Reporting quality and metrological traceability of 222-nm far-UVC dosimetry: a systematic review and metascientific audit

Plain-language summary

This systematic review examined dosimetric documentation in 84 reports on 222-nm far-UVC. Mean completeness was 45.72%, and calibration and traceability details were least frequent. AI-assisted source checking resolved 223 alerts and corrected 17 scores. Two reviewers checked scores against their sources. The analysis identifies information needed to reconstruct an exposure using an exploratory dosimetric-documentation instrument.

Introduction

Health-related research using 222-nm far-UVC encompasses controlled experiments with microorganisms and tissues, animal and clinical observations, implementation in occupied spaces, environmental measurements, indoor-air chemistry and dosimetry. Across these designs, administered optical exposure is a central input. Efficacy, biological response and comparison with exposure guidelines are interpreted against the amount and spectrum of radiation reaching a target. If the measurement chain or geometry is not described, the same nominal radiant exposure may not represent the same physical exposure across studies.

Ultraviolet radiometry depends on the source and detector. A broadband radiometer combines the source's spectral irradiance with the detector's wavelength-dependent responsivity. Out-of-band response, spectral mismatch between calibration and test sources, nonlinearity, geometry and field nonuniformity can alter the displayed irradiance [3–5]. CIE 220:2016 describes methods for characterising and calibrating UV radiometers [3]. Spectral-mismatch experiments with commercial UV-C radiometers show that source-specific correction can materially change reported readings [5]. For KrCl sources, nominal 222-nm emission, optical filtering, weaker emission at longer wavelengths, reflective boundaries and target-plane location can affect interpretation [6,7].

Arithmetic checking is one component of reproducibility. The identity

H (mJ/cm²) = E (mW/cm²) × t (s)

allows a linked radiant-exposure scheme to be recalculated, but agreement does not establish that irradiance was measured in the correct plane, with a suitable calibrated detector, or with quantified uncertainty. Reproducibility encompasses both arithmetic and the measurement model.

This systematic review assesses dosimetric reporting completeness through a cross-sectional metascientific audit of a curated corpus. The corpus was reconstructed retrospectively and cleaned by families of full-report versions. The design examines information in the selected documents; search coverage and its relationship to the wider literature are described in Methods.

Objectives were to: (1) quantify completeness by report, domain and item using FDRI-28; (2) identify reports meeting retrospectively defined joint reproducibility and traceability criteria; (3) recalculate locally linked irradiance–time–radiant-exposure schemes and record dimensional or unit-labelling problems; and (4) provide a source-linked package enabling independent human verification. The audit does not assess biological internal validity, efficacy, safety or compliance with an exposure limit.

Methods

Design and unit of analysis

A retrospective systematic review and cross-sectional metascientific audit of a curated corpus were conducted. Reporting completeness was assessed by report after its versions were unified. The protocol and statistical plan were documented on 17 August 2026 to establish rules for corpus reconstruction and analysis. Both were retrospective, without registration or specification before the original screening and extraction.

Corpus selection and version control

The final corpus was obtained from 81 reports, four additions and exclusion of HF06-0710 because it lacked an arm measured specifically at 222 nm, giving 84 reports in total.

The initial set contained 83 records. We excluded HF06-1010 as a conceptual technical note without direct empirical measurement and removed HF06-1013 because the bioRxiv record duplicated the final article retained as HF06-1203. We replaced the metadata and source of the obsolete HF06-0673 preprint with the final Scientific Reports article (doi:10.1038/s41598-022-08462-z), and replaced HF06-0931 with the final Building and Environment article's metadata (doi:10.1016/j.buildenv.2025.112734). Extraction for HF06-0931 used the final article's public accepted manuscript under a CC BY licence, deposited in White Rose Research Online. HF06-0993 (NIST IR 8550) met empirical eligibility criteria and was retained as a technical report, not a peer-reviewed article; its overlap with HF06-0849 and HF06-0980 was documented.

For HF06-0729 and HF06-1137, initial extraction used the corresponding public preprints under final-publication metadata. Final versions were located and checked against the extractions; locators were retained in records B029–B031. Initial deduplication produced an 81-report database; the four additions and subsequent exclusion determine the current 84 reports (Figure 1).

