Gender and risk‐taking behaviors influence the clinical presentation of oral squamous cell carcinoma

Abstract Objective The common risk factors for oral squamous cell carcinoma (OSCC) are smoking and alcohol abuse. A small percentage of patients, mostly women, are demonstrating oral cancer without the common risk behavior. This study investigates how gender and different patterns of lifestyle factors influence the clinical presentation of OSCC. Patients and Methods From this retrospective study, demographical and tumor‐specific data and lifestyle factors were analyzed. Statistical analyses were performed using the χ 2 test or Fisher's exact test for categorical analysis and the t test, ANOVA test, or Kruskal–Wallis test for continuous variables. The influence of the respective lifestyle factors together with their interactions with the gender on tumor characteristics has been tested using logistic and ordinal cumulative link regression models. Results Among a total of 308 patients, men represented the majority of smokers (87.2%) and the female cohort were largely non‐smokers and non‐drinkers (64.9%). For age, tumor site and N‐stage it looks like that differences of men and women are driven by the different risk behavior. But if the lifestyle factors are taken into account, we observe contrary effects between men and women for T‐, N‐, and UICC‐stage. For different cancer locations we saw opposite effects with gender and risk profile. These effects are not dose‐dependent explainable for gender. Conclusion Some but not all differences in the development of OSCC for men and women are explainable by the respective difference in lifestyle behavior. Some further investigations are necessary to find explanations for the obvious differences between men and women in developing OSCC.


| INTRODUCTION
Cancers of the oral cavity and the lips are common malignant tumors of the head and neck. These cancers accounted for approximately 2% of all cancers and cancer deaths worldwide, with an increase from 300,000 to 350,000 new cases and from 145,000 to 177,000 deaths in 2012 and 2018. (Bray et al., 2018;Ferlay et al., 2015;Torre et al., 2015) Cancers of the oral cavity had the highest frequency, with approximately 200,000 cases (Shield et al., 2017). In Germany, approximately 13,000 people per year develop cancers of the oral cavity, and more than 95% of them are squamous cell carcinomas (Wolff et al., 2012). However, oral cavity cancer can be divided into multiple tumor locations, such as floor of the mouth (FOM), tongue, alveolar rim, palate, and buccal mucosa. Oral squamous cell carcinomas (OSCC) are found more frequently in men than in women, and it is widely recognized that there is a strong association between oral cancer and smoking and alcohol consumption (Bray et al., 2018;Ferlay et al., 2015;Poeschl & Seitz, 2004;Shaw & Beasley, 2016;Shield et al., 2017;Wolff et al., 2012). In many studies, factors such as gender and risk behavior in relation to the development of OSCC were considered, but considered separately (Pires et al., 2013;Shaw & Beasley, 2016). It is well known that men make up the greater proportion of patients and also display more risk behavior. And many women show no risk behavior at diagnosis. It was also shown, that women even exhibited larger increasing changes in incidence compared with men (Chaturvedi et al., 2013;Du et al., 2020). Investigating the differences of men and women should include a more detailed analysis of the well-known factors. Thus, the interactions of the known risk factors with the sexes should be examined in order to get to the bottom of the obvious difference in development of OSCC between men and women and also to determine equality.
The aim of this study is therefore to analyze the demographic and clinical data, with a focus on gender, and lifestyle in general and in specific to the interaction effect of risk behavior and gender to get information how it influence the clinical presentation of cancers, especially in the oral cavity.

| PATIENTS AND METHODS
The investigators designed and implemented a retrospective clinicstatistical study. The influence of the risk factors together with their interactions with gender on tumor location, T-stage, N-stage, UICC-stage, and differentiation have been tested using logistic and ordinal cumulative link models (Agresti, 2013). Resulting estimates (on the log scale) are reported with their 95% confidence intervals and associated p value.
The modeled marginal effects have been predicted and visualized.
The significance level was set to α = 5% for all statistical tests. All analyses were performed with the statistic software R (Version 3.6.2) (R Core Team, 2018) using the R-packages ordinal (Version 2019.12.10) (Christensen, 2019) for the cumulative link models and ggeffects (Version 0.14.1) (Lüdecke, 2019) for the visualization of the marginal effects. The study was approved with the reference number 4434-05/15 in May 2015 by the local Ethics Committee.

| Demographic analysis
A total of 308 patients with histologically confirmed OSCC, who met the inclusion criteria, were seen within the period under review in this study. The male-to-female ratio was 3:1, with 231 men and 77 women. The patients' ages ranged from 35 to 93 years, with a median of 58 years (mean 61 ± 12 years) in general. The male mean age was 59 ± 11 years, with a range from 35 years to 89 years. The female mean age was 65 ± 13 years, nearly 6 years higher than the male mean age (p < .01). The youngest woman was 38 years old and the oldest in this study was 93 years old. All demographic and clinical data regarding gender are shown in Table 1.

