CENTER FOR OEVICES & RADIOLOGICAL HEALTH

Tribal code

Ask Donna

What actually matters in this document.

Text

U.S. FOOD & DRUG

CENTER FOR OEVICES & RADIOLOGICAL HEALTH

Review of Published Literature between

2008 and 2018 of Relevance to

Radiofrequency Radiation and Cancer

February 2020

Review of Published Literature between

2008 and 2018 of Relevance to

Radiofrequency Radiation and Cancer

Page 1

www.fda.gov

TABLE OF CONTENTS.

EXECUTIVE SUMMARY.................................................................................................................................... 4

II.

ABBREVIATIONS ......................................................................................................................................... 7

III.

INTRODUCTION .......................................................................................................................................... 8

IV.

IN VIVO STUDIES....................................................................................................................................... 11

SEARCH STRATEGY ............................................................................................................................................11

SUMMARY AND LIMITATIONS OF IN VIVO STUDIES...................................................................................................14

National Toxicology Program RF Reports 2018.......................................................................................14

Papers Meeting the In vivo Review Criteria. ...........................................................................................18

CONCLUSIONS FROM IN VIVO REVIEW. .................................................................................................................32

EPIDEMIOLOGICAL STUDIES...................................................................................................................... 35

SEARCH STRATEGY............................................................................................................................................35

SUMMARY OF RESULTS. .....................................................................................................................................36

Studies reporting overall evidence that does not counter a global null hypothesis on tumor risk in

relation to cell phones...................................................................................................................................................36

Studies Reporting Cvidence on Possible Tumor Risk Modification in Relation to Tumor Types and/or

Cell Phone Usage Practices. ..........................................................................................................................................37

Outcome of Interest – Brain Tumors (Glioma and Meningioma)............................................................38

Outcome of Interest – Acoustic Neuroma, or Vestibular Schwannoma. .................................................39

Outcome of Interest – Parotid Gland Tumors. ........................................................................................39

Outcome of Interest – Skin Cancers. .......................................................................................................40

Outcome of Interest – Leukemia. ............................................................................................................40

Putative Effect Modifier – Laterality of Tumor in Relation to Laterality of Phone Use. ..........................40

Putative Effect Modifiers – Location and Size of Tumors in Relation to Mobile Phone Use....................41

10.

Evidence on Tumor Risk in Relation to Cell Phones from Studies Using RF Exposure Estimates Other

Than Phone Usage Practices.........................................................................................................................................42

OVERALL LIMITATIONS OF EPIDEMIOLOGY DATA.....................................................................................................44

Data Quality Affected by Limitations of RF Exposure Measurements.....................................................44

Strong Biases Affecting Data Analysis and Interpretation. .....................................................................44

Lack of Evidence Based on the U.S. Population.......................................................................................45

No Overall Risk Increase, with Some Evidence Pointing at Possible Subgroup Effects. ...........................46

SUMMARY AND LIMITATIONS OF INDIVIDUAL EPIDEMIOLOGY STUDIES REVIEWED..........................................................48

CONCLUSIONS FROM REVIEWED EPIDEMIOLOGICAL STUDIES.....................................................................................87

VI.

REFERENCES ............................................................................................................................................. 88

IN VIVO ARTICLES, INITIAL SEARCH.......................................................................................................................88

IN VIVO ARTICLES INCLUDED IN EMF PORTAL SEARCH. ............................................................................................96

IN VIVO ARTICLES FOUND BY GENOTOXICOLOGY RELEVANT SEARCH. ..........................................................................96

IN VIVO ARTICLES FOUND IN NULL RESULT SEARCH. ................................................................................................98

IN VIVO ARTICLES SUGGESTED BY PEER REVIEWER. .................................................................................................98

IN VIVO ARTICLES MEETING SEARCH CRITERIA......................................................................................................100

Page 2

www.fda.gov

IN VIVO ARTICLES NOT MEETING SEARCH CRITERIA...............................................................................................104

REFERENCES USED FOR EPIDEMIOLOGICAL REVIEW. ..............................................................................................106

I. OTHER EXTERNAL (NON-FDA) REVIEWS. ................................................................................................................112

Page 3

www.fda.gov

EXECUTIVE SUMMARY

As part of our core mission to protect and promote the public health, the FDA continually

monitors new scientific evidence that might impact our understanding of the safety profile of

medical devices and radiation-emitting electronic products that Americans are exposed to every

day. The Agency has taken a comprehensive approach to evaluating the available scientific

evidence regarding the impact of radiofrequency radiation (RFR) exposure on human health.

The FDA has completed an updated radiofrequency (RF) exposure risk analysis based on

relevant peer-reviewed in vivo (animal) and epidemiological studies published from January 1,

2008 to August 1, 2018 for in vivo studies, and from January 1, 2008 to May 8, 2018 for

epidemiological studies. This risk analysis was scoped to assess any possible causal relationship

between RFR exposure and the formation of tumors.

In this technical report we provide a detailed summary of the substantial body of scientific

evidence that has informed our determination regarding potential adverse health effects in

humans caused by RFR exposures and the risk analysis we performed. As discussed in this report,

this determination is supported by the considerable body of peer-reviewed scientific publications

as well as public registries of, for example, cancer rates that show a slight decrease in brain

tumors despite the enormous increase in cell phone use over the last two decades. This technical

report summarizes the information on which the FDA has based its conclusion and was peerreviewed by scientists specialized in areas such as epidemiology, health physics and public health,

and the report incorporates comments by these external scientific reviewers. Subsequently, the

report and our analysis were updated with more recent, relevant peer-reviewed publications

through August 2019.

In the period covered by this risk analysis, there have been approximately 125 articles that

are most relevant for the study 1 of any effects of RFR on animals. However, none have adequately

demonstrated that localized exposure of RFR at levels that would be encountered by cell phone

users can lead to adverse effects. Section IV of the report provides a detailed description of the in

vivo studies.

In vivo studies assessing possible adverse or other effects of RFR are extremely challenging

studies to design and undertake for several reasons, including the engineering considerations of

applying an RFR field to animals that may specifically simulate, for example, the localized

Note that in this technical report, the terms study, experiment, investigation, and similar terms are used

interchangeably.

Page 4

www.fda.gov

exposure of tissue to a cell phone held to the ear. Many researchers therefore undertake wholebody exposure of the animals to RFR exposure to avoid the need to restrain the animals during

each exposure. However, the effects of whole-body exposure do not reflect the real-world

situation of localized exposure to the ear and head from a handset as used by humans.

It is difficult to measure and model with adequate accuracy and precision the specific dose

of RFR that each animal absorbs in each study. Additionally, it is difficult to separate the effects of

direct RFR exposure, if there are any that occurred, from the well-documented indirect effects of

temperature rise (the only established biological effect of RFR on tissue) and the stress

encountered by experimental animals, which confounds reported outcomes.

Given the difficulties of conducting in vivo studies on the effects of RFR exposure

experienced by humans as described above and the widespread use of cell phones, strong

epidemiological studies generally provide more relevant and accurate information. In vivo studies

are of immense value in medical science, but they are less useful than studying effects on the

human population, where that is feasible.

Based on the FDA’s ongoing evaluation, the available epidemiological and cancer incidence

data continues to support the Agency’s determination that there are no quantifiable adverse

health effects in humans caused by exposures at or under the current cell phone exposure limits.

In the last decade, there have been approximately 70 relevant epidemiological studies that

have been published as peer-reviewed scientific evidence. As part of our ongoing monitoring

activities, we have analyzed these publications for specific outcomes, including brain and other

tumors as well as other potential adverse events. While some studies suggest a possible link

between, for example, “heavy” users of cell phones and some tumors, there is no clear and

consistent pattern that has emerged from these studies and these studies were subject to flaws

and inaccuracies.

Epidemiological studies can also be subject to significant limitations; for instance, exposure

assessment is understandably limited and potentially problematic. Specifically, cell phone usage is

not necessarily an appropriate proxy for total RF exposure compared with, for example, the ability

to capture total cumulative RF exposure from all relevant sources, including for example other

communications devices. There are no direct measurements of RF exposure outside of a

laboratory setting; rather, studies generally rely on the participants to track and self-report.

Therefore, the actual RFR exposure remains an estimate at best. Based on the available data,

there is no clear evidence as to whether frequency, duration, or intensity of exposure is the most

important factor in assessing exposure and, as discussed above, there is no clear way to measure

or estimate each of these factors. Additionally, there are generally various biases (e.g., recall

Page 5

www.fda.gov

accuracy) and misclassification issues when individuals are asked to recollect behavior patterns.

See Section V of the report for a summary of the epidemiological studies.

Based on the studies that are described in detail in this report, there is insufficient evidence

to support a causal association between RFR exposure and tumorigenesis. There is a lack of clear

dose response relationship, a lack of consistent findings or specificity, and a lack of biological

mechanistic plausibility.

Page 6

www.fda.gov

II.

ASP

APC

APD

CDMA

CNS

DCS

DMBA

EMF

FDA

GFAP

GSM

IARC

ICNIRP

IEI- EMF

IRR

NAS

NeuN

NTP

RFR

ROS

SAR

SCENHIR

SEER

SSM

TCSE

UMTS

USC

WCDMA

ABBREVIATIONS

Annual Summarized Power

Annual Percentage Change

Annual Power Density

Base Station

Code-Division Multiple Access

Central Nervous System

Digital Cellular System

12-Dimethylbenz(a)anthracene

Electromagnetic Field

US Food and Drug Administration

Glial Fibrillary Acidic Protein

Global System for Mobile Communications

Hazard Ratio

International Agency for Research on Cancer

International Commission on Non-Ionizing Radiation Protection

Idiopathic Environmental Intolerance attributed to Electromagnetic Fields

Incidence Rate Ratio

National Academy of Sciences

Neuron-Specific Nuclear Protein

National Toxicology Program

Odds Ratio

Radiofrequency

Radiofrequency Radiation

Reactive Oxygen Species

Relative Risk

Specific Absorption Rate

Scientific Committee on Emerging and Newly Identified Health Risks

Surveillance Epidemiology, and End Results

Swedish Radiation Safety Authority

Total Cumulative Specific Energy

Universal mobile telecommunications system

United States Code

Whole Body

Wideband Code Division Multiple Access

Page 7

www.fda.gov

III.

INTRODUCTION

The FDA is responsible for, among other things, ensuring the safety of electronic products

that emit radiation, such as televisions and cell phones. These types of products are part of

Americans’ daily life, and we take our duty to protect consumers with the utmost gravity.

The FDA is required under 21 USC § 360ii(b)(1) to collect and make available, through

publications and other appropriate means, the results of, and other information concerning,

research and studies relating to the nature and extent of the hazards and control of electronic

product radiation; and make such recommendations relating to such hazards and control, as

appropriate. The FDA continually monitors new information and, when appropriate, undertakes

formal risk analysis.

The FDA’s doctors, scientists, and engineers have reviewed many sources of scientific and

medical evidence related to the possibility of adverse health effects from RFR exposure in both

humans and animals. We have relied extensively on a myriad of scientific evidence developed

over many years to help inform our regulatory thinking regarding RFR exposure. Standard practice

in scientific evaluation is to use the broadest set of credible information available and then to

assess the significance of that information to the question at hand. The most commonly used

source of information is the set of peer-reviewed publications that are indexed through Medline

and typically retrieved through PubMed. The FDA uses this source as well as more specific sources

of information, where appropriate. For the ongoing monitoring of possible effects of RFR, for

example, we also use the Electromagnetic Frequency (EMF) Portal 2 as a potential source for peerreviewed papers to ensure as wide a coverage as possible.

The FDA also considers independent studies that are separately published, although the

Agency often undertakes its own review of the papers analyzed in those reports in order to assess

validity and applicability to the United States. Recent examples of independent studies that the

FDA considered include the 2013 International Agency for Research on Cancer (IARC) study3, the

European Commission’s Scientific Committee on Emerging and Newly Identified Health Risks 4, the

EMF Portal (https://www.emf-portal.org/en/article/overview/mobile-communications-med-bio/g-85/t85002?pageSize=50&pageIndex=0&view=list#level-4).

IARC (2013). “Non-Ionizing Radiation, Part 2: Radiofrequency Electromagnetic Fields” Monograph 102.

Scientific Committee on Emerging and Newly Identified Health Risks (SCENIHR). 2015. 'Potential health

effects of exposure to electromagnetic fields (EMF) ', European Commission Report.

Page 8

www.fda.gov

Swedish Radiation Safety Authority’s (SSM) Scientific Council reports on EMF 5, and the 2017

reports from the National Toxicology Program studies on the effects of whole body irradiation of

rodents6, as well as other sources as discussed in this report.

In this peer-reviewed technical report, we are providing a detailed summary of our updated

RF exposure risk analysis based on relevant and available peer reviewed scientific information

published from January 1, 2008 to August 1, 2018 for in vivo studies, and from January 1, 2008 to

May 8, 2018 for epidemiological studies. This technical report is peer-reviewed by scientists

specialized in areas such as epidemiology, health physics and public health, and the report

incorporates comments by these external scientific reviewers.

The conclusions presented in this report are the result of a comprehensive analysis of

information published within the following scope:

The main health outcome focused on relates to the onset of cancer formation, known

as tumorigenesis.

We reviewed articles on RF exposure in general and did not limit the analysis to any

specific power levels or modulations used by cell phones.

The frequency range considered was 100 KHz to 6 GHz 7.

These articles were indexed on Medline and typically retrieved through PubMed. The

EMF portal (https://www.emf-portal.org/en) was also used to ensure as wide a coverage

of the search as possible.

Only articles that had been peer-reviewed were included, other than the final NTP in

vivo reports which were published in November 2018, and are available from

ii.

iii.

iv.

https://ntp.niehs.nih.gov/

vi.

The literature review was conducted using articles published from January 1, 2008 to

August 1, 2018 for in vivo studies, and from January 1, 2008 to May 8, 2018 for

See e.g., Swedish Radiation Safety Authority (SSM)’s 2018. "Recent Research on EMF and Health Risk

Twelfth report from SSM’s Scientific Council on Electromagnetic Fields, 2017." In.: Swedish Radiation Safety

Authority (SSM); Swedish Radiation Safety Authority (SSM)’s SSM's Independent Expert Group on

Electromagnetic Fields. 2010. "Recent Research on EMF and Health Risk Seventh annual report from SSM:s

Independent Expert Group on Electromagnetic Fields, 2010."

National Toxicology Program (2018) Technical Report TR-595. 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; and National Toxicology Program (2018) Technical Report

TR-596. 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.

This is the frequency range in which limits are based on specific absorption rate (SAR). Above 6

GHz power density is the appropriate metric for determining safe exposure limits.

Page 9

www.fda.gov

vii.

epidemiological studies. These time periods overlap with the timeframe from the prior

review on RF research needs undertaken by the National Academy of Sciences

(available from https://www.nationalacademies.org/) and published in 2008.

In vitro studies were excluded as the analysis was focused on assessing evidence for

tumorigenesis in vivo.

There were two main foci for the review: i) epidemiological evidence for the existence of

any tumor risk from cell phone usage, and ii) in vivo (animal) studies assessing any causality of

tumorigenesis from of RFR exposure.

The following sections of this technical report provide a summary of the currently

available scientific evidence related to RF exposure safety and any causal relationship

between such exposure and the formation of tumors that might impact the previous FDA

findings.

Page 10

www.fda.gov

IV.

In vivo Studies

This section provides a summary of the in vivo studies reviewed. Section IV.A. describes the

search strategy and search terms we used to find relevant in vivo studies. Section IV.B. provides a

summary of the studies reviewed along with a description of the limitations. In Section IV.C, the

FDA’s conclusions based on the reviewed and assessed in vivo studies is provided.

A. Search Strategy

The literature search to identify all relevant in vivo studies was conducted in 3 parts. The

goal of the review was to consider all evidence from in vivo animal exposure to RF from peerreviewed papers which had endpoints related to tumorigenesis and involved a broad set of

search terms as described below. The date range used for in vivo studies was January 1, 2008

through August 1, 2018. A wide search was initially conducted, which identified only a few

relevant articles. The criteria were then expanded to include in vivo genotoxicity studies

where the protocols involved exposure of animals to RF. Finally, a search identifying null

results was undertaken to identify studies involving tumorigenesis, but which failed to find

evidence for any link. This review is focussed on human cancer risk and consequently does not

directly investigate RF mechanistic studies (in vitro) or review articles that pertain to in vitro

RF exposure. The pool of potential papers was identified using the following search on

PubMed:

("radio waves"[MeSH Terms] OR ("radio"[ALL FIELDS] AND "waves"[ALL FIELDS]) OR "radio waves"[ALL

FIELDS] OR "cell phone?"[ALL FIELDS] OR “mobile phone”[ALL FIELDS] OR "Radio Waves/adverse

effects"[MeSH] OR "Electromagnetic Fields/adverse effects"[MAJR] OR "Electromagnetic Fields"[MeSH] OR

“radiofrequency"[ALL FIELDS] AND (cancer[sb] OR "Neoplasms"[MeSH] OR “carcinogenicity”[ALL FIELDS])

AND ("Rodentia"[MeSH] OR "Lagomorpha"[MeSH] OR rat? OR mouse OR mice OR rabbit?) AND

English[lang]) NOT ("Humans"[MeSH] OR "MRI"[ALL FIELDS] OR "ablation"[ALL FIELDS] OR "Ablation

Techniques"[MeSH] OR "Imaging"[ALL FIELDS] OR "Therapy"[ALL FIELDS] OR "ELF"[ALL FIELDS] OR

"Synthesis"[ALL FIELDS] OR "cognitive"[ALL FIELDS] OR "60 Hz"[ALL FIELDS] OR "50 Hz"[ALL FIELDS] OR

"hyperthermia"[ALL FIELDS] OR "healing"[ALL FIELDS] OR "adaptive"[ALL FIELDS] OR "In vitro"[ALL FIELDS]

OR "extraction"[ALL FIELDS] OR "microwave-assisted"[ALL FIELDS] OR "paramagnetic"[ALL FIELDS] OR

"Expression of "[ALL FIELDS]) AND ("2008/01/01"[Date - Publication] : "2018/08/01"[Date - Publication])

This initial search found 93 papers for consideration. The parameters after the ‘NOT’

term removed many, but not all, results that were outside the scope of the criteria for

inclusion. The 93 articles were evaluated according to the search criteria. Articles not included

in this review included many that were not assessing evidence of an adverse health effect

Page 11

www.fda.gov

from RF exposure, many others had endpoints that were unrelated to tumorigenesis, and

smaller sets of papers that either did not include exposure of live animals or in which the

exposures were to extremely low frequency electromagnetic radiation rather than to RF.