Information sources and retrospective search reconstruction

The search was reconstructed through PubMed and OpenAlex API queries on 17 August 2026 because the original dated exports and complete title-and-abstract screening history were unavailable. Search strings and query URLs were retained in the supplement. Reconstruction exported 905 PubMed records and 4,009 unique OpenAlex records (4,914 before cross-database deduplication), with 4,447 records after exact deduplication by DOI, PMID and normalised title. Automated exact-scope matching flagged 1,334 records. Of these, 81 linked to included HF06 reports, 21 to documented exclusions after full-text review, one was a replaced version and 1,233 did not receive independent human screening.

Bibliographic reconstruction was documented separately from original selection because indexing and availability changes can alter query results. The review and audit are restricted to the selected corpus and do not presume an exhaustive search. PRISMA 2020 and PRISMA-S were used as references for describing sources, queries and selection [1,2].

Eligibility

Reports were eligible when they (1) described a directly measured 222-nm/far-UVC exposure; (2) addressed a healthcare, care-delivery, human, animal, tissue, microbial, environmental-implementation, indoor-air-chemistry or dosimetry/metrology question relevant to health interpretation; and (3) had publicly available full text or full text supplied for the audit, sufficient for an item-level audit.

Reviews, editorials, corrections without primary data, protocols, modelling-only work without direct measurement, component-engineering articles without a health-relevant exposure application, reports focused on another wavelength, conceptual device proposals without measured methods and results, and duplicate versions of the same study were excluded. Within a family, the final article was preferred to a preprint. Composite technical reports could remain eligible when they contained relevant empirical work, but their overlap with component articles was recorded.

Full-text retrieval and provenance

New extractions were retrieved for 83 of the 85 historical texts. After excluding HF06-0710, 82 of the 84 included texts were retained locally. Final versions of HF06-0729 and HF06-1137 were consulted through public or publisher sources without redistributing the complete files. The provenance record retains URLs, retrieval attempts and fingerprints, distinguishing historical files from subsequent extractions.

FDRI-28 instrument and scoring

FDRI-28 contains 28 reporting-presence items across five domains:

- Source/spectrum (S1–S6): exact wavelength; source type; brand/model; filter status; measured spectrum; and quantified out-of-band emission.

- Radiometry/calibration (M1–M6): instrument brand/model; calibration statement; calibration date/certificate; metrological traceability; spectral responsivity or mismatch handling; and measurement uncertainty.

- Geometry/field (G1–G6): distance; angle/orientation; beam, field or aperture; uniformity/spatial map; target-plane location; and reflection, shadow, line of sight or boundary conditions.

- Exposure arithmetic (E1–E6): numerical irradiance; exposure time; numerical radiant exposure/fluence; delivery mode/duty cycle; output stability; and a locally linked, evaluable irradiance–time–radiant-exposure scheme.

- Experimental context (B1–B4): target/matrix; relevant conditions; replication/sample size; and distinction between incident, absorbed, effective/weighted or target-plane quantities.

An item received one point only when explicit text in the primary report met its operational definition. Applicable but absent items received zero. B4 was considered not applicable only when the report contained no numerical exposure quantity to which the distinction could apply. All other items were considered applicable.

For report i, primary completeness was 100 × reported applicable items/applicable items. Fixed-denominator sensitivity was 100 × reported items/28, treating NA as zero only for that sensitivity analysis. Domain percentages used the same applicability rule. Scores were categorised descriptively as low (<50%), moderate (50% to <75%) or high (≥75%). These cut-offs are not validated quality thresholds.

Joint reproducibility and traceability criteria

The joint basic-reproducibility criterion required items S1, S2, S4, M1, G1, G5, E1 and E3, plus reported exposure time (E2) or an evaluable linked scheme (E6). The joint traceability criterion required instrument identity (M1), calibration (M2), metrological traceability (M4) and measurement uncertainty (M6). Both describe simultaneous presence of reporting elements and were not used as risk-of-bias instruments.

Arithmetic and dimensional audit

We used H (mJ/cm²) = E (mW/cm²) × t (s). A scheme was evaluable only when irradiance, exposure time and radiant exposure were locally and unambiguously linked, without contextual inference. Relative error was 100 × |Hreported − Hrecalculated|/|Hreported|. Results were classified as consistent (≤5%), minor (>5% to 10%) or major (>10%). Dimensional and unit-labelling problems were recorded separately from multiplication consistency, so a numerically consistent scheme could still be flagged for incompatible terminology or units.

