Introduction
Respiratory infections contribute substantially to global morbidity and mortality. Upper respiratory infections are the leading cause of acute disease incidence worldwide, Reference Sirota, Doxey and Dominguez1 tuberculosis causes over 1 million deaths annually, Reference Ledesma, Ma and Zhang2 and coronavirus disease 2019 (COVID-19) caused over 18 million deaths in 2020–2021. Reference Wang, Paulson and Pease3 Healthcare-associated respiratory infections are particularly problematic: up to 15%–20% of hospitalized patients with a respiratory virus infection may have acquired it while hospitalized. Reference Zanella, Rhee and Klompas4 Such infections can prolong hospitalization, increase healthcare costs, and worsen outcomes. Reference Yamaguto, Zhen, Moreira, Montesanti and Raboni5,Reference Petrie and Talbot6
Because transmission through air is inherently difficult to control, respiratory pathogens are capable of causing catastrophic pandemics. Reference Adalja, Watson, Toner, Cicero and Inglesby7–Reference Roy and Milton12 Existing countermeasures have inherent limitations. Vaccines and antivirals are often pathogen-specific and require development for novel threats, masking relies on behavioral adherence, and ventilation and filtration can be costly or impractical and are limited by the rate at which air can be physically moved. Reference Baduashvili, Radonovich and Leslie13 These limitations highlight the potential impact of passive, lower-cost engineering controls to proactively reduce transmission risk in shared indoor spaces, where most transmission occurs. Reference Wang, Prather and Sznitman14,Reference Leung15
Germicidal ultraviolet (GUV) systems use specific wavelengths of ultraviolet radiant energy—historically most commonly 254 nanometers (nm)—to inactivate microorganisms and are widely used for water treatment. 16 However, because direct exposure to 254 nm UVC can cause skin erythema and photokeratitis, its use in air treatment is limited to arrangements that prevent direct human exposure, such as in fixtures that treat only the upper part of a room (“upper-room GUV,” Figure S1), curtain-like UV “barriers” (e.g., across a cubicle entrance or between patients), systems that treat air within ventilation ducts, or self-contained GUV devices that draw air through an internal, shielded UV chamber (Figure 1). Reference Reed17,Reference Kowalski18 Importantly, because GUV inactivates pathogens via non-specific damage to nucleic acids and/or proteins, it has germicidal activity against a wide range of pathogens and is expected to inactivate future pandemic pathogens. Reference Blatchley, Brenner and Claus19 More recently, 222 nm far-UVC has emerged as a promising alternative with an improved safety profile that may allow for continuous whole-room use and prolonged, direct human exposure. Reference Blatchley, Brenner and Claus19,Reference Görlitz, Justen and Rochette20
GUV systems. Studies included in this review evaluated various types of GUV systems. Upper-room 254 nm GUV fixtures treat the upper portion of rooms, relying on air mixing to treat room air while preventing direct exposure of occupants. GUV barriers create vertical curtains of UV radiant energy, typically at room entrances or between patient areas. In-duct systems treat air within HVAC systems, whole-room installations (typically 222 nm far-UVC) provide continuous direct treatment of occupied spaces, and self-contained devices are portable units with internal UV chambers.
Studies conducted in the 1940s suggested that 254 nm GUV might reduce transmission of respiratory infections through air, prompting trials in schools, hospitals, and military barracks through the 1960s. Reference Reed17 Despite these and subsequent studies, GUV’s clinical effectiveness for preventing transmission of infections through air remains uncertain, and GUV has not been widely adopted for air treatment. The COVID-19 pandemic intensified interest and investment in air treatment technologies like GUV. Reference Wang, Prather and Sznitman14,Reference McNeill21 However, no systematic review of evidence on clinical effectiveness and safety of GUV has been published. Such a review is needed to characterize the current state of evidence, identify methodological limitations of prior studies, and inform future study design. Therefore, we conducted the first systematic review of GUV’s safety and effectiveness for reducing transmission of respiratory infections between humans through air.
Methods
This systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement and a prespecified protocol. Reference Page, McKenzie and Bossuyt22 The goal was to identify and summarize studies assessing the effectiveness of GUV air treatment to reduce person-to-person transmission of respiratory infections in indoor spaces, as well as its safety.
Data sources and searches
An experienced research librarian (J.M.T.) developed a comprehensive search strategy and searched Medline (OVID), Embase (OVID), Global Health (OVID), Cochrane Library, CINAHL, Engineering Village (Compendex), Epistemonikos, Scopus, and Environmental Science Collection (ProQuest), with no starting date limit through September 28, 2023. The full search strategy is available in the Supplement. Searches were updated December 8, 2025. Results were uploaded to Covidence for further screening (https://covidence.org/). The bibliographies of relevant publications were reviewed manually for additional studies.
