Objectives Inhalation of secondhand smoke (SHS) causes several diseases, including lung cancer. Tobacco smoking is a known cause of oral cancer; however, it has not been established whether SHS also causes oral cancer . The aim of this study was to evaluate the potential association between SHS exposure and the risk of oral cancer.
Methods A systematic review and meta-analysis study (following the PRISMA guidelines) was developed to examine the studies reporting on the associations of SHS and the risk of oral cancer, employing a search strategy on electronic databases (PubMed, Web of Science, Scopus, Cochrane Library, Open Grey, and ProQuest databases for dissertations) until 10 May 2020. Meta-analyses and sensitivity analyses were performed using random-effect models. The protocol was registered in PROSPERO (CRD42020189970).
Results Following the application of eligibility criteria, five studies were included, comprising a total of 1179 cases and 5798 controls, with 3452 individuals exposed and 3525 individuals not exposed to SHS. An overall OR of 1.51 (95% CI 1.2o to 1.91, p=0.0004) for oral cancer was observed, without significant heterogeneity (I2=0%, p=0.41). The duration of exposure of more than 10 or 15 years increased the risk of oral cancer (OR 2.07, 95% CI 1.54 to 2.79, p<0.00001), compared with non-exposed individuals, without significant heterogeneity (I2=0%, p=0.76).
Conclusions This systematic review and meta-analysis supports a causal association between SHS exposure and oral cancer. Our results could provide guidance to public health professionals, researchers, and policymakers to further support effective SHS exposure prevention programs worldwide.
- secondhand smoke
- global health
Data availability statement
Data are available in a public, open access repository. Data sharing not applicable as no datasets generated and/or analysed for this study. Data are available upon reasonable request. Additional information is provided in supplemental documents.
Statistics from Altmetric.com
Oral cancer is considered to be an important public health problem worldwide. Lip, oral cavity and oropharynx cancer grouped together accounted for an estimated 447 751 new cases and 228 389 deaths according to the GLOBOCAN database.1 The major risk factors for oral cancer include tobacco smoking and use of smokeless tobacco, consumption of alcohol, and betel quid chewing.2 Joint tobacco smoking and alcohol drinking result in a more than additive risk for oral cancer.3
Tobacco smoke constitutes the largest exposure of humans to chemical carcinogens and it causes one out of five cancer-related deaths in the world.4 However, it is not only active smokers who are in contact with tobacco smoke. According to data from 192 countries, 33% of male non-smokers, 35% of female non-smokers and 40% of children were exposed to involuntary smoking, by inhalation of secondhand tobacco smoke.5
Secondhand smoke (SHS) includes the ‘mainstream smoke’, that is, smoke exhaled by a smoker after inhaling cigarette smoke, and the ‘sidestream smoke‘, which is smoke released between puffs from the tip of a cigarette or other burned tobacco product.6 Secondhand tobacco smoke is also referred to as ‘environmental tobacco smoke’, ‘passive smoking’ or ‘involuntary smoking’,7 considering that non-smokers do not wish to inhale tobacco smoke.
The exposure to SHS may cause several adverse health effects. There is evidence that SHS is a cause of both non-fatal and fatal heart disease, increasing the risk of death by about 30% or more. There is also evidence that SHS causes respiratory diseases including asthma in both adults and in children. In women during pregnancy, SHS exposure can result in low birth weight and preterm delivery, and cause sudden infant death syndrome (SIDS).8–10 An evaluation by the International Agency for Cancer Research (IARC) in 2009 confirmed an earlier evaluation that there was sufficient evidence that secondhand (tobacco) smoke was carcinogenic to humans and SHS causes cancers of the lung.8 Also, a positive association was reported between exposure to secondhand tobacco smoke and cancers of the larynx and the pharynx, but not for any other cancer sites.6 There is strong evidence for tobacco smoking as a major risk factor for oral cancer.4 6 11 This suggests that the oral cavity could also be a target site for SHS. At the time of the latest IARC evaluation, for head and neck cancer only four studies12–15 were available for evaluation by the working group; however, only one study (Lee et al12) reported the effect estimates for oral cancer, taking tobacco smoking or alcohol drinking into account. The other three studies reported effect estimates for head and neck grouped together or for larynx separately, and some presented unadjusted data making them unevaluable. Since the latest IARC evaluation, several epidemiological studies have further examined the risks associated with SHS and oral cancer, but a systematic review or a meta-analysis of these recent studies have not been conducted so far. The aim of our systematic review and meta-analysis is to evaluate any potential association between SHS exposure and the risk of oral cancer.