Extraction and source verification

Extraction was based on identifiable passages from each source and the operational FDRI-28 definitions. Quotations and locators were retained, together with the rationale and search terms used when an item was not identified. Document control detected positive scores without a quotation or verifiable locator for additional checking. Automated checks assessed unique identifiers, permitted values, denominators, score reproduction, joint criteria and agreement between files.

Extraction used AI assistance and explicit rules. Two reviewers checked scores against sources and ratified eligibility and corpus versions, including HF06-0729 and HF06-1137. Both reviewed the definitions and application of FDRI-28 and agreed on final decisions. Agreement was described qualitatively without calculating an inter-rater reliability coefficient.

Statistical analysis

We summarised total and domain percentages using the mean, sample standard deviation, median, first quartile and third quartile. Item proportions use two-sided 95% Wilson intervals and show the applicable denominator. Absent items were not imputed. The applicable-item score is reported as primary and the fixed score out of 28 as sensitivity.

The relationship between completeness and publication year was explored using Spearman correlation and an unadjusted ordinary least-squares slope, expressed in percentage points per year. Analysis was retrospective and descriptive, with incomplete 2026 coverage. No confirmatory comparisons by application or peer-review status were performed because strata were defined retrospectively and some categories were small. Results and figures were generated using reproducible code from the retained analysis files.

Bibliographic update and corpus closure

The evidence cut-off was 2 October 2026; queries were executed on 3 October UTC using publication filters on or before the cut-off. Four reports identified in the documentary update were incorporated: Hilge, Su, Chen and Sheng [8–11]. Definition code v2.0 was retained. Hilge studies artificial human skin in vitro; Su combines laboratory and field textile measurements; Chen studies algae-containing water and Fe(VI)/UF; Sheng examines food microbiology using plasma and UV. Su and Sheng were extracted from publisher versions preceding the version of record. Chen includes the optical supplement and is coded to 2026 based on online publication, although its volume is 2027. No clinical inferences were drawn from these models.

The update retrieved 1,001 Europe PMC records, 21 OpenAlex records and 100 of the 672 results from a partial Crossref query. Deduplication produced 1,118 unique records: 960 already linked to the study record and 158 new candidates. The Crossref query was constrained by the permitted number of requests. New candidates were retained in the discovery archive, separate from the 84-report corpus ratified by the reviewers.

The supplementary archive retains the initial screening of 1,233 records, 35 full-text checks and assisted decisions on the 158 new candidates. These stages are identified separately to avoid adding them as unique studies. Historical spreadsheets retain their original R1 and R2 fields; subsequent human review concerns the final corpus's scores and decisions.

Retrieval of Chitraningrum's text (HF06-0773) identified a measured peak of 219 nm, insufficient to confirm wavelength eligibility. Ran's text remained inaccessible and was retained in the retrieval record. Both cases remained outside the scored corpus. Final versions of HF06-0729 and HF06-1137 were checked against their extractions and retained in the corpus.

Corpus size and analytical scope

Corpus size corresponded to reports selected and available for the documentary audit. No clinical power calculation was performed because the study assesses exposure reporting and includes neither new interventions nor adverse-event follow-up.

Results

Source verification and eligibility

AI-assisted source checking reviewed all 223 report–item pairs in the cumulative record: 206 positive scores were confirmed and 17 corrected; checking of all flagged pairs was completed. The 223 alerts concerned positive scores selected for checking their documentary support. HF06-0710 measures nine wavelengths in five-nm increments between 215 and 255 nm, without a separate arm measured at 222 nm. It was excluded under the corpus eligibility criteria; its record and 28 scores are retained in the history. Decisions, sources and locators are preserved in blocks B001–B031. Two reviewers ratified scores and selection decisions.

Corpus composition and reporting completeness

The corpus included 84 reports: 81 peer-reviewed articles, two preprints and one technical report. Mean completeness was 45.7168% (SD 12.2571) and the median was 46.4286% (Q1–Q3 39.2857–53.5714%). Descriptive categories comprised 52 reports with low completeness, 31 moderate and one high. Corpus composition and score distribution are shown in Figures 1 and 2; item-level results are shown in Figure 3.

