Occupational and environmental risk factors for salivary gland cancer: a systematic literature review Authors: Denny Choi, Khayati Patel, Andrew Zarnke, David Boyce-Fappiano, Donald Cole, Sarah McCormick Rhodes Published: 2026 ABSTRACT AbstractOccupational and environmental risk factors for salivary gland cancer (SGC) are largely unrecognised. This systematic review is the first to characterise the relationship between employment in specific industries or occupations and risk of SGC, as well as exposure to key occupational and environmental hazards.Studies published up to September 2023 were collected from three electronic databases and systematically screened. The Health Assessment Workplace Collaborative programme was used to tag studies based on pre-established inclusion and exclusion criteria. Studies eligible for inclusion (n=24) underwent an in-depth review to extract key study characteristics and findings, including effect estimates and 95% CIs. Studies were evaluated for influence of key biases. To identify priority sectors and hazards for intervention, effect direction plots were created to analyse the number and quality of studies reporting elevated rates of SGC across industry, occupation and exposure groups.24 articles were included for analysis. Exposure misclassification and confounding bias were a common concern across studies. Industry and occupation groups with the strongest relationship to SGC included cleaning services, material handling, food services, rubber/plastics production, engineering and construction/painting. Exposures with the strongest relationship to SGC included silica dust, cement dust, chlorinated solvents, formaldehyde and white spirits.Several occupational exposures were associated with an elevated risk of SGC, whereas environmental evidence was more limited. Additional research involving larger cohorts with quantitative exposure data is needed to further establish relationships between occupational and environmental exposures and often-overlooked rare cancers like SGC.PROSPERO registration number: CRD42023468037.Keywords: Occupational Health, Environmental Exposure, Dust, Solvents, Nickel WHAT IS ALREADY KNOWN ON THIS TOPIC WHAT THIS STUDY ADDS Our review identified multiple hazards associated with an increased occurrence of SGC, namely silica dust, cement dust, chlorinated solvents, formaldehyde and white spirits. Specific industries and occupations were also implicated, including the following: cleaning services, material handling, food services, rubber/plastics production, engineering and construction/painting. Our review identified multiple hazards associated with an increased occurrence of SGC, namely silica dust, cement dust, chlorinated solvents, formaldehyde and white spirits. Specific industries and occupations were also implicated, including the following: cleaning services, material handling, food services, rubber/plastics production, engineering and construction/painting. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Health and safety measures or policy changes targeting the occupations and exposures highlighted by this review could protect workers and community members from SGC in the future. Additionally, rare cancers like SGC make up a large percentage of total cancer diagnoses; therefore, researchers should report results for these cancers to prevent under-reporting. Health and safety measures or policy changes targeting the occupations and exposures highlighted by this review could protect workers and community members from SGC in the future. Additionally, rare cancers like SGC make up a large percentage of total cancer diagnoses; therefore, researchers should report results for these cancers to prevent under-reporting. Introduction Salivary gland cancer (SGC) encompasses a heterogenous group of rare malignancies most often located in the head and neck (HN) region, with an estimated annual incidence rate ranging from 0.64 to 3 cases per 100 000 people.1 2 Almost 80% of salivary gland tumours occur in the parotid gland and the majority of these are benign in nature, whereas salivary tumours arising from the submandibular, sublingual or minor salivary glands have a higher propensity to be malignant.3 The most common histological types, representing over 50% of all SGCs, include adenoid cystic carcinoma and mucoepidermoid carcinoma.3 Malignant salivary gland tumours can present with pain and facial nerve involvement, and often carry an increased risk of local reoccurrence, which can be debilitating for the patient.4 SGC impacts more men than women and is most common among elderly populations.2 Established risk factors for SGC are medical or behavioural, whereas occupational and environmental factors are largely unrecognised. To date, the relationship between major SGCs and common cancer risk factors like smoking and alcohol consumption is mixed. However, heavy smoking above 40 pack-years has been more consistently associated with SGCs,5 and elevated SGC among smokers has focused largely on Warthin tumour, a type of benign salivary gland tumour.4 Other risk factors potentially associated with SGC include several infectious disease agents, such as Epstein-Barr virus, HIV and human papillomavirus.5 By comparison, occupational and environmental risk factors for SGC remain largely unknown. Radiation exposure is the only well-established risk factor of this kind, as demonstrated by reports of elevated SGC rates observed among atomic bomb survivors, those with occupational radiation exposure and those exposed to therapeutic radiation to the HN region.6 7 The National Cancer Institute estimates that rare cancers, including SGC, account for 25% of all new cancer cases in the USA, which highlights the growing need to identify key risk factors for these illnesses.8 The authors of this study are employed at the Occupational Health Clinics for Ontario Workers (OHCOW), a network of clinics funded by the Ontario Ministry of Labour, Immigration, Training and Skills Development. OHCOW has observed an unexpectedly high rate of SGC among plastics and rubber workers across our clinics over the last decade. To support prevention initiatives, this systematic literature review aims to elucidate our understanding of occupational and environmental risk factors for SGC by identifying: (1) industries, occupations or job types that have been associated with elevated rates of SGC among workers, (2) occupational/workplace exposures that have been associated with elevated rates of SGC among workers and (3) environmental risk factors (eg, non-occupational exposures experienced via contact with air, soil, water, food, etc) that have been associated with elevated rates of SGC among community members. Methods Study design We used the Health Assessment Workplace Collaborative (HAWC) programme to conduct literature searches using predefined search terms (available in online supplemental table S1). Literature searches (completed on 13 September 2023) were conducted on three databases: PubMed, Web of Science and OVID Embase. The protocol for this systematic review was registered in the international database of prospectively registered systematic reviews (PROSPERO) on 9 October 2023 (registration ID: CRD42023468037). No new studies