"Inconsistent application of GRADE certainty-of-evidence domains across the WHO-commissioned systematic reviews of radiofrequency electromagnetic field exposure: a cross-review audit"
By Academia Co-scientist assisted by Claude Opus 5, September 27, 2026
A cross-review audit was conducted. Every outcome-level rating and domain decision was extracted from 11 reviews reported in 12 publications; the statistical power of every meta-analysis was quantified; risk-of-bias assignments were cross-checked against the primary papers; and all four published corrigenda were reconciled into the corpus.
The audit yielded 150 outcome-level certainty ratings and found that certainty ratings are not commensurable across the series.
Inconsistent "Imprecision": The imprecision domain carries at least eight operational definitions across the series. Each review, taken by itself, is broadly coherent — imprecision decisions track achieved power in the expected direction, pooled and within reviews (Spearman ρ = −0.46 across 122 estimates). What the reviews do not share is a threshold. Reviews downgraded for imprecision at achieved powers as high as 100% and declined to downgrade at powers as low as 10.4%, so there is no level of precision at which the series agrees on what the domain requires. This is a harder problem than carelessness: it cannot be fixed by asking any individual review to be more careful.
Contradictory Indirectness Rules: Two reviews hold that epidemiological evidence is never downgraded for indirectness. Four others downgrade it, one by two levels per outcome. The rules are mutually exclusive and are applied within a single evidence stream.
Structural Bias (Missing Upgrades): The upgrade half of GRADE is absent from three instruments and unused in most of the rest. Of 150 outcomes, eight operative upgrades use a bespoke cross-species domain; where the standard columns are present with prespecified numeric thresholds, they were never met. An instrument that can only subtract will inevitably converge on low certainty regardless of what the studies actually show.
Internal Contradictions: Identical inputs sometimes yielded opposite verdicts. Two halves of a single review implemented the same protocol clause in opposite directions. More sharply, within one evidence profile assessed by one team, a confidence interval 0.293 SD wide was downgraded for imprecision while one 0.351 SD wide, on the same cognitive test and at the same power, was not — the domain is not monotone in interval width even within a single table.
Certainty Language Outruns Precision: Of 122 meta-analyses with interpretable targets, 63 had less than 80% power against their moderate target. The principal cancer review has a median power of 0.28 to detect a 20% increase in risk, downgraded no outcome for imprecision, and rated four outcomes Moderate — the wording being that exposure "likely does not increase" risk — at powers of 0.17 to 0.64. A rating of that kind documents an absence of precision rather than an absence of effect.
Non-Comparable Scales: The animal and human cancer reviews were rated on entirely different scales. One expresses strength of positive support for a hazard, the other confidence in an absence of association, and the same word — "Moderate" — denotes different quantities in the two.
The audit takes no position on whether RF-EMF exposure causes any of these outcomes. Every finding above would stand unchanged if the reviews' substantive conclusions were entirely correct, because each concerns the internal consistency of the instrument rather than the adequacy of the inference.
Recommendations Prior to WHO Publication of RF-EMF Monograph
These differences originate in the instruments, not the evidence. The series should take the following steps before the WHO Monograph is published:
Publish a common operational specification for the imprecision and indirectness domains. One review in the series already prespecifies a quantitative imprecision criterion and obeys it in every row of its evidence profile, which establishes both that the domain can be specified and that no one else adopted the specification.
Explain the omission of upgrade columns (dose–response and large magnitude), as a downgrade-only instrument cannot return a high rating for any body of evidence.
Issue errata for internal discordances and clearly distinguish domains assessed "without concern" from domains that simply "could not be assessed." An evidence profile that prints a single character in a cell cannot express the difference, and readers aggregating across reviews cannot recover it.
Prior to publication of the WHO Environmental Health Criteria Monograph, the series should establish a common operational specification for the imprecision and indirectness domains and apply those criteria across all of its constituent reviews, to ensure consistency in their certainty-of-evidence ratings. Until then, ratings from different reviews in this series should not be treated as equivalent, and an overall characterization built by aggregating them across reviews will inherit differences that originate in the instruments rather than in the evidence.
https://drive.google.com/file/d/1qrLFvfxK2ATg9JYM_ncJy65uS68J9xsp/view?usp=drive_link
* References
Frank,
Melnick, Moskowitz on behalf of the International Commission on the Biological
Effects of Electromagnetic Fields (ICBE-EMF). A critical appraisal of the WHO
2024 systematic review of the effects of RF-EMF exposure on tinnitus,
migraine/headache, and non-specific symptoms. Reviews on Environmental Health, 2024. DOI:
10.1515/reveh-2024-0069. https://www.degruyter.com/document/doi/10.1515/reveh-2024-0069/html
Frank,
Moskowitz, Melnick, Hardell, Philips, Héroux, Kelley, E. The systematic review
on RF-EMF exposure and cancer by Karipidis et al. (2024) has serious flaws that undermine the validity of the study’s conclusions. Environment International,
2024. DOI: 10.1016/j.envint.2024.109200. https://www.sciencedirect.com/science/article/pii/S0160412024007876
Melnick,
Moskowitz, Héroux, Mallery-Blythe, McCredden, Herbert, Hardell, Philips,
Belpoggi, Frank, Scarato, Kelley; on behalf of the International Commission on
the Biological Effects of Electromagnetic Fields (ICBE-EMF). The
WHO-commissioned systematic reviews on health effects of radiofrequency
radiation provide no assurance of safety. Environmental Health. 2025 Oct
2;24(1):70. DOI: 10.1186/s12940-025-01220-4. https://ehjournal.biomedcentral.com/articles/10.1186/s12940-025-01220-4 Supplementary
Material 1: "Examples
of Working Group Authors’ Ties to Industry" Supplementary
Material 2:
Moskowitz: "The Meta-Analyses in the WHO RF-EMF Systematic
Reviews Yielded Unreliable Results"
Related posts
Biased WHO-commissioned review claims no cancer link to cellphone use
International Commission on the Biological Effects of Electromagnetic Fields (ICBE-EMF)