The search results were compared to the bibliography of the International Agency for

Research on Cancer’s (IARC) Monograph 102’s, Non-ionizing radiation, Part 2: Radiofrequency

electromagnetic fields, animal experiments chapter (IARC 2013: 281 - 83). There were 4

papers in that bibliography that were not found by the search, but they were all published

before 2008. Similarly, the search results were compared with the articles in an EMF portal

search 8 for experimental studies relevant to RF and cancer and found 3 papers published after

2008 that were missed by the search. The 3 EMF portal papers were added to the pool of

papers for consideration.

Search terms for the two supplementary searches are below:

Genotoxicity Search, which added 20 papers.

(("radio waves"[MeSH Terms] OR ("radio"[All Fields] AND "waves"[All Fields]) OR "radio

waves"[All Fields] OR "cell phone?"[all fields] OR “mobile phone”[all fields] OR "Radio Waves/adverse

effects"[Mesh] OR "Electromagnetic Fields/adverse effects"[MAJR] or "Electromagnetic Fields"[Mesh]

or “radiofrequency”[all fields]) AND (micronucleus or comet or genotox*) AND ("Rodentia"[Mesh] OR

"Lagomorpha"[Mesh] or rat? or mouse or mice or rabbit?) AND English[lang]) NOT ("Humans"[Mesh]

OR "MRI"[ALL Fields] OR "ablation"[ALL FIELDS] OR "Ablation Techniques"[Mesh] OR "Imaging"[ALL

FIELDS] OR "Therapy"[ALL FIELDS] OR "Synthesis"[ALL FIELDS] OR "cognitive"[ALL FIELDS] OR "60

Hz"[All Fields] or "50 Hz"[All Fields] or "ELF"[All Fields] OR "hyperthermia"[all fields] OR "healing"[All

Fields] or "adaptive"[all fields] or "In vitro"[all fields] or "extraction"[all fields] or "microwaveassisted"[all fields] or "paramagnetic"[all fields] or "Expression of "[all fields]) AND ("2008/01/01"[Date

- Publication] : "2018/08/01"[Date - Publication])

Studies null (negative) results, which added 6 papers.

(("radio waves"[MeSH Terms] OR ("radio"[All Fields] AND "waves"[All Fields]) OR "radio

waves"[All Fields] OR "cell phone?"[all fields] OR “mobile phone”[all fields] OR "Radio Waves/adverse

effects"[Mesh] OR "Electromagnetic Fields/adverse effects"[MAJR] or "Electromagnetic Fields"[Mesh]

or “radiofrequency”[all fields]) AND (“no evidence”[ALL Fields] or “does not alter”[ALL Fields] or “not

provide any evidence”[all fields] or ((“not” OR “no”) NEAR “effect”) or ((“not” OR “no”) NEAR

“evidence”) or “not promote”[ALL FIELDS] or (“not” NEAR “cause”)) AND ("Rodentia"[Mesh] OR

"Lagomorpha"[Mesh] or rat? or mouse or mice or rabbit?) AND English[lang]) NOT ("Humans"[Mesh]

OR "MRI"[ALL Fields] OR "ablation"[ALL FIELDS] OR "Ablation Techniques"[Mesh] OR "Imaging"[ALL

https://www.emf-portal.org/en

Page 12

www.fda.gov

FIELDS] OR "Therapy"[ALL FIELDS] OR "Synthesis"[ALL FIELDS] OR "cognitive"[ALL FIELDS] OR "60

Hz"[All Fields] or "50 Hz"[All Fields] or "ELF"[All Fields] OR "hyperthermia"[all fields] OR "healing"[All

- Publication] : "2018/12/31"[Date - Publication])

The list was compared with bibliographies from recently published expert reports to

ensure a thorough assessment. The expert reports used for bibliographic comparison included

the European Scientific Committee on Emerging and Newly Identified Health Risks report,

Potential health effects of exposure to electromagnetic fields (EMF)(SCENIHR 2015), the

Swedish Radiation Safety Authority (SSM)’s Scientific Council on Electromagnetic Fields’ 7th

(2010) through 12th (2018) reports (SSM's SCEF 2014, 2010, 2013, 2015, 2016, 2018), and the

series of reports from the Health Council of the Netherlands (Health Council of the

Netherlands 2013, 2014, 2016). No other animal experiment papers that might fit the criteria

from the January 1, 2008 through August 1, 2018 timeframe were identified.

The pool of papers identified via the above process was then narrowed down to papers

that considered experiments on live mammals (rats, mice, or rabbits) with endpoints related

to cancer, carcinogenesis, tumorigenesis, oncogenesis, or genotoxicity. This reduced the list to

26 papers that were reviewed. After an initial review, 3 papers were found to be outside the

report scope and were eliminated from further analysis (one had been retracted, one was an

exposure of excised bones, and the third involved exposure of animals to pulsed magnetic

fields not RF).

A peer reviewer identified 31 papers more papers as possibly relevant. About half of

these papers fell outside our inclusion criteria. There were a few papers regarding RFR

exposure of non-mammal animals. A few papers used frequencies above 6 GHz. A couple of

papers had endpoints outside of our scope. The other papers that were excluded used cell

phones as the source of RFR which renders these studies results uninterpretable. Dosimetry

and modeling to calculate the SAR of the exposed animas animals is difficult with a defined

RFR source. A normal cell phone is not a reproducible source of RFR.

A review of the 39 papers that met the inclusion criteria is described below. Additionally,

the final National Toxicology Program reports on exposure of rats and mice to radiofrequency

radiation and an associated paper, Effect of cell phone radiofrequency radiation on body

temperature in rodents: Pilot studies of the National Toxicology Program's reverberation

chamber exposure system (Wyde et al. 2018) were also reviewed. A complete list of the

papers found by the searches for this section appears in Section VI References.

Page 13

www.fda.gov

B. Summary and Limitations of In vivo Studies.

1. National Toxicology Program RF Reports 2018.

The National Toxicology Program released reports on whole-body exposures of rats and

mice to radiofrequency radiation that was modulated using the signal patterns commonly used by

third generation wireless mobile telecommunications technology (3G). The NTP study is the most

comprehensive study of the impact of RFR on tumorigenesis to date. While the FDA does not

agree with all the conclusions of the NTP RFR study, we do acknowledge that the study has added

to the collective database and provided possible pathways for further investigation. The effects of

a whole-body RFR exposure are not predictive of the results for a local RFR exposure in the same

animal. Thus, results of whole-body RFR exposure to rats and mice cannot be directly related to

the results of the local RFR exposure a human receives while using a cell phone.

There were three components to the NTP study. The first component was a 5-day thermal

pilot study at specific absorption rates (SARs) of 4 – 12 W/kg (Wyde et al. 2018) which was

published separately from the remaining portions of the study covered in NTP’s reports. The

second component was a 28-day pre-chronic toxicology study in mice and rats. The last

component was a 2-year toxicology and carcinogenicity study in mice and rats9. The exposure

system used in all three components of this study was a specially engineered reverberation

chamber designed with the goal of providing nearly uniform whole-animal exposure without the

need for stress inducing animal restraint systems (Capstick et al. 2017). The exposed rats and

mice were exposed to either global system mobile communication (GSM) or code division

multiple access (CDMA) modulated signals for 18 hours and 20 minutes / day in 10 minutes on

and off cycles. The rats were exposed starting in utero to GSM- or CDMA-modulated signals at 900

MHz. The mice were exposed to either GSM- or CDMA-modulated signals at 1900 MHz. The

results discussed here are from Wyde et al. 2018 and the NTP technical reports released to the

public in November 2018.

The 5-day pilot study examined the impact of animal size on the subcutaneous

temperatures in mice and rats, aged mice and rats, as well as pregnant rats. The aged rats in the

NTP report on Rats https://ntp.niehs.nih.gov/ntp/htdocs/lt_rpts/tr595_508.pdf?utm_source=direct&utm_medium=prod&utm_cam

paign=ntpgolinks&utm_term=tr595 &

NTP report on Mice https://ntp.niehs.nih.gov/ntp/htdocs/lt_rpts/tr596_508.pdf?utm_source=direct&utm_medium=prod&utm_cam

paign=ntpgolinks&utm_term=tr596

Page 14

www.fda.gov

pilot study were at most 270 days old while rats in the 2-year study reached 763 days old. The

effect of animal age increasing from 270 days to 763 days was not investigated and could have

resulted in further age-related weakening of the rats’ thermoregulatory capacity. The

temperatures measured were subcutaneous and not a measure of the core temperature of the

animal. No temperature measurements were taken in the animals housed in RF environments

during the 2-year study. The pilot study confirms that aged animals or pregnant animals cannot

thermoregulate as well as young, non-pregnant animals. The study revealed that there were

some incidences of increased subcutaneous temperature (>1°C) in aged rats exposed to 6 W/kg

GSM radiation or 900 MHz CDMA RF radiation but not in young rats. Likewise, there were

sporadic temperature increases in young or old mice exposed up to 12 W/kg GSM and CDMA

1900 MHz RFR. Measurements were made when RFR was off, and as such, allowed for external

surface cooling which likely reduced the measured subcutaneous temperatures.

Additionally, NTP’s assessment of organ SAR showed that local SAR in skin and fat was less

than the whole-body and internal organ SAR which was higher. An animal’s core temperature

cannot be accurately estimated from subcutaneous temperature measurements. Considering the

information, the animals’ core temperature was likely to have exceeded the measured

subcutaneous temperature. It is probable that during the 2-year study, as the male rats aged their

core temperatures eventually routinely exceeded a 1 °C rise in the 6 W/kg SAR exposure group.

We believe that this study indicates a need for a greater understanding of the impact of cyclic

heating and declining thermoregulation capacities of aged mammals.

The NTP has established 5 levels of evidence of carcinogenic activity that are used in NTP

reports. The highest level is Clear Evidence. This is a dose-related (i) increase of malignant

neoplasms, (ii) an increase of a combination of malignant and benign neoplasms, or (iii) marked

increase of benign neoplasms if there is an indication from this or other studies of the ability of

such tumors to progress to malignancy. The next lower level of evidence is called Some Evidence.

For this level the test agent-related increased the incidence of neoplasms in which the strength of

the response is less than that required for clear evidence. Equivocal Evidence is the next lower

level. This is a marginal increase of neoplasms that may or may not be test agent-related. The two

lower levels are No Evidence and Inadequate Study.

The histopathological analysis of the RF-exposed mice was reported as Equivocal Evidence.

For GSM-modulated exposed male mice, the combined incidence of fibrosarcoma, sarcoma, or

malignant fibrous histiocytoma of the skin and the combined incidence of alveolar/bronchiolar

adenoma or carcinoma in the lung were described as marginally increased. For GSM-modulated

exposed female mice, there was a marginally increased incidence of malignant lymphoma. For

CDMA-modulated exposed male mice there was increased incidence of hepatoblastoma of the

liver. For CDMA-modulated exposed female mice malignant lymphoma incidence was marginally

Page 15

www.fda.gov

increased. All other tissues examined showed no test agent-related increase in neoplasms. The

results shown for some of these equivocal results are above the rate for the sham control, but

many are within the historical control range. This suggests that the equivocal results are

describing fluctuations in the natural background of these tumors and not an increase caused by

exposure to RF radiation.

Genetic toxicology testing was performed using the micronucleus assay and the comet assay

in the RF-exposed mice and rats as part of their 14-week intermediate analysis. There was an

increase in apparent DNA damage in male rats exposed to CDMA RFR, but the increase was not

statistically significant. Other noted increases in DNA damage were also considered to be

equivocal, and no statistically significant increases in DNA damage was observed in female rats or

in either mouse sex. All results from the micronucleus assay in rats and mice were negative.

The two-year carcinogenesis and toxicology study showed an unexpected finding in male

rats. The exposed groups of animals had greater survival rates than the animals in the sham

control. This might be attributed to amelioration of chronic progressive nephropathy: animal core

temperature elevation resulting in reduced food consumption and increased water consumption

in the RFR exposed animals could have improved the survival of the exposed animals. NTP’s

stated that their statistical analysis took these differences in survival rate into account during

study analysis. The public health impact of these observations is unclear.

NTP’s study conclusions in their final report for the two-year exposures in rats were:

Clear evidence of carcinogenic activity

Incidences of malignant schwannoma in the heart of male rats for both GSM and CDMA.

Some evidence of carcinogenic activity (considered to be related to cell phone RFR

exposure)

Incidences of malignant glioma in the brain of male rats for both GSM and CDMA

Incidences of pheochromocytoma (benign, malignant, or complex combined) in the adrenal

medulla of male rats for GSM

Equivocal evidence of carcinogenic activity (may have been related to cell phone RFR

exposure).

Incidences of adenoma or carcinoma (combined) in the prostate gland of male rats for GSM.

Incidences of benign or malignant granular cell tumors in the brain of male rats for GSM.

Incidences of adenoma in the pars distalis of the pituitary gland of male rats for GSM and CDMA

Incidences of pancreatic islet cell adenoma or carcinoma (combined) of male rats for GSM.

Page 16

www.fda.gov

Incidences of adenoma or carcinoma (combined) in the liver of male rats for CDMA.

Incidences of malignant schwannoma in the heart of female rats for GSM

Incidences of malignant glioma in the brain of female rats for CDMA

Incidences of malignant schwannoma in the heart of female rats for CDMA

Incidences of pheochromocytoma (benign, malignant, or complex combined) in the adrenal

medulla of female rats for CDMA.

NTP reported that two-year exposure to GSM- or CDMA-modulated RFR revealed an

increase in malignant schwannoma in in the heart of male rats. NTP also reported that the

incidence of schwannomas in all locations was not statistically significant compared to the control

group. Cardiomyopathy was also seen in rats exposed to higher SARs (3 or 6 W/kg). The historical

controls reported in NTP’s preliminary reports published in 2016 (tumor incidence over time in a

larger population of similar control rats in NTPs experiments) had approximately 9 out of 699

control rats that had malignant heart schwannomas (1.3% and a range of 0-6%). Only male rats

exposed to higher SAR (6 W/kg) have higher percentage of malignant schwannoma then the

maximum range of historical controls. However, NTP chose to use only three studies for historical

controls in the final report. This set of studies had control groups of 50 rats each. There was a 2%

incidence of cardiac schwannoma in 2 of the studies and no occurrences in the third study. NTP

also determined (Wyde et al. 2018) that old, large male rats did not thermoregulate as well as

younger rats. This is a possible indication of thermal confounding at 6.0 W/kg. It is not clear if

there is a relationship between malignant schwannoma and cardiomyopathy.

NTP reported weaker responses to both modulations were found in other tissues. Glioma

was one of the weaker responses and was not statically significant. The sham controls for gliomas

show 0% incidence while the tumor incidence over time in a larger population of similar control

rats in NTP experiments has a percent incidence of approximately 2-4%. Only the lowest RF

exposure, 1.5 W/kg CDMA-modulation exceeds this historical control. Other tumors with

increased incidence included pituitary tumors, benign and malignant pheochromocytoma,

adenoma and carcinoma of the prostate. However, the highest increases were not consistently in

the highest SARs and in some cases (adrenal medulla) the tumor incidence over time in control

animals from other NTP experiments was as high as 28%.

The specific criteria for the categorization of data into the specific outcomes (e.g., clear

evidence, etc.) remains unclear and there are inconsistencies in some of the interpretation of the

data and the conclusions that have been drawn. As an example, the incidence of adenoma or

carcinoma of the prostate gland in male rats subjected to GSM modulated RF was characterized

as showing equivocal evidence of carcinogenicity with an incidence of 2.2%, 2.2%, 7.8% and 3.3%

for the four doses of RF, whereas fibromas of the skin which showed incidences that are higher

than the above (2.2%, 5.6%, 7.8% and 5.6%) were not classified as equivocal or any level above

that.

Page 17

www.fda.gov

In addition to the inconsistencies in the data interpretation, the importance attributed to

dose response was also unclear in the final conclusions of the NTP reports. The incidence of

benign, complex, or malignant pheochromocytoma in the adrenal medulla in male rats subjected

to GSM modulated RF (12.5%, 26.7%, 31.5%, 16.1%) were classified as some evidence of

carcinogenicity despite there being no clear dose response and no statistical difference between

the highest RF dose and the control animals (P=0.472).