Protocol and registration
We followed the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta‐Analysis (PRISMA) checklist for the elaboration of this systematic review and meta-analysis.16 The protocol for this study was registered prospectively in the International Prospective Register of Systematic Reviews (PROSPERO), with registration number of CRD42020189970.17
The systematic review and meta-analysis were designed to respond to the focused question: Does exposure to SHS increase the risk of oral cancer in humans? This was structured according to the PECOS format: P (participants), studies restricted to human subjects of any age, gender or place of origin; E (exposure), individuals who were exposed to SHS (environmental tobacco smoke, involuntary smoking, passive smoking), including dose–response analyses by duration and intensity of exposure; C (comparison), to actively compare with individuals not exposed to SHS; O (outcomes), primary outcome being a pathology-confirmed oral cancer; S (types of study), case–control and cohort studies without restriction to language and time period. The exclusion criteria were applied to: (1) studies analysing cancers other than oral cancer or studies on head and neck cancer without data specifically for oral cancer; (2) studies without a control group (not exposed to SHS); (3) studies not reporting effect estimates for SHS exposure and oral cancer, and not provided by the author on request; (4) literature reviews, pilot studies, letters, protocols, conference abstracts, and case reports; and (5) non-human animal models and in vitro studies.
Information sources and search strategy
The search strategy was performed using the electronic databases PubMed, Web of Science, Scopus and Cochrane Library (online supplemental table S1). Additionally, a search was carried out in the grey literature databases of the Open Grey, as well as ProQuest databases for dissertations. An extensive hand-search was also performed encompassing the bibliographies of the included papers and other narrative and systematic reviews. The initial search was performed on 9 April 2020 and was updated on 10 May 2020.
Selection of studies
Following the initial literature search, all article titles and abstracts were screened to identify relevant articles. Next, the inclusion and exclusion criteria were applied to the information given in abstracts or when any information was missing, on the full-text reading. At the final stage, all the eligible full-text articles were carefully screened, and only relevant articles were included for further analysis.
The articles (during abstract or full-text assessment) were independently reviewed by two authors (LCM and LM) to confirm each study’s eligibility. Any discrepancies between reviewers were resolved by discussion and consensus, and by consulting a third author (SW). The EndNote X9 software program (Thomson Reuters, New York, USA) was used for electronic management.
Data extraction was performed independently by two investigators (LCM and LM) and checked by a third investigator (SW). The data were drafted into the ‘Characteristics of included studies’ table of the RevMan version 5.3 software. The following information was extracted from all eligible studies and included: study design and location; number, sex and age distribution of participants (by case status); exposures to SHS (location and exposure metrics, hours/day and duration of exposure); definition of outcomes (International Classification of Diseases (ICD) codes, histology confirmation); control of potential confounders; and odds ratios (ORs) and their corresponding 95% confidence intervals (95% CIs). In addition, we tried to assess any potential overlap between studies.
Quality assessment of selected studies
For each included study, two reviewers (LCM and LM) independently assessed the quality of included studies using the Newcastle-Ottawa scale.18 This scale judges a study quality based on selection, comparability, and ascertainment of outcome through a star-based scale ranging from 0 to 9 stars. Any disagreements were resolved through discussion between the two authors and with a third author (SW). For anticipated sensitivity analyses, specificity of outcome assessment and control for potential confounders were judged as being particularly important.
Data analysis and syntheses
We used the software program Review Manager (RevMan) version 5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2012) to perform all meta-analyses. The first step involved a meta-analysis of the OR and corresponding 95% CIs, and p<0.05 was accepted as being statistically significant. We performed Q tests and I2 tests to evaluate the heterogeneity among selected studies, defining a significant heterogeneity as Cochrane Q <0.10 and/or I2 >50%. Fixed effect or random-effect models were used according to the absence or presence of heterogeneity and/or methodological diversity among the included studies following the Cochrane Collaboration guidelines.19 For studies that reported both crude and adjusted OR estimates, the adjusted risk estimate was selected for the meta-analysis. Additional meta-analyses were performed by categories of duration or intensity of exposure. Sensitivity analyses were carried out by dropping studies to address issues such as potential overlap between study populations, specificity of outcome assessment, and residual confounding. Publication bias was assessed graphically using a Begg’s funnel plot. Begg and Mazumdar rank correlation and Egger’s regression interception methods were used to statistically test the presence of publication bias (significant publication bias was considered positive when p<0.05).
