Introduction
Most salivary gland tumours are benign, with malignant pathology representing a small fraction of overall head and neck cancers.Reference Roland, McRae and McCombe1 The incidence worldwide of malignant salivary gland cancers is estimated at 0.5 to 3 per 100 000 per year.Reference Roland, McRae and McCombe1 These relatively small numbers are likely to contribute to the current limited evidence regarding both geographical distribution of different salivary gland malignancies as well as the understanding of any possible risk factors.
There is limited high quality literature regarding the potential lifestyle, occupational, geographical and exposure risk factors for salivary gland cancers; however, the current evidence is discussed below. One small French epidemiological study found no significant association between salivary gland cancers and smoking or alcohol but did find an association with previous radiotherapy and certain manual occupations.Reference Radoï, Barul, Menvielle, Carton, Matrat and Sanchez2 The finding of an association between previous radiotherapy and development of salivary gland cancers has been echoed in further large population-based studies of lymphoma patients,Reference Boukheris, Ron, Dores, Boukheris, Ron and Dores3,Reference Chowdhry, McHugh, Fung, Dhakal, Constine and Milano4 and a large American case-control study also found a link with salivary gland cancers and outdoor occupations (although this was only in certain ethnic groups).Reference Wilson, Moore and Dosemeci5 Multiple further case-control studies have attempted to prove links with salivary gland cancers and common risk factors such as smoking,Reference Pan, de Groh and Morrison6–Reference Muscat and Wynder8 alcohol,Reference Pan, de Groh and Morrison6,Reference Muscat and Wynder8 obesityReference Pan, de Groh and Morrison6,Reference Muscat and Wynder8 and radiation exposure;Reference Pan, de Groh and Morrison6,Reference Muscat and Wynder8 however, although positive correlations were found, the results were not statistically significant.
Radon, an odourless, colourless radioactive gas forms from the decay of uranium emitted by natural soil and rocks. Because of its radioactivity, radon has some potential harmful consequences on the human body. It has been proven to be a risk factor for the development of lung cancer.Reference Krewski, Lubin, Zielinski, Alavanja, Catalan and Field9,Reference Schoenberg, Klotz, Wilcox, Nicholls, Gil-del-Real and Stemhagen10 However, several papers have attempted to identify a significant correlation with increased background radon levels and a variety of extrapulmonary cancersReference Turner, Krewski, Chen, Pope, Gapstur and Thun11–Reference Darby, Radford and Whitley14 as well as salivary gland cancers specificallyReference Miller, Harwick, Alfaro-Miranda and Sundararajan15 and, although some found positive correlations, there were no statistically significant results linking radon exposure to the development of any extrapulmonary cancers, including salivary gland cancers. In contrast to these findings, one large German cohort studyReference Kreuzer, Walsh, Schnelzer, Tschense and Grosche16 did find increased incidence of a variety of extrapulmonary cancers in miners exposed to radon but a statistically lower incidence of head and neck cancers (including salivary gland cancers). The authors of this study did however admit that several confounding factors may have influenced these results.
In summary, although there do appear to be correlations with certain risk factors (previous radiotherapy) and the development of salivary gland malignancy, no studies to date have firmly identified any statistically significant contributing environmental or lifestyle factors. There is therefore a clear epidemiological need to investigate this further.
Aim and objectives
The aim of this study was to identify the risk factors which may contribute to the development of salivary gland malignancy. This was to be achieved by addressing the following objectives: (1) identify the incidence of common subtypes of salivary gland malignancies across a variety of geographical areas; (2) identify populational rates of common risk behaviours including, smoking, mean activity levels, obesity levels and mean background radon levels across the same geographical areas; and (3) undertake a correlational analysis to establish or refute a link between a variety of risk factors and incidence of different types of salivary gland malignancy.