B4 applies to 70 reports (14 NA). The mean using the fixed denominator of 28 was 45.4932%. HF06-0559 and HF06-0626 met the joint basic-reporting criterion; no report met the traceability criterion. Zeros for S6 and M5 indicate absence of explicit evidence under the operational definition, not universal absence of the practice.

Arithmetic and temporal exploration

Of the 23 evaluable exposure schemes, 21 agreed arithmetically within 5%; HF06-0858 and HF06-0927 had differences exceeding 10%. Four additional unit or dimensional discrepancies were identified and recorded separately. Temporal exploration showed no clear monotonic association between completeness and publication year (Spearman ρ = −0.138; p = 0.210).

Assessment of the supplementary bibliographic archive

Assisted assessment of 35 full texts identified 26 candidates without an FDRI-28 score, five proposed exclusions, one linked correction and three scope queries. Among the 158 new candidates, there were 137 proposed exclusions, 15 full-text requests, four scope queries, one related version and one link to an earlier candidate. These records remain in the discovery archive and do not contribute to final-corpus scores. The supplementary set of 1,233 records did not receive independent human screening.

Lexical checking of the 2,352 decisions identified 223 positives for additional documentary verification. Checking documented in records B001–B031 confirmed 206 scores and corrected 17, completing review of the alerts. Ratified scores describe reporting completeness.

Original figure; its source labels are preserved. The caption is translated into English.

Original figure; its source labels are preserved. The caption is translated into English.

Original figure; its source labels are preserved. The caption is translated into English.

Discussion

Main findings

Across the 84 reports, descriptions of the source and experiment were more frequent than information needed to reconstruct a metrologically traceable exposure. The median report met 46.4% of applicable FDRI-28 items, and radiometry/calibration was the least complete domain. No report combined instrument identity, calibration, traceability and uncertainty in the corrected matrix. The result concerns explicit presence of information under the documentary rules applied.

FDRI-28 records presence of reporting elements: a zero indicates that no explicit evidence was located, and a one indicates that the element was documented. This definition keeps documentary completeness separate from experimental implementation. The instrument is exploratory and does not measure risk of bias, efficacy, safety or compliance with exposure limits.

Arithmetic consistency and measurement traceability

Among the 23 evaluable schemes, 21 were arithmetically consistent within 5%. However, approximately 73% of the corpus did not supply a linked exposure scheme permitting this check. Agreement of H = E × t describes calculation consistency; its interpretation also depends on spectral mismatch, calibration, measurement-plane location, field uniformity, reflections, shadows and source stability. The low frequency of items M3–M6 and E5 identifies the least documented metrological components.

The language of units and dimensions also matters. Calling power per area 'fluence', using energy per area when irradiance is intended, or printing W/cm² when a table implies mW/cm² can cause errors in replication, evidence synthesis and implementation. Recording these problems separately from arithmetic agreement preserves the distinction between a computational discrepancy and a measurement-definition problem.

Minimum reporting set for 222-nm studies

Observed gaps support a concise minimum set that can be reported in Methods, a structured exposure table or supplementary material:

1. Source and spectrum: exact wavelength, physical source type, manufacturer/model, filter status, measured spectrum and numerical out-of-band emission.

2. Measurement chain: radiometer or detector brand/model, relevant spectral responsivity, calibration method and date/certificate, traceability chain, source-specific mismatch correction where appropriate, and uncertainty.

3. Geometry and field: source–target distance, angle/orientation, field or aperture dimensions, uniformity or spatial map, measurement position relative to the target plane, and reflection/shadow/boundary conditions.

4. Exposure scheme: numerical irradiance, duration, radiant exposure, continuous/intermittent/pulsed operation, warm-up or stability checks, and at least one locally linked E–t–H example.

5. Context: target or matrix, relevant environmental and biological conditions, replication or sample size, and whether the reported quantity is incident, absorbed, effective/weighted or target-plane.

This set does not prescribe a particular instrument or exposure. It enables examination of the measurement model and its uncertainty.

Implications for research and editorial practice

Dose–response comparison assumes that exposure quantities are physically comparable. Incomplete reporting of spectrum, geometry and metrology can create hidden heterogeneity that conventional evidence synthesis might misclassify as biological variation. The problem is particularly relevant when transferring laboratory exposures to occupied spaces, where reflection, shadow, source distribution, time-varying output and target location can produce very different human, microbial and environmental exposures [6,7].