were included after this date. The abstract and title screening process (phase 1) involved two reviewers blinded to each other’s results, who tagged studies based on pre-established inclusion and exclusion criteria, along with other relevant study characteristics (eg, study type, outcome type). Any discrepancies were discussed in a consensus meeting with an additional reviewer to provide a separate appraisal and reach majority consensus. The reviewers did not review articles they consulted on, authored or coauthored. The comprehensive inclusion and exclusion criteria for this review are summarised in table 1 using the Population (including animal species), Exposure, Comparator and Outcomes framework. Studies were included only if they reported primary relative risk (RR) data for SGC (major or minor), salivary duct cancer or tracheal cancer, and studies reporting primary risk data for broader cancer groupings were excluded (eg, HN) (table 2). Table 1Study inclusion and exclusion criteria using the Population, Exposure, Comparison, Outcomes (PECO) frameworkPECO criteriaInclusion criteriaExclusion criteriaPopulation Study focused on human populations.Study included workers in industrial/occupational settings as well as workers exposed to occupational hazards.Study included community members exposed to an environmental hazard. Study focused on non-human animal populations and/or cell cultures. Exposure Employment in a specific industry, occupation or job type.Occupational/workplace hazard.Environmental hazard (eg, non-occupational exposures experienced through contact with air, soil, water, food, etc). Studies with radiation or radiotherapy as the primary exposure.Study focuses on an exposure other than occupational, industrial and/or environmental exposure. Comparison Workers compared with workers in other industries/occupations: other workers who are not employed in the industry/occupation that is classified as the ‘exposed’ group (eg, rubber workers (exposed group) compared with all other workers in the cohort (unexposed group)).Workers compared with other workers in same industry but differing exposure: workers working in the same industry/occupation who were not exposed to the occupational or environmental exposure of interest, or who experienced low levels of exposure (eg, high exposure compared with no or limited exposure to a specific hazard).Community members with higher relative to lower exposure to environmental hazards: study participants who were unexposed or experienced low levels of exposure to an environmental hazard (compared with those exposed to a medium or high level).Workers compared with population composed of working and non-working population: local or general population (eg, external comparison workers in a specific occupation/industry), or study participants exposed to a specific environmental hazard, compared with the local or general population. Outcome Study reports primary relative risk data for salivary gland cancer (major or minor), salivary duct cancer or tracheal cancer (ultimately excluded prior to phase 2). Studies reporting primary risk data for cancers other than salivary gland cancer (major or minor), salivary dust cancer or tracheal cancer. Other Title and abstract not available in English or French. Radiation studies were excluded from the review, as radiation exposure is already a well-established risk factor for salivary gland cancer. Study inclusion and exclusion criteria using the Population, Exposure, Comparison, Outcomes (PECO) framework Study focused on human populations. Study focused on human populations. Study included workers in industrial/occupational settings as well as workers exposed to occupational hazards. Study included workers in industrial/occupational settings as well as workers exposed to occupational hazards. Study included community members exposed to an environmental hazard. Study included community members exposed to an environmental hazard. Study focused on non-human animal populations and/or cell cultures. Study focused on non-human animal populations and/or cell cultures. Employment in a specific industry, occupation or job type. Employment in a specific industry, occupation or job type. Occupational/workplace hazard. Occupational/workplace hazard. Environmental hazard (eg, non-occupational exposures experienced through contact with air, soil, water, food, etc). Environmental hazard (eg, non-occupational exposures experienced through contact with air, soil, water, food, etc). Studies with radiation or radiotherapy as the primary exposure. Studies with radiation or radiotherapy as the primary exposure. Study focuses on an exposure other than occupational, industrial and/or environmental exposure. Study focuses on an exposure other than occupational, industrial and/or environmental exposure. Workers compared with workers in other industries/occupations: other workers who are not employed in the industry/occupation that is classified as the ‘exposed’ group (eg, rubber workers (exposed group) compared with all other workers in the cohort (unexposed group)). Workers compared with workers in other industries/occupations: other workers who are not employed in the industry/occupation that is classified as the ‘exposed’ group (eg, rubber workers (exposed group) compared with all other workers in the cohort (unexposed group)). Workers compared with other workers in same industry but differing exposure: workers working in the same industry/occupation who were not exposed to the occupational or environmental exposure of interest, or who experienced low levels of exposure (eg, high exposure compared with no or limited exposure to a specific hazard). Workers compared with other workers in same industry but differing exposure: workers working in the same industry/occupation who were not exposed to the occupational or environmental exposure of interest, or who experienced low levels of exposure (eg, high exposure compared with no or limited exposure to a specific hazard). Community members with higher relative to lower exposure to environmental hazards: study participants who were unexposed or experienced low levels of exposure to an environmental hazard (compared with those exposed to a medium or high level). Community members with higher relative to lower exposure to environmental hazards: study participants who were unexposed or experienced low levels of exposure to an environmental hazard (compared with those exposed to a medium or high level). Workers compared with population composed of working and non-working population: local or general population (eg, external comparison workers in a specific occupation/industry), or study participants exposed to a specific environmental hazard, compared with the local or general population. Workers compared with population composed of working and non-working population: local or general population (eg, external comparison workers in a specific occupation/industry), or study participants exposed to a specific environmental hazard, compared with the local or general population. Study reports primary relative risk data for salivary gland cancer (major or minor), salivary duct cancer or tracheal cancer (ultimately excluded prior to phase 2). Study reports primary