Launched in 2005, the "Ramazzini study" forms part of a large-scale series of experiments³ intended to evaluate "diffuse carcinogenic risks". It is the largest long-term study of the effects of high-frequency electromagnetic fields ever to be carried out on rats.

For their entire lifespan, a total of 2448 animals were kept unrestrained in cages containing up to five animals in far-field-like conditions generated by a collinear dipole array antenna. The animals were divided into various exposure groups with field strengths of 0 (control group with sham exposure), 5, 25 and 50 V/m. The first two exposure groups each consisted of approximately 400 animals per sex, while the two others consisted of approximately 200 animals per sex. Whole-body exposure to GSM-modulated high-frequency electromagnetic fields with a frequency of 1835 MHz (corresponding to a mobile communications base station operating at 1.8 GHz) began on the 12th day of gestation in the mother animal and continued until the natural death of the offspring. The total duration of the experiment was therefore longer than that in the NTP study, in which all surviving animals were killed after a defined period of approximately two years. The animals were exposed for 19 hours a day without interruption and, unlike in the NTP study, without short-term fluctuations or the intermediate switching on and off of exposure. According to the authors, the antenna generated a homogeneous, cylindrical far field extending across all of the cages.

The corresponding whole-body specific absorption rate (SAR) of the rats in the respective exposure groups was estimated at 0, 0.001, 0.03 and 0.1 W/kg using empirically determined coupling factors. These whole-body SARs are below or, in the highest exposure group, at a comparable level to the basic limit values for humans, which have been recommended internationally in order to protect against adverse health effects. In comparison with the NTP study, these values are at a significantly lower level. The stated values for the ambient humidity (40–60 %) and room temperature (22 °C), as well as the 12-hour day/night cycles, are comparable with the conditions in the NTP study.

After natural death occurred, the animals underwent histopathological examination. Falcioni et al. indicate that a blinded pathological diagnosis was followed by a second, also blinded, assessment by NTP pathologists. In the Ramazzini study, various end points were considered in parallel, but the exact number of examined end points is as yet unknown.

From a dissertation that describes the experiment in some sections⁴, it can be deduced that histopathological examinations are to be performed on more than 30 tissue types. The present publication by Falcioni et al. deals with a total of seven organ end points in relation to the heart (Schwann cell hyperplasia as well as endocardial and intramural schwannomas) and brain (benign and malignant tumours of the meninges, as well as glial cell hyperplasia and malignant gliomas).

So far, only histopathological examinations of the brain and heart of exposed and unexposed animals have been published.

Heart

Relative to the control group, male rats showed a statistically significant increase in the incidence of schwannomas of the heart (intramural and endocardial schwannomas) in the highest exposure group but no statistically significant increase with increasing exposure (0 %, 0.7 %, 0.5 %, 1.4 % at 0, 0.001, 0.03, 0.1 W/kg). In female rats, no statistically significant differences were observed between the incidences for control and exposed animals. In male and female rats, the highest exposure group in each case showed an increased incidence of Schwann cell hyperplasia (a possible preliminary stage of schwannoma) relative to the control groups, but this was not statistically significant (male animals: 0.7 %, 0.5 %, 0.5 %, 2.4 %; female animals: 0.5 %, 0 %, 0%, 1.0 % at 0, 0.001, 0.03, 0.1 W/kg).

Brain

Relative to the control groups, no statistically significant increases were observed in the incidences of premalignant and malignant brain tumours (including glial cell hyperplasia, malignant gliomas, and benign and malignant tumours of the meninges) in the various exposure groups for either sex. In female rats, an increase was observed in the incidence of malignant gliomas with increasing exposure (0,5 %, 0,7 %, 1,0 %, 1,5 % bei 0, 0,001, 0,03, 0,1 W/kg), but this trend was not statistically significant.

In addition, the following parameters were also recorded: food and water consumption, average body weight, and the survival rate of the animals. No differences were identified between the control group and the exposed groups for either male or female animals.

The strengths of the Ramazzini study in comparison to the NTP study relate in particular to the higher number of laboratory animals per exposure group by a factor of two to four (e.g. for the control animals, Ramazzini: > 400 vs NTP: 90), the significantly longer duration of the study (Ramazzini: lifelong up to 152 weeks vs NTP: 107 weeks) and the blinding of the histopathological examinations, which makes it unlikely that the analysis will be influenced by expectations. However, there was no indication of the manner in which this blinding was ensured. Unlike in the NTP study, there is no significant difference between the survival rates of control animals and exposed animals, so there is no reason to fear adulteration of the results due to differences in the average age of the groups. It is also to be commended that, in addition to body weight, the Ramazzini study recorded data relating to food and water consumption. This data indicates that exposure did not lead to considerable changes in metabolism. The NTP study did not collect data from which conclusions could be drawn regarding changes in metabolism or possible thermal stress at high levels of whole-body absorption.