The search strategy initially identified 480 publications. From these, 31 articles were removed because they were duplicates, and another 413 articles were excluded after evaluating their titles and abstracts. Subsequently, 36 full-text articles were assessed for eligibility, of which 31 did not have sufficient information regarding the topic of interest (online supplemental table S2). The process of identification and eligibility of studies is shown in a flowchart in figure 1. Finally, five studies (Lee et al,12 Hashibe et al,20 He et al,21 Lee et al,22 Yan et al23) were selected for inclusion in the meta-analysis.
Eligible studies and rationale for their selection
A total of five studies were initially included,12 20–23 all of them case–control studies. Among them, three were conducted in Asia,20 21 23 one in Europe22 and one was a multicentre study in North America, Latin America and Europe.12 The period covered by the included studies ranged from 2008 to 2019. The main characteristics of the eligible studies are listed in table 1 (and online supplemental table S3). The definitions of SHS exposure used in each study are given in table 2.
The eligible studies included 1179 cases and 5798 controls, with exposure to SHS in 3452 individuals; 3525 individuals had not been exposed to SHS. Overall, four studies included both men and women12 20 22 23 and one study included only women.21 The age of the population of interest in the included studies varied as shown in table 1 (and online supplemental table S3). Four of the studies were on oral cancer12 20 21 23 and one included both oral and oropharyngeal cancers for analyses by duration of exposure to SHS.22
We identified the likelihood of some overlap of the study populations between three of the initially included articles: Hashibe et al,20 He et al,21 and Yan et al.23 (online supplemental table S3). The similarities between the reports of He et al21 and Yan et al,23 both performed in the region of Fujian, China, were high, with the latter probably reporting on an extended study population. When the necessary data were available, we preferred Yan et al23 over He et al21 because of the higher precision of their effect estimates. However, because Yan et al23 did not report results from dose-response analyses, we had to revert to He et al.21 Further, Hashibe et al20 reported on study populations from several geographic locations, also including a study population from the Fujian region where the two other studies had been conducted, therefore some overlap between the Fujian study populations is likely. We performed sensitivity analyses to address an overlap by dropping one of the two studies according to the validity and precision of the studies. We present the summary results showing the effects of ever versus never exposure to SHS, preferring results for oral (vs oral and oropharyngeal) cancer and results for subjects who did not report tobacco smoking and alcohol consumption or with an OR adjusted for alcohol consumption as indicated in table 1 (or online supplemental table S3) and table 2). For analyses by duration and intensity of exposure to SHS some of the original studies distinguished between exposure at work or at home. We aimed to include the effect estimate for the most comprehensive exposure metric and the most precise effect estimate (see table 1 for details).
Quality assessment of the studies
Quality assessment of included studies revealed an overall mean of 5 star-points on the Newcastle-Ottawa scale, with the same grading for all studies except for Lee et al22 with 4 star-points (online supplemental table S4). Most of the included articles presented data excluding potential confounders such as tobacco smoking and alcohol drinking (Lee et al,12 Yan et al,23 Hashibe et al20). Lee et al22 presented cases without exposure to tobacco smoking but included subjects drinking alcohol, although with an adjusted OR for alcohol drinking. He et al21 included tobacco smokers and alcohol drinkers in their samples but presented an OR adjusted for these two variables. All included articles presented cases of oral cancers as outcome; however, in the study by Lee et al,22 data on duration of exposure also included oropharynx cases.
SHS exposure and the risk of oral cancer
The meta-analysis of the four included studies using the random-effect model showed a strong positive association with an OR of 1.51 (95% CI 1.20 to 1.91, p=0.0004) for oral cancer risk associated with SHS exposure. The magnitude of heterogeneity was not significant across studies (χ2 test 2.87, I2=0%, p=0.41) (figure 2). Among these four primary studies, Lee et al22 with an OR of 2.45 and a relatively broad CI (95% CI 1.2 to 5) had the lowest weight (10.5%), while Yan et al23 had the highest weight (50.6%) (figure 2).
To address certain issues of validity of the original studies and possible overlap between study populations, sensitivity analyses were undertaken for exposure versus non-exposure to SHS. First, we explored the impact of a possible overlap by dropping the publication by Hashibe et al.20 Compared with the overall meta-OR, the OR changed only very marginally and was still significant (OR 1.49, 95% CI 1.08 to 2.06, p=0.02) with a value for I2 of 25% (p=0.26). In the sensitivity analysis dropping the results by Lee et al22 (that reported OR adjusted alcohol consumption), the pooled OR for SHS and oral cancer in never smokers and non-alcohol users was 1.43 (95% CI 1.12 to 1.83, p=0.004) (figure 2), and heterogeneity was not significant observed across the studies (I2=0%, p=0.63).
