This study was designed as an integrative narrative review with a systematized and traceable compilation of evidence on the use of IRT in WMSDs. A narrative approach was selected to enable critical integration and applied interpretation across physiological foundations relevant to interpreting the thermal signal, standardization, and quality-control requirements, and occupational/ergonomic scenarios, while preserving a study identification and selection process that is sufficiently transparent for methodological audit. Manuscript preparation followed SANRA (Scale for the Assessment of Narrative Review Articles) principles [42], emphasizing an explicit rationale, a reproducibly described search, a consistent logical structure, and balanced use of directly relevant evidence. The identification, screening, and eligibility process was documented using Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 [43] with the specific purpose of recording the selection pathway; this flow documentation does not, in itself, imply quantitative synthesis. Search strategies were reported in a PRISMA extension for reporting literature searches (PRISMA-S)-compatible format [44], with full search strings and parameters provided in the Supplementary material.

The review question was structured using the Population-Concept-Context (PCC) framework [45] to explicitly define scope and eligibility within an integrative narrative review with traceable study selection. The guiding question was:

In workers exposed to occupational risk factors for WMSDs, to what extent can IRT—when acquired under standardized protocols—support screening and risk assessment/stratification, as well as monitoring of interventions, either as a stand-alone measure or integrated with functional/structural measures and patient-reported outcomes?

For the P (Population) component, we considered workers and populations in occupational settings (repetitive activities, intensive computer use, industrial sectors, and services), including asymptomatic individuals at risk, symptomatic individuals, and those diagnosed with WMSDs. For the C (Concept) component, the focus was on IRT applied to MSK conditions/WMSDs for screening, risk assessment, and monitoring, including studies employing thermal metrics and/or static/dynamic protocols under controlled conditions. For the C (Context) component, the review was delimited to occupational health and ergonomics applications, including monitoring of interventions (e.g., ergonomic measures, active breaks, and rehabilitation) and, when available, integration with complementary modalities (sEMG, strength/dynamometry, and US with or without Doppler) and PROs.

To maximize coverage and sensitivity, international and regional bibliographic databases with complementary scope were searched: PubMed/MEDLINE (Medical Literature Analysis and Retrieval System Online), Embase (Elsevier), Scopus, Web of Science Core Collection, and SciELO. In addition, technical reference documents and consensus statements focused on standardization and quality assurance in clinical thermography were included because they are directly pertinent to reproducibility and the operational validity of thermal-signal acquisition and interpretation in neuromusculoskeletal (NMSK) applications, including the International Academy of Clinical Thermology (IACT) Quality Assurance Guidelines (2015) and the American Academy of Thermology (AAT) NMSK Thermography Guidelines (2024) [32, 33]. As an additional identification method, snowballing (citation tracking) was performed, including systematic screening of reference lists (backward citation searching) from eligible studies and key reviews/guidelines, as well as forward citation searching in citation-indexed databases (Scopus/Web of Science) [46–48].

Bibliographic searches were conducted on February 12, 2026, with no lower date limit and with pre-specified language eligibility for English, Portuguese, and Spanish. Strategies combined controlled vocabulary (MeSH–Emtree) and free-text terms, organized into four conceptual blocks:

a)

IRT/IRT;

Searches were run in PubMed/MEDLINE, Embase, Scopus, Web of Science Core Collection, and SciELO, using platform-specific syntax and fields compatible with each search engine (ti/ab/kw; TITLE-ABS-KEY; TS). In PubMed/MEDLINE, the strategy was operationalized in three complementary sets (CORE; IRT multimodality with sEMG/strength/US; and standardization/protocols). To reduce document noise without compromising sensitivity, database-specific filters and exclusions were applied: in PubMed/MEDLINE, animal-only records were excluded using the NOT operator (animals[MeSH] NOT humans[MeSH]), while preserving records indexed with humans; in Embase, limits were applied to humans and conference publication types were excluded (conference abstract, conference paper, and conference review); in SciELO, Portuguese and Spanish strategies were used with semantic equivalents of the conceptual blocks.

Identification was supplemented by snowballing, and records retrieved through this method were subjected to the same import, normalization, and deduplication workflow prior to screening, and were counted in the PRISMA 2020 [43] flow diagram as records identified from other sources. Full, verbatim search strings—by database and by set (including fields, operators, filters, and original syntax)—were preserved and made available in Supplementary material. All retrieved records were exported in raw format (plain text or .csv when available) and integrated into a deterministic R-based pipeline [49] prior to screening and eligibility steps, preventing double-counting and ensuring consistency with the PRISMA 2020 flow diagram.