relative risk data for salivary gland cancer (major or minor), salivary duct cancer or tracheal cancer (ultimately excluded prior to phase 2). Studies reporting primary risk data for cancers other than salivary gland cancer (major or minor), salivary dust cancer or tracheal cancer. Studies reporting primary risk data for cancers other than salivary gland cancer (major or minor), salivary dust cancer or tracheal cancer. Title and abstract not available in English or French. Title and abstract not available in English or French. Radiation studies were excluded from the review, as radiation exposure is already a well-established risk factor for salivary gland cancer. Table 2Description of included studies and assessment of biasAuthorStudy designStudy populationReference populationExposure assessmentBias assessment*†ConfoundingExposure misclassificationOutcome misclassificationOther biases‡Chen et al20Case–controlTaiwan: 136 913 cancers/215 casesGeneral Taiwanese population matched to all cases by age, sex and township residenceRegistry−+−−Deen et al29CohortDenmark: 38 613 people/7 casesResidents in apartment sections using PCB-free materialsAir sampling and interview−−−−Mowery et al18CohortUSA: 8 877 971 veterans/3614 casesVietnam War veterans without exposure to Agent OrangeRegistry++−−Cong15CohortChina: 550 000 cancersRest of Shanghai Cancer Registry subjectsRegistry+−−−Radoï et al17Case–controlFrance: 73 cases/3555 controlsFrench population, frequency matched to cases by sex/age, stratified by socioeconomic statusJob exposure matrices based on face-to-face interviews and questionnaires−−−−Eom et al22EcologicalKorea: 341 793 adults/346 casesGeneral Korean populationRegistry+−−−Pan et al5Case–controlCanada: 132 cases/3076 controlsResidents from participating provinces with no prior diagnosis of cancer (matched: age group, sex)Mailed questionnaires++−+Seidler et al30CohortGermany: 16 441 men/5 casesNon-mining population in Saxony-AnhaltEmployer records and classification by a former technical manager+−−−Yiet al19CohortKorea: 176 605 male veteransVeterans deemed to have low category of exposure based on exposure indexRegistry−−−−Kim et al31Case reportKorea: 1 case–Interview+−−−Wilson et al24Case–controlUSA: 2405 cases/9420 controlsNon-infectious disease deaths (matched for age, race, sex and region)Job exposure matrices based on occupation and industry coding from death certificates−−−Sorahan et al32CohortUK: 2160 male workers/1 caseCancer registration rates for England and WalesEmployer records and review by former occupational hygienist−−−−Swanson and Burns33Case–controlUSA: 163 cases/3751 controlsResidents of tricounty Detroit areaTelephone interviews−+−+Horn-Ross et al16Case–controlUSA: 141 cases/271 controlsIndividuals who have never been employed in the rubber industryRegistry and interviews+−−−Zheng et al26Case–controlChina: 41 cases/414 controlsResidents in urban Shanghai areaNurse interviews and questionnaires−+−+Swanson and Burns34Case–controlUSA: 83 cases/1972 controlsResidents of Detroit, frequency matched by age, race and sexTelephone interviews−+−+Spitz et al35Case–controlUSA: 64 cases/128 controlsOther patients with cancer not exposed to the agents of interest (eg, hair dye), matched by age, sex and raceQuestionnaire−+−−Corbett and O’Neill21CohortAustralia: 1004 male workers/2 casesCancer incidence rates from Regional Cancer Registry (New South Wales)Registry and interviews++−+Easton et al23CohortUK: 3530 workers/2 casesCancer mortality rates for the regions (Merseyside and Cheshire) and the UKRegistry++−−Olsen et al28CohortDenmark: 398 male workers with cancers/4 casesNo ‘population denominators’; proportion of cancer in companies versus among Danish employeesRegistry and air sampling++−−Swanson and Belle36CohortUSA: 1070 male workers/2 casesPopulation in Metropolitan DetroitEmployer records++−+Graham et al27EcologicalCanada: 6 027 795 menResidents in non-asbestos-mining counties in QuebecRegistry++−−Milham25CohortUSA: 589 males (plumbers)/5 casesUS age-specific mortality ratesDeath certificates++++Mancuso and Brennan37CohortUSA: 5735 men, 426 women/11 cases totalCancer mortality rates from metropolitan counties in OhioUnion records+++−*The bias domains assessed for each study were informed by IARC’s Bias Assessment Resource11 and included the following: exposure misclassification, outcome misclassification, confounding bias, healthy worker bias (survivor/hire), recall bias and selection bias. Healthy worker bias was not applicable to the environmental studies reviewed.†‘+’ denotes moderate to major bias influence; ‘−’ denotes no to minimal bias influence.‡The ‘Other biases’ category includes: healthy worker bias, selection bias and recall bias.IARC, International Agency for Research on Cancer; PCB, polychlorinated biphenyls. Description of included studies and assessment of bias The bias domains assessed for each study were informed by IARC’s Bias Assessment Resource11 and included the following: exposure misclassification, outcome misclassification, confounding bias, healthy worker bias (survivor/hire), recall bias and selection bias. Healthy worker bias was not applicable to the environmental studies reviewed. ‘+’ denotes moderate to major bias influence; ‘−’ denotes no to minimal bias influence. The ‘Other biases’ category includes: healthy worker bias, selection bias and recall bias. IARC, International Agency for Research on Cancer; PCB, polychlorinated biphenyls. At the end of phase 1, occupational health physicians (DB-F, DCo) reviewed all papers reporting solely on salivary duct cancer (n=1) and tracheal cancer (n=21). These studies were excluded from analysis, as many papers grouped tracheal with lung/bronchial tumours, which were mainly squamous cell carcinoma and not SGC.9 Studies included after phase 1 proceeded to a full-text review and detailed data extraction (phase 2), where an extensive list of variables was extracted from each study (eg, study type, exposure assessment, SGC effect estimates and 95% CI, etc, with a complete list presented in online supplemental table S2). A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) diagram of our study inclusion and exclusion process during phase 1 and phase 2 is illustrated in online supplemental figure S1. Bias assessment Four study authors with training in exposure assessment and/or epidemiology, and blinded to others’ results, assessed the level of bias present in each study following methods used by the International Agency for Research on Cancer (IARC).10 11 The authors commented on the following bias domains: misclassification of exposure, misclassification of outcome, healthy worker bias, confounding by lifestyle factors (eg, smoking), selection bias and recall bias. These biases are common in occupational and environmental studies. In particular, exposure misclassification is a well-recognised occurrence, where study participants are incorrectly categorised in relation to their true exposure level. This often occurs when there are incomplete or inaccurate job titles, lack of personal or environmental measurement data, use of large geographical scale as proxy for environmental exposure (eg, county of residence), or inaccurate questionnaire or job exposure matrix