The strongest evidence found in the study was for malignant schwannomas of the heart in

male rats, which were characterized originally as some evidence and after peer review changed to

clear evidence and showed statistical significance at some doses of RF. It is notable that there was

no incidence of heart schwannoma in the control group. An NIH reviewer noted that if there had

been even one case in the control group then the statistical significance of this finding would not

exist. It has also been noted that the histology review of samples was only partially blind in that

the pathologists knew which group of samples was from the control group. This result was not

consistent between the male rats and female rats or either sex of mice. As has been noted before,

these extremely large whole-body RFR exposures (6 W/kg is 75 times higher than the limit for the

public which is 0.08 W/kg) in rodents cannot be directly compared to real world partial body

exposures to people using mobile phones 10.

We do not know if there is a causal effect or if these results are due to weakening of the

immune response due to animal stress from cyclic heating and thermoregulation decline in aging

animals leading to whole-body temperature increase, possible sleep disruption due to the cyclic

heating, or due to an RF-specific effect that has not been identified and has an adverse effect

before heating becomes the dominant safety issue. It is possible that any form (ambient, IR,

ultrasound) of cyclic whole-body heating of this magnitude may cause similar findings, but no

such studies have been conducted to date.

2. Papers Meeting the In vivo Review Criteria.

In this section we summarize other relevant in vivo studies that we carefully considered

and reviewed. Below is a summary of the results and limitations of these studies.

The largest animal exposure study published to date was conducted at the Ramazzini

Institute’s Cesare Maltoni Cancer Research Center in Italy (Falcioni et al. 2018). The goal of the

Note that the terms “wireless phone,” “cell phone,” and “mobile phone” are used interchangeably in

this and other studies.

Page 18

www.fda.gov

study was to examine the carcinogenic effect of 1.8 GHz simulated GSM in the far field whole

body exposure for 19 hours/day for life on 2,448 Sprague Dawley rats. The size of the stacked

dipole array antenna was not given but based on FDA expert analysis of the picture provided, the

exposure may have been in the near-field. Exposure was based on fixed field strengths (V/m) and

SAR (dose rate) estimates were based on a fixed “coupling factor”. However, for a given electric

field SAR changes with the change of mass, which will change over the animal’s lifetime. As such,

the estimation of the dose using a fixed “coupling factor” was not accurate.

The exposure groups were reported to be 0 (0 W/kg), 5 (0.001 W/kg), 25 (0.03 W/kg) or 50

V/m (0.1 W/kg). The animals were irradiated in their home cages, initially 5 rats per cage, but

pictures show cages containing only one rat. The number of rats per cage is an important factor

for the evaluation of potential stress and the possibility of localized RF current flow when animals

touch. Because cages were exposed horizontally, rats could shield each other during exposure. As

rats die and fewer are left in a cage their potential exposure increases unless survivors are

rearranged to maintain full cages. Moreover, it was unclear if five animals per cage resulted in

conflict and fights between animals, hence inducing stress.

There was no indication that animal core or other temperature was measured or recorded.

Tissues from the heart and the brain were examined for malignancies and pre-malignancies.

There was no mention of any other tissue being assayed. The authors state that the expected

spontaneous tumor incidence and its fluctuations were based upon data derived from more than

20,000 historical controls; however, no reference to these data is provided. There were no

differences in food or water consumption, mean body weight or survival index among the four

exposure groups. This finding is different from the NTP rat exposure results where the water

consumption was not monitored, and RF-exposed rats lived longer than control rats.

Histopathological evaluation of the hearts and brains detected a statistically significant increase in

the tumor incidence of heart schwannoma in the highest exposure for male rats (1.4%). The

authors refer to historical controls with an incidence of schwannoma in male rats of 0.6% (10 out

of 3,160 rats). Again, these historical control data were not available for analysis, and

demonstrate very high inconsistencies with other reported historical rates of rodent

schwannomas.

No statistically significant increases in malignant brain tumors and pre-malignant brain

tumors were observed. The authors found a non-statistically significant exposure dependent

increase in the incidence of malignant glial tumors in exposed female rats. After comparing this

non-significant trend to historical control data where the amount of malignant glial tumors in

Page 19

www.fda.gov

female rats was lower: 15 glial cell tumors out of 3,165 rats, or 0.5% 11, it was determined that the

50 V/m exposure group had a greater incidence of glial cell tumors than the historical (but not the

experimental) control.

Bartsch et al. (2010) reported on two long term and two life-long experiments that

examined the effects of chronic exposure of rats to 900 MHz, pulsed at 217 Hz to stimulate a GSM

signal on unrestrained animals. The exposure system was far-field simulated GSM 900 MHz

exposure via a circularly polarized antenna. Fixed field strength, rather than SAR, was maintained

throughout the experiments and thus the effective SAR changes over time as animals age and

grow larger. The mean whole-body SAR was estimated to be 80 mW/kg by means of

computational modeling and simulations based on an approach that has been extensively used in

RF dosimetry literature. Twelve female rats per cage were exposed vertically so resonance effects

are more likely. The exposure was for 23 hours and 45 minutes/day. Radiation exposure started

at 52-70 days of age. The experiment was designed with animals irradiated in four different

experiments with different time lengths, I. 24 months, II. 17 months, III. 36 months and IV. 37

months. The experimental endpoints included: animal general health, necropsy and

histopathology but there was no mention of temperature monitoring, which could lead to

confounding results. Experiments I and II recorded adverse health effects. Experiments III and IV

showed, unlike the NTP study, that survival was decreased with RF exposure. It is unclear how the

sham controls were treated or what elements the sham setup duplicated of the real exposure.

Animal temperature measurement was not mentioned. The investigators concluded that chronic

exposure to RF may exert negative health effects and shorten survival if exposure is applied

sufficiently long. The investigators also noted that the month of birth may affect life span and

stated that an experiment should be run to determine if the 11-year sunspot cycle could affect a

rat’s lifespan.

A series of investigations from Korea examined the impact of radiofrequency exposure on

laboratory animals (Jin et al. 2011; Kim et al. 2008; Lee et al. 2011). Kim et al. (2008) exposed the

heads of mice to 849 or 1763 MHz RFR for up to 12 months in a carousel-type head exposure

system with a specifically-designed exposure chamber with real CDMA signal. The chamber was

made of aluminum, the external size of which is 745 mm × 1,060 mm × 1,980 mm. To reduce

reflections from the walls, absorber foam with a thickness of 4.5 inches was used. The SAR

measurement was simulated. The SAR estimate was validated against thermal probe temperature

measurements. The actual temperature within the animal was not validated. There was no animal

core or head temperature monitoring to evaluate thermal impact. There was also no mention of

blinding. The heads of the restrained mice were irradiated 1 hour/day, 5 days a week for 6 or 12

No reference was provided for these data

Page 20

www.fda.gov

months with SAR of 7.8 W/kg. The weights of the irradiated animals were not significantly

different from the controls. Furthermore, brain tissue taken from the animals showed no

difference in histology, cell proliferation rate, cell death by TUNEL assay, and in the expression

and distribution of NeuN and GFAP (markers of specific neuronal cell types) in the hippocampus

and cerebellum between exposed and control animals. No evidence of brain tumorigenesis

associated with RFR was found under microscopic examination. The authors concluded that

exposure to the RFR at two different wavelengths did not result in alterations in cell proliferation,

cell death, or gliosis.

Jin et al. (2011) whole-body irradiated rats in a reverberation chamber with two combined

RF modulations, but without any core body temperature monitoring. The first was code division

multiple access (CDMA), at 849 MHz and the second represents wideband code division multiple

access (WCDMA), at 1.95 GHz. The animals were irradiated for 45 minutes/day, 5 days a week for

one year. The SAR for these experiments was 2 W/kg for CDMA and 2 W/kg for WCDMA for a

total combined SAR equal to 4 W/kg. No core temperature measurements were made to assess

thermal effects. To prevent treatment bias, all the exposure and data analysis were performed

using blinded methods. No significant differences in body weight or tumor incidence between

exposed and control rats were found. There were some notable alterations in physiological

parameters (complete blood count and serum chemistry) among exposed animals. However, the

authors concluded that a 1-year simultaneous exposure to 4 W/kg CDMA and WCDMA did not

increase chronic illness.

Lee et al. (2011) exposed AKR/J mice to CDMA and WCDMA RFR for 45 min/day, 5

days/week for 42 weeks at a total SAR of 4 W/kg. Exposure was given by using a reverberation

chamber. SAR was calculated in rat voxel models by means of computational modeling and

simulations based on the Finite Difference Time Domain (FDTD). Only whole-body SAR values

were reported and no information on local SAR was included. Rectal temperature measurements

were made before and after exposure and there was no elevation of temperature. Notably, AKR/J

mice showed a high incidence of spontaneous lymphoma that was followed in these experiments.

After examining body weight, survival, lymphoma incidence and splenomegaly, no differences

were found between sham irradiated and irradiated animals. There were observed differences in

infiltration of the brain by lymphoma: RFR-exposed male mice had lower brain infiltration

compared with the controls and female mice had higher levels of brain infiltration compared with

the controls. However, the authors concluded that RFR exposure did not affect lymphoma

development in AKR/J mice.

Several investigations examined the impact of RFR exposure on the brain. Kesari, Behari,

and Kumar (2010) examined the mutagenic response of 2.45 GHz RFR exposure on the brains of

12 male Wistar rats. The animals were whole-body exposed to a horn-antenna for 2 hours per day

Page 21

www.fda.gov

for 35 days and sacrificed immediately after the last exposure and the brains of exposed and

control animals assayed for DNA strand breaks (comet assay), antioxidant enzyme activity

(glutathione peroxidase, catalase and superoxide dismutase) and histone kinase activity. The

whole-body SAR was reported to be 0.11 W/kg. They detected an increase in DNA strand breaks,

a decrease in glutathione peroxidase, superoxide dismutase and histone kinase activity and an

increase in catalase activity. They concluded that chronic exposure may cause significant damage

to the brain. Animal temperature measurement was not mentioned.

Chaturvedi et al., (2011) examined the impact of 2.45 GHz RFR exposure on circadian

organization, spatial memory, DNA damage in the brain and blood counts of Parks strain male

mice. The mice were exposed to RFR 2 hours per day for 30 days. RFR exposure was from an

analog signal generator and horn antenna. Five of the ten animals were exposed individually and

the whole-body SAR was 0.035 W/kg. The remaining 5 animals were sham exposed. Locomotor

activity via a running wheel was recorded prior to and during RFR exposure. The Morris water

maze was used from the 17th to the 22nd day of RFR exposure. After completion of the exposure

blood was collected for hematological parameters, brain tissue for DNA strand breaks (comet

assay), epididymis for sperm count and motility and serum for serum gluSGOT and SGPT. The

authors concluded that RFR caused an increase in leukocyte and erythrocyte counts as well as

altered the rate of DNA damage in exposed mice. Furthermore, the authors believe that RFR

exposure altered the circadian system.

Deshmukh et al. (2013) studied the impact of 900 MHz (SAR= 0.595 mW/kg), 1800 MHz

(SAR= 0.58 mW/kg) and 2450 MHz (SAR=0. 6.7 mW/kg) on the brain of rats. There were 24 rats

that were organized into 4 groups (6 rats each). Three groups were exposed, and one group was

sham exposed. The restrained rats were exposed for 2 hours/day, 5 days/week for 30 days. There

were no statements regarding measurement of the animals’ temperatures. Using the results from

their modified alkaline comet assay, the authors concluded that low SAR exposures with any of

the three frequencies may induce DNA damage.

Two other studies in rats from the same group evaluated the impact of RFR on cognitive

function, heat shock protein levels and DNA damage (comet assay) in the brains of Fischer rats

(Deshmukh et al., 2015, Deshmukh et. al., 2016). This group used a Gigahertz Transverse

Electromagnetic cell as their exposure system. There were 24 rats in each study that were

organized into 4 groups (6 rats each). In each study three groups were exposed, and one group

was sham exposed. Restrained animals were irradiated 2 hours / day 5 days / week for 90 days

(Deshmukh et. al., 2016) or 180 days (Deshmukh et al., 2015). The SARs were 0.5953, 0.5835 and

0.6672 mW/kg for exposures of 900, 1800 or 2450 MHz respectively. Body temperature of rats

was noted by rectal measurements immediately before and after the MWR exposure in all the

groups. There was no change in body temperature. For both exposures, (90 or 180 days) the

Page 22

www.fda.gov

authors found cognitive function decline, elevated HSP synthesis and DNA damage and they

concluded that chronic low-intensity microwave exposure may cause hazardous effects on the

brain. The authors state that HSP synthesis can be from non-thermal effects. However, thermal

effects even at these low SARs cannot be ruled out.

Sahin et al. (2016) evaluated the impact of 2100 MHz RFR on the brain of rats. The exposure

system was composed of a vector signal generator and horn antenna. There were 30 rats divided

into 4 groups. Exposure groups had 9 animals and sham exposure groups had 6 animals each. The

exposed animals were exposed for 6 hr/day 5 days a week for 10 or 40 days. The SAR was

reported to be 0.4 W/kg. No core temperature measurements were made to assess thermal

effects. The authors assessed oxidative DNA damage (8-hydroxy-2’deoxyguanosine

determination) and the level of lipid peroxidation (malondialdehyde determination) in the brain

of RFR exposed rats. The authors found no difference in oxidative DNA damage when exposed

animals were compared to their individual controls, however, increased oxidative DNA damage

was found in animals exposed for ten days compared to the 40-day exposed animals. Decreased

lipid peroxidation was observed after 40 days of exposure. The authors postulated that the longer

exposure may have increased DNA repair mechanisms.

Ibitayo et al., (2017) investigated the effect of RFR on the brains of male rats. There were 16

rats in 4 groups of 4. The authors described the exposure set up as follows: “The animals were

pair housed in groups of four in steel mesh cages and placed about 10 cm away from installed

devices that emitted RF EMF. The control group (𝑛𝑛 = 4) was kept in a steel mesh cages with a glass

barrier separating the control from the experimental animals.” It appears the animals were whole

body exposed to 2.5 GHz RFR for 30, 45 or 60 days. No description of the RF source nor a SAR

determination was provided. No core temperature measurements were made to assess thermal

effects. Histopathology and DNA damage was determined. The authors concluded that vascular

congestion and DNA damage indicative of oxidative stress are evident in RFR exposed animal

brains. The authors found that the 45-day exposed samples did not show severe histopathological

changes seen in the 30- or 60-day exposed samples. There was also no report of malignant

change in the investigation.

Three investigations examined the impact of 1800 or 2100 MHz on bladder cells of rats

(Gurbuz et al. 2010; Gurbuz et al. 2014; Gurbuz et al. 2015). In these experiments, the rats were

exposed to RF radiation and the bladder cell micronuclei were assayed for micronuclei in vitro.

Gurbuz, et. al. (2010) exposed 6 female Wistar rats to 1800 MHz GSM-like RF for 20 minutes/day,

5 days/week for a month and assigned 6 others to a control group. Near field exposure was used

to simulate cell phone exposure. Field strengths were controlled with a Narda EMR 300 meter and

probe. The E-field was 4 V/m. The authors claim that because E-field values are much lower than

the 41 V/m of ICNIRP, exposure is assumed to be non-thermal. However, temperature was not

Page 23

www.fda.gov

monitored, and those statements cannot be confirmed. RFR exposure did not increase the

frequency of micronuclei in exposed rats compared with the controls.

Gurbuz et al (2014) examined exfoliated rat bladder cells for micronuclei after whole body

exposure (30 min/day, 6 days/week for one and two months) to 1800 MHz and 2100 MHz RF

radiation. The animals were free to move in their cages during the RF exposure. The control

animals were housed in their home cages during the entire experimental period without being

subjected to any experimental manipulation. A vector signal generator was used to create a GSM

modulated RF radiation in the experimental setup. The E-field was 17 V/m. The calculated

average SAR was estimated to be 0.23 W/kg. No significant increase in micronuclei found in the

exfoliated rat bladder cells was detected in cells from RF exposed rats compared to control rats.

Finally in a third study, the same group (Gurbuz et al. 2015) examined the influence of RFR

on micronucleus formation in bladder cells of diabetic rats. They postulated that diabetes may

accelerate oxidative damage in DNA molecules through glucose auto-oxidation and nonenzymatic glycation. The diabetic rats were exposed to 2100MHz RF for 30 min/day, 5 days/week

for 30 days at a low SAR (0.24 W/kg). As in previous experiments, they used a near field exposure

and the dosimetry was unclear. Again, this group found no statistically significant difference

between the exposed population and the diabetes control or the non-diabetes control. Overall,

based on the results presented, the studies indicate that rat bladder cells are not sensitive to RF

exposure at low SARs.

Two separate experiments examined the co-carcinogenic potential of RF exposure in

combination with a known carcinogen (Lerchl et al. 2015; Tillmann et al. 2010). In both cases the

known carcinogen was ethylnitrosourea. Tillmann et al. (2010) exposed mice over the whole-body

to a universal mobile telecommunications system (UMTS) signal. RF exposure lasted up to two

years starting at embryo-fetal stage. Whole body SARs were estimated to be up to 4 W/kg. The

animals were exposed once per day (2 GHz, continuous wave signal, 2 h/d) with a stepwise

increased power density in one radial waveguide of a three-decker exposure system. Rectal

temperature measurements simultaneously obtained in 10 mice (2 mice per cage) directly after

termination of the RF exposure revealed regular body temperatures (37.1°±0.38°C to

37.7°±0.68°C) regardless of the RF power density applied. The low field intensity group was also

given 40 mg ethylnitrosourea. The high UMTS exposure (48 W/m2), the sham exposure and the

cage control all showed comparable tumor incidence. Animals exposed to ethylnitrosourea and

4.8 W/m2 RF showed an increased incidence of lung carcinoma. Notably, the authors identified a

Helicobacter hepaticus infection in these animals and discounted the high RFR liver tumor results.

