Karipidis et al. [1] (hereinafter: Karipidis) published a scoping review investigating
radiofrequency (RF) studies in the range >6 GHz, with a particular focus on the millimetre
wave (MMW) band. The Karipidis review was performed against a backdrop of rising public
concerns associated with the health and safety of 5th generation (5G) wireless technology
[2]. Subsequently, the telecommunications industry is now using the Karipidis review
to suggest “no evidence of adverse health effects from the radio waves used in 5G
including mmWave” [3]. Notwithstanding the fact that no studies have investigated
specific 5G frequencies and modulations, does the Karipidis review stand up to scrutiny
in providing assurances of safety (no evidence of harm) that industry is suggesting?
The analysis herein reveals that it does not.
A host of study design weaknesses in the existing literature were critiqued throughout
the Karipidis review. In spite of the apparent lack of rigour attributed to many papers,
Karipidis concluded that “experimental studies provided no confirmed evidence that
low-level MMWs are associated with biological effects relevant to human health” and
similarly, that radar-related epidemiological studies “presented little evidence of
an association between low-level MMWs and any adverse health effects”.
This line of reasoning parallels that used previously by scientists working for the
tobacco industry, whose studies repeatedly arrived at conclusions suggesting no clear
determination of harm could be made [4]. This was part of a broader strategy of manufacturing
doubt about the potential negative health effects of their product, as summarised
by Gilbert [5]:
“The very nature of scientific exploration is to ask and answer the next question.
But rather than accepting the process of scientific discovery, business interests
press to have every tiny bit of uncertainty explored before any policy decision can
be made, demanding proof rather than precaution—in fact, they even manufacture uncertainty.
As a result, decisions are not made; policy is not advanced; problems are not addressed.”
A similar ethos is observed with the handling of scientific evidence by some governments
and associated regulatory bodies in regards to radiofrequency exposures and health
risks [6, 7]. The same was noted by the US Court of Appeals in the recent case against
the FCC [8].
Surprisingly, the Karipidis review did not identify and discuss potential risk implications.
This is of significant importance, because as Karipidis noted, the use of RF frequencies
above 6 GHz is only just beginning. Best practice demands a risk management approach
for the identification of all potential hazards and implementation of mitigation strategies
to address these risks. This is already the case with low-dose ionizing radiation
[9], but is sorely neglected for non-ionizing RF radiation [10]. Rather than waiting
for harm to be established before acting, a precautionary approach to risk management
is necessary [6, 11].
Detailed analysis of Karipidis tables and selection of papers
In order to conduct an independent assessment of the Karipidis review, we performed
our own literature search using the same international research libraries as Karipidis,
and also accessed the Oceania Radiofrequency Scientific Advisory Association (ORSAA)
database (ODEB) [12].
To perform our assessment, the Karipidis review was critiqued and classified into
different categories as summarised in Tables 1 and 2. The full set of Karipidis tables,
our corresponding review comments and analysis can be downloaded from the ORSAA website
[13]. The results reveal issues of potential bias as well as questions around the
completeness and thoroughness of the work conducted by Karipidis.
Table 1
Analysis summary of Karipidis et al. experimental study review (covering Tables 1–6 in
their review publication).
Critique Category
Description
Incorrect Biological System
Karipidis specified “Bacteria and Yeast” in more than thirty experimental studies
that were reviewed. This classification is not entirely appropriate because bacteria
and yeast are different species and because most experiments typically expose either
yeast (fungi) or bacteria, not both. A more generic description, if needed, could
have been “Microbes”. However, such a level of abstraction would prevent a detailed
analysis from identifying potential study replications. This is also important when
it comes to potential resonance effects, as an example: DNA of yeast will have a different
molecular weight compared to DNA from bacteria and so will likely respond to different
but specific resonance frequencies.
Incorrect Exposure Time or Exposure Time Range
There were twenty instances where a discrepancy between the exposure duration that
Karipidis indicated in their review tables and the duration specified in the reviewed
papers was found.
Incorrect Frequency/ Incorrect Frequency Range/Missing Frequency
There were thirteen instances where discrepancies were identified between the exposure
frequency Karipidis indicated within their tables and the exposure frequency specified
in the papers they reviewed.