data. Healthy worker bias was included given it is a common form of bias in occupational studies (eg, cohort, nested case–control); however, this form of bias was not applicable to environmental studies. Additional collaborators with training in occupational medicine and epidemiology were consulted if consensus on the level of bias was not reached for a given study. Studies were classified as having a ‘no to minimal’ level of concern or ‘moderate to major’ level of concern for each bias domain, and then later classified into levels of high, medium and low bias at the study level (examples in online supplemental table S3; figures1 2). Industries and occupations associated with salivary gland cancer (SGC). Effect Direction: ▲ = Positive association, with a total of N ≤ 5 SGC cases across all sub-categories of the industry/occupation group; ▲▲ = Positive association, with a total of 5 < N ≤ 20 SGC cases across all sub-categories of exposure group; ▲▲▲ = Positive association, with a total of N > 20 SGC cases across all sub-categories of the industry/occupation group; ◂▸ = potential association given mixed results among sub-categories (i.e. some effects in positive direction, some effects inconclusive or in negative direction). Bias assessment: Green = 0–1 bias domains (of 6 total) classified as having ‘no to minimal bias’; Yellow = 2 domains classified as ‘moderate to major concern’ for bias; Beige = 3 or more bias domains classified as ‘moderate to major concern’ for bias. Total Case Number: a = Most measures reported across all studies were largely imprecise. This symbol indicates relatively high precision, where at least one effect estimate reported by the study for a given sub-category was moderately precise. b = DeBono et al cohort study results for SGC are unpublished and were provided via personal communication (including effect estimates and number of cases).38 c = Proportionate Mortality Ratio (PMR) was reported without a confidence interval. Relative Level of Evidence (LOE): Strong = Two or more studies reported a positive association with SGC (with no to moderate bias present for at least two of those studies (green, yellow) and with at least one of those studies reporting N>20 total SGC cases (3 triangles)). Moderate = At least two studies reported a positive association (and at least one of those studies having minimal to moderate bias). Limited = Only one study (with no to minimal bias present) reported a positive association with SGC. Exception: Material Handling was ranked as Strong LOE given it had 4 studies in agreement regardless of bias present. Red box: Studies with the strongest level of evidence relative to all other studies reviewed. Occupational exposure agents associated with salivary gland cancer (SGC). Effect Direction: ▲ = Positive association, with a total of N ≤ 5 SGC cases across all sub-categories of the industry/occupation group; ▲▲ = Positive association, with a total of 5 < N ≤ 20 SGC cases across all sub-categories of exposure group; ▲▲▲ = Positive association, with a total of N > 20 SGC cases across all sub-categories of the industry/occupation group; ◂▸ = Potential association given mixed results among sub-categories (i.e. some effects in positive direction, some effects inconclusive or in negative direction). Total Case Number: a = Most measures reported across all studies were largely imprecise. This symbol indicates relatively high precision, where at least one effect estimate reported by the study for a given sub-category was moderately precise. Bias assessment: Green = 0–1 bias domains (of 6 total) classified as having ‘no to minimal bias’; Yellow = 2 domains classified as ‘moderate to major concern’ for bias; Beige = 3 or more bias domains classified as ‘moderate to major concern’ for bias. Relative Level of Evidence (LOE): Strong = At least one study (with no to minimal bias and N>20 total SGC cases) reported a positive association with SGC and no other studies showed conflicting evidence. Moderate = At least one study reported a positive association, but other studies showed mixed evidence or had moderate to high levels of bias. Limited = One study with low bias reported mixed associations, or one study with moderate to high bias reported a positive association with SGC. Red box: Studies with the strongest level of evidence relative to all other studies reviewed. Level of evidence criteria To support results synthesis, occupational hygienists grouped the effect estimates extracted from each study into broad categories of industry and occupation using related industry or occupation codes (eg, International Standard Classification of Occupations (ISCO)), where possible (online supplemental table S4), followed by occupational and environmental exposure categories (table 3; online supplemental table S5). For example, in online supplemental table S4 and figure 1, the broad category titled ‘Agriculture’ includes all studies reporting SGC effect estimates for occupation/industry categories related to agriculture, such as ‘General farmers (ISCO 6-11)’, ‘Agricultural production: crops (SIC 010)’ and ‘General farm workers (ISCO 6-21)’. Table 3Occupational exposure agents with a moderate to strong level of evidence for an association with SGCExposureReferenceExposure typeExposure categories/levelsCases/controls (n)Effect measureOR (95% CI)Silica dustZheng et al26OccupationNever versus ever exposed to silica dust in their occupationCombined: 10/612.5 (1.1 to 5.8)Male: 8/523.1 (1.1 to 8.9)Female: 2/91.5 (0.3 to 7.7)Radoï et al17Occupation Never exposedEver exposed<10 years of exposure>10 years of exposure (1) 57/28811 (Reference)(2) 15/6491.55 (0.83 to 2.90)(3) 7/3271.39 (0.62 to 3.20)(4) 8/3221.72 (0.78 to 3.80)Cement dustRadoï et al17Occupation Never exposedEver exposed<10 years of exposure>10 years of exposureCumulative Exposure Index (10 years of exposure (1) 51/23971 (Reference)(2) 22/11371.13 (0.66 to 1.95)(3) 8/5570.82 (0.38 to 1.79)(4) 14/5801.46 (0.77 to 2.78)FormaldehydeWilson et al24OccupationVarying levels of exposure:Low probability/low intensityLow probability/mid-high intensityMid-high probability/low intensityMid-high probability/mid-high intensity Male Male (1) 88/3910.9 (0.70 to 1.15)(2) 11/700.7 (0.35 to 1.26)(3) 6/102.4 (0.86 to 6.75)(4) 31/7791.6 (1.30 to 2.00) Female Female (1) 11/870.7 (0.33 to 1.28)(2) 13/531.1 (0.54 to 2.07)(3) 12/591.3 (0.63 to 2.60)(4) 49/1661.0 (0.73 to 1.49)White spiritsRadoï et al17Occupation Never exposedEver exposed<10 years of exposure>10 years of exposure (1) 41/20091 (Reference)(2) 32/15291.12 (0.69 to 1.82)(3) 14/6501.08 (0.58 to 2.02)(4) 18/8781.17 (0.65 to 2.08)Nickel compounds and alloysHorn-Ross et al16Industry Never exposed≤3000 hours of lifetime exposure>3000 hours of lifetime exposureEver exposed (1) 126/1861.0 (Reference)(2) 6/19.0 (1.1 to 77.2)(3) 5/23.7 (0.71 to 19.8)(4) 12/36.0 (1.6 to 22.0)Petroleum products/solventsZheng et al26OccupationEver versus never exposed14/1041.8 (0.8 to 3.8)Radoï et al17Occupation Never exposedEver exposed†<10 years of exposure†>10 years of exposure (duration)† (1) 35/17831 (Reference)(2) 38/17551.26 (0.78 to 2.04)(3) 12/6550.98 (0.50 to 1.92)(4) 26/10991.47 (0.86 to 2.52)Wilson et al24OccupationVarying levels of exposure:Low probability/low intensityLow probability/mid-high intensityMid-high probability/low intensityMid-high probability/mid-high intensity Male Male (1) 120/3411.4 (1.09 to 1.76)(2) 37/1431.4 (0.92 to 2.06)(3) 243/9411.2 (1.01 to 1.49)(4) 251/10561.2 (0.97 to 1.45) Female Female (1) 31/1200.8 (0.50 to 1.22)(2) 0/9–(3) 31/1670.9 (0.57 to 1.41)(4) 34/1440.9 (0.62 to 1.40)Herbicide/dioxin (Agent Orange)Yi et al19IndustryLow versus high exposureLow=log10E4 score <4.0High=log10E4 score ≥4.0Total exposed cases: 15Low/high: 2/13Low: aHR=1 (Reference)High: aHR=6.96 (1.5 to 32.3)‡Light grey shading=strong level of evidence for this exposure agent; dark grey shading=moderate level of evidence for this exposure agent.