Impact on patients with oral squamous cell carcinoma in different anatomical subsites: a single-center study in Taiwan

The incidence of oral cavity squamous cell carcinoma (OSCC) is particularly high in South Asia. According to the National Comprehensive Cancer Network, OSCC can arise in several subsites. We investigated survival rates and the clinical and pathological characteristics of OSCC in different anatomical subsites in the Taiwanese population. We retrospectively analyzed data for 3010 patients with OSCC treated at the Changhua Christian Hospital. Subsequently, we compared clinical and pathological features of OSCC in different subsites. Pathological T4 stage OSCCs occurred in the alveolar ridge and retromolar trigone in 56.4% and 43.7% of cases, respectively. More than 25% of patients with tongue OSCC and 23.4% of those with retromolar OSCC had lymph node metastasis. The prognosis was worst for hard palate OSCC (hazard ratio 1.848; p < 0.001) and alveolar ridge OSCC (hazard ratio 1.220; p = 0.017). Retromolar OSCC recurred most often and tongue OSCC second most often. The risk for cancer-related mortality was highest for hard palate OSCC, followed by alveolar ridge and retromolar OSCC. We found distinct differences in survival among the different subsites of OSCC. Our findings may also help prompt future investigations of OSCC in different subsites in Taiwanese patients.

Comparison of the patients according to anatomical subsites revealed that those whose primary tumor was at the FOM appeared to have greater association with alcohol consumption. Pathological T4 stage OSCCs occurred in the alveolar ridge and retromolar trigone in 56.4% and 43.7% of the patients, respectively. Further, > 25% of the patients with tongue OSCC and 23.4% of those with retromolar OSCC had pathological lymph node metastases. Moreover, lymph node skip metastases were highly likely to occur with tongue OSCC. More details of these differences are presented in Table 1.
The relationships between clinicopathological characteristics and recurrence or death are presented in Table 2. Age at diagnosis was significantly associated with cancer-related death (p < 0.001), and the patients aged > 71 years had the highest mortality (53.8%). Regarding pathological stages, those with advanced T and N stages presented with worse prognosis (both p < 0.001 in terms of recurrence and death). In terms of pathological features, ECS (both p < 0.001), level IV or V positive lymph nodes (p = 0.009 and p < 0.001, respectively), and pathological grade (both p < 0.001) were significantly associated with worse prognosis.
When the patients were compared according to anatomical subsite, OSCC arising from the retromolar trigone appeared to be more related to recurrence, and OSCC arising from the hard palate appeared to be more closely associated with cancer-related death (p = 0.044 and p < 0.001, respectively).
The results of the univariate and multivariate Cox regression analyses for disease-free survival (Table 3) revealed that the patients with OSCCs in the hard palate and alveolar ridge had the poorest disease-free survival outcomes, with the hazard ratios of 1.848 (p < 0.001) and 1.22 (p = 0.017), respectively. Figure 1 shows the cumulative rates of OSCC recurrence according to the different subsites. OSCC that arose from the retromolar trigone, tongue, and alveolar ridge had the highest rates of recurrence, and those of the lip mucosa and body of the lip had the lowest rates of recurrence (p = 0.042). Figure 2 shows the cumulative rates of cancer-related death among patients according to different subsites of OSCC. OSCC that arose from the hard palate, alveolar ridge, and retromolar trigone caused the highest rates of death, and those of the body of the lip, buccal mucosa, and lip mucosa caused the lowest rates of death (p < 0.001).