Incorrect Intensity/ Incorrect Intensity Range
Eighteen discrepancies were found between the exposure intensity Karipidis documented
and the actual exposure intensity specified in the papers reviewed.
Misclassified/ Questionable Classification
There were seventeen instances where the inclusion of a study in a particular table
in relation to biological endpoint relevance was questionable. Examples include gene
expression studies being included in the genotoxicity table and vice versa.
Misstatements
There were fourteen instances where Karipidis has incorrectly stated a study finding
or parameter. This is a serious issue particularly in the cases where a statistically
significant finding was mis-reported as a no effect. This has direct implications
for a linked study from Wood et al. [26].
Nonsensical Quality Issues
The validity of a number of quality issues raised by Karipidis can be challenged.
This has direct implications to a linked study from Wood et al. [26]. Our analysis
shows Karipidis has performed a quality assessment of other’s work yet their own publication
suffers from serious quality deficiencies.
Findings Not Reported/ Incomplete Results
Our analysis identified forty two instances where important statistically significant
biological effect findings are not disclosed by Karipidis in the results column or
included in the 5G health review discussion. This has important implications for understanding
biological effects that RF exposure (>6 GHz) has on biological entities and the health
implications, if any, that may arise.
Table 2
Analysis summary of Karipidis et al. epidemiological study review (Table 7 in their
review publication).
Critique Category
Description
Undisclosed Disease Risk
There were a total of eight instances identified in our analysis where epidemiological
studies that identified important health risks, such as specific cancers, were not
disclosed in the disease column for studies in Table 7 of the Karipidis 5G Health
review.
Incorrect Case/Personnel Numbers
There were a number of instances where studies presented by Karipidis had case/personnel
number discrepancies not matching actual published numbers or included case and control
counts not included in the scope for review (i.e., <6 GHz).
Incorrect Odds Ratio Assignment
One instance was found where an Odds Ratio (OR) for an occupational exposure provided
by Karipidis was not in the scope of review (wrong exposure type i.e., <6 GHz).
Limitations Misstatement
There were two instances where Karipidis claimed “no information on confounding factors”
was found to be incorrect.
Risk Estimate Issue
Karipidis has incorrectly specified Odds Ratio in three instances where a different
type of rate or ratio was used.
Critique summary
Examination of the Karipidis 5G health review reveals many errors in classification
and analysis. Some are minor, and although indicating a lack of diligence, they have
no substantial implications for the outcomes identified in the papers reviewed. Of
much greater concern are the number of misstatements, misclassifications, and exclusions
of important findings from sound research.
The Karipidis review is at best a superficial analysis of a restricted set of available
publications investigating exposures to radio frequencies in the >6 GHz range. No
attempt has been made to understand or reconcile differing study outcomes. Karipidis
has simply restated the results for specific endpoints, showing papers that have demonstrated
statistically significant effects and those that have not. Divergent findings have
been used to suggest ‘inconsistency’ as a problem, thereby diminishing the importance
of biological effect findings. In contrast, our assessment [13] provides rational
justifications to explain some of the divergent findings. We have also previously
discussed a number of physical and biological variables, which underlie the different
outcomes from studies investigating biological effects of RF exposures in general
[14], and MMW exposures in particular [15]. Karipidis also attributed ‘quality’ deficiencies
to a number of studies that are unjustified [13].
A literature search identified a significant number of relevant papers (at least 70
experimental papers and 16 epidemiological papers available from PubMed and ODEB)
were missing from the Karipidis collection. These papers cover all major themes presented
by Karipidis and more, with the majority showing statistically significant effects.
By restricting the paper selection criteria, the balance of evidence can be skewed.
A lack of transparency regarding papers found and ultimately discarded by Karipidis
means that selection bias cannot be excluded.
Also missing from the Karipidis review is an analysis of potential publication and
funding biases, which would allow the reader to assess how such influences affect
study outcomes. This is often very obvious. For example, on a related topic, Carpenter
[16] found that evidence for magnetic fields increasing the risk of cancer is neither
inconsistent nor inconclusive (from government or independent studies), yet almost
all industry supported studies fail to find any significant or even suggested associations.