A review of the neoplastic and pre-neoplastic lesions in the protocol organs shows that UMTS RF

exposure (48 W/m2) for up to two years did not produce tumors or other lesions of the cerebrum.

Also, other tissues examined such as the lungs, kidneys, spleen and hematopoietic /

Page 24

www.fda.gov

lymphopoietic system did not show increased tumors or increased pre-neoplastic lesions when

the UMTS exposed animals were compared to the cage control or sham control animals.

Lerchl et al. (2015) also examined the co-carcinogenic potential of RFR exposure in

combination with a known carcinogen (ethylnitrosourea). This investigation used 5 groups of 96

mice. Four groups were treated with ethylnitrosourea and exposed at SAR equal to 0 W/kg

(sham), 0.04, 0.4 or 2 W/kg SAR for 23.5 hours per day up to 72 weeks. The fifth group was an

untreated unexposed cage control. Details of the RFR exposure and exposure system were

previously published (Reinhardt et al. 2007), presenting the 900 MHz UMTS waveguide exposure

system and simulations to estimate whole body SAR based on typical arrangement of mice in the

exposure area. The E-field inside waveguides was monitored. There was no data, either simulated

or measured that correlated E-field with the desired treatment levels of whole-body SAR. There

was no temperature monitoring. Further, this large experiment did not include an RFR only

control. The authors concluded that the combination of ethylnitrosourea and RFR produces a

greater number of tumors in the lungs and livers of exposed animals compared to the sham

irradiated controls. Also, the results found lymphoma in RFR-exposed animals. The results do not

demonstrate a clear dose response. A review of the neoplastic and pre-neoplastic data shows no

significant tumors or pre-neoplastic lesions of the cerebrum compared to the cage control or the

sham irradiated control.

Hruby (Hruby et al. 2008) examined the impact of 902 MHz GSM-type wireless RF radiation

on 7, 12-Dimethylbenz(a)anthracene (DMBA)-induced mammary tumors in rats. DMBA is widely

used in animal research as an immunosuppressor and a powerful organ-specific carcinogen. All

animals received DMBA, there were two control groups (cage control, sham irradiation control)

and three RFR exposure groups (0.4, 1.3 and 4.0 W/kg). Each group had 100 animals and another

20 were kept as sentinel animals for disease surveillance. RF exposure was simulated GSM at 902

MHz for 4hr/day, 5 days/week for 6 months. The animals were exposed individually, held in

restraining tubes fitted into a 17 spoked waveguide exposure system. The authors cited

references on the exposure system which provide credible evidence that the system was

accurate, monitored in real-time, and well supported by simulations to estimate whole body SAR.

The authors found the following statistically significant results: All RFR exposed groups, at

different times had significantly more palpable masses; There were fewer animals with benign

lesions in the RFR exposed groups and more malignant tumors in the highest RFR exposure group;

The cage control group had more palpable tissue masses and more benign and malignant tumors

compared with the sham irradiated control. However, the RFR exposure groups did not show a

dose response. The authors concluded that, “In the context of the results of the cage-control

group, in the light of controversial results reported in the literature, and given the fact that the

DMBA-mammary tumor model is known to be prone to high variations in the results, it is the

authors’ opinion that the differences between the groups are rather incidental ones.”

Page 25

www.fda.gov

Paulraj and Behari (2011) investigated the impact of 2 frequencies of RFR on carcinogenesis

in mice. There were 18 mice in each of the 7 groups. The study was divided into two parts. The

first, involved the topical administration of DMBA on the skin followed by exposure to 112 MHz or

2.45 GHz 2 hours/day, 3 days / week for 16 weeks. There was insufficient information to

determine how SAR was estimated but reported SARs were 0.75W/kg for 112 MHz and 0.1 W/Kg

for 2.45 GHz. No core temperature measurements were made to assess thermal effects. The

study also employed a positive control, DMBA + croton oil (a well-known tumor promoter). The

positive control produced tumors in animal skin. However, no tumors were detected in the

controls or the exposed populations. The second part consisted of transplanting Ehrlich ascites

cells (originally established to produce ascites tumors in mice) intraperitoneally followed by

exposure to 112 MHz RF for 14 days. There were no significant changes in mortality or cell

proliferation between the control and the exposed groups. It is unclear if this was a sham or cagecontrol.

Two studies from the same group investigated the cytotoxic effects of 1800 or 900 MHz RFR

exposure to rat bone marrow cells (Atli Sekeroglu, Z., Akar, and Sekeroglu 2013; Sekeroglu, V.,

Akar, and Sekeroglu 2012). The first study (Sekeroglu, V. , Akar, and Sekeroglu 2012) examined

the effects of 1800 MHz RFR on immature and mature bone marrow cells (RIM and RMM). In each

study a total of 48 animals were divided into 6 groups of 8 animals each. Rats with RIM and RMM

were exposed to 1800 MHz RFR for two hours/day for 45 days. Some of the animals were tested

immediately and others returned to their cage for 15 days before testing. Insufficient information

was provided on dosimetry and there was no discussion about measuring the animals core

temperature. The reported whole-body average SARs were 0.37 W/kg and 0.49 W/kg for RIM and

RMM cells, respectively. The bone marrow cells were scored for common signs of bone marrow

cell toxicity including differences in mitotic index, chromosomal aberrations, and micronucleus

formation. Significant differences between control and exposed bone marrow cells were found in

chromosomal aberrations, micronucleus formation, mitotic index, and the ratio of polychromatic

erythrocytes. The authors found that the damage detected was greater in RIM cells than RMM

and that an unirradiated recovery period after exposure did not improve the RIM cells.

Additionally, it is not clear if the control animals were sham irradiated while in the irradiation

restraining device.

The same group, (Atli Sekeroglu, Z., Akar, and Sekeroglu 2013) repeated the experiment

with exposure to 900 MHz RF. The animals were exposed for 2 hours /day for 45 days. The

estimated SARs were 0.38-0.78 W/kg for RIM and 0.31-0.52 W/kg for RMM. Significant

differences between control and exposed bone marrow cells were found in chromosomal

aberrations, micronucleus formation, mitotic index, and the ratio of polychromatic erythrocytes.

The comet assay was not blinded, and the OECD approved %tail DNA was not used as the comet

Page 26

www.fda.gov

assay damage metric. The authors found that the damage detected was greater in RIM cells than

RMM and that an unirradiated recovery period after exposure did not improve for bone marrow

cells. Conclusions from this study must be made with caution because no positive control with a

known cytotoxic agent was included in this experiment and there was only one timepoint for the

recovery experiment reported.

Ziemann (Ziemann et al. 2009) examined the genotoxic potential of 902 MHz GSM and 1747

MHz digital cellular system (DCS) communication signals in a two-year study in mice. Each group

contained 65 animals, with the 15 animals used for interim examinations. There were additionally

30 male and 30 female mice assigned as sentinel animals. The animals were chronically exposed

to RFR for 2 hours/day, 5 days/week for two years at whole-body averaged SARs of 0.4, 1.3 and 4

W/kg. There was no indication that animal core temperature was measured or recorded.

Micronuclei frequency was determined in blood smears of cage control animals, sham irradiated

animals, with animals exposed to SAR equal to 0.4, 1.3 and 4 W/kg as well as mitomycin C (MMC)

as a positive control. MMC alkylates and cross-links DNA. No significant differences in micronuclei

frequency were found between RFR exposed animals and cage controlled or sham irradiated

animals. Only animals in the positive control group showed significant increase in micronucleus

frequency.

Furtado-Filho (Furtado-Filho et al. 2015) examined the impact of 950 MHz RFR on ROS

metabolism and DNA damage in the brains of young rats. RFR exposure was applied during

pregnancy to 6 animals and to 6 pups for 6 days after birth, with estimated SAR ranging from

1.14-1.32 W/kg. Another 6 pregnant rats were sham exposed. 6 pups from both exposed parents

and from sham exposed parents were dissected at 0 days old. Another 6 pups that continued

exposure and those that had no exposure were dissected after 6 days. There was no temperature

monitoring and no apparent blinding for the endpoint analysis. No positive control was included

in this study. No genotoxicity and oxidative stress were identified in neonates and 6-day old mice.

Another small study investigated the impact of 915 MHz with GSM modulated RFR on DNA

of brain, liver and kidney of rats (Trosic et al. 2011). 9 rats were in the exposed group and 9 rats in

the sham exposed group. The animals were irradiated at 37oC for one hour/day, seven days a

week for two weeks at a SAR of 0.6 W/Kg. Exposure was generated by a Gigahertz Transversal

Electromagnetic Mode Cell and a signal generator. The elevated environmental temperature

during exposure may have influenced thermal regulation of the exposed animals and thus

increased the chance of thermal confounding. Before and after exposures a ThermoScan

thermometer was used to measure body temperature. No significant changes were observed in

the treated animals compared to the sham exposed animals. Genotoxicity was assessed via the

comet assay used tail length and not the OECD approved endpoint %tail DNA to detect DNA

Page 27

www.fda.gov

damage, which demonstrated no statistically significant correlation between dose and comet tail

length.

Akdag et al. (2016) investigated the impact of long-term exposure to 2450 MHz RFR

exposure on DNA damage in the brain, liver, kidney, skin, and testicular tissue of rats. There were

two groups of 8 rats each. The control group was sham exposed. The animals were exposed 24

hours/day for 12 months using a signal generator. The animals’ core temperature was not

measured or not documented. There was no mention of blinding. Results from the comet assay

appeared to show more DNA damage in the RFR-exposed rats compared to the sham irradiated

rats. However, the results were not statistically significant for the DNA damage in brain, liver,

kidney and skin. The only tissue in which RFR significantly increased DNA damage was in the

testis.

Pandey et al. (2017) investigated histology and DNA damage in testicular cells in mice after

900 MHz RFR exposure. There were 5 groups of 15 mice each. The animals were exposed for

either 4 hours or 8 hours/day for 35 or 70 days and the SAR was calculated to be from 5.4 to 5.16

mW/kg. The animals’ temperatures were not measured. There was no blinding of quantitative

endpoints. There is no description of a sham exposure for the control group. The authors

concluded that oxidative stress caused DNA damage in germ cells of exposed mice without

measuring oxidative stress, and the precise link between the apparent oxidative stress and DNA

damage remains unclear.

Pandey and Giri (2018) investigated DNA damage with the comet assay and other oxidative

stress endpoints in germ cells of male mice exposed to 900 MHz RF radiation. There were 60

animals in 4 groups of 12 animals each. The groups were: controls, mice dosed with melatonin,

mice exposed to RFR, and mice dosed with melatonin and exposed to RFR. There was no mention

of sham controls. The animals were irradiated for 3 hours/day, twice a day for thirty-five days.

The SAR ranged between 0.052 and 0.0054 W/kg. The authors state that the samples and slides

were carefully blind coded before analysis; however, blinding is mentioned specifically only in the

histology section. The authors found that melatonin administration reduced DNA damage that

they attributed to RFR exposure.

Megha et al. (2015) examined the impact of microwave exposure on oxidative stress and

DNA damage in the hippocampus of Fisher Rats. There were 24 animals in 4 groups of 6 rats each.

The rats were exposed in a GHz transverse electromagnetic cell to 900, 1800 and 2450 MHz at

SARs of 0.59, 0.58 and 0.66 mW/kg respectively. The control group was sham exposed. Body

temperature of rats was noted by rectal measurements before and after the microwave exposure

in all the groups. The exposures were given to restrained animals for two hours/day, five

days/week for 60 days. The slides prepared were randomized and coded to blind the comet

Page 28

www.fda.gov

scorer, but no other blinding was done. The authors report that DNA damage was significant in all

microwave-exposed animals and they believe that the increase in DNA damage is frequency

dependent.

Guler et al., (2016) investigated the impact of 1800 MHz RFR on the brains of developing

rabbits. The endpoints were oxidative DNA damage (8-hydroxy-2’deoxyguanosine determination),

lipid peroxidation (malondialdehyde determination) and apoptotic cell formation TUNEL assay). A

signal generator with an integrated pulse modulation unit and horn antenna was used to expose

the animals. The estimated SAR was 18 mW/kg. 36 female rabbits were exposed for 15 minutes

/day for 7 days. 36 males were exposed for 15 minutes / day for 14 days. 36 of the infant rabbits

were exposed for 15 minutes / day in the intrauterine gestational period. Animals were also

exposed a month after birth. The animals were divided into 4 groups. Group I: no intrauterine +

no extrauterine exposure; Group II: no intrauterine + extrauterine exposure; Group III:

intrauterine exposure + no extrauterine exposure; Group IV: intrauterine and extrauterine

exposure. Blinding was used to prevent bias during sample analysis. There was no measurement

of the animals’ temperatures reported. The authors found that intrauterine and extrauterine

1800 MHz RFR exposure may cause oxidative and DNA damage in the brain tissue of baby rabbits

which are exposed while they are fetus and after they are 1-month old. However, there was no

evidence of apoptosis in any of the groups.

Turedi et al. (2014) studied the effect of 900 MHz RFR prenatal exposure on the 21-day old

male rat heart. The exposure system used was an ultra-high frequency oscillator attached to an

uninterrupted power supply and a dipole antenna installed into a plexiglass box. The whole-body

SAR was 0.025 W/kg. Three pregnant rats were exposed for 1 hour / day on days 13 to 21 of their

pregnancy. Three other pregnant rats were assigned to the control group and there was no sham

exposure. Six control and six exposed male pups were sacrificed on day 21 and hearts collected

for microscopy, malondialdehyde, superoxide dismutase, catalase, and glutathione activity

determination. There was blinding used to prevent analysis bias. Apoptotic changes and

irregularities in the heart muscle of exposed animals were shown by microscopy. Crista loss,

mitochondrial swelling, degeneration in myofibrils and structural impairments in z-bands were

shown by electron microscopy. Malondialdehyde, superoxide dismutase and catalase values were

found to be significantly higher. Glutathione activity was found to be lower. The authors

concluded that RFR exposure in the prenatal period causes oxidative stress and histopathological

changes in the heart, although the impact of any potential temperature increase during exposure

was not determined and there was no measurement of the animals’ core temperatures.

The genotoxic effects of 2.45 GHz RFR on the testis and ovary of rats was investigated by

Usikalu et al., (2013). The exposure system was a microwave generator and all exposed animals

were whole body irradiated for 10 minutes at various SARs. There were 30 male and 30 female

Page 29

www.fda.gov

rats used in this study and one control and 9 experimental groups. There were 3 male and 3

female rats in each group. SARs were measured by insertion of a thermistor into the rectum. The

animals were exposed to 0, 0.48, 0.95, 1.43, 1.91, 2.39, 2.9, 3.4, 3.8 and 4.3 W/kg. DNA damage

was assessed by direct amplification of length polymorphism and the comet assay. Organ damage

was assessed by histopathology. There was no blinding described in the paper nor was there a

sham control described. The authors conclude that 2.54 GHz RFR exposure alters DNA bands

patterns, induces single strand DNA breaks, and reduces the number of male germ cells in rats.

Furthermore, the authors concluded that RFR exposure at 0.48 W/kg and above produces

genotoxic effects on the testis or the ovary.

Ozturk Okatan et al., (2018) evaluated ovarian changes in adolescent rats exposed to 900

MHz RFR. The exposure system consisted of an oscillator, a continuous fixed power source and an

EMF cage. Rats were exposed to 900 MHz for 1 hour a day between postnatal days 35 and 59.

The calculated whole-body SAR was 0.0098 W/kg. Animals were divided into 3 groups: control,

sham-exposed and 900 MHz exposed. Each group contained 8 rats. Animals’ core temperatures

were not measured. Ovaries and blood were collected at the end of the exposure for

histopathological and biochemical analysis. Blinding was used to avoid analysis bias. Among the

results of the histopathological examination was thinning in the zona granulosa and theca layers,

shrinking in granulosa cells, reduced mitotic activity and leukocyte infiltration in the follicles and

stroma of RFR exposed animals. Among the results of the biochemistry examination was an

increase in superoxide dismutase, catalase and anti-Mullerian hormone levels in the RFR exposed

and sham exposed animals. The authors conclude that RFR exposure may cause changes in the

morphology and biochemistry of the ovary, although with changes also seen in the biochemistry

of sham exposed animals, the conclusions appear unclear.

The Influence of 1800 MHz GSM RFR exposure on oxidative DNA and Lipid damage in liver of

pregnant, non-pregnant, and new born rabbits was studied by Tomruk et al. (2010). Six groups of

nine rabbits each were used. A signal generator, pulse modulator and horn antenna were used as

the exposure system. Pregnant and non-pregnant rabbits were exposed whole-body to 1800 MHz

GSM-like RFR signal for 15 minutes / day for a week. No estimation of SAR was provided. Core

temperature measurements and blinding are mentioned. Oxidative DNA damage (8 OHdG/106DG

level) and lipid damage (malondialdehyde assay and ferrous oxidation in xylenol orange level) was

determined in control and exposed pregnant and non-pregnant as well as in new born control and

exposed rabbits. There was no difference in oxidative DNA damage between exposed and control

livers. However, lipid damage was detected in the exposed groups. No increased oxidative DNA

damage was detected in new born rabbits exposed to 1800 MHz RFR in utero. Increased lipid

damage was detected in new born rabbits exposed to 1800 MHz RFR in utero.