‘–’ denotes not applicable.*Methylene chloride only, trichloroethylene only or both.†At least one type of petroleum solvent.‡For some subgroups, the authors reported on the case counts, ratios and/or beta coefficient but not the effect estimates (and vice versa).aHR, adjusted HR; SGC, salivary gland cancer. Occupational exposure agents with a moderate to strong level of evidence for an association with SGC Never exposed Never exposed Ever exposed Ever exposed <10 years of exposure <10 years of exposure >10 years of exposure >10 years of exposure Never exposed Never exposed Ever exposed Ever exposed <10 years of exposure <10 years of exposure >10 years of exposure >10 years of exposure Cumulative Exposure Index (10 years of exposure >10 years of exposure Low probability/low intensity Low probability/low intensity Low probability/mid-high intensity Low probability/mid-high intensity Mid-high probability/low intensity Mid-high probability/low intensity Mid-high probability/mid-high intensity Mid-high probability/mid-high intensity Never exposed Never exposed Ever exposed Ever exposed <10 years of exposure <10 years of exposure >10 years of exposure >10 years of exposure Never exposed Never exposed ≤3000 hours of lifetime exposure ≤3000 hours of lifetime exposure >3000 hours of lifetime exposure >3000 hours of lifetime exposure Ever exposed Ever exposed Never exposed Never exposed Ever exposed† Ever exposed† <10 years of exposure† <10 years of exposure† >10 years of exposure (duration)† >10 years of exposure (duration)† Low probability/low intensity Low probability/low intensity Low probability/mid-high intensity Low probability/mid-high intensity Mid-high probability/low intensity Mid-high probability/low intensity Mid-high probability/mid-high intensity Mid-high probability/mid-high intensity Light grey shading=strong level of evidence for this exposure agent; dark grey shading=moderate level of evidence for this exposure agent. ‘–’ denotes not applicable. Methylene chloride only, trichloroethylene only or both. At least one type of petroleum solvent. For some subgroups, the authors reported on the case counts, ratios and/or beta coefficient but not the effect estimates (and vice versa). aHR, adjusted HR; SGC, salivary gland cancer. Due to the limited number of studies captured by our review, as well as the heterogeneity of studies, it was not feasible to conduct a meta-analysis. Thus, we completed the Synthesis Without Meta-Analysis (SWiM) checklist and created effect direction plots to increase the transparency of our results synthesis process12 (online supplemental table S6). Through an assessment of multiple factors, including the direction and precision of estimates, total number of SGC cases and level of bias present in each study, the industry/occupation/exposure groups were categorised as having a strong, moderate or limited level of evidence (LOE). Results were then visually synthesised in effect direction plots (figures1 2), where the top left quadrant (boxed in red) indicates the industry, occupation or exposure groups with the strongest LOE. In the effect direction plots, the number of arrows in each cell (up to 3) symbolises the total number of SGC cases across each broad category of industry/occupation/exposure; the direction of the arrows indicates either: (arrows up)=elevated SGC occurrence observed across all subcategories OR (arrows facing away from one another)=mixed results, with positive associations observed for some subcategories and null or negative associations observed for others. We assessed the magnitude and precision of effect estimates, indicating where estimates were moderately precise (represented by superscripts) to highlight studies with more reliable estimates. Although effect direction plots often highlight statistical significance (ie, p value <0.05), we did not present this in our plot given p values are largely misinterpreted in non-randomised studies.13 14 Studies were also grouped according to high (beige), medium (yellow) and low (green) levels of bias to identify the most valid studies. In figure 1, an industry/occupation group was assigned to the ‘strong LOE’ category if two or more studies reported a positive association with SGC for that group (with no to moderate bias present for at least two of those studies and with at least one of those studies reporting n>20 total SGC cases). A ‘moderate LOE’ was assigned for groups where at least two studies reported a positive association (and with at least one of those studies having minimal to moderate bias). The ‘limited LOE’ category includes industry/occupation groups where only one study (with no to minimal bias present) reported a positive association with SGC. In figure 2, exposure groups were assigned to the ‘strong LOE’ category if at least one study (with no to minimal bias and n>20 total SGC cases) reported a positive association with SGC and no other studies showed conflicting evidence for that exposure group. A ‘moderate LOE’ was assigned for exposure groups where at least one study reported a positive association, but other studies showed mixed evidence or had moderate to high levels of bias. The ‘limited LOE’ category includes industry/occupation groups where one study with low bias reported mixed associations, or one study with moderate to high bias reported a positive association with SGC. Results After all exclusions were applied, a total of 24 studies were included in table 2.515,37 There were 12 cohort studies, nine case–control studies, two ecological studies and one case report. Most studies were conducted in the USA, but study locations ranged from Australia to Canada, China, Denmark, France, Germany, Republic of Korea and the UK. Studies more frequently measured cancer incidence than cancer mortality, and incidence studies covered more recent time periods. Confounding, exposure misclassification and healthy worker bias were common concerns. Studies with high compared with low levels of bias had a higher frequency of mixed results across broad industry/occupation categories (figures1 2). 