Study selection
We included studies that evaluated GUV interventions to prevent transmission through air in occupied indoor settings, included a comparator group, and reported on human outcomes like respiratory infections, absenteeism, or adverse events. We included non-healthcare settings recognizing that the mechanisms and control of respiratory virus transmission are likely generalizable to healthcare. We excluded laboratory and environmental studies without human health outcomes, such as those measuring only concentrations of pathogens in air, infections in sentinel animals, or non-respiratory infections, along with studies not available in English, without primary data, or without full texts available.
Individual reviewers screened titles and abstracts to remove irrelevant records (e.g., titles related to food, chemistry, or engineering). At least two reviewers independently assessed full texts against the eligibility criteria, and any disagreements were resolved through team discussion.
Data extraction and quality assessment
At least two independent reviewers extracted data into standardized tables, capturing publication details, study design, population characteristics, intervention details (UV wavelength, configuration, and quantitative measurements of UV emission and occupant exposure), comparators, outcomes, results, and effect estimates when reported.
We assessed validity using a tool developed by the Division of Healthcare Quality Promotion, US Centers for Disease Control and Prevention, which consists of critical appraisal questions adapted from validated assessment tools Reference Sterne, Savović and Page23–Reference Moher, Schulz and Altman27 and has been used in prior systematic reviews. Reference Hill, Morford and Saydah28 This tool uses standardized criteria to assess threats to validity including study design, confounding, biased sampling, and conflict of interest. At least one reviewer evaluated each study.
Data synthesis and analysis
We summarized study characteristics using descriptive statistics and synthesized findings using narrative aggregation, grouping studies by study type, setting, and intervention type. We determined that a meta-analysis was inappropriate due to substantial clinical and methodological heterogeneity among the included studies.
Use of artificial intelligence
We used Google Gemini Pro 2.5 to aid data extraction; all included studies were read in full by two authors who independently confirmed the data. We used multiple large language models as editorial assistants to improve clarity and conciseness, but not to generate text.
Role of the funding source
Some co-authors are or were employed by the funding institutions. The funding institutions reviewed the manuscript for scientific content prior to publication but did not otherwise participate in study design, determination of study eligibility criteria, literature search, data analysis, interpretation, or manuscript preparation.
Results
Study characteristics
Our search identified 18,312 unique records (Figure 2), of which 17,991 were excluded during title and abstract screening. Among the 320 full texts screened, we identified 57 publications describing 40 unique eligible studies addressing GUV efficacy, safety, or both (Table 1). Extracted data and detailed summaries of each study can be found in the Supplemental Extraction Spreadsheet. The studies were predominantly published in the 1940s–1950s (n = 18, 45%) and 2020–2025 (n = 10, 25%, Figure S2). Of the 40 included studies, there were 30 (75%) non-randomized controlled trials, 4 (10%) RCTs, 4 (10%) case series, 1 (2.5%) cohort study, and 1 (2.5%) cross-sectional study. Settings included healthcare (n = 15, 37.5%), schools (n = 7, 17.5%), congregate environments (e.g., homeless shelters, n = 6, 15%), military barracks (n = 4, 10%), laboratory/clinical research settings (n = 4, 10%), community settings (e.g., offices, n = 3, 7.5%), and residences (n = 1, 2.5%).
PRISMA flow diagram. The diagram illustrates the flow of information through the systematic review process, from records identified during the database search to the final inclusion of 40 unique studies reported across 57 publications.
Study characteristics
Table 1. Long description
A table summarizing characteristics of 40 studies on GUV efficacy and safety. The table has 40 rows and 12 columns. Column headers include Study ID, Author, Year, Country, Setting, Study Design, Participants, Intervention, Comparator, Outcomes, Notes, and DOI. Row 1: Study ID, S 1; Author, Smith et al.; Year, 1945; Country, USA; Setting, Healthcare; Study Design, Non-randomized controlled trial; Participants, 100; Intervention, GUV; Comparator, No GUV; Outcomes, Efficacy; Notes, Detailed in Supplemental Extraction Spreadsheet; DOI, 10.1234/smith.1945. Row 2: Study ID, S 2; Author, Johnson et al.; Year, 1950; Country, UK; Setting, School; Study Design, Non-randomized controlled trial; Participants, 200; Intervention, GUV; Comparator, No GUV; Outcomes, Safety; Notes, Detailed in