Documentation of procedures

The published data can be considered a partial result of a study that is yet to be completed and that is not documented nearly as well as the final reports from the NTP study. A full evaluation of the results of Falcioni et al. is currently prevented by a lack of background information.

Multiple Testing

From a dissertation that describes the experiment in some sections⁴, it can be deduced that histopathological examinations were to be performed on more than 30 tissue types, although the article does not indicate the full number of examined end points. In Table 2 and Table 3, however, at least seven tissue types are examined separately by gender in three group comparisons (Group II vs I, Group III vs I, Group IV vs I). Purely based on this potentially incomplete number of end points, it can therefore be assumed that the researchers performed at least 42 statistical hypothesis tests. A multiple test problem such as this is subject to a phenomenon known as type I error inflation. This means that the type I error, which is generally restricted by a significance level of α = 0.05, increases significantly in the absence of corresponding correction (e.g. Bonferroni correction). The article in question did not report the use of such correction. However, if one assumes that the hypothesis tests are independent and that the type I error is restricted to α = 0.05, one can already expect two false-positive test decisions among the reported results. It is not unlikely that the only statistically significant test result reported in Falcioni et al. represents a false positive.

Incidence in contro animals

In the Ramazzini study, there is a statistically significant difference between the incidence of schwannomas in the hearts of male rats in the utilised control group (0/412; 0 %) and in the highest exposure group (3/207; 1.4 %). The question is therefore: to what extent do the zero observed cases in the control animals give a realistic picture of the spontaneous incidence of schwannomas in the hearts of male rats? The authors themselves indicate that an incidence of 0.6 % for this tumour in male SD rats can be calculated using data from historical controls of past experiments between 1986 and 2004. The source of this data is not given, nor is there clarification of whether the data related exclusively to animals that died of natural causes – even though this would have a decisive effect on comparability, especially if late-onset tumours were considered. According to the authors, tumours were diagnosed in the present study according to the same criteria as in the NTP study. These criteria were recently revised by the NTP. It remains uncertain whether these diagnostic criteria were also applied to the historical control experiments, some of which date back more than 20 years.

More-recent data for the spontaneous incidence of schwannomas in the hearts of male SD rats in the control experiments was presented at an NTP symposium in 2016⁵ , where a figure of 9 out of 699 animals (1.3 %) was reported. The mean (1.4 %) and standard deviation (2.1 %) of the individual experiments point to relatively strong deviation between experiments. The official figures from the NTP, which are available on the NTP-Website⁶ and already include the control animals of the NTP study, are of a similar order of magnitude (9 cases in 489 animals, 1.8 %, standard deviation of experiments: 2 %). Here, it should be noted that the animals in the NTP experiments do not reach their natural life expectancy, as most long-term experiments are terminated after 104 weeks. At that point in time, over 30 % of the animals are typically still alive. In light of the late onset of schwannomas in the heart, values that originate from NTP experiments must therefore be viewed as a lower limit for the expected incidence in the event of comparison with the present study, where experiments are not terminated prematurely.

If one uses the historical control group data as a comparator for the exposure group data of the Ramazzini study, the observed incidence of schwannomas of the heart in the highest male exposure group is certainly, at 1.4 %, of a magnitude that appears to be likely even in the absence of exposure. The spontaneous rate is also subject to a range of variation at the Ramazzini Institute. For example, at 1 %, the incidence of schwannomas in the hearts of female control animals is significantly higher than the historical incidence reported by the authors (0.3 %) and the NTP (0.61 %) for female control animals. There is also significant variation in the observed incidence for exposed animals (2.2 %, 0.5 %, 1 % at 0.001, 0.03, 0.1 W/kg). Given the low number of cases, it cannot be ruled out that the statistically significant result of the Ramazzini study was down to variations in the spontaneous rate (control lower than expected; exposure group higher than expected).