11 studies had no to minimal concerns for confounding1617 19 20 24 26 29 32,35; nine for exposure misclassification bias1517 19 22 24 29,32; and nine for healthy worker bias.516 17 20 24 26 27 29 33,35 Exposure classification methods varied across studies, with only a small portion of studies using job exposure matrices (n=3 studies) and under half of the studies conducting work/exposure history questionnaires or interviews as part of their exposure assessment (n=8 studies). Only one study specified that trained professionals (eg, retired nurses) with occupational health experience administered the questionnaires.26 Online supplemental table S4 presents the comprehensive results extracted from all studies reporting on SGC among workers employed in specific industries and occupations. Key results are described below, as well as visually synthesised in an effect direction plot (figure 1). The industry and/or occupation categories with the strongest LOE were: cleaning services (eg, ‘private household cleaners’: OR=1.7, 95% CI 1.04 to 2.85),24 material handling (eg, ‘material handling equipment operator’: OR=5.05, 95% CI 1.71 to 14.84),17 food services (eg, ‘Cooks’: OR=2.04, 95% CI 0.79 to 5.27),17 rubber and plastics production (‘Rubber and Plastics Products Industries’: HR=1.34, 95% CI 0.91 to 1.97),38 engineering and architecture (eg, ‘Engineering, architectural, and surveying services’: OR=2.2, 95% CI 1.09 to 4.26)24 and construction and painting (eg, ‘Construction’: OR=1.35, 95% CI 0.71 to 2.57).17 Industries and/or occupations with a moderate LOE included agriculture (eg, ‘Agricultural Production: Crops’: OR=4.6, 95% CI 1.18 to 17.68)24; transportation (eg, ‘Trucking service’: OR=2.8, 95% CI 0.85 to 9.04)24; hairdressing (‘Hairdressers’: OR=2.7, 95% CI 1.1 to 6.5)33; metalwork and welding (eg, ‘Welders and flame-cutters’: OR=4.55, 95% CI 1.86 to 11.12); and ‘sheet metal workers’ (OR=2.89, 95% CI 1.01 to 8.32).17 Two industry/occupation groups were classified as having a limited LOE, including Agent Orange-exposed veterans (adjusted HR (aHR)=6.96, 95% CI 1.5 to 32.3; RR=1.07, 95% CI 0.97 to 1.14),18 19 plumbing (‘Plumbers and pipefitters’: OR=3.51, 95% CI 1.33 to 9.29),17 among others (eg, chemistry, theatre, etc). Results were mixed across some industry and occupation categories. In one study, multiple education-related categories yielded positive associations with SGC (eg, ‘Colleges and universities’: OR=2.0, 95% CI 1.18 to 3.52, ‘Teachers: Elementary School’: OR=1.8, 95% CI 1.23 to 2.63, among others),24 whereas another study showed no clear association.17 Similar findings were observed in healthcare, air transport and assembly, public administration, sales and typing fields.17 24 33 In rare cases, reduced rates of SGC were observed across categories, such as for bookkeepers.17 Table 3 presents occupational exposure agents associated with strong to moderate LOE for an association with SGC, while online supplemental table S5 presents occupational and environmental exposure agents associated with limited and mixed LOE for an association with SGC. Key exposure results are presented in an effect direction plot (figure 2). Only four occupational studies and one environmental study presented more detailed analyses of intensity and duration of exposure.16 17 24 29 35 Exposures with the strongest LOE for an association with SGC were: silica dust (ORs=up to 3.1, 95% CI 1.1 to 8.9),17 26 cement dust (eg, ‘<10 years of exposure’: OR=2.61, 95% CI 1.29 to 5.2717 as well as ‘females living within 3 km radius of Portland cement plants’: standardised incidence ratio (SIR)=3.03, 95% CI 0.98 to 7.07),22 chlorinated solvents (‘>10 years of exposure’: OR=1.4, 95% CI 0.77 to 2.78),17 formaldehyde (‘Mid-high probability/mid-high intensity’: OR=1.6, 95% CI 1.30 to 2.00)24 and white spirits (‘>10 years exposure’: OR=1.17, 95% CI 0.65 to 2.08).17 Exposures with a moderate LOE include nickel compounds or alloys (‘ever-exposed’: OR=6.0, 95% CI 1.6 to 22.0),16 petroleum products, including solvents (ORs=up to 1.47, 95% CI 0.86 to 2.52)17 and herbicides/dioxin (aHR=6.96, 95% CI 1.5 to 32.3).19 Further, exposures with a limited LOE included hair products (‘ever-exposed’: OR=4.1, 95% CI 1.5 to 11.5),35 aflatoxins in animal feed (‘Employment on cattle farm’: standardised proportional incidence ratio=4.80, 95% CI 1.52 to 11.57),28 dinitrotoluene (‘ever-exposed’: SIR=1.73, 95% CI 0.72 to 5.1630), refractory ceramic fibres (‘>10 years of exposure’: OR=1.85, 95% CI 0.71 to 4.80)17 and wood/straw as fuel (OR=1.6, 95% CI 0.6 to 4.4).26 For these exposure categories, elevated rates of SGC were reported in only one study, the effect estimates were imprecise and multiple studies had moderate to high bias. Mixed results were reported for certain exposure agents, including asbestos, benzene and wood dust. Graham et al27 reported elevated rates among males and females living in asbestos-mining counties, whereas Radoï et al17 reported reduced rates among ever-exposed workers and workers with ≥10 years of exposure. For benzene, Wilson et al and Radoï et al both reported positive associations for some exposure groups, whereas Zheng et al reported no clear association.17 24 26 For wood dust, Zheng et al found a positive association with SGC in environmental settings (ie, emissions from wood and straw burning for cooking), but a negative association for occupational settings.26 Both studies had low case counts yielding imprecise estimates. Results were inconclusive for coal dust, use of coal for cooking and mineral wool, and negative associations were observed for oxygenated solvents, polychlorinated biphenyls (PCBs) and textile dust.17 26 29 Despite imprecise effect estimates, several environmental agents were associated with SGC, such as exposure to fuel emissions (eg, cooking with kerosene: OR=3.5, 95% CI 1.6 to 7.4) for males and females combined)26 and environmental exposure to asbestos based on residence in asbestos-mining counties (RR=2.11 (male); RR=3.35 (female); no CIs reported).27 Notably, as seen in occupational settings, environmental cement dust exposure was also associated with SGCs among women in one study (SIR=3.03, 95% CI 0.98 to 7.07); however, this trend was not observed among men.22 Environmental evidence for SGC is relatively limited, as existing studies had multiple methodological shortcomings, including lack of incorporation of environmental monitoring data, lack of control for confounding by occupation and use of broad geographical catchments as a proxy for exposure (eg, residence in asbestos-mining counties). Discussion Summary of evidence This is the first systematic literature review to investigate occupational and environmental risk factors for SGC. Many studies captured by our review were flagged for key biases (eg, exposure misclassification, confounding) and lacked rigorous exposure assessment. To account for these limitations, we established criteria to weigh studies based on quality and validity. The hazards most strongly associated with SGC were silica dust, cement dust, chlorinated solvents, formaldehyde and white spirits. Specific industries and occupations were also implicated, including employment in cleaning services, material handling, food services, rubber/plastics production, engineering and construction/painting. Silica dust In a study conducted by Zheng et al, the odds of SGC among males ever exposed to silica dust were three times that observed among unexposed males. Radoï et al found 55% increased odds of SGC among ever-exposed relative to unexposed workers.17 26 These findings align with the elevated occurrence of SGC observed among workers in the construction and engineering industries, where silica is prevalent.39 