| T-and N-stage
Mostly T1 (35.2%) and T2 (31.6%) tumors were seen. More than half of the patients had N0 necks (56.0%). T-stage showed no differences in gender whereas in N-stage men had more N ≥ 2 than women who had mostly N0 stage (p = .01) ( Figures S1 and S2).

| UICC and tumor differentiation
Most OSCC were presented at diagnosis in UICC status IV with 41.3%. A high proportion of males presented UICC status IV with 43.8%, whereas women showed cancers with UICC status I with 33.8%. There were no statistical differences in gender regarding UICC distribution and tumor differentiation ( Figure S3).

| Tobacco and alcohol consumption
Among all patients, 213 patients (69.2%) had a tobacco-and/or alcohol-positive history. We saw 177 men (76.6%) but only 26 women (33.8%) who were smokers (p < .01). However, within the group of smokers, there was no difference in gender in terms of the intensity measured by pack years (p = 1.00). Men reported a tobacco consumption of approximately 34 ± 14 pack years and women of 33 ± 12 pack years at diagnosis. So, men and women smoke equally intensely.
In general, 134 patients (43.5%) had a positive alcohol history.
Females were more numerous in the NSND group (64.9%) whereas the majority of male patients present at least one risk factor (80.5%). We saw that SD patients had a mean age of 56 ± 9 years whereas the NSND patients had a mean age of 67 ± 13 years, respectively (p < .01). The groups with just one risk factor show a mean age of 60 ± 11 years. Therefore, with increasing risk behavior, the age decreases for OSCC patients. In gender-separated analysis, we saw the same effect. Within the single risk groups, there were no differences regarding age between male and female patients ( Figure 1).
Patients of different risk factor groups show significantly different tumor locations (p < .01). The main tumor location for patients with no risk factors (NSND) was the tongue with 33.7%. For patients with one (SND and NSD) or even two (SD) risk factors the main location was the FOM with 46.1% and 59.7%, respectively. For smoking and non-smoking we see the same with no differences in gender ( Figure S4).
Regarding N-stage, patients with two risk factors developed more N ≥ 2 than did patients with no or one risk factor (p = .01). There was a tendency for more T1 in the NSND group but with no statistical differences in T-stage between the risk groups. SD patients presented more UICC IV stages in comparison to NSND and SND and NSD patients but without reaching clear statistical significance (p = .08).
We fitted logistic regression models to predict each of the different main locations, FOM and tongue, with gender and the risk factors. For FOM in general the effect of the female gender is negative and the effect of smoking and alcohol is positive. Especially for women who smoke, the interaction effect to get a carcinoma at FOM is positive. The opposite was seen for tongue cancers. Here the effect for women to get tongue cancer is positive but in general for smoking and alcohol negative. The interaction effect for smoking women for tongue cancer is negative. None of these effects was significant (Table S1). There is also an opposite effect evident when comparing non-smokers with smokers for FOM and tongue cancers ( Table 3).
The influence of the risk factors together with their interactions with the gender on T-stage, N-stage, UICC-stage, and differentiation were investigated. In particular, for the risk factor of smoking, we see an adverse interaction effect for the T1 and T4 stage, for N0 and N ≥ 2 stage, and for UICC I and IV stage for men and women but without reaching statistical significance. Regarding differentiation, there were no differences of effects ( Figure 2).
We saw some different effects in gender and lifestyle factors, but we further wanted to investigate the influence of the number of pack years. For the main locations FOM and tongue and for N-stage, our findings show, that for FOM the effect of the number of pack years is positive and the effect for female gender is negative, but neither effect was statistically significant. But the interaction effect of female gender on pack years is negative with statistical significance (p = .037). For patients with tongue cancer, we see the opposite.
The effect of pack years is negative and the interaction effect of female gender on pack years is positive, but without reaching statistical significance (Figure 3). We also fitted a logistic model to predict lymph node positivity (N+) with pack years and gender. We see that the effect of pack years is positive and the effect of the female gender is negative for lymph node-positive status. The interaction effect of the female gender on pack years is positive. All the mentioned effects for the association with lymph node positivity did not reach statistical significance. This was similar in male and female groups ( Figure S5 and Table S2).