Discussion
From the perspective of mortality, Farhood et al. reviewed data for 20,647 patients from the Surveillance, Epidemiology, and End Results Program (SEER) 9 database and reported that OSCC was most commonly diagnosed in the FOM and the tongue 19 . Furthermore, they found that tongue OSCC was associated with more cause-specific mortality than were OSCCs at other subsites 19 . Different cultures and dietary habits could account for the difference in the predilection site between Taiwan and other places 22,23 , and this could be the major reason for the difference between their results and ours. However, rates of survival among patients with OSCC in different subsites did differ, and NCCN treatment guidelines do not take this point into account. Su et al. studied a large Taiwanese population and reported that the 5-year rate of survival was poorest among patients with hard palate OSCC, followed by those with gingival and FOM OSCCs, and gingival and hard palate OSCCs were most likely to be diagnosed at an advanced stage 18 . They pointed out that OSCC at different subsites necessitated specific surveillance strategies and tailored treatment. Our results were consistent with theirs; however, we subclassified gingival OSCC as retromolar trigone and buccal mucosa OSCC according to the NCCN treatment guideline and found that survival and recurrence rates of the two did indeed differ. Second, our data obtained were from a single center, and all our patients underwent surgery and adjuvant therapy, performed by a single team of head and cancer specialists. Finally, apart from overall survival, we also found different recurrence rates among OSCCs at different subsites.
In our study, the risk of lymph node skip metastases was greatest with tongue OSCCs (3.6%), followed by retromolar trigone and FOM OSCCs. According to several reports, supra-omohyoid neck dissection was not sufficient for OSCC, and neck dissection at level IV should be routine to prevent neck metastasis 24,25 . However, other investigators have reported conflicting opinions 26 . Warshavsky et al. conducted a meta-analysis about the rationality of prophylactic level IV neck dissection for OSCC and reported that the rate of skip metastasis ranged from 0 to 5.50% with a fixed-effects model of 0.50% (95% confidence interval 0.09-1.11%), and staging and subsites of OSCC did not notably affect the rate of skip metastasis 27 . Weiss Shabtay and Ronen also conducted a meta-analysis about prophylactic level IV neck dissection for tongue OSCC and reported a 2.8% rate of skip metastasis 28 . In our study, the rate of lymph node skip metastasis in tongue cancer was 3.6%, which  www.nature.com/scientificreports/  www.nature.com/scientificreports/ 665 patients with buccal mucosa OSCC and reported that the former had more recurrence factors, including perineural invasion, lymphovascular emboli, and poorer pathological grade 17 . In our study, retromolar trigone OSCC was most likely to recur, followed by tongue, alveolar ridge, FOM. Compared with the previous studies, our study was conducted with a larger sample size and included all American Joint Committee on Cancer (AJCC) anatomic subsites in the analysis. Hard palate and alveolar ridge OSCCs account for a small percentage of OSCCs in comparison with OSCCs in other subsites [29][30][31] . We found that hard palate and alveolar ridge OSCCs carried a higher risk of mortality than OSCCs at the other subsites. This finding has several explanations: OSCCs adjacent to the mandibular or maxillary bone could be more likely to be at an advanced T stage when diagnosed, and > 40% of both hard palate and alveolar ridge OSCCs occurred in patients aged > 60 years; age was an independent factor of disease-free survival in our study. Several investigators have reported that different strategies should be tailored to OSCCs at various subsites [17][18][19]32 , consistent with our findings.
From a biological perspective, different subsites of OSCC may have their own biomarkers, demonstrating discrepancies in survival and prognosis. Fu et al. reported that if patients with tongue and lip OSCC have high DEXD/H box helicase 60 (DDX60) expression, the outcome will be poor 33 . They concluded that DDX60 is a novel but subsite-specific biomarker for OSCC. Boldrup et al. also reported that patients with tongue OSCC demonstrate significant deregulation of miR-21, miR-125b, and miR-203, but those with gingival OSCC demonstrate significant downregulation of only miR-125b 34 . In the future, biomarkers for different subsites of OSCC should be analyzed to further explain the differences in survival among patients with OSCCs at all subsites.
Our study had several limitations. First, the retrospective study design may have contributed to bias. Second, the data for our study were collected from a single medical center in Taiwan. Oral cavity cancer is strongly associated with betel nut chewing, which is popular in Taiwan; therefore, our results may differ from findings in other geographical regions 5,35 . Finally, in our result, the tissue adjacent to the metastatic patterns of the tumor and lymph nodes cannot fully explain why the patients with hard palate OSCC had the worst disease-free survival and why alveolar ridge OSCC was most likely to recur. Further investigations may focus on discrepancies among biomarkers for OSCCs at different subsites.
In conclusion, survival distinctly differed among patients with OSCCs at different subsites, although the NCCN treatment guideline did not account for subsites of OSCC. Our results not only were consistent with the findings of previous studies but also may encourage future investigations on Taiwanese patients with OSCC at different subsites.