A similar industry funding study bias was observed with mobile phones [17].
The biased selections and assessments that have been uncovered in the Karipidis review
create an unbalanced view of the science, and skew the final conclusion towards uncertainty.
In contrast, when appraising all relevant findings, the evidence found in our review
points to risks not fully considered by the International Commission on Non-Ionizing
Radiation Protection (ICNIRP) or the Australian Radiation Protection and Nuclear Safety
Agency (ARPANSA) in their respective RF guidelines and RF standards. These guidelines
do not reflect the current state of scientific knowledge and are based on acute heating
protection only [10], which is purely for regulatory convenience. The gulf between
thermal and non-thermal evaluative frameworks has previously been discussed [6, 18].
Other important works have been omitted from the Karipidis review. Epidemiological
studies suggest that RF exposures from other technologies such as radar are associated
with an increased risk of hemolymphatic cancers [19], and experimental studies investigating
genotoxicity in blood cells [20] have found the same. Such converging evidence requires
an immediate focused investigation into RF bioeffects rather than dismissal. Other
health risks potentially linked to RF exposures include pregnancy complications, fertility
impairment, testicular cancer and brain cancer. These are identified in our analysis
[13] and will be discussed in a future paper.
Other researchers [21] agree that the current peer reviewed science points to “predictable
harm to life forms within mixed frequency mesh networks with negative consequences
likely over time”. Russell assessed the literature on MMW effects on skin and eyes,
the immune system, gene expression, and bacterial antibiotic resistance. Because of
the shallow penetration of MMW, the skin and eyes are of significant concern. More
than a decade ago, research by Feldman et al. [22] indicated that sweat ducts in the
skin could behave as antennas and thus respond to MMW. The same group [23] later stated
that there is enough evidence suggesting that helical sweat ducts in conjunction with
wavelengths approaching the dimensions of skin layers could lead to non-thermal biological
effects.
Finally, the Karipidis review lacks representation of many species, including plants,
amphibians, birds, domestic animals and most importantly, insects. Therefore, readers
are provided with little to no understanding of how MMWs impact these important ecological
entities. This is a significant gap.
Discussion
The above critique of the Karipidis review raises a number of ‘red-flags’. These require
clarification and clear justification before telecommunications companies are given
carte blanche to begin rolling out novel modulated signals to which biological systems
have never been exposed.
Karipidis has conducted an investigation resulting in the exclusion of important findings,
while also overemphasising quality deficiencies and inconsistencies in the data, thereby
suggesting confirmation bias. Di Ciaula [24] argues that underestimating the relevance
of available results (in particular those from in vitro and animal models) is ethically
unacceptable, and is equivalent to saying that potential hazardous effects can only
be assessed after the agent has had time to exert its harmful effects.
In this regard, Gee’s discussion [25] of risk assessment is pertinent. In “late lessons
from early warnings” a variety of case studies spanning chemicals, physical agents,
pathogens, and environmental issues illustrate how timing is critical for risk analysis
and application of the precautionary principle. In all cases, precautionary action,
or foresight based on a lower strength of evidence, would have lowered the burden
of disease, reduced unnecessary suffering and prevented many premature deaths.
Conclusion
In our opinion, the Karipidis review provides insufficient evidence of safety, which
is being used by Industry [3] as justification for the planned densification and ubiquitous
use of radiofrequencies >6 GHz as part of the 5G rollout. However, we concur with
Karipidis that future experimental studies “should improve the experimental design”
and “epidemiological research should continue to monitor long-term health effects
in the population related to wireless telecommunications”.
The Karipidis review seemingly equates risk management with the need to confirm evidence
of harm. The point at which harm becomes a public issue is far too late, given the
size of the population being exposed without formal consent. We consider that risks
to humans and the environment identified in past epidemiological studies [13], as
well as unknown risks yet to be identified, warrant the application of a precautionary
approach.
We find the Karipidis review to be both inadequate and incomplete, sending the wrong
messages regarding safety assessment and public health.