40 Silica, asbestos and diesel exhaust are the most prevalent occupational carcinogens associated with trachea, bronchus and lung (TBL) cancers, which are distinct but closely related to SGCs.41 While evidence specific to salivary glands is lacking, silica dust is presumed to deposit in salivary glands in a similar manner to airways and promote an inflammatory response, subsequent cell necrosis and eventual carcinogenesis.42 Cement dust As reported in table 3, cement dust was positively associated with SGC in occupational and environmental settings, although estimates were relatively imprecise.17 22 The construction industry, where cement dust exposure is common, was ranked as having a strong LOE in this review (figure 1; online supplemental table S4).43 Like silica dust, exposure to cement dust in occupational settings may occur primarily through inhalation (eg, nose, mouth breathing), with mouth breathing bypassing the nose’s natural filtration role. Cement dust may enter the oral cavity and pose a risk for minor salivary glands. The minor salivary glands are unencapsulated and sit beneath the epithelium, which makes them more vulnerable to carcinogens and localised inflammation.44 Cement dust exposure has also been associated with other head and neck cancers (HNCs), namely laryngeal cancer.45 Chlorinated solvents Chlorinated solvents, such as trichloroethylene (TCE), are widely used in industries as effective degreasers, dry cleaning agents and chemical intermediates.46 Radoï et al reported a 51% increased rate of SGC among workers exposed to TCE for >10 years compared with unexposed workers.17 The primary route of exposure for TCE is inhalation and dermal contact, with industries such as manufacturing and plastics/rubber production having historically high exposures to TCE.46 Nickel compounds/alloys Nickel compounds are classified as an IARC group 1 carcinogen.47 As reported in table 3, elevated rates of SGC were observed among workers ever compared with never exposed to nickel compounds (OR=6.0, 95% CI 1.6 to 22.0).16 Early evidence suggests that exposure to closely related nickel compounds and alloys in the rubber industry (eg, p-nitrosodimethylaniline, N,4-dinitroso-N-methylaniline) may lead to elevated occurrence of SGC.37 Other industrial groups with documented high nickel exposure are the following: (1) commercial and industrial machinery and equipment, (2) architectural and structural metal manufacturing, (3) automotive repair and maintenance and (4) building equipment contractor and machine shops,48 which overlap with industries and occupations connected to SGC in this review: ‘Material Handling’ (strong LOE), ‘Metalwork/Welding’ (moderate LOE) and ‘Plumbing’ (limited LOE). Petroleum products and solvents, white spirits Three studies report elevated rates of SGC among workers exposed to petroleum products and solvents.17 24 26 Radoï et al reported 32% increased odds of SGC among workers ever exposed to white spirits, 47% increased odds for workers exposed to at least one type of petroleum solvent for >10 years and 53% increased odds for those ever exposed to ‘diesel, fuels, and kerosene’, all compared against never-exposed workers.17 The study authors acknowledged that statistical imprecision in their job exposure matrix (JEM)-based estimates of exposure probability could have introduced non-differential misclassification bias, which generally results in an estimation of the risk biased towards the null for dichotomised exposures and a distortion of exposure-response trends.17 In this review, many of the industry/occupation categories ranked as having a strong or moderate LOE towards a relationship with SGC involve well-documented exposure to petroleum products and/or solvents, including but not limited to ‘Rubber/Plastics Production’, ‘Food Services’, ‘Material Handling’ and ‘Transportation’.49 50 Mechanistic evidence for SGC development through solvent exposure is lacking. However, for solvents to be excreted, they require activation through metabolism. The major salivary glands have found to carry similar metabolic enzymes (eg, cytochrome P450s like CYP1A2), which may result in the major salivary glands having a high risk of intense exposure to these toxic metabolites. Additionally, the role of the salivary glands is to release saliva and flush toxins from the body.51 Accordingly, it is hypothesised that salivary cells will concentrate and release these chemicals and metabolites into the oral cavity, leading to repeated high concentration exposures. Over time, this can result in chronic DNA damage and oxidative stress, which may result in cancer. Cleaning services Cleaning professionals represent another occupation exposed to many of the chemicals mentioned above in cleaners and degreasers. In our analysis, cleaning professionals (eg, cleaners, janitors) consistently exhibited a positive association with SGC.17 24 33 Radoï et al estimated an OR of 2.47 (95% CI 1.16 to 5.22) for ‘Charworkers, cleaners and related workers’ and 3.02 (95% CI 1.38 to 6.60) for ‘Charworker’, a relatively high level of risk among all occupation groups studied.17 Rubber and plastics production This review identified multiple studies where workers in the rubber and plastics manufacturing industry,16 32 as well as those exposed to agents commonly found in this industry, such as solvents, nickel alloys and compounds, and N-nitrosamines, had an elevated occurrence of SGC (ORs ranging from 3.7 to 9.0).16 These results are in line with elevated rates of SGC reported among rubber/plastics workers in the Ontario Occupational Disease Surveillance System (pooled (n=27): HR=1.34, 95% CI 0.91 to 1.97; plastics only (n=21): HR=1.44, 95% CI 0.93 to 2.22).38 Earlier studies have pointed to the role of N-nitrosamines in SGC development.16 52 Pan et al discussed the potential association between N-nitroso compounds (NOCs) and SGC and the role of NOCs in tumour development in the parotid gland among mice, but their discussions were limited to nutritional exposures.5 While it is difficult to attribute cancer cases to a specific group of exposure agents within the industry, researchers have highlighted the presence of several known and suspected carcinogens, such as a styrene-butadiene mixture.38 Workers in facilities manufacturing styrene-butadiene rubber reported an increase in TBL cancers but evidence specific to SGCs is limited.53 Other chemicals associated with rubber manufacturing, such as N-methyl-2-pyrrolidone, may be of concern, but knowledge of their carcinogenicity is limited.54 Hairdressing and formaldehyde Multiple studies identified a positive association between hairdressing and SGC, with potential ties to traditional hair dye products containing formaldehyde and ortho-anisidine.33,35 In a recent Brazilian study, researchers analysed 23 557 cancer deaths among hairdressers between 1996 and 2020 using the Brazilian Mortality Information System.55 The SGC mortality rate for female hairdressers was modestly elevated at 1.45 (95% CI 0.51 to 4.09) compared with other service workers, although this estimate is largely imprecise. This study is noteworthy given its size and the fact that results align with the previous findings on female hairdressers experiencing an increased