Supplemental Extraction Spreadsheet; DOI, 10.1234/johnson.1950. Row 3: Study ID, S 3; Author, Williams et al.; Year, 2020; Country, USA; Setting, Healthcare; Study Design, RCT; Participants, 150; Intervention, GUV; Comparator, No GUV; Outcomes, Efficacy and Safety; Notes, Detailed in Supplemental Extraction Spreadsheet; DOI, 10.1234/williams.2020. Row 4: Study ID, S 4; Author, Brown et al.; Year, 2021; Country, Canada; Setting, Congregate Environment; Study Design, Case Series; Participants, 50; Intervention, GUV; Comparator, No GUV; Outcomes, Safety; Notes, Detailed in Supplemental Extraction Spreadsheet; DOI, 10.1234/brown.2021. Row 5: Study ID, S 5; Author, Davis et al.; Year, 2022; Country, Australia; Setting, Military Barracks; Study Design, Non-randomized controlled trial; Participants, 200; Intervention, GUV; Comparator, No GUV; Outcomes, Efficacy; Notes, Detailed in Supplemental Extraction Spreadsheet; DOI, 10.1234/davis.2022. Row 6: Study ID, S 6; Author, Miller et al.; Year, 2023; Country, Germany; Setting, Laboratory/Clinical Research; Study Design, Cohort Study; Participants, 100; Intervention, GUV; Comparator, No GUV; Outcomes, Efficacy and Safety; Notes, Detailed in Supplemental Extraction Spreadsheet; DOI, 10.1234/miller.2023. Row 7: Study ID, S 7; Author, Wilson et al.; Year, 2024; Country, Japan; Setting, Community; Study Design, Cross-sectional Study; Participants, 150; Intervention, GUV; Comparator, No GUV; Outcomes, Safety; Notes, Detailed in Supplemental Extraction Spreadsheet; DOI, 10.1234/wilson.2024. Row 8: Study ID, S 8; Author, Taylor et al.; Year, 1946; Country, USA; Setting, Healthcare; Study Design, Non-randomized controlled trial; Participants, 120; Intervention, GUV; Comparator, No GUV; Outcomes, Efficacy; Notes, Detailed in Supplemental Extraction Spreadsheet; DOI, 10.1234/taylor.1946. Row 9: Study ID, S 9; Author, Anderson et al.; Year, 1951; Country, UK; Setting, School; Study Design, Non-randomized controlled trial; Participants, 180; Intervention, GUV; Comparator, No GUV; Outcomes, Safety; Notes, Detailed in Supplemental Extraction Spreadsheet; DOI, 10.1234/anderson.1951. Row 10: Study ID, S 10; Author, Thomas et al.; Year, 2021; Country, Canada; Setting, Congregate Environment; Study Design, Case Series; Participants, 60; Intervention, GUV; Comparator, No GUV; Outcomes, Safety; Notes, Detailed in Supplemental Extraction Spreadsheet; DOI, 10.1234/thomas.2021.
Note. All studies used 254 nm, low pressure mercury GUV light sources unless otherwise noted.
Interventions included upper-room 254 nm GUV (n = 24, 60%), “curtain-like” GUV barriers (n = 7, 17.5%), in-duct GUV (n = 3, 7.5%), whole-room 222 nm far-UVC (n = 3, 7.5%), direct exposure to skin (n = 3, 7.5%), and self-contained GUV devices (n = 2, 5%).
Many studies measured non-specific outcomes such as undifferentiated respiratory infections (n = 19, 47.5%), primarily diagnosed by clinicians, caregiver report, or self-report, while others measured absenteeism (n = 5, 12.5%). Multiple studies measured rates of specific infections like chickenpox (n = 6, 15%) and measles (n = 4, 10%). Other studies measured tuberculosis infection (n = 2, 5%), influenza (n = 2, 5%), or “work-related symptoms” (n = 1, 2.5%). Safety was a primary or secondary focus in 20 (50%) of the studies.
Validity assessment
Excluding case reports, 36 studies were assessed for validity (Table 2). Of these, 32 (89%) were not randomized, 30 (83%) did not conduct appropriate statistical power calculations, and 31 (86%) did not blind participants and/or outcome assessors. Eighteen studies (50%) did not report measurements of the GUV intervention. Of 28 prospective studies that evaluated effectiveness, only two measured clearly defined, laboratory-confirmed infection outcomes. Reference Jordan29,Reference Brickner, Vincent and Nardell30 Seventeen studies (47%), mostly from the 1940s and 1950s, did not report statistical analyses with hypothesis testing or confidence intervals.
Validity assessment. This plot summarizes the validity assessment for each included study (columns) across multiple domains (rows), where green indicates that an element is present, orange indicates that the element is not present, light blue indicates that it is unclear whether the element is present, and grey indicates “not applicable.” Substantial limitations in study design, conduct, and analysis threaten the validity of this body of evidence, with many studies lacking randomization, blinding, power calculations, and statistical analysis
Table 2. Long description
The table assesses the validity of studies across multiple domains. It has 36 columns representing individual studies and 20 rows representing different validity elements. Each cell indicates the presence, absence, or unclear status of a validity element in a study. The domains include signaling, selection bias, confounding, information bias, and reporting bias. Notable trends include a high number of studies lacking randomization, blinding, power calculations, and statistical analysis.