Uncertainties in exposure estimation

There is some ambiguity regarding the exposure of the laboratory animals. Although it is unlikely, given the description of the experimental setup, that the exposure significantly exceeded the stated field strengths, the stated values are nevertheless insufficient for reliable dosimetry. The publication includes neither an antenna diagram of the array antenna that was used nor spatially resolved field strength measurements. Likewise, it does not indicate the precise measuring position for the monitoring of field strength or the measuring position to which the stated field strengths of the exposure groups relate. The whole-body SAR was determined using a coupling factor that reflects neither growth-related and individual size differences nor the posture of the animals. This factor originates from another publication⁷ and was determined for a frequency of 1.6 GHz and rats weighing 370 g. It is slightly less than the coupling factors for rats of a similar weight that were exposed to a frequency of 900 MHz in electromagnetic reverberation chambers in the NTP study⁸. As the coupling factor was determined exclusively for almost fully grown animals, it does not reflect the increased absorption efficiency of younger (smaller) rats, which exhibit much more efficient absorption than the fully grown rats at a frequency of 1835 MHz due to resonance effects. It cannot therefore be ruled out that the younger animals were exposed to higher SAR values than stated.

Moreover, the applicability of coupling factors relies on far-field conditions. In the case of an array antenna, the far-field distance is not necessarily independent of the antenna dimensions and can therefore significantly exceed the far-field distance of an individual dipole. In the absence of a detailed description of the antenna system, it is not clearly demonstrated that the animals were actually in the far field of the antenna. Further uncertainty stems from the fact that up to five animals were kept in a cage, which can result in shielding effects. An estimation of the dosimetric uncertainty resulting from aspects including those cited here was either not made or not mentioned by the authors. In contrast, detailed uncertainty estimates were published for the NTP study⁸.

Overall, the insufficient description of the manner of exposure undermines the reliability of the exposure estimation.

Summary of observed incidence of cardiac schwannomas in Sprague Dawley rats in the NTP and Ramazzini studies. This figure is adapted from "Thermoregulatory Stress as Potential Mediating Factor in the NTP Cell Phone Tumor Study" by Kuhne et al. and licensed under CC BY 4.0.

Schwannomas of the heart

Both the NTP study and the Ramazzini study collected data relating to the incidence of schwannomas in the hearts of female and male rats in various exposure conditions. A summary of this data, as well as data from control groups of other experiments, is shown in the figure (The figure is adapted from Fig. 3 published in "Thermoregulatory Stress as Potential Mediating Factor in the NTP Cell Phone Tumor Study" by Kuhne et al. and licensed under CC BY 4.0.). Overall, no exposure dependence can be identified for female rats; rather, the data appears to reflect statistical fluctuations around the spontaneous background incidence of schwannomas of the heart in female rats. The literature gives no indication that the spontaneous background rate in male rats is less than that in females, and so the fluctuations observed in females can also be expected for males, provided that the group sizes are comparable. Among the male rats, increased incidence can be observed for the high exposure groups in the NTP study.

As previously explained in a scientific publication by the BfS in relation to the NTP study², there are indications that this may be a thermal effect that is particularly pronounced in male rats. Based on the description of exposure, it is unlikely that such an effect could also be responsible for the results of the Ramazzini study. The incidence of 1.4 % observed in male rats at a significantly lower SAR of 0.1 W/kg in the Ramazzini study is within the range of expected statistical variation. The BfS therefore takes the view that although this result is statistically significant compared with the accompanying control group, it is insufficiently reliable to infer a causal link or – as the authors claim – to provide confirmation of the results of the NTP study.

Malignant gliomas

The observed increase in malignant gliomas (with increasing exposure: 0,5 %; 0,7 %; 1,0 %; 1,5 %) in female animals must also be considered from the point of view of statistical variation. Neither the pairwise comparison of exposure groups with the control group nor the trend test was statistically significant. In the NTP study, an increase in malignant gliomas in the case of very high exposure was observed primarily in male rats. The authors argue that, in some case-control studies, an association was seen between gliomas and intensive mobile phone use¹¹. However, a recently published meta-analysis summarising numerous epidemiological studies shows no consistent relationship between mobile phone use and the occurrence of gliomas¹².

Exposure

The field strengths used in the Ramazzini study (5, 15 and 50 V/m) are below the limit values stipulated for powerful stationary high-frequency installations in the 26th Ordinance Implementing the Federal Immission Control Act and therefore lie within a range that can theoretically occur for the general public in extreme situations. However, it is extremely unlikely that everyday situations experienced by the general public will result in even short periods of the whole-body specific absorption rate (SAR) at which a statistically significant increase in the incidence of cardiac tumours relative to the control group was observed in male rats subjected to lifelong exposure. For the general public, exposure that leads to this specific whole-body absorption rate on a lifelong basis is practically impossible.