Page 30

www.fda.gov

Furtado-Filho et al., (2014) also investigated the impact of RFR on the liver of rats. These

investigators used a generator and a half-wave dipole antenna to produce RFR at a power of 1

Watt and a frequency of 950 MHz. The SAR was 0.88 W/kg for neonate rats, 0.51 W/KG for 6-day

old rats, 0.18 W/Kg for 15-day old rats and 0.06 W/kg for 30-day old rats. Six pregnant rats were

exposed to SARs ranging from 0.01 to 0.03 W/kg. 6 other pregnant rats were assigned to the

control group. The animals were exposed to RFR for thirty minutes / day for up to 51 days. The

exposure was divided into 21 days of gestational exposure and up to 30 days outside the womb.

Temperature was not measured and there is no mention of blinding. Damage to lipids, proteins

and DNA (comet assay) was assessed. The authors found 950 MHz exposure does not cause

oxidative stress and is not genotoxic to neonates or rats up to 15-days old. However, the authors

also found that RFR exposure is genotoxic to 30-day old animals.

The excretion of 8-oxodG, an oxidative DNA marker in the urine of rats exposed to 1800

MHz RFR was conducted by Khalil et al. (2012). A GSM signal generator and antenna were used to

expose the rats to 1800 MHz for 4 hours. There were 24 rats in two groups of 12. Sample analysis

was blinded. The SAR was 1.0 W/kg. The results indicate that RFR exposure increased the urinary

excretion of 8-oxodG compared to the control. The authors concluded the following: “the present

results suggest that (a) exposure to RFR may augment oxidative stress; (b) urinary excretion of 8oxodG may be a good and specific marker of oxidative stress. Further work is also required to

understand the origin and validity of excreted urinary 8-oxodG as a marker of oxidative stress.”

Two investigations from the same group investigated the adaptive response in mice exposed

to 900 MHz (Jiang et al., 2012; Jiang et al., 2013). Jiang et al., (2012) examined the impact of prior

exposure of 900 MHz RFR for 4 hours / day for 1, 3, 5, 7 and 14 days to subsequent exposure to 3

Gy of ionizing radiation. There were 3 exposure groups containing 5 mice each. There was no

blinding. The exposure system consisted of a GTEM chamber, power amplifier and an antenna. The

calculated SAR was 0.548 W/kg. The exposed mice were restrained during exposure. DNA damage

was assessed with the alkaline comet assay. One exposure for 4 hours to 900 MHz plus 3 Gy was

not significantly different from 3 Gy alone. Mice pre-exposed for 3 or more days to 900 MHz show

progressively less damage to 3 Gy suggesting adaptive response triggered by the conditions of the

RFR exposure or the RFR.

Jiang et al. (2013), conducted a similar experiment that investigated the impact of prior

exposure to 900 MHz RFR for 4 hours / day for 7 days on subsequent exposure to 3Gy of ionizing

radiation. There were 6 groups of 10 mice each used in this study. The exposure system consisted

of a GTEM chamber, power amplifier and an antenna. The calculated SAR was 0.548 W/kg.

Genotoxicity was assessed via the micronucleus assay in immature erythrocytes in peripheral blood

and bone marrow. Results of the micronucleus assay were similar for the non-exposed control mice

and the 900 MHz RFR exposed mice. Exposure to 900 MHz RFR for 7 days before 3 Gy of ionizing

Page 31

www.fda.gov

radiation resulted in a significant decrease in micronuclei compared to 3 Gy ionizing radiation only.

This may indicate the conditions of RFR exposure or the RFR exposure activated an adaptive

response.

Meena et al., (2013) studied the protective effects of Melatonin (antioxidant) on oxidative

stress-mediated testicular impairment due to long- term exposure from RFR. The exposure system

consisted of an anechoic chamber in a far field region from a horn antenna. The rats were exposed

to 2.45 GHz RFR for 2 hours per day for 45 days. There were 24 rats in 4 groups of 6 rats each. The

samples and analysis were blinded. There was no data regarding the animals’ core temperature.

The SAR was estimated to be 0.14 W/kg. Four groups were tested in this investigation: sham

exposed, melatonin treated (2g/kg), 2.45 GHz RFR and 2.45 GHz plus melatonin. Oxidative stress

related parameters were documented including lactate dehydrogenase activity, xanthine oxidase

activity, reactive oxygen species, protein carbonyl content, DNA damage (comet assay) and

malondialdehyde concentration. The authors found that melatonin reduced oxidative damage by

increasing lactate dehydrogenase, decreasing malondialdehyde and reactive oxygen species.

Melatonin also reversed the effects of 2.54 GHz RFR on xanthine oxidase, protein carbonyl content,

sperm count, testosterone level and DNA strand breaks. The authors concluded that melatonin has

a strong effect on the RFR mediated oxidative stress in testicular cells.

The effect of RFR adaption on bleomycin-induced DNA and oxidative damage and repair was

studied by Zong et al. (2015). The exposure system was a GTEM chamber. 900 MHz RFR exposure

was for 4 hours / day for 7 days. The mice were restrained. The SAR was estimated to be 50 mW/kg.

There were 48 mice total. Six groups were tested. The animals’ core temperatures were not

measured. The samples were blinded. The groups included unexposed controls, RFR only, sham

exposure only, Bleomycin only, RFR exposure plus bleomycin and sham exposure plus bleomycin.

DNA damage was assessed via the comet assay. Oxidative damage was assessed via

malondialdehyde levels and antioxidant status by superoxide levels. There was no indication for

increased DNA and oxidative damage in mice exposed to RFR alone. Mice exposed to RFR plus

bleomycin significantly reduced bleomycin damage and decreased malondialdehyde concentration

in the plasma and lung and increased SOD levels in lung. The conclusion was that RFR exposure

induced an adaptive response to bleomycin DNA and oxidative damage by activating certain cellular

processes. There was no evidence for DNA strand breaks or oxidative damage caused by RFR

exposure in this experiment. The authors suggested that RFR exposure activates an adaptive

response through cellular processes that are not DNA or oxidation dependent.

C. Conclusions from In vivo Review.

Page 32

www.fda.gov

The in vivo studies conducted between January 1, 2008 and August 1, 2018 and summarized

here have contributed to our collective understanding of the potential effects of RFR on

mammals. Overall, based on certain limitations, these studies have not produced any clear

evidence that RFR exposure has any tumorigenic effect. In some cases, the authors of these

studies suggested the need for more research based on the reported results. Other authors

stated that RFR exposure does not result in tumor-initiating or -promoting effects.

Most of the referenced literature did not include measurement of animal temperatures. The

impact of potential temperature elevation from RFR exposure is a confounding factor in such

studies and critical to assess. The NTP reports examined temperature increases in their 5-day pilot

study. However, they did not adjust their highest exposure groups, exposures up to 6 W/kg in rats

and 10 W/kg in mice, to address concerns with geriatric thermoregulation. The NTP pilot study

confirms that pregnant animals and older animals have difficulty regulating body temperature

when they are exposed to sources of heat such as RFR exposure. NTP reports indicate a need to

further evaluate how mammals’ thermoregulatory abilities decline as they age and how to

address this issue to remove confounding heating as a factor in any future RFR exposure research.

Similarly, some of the other literature we reviewed have results that are confounded by heating

issues.

As with many areas of active investigation, the variety of experimental methods employed

can result in significant differences in conditions experienced by the experimental animals, and in

turn this can lead to a high degree of variability in study conclusions. For RFR studies, there are

large differences in, for example, methods for dosing and restraining animals during the

experimental procedure that will affect the specific stress levels of the test animals, confounding

some conclusions. There are also important differences in the overall experimental designs,

including the use (or lack) of both a control and sham exposed control group. Such differences

may be one reason why there are many diverging and conflicting conclusions of the in vivo effects

of RFR.

Another experimental issue that has an impact on the understanding of the interaction

between RFR with animals and needs further research is how samples are handled and evaluated.

The method of sacrifice (time since last exposure) and time from an animal’s death to sample

preparation have a large impact on the results of sensitive tests like the comet assay. In these

experiments, there were differences in how biological samples were collected and processed that

added confounding factors. One of the most prevalent problems we observed in these studies

was the lack of adequate blinding of samples which generally leads to an unconscious bias in the

evaluation of data. More clarity regarding potential confounders is needed in this area of

research.

Page 33

www.fda.gov

The FDA has followed RFR research developments for over 20 years. More than 100 peer

reviewed articles about in vivo animal exposures were assessed for to determine if they met the

scope of this review. 37 peer reviewed articles met our criteria and are included in this review. In

vivo studies are of great value and contribute to our understanding of this topic. However, as

described above, due to the critical limitations of in vivo studies in assessing the effects of RFR

exposure to humans (e.g., whole-body RFR exposure), we cannot draw conclusions about the

impact of such exposure to humans based on these in vivo animal studies. The results from such

studies should not be applied to human cell phone usage as further research is needed.

Page 34

www.fda.gov

Epidemiological Studies

Section V.A. below discusses the search strategy and search terms we adopted to find

relevant epidemiological literature. Section V.B. of the report provides a high-level narrative

summary of the epidemiological studies reviewed and then followed by a description of the

overall limitations of such studies in Section V.C. Specifically, Section V.B. summarizes general

findings stratified by overall risk trends (e.g., increasing or decreasing), study types and/or sources

(e.g., ecological trend analysis or case-control study), and outcomes of interest represented by

different tumor types (e.g., glioma, meningioma, vestibular schwannoma, etc.). Section V.D.

includes a more detailed description of the individual studies reviewed, including summary of the

methodology and main results, along with the limitations for each referenced study. Section V.E

describes the FDA’s findings based on the review of the available epidemiological literature.

A. Search Strategy.

As the primary source for Medline-indexed journals and peer-reviewed publications,

PubMed was queried using the following search terms and limits:

("cell phone*" OR "mobile phone*" OR "non-ionizing radiation" OR "radiofrequency radiation")

AND (cancer OR cancers OR carcinogen* OR tumor OR tumors OR tumorigenesis)

Filters: Full text, English and Humans

The initial PubMed query with no limits resulted in a total of 1,180 records. After applying

PubMed filters for Full Text, English, and Humans, the number of records was reduced to 792

publications, which constituted the pool for 1st pass (i.e., abstract- based) review. The 1st pass

review excluded articles not directly related to the review’s subject (e.g., main outcomes other

than cancer). Note: a relatively high number of excluded articles (343 of the 388) was due to the

fact that in many studies, RF-related wireless technology was a part of the study design, but not

the main research topic (e.g., mobile phones were used for collecting data from cancer patients).

The 1st pass review identified 449 records, which were focused on the RF- related cancer risk, but

included animal (in vivo) studies as well as article formats such as meta-analyses, expert opinions,

etc. The 2nd pass (full-text based) review was aimed to select the original clinical/epidemiologic

studies assessing RF-related cancer risk. Per RF Working Group’s request, the 2nd pass review also

applied the publication date filter (January 2008 – May 2018) for deriving more recent evidence.

After the 2nd pass review, a total of 69 records (with one cross-reference) were selected for the

final review with qualitative evidence synthesis.

Note: this technical report includes some additional references, which have been used to

support the overall data analysis and interpretation. In order to distinguish them from the

Page 35

www.fda.gov

selected clinical/epidemiological publications (which are listed in the References section), the

additional references cited are presented in the footnotes.

B. Summary of Results.

This section only provides a high-level narrative summary of the epidemiological studies

reviewed. For a description of the limitations of these studies see Section V.C below and for a more

detailed description of each of the individual studies referenced here, see Section V.D.

1. Studies reporting overall evidence that does not counter a global null

hypothesis on tumor risk in relation to cell phones.

Negative evidence on associations between use of cell phones and tumor incidence was

shown in a number of ecological trend analyses reporting no population-level risk increases

for the brain and central nervous system (CNS) tumors (Inskip 2010; de Vocht 2011b; Ding and

Wang 2011; Benson 2013; Hsu 2013; Kim 2015; Sato 2016), which included glioma and

meningioma (Deltour 2009, 2012; Chapman 2016), acoustic neuroma 12 (Benson 2014), and

parotid gland neoplasms (de Vocht 2011a). In the study by Barchana (2012), no increasing

incidence in the period after introduction of mobile phones was reported for low-grade (but

not high-grade) gliomas.

Similarly, no evidence for overall tumor risk increase in relation to cell phone use was

shown in a number of case-control and cohort studies on acoustic neuroma (Corona 2012; Han

2012; Schüz 2011), glioma (Gousias 2009), pituitary adenoma (Shoemaker and Swerdlow 2011),

uveal melanoma (Stang 2009), and salivary gland tumors (Soderquist 2012). Some studies

(Lankola 2008; Shrestha 2015) reported overall reduced risk for meningioma (OR=0.76 [95% CI:

0.65; 0.89]) or pituitary tumors (OR=0.39 [95% CI: 0.21; 0.72]), which however indicated the

presence of strong biases as discussed below.

However, some studies (Aydin 2011, 2012; Cooke 2010; Coureau 2014; Frei 2011; Kaufman

2009; Poulsen 2013; Yoon 2015) reported no overall risk increases in the entire study populations

but found elevated risks in certain study subpopulations (for more details, see section IV.D.

below). Risk increases among certain cell phone user categories were found even in the largesample multi-country INTERPHONE (2010, 2011) studies that reported overall reduced risks for

glioma and meningioma (OR=0.81 [95% CI: 0.70; 0.94] and OR=0.79 [95% CI: 0.68; 0.91],

respectively) and acoustic neuroma (OR=0.85 [95% CI: 0.69; 1.04]).

Acoustic neuroma is also known as vestibular schwannoma; in this memo, both terms are being used interchangeably, as presented

in original publications.

Page 36

www.fda.gov

No overall risk increase corresponding to the increasing time of cell phone use was found in

the Swedish population-based study on acoustic neuroma, where controls were assigned a

reference date and a laterality for “fictive tumor” corresponding to their matched case

(Pettersson 2014).

No significant risk changes in relation to cell phone use were found in the study on

malignant brain tumors by Spinelli (2010). Positive risk association, however, was reported for

occupational computer exposure for ≤ 4h/day. The lower risk reported for residing near cell

phone towers was speculatively attributed by the authors to possible lack of EMF on the lower

part of the base stations (BS), or to lower emissions from cell phones due to better connections in

the surrounding areas.

2. Studies Reporting Cvidence on Possible Tumor Risk Modification in

Relation to Tumor Types and/or Cell Phone Usage Practices.

Tumor risk increases in relation to use of cell phones were extensively reported by Hardell

and Carlberg (2008-2017, Sweden). A recent ecological trend analysis by this group suggested an

increase in thyroid cancer incidence concordant to mobile phone use (Carlberg 2016). Increasing

rates of brain tumors of unknown type (with joinpoint 2007) were speculatively attributed to the

glioma risk increase due to use of cell phones (Hardell and Carlberg 2015). As a further

speculation, the increased brain tumor risk among users with the first use of a cell phone before

the age of 20 was suggested as possible explanation for the increasing rates (2007–2015) of

unknown type CNS tumors that were shown to be specifically elevated in the age group of 20–39y

(Hardell and Carlberg 2017).

However, possible risk increases in relation to use of cell phones were also suggested in

other ecological trend analyses based on non-Swedish populations. The UK-based study by de

Vocht (2016) showed a faster than expected incidence increase for malignant neoplasms of the

temporal lobe. The Israel-based study by Barchana (2012) found an increasing trend in the period

after mobile phone introduction for high-grade gliomas (including glioblastoma multiforme),

especially among females. Using worldwide GLOBOGAN 2008 data, de Vocht (2013) identified the

rate of mobile/ cell communications subscriptions as a risk indicator for brain/nervous system

cancer. Using data from cancer registries and GLOBOGAN 2012, Neupane (2017) identified

positive associations between the mobile phone and personal computer densities and the

elevated incidence of mortality in prostate cancer. According to the single available ecological

trend analysis involving the US population (Little 2012), glioma incidence trends per US

Surveillance, Epidemiology, and End Results (SEER) program did not confirm the risk increase

Page 37

www.fda.gov

inferred from Swedish studies by Hardell (2010) but were consistent with the modest risk

estimate derived from INTERPHONE (2010) study.

In addition to the aforementioned ecological trend analyses, evidence on possible tumor risk

increases was found in numerous case-control and cohort studies, as presented below in the

sections stratified by outcomes of interest.

3. Outcome of Interest – Brain Tumors (Glioma and Meningioma)

Combined analysis from case-control studies on brain tumors by Hardell and Carlberg

(Carlberg and Hardell 2009, 2014; Hardell 2010, 2011, 2013a,b) showed increased hazards ratios

(HR) for all gliomas with the latency >20 years (HR=1.7 [95% CI: 1.2; 2.3]) and astrocytoma grade

IV (glioblastoma) with the latency >10 years (HR=1.3 [95% CI: 1.03; 1.7]). Highest HR of 2.3 [95%

CI: 1.1; 4.7] was reported among individuals who started using mobile phones before the age of

20. Risk for malignant brain tumors was reported to vary, increasing with the latency 13 >1-5 years,

decreasing in the next latency categories, and then increasing again with the latency >15- 20

years. Per analysis of deceased cases, risk for malignant brain tumors (mostly astrocytoma grade

III-IV) was shown to increase with cumulative lifetime use (Ptrend=0.02). A borderline increased risk

for meningioma (Carlberg and Hardell 2015) was found in the 4th quartile of cumulative phone

use (>1,436 h).