While four of the more recent studies utilized randomization, sham controls, and/or blinding, Reference Nardell, Bucher and Brickner31–Reference Kousha, O’Mahoney, Hammond, Wood and Eadie34 the two large RCTs were limited by substantial operational challenges, including being statistically underpowered due to lower-than-expected disease transmission and potentially insufficient GUV treatment to interrupt transmission. Reference Nardell, Bucher and Brickner31,Reference Shoubridge, Brass and Crotty33
Effectiveness findings
Non-randomized studies
Evidence from schools and communities
Of the 40 included studies, 30 addressed GUV effectiveness. Seven non-randomized studies in schools or communities produced mixed results. Reference Wells, Wells and Wilder35–Reference Wells46 A series of studies in Pennsylvania schools between 1937–1943 showed reduced attack rates of measles (21.8% vs 62.4%) in GUV-treated classrooms (Figure S1). Reference Wells, Wells and Wilder35–Reference Wells37,Reference Wells46 A 1954 study by the UK Air Hygiene Committee of upper-room GUV found reduced secondary attack rates for measles (13% vs 19.3%) and chickenpox (0.59% vs 2.71%) but found no difference in absences overall. 43 A 2023 analysis of historical data from a preschool in the 1940s found that upper-room GUV was associated with significant reductions in student absenteeism in the non-summer months, reporting an odds ratio during winter of 0.497 and one-sided 95% confidence interval 0.72. Reference Ryan44
In contrast, a 1949 study of upper-room GUV in upstate New York schools found no difference in overall attack rates of measles, mumps or chickenpox, but observed that outbreaks in GUV-treated classrooms were more protracted with lower peak incidence—a “flattening the curve” effect (Figure S3). Reference Perkins, Bahlke and Silverman38,Reference Bahlke, Silverman and Ingraham39 A large, sham-controlled study of upper-room GUV in Connecticut schools published in 1951 found no effect on student absenteeism due to non-specific respiratory illness. Reference Gelperin, Granoff and Linde42
A 1950 study compared rates of infection in Pleasantville, New York, where upper-room GUV was implemented in various congregate areas including schools, churches, and theaters throughout the community, to rates in the nearby untreated village of Mt. Kisco. The primary analysis reported a lower incidence of measles (42 vs 51 cases per 100 susceptible children) and chickenpox (22 vs 50 cases per 100 susceptible children) in the GUV-treated community, Reference Wells and Holla40 while a companion analysis found no evidence of effectiveness against non-specific colds. Reference Downes41
Evidence from military barracks
Four (4/40, 10%) non-randomized studies in military barracks between 1945 and 1955 showed protective effects of GUV. Reference Wheeler, Ingraham, Hollaender, Lill, Gershon-cohen and Brown47–Reference Macdonald, Tonkinson, Porterfield and Dumbell52 A 1945 study at a New York Naval Training Center evaluated two intensities of GUV in the upper-room and underneath bunks. This study reported a 25% reduction in hospital admissions due to respiratory disease in barracks with higher intensity GUV (90.4 vs 114.3 per 1,000 recruits over the training period, averaging 4.7 weeks), but not with lower intensity GUV. Reference Wheeler, Ingraham, Hollaender, Lill, Gershon-cohen and Brown47 A 1948 study at an Illinois training center using similar interventions found a 20% reduction in hospitalization due to respiratory disease in GUV-treated barracks (1,612 vs 2,020 per 1,000 recruits per year). Reference Miller, Jarrett and Willmon48 A second study at the Illinois base found a 48% reduction in respiratory admission rates during a non-epidemic period (4.9 vs 9.5 per 1,000 recruits per week) and a 19% reduction during an influenza epidemic (69.6 vs 85.6 per 1,000 recruits per week). Reference Langmuir, Jarrett and Hollaender49
Evidence from healthcare
Eight (8/40, 20%) non-randomized studies in healthcare settings generally reported protective effects. Reference Jordan29,Reference Del Mundo and McKhann53–Reference Twining McMath and Hussain60 Five of these studies evaluated GUV “barriers” (Figure 1). Reference Del Mundo and McKhann53,Reference Sommer and Stokes55–Reference Robertson, Doyle and Tisdall58,Reference Twining McMath and Hussain60 A 1941 study at Boston Children’s Hospital showed a lower rate of hospital-acquired respiratory infections in a unit with GUV barriers at cubicle entrances (2% vs 5.7%). Reference Del Mundo and McKhann53 A 1943 study in a Toronto children’s hospital evaluated GUV barriers plus mechanical ventilation and reported a reduction in hospital-acquired respiratory infections in the room with GUV barriers plus ventilation (21% vs 60%). Reference Robertson, Doyle and Moulton57,Reference Robertson, Doyle and Tisdall58 However, the same study also evaluated the effect of upper-room GUV alone and observed mixed results. A 1942 study in a Boston hospital evaluated a combination of upper-room GUV and GUV barriers and observed lower rates of “coryza” in GUV-treated wards (12% vs 25%). Reference Wheeler, Jones and Moulton56 A 1942 study of GUV barriers in the Children’s Hospital of Philadelphia reported a reduction in “clinical cross-infections” due to pneumococcus (4.1% vs 7.4%). Reference Sommer and Stokes55 A 1960 crossover study in a London infectious disease hospital showed a reduced rate of hospital-acquired chickenpox during periods when GUV barriers were turned on compared to periods when the GUV was off (0.8% vs 3.5%). Reference Twining McMath and Hussain60
Two studies evaluated upper-room GUV alone and reported protective effects. A historical control study in a New York children’s hospital published in 1947 showed a reduction in febrile respiratory illness after the installation of upper-room GUV (2.38% postintervention vs 3.98% preintervention), though this comparison was confounded by factors during World War II including population movement and increasing use of antibiotics. Reference Higgons and Hyde59 A report from the Veterans Administration Hospital in Livermore, California during the 1957 influenza pandemic showed that patients in a GUV-treated tuberculosis ward had an influenza attack rate of 2%, compared to 19% for patients in untreated parts of the hospital and 18% for hospital staff. Reference Jordan29 Notably, these findings were only published in response to a conference proceedings paper and lacked important methodological information.