Materials and methods
A search of two national databases, Patient Episode Database Wales and Cancer Network Information System Cymru, using 14 possible diagnostic codes was undertaken to identify all patients with a recorded episode of submandibular or parotid malignancy in Wales from 1992 to 2019. The search was undertaken in conjunction with and the approval of NHS Wales Informatics Services Information Governance Department. Histological diagnosis and partial postcode data were recorded for each patient. The search of the two databases initially yielded 862 recorded episodes. There were 181 cases overlapping both databases, and these were excluded to leave 681 recorded episodes.
Histological diagnoses of squamous cell carcinoma (n = 151) were excluded, as were other non-primary parotid malignancies (lymphoma, melanoma, sarcoma; n = 37). Non-specified malignancies (data coded as malignancy without histological subtype; n = 137) were also excluded to leave a total of 356 patients with a confirmed primary salivary gland malignancy. The cases were then stratified into groups based on the following histological subtypes: acinic cell carcinoma, adenocarcinoma, adenoid cystic carcinoma, carcinoma ex-pleomorphic, epithelial-myoepithelial carcinoma or mucoepidermoid carcinoma (Figure 1).
Fig. 1. Flowchart of study population selection.
Wales is divided into 22 local governmental authorities. The total numbers of each subtype of salivary gland cancers were recorded for each local authority using partial postcode data. Population numbers for each local governmental authority were retrieved from Office for National Statistics data17 and incidence was calculated per million population over the entire study period.
Public Health England data from 2016Reference Miller and Rees18,19 was used to establish mean smoking rates, obesity and mean activity levels as well as mean household radon levels for each of the 22 local governmental authorities.
A multivariate analysis of variance study was undertaken comparing the risk factors in each local governmental authority with incidence of each subtype of salivary gland cancers in that area.
Results
The total numbers of all six salivary gland cancer subtypes across all 22 local governmental authorities for the entire study period can be found in Table 1 and the incidence per million over the 27-year period that the study looked at is shown in Table 2. A confidence interval was calculated for each incidence and plotted on a box and whisker plot (Figure 2). The incidence (per million) of each salivary gland cancer subtype for each local governmental authority over the study period is presented in Figure 3, and the incidence of all salivary gland cancers across the same regions can be found in Figure 4.The results of the multivariate analysis of variance analysis are displayed in Table 3.
Fig. 2. Box and whisker plot of salivary gland cancers subtype incidence.
Fig. 3. Bar graph of incidence of salivary gland cancers subtypes for each local governmental authority. RCT = Rhondda Cynon Taff
Fig. 4. Bar graph of incidence of salivary gland cancers subtypes in Wales. RCT = Rhondda Cynon Taff
Table 1. Total cases of salivary gland cancers by subtype for all 22 local governmental authorities in Wales for the study period
Table 2. Incidence of salivary gland cancers subtype
Table 3. Results of multivariate analysis of variance study
A significance level of less than 0.05 was used.
This analysis has identified a significantly higher incidence of mucoepidermoid salivary gland cancers in areas with higher mean radon levels (p = 0.006), a significantly higher incidence of epithelial-myoepithelial carcinoma in areas with higher smoking levels (p = 0.029) and a significantly higher incidence of adenoid cystic carcinoma in areas with increased obesity (p = 0.028). There was no correlation found between any of the salivary gland cancer subtypes and mean activity levels.
Discussion
Significance of results
To our knowledge, there are no previous studies that have identified significant links between salivary gland carcinoma and the lifestyle risk factors of smoking and obesity, nor the environmental risk factor of background radon level. One reason for this may be that, unlike our methods, most previous studies have attempted to identify links with the development of salivary gland malignancy as a whole,Reference Radoï, Barul, Menvielle, Carton, Matrat and Sanchez2,Reference Chowdhry, McHugh, Fung, Dhakal, Constine and Milano4,Reference Pan, de Groh and Morrison6,Reference Muscat and Wynder8 rather than with individual histological subtypes.
Undertaking a multivariate analysis of variance analysis for each histological subtype against each individual potential risk factor has allowed us to identify that the incidence of mucoepidermoid salivary gland cancers is significantly higher in areas of increased background radon, incidence of epithelial-myoepithelial carcinoma is significantly higher in populations with increased smoking levels and incidence of adenoid cystic carcinoma is higher in obese populations.
