The methodology for systematic reviews for in vivo studies, described in the ‘Handbook for Conducting a Literature-Based Health Assessment Using OHAT Approach for Systematic Review and Evidence Integration (https://ntp.niehs.nih.gov/ntp/ohat/pubs/handbookmarch2019_508.pdf)edited by the National Toxicology Program (NTP) - Office of Health Assessment and Translation (NTP-OHAT, 2019), will be applied.
Experts from the Cochrane Collaboration (https://www.cochrane.org/),
Navigation Guide (https://ehp.niehs.nih.gov/doi/10.1289/ehp.1307175),
GRADE Working Group (https://www.gradeworkinggroup.org/),
CAMARADES (http://www.dcn.ed.ac.uk/camarades/default.htm),
SYRCLE 1(https://norecopa.no/3r-guide/systematic-review-centre-for-laboratory-animal-experimentation-syrcle)
and others were consulted to write this guide, aimed to assess new methods for evidence-based evaluation of nonhuman toxicological studies, including mechanistic studies, starting from the aspects of existing methods for a systematic review. Other handbooks will be consulted to upgrade the methodology regarding specific aspects of experimental studies (Higgins and Green 2011; Hooijmanset al 2014).
This protocol adheres with the preferred reporting items for systematic review and meta-analysis protocols statement (PRISMA-P) (Moher et al., 2015; Shamseer et al., 2015).
4.1 Eligibility criteria
The eligibility criteria were defined using the Population, Exposure, Comparison, Outcome (PECO) strategy (NTP, 2015 a, b).
4.1.1.Types of populations
Only articles reporting experimental studies on rodents of both sexes, of all ages and species, of all genetic background (wild type, transgenic and tumor-prone animal models) will be included in this review.
4.1.2. Types of exposures
We will include studies where experimental animals were exposed to electromagnetic fields (100 kHz-300 GHz) alone, or in combination with other physical or chemical agents. The exposure systems details, exposure modality and dosimetric assessment should be reported. Studies regarding exposure to extremely low frequency (ELF), infrared, visible and ultraviolet (UV) radiations, as well as theranostic applications, will be excluded from this review. Papers with lack of dosimetric information or with the use of inappropriate sources for the generation of the incident field will be considered not eligible.
4.1.3. Types of comparators
To be eligible for inclusion, studies should report outcome data in the exposed groups and sham control group (a control group simulating all environmental conditions and stress factors of exposed animals, but in absence of RF-EMF exposure).
4.1.4. Types of outcomes
All types of cancer will be considered as well as all tumor-related outcome measures (e.g. incidence, tumor multiplicity, tumor volume, progression), while articles concerning genotoxicity and oxidative stress only will be excluded.
4.1.5. Types of studies.
Only peer-reviewed articles written in English will be considered; all publication dates will be included. Reviews will be excluded but retained as a check of the bibliographic research.
4.2. Information sources and search strategy
4.2.1.Electronic academic databases.
The following databases will be used for the search:
PubMed: www.pubmed.ncbi.nlm.nih.gov;
EMF-Portal: www.emf-portal.org/en.
Only English language peer-review papers, without restrictions on the year of publication, will be included in the review.
The search strategy (query) will be composed of seven distinct elements, separated by appropriate logical operators:
-
RF-EMF (identifying exposure);
-
In-vivo studies, animal studies, rodents, mice, rats (identifying population);
-
Carcinogenicity, cancer/tumor/neoplasia, neoplastic and non-neoplastic lesions, tumor induction/co-promotion (identifying outcome);
-
Body organs and tissues (identifying biological target);
-
Elements intended to exclude sources different from RF-EMF emissions: static and ELF electric and/or magnetic fields, UV radiations alone, infrared and visible radiation, ionizing radiation (identifying exclusion criteria for exposure item);
-
Elements intended to exclude theranostic applications: RF and microwave (MW) ablation, hyperthermia and MW imaging (identifying exclusion criteria for study type);
-
Elements intended to exclude other studies (i.e. observational).
4.2.2.Hand-searching.