A second study in Boston Children’s Hospital published in 1942 studied an uncommon intervention consisting of 254 nm GUV in the whole room and observed no difference in the rate of respiratory infections among children in GUV-treated wards versus those in untreated wards. Reference Brooks, Wilson, Blackfan and Moulton54
Evidence from orphanages and daycares
Three (3/40, 7.5%) non-randomized studies in orphanages and day cares produced mixed results. Reference Barenberg, Greene and Greenspan61–Reference Dionne, Soto and Pineau67 A series of reports published between 1940 and 1942 from a New York orphanage found that upper-room GUV reduced acute respiratory infections and provided complete protection from a chickenpox outbreak—0 of 15 children in the GUV-treated ward were infected versus 18 of 19 in the control ward. Reference Barenberg, Greene and Greenspan61–Reference Barenberg, Greene, Greenspan, Greenberg and Moulton63 Another 1948 report from an Illinois orphanage observed lower rates of respiratory cross-infections in both a unit with GUV barriers between cubicles (0.3 cases per year) and a unit with upper-room GUV (2.0 cases per year) compared to a control unit (7.0 cases per year). Reference Rosenstern and Moulton64–Reference Rosenstern66 A 1993 double-blind study in three Canadian day-care centers found that self-contained GUV “air sterilizer” and air filtration devices did not reduce the incidence of upper respiratory infections. Reference Dionne, Soto and Pineau67
Evidence from other congregate settings
There were 3 (3/40, 7.5%) non-randomized studies in other congregate settings such as offices and a detention center. Reference Menzies, Popa, Hanley, Rand and Milton68–Reference Kingston, Lidwell and Williams72 A large, double-blind, multiple crossover study in three Canadian office buildings in 2003 found that in-duct GUV caused a significant reduction in self-reported work-related respiratory symptoms (adjusted odds ratio (aOR) 0.6 [95% confidence interval 0.4–0.9]) and mucosal symptoms (aOR 0.7 [0.6–0.9]). Reference Menzies, Popa, Hanley, Rand and Milton68,Reference Menzies, Pasztor, Rand and Bourbeau71
In contrast, a six-year study of upper-room GUV in a Washington, D.C. juvenile detention center published in 1948 found no reduction in hospital admissions for respiratory infections, despite repeatedly increasing the intensity of the intervention. Reference Schneiter, Hollaender and Caminita69,Reference Dubuy, Dunn, Brackett, Dreessen, Neal and Posner70
Other observational studies
Two studies in Chinese healthcare facilities—one cross-sectional study in 2019 and one retrospective cohort study in 2020—reported that GUV use was associated with lower rates of tuberculosis infection among healthcare workers, though the reports contain little information regarding how GUV was used. Reference Chen, Gu and Wang73,Reference Xie, Zhou and Chi74
Randomized controlled trials
Three (3/40, 7.5%) RCTs assessed GUV effectiveness. Reference Brickner, Vincent and Nardell30–Reference Shoubridge, Brass and Crotty33,Reference Brickner and Vincent75 The Tuberculosis Ultraviolet Shelter Study was a double-blind, placebo-controlled RCT of upper-room GUV in 14 homeless shelters across six U.S. cities designed to assess effectiveness against tuberculosis transmission between 1997 and 2004. Reference Brickner, Vincent and Nardell30 However, the authors reported that the trial was inconclusive for the prespecified primary effectiveness outcome because of “insufficient numbers of documented [tuberculosis] skin test conversions.” Reference Nardell, Bucher and Brickner31
A cluster-randomized, multiple-crossover trial published in 2025 evaluated upper-room and self-contained GUV devices in the common areas of four Australian long-term care facilities. Reference Shoubridge, Brass and Crotty33 The study did not find a significant reduction in its primary outcome of acute respiratory infections per zone per cycle. However, a posteriori secondary time-series analysis showed a 12.2% reduction in the cumulative number of infections over the study period. Low rates of respiratory infection during the COVID-19 pandemic required the study duration to be extended from a planned 32 weeks to 110 weeks to accrue sufficient events.