Despite reporting the reduced overall risk as the main finding, INTERPHONE (2010) study

found an increased risk for glioma and meningioma among ‘heavy’ mobile phone users with the

latency of 1–4 (but not >4) years: OR=3.77 [95% CI: 1.25; 11.4] and OR=4.80 [95% CI: 1.49; 15.4],

respectively. Note: some reported values in the ‘heavy’ use (≥1,640h) category, defined as the

10th decile of cumulative call time, were deemed “implausible” by the study authors.

The possibility for increased risks of glioma and meningioma (OR=2.89 [95% CI: 1.41; 5.93]

and OR=2.57 [95% CI: 1.02; 6.44], respectively) among ‘heavy’ users (≥896 h) was also supported

in the CERENAT-based study by Coureau (2014).

In the CEFALO-based study on children and adolescents (Aydin 2011) which did not find the

overall risk increase for brain tumors, elevated risk was found among young individuals with >2.8

years of use (per operator-recorded data): OR = 2.15 [95% CI; 1.07; 4.29].

Latency is defined here and in other studies as a time period from first use of mobile phone to diagnosis of tumor.

Page 38

www.fda.gov

In the large cohort study by Frei (2011) which found no indication of dose-response relations

either by subscription years or by tumor location, a borderline increased risk for glioma was found

among men with mobile phone use of 1-4 years: OR=1.20 [95% CI: 0.96; 1.22].

4. Outcome of Interest – Acoustic Neuroma, or Vestibular Schwannoma.

The Swedish studies by Hardell and Carlberg showed a risk increase for acoustic neuroma

among mobile and cordless phone users, especially with the latency >20 years (Hardell and

Carlberg 2009; Hardell 2013c).

INTERPHONE (2011) study found an increased risk for acoustic neuroma in ‘heavy’ users

(≥1640 h) with the latency of 5 years: OR=2.79 [95% CI: 1.51; 5.16]. Note: some values reported in

the ‘heavy’ use category were deemed implausible by the study authors.

Overall increasing risk trends for acoustic neuroma among mobile phone users with >5 min

per call were found in the study by Sato (2011). Substantial risk increases were reported among

users with >20 min/day on average: OR=2.74 [95% CI: 1.18; 7.85] and OR=3.08 [95% CI: 1.47; 7.41]

with the latencies for 1 year and 5 years, respectively.

The study on acoustic neuroma by Pettersson (2014) showed a risk increase among cordless

phone users with higher cumulative use: OR=1.67 [95% CI: 1.13; 2.49]. Among histologicallyconfirmed cases, a risk increase was shown among users of analog phones for 5-9 (but not >10)

years: OR=4.03 [95% CI: 1.07; 15.2].

The case-control study by Moon (2014) did not find evidence for increased incidence of

acoustic neuroma in relation to mobile phone use; however, the case-case analysis from the same

study found correlations between tumor volume and phone usage estimates (see below).

5. Outcome of Interest – Parotid Gland Tumors.

In the study by Sadetzki (2008), elevated risk for parotid gland tumors was associated with

regular use of cell phones and high-exposure conditions. The risk elevation was confined mostly to

rural areas, reaching a 2-fold increase among users with ≥1,035h of cumulative call time: OR=1.96

[95% CI: 1.11; 3.44].

Increased likelihood of parotid gland tumors in relation to ‘heavy’ use of cell phones was

reported by Duan (2011). Risk for epithelial parotid gland malignancy was independently

associated with 13 phone usage variables, with the greatest OR for the user category with >42,000

calls since first use: OR=15.36 [95% CI: 13.34; 17.38]. Possible risk increase was particularly

Page 39

www.fda.gov

suggested for mucoepidermoid carcinoma with an average daily use of >2.5h: OR=31.25 [95% CI:

10.79; 90.45].

6. Outcome of Interest – Skin Cancers.

In the large cohort study on skin cancers (Poulsen 2013), a borderline increase of adjusted

incidence rate ratio: IRR=1.18 [95% CI: 0.98, 1.42] was reported for basal cell carcinoma among

women with 5-9 years (but not 10-12y or >13y) of mobile phone subscription.

7. Outcome of Interest – Leukemia.

Although the case-control study by Cooke (2010) did not find overall risk increase for

leukemia, a borderline increased risk for acute myeloid leukemia was associated with ≥15 years of

use of mobile phones: OR=2.08 [95% CI: 0.98; 4.39].

Possible risk modification due to certain usage practices was also suggested in the casecontrol study by Kaufman (2009) which found no clear overall association with use of cell phones.

Increased risk for leukemia was found among subjects using Global System for Mobile

communications (GSM): OR=2.1 [95% CI: 1.1; 4.1]. Risk increase in relation to cell phones was

shown for myeloid leukemia (acute and chronic combined): OR=1.7 [95% CI: 1.0; 2.9]. Further,

increased risks for combined myeloid leukemia and especially for acute myeloblastic leukemia

(OR=4.3 [95%: 1.3; 15] and OR=5.5 [95% CI: 1.4; 21], respectively) were found among subjects

whose work was affiliated with powerlines. Substantially elevated odds of having myeloid

leukemia were found in cases when ≥2 putative etiological factors (e.g., benzene, pesticides) were

involved.

8. Putative Effect Modifier – Laterality of Tumor in Relation to Laterality

of Phone Use.

The Swedish studies by Hardell and Carlberg (Carlberg and Hardell 2009, 2014; Hardell

2010, 2011, 2013a,b) associated the highest risk for astrocytoma (a type of glioma) with the

latency >10 years of ipsilateral use of mobile and cordless phones. The same authors (Hardell

and Carlberg 2009; Hardell 2013c) also suggested the highest risk for acoustic neuroma among

ipsilateral mobile phone users with >10 years of use.

A greater likelihood for ipsilateral vs. contralateral gliomas among regular users (OR=1.27

[95% CI: 1.19; 1.37]) was also reported in the INTERPHONE (2010) study.

In the case-case and case-specular analyses (Larjavaara 2011) where distances from actual

and specular locations were compared using a hypothetical reference location assigned for each

Page 40

www.fda.gov

tumor, a significant excess of gliomas was found on the self-reported side of use (this finding,

however, was attributed by the authors to recall bias).

In the study by Yoon (2015) which did not find an overall risk increase, increased relative

risks (RR) were reported for among regular users with the same laterality for glioma and phone

use: RR=1.26 (P=0.05) and RR=1.43 (P=0.01) for all respondents and self-respondents,

respectively.

In addition to intracranial tumors such as glioma and meningioma, laterality analysis was

also performed for acoustic neuroma and parotid gland malignancies. INTERPHONE (2011) study

found risk increases for ipsilateral tumors in ‘heavy’ users (≥1640 h) with the latencies for 1 year

and 5 years: OR=2.33 [95% CI: 1.23; 4.40] and OR=3.53 [95% CI: 1.59; 7.82], respectively.

In the study by Sato (2011), a more frequent use of the affected ear was reported in all but

one case (15/16) among ‘persistent heavy’ users (>20 min/day) with acoustic neuroma: OR=5.0

[95% CI: 1.3; 24.8].

In the study by Sadetzki (2008), elevated risk of parotid gland tumors was found among

regular – ipsilateral or bilateral – users with the latency of 5 (but not 10) years. The risk elevation

was especially noticeable among ipsilateral users in the highest categories of cumulative number

of calls and call time: OR=1.58 [95% CI: 1.11, 2.24] and OR=1.49 [95% CI: 1.05, 2.13], respectively.

9. Putative Effect Modifiers – Location and Size of Tumors in Relation to

Mobile Phone Use.

The Swedish studies by Hardell and Carlberg (Carlberg and Hardell 2009, 2014; Hardell

2010, 2011, 2013a,b) suggested increasing risk trends for brain tumors located in the temporal

lobe or overlapping tumors in the temporal and adjacent lobes.

INTERPHONE (2010) study also reported higher likelihood of gliomas located in the temporal

lobe, with the higher risk estimate among ‘heavy’ (≥1,640) users: OR=1.87 [95% CI: 1.09; 3.22].

On the other hand, significantly increased risks for gliomas with locations other than

temporal/ frontal lobe were shown in studies by Coureau (2014) and Aydin (2011): OR=3.61 [95%

CI: 1.00; 12.96] and OR=1.92 [95% CI; 1.07; 3.44], respectively. The cohort-based study on brain

tumors by Frei (2011) reported increased risks for other/unspecified type tumors: OR=1.35 [95%

CI: 1.05; 1.75], tumors in the cerebral ventricle: OR=2.58 [95% CI: 1.08; 6.15] as well as tumors in

the occipital lobe among users for 1-4 years: OR=2.50 [95% CI: 1.18; 5.31].

Page 41

www.fda.gov

The aforementioned analysis on the actual and specular tumor locations (Larjavaara 2011)

did not show predominance of gliomas in the brain parts expected to have the highest RF fields.

In the study by Moon (2014) which did not find increased incidence for acoustic neuroma in

relation to mobile phone use, the MRI-based case-case analysis showed a trend for larger tumors

among regular users (p=0.001), indicating correlation between tumor volume and cumulative

hours of use (r2=0.144, p=0.002).

10. Evidence on Tumor Risk in Relation to Cell Phones from Studies Using

RF Exposure Estimates Other Than Phone Usage Practices 14.

In addition to the aforementioned studies using phone usage practices as the proxies for RF

exposure, some studies assessed tumor-related outcomes based on the RF exposure estimates

such as total cumulative specific energy (TCSE) and SAR.

No increased risk for brain tumors (glioma and meningioma) was found in the case-control

study by Takebayashi (2008), which included assessment of SAR-based spatial relationships

between tumor location and intracranial RF distribution.

The TCSE-based analysis using INTERPHONE study components (Cardis 2011) showed an

increased glioma risk (Ptrend=0.01), reaching OR=1.91 [95% CI: 1.05; 3.47] for the highest TCSE

quintile category with mobile phone use for >7 years. The risk estimate was further increased for

gliomas in the most exposed areas with >10 years of use: OR=2.80 [95% CI: 1.13; 6.94].

The INTERPHONE-based analysis using both SAR and TCSE (Grell 2010) showed an increased

risk of observing a glioma within the shortest distance from preferred ear to gravity center of

tumor (0-55 mm) with cumulative use ≥200 h: OR=4.06 [95% CI: 2.03; 11.6]. Overall, shorter

distances were associated with larger tumors: OR=4.09 [95% CI: 1.90; 12.0].

In the SAR-based case-case analysis by Hartikka (2009), a significant risk increase was shown

for gliomas among contralateral (but not ipsilateral) mobile phone users: OR=4.93 [95% CI: 1.13;

21.5]. However, a risk change (OR=0.51 [95% CI: 0.27; 0.96]) based on the distance between

handset and tumor midpoint among regular vs. non-regular users indicated that a distance

increase by 1cm decreased the odds of glioblastoma by 49%.

In the study by Dode (2011), which was conducted in the Brazilian region of Minas Gerais

with a major concentration of cell phone transmitter antennas, SAR-based analysis of spatial

Evaluation of the evidence presented in this section is limited to epidemiological aspects only and it does not include the

methodological validity of dosimetry assessments in the cited publications

Page 42

www.fda.gov

correlation between BS clusters and corresponding mortality estimates (1996-2006) showed an

increased risk of death by neoplasia (more details including the subsequent critical comments and

authors’ response can be found in the individual review of this publication in the section below).

In the population-based case-control study on RF exposure in relation to childhood

neoplasms (Li 2012), annual summarized power (ASP, watt-year) was calculated for mobile phone

BSs and annual power density (APD, watt-year/km2) of each township was computed as a ratio of

the total ASP of all BSs in a township. As a result, typically encountered BS-emitted RF levels were

found not to pose substantial childhood cancer risk. However, higher than median APD (168 wattyear/km2) was associated with borderline risk increases for all neoplasms and leukemia in

particular: OR=1.13 [95% CI: 1.01; 1.28] and OR=1.23 [95% CI: 0.99; 1.52], respectively.

Page 43

www.fda.gov

C. Overall Limitations of Epidemiology Data.

As described in this section, despite the sizeable amount of retrieved and reviewed data,

cross-cutting limitations that involved almost all manuscripts impacted the overall data

interpretation and limited comprehensive conclusions on all possible aspects of RF-related tumor

risk for humans.

1. Data Quality Affected by Limitations of RF Exposure Measurements.

Epidemiologic evidence was mostly derived from observational studies that used phone

usage practices as the proxy estimates for RF exposure. None of the available

epidemiologic/clinical studies, including the ones that used SAR and TCSE based RF exposure

measurements, attempted to estimate the entire accumulated RF dose from all possible EMFs in

the individual environments of study subjects. The accuracy of indirect RF assessment per phone

usage practices was further limited by use of self (or kin)-reported questionnaires that had varying

definitions for measurables and that were mostly collected in the non-blind setting. In addition to

possible inaccuracies in the RF exposure measurements based on phone usage patterns (which

are not necessarily directly correlated), exposure misclassification was virtually unavoidable in

ecological trend analyses (de Vocht 2013; Neupane 2017) or cohort studies (Poulsen 2013) that

used mobile phone subscription data, without the means for further classification by actual phone

usage practices at individual levels.

2. Strong Biases Affecting Data Analysis and Interpretation.

Due to inevitable errors in the measurements of RF exposure, all studies were affected, to

varying degrees, by the exposure classification bias. In addition to possible nondifferential bias

with a systematic measurement error (similar between cases and controls), exposure

misclassification, to an even greater extent, was likely caused by differential errors due to recall

bias (e.g., better memory of phone use among study cases vs. healthy controls) or ascertainment

bias (e.g., phone use information was influenced by tumor diagnosis). Recall bias is avoided in

cohort studies as exposure is assessed prior to disease occurrence. Furthermore, the two

available cohort studies, the million women study of Benson et al (2013) and the Danish cohort

study of Schuz et al (2011) both failed to find an association between cell phone use and glioma

or meningioma. Information on estimated RF exposure might have been also affected by mental

status of interviewees with brain tumors or by hearing impairment in cases with acoustic neuroma

(with the latter causing non-random variations in laterality analyses in particular). Many case-case

studies were prone to selection and classification biases due to heterogeneity or limited

availability of cases and/or controls, resulting in the study populations that were not fully

Page 44

www.fda.gov

representative of cell phone use practices or tumor incidences. Some studies were also prone to

possible assumption and interpretation biases due to their study designs based on uncertain

postulations (e.g., hypothetical tumor location assigned for controls, or baseline risk estimates in

ecological trend analyses). Exposure assessment is further limited by the fact that there is an

absence of experimental evidence that can determine if frequency, duration, or intensity of

exposure is the most important factor in potential cancer risk. The component of exposure that

epidemiologists should target is unknown.

Even the most cited international studies such as INTERPHONE (2010, 2011) which collected

individual phone usage data were affected by heavy multi-composite biases that resulted in the

overall reduced tumor risk, undermining the validity of derived evidence. Although the

INTERPHONE study authors rendered a genuine protective effect as implausible and referred to

strong bias as the main explanation, their reports of reduced tumor likelihood (along with a

number of elevated ORs) spurred widespread discussions, involving social and popular science

media and resulting in some ungrounded statements such as “for most people cell phone use

appeared to protect against brain tumors” 15. On the contrary, increased risk trends that stemmed

from multiple publications by the same group (Hardell and Carlberg, 2009-2017, Sweden) were

likely affected by limitations of a single data source limited to one population.

3. Lack of Evidence Based on the U.S. Population.

The vast majority of reviewed studies, including multi-country studies (INTERPHONE 2010,

2011; de Vocht 2013), were conducted outside the US (mostly, in Europe), while very few studies

assessed the risk of tumor in relation to mobile phone use based on the US population (Inskip

2010; Han 2012; Little 2012). There is compelling evidence from the Surveillance, Epidemiology

and End Results (SEER) program, a highly regarded cancer registry, that the incidence of brain

cancer has not increased over the last 20 years despite cell phones becoming nearly ubiquitous16.

It might also be noted here that the failure to detect a real increase in brain cancer rates is

exceedingly unlikely because of the high specificity and sensitivity of the non-invasive diagnostic

modalities of computerized tomography and magnetic resonance imaging.

Finally, it is worthy of note that there are currently no primary or secondary prevention

efforts to reduce brain cancer incidence. There are no countervailing forces that would offset a

potential increase in the rate of brain cancer from cell phone use. With the findings from these

15 https://www.sciencenews.org/blog/science-public/interphone-study-finds-hints-brain-cancer-risk-heavy-cell-phone-users

16 https://www.arpansa.gov.au/news/new-arpansa-study-finds-no-link-between-mobile-phone-use-and-brain-cancers

Page 45

www.fda.gov

three cited US studies mentioned in the Summary, “the very linear relationship between cell

phone usage and brain tumor incidence” reported in an additional US study (Lehrer et al 2011)

cannot be accepted as valid due the study’s flawed statistical assessment (for more details see

Section V.D below).

4. No Overall Risk Increase, with Some Evidence Pointing at Possible

Subgroup Effects.