The meta-analyses by duration and intensity of exposure include a different set of studies due to availability of reported effect estimates. Therefore, as a point of reference, we performed another meta-analysis for ever versus never exposure to SHS, including the study of He et al21 instead of the study of Yan et al.23 The meta-OR was slightly higher than the primary ever versus never meta-analysis (see online supplemental figure S1).
For the meta-analysis by duration of SHS exposure, we observed an increased risk for oral cancer when exposed to SHS for more than 10 or 15 years compared to non-exposed (OR 2.07, 95% CI 1.54 to 2.79, p<0.00001, heterogeneity of I2=0%, χ2 test 1.17, p=0.76) (figure 3). The cases exposed to less than 10 or 15 years presented an OR of 1.56 (95% CI 1 to 2.43, p=0.05, heterogeneity of I2=0%, χ2 test 2.77, p=0.43) compared with non-exposed cases.
Sensitivity analyses were carried out to assess the impact of possibly overlapping studies (He et al21 and Hashibe et al20), different cut-points of exposure categories, adjustment for potential confounders (instead of restriction to never users) and less specific outcomes. By dropping He et al,21 who reported SHS exposure among cases for more than 10 years and adjusted for potential confounders, the meta-OR was 1.93 (95% CI 1.35 to 2.77, p=0.0003) without heterogeneity (I2=0%, p=0.7) compared with non-exposed cases. By dropping Lee et al22 (who included oropharynx cancer cases for their analyses by duration of exposure) comparing cases with SHS exposure of more than 10 or 15 years with non-exposed cases we obtained an OR of 2.05 (95% CI 1.45 to 2.89, p<0.0001) without heterogeneity (I2=0%, p=0.56).
We repeated the sensitivity analyses for cases exposed to less than 10 or 15 years. Dropping He et al21 resulted in an OR of 1.39 (95% CI 0.84 to 2.31) comparing to non-exposed individuals, and by dropping Lee et al22 we observed an OR of 1.67 (95% CI 0.87 to 3.21, p=0.12, heterogeneity I2=23%, χ2 test. 2.59, p=0.27) (figure 3).
Finally, we performed meta-analyses by intensity of SHS exposure (figure 4). We observed an OR of 2.15 (95% CI 1.22 to 3.77, p=0.008) for cases exposed to more than 2 or 3 hours compared with non-exposed cases with a magnitude of heterogeneity of I2=33% (χ2 test 2.99, p=0.22). Cases with an exposure less than 2 or 3 hours yielded an OR of 1.65 (95% CI 1.11 to 2.45, p=0.01) compared with non-exposed cases and without significant heterogeneity (I2 of 0%, p=0.53). A sensitivity analysis for overlapping, adjustment for potential confounders and cut-point values was conducted by dropping the study by He et al,21 observing an OR of 1.69 (95% CI 0.91 to 3.13, p=0.1) for cases exposed to more than 2 or 3 hours compared with non-exposed cases and an OR of 1.41 (95% CI 0.82 to 2.42, p=0.21) for cases with an exposure less than 2 or 3 hours; both analyses had no significant heterogeneity (I2=8%, χ2 test 1.09, p=0.3 and I2=0%, χ2 test 0.59, p=0.44, respectively).
Publication bias was visually assessed by Begg’s funnel plot (online supplemental figure S2). Begg’s test (z=0.67, p=0.248) and Egger’s test (t=0.836, p=0.246) did not reveal significant publication bias.
Our overall aim was to evaluate whether exposure to SHS does increase the risk of oral cancer in humans. On the basis of the included studies we observed a statistically significant association of SHS exposure with the risk of oral cancer with an OR of 1.51 (95% CI 1.2 to 1.91, p=0.0004). This association was corroborated by relationships with the duration and intensity of exposure where individuals with more time of exposure to SHS presented higher risks for oral cancer. Our results were very similar when sensitivity tests were performed addressing potential confounder variables. Overall, we demonstrated good homogeneity across the included population-based studies, a very consistent finding despite wide diversity of underlying study populations, including different ethnicities and different smoking and drinking habits, gender differences, and SHS policies in WHO member countries. For these reasons, we chose to use the random-effects model even when there was no significant heterogeneity.6
We identified possible overlapping of data between some cases and controls recruited from the region of Fujian, China.20 21 23 Such potential overlap was more evident in the studies of He et al21 and Yan et al23 and for these reasons we did not use these studies together in the same meta-analysis. In the Hashibe et al20 study, no more than 32 cases or 27 controls (from 243 head and neck cancers and 403 control never smokers/never alcohol drinkers) presented possible overlap with the studies of He et al21 and Yan et al.23 As a precaution, we performed sensitivity analyses to assess the possible effect of this overlap by removing the article by Hashibe et al20 during the meta-analysis. The results were consistent with our previous analysis showing a statistically significant association of SHS exposure with the risk of oral cancer and with minimal and non-significant heterogeneity.