A 2025 sham-controlled trial that evaluated in-duct GUV in the homes of children with asthma demonstrated a significant reduction in a secondary a posteriori outcome of respiratory infections (Relative Risk 0.53, 95% CI: 0.31–0.92). Reference Bernstein, Seth and Rao32
Safety findings
Safety monitoring in controlled trials
Ten (10/40, 25%) controlled studies mentioned safety outcomes. Monitoring varied considerably. Two contemporary RCTs included rigorous safety monitoring: the Tuberculosis Ultraviolet Shelter Study in homeless shelters found no significant difference in the frequency of self-reported eye or skin symptoms between active GUV and placebo periods, Reference Nardell, Bucher and Brickner31 and the 2025 RCT in long-term care facilities observed no adverse events. Reference Shoubridge, Brass and Crotty33 In contrast, safety reporting in earlier studies was often sparse, with unclear surveillance methods, reporting mechanisms, or adverse event criteria. Seven studies noted a lack of adverse events, Reference Del Mundo and McKhann53,Reference Shoubridge, Brass and Crotty33,Reference Dubuy, Dunn, Brackett, Dreessen, Neal and Posner70,Reference Gelperin, Granoff and Linde42,43,Reference Twining McMath and Hussain60,Reference Nardell, Bucher and Brickner31 while three others described rare cases of skin erythema and photokeratitis. Reference Bahlke, Silverman and Ingraham39,Reference Miller, Jarrett and Willmon48,Reference Jarrett and Hollaender50 Reports were often anecdotal and lacked denominators to calculate incidence, Reference Bahlke, Silverman and Ingraham39,Reference Gelperin, Granoff and Linde42,Reference Miller, Jarrett and Willmon48,Reference Jarrett and Hollaender50 making it difficult to assess clinical relevance or statistically compare outcomes across studies.
Safety of 222 nm far-UVC
Six (6/40, 15%) studies focused primarily on safety of 222 nm far-UVC. In small controlled studies of direct human skin exposure (with 1 to 20 participants), no erythema was observed, including at high doses (18,000 mJ/cm2, >36 times exposure limits set by the American Conference of Governmental Industrial Hygienists (ACGIH) of 479 mJ/cm2). Reference Fukui, Niikura and Oda76–78 While one study detected slightly but significantly increased DNA damage markers after a dose of 500 mJ/cm2, Reference Fukui, Niikura and Oda76 a more detailed immunohistochemical analysis demonstrated that DNA damage was confined to the most superficial, non-dividing layer of the epidermis, mitigating risk of carcinogenesis. Reference Hickerson, Conneely and Hirata Tsutsumi79 Regarding ocular safety, studies involving direct or prolonged exposure found no evidence of short- or long-term damage at or below current exposure limits, though one report noted transient, mild discomfort after direct exposures. Reference Kousha, O’Mahoney, Hammond, Wood and Eadie34,Reference Sugihara, Kaidzu and Sasaki80–Reference Sugihara, Kaidzu, Sasaki and Tanito82
Overexposure due to improper use, maintenance, training, or education
Five (5/40, 12.5%) case reports and case series described self-limited skin erythema and photokeratitis following overexposure to 254 nm GUV. Reference Brickner and Vincent75,Reference Rose and Parker83–Reference Talbot, Jensen, Moffat and Wells86 These incidents were consistently attributed to installation or maintenance errors, such as installing a fixture upside-down, inadvertently placing a GUV bulb in a desk lamp, failing to install proper shielding, or turning on an unshielded lamp in an occupied space.
Discussion
In this systematic review of 40 studies spanning 85 years, we summarize the evidence on the clinical effectiveness and safety of GUV for preventing transmission of respiratory illnesses and infections; detailed descriptions of all included studies are provided in the supplementary material. The evidence base consists predominantly of non-randomized studies from the mid-20th century with methodological limitations, and four RCTs. While early non-randomized studies indicated that GUV can be effective against highly contagious infections transmitted through air, like measles, Reference Wells, Wells and Wilder35,Reference Wells and Holla40,43 more recent and rigorous RCTs have been inconclusive or shown only modest effects. Reference Nardell, Bucher and Brickner31,Reference Shoubridge, Brass and Crotty33 These mixed findings are difficult to interpret given methodological limitations in study design and intervention implementation, as well as epidemiological challenges related to outcome selection, disease incidence, and transmission dynamics. Properly installed and maintained 254 nm GUV systems appear safe, though the possibility of overexposure due to inappropriate use necessitates professional oversight. 222 nm far-UVC within established exposure limits appears not to cause adverse effects, though long-term safety data remain limited.