All records retrieved from electronic searches and additional methods (including snowballing) were processed through a deterministic pipeline designed to ensure provenance, traceability, reproducibility, and Quality Control (QC) prior to screening. Outputs from each source were preserved as immutable raw files, maintaining fidelity to the format and content exported by each platform. Input integrity was documented via a cryptographic manifest containing file metadata (name/relative path, size, and timestamp) and a Secure Hash Algorithm 256-bit (SHA-256), enabling audit of changes and a verifiable chain linking inputs to intermediate workflow products.

Import and parsing were performed using source-specific routines to accommodate heterogeneity in structures and fields. Each record was assigned explicit provenance fields (source/database, input file, and import position/line), preserving intra- and inter-source traceability. Next, canonical normalization of identifiers and bibliographic metadata [DOI (Digital Object Identifier), PMID (PubMed identifier), title, year, and journal] was applied, including syntactic standardization (removal of DOI prefixes/URLs when present, normalization of case and whitespace, harmonization of types/formats, and consistent handling of characters), to maximize matching sensitivity without inducing spurious merges. Records with missingness in critical fields were flagged for review (QC file), and, when strictly necessary for identifiability, targeted corrections to critical metadata could be made after verification against the original record, while maintaining traceability and consistency of the analytic dataset.

Deduplication was conducted hierarchically and deterministically using matching keys with priority given to persistent identifiers: DOI > PMID > normalized title + year > normalized title. Records were grouped into duplicate clusters by key type, and each cluster received a stable identifier derived from a hash. To select the retained records within each cluster, a deterministic retention rule based on bibliographic completeness and quality was applied, combining:

a)

a metadata completeness score (prioritizing identifiers, year, journal, authorship, and abstract when available); and,

If ties persisted, a stable ordering based on provenance fields (source/file/import order) was used, ensuring reproducibility of the retained-record selection.

To ensure auditability, the pipeline generated:

a)

a deduplicated master table for screening;

This design ensures that each study contributes only once to subsequent stages (screening and eligibility), preventing count inflation and preserving consistency with the PRISMA flow diagram (including the contribution of snowballing).

Study selection was conducted in two stages [TA (Title/Abstract) and Full Text] using the master dataset generated after cross-source integration and deduplication, while preserving traceability from each record to its unique identifier and retrieval source (bibliographic databases vs. snowballing). All decisions were recorded in canonical spreadsheets and processed through deterministic routines, such that selection-flow counts were programmatically derived from the decision files and reported in the PRISMA 2020 flow diagram; final selection numbers are presented in the Results section (Figure 1).

Data extraction was performed from the included full-text reports using a canonical spreadsheet with pre-specified fields and an associated data dictionary. For each included study, the domains and variables listed in Table 1 were extracted.

Missingness was handled using explicit, non-overlapping codes. Fields not reported in the source article were coded as not reported (NR), with no imputation or inference, thereby preserving the distinction between absence of evidence and evidence of absence. Fields that were not applicable (NA) by study design or context (i.e., the variable could not logically be collected or did not apply) were coded as NA. Information that was sparsely and inconsistently described across the corpus (adverse events/intercurrent events) was captured only when explicitly stated in the full text, but it was not operationalized as a structured domain for cross-study synthesis.

Critical appraisal in this review was conducted at two complementary levels:

a)

reporting quality and editorial consistency of the manuscript; and,

At the manuscript level, SANRA was used as an editorial checklist for narrative reviews. Because this review is an integrative narrative review with a systematized and traceable evidence compilation, without quantitative synthesis, a formal risk-of-bias instrument with high/low judgments by domain was not applied across all studies. Instead, a descriptive appraisal oriented toward reproducibility and plausible sources of systematic variation in the thermal signal was adopted, focusing on the practical applicability of IRT in WMSDs.

At the primary-evidence level, descriptive appraisal of studies prioritized in the synthesis (occupational/ergonomics, standardization, and/or multimodal studies) was guided by pre-specified criteria directly linked to IRT validity and interpretability, including:

a)

reported environmental control and acclimatization;

This appraisal informed the interpretive weight assigned to findings in the compilation, without aggregated scoring and without automatic exclusion based solely on reporting limitations.

