specific absorption rate of whole-body exposure comparatively constant at a fixed value over the entire duration of the study per exposure group (time-averaged). In long-term toxicological studies, it is crucial that the dose (SAR) is kept constant. However, the specific absorption rate alone is not representative of the expected temperature increase of the animals, which, if too high ( > 1 °C), can lead to proven health effects.

• Missing temperature measurement

  A major weakness of the study is that body temperature was not recorded in the main study. Nor were any parameters collected to record the metabolism of the animals (e.g. water consumption, CO^₂ production) and thus to obtain indications of increased body temperatures. In the first phase of the study (Annex 1), adult male rats had significant body temperature increases (some measurements significantly > 1 °C) from a whole body SAR of 6 W/kg onwards. There was also a clear correlation between body weight and temperature increase. This can be explained by the fact that large, heavy animals have a lower thermally radiated power (cooling effect) than small animals because of their small surface area compared to their body mass. This is also compatible with the small temperature increases measured in phase 2 in younger and thus lighter male rats. Because of the lack of temperature measurement in the main study, it cannot be excluded that an excessive warming of > 1 °C was observed in male (older and heavier) rats. These increases in body temperature above the limits are known to have health effects. It is therefore possible that the increased tumour incidences observed in male rats were the result of thermal stress.

• Multiple testing

  Statistical tests were performed for a number of endpoints: separately for male and female animals, GSM and CDMA, intermediate results and two-year study as well as three exposure groups. Because of the large number of tests, significant results of a purely random nature can be expected. For continuous variables (e.g. body weight), the researchers corrected for multiple testing. However, discrete variables (e.g. incidences of tumours) were not corrected for multiple testing according to standardised NTP protocols. There is thus a higher likelihood of false positive results. In the statistical sense, none of the significant results can be individually evaluated as actually "significant" (i.e. higher than random).

• Blinding

  The histopathological examinations were carried out in a multi-stage process by several pathologists. However, the initial evaluation of the tissue sections to identify possible lesions in mice and rats was unblinded. This means that the pathologists in the initial evaluation were always aware of whether the tissue section to be evaluated came from an exposed or sham-exposed mouse or rat. This is unusual in scientific research but corresponds to the recommendations of the American Society for Toxicological Studies [5]. A lack of blinding can lead to a distortion of the results because individual – also unconscious – expectations of the analysing scientists can influence the number and severity of the tissue changes found in the respective groups.

Malignant cardiac tumours of the heart are extremely rare in humans. However, the results of the NTP study are relevant insofar as benign acoustic neuroma (tumours of the auditory nerve), which are more frequent in humans, belong to the group of schwannomas. Epidemiological studies have revealed isolated indications of a possible connection between acoustic neuroma and mobile phone use. In this context, it is worth mentioning that in the NTP study, no significantly increased incidence rates were found in any of the schwannomas (of all organs) examined. However, a histopathological examination for acoustic neuroma was not performed in the NTP study.

Previous animal studies (e.g. the multi-generational studies carried out in the German Mobile Telecommunication Mobile Radio Research Programme (DMF)) had the aim of verifying the existing limit values. These were therefore carried out at lower radiation intensities. A carcinogenic effect of radio-frequency fields with cell phonemobile radio -relevant frequencies and modulations was not seen in most DMF studies and some other animal studies [6] (exception: Falcioni et al. (2018) [3]).

The results of the NTP study thus differ from previous study results and raise the question of whether thermal stress may have led to the findings observed. This should be clarified in future research.

The SAR distributions in human and animal tissues vary widely: significant differences exist with regard to body sizes, organ sizes, and organ arrangements, each of which affects the SAR distributions. In mice and rats, internal organs are reached more strongly by radio-frequency electromagnetic fields acting from outside than in humans.

The whole-body SAR values used in the main study (≥ 1.5 W/kg) are significantly (approx. 20 times) higher than the nationally and internationally recommended limit value of 0.08 W/kg for whole-body exposure for the general population to protect against proven health effects. This value is used to derive the limit values applicable in Germany for fixed radio-frequency systems; these are far from being exhausted in everyday situations. For example, studies conducted by the DMF (German Mobile Telecommunication Research Programme) have found that in the proximity of GSM and UMTS base stations the limit value for the electric field strength has been exhausted to a maximum of 10 %. Under these conditions, it can be estimated that the whole-body SAR values induced in the human body do not exceed the order of 0.001 W/kg and are therefore at least three orders of magnitude (factor 1000) below the whole-body SAR values of the exposed rats in the NTP study.

Only the lowest exposure level used for rats (1.5 W/kg) is below the recommended local exposure limit (for the general population, the recommended SAR limit for the local exposure of head or torso is 2 W/kg). However, the dosimetric analyses [7] carried out by the authors at organ level show that the exposures of the hearts of male rats that were striking in the histopathological examinations were on average 1.9 times higher than in the body average (i.e. in the lowest exposed group at 2.8 W/kg and thus already above the recommended limit value for head and trunk).

[3] Falcioni, L., et al., Report of final results regarding brain and heart tumors in Sprague-Dawley rats exposed from prenatal life until natural death to mobile phone radiofrequency field representative of a 1.8GHz GSM base station environmental emission. Environ Res, 2018. 165: p. 496-503.

[4] Kadekar, S., et al., Gender differences in chemical carcinogenesis in National Toxicology Program 2-year bioassays. Toxicol Pathol, 2012. 40(8): p. 1160-8.

[5] Crissman, J.W., et al., Best practices guideline: toxicologic histopathology. Toxicologic Pathology, 2004. 32(1): p. 126-131.

[6] SCENIHR 2015. Potential health effects of exposure to electromagnetic fields (EMF). 27 January 2015.

[7] Gong, Y., et al., Life-Time Dosimetric Assessment for Mice and Rats Exposed in Reverberation Chambers for the Two-Year NTP Cancer Bioassay Study on Cell Phone Radiation. IEEE transactions on electromagnetic compatibility, 2017. 59(6): p. 1798-1808.

For future animal studies we recommend a continuous recording of the following parameters:

• Core and peripheral (e.g., skin) temperatures
• Infrared thermography to assess peripheral vasodilation, indicating heat stress
• Heart & respiratory rate
• Metabolic parameters (e.g., CO² production, O² consumption, Water- and food intake)
• Behavioral changes indicative of thermal stress (e.g., sprawled body posture, saliva grooming of the fur and tail)

We recognize that experiments can never control for every parameter due to limiting factors such as funding. Especially the monitoring of heart rates might be challenging if one deals with free roaming animals. In recent years, image processing has developed rapidly allowing to extract heart and breath rates from video data [Zhao et al., 2013]. For future studies we therefore suggest to continuously videotape animals in order to gain information on behavior, heart and respiratory rate.