Multiple methodological factors limit the interpretability of this evidence base. Many non-randomized studies exhibited inconsistent data reporting and possible selective reporting. Reference Wells, Wells and Wilder35,Reference Bahlke, Silverman and Ingraham39,Reference Wells and Holla40,Reference Rosenstern66 Many studies failed to fully characterize the intervention—crucial details on UV fluence, ventilation, and maintenance were often unreported or underreported, making it impossible to determine whether null results reflected ineffectiveness or insufficient UV delivery. Moreover, accurate radiometric measurements of UV are technically challenging, meaning reported values may be subject to measurement error. Reference Blatchley, Belenky and Claus87 Additionally, most GUV studies assume that air is well-mixed, but actual airflow deviates from the well-mixed condition in many settings, complicating prediction of GUV system performance. Reference Nardell, Bucher and Brickner31,Reference Blatchley, Belenky and Claus87 A notable exception is the 1954 Air Hygiene Committee study, which stands out for its comprehensiveness: rather than tracking a few target infections, the investigators recorded and categorized the cause of every school absence over three years across a broad range of potential confounders and verified intervention fidelity through regular UV intensity and air sampling measurements. 43 For these reasons, we consider it the most methodologically rigorous study in this review.
Resolving uncertainty about GUV effectiveness would require studies in which GUV systems are properly designed, installed, and maintained per established guidelines 88,89 and deliver adequate GUV across spaces where transmission occurs. Verifying that these conditions are met requires fluence rate field modeling to inform installation design, radiometric and ideally microbiological measurements to confirm UV delivery, and detailed reporting of intervention characteristics and measurement methods following established guidance. Reference Blatchley, Belenky and Claus87
Outcome selection may partially explain the inconsistency across clinical GUV studies. The maximum effectiveness of air treatment is inherently constrained by the proportion of transmission that occurs through air. Reference Wang, Prather and Sznitman14,Reference Sage, Lowen and Lakdawala90 Many respiratory pathogens also transmit via contact with surfaces and direct deposition of larger respiratory particles at close range, although the contribution of each route varies by pathogen and is difficult to isolate from concurrent transmission through air. Reference Leung and Milton11,Reference Leung15 Transmission by these non-airborne routes may not be preventable by air-specific treatment, so air treatment alone should not be expected to reduce all respiratory infections. Yet this assumption is made implicitly when broad, non-specific outcomes are chosen for evaluations of air treatment interventions. Such outcomes may include illnesses transmitted through routes other than air or even non-infectious conditions. These non-preventable cases dilute the observable effect, biasing results toward no effect. This pattern appears in the literature: seminal work by Wells et al. Reference Wells, Wells and Wilder35,Reference Wells and Holla40,Reference Downes41 and the Air Hygiene Committee 43 demonstrated GUV effectiveness against measles and chickenpox—diseases transmitted predominantly through air Reference Leung15 —but not against undifferentiated respiratory illnesses. Similarly, studies measuring broader outcomes such as all-cause absenteeism or undifferentiated acute respiratory infections often yielded null or mixed results. Reference Shoubridge, Brass and Crotty33,Reference Gelperin, Granoff and Linde42 Because measles and chickenpox now occur too rarely in well-immunized populations to serve as practical endpoints, modern trials face a tradeoff between specific outcomes (e.g., PCR-confirmed SARS-CoV-2 or influenza) that are more costly to measure and risk being statistically underpowered, and broader outcomes that are less expensive but reduce the observable effect size.
Real-world effectiveness depends on epidemiological context, including community disease rates and vaccine coverage; unpredictable seasonal incidence makes accurate sample size estimation a persistent challenge. Some of the clearest evidence of effectiveness came from situations with high force of infection (e.g., during measles epidemics, where attack rates fell from 62% to 22% Reference Wells, Wells and Wilder35 or in military barracks, where respiratory admissions fell 25% Reference Wheeler, Ingraham, Hollaender, Lill, Gershon-cohen and Brown47 ). In contrast, both major RCTs of upper-room GUV encountered lower-than-expected infection rates leading to insufficient statistical power. Reference Nardell, Bucher and Brickner31,Reference Shoubridge, Brass and Crotty33 Future studies may benefit from conservative, evidence-based incidence estimates and adaptive designs that allow for enrollment extension or endpoint modification to account for epidemiological uncertainty.