Study design
The size of the study population was limited by the information available in the databases that were used. We could not establish a source of reliably populated histological data prior to 1992. It was therefore decided that this should dictate the earliest source of data for this study. The authors considered increasing the sample size to the whole of the UK. However, the potential benefits of an increased study population were likely to be outweighed by logistical difficulties in obtaining data, heterogeneity of histology reporting and differences in recording techniques for each risk factor that was investigated.
Non-primary salivary gland malignancies such as squamous cell carcinoma and lymphoma were excluded from this study as they already have well established risk factors, and the numbers were relatively small. It was therefore deemed that investigating these pathologies was unlikely to yield significant results or meaningfully add to the current literature base.
With regard to the group of ‘non-specified malignancy’, the authors acknowledge that with a number of 137 it could be interpreted that if the subtypes of this group were known, it may have the potential to influence the results. We argue that this is unlikely as the patients from this group were distributed evenly across the 22 local governmental authorities with no significant cluster in one particular area. This group is likely to represent an evenly distributed sample of malignancies that were either difficult to interpret histologically or were recorded with insufficient information.
A clear pitfall of this study is that we have used populational data to establish lifestyle risk factor levels. Interviewing individual cases regarding their smoking and activity levels as well as establishing a body mass index for each case would have been a more accurate method of collecting data. However, given the 27-year time span of the study, it was not possible to establish this information accurately for individual cases. Rather than only interview more recent cases and therefore decrease the sample size, it was deemed that a larger study population over a longer timeframe would be more valuable. Furthermore, the environmental risk factor of background radon levels is best measured using populational data.
We acknowledge that populations change over time. The incidence per million of each subtype of salivary gland cancers as well as smoking, obesity and radon levels have changed over the study period. The multivariate analysis of variance analysis that we performed accounts for these changes and matches data accordingly.
Interpretation of results
To our knowledge, this is the first study to establish a significant link between increased background radon levels, smoking, obesity and certain types of salivary gland carcinoma.
It is possible that the same cellular processes that take place in the development of salivary gland cancers after iatrogenic radiationReference Boukheris, Ron, Dores, Boukheris, Ron and Dores3,Reference Chowdhry, McHugh, Fung, Dhakal, Constine and Milano4 are responsible for the increased incidence of mucoepidermoid carcinoma in areas of increased background radon.
Epithelial-myoepithelial cancers are rare and make up less than 1 per cent of salivary gland tumours;Reference Krane, Farquin, Cibas and Ducatman20 there is therefore a paucity of data regarding risk factors. Within salivary glands, they tend to originate from ductal cells,Reference Krane, Farquin, Cibas and Ducatman20 and the impact of smoking on the function of the salivary ducts at a cellular level may play a part.
• A 27-year population-based study was conducted to identify possible environmental and lifestyle risk factors for salivary gland carcinoma
• A link between background radon levels and mucoepidermoid cancer, smoking and epithelial-myoepithelial cancer, and obesity and adenoid cystic cancer was identified
The link between obesity and a variety of cancers including endometrial,Reference Setiawan, Yang, Pike, McCann, Yu and Xiang21 oesophagealReference Hoyo, Cook, Kamangar, Freedman, Whiteman and Bernstein22 and gastricReference Chen, Liu, Wang, Wang, Yan and Cheng23 adenocarcinomas is well established. Although adenoid cystic cancers are histologically distinct to adenocarcinomas in the salivary glands, similar pathological processes may contribute to the development of these types of cancers.
Following our identification of significant risk factors for different salivary gland malignancies, more work is now needed to investigate these processes at a cellular level.
Conclusion
Background radiation exposure, smoking and obesity are known risk factors for a variety of cancers. Until now, however, there has been no convincing evidence to prove a link between these risk factors and salivary gland cancer. Our study has identified a significant connection between background radiation and mucoepidermoid cancer, smoking and epithelial-myoepithelial cancer and obesity and adenoid cystic cancer of the salivary glands.
Competing interests
None declared