QC of the corpus and consistency between decisions and counts were ensured by the deterministic R pipeline, with explicit normalization of identifiers and fields, hierarchical deduplication, canonical decision files (TA and full text), and automated consistency validations (preventing logically invalid combinations between full-text retrieval status and eligibility decision; requiring standardized reasons for exclusions; and recomputing PRISMA counts from the decision spreadsheets). With respect to blinding, reviewers were not blinded to authors, journals, or affiliations; mitigation focused on structured screening, explicit recording of decisions/reasons, and deterministic consistency checks.

Synthesis of the included studies was conducted through thematic integration, structured a priori into axes aligned with the PCC framework and the study objectives, while preserving linkages among IRT acquisition conditions and technical variables, the thermal metrics used, and the occupational/ergonomic context (exposure, task, anatomic region, and screening/monitoring purpose). Findings were organized into four main axes:

Axis A: physiological foundations and interpretation of the thermal signal;

A meta-analysis was not planned because the research question and scope encompass study designs, populations/occupations, anatomic regions, acquisition protocols, and thermal metrics that are often heterogeneous and not directly comparable under a single clinical/methodological estimator. The thematic strategy was adopted as the core approach to enable applied interpretation while maintaining traceability between inferences and source studies.

To ensure audit-ready transparency and end-to-end traceability, all workflow stages (searching, import/parsing, normalization, deduplication, screening, and eligibility) were implemented deterministically, with preservation of canonical artifacts and an audit trail. Complete, verbatim search strategies by database and by set were provided in Supplementary material in a PRISMA-S-compatible format. PRISMA flow counts and summaries of full-text exclusion reasons were automatically recomputed from the canonical decision files, avoiding manual transcription of numbers; final values are presented in the Results. Scripts are available as Supplementary material via a repository.

A total of 1,128 records were identified, including 1,083 from bibliographic databases and 45 from other methods (snowballing). After deterministic deduplication (n = 66), 1,062 records remained for TA screening. At this stage, 761 records were excluded, and 301 reports were sought for retrieval. During full-text eligibility assessment, 27 reports were not retrieved (i.e., unavailable for full-text evaluation), resulting in 274 full-text reports assessed for eligibility. Of these, 27 reports were excluded after full-text review, using standardized exclusion reasons; out_of_scope was the predominant reason (n = 25), followed by duplicate (n = 2). Ultimately, 247 studies were included in the synthesis (Figure 1).

IRT should be interpreted primarily as a functional marker of superficial thermal phenomena related to cutaneous perfusion and autonomic regulation, with greater utility when grounded in ROI-based patterns (side-to-side asymmetry, ΔT, and dynamic responses) rather than in isolated absolute thresholds [31–33, 61]. From an applied standpoint, the included studies suggest that IRT may function as a complementary tool for operational screening and longitudinal monitoring of responses associated with repetitive exposures and sustained postures, with evidence in computer-based work settings (continuous mouse/keyboard use) and in work-organization interventions (active breaks) monitored through segmental thermal maps [9, 13, 38]. Accordingly, the primary practical implication is to reposition IRT as an adjunct, repeatable, protocol-driven objective measure, for which environmental control, participant preparation, and procedural documentation constitute minimum requirements for validity and comparability across assessments [31–33].

IRT is positioned as an indirect functional marker of superficial thermal phenomena, with interpretation dependent on the physiological context (microperfusion/vasomotor regulation), the definition of ROIs, and, most critically, acquisition standardization [19, 31–33, 37, 61]. Across reviews and normative documents included in the corpus, IRT’s primary comparative advantage lies in its ability to capture relatively large-area thermal maps rapidly and without contact, which favors pattern-based interpretation (asymmetry, regional distribution, and changes relative to baseline) in screening and monitoring scenarios [19, 36, 37, 69]. In terms of applicability, included reviews also describe the expansion of portable platforms (including smartphone-based thermography) as a potential logistical enabler for serial monitoring, provided that minimum requirements for environmental control, device configuration, and protocol documentation are met [36, 69].

These operational advantages, however, coexist with structural limitations that clearly delineate what IRT does not do. The corpus converges in recognizing that IRT does not provide deep anatomical information and should not be interpreted as a substitute for structural imaging; moreover, because it is highly sensitive to environmental and technical variables (ambient temperature and stability, humidity, airflow, external IR sources, emissivity, distance/angle, and sensor parameters), IRT can yield false-positive findings when applied without rigorous standardization, with incomplete reporting, or in uncontrolled environments [19, 20, 31–33, 37, 52]. Controlled studies and methodological reviews included in the corpus further emphasize that technical differences and cutaneous characteristics can influence thermal measurements and their comparability, reinforcing the need for complete documentation and caution when extrapolating thresholds across devices and subgroups [31, 52].