The 28 items organise source, measurement, geometry, exposure and context components. Exposure reconstruction, calibration and geometric description enable dosimetric reporting to be examined. Exploratory use of FDRI-28 and its validation status are considered under Scope and limitations.

Strengths

The audit applied explicit rules to identify versions and classify reports, and completed documentary checking of all 223 alerts. Sources, locators and decisions allow each score to be reconstructed. Applicable-item analysis was compared with a fixed denominator of 28, while arithmetic and dimensional discrepancies were recorded separately. Review by two assessors and reproducible code support traceability of the corpus results.

Scope and limitations

Frequencies describe the 84 reports in the selected corpus. Retrospective reconstruction, absence of original exports and the 1,233 supplementary records without complete human screening limit representativeness of the wider literature. HF06-0993 shares content with two included articles, so report-level deduplication does not remove all documentary dependence.

FDRI-28 was applied as an exploratory documentary-completeness instrument, with definitions and scores reviewed by two assessors. The analysis did not estimate inter-rater reliability, construct validity or association with external outcomes. Binary scoring summarises the presence of information and may omit differences in detail between experimental arms. Fixed-denominator sensitivity using 28 yielded almost identical results. Arithmetic consistency concerns published values and does not constitute calibration of the original measurements.

Year-based comparisons were retrospective and unadjusted, with changes in corpus composition and incomplete 2026 coverage. They are interpreted as temporal descriptions of the available reports.

Conclusions

Across the 84 reports, mean dosimetric completeness was 45.72% and the median was 46.43%. The distribution—52 reports below 50%, 31 from 50% to less than 75%, and one at least 75%—identifies reporting gaps using the exploratory FDRI-28 instrument.

AI-assisted source checking resolved 223 alerts: it confirmed 206 scores and corrected 17. All 2,352 decisions retain sources and locators for reproducing the aggregates. The audit provides a verifiable route for examining exposure documentation and metrological traceability.

Disclosures

Ethics approval and consent

Not applicable. This systematic review and metascientific audit used published reports, public or supplied for analysis. It collected no individual data, contacted no participants and accessed no private identifiable information. No new human or animal intervention was performed.

Patient and public involvement

Patients and members of the public were not involved in the design, conduct, interpretation or reporting of this audit. The work evaluates reporting in published documents.

Protocol, registration and analysis plan

The protocol and statistical plan dated 17 August 2026 retrospectively document reconstruction of the v1.2 package. They were neither prespecified nor prospectively registered. Live API queries are retained as search reconstruction and distinguished from the original historical exports.

Data, code and material availability

The reproducible HF06 v2.9 material contains the final corpus, FDRI-28 v2.5 matrix, evidence record, definitions, arithmetic audit, decisions B001–B031, provenance records and code. The programs code/run_reproduce.py and code/validate_outputs.py reproduce all 17 corrections, scores, summaries and arithmetic checks. Historical spreadsheets retain their original state; subsequent review by the two assessors is described in Methods. Full-text distribution respects each source's permissions.

Funding and sponsor role

The work received internal research funding from MedETechni SpA, with which the authors are affiliated. There was no external funding. The institutional and financial relationship is disclosed under Conflicts of interest and editorial independence.

Conflicts of interest and editorial independence

Kevin Virgilio Morales Chamorro is founder and Chief Executive Officer of MedETechni SpA, developer and marketer of 222-nm far-UVC technologies including Virgilio Rays. Gricell Jeraldina Báez Arévalo is Chief Medical Officer of MedETechni SpA. MedETechni SpA funded the work and publishes MedETechni Journal of Translational Research. These institutional, financial, commercial and editorial interests are disclosed.

Author contributions

CRediT contribution distribution. Kevin Virgilio Morales Chamorro: conceptualisation, methodology, software, formal analysis, investigation, data curation, resources, visualisation, original draft, supervision, project administration and funding acquisition. Gricell Jeraldina Báez Arévalo: methodology, investigation, and review and editing of the text.

AI-assisted work statement

Generative artificial intelligence based on a language model, through Codex (OpenAI; GPT-6.1 Sol model), was used for text retrieval and normalisation, evidence location, application of extraction rules, calculations and writing. Decisions retain their sources and locators. Two reviewers checked scores and corpus selection. The authors retain scientific responsibility for interpretation and final content.