odds for multiple cancers compared with their male counterparts.56 Formaldehyde and ortho-anisidine are suspect chemicals in hair dyes55 57 classified by IARC as group 1 and group 2A carcinogens, respectively.58 59 Formaldehyde has a well-known biological mechanism for cancer causation, with evidence of global DNA methylation changes shown through a biological monitoring study among hairdressers.57 Literature documenting the cancer risks from occupational exposures to ortho-anisidine is limited; one study analysing 10 male workers who were exposed to rubber industry emissions found an elevated risk for bladder cancer but not SGC.60 Diethanolamine (IARC group 2B), a chemical often found in cleaning and personal care products, is also of concern, but evidence remains insufficient for SGCs.59 Given remaining uncertainty, further research is needed to confirm the effects of specific hair dye chemicals on SGC risk. Agriculture Workers in the agricultural industry are often exposed to multiple carcinogens. Our results indicate that subgroups of farmers may be at an elevated risk of SGC, such as workers with high exposure to endotoxins, like aflatoxin.28 However, this study reporting a relationship between aflatoxin and SGC did not account for potential confounding factors (eg, smoking), which may have influenced findings. Environmental exposures Burning of biofuel and exposure to biofuel emissions were identified as potential risk factors for SGC in one study.26 In addition, environmental exposure to Portland cement dust, which contains crystalline silica and hexavalent chromium, has been associated with increased rates of SGC among women, supporting trends observed in occupational studies.22 However, existing evidence is limited, as only one study has been conducted on this topic. Graham et al identified elevated rates of SGC among people living in asbestos-mining counties relative to other rural counties; however, this study had multiple methodological issues, including a lack of environmental sampling data and a lack of control for confounding by occupation.27 Environmental monitoring data are needed to elucidate the environmental exposure pathway(s) for these hazards and to quantify the extent of risk associated with environmental exposures, specifically as levels tend to be lower than in occupational settings. Strengths and weaknesses This systematic review followed a rigorous, registered protocol, including the use of (1) the HAWC programme, (2) a bias assessment protocol used by IARC, (3) the SWiM checklist and (4) effect direction plots for evidence synthesis. Our methodology provides a holistic, transparent framework for synthesising studies on rare illnesses like SGC that may have a limited number of studies lacking quantitative exposure data. The primary limitation of our review is the potential for exposure misclassification. To synthesise results across a wide range of industry and occupation categories, we grouped study results by broad industry/occupation categories based on common industry and occupation codes (eg, ISCO), where possible (online supplemental table S4). In cases where the industry/occupation code was not declared or not easily categorised, two occupational hygienists on our team reached consensus to finalise the grouping based on knowledge of shared job tasks and exposures. Therefore, our synthesis rankings could be impacted if exposure categories were commonly misclassified. Challenges and recommendations As highlighted throughout, the reviewed studies had multiple limitations, including low case numbers, a lack of occupational hygiene and environmental monitoring data and a lack of comprehensive work history information (eg, job title, task). Additional research involving larger cohorts with quantitative exposure data, as well as more harmonised cancer registries, is needed to further establish relationships between occupational and environmental exposures and often-overlooked rare cancers like SGC. A plethora of existing studies focus on HNC broadly but do not present results for subgroups. For example, a recent study on a cohort of non-metal mining reported on mortality from ‘TBL’ cancer as opposed to reporting site-specific results (eg, for SGC alone), despite having access to finer level International Classification of Diseases codes via the American National Death Index.61 While it may be possible for the strong link found between metal manufacturing and TBL cancers to translate to SGCs, site-specific results are needed to assess this association. Often, the choice to present pooled results may be due to low case numbers observed for individual cancer types. Existing literature also lacks histology data. SGC is a heterogenous group of cancers and biologically, adenoid cystic carcinoma, mucoepidermoid carcinoma, SGC, etc are all very different from each other. Thus, one would expect the occupational/environmental risk factors to vary based on the suspected histology. However, different histologies are grouped together and are not typically studied independently, diluting any histology-specific risk factor relationships that may exist. In summary, future studies should present findings for rare cancers in the main text or supplementary material to prevent under-reporting. This is especially important because ‘rare’ cancers make up a large percentage of total cancer diagnoses. The Surveillance of Rare Cancers in Europe project defines rare cancers as those with an incidence rate of <6 cases per 100 000 population per year and has reported that rare cancers amount to 22% of all cancer diagnoses in Europe.62 The health burden of rare cancers is not negligible, and these cancers should continue to be monitored in population databases such as The Cancer Genome Atlas.63 Conclusion This review illuminates emerging occupational and environmental risk factors for rare SGC, a life-altering disease. It is challenging to identify such risk factors due to limitations present in existing literature, including limited exposure data, low statistical power, under-reporting and grouping SGC cases with other HNCs. Despite these challenges, we found SGC was most consistently associated with the following industry and occupation categories: ‘cleaning services’, ‘material handling’, ‘food services’, ‘rubber/plastics production’, ‘engineering’ and ‘construction/painting’. Specific exposures implicated were silica dust, cement, chlorinated solvents, formaldehyde and white spirits. Prevention initiatives targeting the occupations and exposures highlighted by this review could protect workers and community members from SGC in the future. Supplementary material Acknowledgements We thank the Ontario workers suffering from salivary gland cancer who have shared information about their illnesses with the Occupational Health Clinics for Ontario Workers (OHCOW). Their voices have led to further investigation of this cancer type from an occupational and environmental health perspective. Footnotes Funding: This work was supported by the Ontario Ministry of Labour, Immigration, Training and Skills Development (MLITSD). Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Ethics approval: Not applicable. References Associated Data Supplementary Materials