As discussed by the INTERPHONE (2010, 2011) authors, while systematic bias in these

studies could result in spurious associations inflating certain risk estimates, the elevated ORs

could also indicate small effects detectable only in subsets of the study population. The possibility

of true risk elevations due to the small subgroup effects (rather than due to the falsely inflated

risk estimates) appears even more plausible, when considering the presence of a strong bias that

resulted in the overall reduced risk estimates in both INTERPHONE studies.

The possibility of subgroup effects was also indicated in studies other than INTERPHONE

2010, 2011. In most studies reporting no overall risk increase, the main conclusion was based on

the absence of upward trends across all cumulative call time categories. However, the absence of

a linear dose-dependent correlation does not exclude the possibility of an intricate risk

modification, especially since the existing epidemiologic studies did not offer sufficient statistical

power to detect relatively small – subgroup-based – effects. As hypothesized by Deltour (2009)

and de Vocht (2011), putative risk modification may be too small to be observed in the overall

study population, as it may concern only certain individuals or tumor subtypes.

As examples indicating possible subgroup effects, some studies (INTERPHONE 2010, 2011;

Pettersson 2014; Sadetzki 2008; Benson 2013 17) reported risk increases among phone users with

shorter vs. longer latencies. Some of these risk fluctuations might have been due to an earlier

tumor detection in ‘heavy’ users, as was particularly suggested for acoustic neuroma (Pettersson

2014). However, risk increases in the intermediate latencies may also indicate earlier tumor

development in the cases where a relatively short but intense RF exposure might have been

sufficient for triggering tumorigenesis in more vulnerable subjects who were inherently

predisposed to tumorigenesis due to various genetic and environmental reasons.

As referenced by Sadetzki (2007), RF-EMF emissions from cell phones do not have enough

energy to break chemical bonds or damage DNA and, therefore, are unlikely to initiate

tumorigenesis. However, the lack of recognized direct genotoxicity or mutagenicity (as referenced

This reference refers to the increased risk for pituitary tumors and not for acoustic neuroma, as the latter finding was retracted in

the subsequent publication (Benson 2014).

Page 46

www.fda.gov

by Takebayashi 2008 for RF-EMF exposure in the range of 800–1900 MHz and an SAR <2Wkg-1)

does not preclude the possibilities for indirect tumor-modifying effects due to RF exposure. As

suggested by de Vocht (2011), rather than acting as a tumor-initiating factor, the RF exposure may

act as a tumor promoter, whose effects may include inhibition of anti-tumor immune responses.

Indicating possible tumor-promoting effects, the MRI-based analysis by Moon (2014) revealed

positive correlation between ‘heavy’ RF exposure and the tumor size of acoustic neuromas.

Although laterality analyses from different studies did not always confirm the tumor location

coinciding with the preferred side of phone use, some studies (Cardis 2011; Grell 2010; Hartikka

2009; INTERPHONE 2010, 2011; Sato 2011; Yoon 2015) found the associations between tumor

location and presumed RF exposure that indicated possible tumor-promoting effects. Further, the

higher risk for myeloid vs. lymphoid leukemia found in some studies (Cooke 2010; Kaufman 2009)

indicated possible RF-related risk modification which, similar to ionizing radiation, may involve the

innate – myeloid cell related – immunity.

As suggested by Sato (2011), possible subgroup effects raise the need for a biologically

plausible explanation for reasons why exposure to RF-EMF may exert adverse effects only on a

small fraction of ‘heavy’ users. Possible explanation, however, is compounded by a variety of

genetic and/or other confounding factors that can underlie the RF-related risk modification. While

comprehensive analysis of possible RF-related biological effects is beyond the scope of this

analysis, the NIH/NCI-based study by Rajaraman (2010) 18 can be mentioned as a cursory

reference on the existence of genetic risk factors for tumors such as glioma, meningioma and

acoustic neuroma that are conventionally considered in the context of RFR-related risk. In

addition to genetic factors, RFR-related confounding may involve behavioral and social patterns

(e.g., cognitive, sleep, eating, physical and emotional stress, exercise, etc.) that accompany use of

mobile phones and other consumer sources of RFR exposure as well as environmental exposure

to various unrelated to RFR carcinogens (e.g., pesticides).

As summarized by Corle (2012) 19, generation of decisive evidence on an association between

cell phones and tumorigenesis has been very challenging due to the evolving nature of wireless

technologies, the relative rarity and long induction periods in purported tumor types, and, the last

but not the least, the varying likelihood of genetic tumor susceptibility among individual subjects.

Due to these challenges, the risk assessment in some studies might have been affected by

opposite risk trends: a higher tumor risk among the genetically (or otherwise) predisposed

Rajaraman P, Hutchinson A, Wichner S, Black PM, Fine HA, Loeffler JS, Selker RG, Shapiro WR, Rothman N, Linet MS, Inskip PD. DNA

repair gene polymorphisms and risk of adult meningioma, glioma, and acoustic neuroma. Neuro Oncol. 2010 Jan;12(1):37-48. doi:

10.1093/neuonc/nop012

Corle C, Makale M, Kesari S. Cell phones and glioma risk: a review of the evidence. J Neurooncol. 2012 Jan;106(1):1-13. doi:

10.1007/s11060-011-0663-9

Page 47

www.fda.gov

individuals might have been counterbalanced by a comparatively reduced risk among other

members of the same study population. As a general limitation that possibly skewed the overall

RF-related risk assessment, the existing studies were likely to have a selection bias due to underrepresentation of susceptible individuals who are expected to be rare in the entire population and

who, most likely, remain unidentified. In fact, some studies (INTERPHONE 2010; Benson 20132014) intentionally excluded subjects diagnosed with neurofibromatosis as potentially prone to

acoustic neuroma.

D. Summary and Limitations of Individual Epidemiology Studies Reviewed.

This section provides a summary of each of the studies reviewed and assessed followed

by a description of limitations of the studies. The studies are listed in alphabetical order based

on the author’s name.

The INTERPHONE Study Group (2010) assessed brain tumor risk in relation to mobile

phone use by conducting an interview-based study (2000-2004, 13 countries) incorporating

cases with glioma and meningioma (2,708 and 2,409, respectively) and matched controls. In the

case- control analysis, regular mobile phone users had reduced likelihood for both glioma and

meningioma: OR=0.81 [95% CI: 0.70; 0.94] and OR=0.79 [95% CI: 0.68; 0.91], respectively.

Similarly, ORs <1.0 were shown for all deciles of lifetime number of phone calls and nine deciles of

cumulative call time. In the 10th decile of recalled cumulative call time (51,640 h), possible risk

increase was shown for glioma: OR=1.40 [95% CI: 1.03; 1.89]; a similar trend for meningioma did

not reach statistical significance. Notably, the excess risk for both glioma and meningioma

(OR=3.77 [95% CI: 1.25; 11.4] and OR=4.80 [95% CI: 1.49; 15.4], respectively) was indicated in the

highest call time category for those who started phone use 1–4 years (but not earlier) before the

reference date. In the case-case analysis of the concordance between tumor side and preferred

side of phone use, the likelihood of glioma was slightly elevated among regular users with the

ipsilateral tumor location: OR=1.27 [95% CI: 1.19; 1.37]. In addition, the likelihood of gliomas

located in the temporal lobe was shown to be greater among users with ≥1,640 h of cumulative

call time: OR=1.87 [95% CI: 1.09; 3.22].

Limitations: Although this study represents one of the main available sources for evaluating

brain tumor risk in relation to mobile phone use, overall validity of the results (including the main

conclusion on no overall risk increase) is affected by a disproportionately high number of ORs <1,

indicating a strong possibility of systematic bias (likely, due to different levels of participation as

well as differential errors in reporting among users vs. non-users). With the possibility for genuine

protective effect regarded as implausible, the authors referred to strong biases as the main

explanation for the overall reduced tumor likelihood. With the study cases recruited mainly

through clinics within the study area, selection bias could result of under- ascertainment due to

Page 48

www.fda.gov

possible cases from non-participating clinics, or a failure to identify tumors that had been

diagnosed but not treated (e.g., due to their small size). Reporting and assessment of mobile

phone use might have been affected by prodromal symptoms in subjects with brain tumor and/or

by timing of interviews. The study results might have been also affected by methodology used for

the imputation of missing data, or the assignment of the tumor location for controls. A finding of

the increased glioma risk among users with highest cumulative call time was attributed by the

authors to a recall bias, based on some ‘implausible’ values of reported use in this group. When

compared with controls, glioma cases had a higher number of imputations for missing values as

well as higher proportions of proxy respondents and subjects described as non-responsive or

having poor memory. However, sensitivity analyses suggested that these differences alone cannot

explain the results seen in the highest decile of cumulative call time. Thus, the main finding on

reduced tumor risk along with a number of elevated ORs failed to produce conclusive evidence,

prompting the need for plausible explanation of the observed effects and spurring further studies

on the RF-related risk assessment.

The INTERPHONE Study Group (2011) assessed risk of acoustic neuroma (or vestibular

schwannoma) in relation to mobile phone use by conducting an interview-based case–control

study involving 13 countries and including 1,105 study cases and 2,145 matched controls. There

was no indication of an overall risk rising with the increasing exposure measured by cumulative

call time or number of calls. However, in the 10th decile (≥1640 h) of cumulative call time, the risk

was borderline increased with censoring at up to one year before the reference date and was

further increased with censoring at 5 years: OR=1.32 [95% CI: 0.88; 1.97] and OR=2.79 [95% CI:

1.51; 5.16], respectively. In the tumor location analysis among regular users in the 10th decile of

cumulative hours of use (≥1640 h), the risk increase was indicated for ipsilateral (but not

contralateral) use when censored at 1 year and 5 years: OR=2.33 [95% CI: 1.23; 4.40] and OR=3.53

[95% CI: 1.59; 7.82], respectively.

Limitations: Similar to the previously discussed INTERHONE (2010) study on glioma and

meningioma, causal interpretation in this study was similarly impaired by several aforementioned

issues of major importance such as participation, selection, and reporting biases. Similar to

INTERHONE 2010, elevated ORs at the highest level of cumulative call time were regarded by the

authors as a chance finding resulting from the reporting bias indicated by some ‘implausible’

values reported in this category. Since acoustic neuroma is a slowly growing tumor, the exposure

interval since introduction of mobile phones might have been too short to observe an effect, if

there is one. As a rare hereditary disorder associated with a high risk of acoustic neuroma,

neurofibromatosis type II could represent an additional confounding concern specific for this and

other studies on acoustic neuromas; however, a subanalysis excluding the diagnosed subjects

(x10) did not alter the main results in this study.

Page 49

www.fda.gov

Aydin et al 2011 examined possible association between mobile phone use and brain tumors

in children and adolescents aged 7-19 years, using an international (Denmark, Sweden, Norway,

and Switzerland) case–control study (CEFALO, 2004-2008). Personal interviews on mobile phone

use were conducted with both study cases (a total of 352 patients with astrocytoma – 46%,

ependymoma – 6% or another glioma – 8.5%, primitive neuroectodermal tumors – 17.6%, other

specified intracranial neoplasms – 15.1%, and unspecified intracranial neoplasms – 6.8%) and

controls (646) randomly selected from registries and matched by age, sex, and geographic region.

Per conditional logistic regression analysis, children using mobile phones for ≤5 years were not at

increased risk compared to non-users: OR=1.26 [95% CI; 0.70; 2.28]. A slight trend for higher

frequency of brain tumors among regular users of mobile phone vs. nonusers did not reach

statistical significance: OR=1.36 [95% CI: 0.92; 2.02]. However, in the subset with available

operator-recorded data on mobile phone use, users with the longest period since first

subscription (>2.8 years) showed a risk increase compared to controls: OR = 2.15 [95% CI; 1.07;

4.29], P < .001. The overall likelihood of brain tumors showed similar borderline increases among

both ipsilateral and contralateral regular users. In the further tumor location analysis, no

increased risk was shown for regular users with tumor locations in the expected higher exposure

areas; however, a significant risk increase was shown for regular users with tumors located in the

areas other than temporal/frontal lobes and cerebellum: OR=1.92 [95% CI; 1.07; 3.44]. In the

analysis focused on tumor morphology, a borderline risk increase (OR=1.65 [95% CI: 0.93; 2.93])

was shown for regular users with tumors other than astrocytomas and other gliomas.

Limitations: Most mobile phone usage data were subject to recall bias, since they were

reported by children and adolescents (or their parents) and were not based on the operatorrecorded data. As a common limitation for population-based studies, this study was not

statistically powered to detect small risk increases. Due to the study population limited to children

and adolescents, amounts and durations of mobile phone use were relatively small. Interpretation

of the results was complicated by the study cases including various brain tumors with different

etiologies which, in case of a true effect, may have been differentially affected by mobile phone

exposure.

In the authors’ response to a commentary (Söderqvist et al 2011) that questioned the lack of

causal association between mobile phone use and brain tumors (as reported in Aydin et al 2011,

see above), Aydin et al 2012 re-examined the results of CEFALO study per age-standardized

incidence rates for brain/CNS tumors (NORDCAN). The relatively stable incidence rates in the age

group of 5–19 years living in the Nordic Countries did not confer an elevated risk from use of cell

phones that would correspond to the recent steep increase of their use by children and

adolescents. However, the authors recommended further monitoring of childhood brain tumor

incidence trends, referring to the possibility of risk increase associated with heavy mobile phone

use, very early life exposure, and rare subtypes of brain tumors.

Page 50

www.fda.gov

Limitations: This study was subject to the previously discussed limitations of CEFALO study

(Aydin et al 2011). The main conclusion was also subject to limitations common for population

based-studies which do not allow for identifying small-scale effects in certain subgroups.

Balekouzou et al 2017 examined lifestyle and behavioral factors as possible risk factors for

breast cancer among Central African women, using a case-control study on 174 histologicallyconfirmed breast cancer cases and 348 age-matched controls. A questionnaire regarding mobile

phone use was limited to the habit to keep cell phones in their bras; per unconditional logistic

regression analysis, the lack of this habit was associated with a reduced likelihood of breast

cancer: OR=0.56 [95% CI: 0.35; 0.89].

Limitations: This relatively small study was not primarily aimed at the assessment of breast

cancer risk in relation to mobile phone use; rather whether it was associated with the

participant’s habit of keeping their cell phones in their bras. A seemingly protective effect of nonhabit to keep cell phones in bras might have been due to chance and/or confounding associated

with some ethnic/ cultural factors.

Barchana et al 2012 examined the incidence and tumor laterality of gliomas in Israeli

population which was described as a population using mobile phone technologies very

extensively. All glioma cases from the national cancer registry (1980-2009) were classified as lowor high-grade per their proliferative potentials. Analysis of tumor laterality in relation to mobile

phone use was performed based on a questionnaire that was sent to randomly-selected adults

(≥18 years). When measured for 5-year intervals, incidences of low-grade gliomas (mostly,

presented by astrocytomas) decreased from the period of 1980-1994 to the period of 1995-2009.

From 1995 onward, a shift towards left-sided location was found for all gliomas (in both grade

categories), especially among subjects aged 20-49 years (p=0.03). This trend appeared

contradictory to the survey results showing that 45% controls from the general population used a

mobile phone only on the right side, compared to only 17% who used the phone (exclusively or

mostly) on the left side. A significant decrease in low-grade gliomas in the 30-years period that

coincided with a sharp increase of mobile phone use did not support the assumption of causal

etiological association. However, in the period after mobile phone introduction (1994-2009) there

was an increasing trend for high-grade gliomas including malignant glioblastoma multiforme,

especially among females (p=0.01).

Limitations: The study lacked individual information on mobile phone use among the study

cases with glioma, as the self-reported questionnaires were available only for randomly-selected

subjects who represented the comparator group. Since the study population was limited to a

single ethnic group, the study results on the increase of high-grade gliomas coinciding with mobile

phone introduction might have reflected possible genetic and ethnic/cultural confounding effects.

Page 51

www.fda.gov

Benson et al 2013 used a UK prospective cohort from the Million Women Study to examine

possible relation between mobile phone use and incidence of intracranial CNS tumors and some

other cancers among middle-aged women. Mobile phone use information was collected for the

periods of 1999-2005 and 2009. In the corresponding 7-year period, 51,680 incident invasive

cancers and 1,261 incident intracranial CNS tumors occurred. Based on adjusted RRs from Cox

regression models, overall cancer risk among ever vs. never users of mobile phones was not

increased when assessed for all intracranial CNS tumors, specified CNS tumor types, or non- CNS

cancers at other (18) sites. No appreciable associations were found among long-term users (>10

years) compared to never users with regards to glioma or meningioma. Risk for acoustic neuroma

was initially reported as elevated among the long-term users (RR=2.46 [95% CI: 1.07; 5.64]), but

this finding was retracted in a subsequent publication by the same authors (see Benson 2014

below). While the rapidly increasing prevalence of mobile phone use in the study cohort was

consistent with the steep increase in mobile phone subscriptions in the UK from the early 1990s

to 2003, little evidence was found to suggest a proportional increase in the incidence of acoustic

neuroma. However, the RR for pituitary tumors was increased in short-term (<5 years) mobile

phone users: RR=2.31 [95% CI: 1.31; 4.06], without a further risk increase with the increasing

duration of use.