In comparison with US, the corpus indicates a functional-structural contrast. US is used for targeted anatomical assessment and, when applicable, for characterizing structural and/or inflammatory/perfusion-related changes within the evaluated segment, whereas IRT adds a superficial functional layer with rapid ROI-based spatial readout [70–72]. Studies combining IRT and US suggest practical complementarity when IRT is used to map segment-level patterns and to guide ROI selection for focused ultrasonographic assessment, rather than to pursue direct diagnostic equivalence [71, 72]. This positioning is consistent with corpus syntheses that treat US and MRI as complementary structural modalities, while reserving for IRT an adjunct role contingent on standardization and on the specific clinical or occupational question [37, 70].

In comparison with sEMG and mechanical measures (dynamometry/strength), the corpus suggests that IRT captures a distinct physiological domain with its own kinetics, which may be useful for contextualizing repetitive demands and regional patterns of superficial response. Included multimodal studies describe protocols in which IRT is analyzed in parallel with EMG (activation/fatigue) and/or strength, indicating that sEMG and dynamometry are more directly aligned with quantifying neuromuscular activity and mechanical performance, whereas IRT provides complementary information when interpreted as ROI-based patterns under standardized tasks and controlled conditions [56, 60, 73]. Accordingly, the corpus-supported critical comparison indicates that integration tends to be more informative than isolated use, particularly when the objective is operational triangulation (superficial functional signal + neuromuscular/mechanical load + self-reported outcomes), rather than structural inference [37, 56, 60, 73].

The contrast with MRI is even more pronounced. The corpus recognizes that MRI provides deep anatomical information and is appropriate when the clinical question requires structural/anatomical characterization, whereas IRT does not access deep tissues and does not provide direct anatomical evidence [37, 70, 74]. In addition, included reviews note that degenerative findings on structural imaging are common in asymptomatic populations and increase with age, requiring clinical correlation and limiting interpretations based exclusively on structural imaging [17]. In the context of low back pain, the corpus includes evidence of limited diagnostic accuracy for thermography relative to clinical reference standards and guidelines that do not recommend its use as a diagnostic test for low back disorders, reinforcing the risk of low specificity when it is applied as a standalone tool [75, 76].

WMSDs represent a multifactorial occupational outcome in which biomechanical determinants (repetition, force, sustained postures, and misalignment), organizational determinants (work pace, breaks, duration of exposure), and psychosocial determinants (occupational stressors) modulate risk by body segment and work context, with substantial economic impact on the workforce [1, 2, 5–7]. Within this framework, IRT’s specific contribution to the theme of “prevalence/risk factors” is not to replace epidemiologic methods or exposure assessment, but to add an objective functional layer that may be useful for characterizing ROI-based patterns in surveillance scenarios and for supporting the prioritization of interventions in tasks or environments with a higher likelihood of superficial physiologic overload [1, 5, 6].

In the domains of surveillance and risk mapping, evidence from the corpus suggests that IRT can be positioned as a complementary component to established approaches for workplace exposure assessment and risk-factor screening tools, provided it is embedded within programs that include explicit acquisition standardization and documentation to reduce variability and misinterpretation [8, 32, 33, 39]. Operationally, such integration is most consistent when:

a)

the unit of analysis is defined by task, department, or job function (rather than diagnosis alone);

Sector- and occupation-specific applicability within the corpus is illustrated by studies in intensive computer users and office work, in which IRT was used to characterize upper-limb thermal patterns under standardized tasks and in association with MSK complaint severity [9, 26, 38]. In scenarios related to postural load and sedentary work, IRT was used to monitor trunk thermal responses during prolonged sitting and to track changes under workstation-level mitigation strategies [13]. In repetitive tasks and production-line analog paradigms, experimental and reliability studies demonstrated the technical feasibility of IRT for monitoring upper-limb ROIs under repetitive demands and overhead tasks, providing a translational basis for task-based risk-mapping programs in industrial sectors [53, 77]. For groups exposed to hand-arm vibration, the corpus included specific evidence for surveillance using cold provocation testing and rewarming assessment, emphasizing protocol heterogeneity and the need for standardized interpretive criteria in occupational health [35, 40, 54, 78]. Finally, the inclusion of studies in high-precision professions with substantial upper-limb demand (e.g., surgeons performing simulated laparoscopic tasks and pianists) broadens the spectrum of applicability and reinforces the potential of IRT as a complementary measure for function- or activity-based functional mapping in specific occupational subgroups [56, 58].