| DISCUSSION
The purpose of this study was to analyze demographic and clinical data to get information on how gender and risk profile, in relation to smoking and alcohol consumption, influence the clinical presentation of OSCC. We show that men and women in the context of lifestyle demonstrate different patterns of risk. Women comprised more NSND in our study, which replicates results from different studies, which, in turn, demonstrates a higher proportion of women without smoking and alcohol consumption with oral cavity tumors (Farshadpour et al., 2007;Kruse et al., 2010;Moyses et al., 2013).
The majority of men present at least one risk factor and represent the majority of SD. Furthermore, we see differences between men and women for age, tumor location, and N-stage. In the case of T-and UICC-stage, there was a tendency seen but without reaching statistical significance.
In our analysis of the lifestyle risk factors we see the same; differences for NSND and SD in age, tumor location and N-stage and a tendency in T-stage and UICC-stage. In view of that, it seems that the differences in men and women are caused by the different risk profiles.
If we look at the same risk profiles within men and women, some Abbreviations: CI, confidence interval; FOM, floor of the mouth; NSD, non-smoker and drinker; NSND, non-smokers and non-drinkers; SD, smoker and drinker; SND, smokers and non-drinker.
F I G U R E 1 Different risk-taking behavior according to the patients age (in years) divided for males and females: With increasing risk behavior age at diagnosis decreases, but with no differences in gender within the risk groups. NSD, non-smoker and drinker; NSND, non-smokers and non-drinkers; SD, smoker and drinker; SND, smokers and non-drinker Note: We fitted a logistic model to predict the tumor sites, FOM and tongue, with different risk behaviors. Male non-smokers served as the reference group. Within this model-FOM: The effect of (m-smoker) is positive and can be considered as small and significant. The effect of (f-non-smoker] is negative. The effect of (f-smoker) is positive. Tongue: The effect of (m-smoker) is negative. The effect of (f-non-smoker) is positive. The effect of (f-smoker) is negative. See the adverse effects between the locations FOM and tongue. The model's intercept is at -0.61 for FOM and at -0.96 for the tongue.  Table S2 were predominantly affected by moderately and poorly differentiated tumors, whereas females presented mostly with moderately and well-differentiated tumors (Pires et al., 2013). This partly con-  were as expected for men and women and also confirm the results presented above.
Oral cancer is described as the third-most significant association between smoking and cancer, following lung cancer and laryngeal cancer (Poveda-Roda et al., 2010). On the flipside, despite a decline in female smoking prevalence, female incidence rates of lung, laryngeal, and oral cavity cancers increased in most parts of Europe . In the United States, an increasing trend in tongue cancer has been observed in young females, often without risk factors (Chi et al., 2015).  Table S3 WOLFER ET AL.

| 149
But different studies describe the incidence of HPV in oral tongue carcinoma and OSCC in general as low and unlikely to play a major role in the etiology (Dahlgren et al., 2004;Iyengar et al., 2014;Salem, 2010;Castellsagué et al., 2016; S3 Guideline Diagnosis and Therapy of Oral Squamous Cell Carcinoma Long Version 3.0, 2021).
Environmental tobacco smoking (ETS) is also being discussed as a risk factor for head and neck cancer, with a dose-dependent increased risk (Zhang et al., 2000). Dahlstrom et al. (2008) also report that ETS may contribute to cancer of the head and neck in NSND women. Furthermore, Koo et al. saw in ETS a possible risk factor with worse disease-specific mortality and a worse prognosis for elderly female NSND patients and discussed etiological and genetic differences between the NSND and SD groups, resulting in more locally aggressive disease or an increased likelihood of nodal and distant spread (Koo et al., 2013 be a lack of information as far as the risk history is concerned. For example, current non-smokers could be former smokers. We also did not collect information about the amount of ETS. In the retrospective setting of the study, it is difficult to differentiate from that point of view.
Furthermore, in this study, there were no investigations done for viral influences and there are no statements to HPV and p16 status because the influence is small in OSCC and is therefore not the subject of this study. However, an influence cannot be ruled out. Secondly, this study was conducted at a single institution, so external validity is limited.
Thirdly, the small number of smoking women, especially the small amount of registered female pack years limits the validity of the analysis.
Prospective studies with larger numbers of patients and a more precise recording of the risk factors are required to eliminate these limitations and investigate this further.

| CONCLUSION
Some but not all differences in the development of OSCC for men and women are explainable by the respective difference in lifestyle behavior. However, the different tumor site distribution is not explainable by lifestyle alone. There must be other reasons for that. For men, there were more consistent results, but for women, there are still some unexplainable facts. Some further investigations are necessary to find explanations for the obvious differences between men and women in developing OSCC. If these could be explained, better and more patient-specific treatment could be developed considering these differences in gender.

ACKNOWLEDGMENTS
None of the authors has a financial interest to declare in relation to the content of this article. No funding was received for this study.

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

AUTHOR CONTRIBUTIONS
Conception and design, acquisition of data, analysis and interpretation of data, drafting the article, revising the article, and final approval: Susanne Wolfer. Acquisition of data, revising the article, and final approval: Annika Kunzler, Tatjana Foos, and Cornelia Ernst. Statistical analysis and interpretation of data, revising the article, and final approval: Andreas Leha.
Revising the article and final approval: Stefan Schultze-Mosgau.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.