The reference lists of the selected papers and reviews on the topic, including descriptive reviews carried out by international panels of experts since 2011, will be screened for potentially relevant papers that may have escaped the first search.
The NTP-OHAT Handbook (March 2019) specifies that it is possible to identify relevant publications that are not commercially published or are not readily available to the public yet. These publications (e.g. technical reports from government agencies or scientific research groups, working papers from research groups or committees) may include or summarize unpublished data. If present, their bibliographies will be scanned to identify other references escaped from database searches too.
All references coming from these last sources will be marked as “provided from other sources” in the study selection flow diagram.
4.3. Selection process
All potentially relevant articles will be screened for eligibility in two stages: a first stage in which the articles will be selected, on the basis of title and abstract, by three authors (RP, PV, PG), a second stage, in which the full text of the remaining papers will be independently reviewed by two groups of investigators, each composed by one biologist and one expert in EMF dosimetry (RP, PG group 1 and LA, PV group 2). Disagreements and technical uncertainties will be discussed and resolved between review authors.
The exclusion criteria will be prioritized according to the following items:
-
Not an original full research paper (e.g. review, editorial, letters);
-
No English paper;
-
Not animal (rodents) studies;
-
No cancer endpoints;
-
Exposure out 100 kHz − 300 GHz range;
-
Lack of dosimetric information;
-
No sham/control group;
-
Theranostic application.
4.4. Data collection process
4.4.1 Data extraction
The data extraction form will be defined and agreed upon before the start of papers analysis. The eligible papers will be equally divided between group 1 and group 2 to independently extract numerical data from text, tables, or figures of each article. In case of missing data of the outcome measures, if possible, the reviewer team will contact the authors at least once by mail.
4.4.2. Data items
The extracted data will include:
-
study design (number of experimental groups, control group(s), number of animals per group, randomization and blind),
-
study type (e.g. cancer induction, co-carcinogenesis studies),
-
animal model, species, strain, sex, type of animal (wild type, transgenic),
-
timing of treatment (i.e. hours for days, days for weeks and total period; age or life stage at the start of dosing),
-
exposure details i.e. frequency, modulation, dose and type of exposure (whole body vs localized exposure, restrained vs freely moving animals), exposure system,
-
primary outcome(s): all tumor-related outcome measures as incidence, tumor multiplicity, tumor volume, progression, latency, survival,
-
secondary outcome(s): all parameters related to animal health conditions evaluated at the end of life/experiment,
-
method to assess the endpoints,
-
data analysis and statistic evaluation,
-
authors, year of publication, title, journal.
We will also extract data on potential conflict of interest in included studies.
4.5. Risk of bias assessment
The internal validity and the quality of eligible studies will be evaluated using the approach recommended by NTP-OHAT (2019) for animal studies. It provides detailed instructions for assessing how potential sources of distortion may have affected the reliability of the results.
The nine included risk of bias criteria are:
-
Randomized exposure level;
-
Allocation concealment of study groups;
-
Evaluation in the study design or analysis of possible important confounding and modifying variables;
-
Blinding of research personnel;
-
Confidence in the exposure conditions;
-
Confidence in the outcomes assessment;
-
All measured outcomes reported;
-
Attrition/exclusion rate;
-
Possible conflicts of interest.
Regarding item 3, possible important confounding factors are, for example, uncontrolled temperature increases in the exposed animals or any other difference in the experimental conditions between exposed and comparison groups. It is worth bearing in mind that the presence of sham group is mandatory for the inclusion in the analysis and, for this reason, it is useless to remark the importance of identical exposure conditions as reported in NTP-OHAT (2019).Two groups of authors (RP, PG and LA, PV) will independently assess these criteria on the individual study level and they will classify the studies according to the following ratings: “++” definitely low risk of bias; “+” probably low risk of bias; “-“probably high risk of bias, or “--“ definitely high risk of bias. Disagreements in the assessment will be discussed between the authors and resolved by consensus. Furthermore, using the OHAT approach, individual studies will be placed into three quality category based on the risk-of-bias ratings. The key criteria for determining the highest weight in the quality of the study are: (1) Confidence in the exposure characterization, (2) Confidence in the outcome assessment and (3) Evaluation in the study design or analysis of possible important confounding and modifying variables. The remaining criteria will be given less weight in determining the quality of the study.