Limitations: Since individual information on mobile phone use was collected in a prospective

cohort (i.e., the Million Women Study), this study is considered to be less prone to the

shortcomings of retrospective reporting on exposure (e.g., recall bias). However, the survey on

mobile phone use might have been still affected by participation and reporting biases. In a survey

conducted between 1999 and 2005, only 65% of women recruited in 1996–2001 replied, and

mobile phone use reported at baseline may have changed subsequently. Considering that brain

tumor risk in general is associated with high socioeconomic status, the risk increase for acoustic

neuroma might have particularly affected by confounding due to a higher proportion of

Socioeconomic group (% in upper third) among ever vs. never users of mobile phone users (35.7%

and 29.6%, respectively). Note: the study excluded 6 women who reported having

neurofibromatosis which is known as possible cause of acoustic neuroma.

In the authors’ response to a commentary (de Vocht et al 2011) that questioned possible risk

elevation for acoustic neuroma among mobile phone users (as reported in Benson et al 2013, see

above), Benson et al 2014 re-examined their previously reported finding based on the updated

follow-up data from 2011. As shown by the re-assessed RRs among users >10 years vs. never

users, an association between long-term mobile phone use and acoustic neuroma was no longer

significant: RR=1.17 (95% CI: 0.60; 2.27), with no trend for risk increase with duration of use

(P=0.3). Note: this publication did not refer to a re-analysis of the previously reported findings on

pituitary tumors.

Page 52

www.fda.gov

Limitations: This commentary is subject to the limitations of the initial study (Benson et al

2013) discussed above.

Cardis et al 2011 examined risk of brain tumors in relation to RF exposure estimates that

were derived from the data on mobile phone usage from Australian, Canadian, French, Israeli, and

New Zealand components of the INTERPHONE study. The study included subjects aged 30-59

years with glioma or meningioma (533 and 676, respectively; diagnosed between 2000 and 2004

and located by neuroradiologists) and corresponding controls (1762 and 1911, respectively)

matched by age, sex, geographic region, and the assigned ‘tumor location’. An algorithm was

developed to estimate individual RF dose as cumulative specific energy absorbed at a given

location in the brain for a given frequency band and communication system. Analysis of putative

modifying factors included laterality of use, hands-free devices, network characteristics and,

where appropriate, frequency of use in urban and rural settings. RF dose was based on multiple

RF exposure determinants and was estimated as TCSE (J/kg) absorbed at the tumor’s estimated

center. Overall likelihood among cases ever having been a regular mobile phone user was not

significantly changed either for glioma or meningioma. However, the likelihood for glioma was

borderline increased among cases with the highest TCSE quintile: OR=1.35 [95% CI: 0.96; 1.90].

The likelihood further increased with the increasing TCSE in cases with >7 years of use before

diagnosis (Ptrend=0.01), reaching OR=1.91 [95% CI: 1.05; 3.47] in the highest quintile. When

compared with their counterparts located elsewhere in the brain, the likelihood of gliomas

located in the most exposed areas of the brain was increased in cases with >10 years of mobile

phone use: OR=2.80 [95% CI: 1.13; 6.94]. No increased likelihood was found for meningioma,

despite some similar to glioma patterns.

Limitations: The study was potentially subject to the initial limitations of INTERPHONE study

as discussed above as well as to additional selection bias since the study sample included only

tumors assessed by neuroradiologists. Accuracy of the RF energy deposition might have been

affected by reporting bias, especially in cases when a subject had no preference for laterality of

use, or when it remained unknown. Although the study claimed improved ascertainment of RF

exposure and no difference in accuracy between computer-based and neurologist-based

assessments of the tumor’s center, the developed algorithm and the study results were subject to

considerable uncertainty and require replication before they can be considered as indication of a

cause-effect relationship.

Carlberg and Hardell 2014 examined survival of patients with glioblastoma multiforme in

relation to use of mobile and cordless phones, using 1,678 glioma patients from their previous

registry-based case-control studies (Sweden, 1997–2003 and 2007–2009). Cell phone use was

assessed using mailed questionnaires. Based on the Cox proportional hazards model, use of cell

Page 53

www.fda.gov

phones with the latency >20 years since first use was associated with increased HR for glioma:

HR=1.7 [95% CI: 1.2; 2.3]. With the same latency, this association appeared to be stronger among

cases with astrocytoma grade IV, or glioblastoma multiforme (n=926), with HRs reaching 2.0 [95%

CI: 1.4; 2.9] and 3.4 [95% CI: 1.04; 11] for use of mobile phones and cordless phones, respectively.

HR was highest for cases with first use of both mobile and cordless phones before the age of 20

years, with the HR for wireless phones reaching 2.3 [95% CI: 1.1; 4.7].

Limitations: In addition to recall bias which is common for retrospective studies and which

may have affected the accuracy of phone usage details, the study was subject to other biases

involving participation, reporting, selection (e.g., exposed cases were younger than unexposed

cases, p < 0.0001) as well as interpretation (e.g., an unlikely “protective” effect of cell phone use

on survival in low-grade glioma). Putative differential survival effects in glioblastoma vs. low-grade

glioma might have reflected a selection bias resulting in overrepresentation of high-grade gliomas

in the study sample. The main finding of decreased survival among cases with high- grade gliomas

may be due to unaccounted confounding factors rather than the alleged use of cell phones.

Carlberg and Hardell 2015 examined risk of meningioma in relation to use of cell phones, by

using their previous case-control studies (Sweden, 1997-2003 and 2007-2009) to perform a

pooled analysis of 1,625 meningioma cases and 3,530 controls randomly selected from the

Swedish Population Registry. No overall associations with use of mobile or cordless phones were

found, per unconditional logistic regression analysis adjusted for sex, age, year of diagnosis, and

socioeconomic index. However, in the fourth quartile of use (>1,436 h), a borderline increased

likelihood was found for both mobile and cordless phones: OR=1.2 [95% CI: 0.9; 1.6] and OR=1.7

[95% CI: 1.3; 2.2], respectively. Similar trends indicating possible risk elevation were also shown

for the highest decile (>3,358 h). No significant risk associations were found with regards to the

longest latency time, ipsilateral use, or anatomical tumor location.

Limitations: In addition to conventional recall bias inherent for retrospective studies, this

study might have been affected by selection bias due to exclusion of clinically-diagnosed cases

with no histopathological confirmation. In addition, the study might have been affected by recall

and classification biases which are typical for this type of studies and which might have resulted in

misclassified exposure to mobile phones.

Carlberg et al 2016 examined incidence of thyroid cancer (NORDCAN; Swedish Cancer

increasing incidence of thyroid cancer in Sweden and the Nordic countries was shown, with

joinpoint regression analysis detecting different joinpoints for men and women and a common

joinpoint in 2006. The incidence of thyroid cancer (mostly, papillary) increased during 2006– 2013

in both women and men, with average annual percentage changes (APC) of +6.16 % [95 % CI:

+3.94; +8.42 %] and +6.84 % [95 % CI: +3.69; +10.08 %], respectively. Comparison of the

Page 54

www.fda.gov

incidences of thyroid cancer in Nordic countries in relation to the available information on mobile

phone use (i.e., total minutes of out-going mobile phone calls) for the same time period showed

concordant increases, with a lag time for the thyroid cancer incidence.

Limitations: The study design did not permit assessment regarding causality and the study

lacked individual information on tumor characteristics and mobile phone use. Use of data derived

from the Swedish Cancer Register might have resulted in underreporting and selection bias due to

the lack of histopathologically confirmed cancer diagnoses in all registry cases.

Chapman et al 2016 examined putative association between brain cancer incidence and

mobile phone use, using national cancer registration data (Australia). Age/sex-specific incidence

rates were calculated among 19,858 males and 14,222 females diagnosed between 1982 and

2012; the derived rates were analyzed in conjunction with mobile phone usage from 1987 to

2012. Brain cancer incidence rates (20–84 years; per 100,000) were stable over 30 years in

females, but slightly increased in males (p < 0.05). Assuming a causal RR of 1.5 and a 10-year lag

period, the expected incidence rates in 2012 were estimated as 11.7 [95% CI: 11; 12.4] and 7.7

[95% CI: 7.2; 8.3] for males and females, respectively. However, this assumption resulted in a

much higher estimate for expected brain cancer cases in 2012 compared the observed cases

(1,867 and 1,434 cases, respectively). Per the age-stratified analysis using modelled rates,

significant increases in brain cancer incidence were observed only in the ≥70 years group (both

sexes), with the increase starting in 1982, i.e., before the introduction of mobile phones. Thus, the

overall brain cancer incidence rates were considered unlikely to be related to mobile phone use;

the observed partial increases were attributed to improved diagnostics due to introduction of

new imaging technologies in the early 1980s.

Limitations: As conventional limitations for this type of studies, no data on individual mobile

phone use and cancer-related outcomes were available. The results were subject to potential

assumption bias due to the assumed RRs and lag period (i.e., 1.5 in ever-users and 2.5 in ‘heavy’

users of mobile phone, with a 10-year lag period).

Cooke et al 2010 examined risk of leukemia in relation to mobile phone use, using a casecontrol study (South-East England) involving 806 incident leukemia cases diagnosed between

2003 and 2009 at ages 18–59 years (acute myeloid leukemia – 449, acute lymphoblastic leukemia

– 125, chronic myeloid leukemia – 154, hairy cell leukemia – 55, acute monocytic leukemia – 7,

and other/mixed – 16) and 585 controls. Information on mobile phone use and other potential

etiological variables was collected via interviews. Based on logistic regression analyses considering

age, sex, ethnicity, area of residence, smoking, socioeconomic status as well as interview lag time,

no significant association was found between overall mobile phone use and risk of leukemia:

OR=1.06 [95% CI: 0.76, 1.46]. However, a borderline risk increase was found among subjects who

Page 55

www.fda.gov

started using a phone ≥15 years ago: OR=1.87 [95% CI: 0.96; 3.63]. In the analysis stratified by

leukemia type, there was a borderline risk increase for acute myeloid leukemia among subjects

with the same latency: OR=2.08 [95% CI: 0.98; 4.39].

Limitations: The study results were likely affected by participation bias (only 50% of eligible

cases participated in the study) and selection bias (controls were comprised of 585 non- blood

relatives provided by 392 cases). Controls were not individually matched to cases, and the

reference dates for controls were constructed by stratifying cases by half-year or whole-year

interview lag times and then randomly allocating controls in each half-year interview period to lag

time strata in the same proportions as the cases. As an indication of possible heterogeneity in the

study sample, controls had less non-white and 55-59-year aged subjects, compared to cases. A

possible recall bias with overestimated phone use in more distant time periods could contribute

to the slightly increased trend in the highest category of cumulative hours of phone use.

Corona et al 2012 examined risk factors associated with vestibular schwannoma, by

conducting a case-control study (Brazil, 2006-2010) involving 44 cases and 104 controls. The study

used a questionnaire which, among various potential risk factors, included exposure to nonionizing radiation such as cell phone use and residing near a cell phone tower. While a history of

chickenpox and exposure to >1 cranial X-ray procedure were identified as potential risk factors, no

significant risk changes were found regarding cell phone use (including laterality) or living near a

cell phone tower.

Limitations: The study results, especially from stratified analyses, were undermined by the

relatively small sample size. Since vestibular schwannoma is not a tumor type that must be

reported in Brazil, the sample size was restricted by difficulties with locating potential study cases

which was likely to cause selection bias. In addition, the study might have been affected by recall

and classification biases which are typical for this type of studies and which might have been

particularly caused by possible memory problems among study cases. The study results might

have been also affected by the long latency period of vestibular schwannoma.

Coureau et al 2014 analyzed possible association between mobile phone exposure and

primary CNS tumors, using the multicenter population-based case-control CERENAT study (France,

2004–2006) on various potential etiologic factors for CNS tumors in adults. A study sample

consisted of a total of 1,339 subjects (253 cases and 504 controls for gliomas; 194 cases and 388

controls for meningiomas), with controls matched on age, sex, and place of residence.

Information on mobile phone use was collected via face-to-face interviews. Per conditional

logistic regression analysis, no significant risk associations were observed when comparing regular

mobile phone users with non-users: OR=1.24 [95% CI: 0.86; 1.77] and OR=0.90 [95% CI: 0.61; 1.34]

for gliomas and meningiomas, respectively. However, ‘heaviest’ users defined by life-long

Page 56

www.fda.gov

duration (≥896 h) were shown to have increased risks for both gliomas and meningiomas:

OR=2.89 [95% CI: 1.41; 5.93] and OR=2.57 [95% CI: 1.02; 6.44], respectively. ‘Heavy’ use defined

by cumulative number of calls (≥18 360) was associated with increased risk for gliomas only:

OR=2.10 [95% CI: 1.03; 4.31]. Elevated risk for gliomas (but not meningiomas) was further

suggested for users in the last decile compared to non-regular users. With the latency of 5 years,

the OR for gliomas reached 5.30 [95% CI: 2.12; 13.23]. The highest risk increase was shown for

gliomas with locations other than temporal and frontal lobes: OR=3.61 [95% CI: 1.00; 12.96]. The

risk of glioma was shown to triple with occupational use of mobile phones (OR=3.27 [95% CI: 1.45;

7.35]) and was increased even greater with exclusively urban setting (OR=8.20 [95% CI: 1.37;

49.07]). A positive risk trend for ipsilateral glioma tumors did not reach statistical significance:

OR=2.11 [95% CI: 0.73; 6.08].

Limitations: The CERENAT study was not exclusively focused on mobile phone use. The study

results might have been affected by participation bias (e.g., the rate for controls was only 45%,

reaching 66% and 75% for glioma and meningioma cases, respectively). As an indication of

participation and selection biases, non-included cases were older than included, and the level of

education was higher in controls than in cases (p<0.001). The delay between index date and

interview was longer for controls. Due to health status of some cases, a number of interviews had

to be conducted using a proxy and a simplified questionnaire. The quality of data was further

affected by the identified recall bias regarding exposure. The study did not include

histopathological and radiological assessment.

Czerninski et al 2011 examined risk of parotid malignant tumors by analyzing data from the

Israel National Cancer Registry (1970–2006). The annual incidence of salivary gland cancer was

estimated as 0.8 cases per 100,000, with the most common subtype being salivary gland

carcinoma (60%). The total number of parotid gland cancers increased 4-fold (with the steepest

increase after 2001), whereas other major salivary gland cancers remained stable. After excluding

population growth as an explanation of the increased incidence of parotid gland cancers, the

authors speculated about possible role of cell phones, referring to a 6-fold increase in their usage

in the Israeli population from 1997 to 2006.

Limitations: The study was methodologically weak: as the incidence analysis for parotid

tumors had no corresponding data on estimated or actual mobile phone exposure.

de Vocht et al 2011(a) examined possible association between use of cell phones and

parotid cancer by using publicly available data from the UK Office of National Statistics (1986–

2008). Rates of malignant neoplasms of the parotid gland increased from 0.5 to 0.8 cases per

100,000 in men (Ptrend <0.01) and from 0.4 to 0.6 per 100,000 in women (Ptrend <0.01). The number

of new cases in the period between 1986 and 2007 increased for both sexes, more than doubling

Page 57

www.fda.gov

in men. Rates of malignant neoplasms of other and unspecified salivary glands remained stable in

men and increased only slightly in women. Meanwhile, cell phone use in England has increased

dramatically as shown by the number of subscribers increasing from 50,000 in 1985 to over 52

million in 2003. However, no causal association was inferred, since the rates of malignant parotid

gland neoplasms started increasing before widespread use of cell phones.

Limitations: Reflecting typical limitations for ecological trend analyses, the study lacked data

on individual mobile phone exposure among subjects with parotid tumors and lacked evidence

inferring causal association.

De Vocht 2016 analyzed the incidence of some brain cancer subtypes (England, 1985-2014)

in comparison to counterfactual ‘synthetic control’ timeseries. Annual incidence of malignant

glioma, glioblastoma multiforme, and malignant neoplasms of the temporal and parietal lobes

were modelled based on population-level covariates, using Bayesian modeling and assuming 5-,

10- and 15-year latency periods. Hypothetical impact of mobile phone use was inferred from

differences between measured and modelled time series. No increasing trends were found with

regards to malignant glioma, glioblastoma multiforme, or malignant neoplasms of the parietal

lobe. However, malignant neoplasms of the temporal lobe showed a faster than expected

increase of 35% [95% CI: 9; 59] during 2005–2014, which was interpreted as an indication of

possible association with the penetration of mobile phone technologies.

Limitations: The incident brain cancer cases were not linked to individual mobile phone use,

and the study results may have been affected by methodological artefacts due to unconventional

study design. As an example of possible reasons for assumption and interpretation bias, the

inference about an effect relies on the assumption that the relationship between covariates and

the incidence time series prior to the cut-off remains stable.

de Vocht et al 2011(b) examined rates of newly diagnosed brain cancer cases (per the UK’s

Office of National Statistics, 1998-2007) in relation to mobile phone use. No time trends were

found in overall incidences of brain cancers in either sex- or age-stratified groups. Unlike

decreased rates for cancers of the parietal lobe, cerebrum, and cerebellum in men, rates for

cancers of the temporal lobe were increased in both men and women (0.04 and 0.02 new

cases/year, respectively). However, no evidence connecting the incidence of brain cancer (19982007) to prior increase in use of mobile phones (1985-2003) was found. If caused by mobile

phone use, the increase of cancers in the temporal lobe was estimated as <1 additional case per

100,000 people.

This text is long and has been trimmed here. Open the source document for the complete record.

This is a copy of a public record, reproduced as it was published. It is not legal advice, and it may not be the version a court would rely on. Check the official source before you cite it.