All included publications reported a well-defined outcome restricted to oral cancer with a defined location, mainly the oral cavity (though some included lip mucosa) and mainly corresponding to squamous cell carcinomas with histology confirmation. In the study by Lee et al,22 some cases of oropharyngeal cancers were included when evaluating the duration of SHS exposure. Similarly, the study by Lee et al12 included various head and neck cancers for the analysis by intensity of SHS exposure. Potential confounding by human papillomavirus virus (HPV) infection, an important etiological factor for oropharyngeal cancers24 25 for the subset of oropharyngeal cancer, has not been adjusted for in these two subgroup analyses. But, as oropharyngeal cancer cases included in the primary studies were few, any potential bias caused by HPV is expected to be minimal. Further, using sensitivity analysis, by dropping the study with some oropharynx cases (Lee et al22), the results were not affected and a statistically significant association of duration of SHS exposure with the risk for the development of oral cancer was observed. Indeed, using this sensitivity analysis, individuals with exposure to 10 or 15 more years revealed an OR of 2.05 (95% CI 1.45 to 2.85, p<0.0001) compared with those not exposed, confirming our results reported for the overall sample.
Most of the included articles adjusted for potential confounding by differential tobacco smoking and alcohol drinking habits in the published results restricted to never smokers and never alcohol drinkers, or as in the study by Lee et al,22 adjusted for alcohol habits among non-smokers. As adjustment could still result in residual confounding, we also performed a sensitivity analysis dropping this study, and still observed a statistically significant association between SHS exposure and oral cancer. Moreover, in a recent meta-analysis, Mello et al3 further corroborated a synergistic (more than additive) effect of alcohol and tobacco consumption on oral cancer risk which could result in model misspecification and residual confounding in studies simply adjusting for tobacco smoking and alcohol drinking habits. Restriction to never smokers and non-alcohol drinkers, or adjustment for potential confounding further corroborated by sensitivity analyses, helps to mitigate potential confounding in our meta-analysis.
We acknowledge as a limitation of this study that the number of the overall available studies with data related to SHS and oral cancer was small. However, several of the original studies had already pooled many individual studies and therefore the overall number of cases and controls for our meta-analyses was high. This may have contributed to the non-significant and minimal heterogeneity observed among all meta-analyses performed. Assessment of exposure to SHS differed across studies and some studies lacked details on the assessment of exposure to SHS (see table 2). None of the studies is of very high quality, as assessed by the Newcastle-Ottawa scale; particularly, recall bias is a concern in these case–control studies. It seems possible that being diagnosed with cancer may affect recall of SHS exposure. Some of the studies used hospital controls, which may introduce selection bias. Finally, we were unable to stratify our meta-analyses by gender, as only one study reported gender-specific results.20 Although no significant publication bias was noted, the low number of studies limits the power of this test. Acknowledging these limitations, to the best of our knowledge this is the first systematic review and meta-analysis of SHS exposure and the risk of oral cancer. Potential future studies may further strengthen the evidence base by overcoming some of the noted limitations.
In conclusion, our systematic review and meta-analysis support a consistent and statistically significant association between SHS exposure and the risk of oral cancer. Moreover, the analyses of exposure response, including by duration of exposure (more than 10 or 15 years) to SHS, further supports causal inference. The identification of the harmful effects of SHS exposure provides guidance to public health professionals, researchers, and policymakers as they develop and deliver effective SHS exposure prevention programmes and adopt appropriate measures to implement guidelines in Article 8 of the WHO’s Framework Convention on Tobacco Control.
What this paper adds
This systematic review and meta-analysis, synthesising and evaluating the evidence on SHS exposure and the risk of oral cancer, show a consistent and statistically significant association between secondhand smoke (SHS) exposure and the risk of oral cancer, with causal inference further strengthened by positive exposure response relationships. Our data and the conclusion add to the last International Agency for Cancer Research evaluation that SHS is carcinogenic to humans.
Data availability statement
Data are available in a public, open access repository. Data sharing not applicable as no datasets generated and/or analysed for this study. Data are available upon reasonable request. Additional information is provided in supplemental documents.
Correction notice This paper has been updated to correct author details.
Contributors LM, SW, and LSM conceived and designed the study. The search, selection, and quality of studies evaluation was performed by LM, SW, and LSM. Analysis and interpretation of the data was performed by all authors. All authors contributed to the manuscript development. All authors approved the final version of the manuscript for submission.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.