4.6. Data synthesis criteria
4.6.1. Strategy
Structured descriptive summary of eligible studies will include the following items:
-
experimental design (e.g. induction/promotion, co-promotion);
-
outcome(s) reported;
-
the animal model used;
-
age of animals at the start of treatment;
-
the developmental stage of animals at treatment and outcome assessment (e.g. mating and/or pregnancy status);
-
exposure type and level;
-
type of data (e.g. continuous or discrete), statistics presented in the paper, ability to access raw data;
-
variation in the degree of risk of bias at the individual study level.
A meta-analysis, if feasible, will proceed according to the following main sequence (Vesterinen et al. 2013):
-
calculating an effect size for each comparison,
-
weighting the effect sizes according to the effect model adopted,
-
calculating the summary effect size
-
calculating the heterogeneity and the extent to which the predefined study design characteristics explain this heterogeneity,
-
evaluating the subgroups analysis;
-
checking the presence of meta-bias (e.g. attrition bias, publication/duplicate bias, detection bias).
If necessary, measures of absolute effect will be converted into relative effect measures, to ensure comparability of effect estimate and facilitate the meta-analysis that will be performed for all outcome measures: a minimum of 3 studies per group/subgroup will be required(Moya et al. 2014, Hooijmans et al. 2014)
4.6.2. Additional analysis
Heterogeneity in the results of the meta-analysis may be observed. Variations in the effect size across sub-sets of studies are expected, for instance, if there are differences between species or between exposure levels and exposure modalities. Thus, subgroup analysis or meta-regression may give insight into the relation between study characteristics and the effect size, to find an explanation.
If meta-analysis is not possible, data will be reported through the descriptive summary.
4.6.3. Effect models
Because of the exploratory nature of animal studies, random effect and fixed-effect models will be used to calculate the cumulative mean difference of studies with significant heterogeneity (60% < I² < 75%) and without significant heterogeneity (I² < 60%), respectively. Extracted data will be analyzed using appropriate software for meta-analysis. Forest plots will be constructed for each useful parameter and the mean difference value, standard error and p-values will be determined (Borenstein et al. 2009 a, b).
4.6.4. Heterogeneity
Heterogeneity among studies will be assessed with Cochran Q test and further quantified by I² statistics. If studies will be affected by high (I² = 75% or greater) heterogeneity, a narrative synthesis will be done.
4.7. Quality of evidence assessment
We will apply the confidence rating approach based primarily on guidance from the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group (Balshem et al. 2011, Guyatt et al. 2011). The GRADE framework is often applied to evaluate the quality of evidence and the strength of recommendations for outcomes reported in systematic reviews (Higgins and Green 2011)on human and animal studies. We will assess the quality of evidence for the entire body of evidence by each outcome, with any disagreements resolved by a third review author. We will downgrade the quality of evidence for the following five GRADE reasons: (i) risk of bias; (ii) inconsistency; (iii) indirectness; (iv) imprecision; and (v) publication bias.
We will grade the evidence, according to the three Navigation Guide standard quality of evidence ratings: "high", "moderate" and "low". Within each of the relevant domains, we will rate the concern for the quality of evidence, using the ratings "none", "serious" and "very serious". We will start at "high" for randomized studies: quality will be downgraded for no concern by nil grades (0), for a serious concern by one grade (− 1) and a very serious concern by two grades (− 2). We will upgrade the quality of evidence for the following other reasons: large effect, dose-response and plausible residual confounding and bias.
4.8. Strength of evidence assessment
We will apply a modified version of the Navigation Guide methods (Woodruff and Sutton, 2014) for the nonhuman evidence (Koustas et al. 2014). The rating will be based on a combination of four criteria (I) sufficient evidence of toxicity; (II) limited evidence of toxicity; (III) inadequate evidence of toxicity; or (IV) evidence of lack of toxicity, which in turn are based on the criteria used by the International Agency for Research on Cancer (IARC 2013).