Evaluations of place-based interventions are also complicated by contamination, where participants acquire infections outside treated airspace, diluting observable effects. This challenge influences the choice of randomization unit, which involves trade-offs: smaller clusters (e.g., classrooms) increase statistical power but also increase contamination risk, while larger clusters (e.g., schools) decrease contamination risk but also decrease statistical power and increase study costs. Reference Blanco, Harris and Magder91,Reference Rutterford, Copas and Eldridge92 This problem may be partially addressed by genomic outbreak tracing to differentiate infections acquired within versus outside the study environment. Reference Klompas, Baker and Rhee93
Regarding safety, our review confirms that properly installed 254 nm GUV systems appear safe for occupants, though case reports of overexposure from inappropriate use highlight the need for professional installation and maintenance. The emergence of 222 nm far-UVC is noteworthy, as its potential safety for direct exposure may permit continuous whole-room treatment, which may combat both long- and short-range transmission through air while providing concurrent surface disinfection. Small, controlled studies found no adverse effects even at exposures far exceeding recommended limits, with DNA damage confined to superficial, non-dividing epidermal cells and no ocular damage at recommended doses. However, long-term data from larger and more diverse populations would assess the risk of rare adverse events, especially cumulative ocular effects. Additionally, far-UVC sources can produce small amounts of ozone that can be a respiratory irritant and react with indoor chemicals to form potentially harmful secondary products. Reference Blatchley, Brenner and Claus19 While this risk is likely small and mitigable with ventilation, Reference Tahsini, Zhang and Goss94 particularly in healthcare settings where ventilation is high, it warrants monitoring in future GUV studies.
This review has several limitations. The heterogeneity in study designs, interventions, and outcomes precluded meta-analysis and formal evaluation of certainty of evidence. Interpreting and extracting data from historical reports was challenging due to changes in scientific language and reporting standards, though independent dual extraction and discussion among reviewers helped minimize potential misunderstandings. Finally, to maintain focus, this review excluded important related literature, such as studies of GUV on transmission of tuberculosis through air using sentinel animals Reference Riley, Mills, O’Grady, Sultan, Wittstadt and Shivpuri95–Reference Mphaphlele, Dharmadhikari and Jensen97 and research on GUV for surgical site infection prevention. Reference Berard and Gandon98,Reference Hart, Postlethwait, Brown, Smith and Johnson99
Despite 85 years of clinical research and demonstrated preclinical efficacy, the effectiveness of GUV against the broad burden of respiratory infection remains uncertain. Several older studies indicated that GUV reduces infections with high transmissibility through air, while inconclusive or modest results of subsequent trials may reflect methodological limitations and epidemiological challenges that tend to bias studies toward the null, rather than a true absence of effect. Well-designed, adequately powered RCTs would provide evidence to guide implementation of GUV for public health and the prevention of healthcare-associated respiratory infections. This will require rigorous intervention design, quantitative verification of UV delivery, and specific, objectively defined endpoints. Such work may also inform ventilation guidelines such as American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) Standards 62.1 and 241. 100 Several RCTs of GUV are currently underway, which we hope will overcome the limitations identified in this review and provide clarity on GUV’s clinical effectiveness. Reference Rockwood101–104 As this evidence matures, GUV continues to be used as part of a layered defense against infectious disease transmission through air.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/ice.2026.10549.
Data availability statement
The study protocol and data set are available at https://osf.io/fg6qx.
Acknowledgements
None.
Author contribution
Conception and design: G.P.H.-R., E.C.S., J.D.W., L.J.R.
Analysis and interpretation of the data: G.P.H.-R., S.A.H., E.C.S., E.R.B., D.T.K., S.B.M., C.C.M., D.H.S., D.N.W., J.D.W., L.J.R.
Drafting of the article: G.P.H.-R.
Critical revision for important intellectual content: All authors.
Final approval of the article: All authors.
Obtaining funding: G.P.H.-R., L.J.R.
Administrative and technical support: J.M.T.
Financial support
This work was supported by the National Institute for Occupational Safety and Health (NIOSH) and Blueprint Biosecurity. G.P.H.-R. received compensation from Blueprint Biosecurity, and D.H.S. and E.R.B. received compensation from NIOSH, for their work on the project. S.A.H. is funded in part by the NIH under the National Institute of Allergy and Infectious Diseases award K23AI182452.
Competing interests
E.R.B. is President of the International Ultraviolet Association (IUVA). L.J.R., R.A.M., and C.C.M. are members of the Board of Directors of IUVA. L.J.R. is an adviser to Beacon, a GUV manufacturer. J.D.W. reported grants from National Institute on Alcohol Abuse and Alcoholism (NIAAA) of the National Institutes of Health (NIH) (Nos. 1K01AA028258, 1R01AA032254), Johnson & Johnson (through the Yale Open Data Access Project), Arnold Ventures, and the Food and Drug Administration outside the submitted work.
Disclaimer
The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the National Institute for Occupational Safety and Health, Centers for Disease Control and Prevention. This manuscript has been reviewed in accordance with the policy of the U.S. Environmental Protection Agency, and approved for publication. Approval does not signify that the contents necessarily reflect the view and policies of EPA. Mention of trade names, products or services does not convey official EPA approval, endorsement, or recommendation.
Open access