With respect to emerging conditions such as telework and hybrid arrangements, the corpus literature does not coalesce as a dedicated subfield; however, findings from office work, continuous computer use, and prolonged-sedentary protocols provide a direct translational basis for designing surveillance in remote contexts [9, 13, 26, 38]. In this extension, the key implication that the intrinsic environmental variability of home settings is likely to amplify confounding and noise, making adoption of minimum standardization requirements (recording temperature/humidity, acclimation, and pre-examination restrictions) and the use of comparative metrics (asymmetry and relative changes) even more critical for exposure-oriented interpretation [32, 33].

This review was designed as an integrative narrative review with a traceable compilation, with the aim of clinically integrating physiological foundations, standardization/quality-control requirements, and occupational applications of IRT in WMSDs. This deliberately broadened scope entailed inclusion of diverse study designs (experimental, observational, diagnostic-accuracy, intervention studies, reviews, and normative documents), which supports thematic synthesis and critical comparison but limits direct cross-study comparability and precludes quantitative outcome-level pooling across much of the corpus [19, 31, 32, 79, 80]. The secondary literature included in the corpus itself reports substantial heterogeneity in devices, ROIs, protocols, environmental conditions, and thermal metrics, with inconsistencies that prevent meta-analysis and constrain the development of universal parameters [79, 80].

At the operational level, a recurrent methodological limitation in the corpus (and therefore inherited by this synthesis) is incomplete reporting of determinants required for reproducibility and for interpretation of the thermal signal. Included guidelines and reviews emphasize that environmental variables, acclimation, acquisition geometry, equipment configuration, and ROI definition are decisive for validity and comparability [1–3]. In the master data extraction spreadsheet (Supplementary material), IRT device reporting contained at least one NR element in 19/247 studies, and comparator/reference information was absent or NR in 18/247; by contrast, key protocol points were available for all included studies (0/247 coded as NR). These reporting patterns restrict critical appraisal of confounding risk factors and increase the likelihood of systematic between-study variation, particularly for a method that is sensitive to technical and environmental factors [19, 31, 32, 52].

With respect to the exclusion of evidence from animal models, the search and eligibility strategy was designed to prioritize literature with direct clinical and occupational applicability, consistent with the objective of guiding the use and interpretation of IRT in WMSDs within occupational health contexts. For this reason, a dedicated translational synthesis based on animal models was not undertaken, acknowledging that such literature may offer mechanistic insights under highly controlled experimental conditions but falls outside the applied and clinical scope that structured this manuscript. Accordingly, this methodological choice entails a deliberate loss of mechanistic depth and constitutes an accepted limitation to be addressed by future targeted translational reviews.

Finally, the absence of meta-analysis resulted not only from heterogeneity in protocols and outcomes, but also from the multimodal and frequently exploratory nature of part of the corpus, with nonuniform measures and comparators [79, 80]. As a methodological choice aligned with this scenario, appraisal was conducted descriptively and oriented toward reproducibility (emphasizing environmental controls, technical documentation, and comparators), rather than through a single standardized, domain-based risk-of-bias judgment applied uniformly across all included designs, recognizing that this option limits quantitative inference and reinforces the interpretive character of the compilation [19, 31, 32, 79, 80]. In addition, although snowballing broadened coverage (41/247 included studies that had snowballing as a contributing identification method), this procedure may favor citation networks and therefore requires cautious interpretation regarding potential visibility bias; this risk was mitigated by applying the same screening/eligibility and deduplication workflow to records retrieved via additional methods.

Establishing IRT as a clinically useful and operationally reproducible tool for WMSDs requires a research program centered on standardization, protocol traceability, and external validation. Included guidelines and methodological reviews converge in emphasizing that environmental variables, participant preparation, equipment specifications, acquisition geometry, and ROI definition determine interpretive validity and comparability across assessments; therefore, future studies should report these components systematically and in an auditable manner [19, 31–33]. In addition, evidence indicates that technical factors and cutaneous characteristics can influence thermographic measurements even under controlled conditions, reinforcing the need for inter-device harmonization and caution when extrapolating thresholds across populations and contexts [52].