Material and methods
Patients. This retrospective cohort study was approved by the institutional review board and ethics committee of Changhua Christian Hospital, Changhua, Taiwan (IRB No. 210210). We obtained all clinical data through a chart review and the cancer registry center of Changhua Christian Hospital. We confirmed that all the methods were performed in accordance with relevant guidelines and regulations. Informed consent was waived owing to the retrospective nature of the study, and the analysis used anonymized clinical data with the approval of the IRB of Changhua Christian Hospital, Changhua, Taiwan (IRB No. 210210). We identified 3620 patients with OSCC who underwent surgery, adjuvant therapy, and follow-up at our center between January 1, 2008, and December 31, 2018. The follow-up duration was from the initial date of diagnosis to December 31, 2019. We excluded patients who did not receive treatment in accordance with the NCCN cancer treatment guidelines, whose initial diagnosis at our hospital was recurrence or distant metastasis, and who did not undergo surgery at our hospital. In total, 3010 patients were enrolled in our study and assigned into subgroups according to the following pathological anatomical sites designated by the AJCC: the buccal mucosa, alveolar ridge, anterior tongue, hard palate, FOM, retromolar trigone, mucosa of the lip, and body of the lip. Treatment protocols. The patients enrolled in our study underwent wide tumor excision and neck dissection according to clinical tumor stage. The patients with clinical stage N0 tumors underwent selective neck dissection, and those with N-positive tumors underwent radical neck dissection. Adjuvant therapy was performed in individual cases by our interdisciplinary head and neck surgery team, which included surgeons, oncology radiologists, a medical oncologist, and a pathologist. In general, postoperative radiotherapy was administered to patients with pathological T3 or T4 primary tumors, N2 and N3 stage nodal disease, N1 stage at levels IV or V, vascular embolism, or perineural invasion, as determined in the final pathological specimens. In our hospital, radiotherapy is considered for patients with one positive node or perineural invasion who do not exhibit other adverse features. Postoperative radiochemotherapy was administered in patients with ECS and positive margins. Radiochemotherapy can also be considered for patients with pT3 or pT4 primary tumors, N2 or N3 stage nodal disease, nodal disease at levels IV or V, perineural invasion, or vascular embolism. Radiotherapy was administered no more than 6 weeks after surgery and was delivered by a linear accelerator at a total dose of 60-66 Gy (1.8-2.0 Gy/fraction). If chemotherapy concurrent with radiotherapy was indicated, cisplatin (80 mg/m 2 ) and 5-fluorouracil (400-500 mg/m 2 ) were administered in two cycles and repeated after 4-5 weeks. The treatment protocol used in this study was previously described 36 . Clinical and pathological parameters. We recorded the patients' sex, age at OSCC diagnosis, survival time, pathological AJCC anatomical site, AJCC (7th edition) TNM stage, pathological grade, recurrence, and positive lymph nodes at each level of the neck. Skip metastasis was defined as positive neck metastasis at levels IV and V without the involvement of higher levels (levels I-III). We also recorded behaviors such as smoking, chewing betel nuts, and alcohol consumption. The anatomical sites were then subclassified as the alveolar ridge, anterior two-thirds of the tongue, buccal mucosa, hard palate, FOM, retromolar trigone, mucosa of the lip, and www.nature.com/scientificreports/ body of the lip. Information about mortality was retrieved from the cancer registry center of Changhua Christian Hospital and from data updated annually by the Health Bureau of Changhua City.

Statistical analyses.
We calculated continuous and categorical variables as mean ± standard deviation and percentage, respectively. We used the Mann-Whitney U test to compare the continuous variables and the chisquare test to compare the differences in the categorical variables among the different patient groups. To examine the effects of the clinicopathological factors on survival in patients with OSCCs, we used univariate and multivariate Cox proportional hazards models. We subsequently calculated hazard ratios and 95% confidence intervals. Rates of outcomes were estimated using Kaplan-Meier analyses. To compare the group survival functions, we used log-rank tests based on survival data. A p value of < 0.05 was considered statistically significant.
To perform all statistical analyses, we used the statistical package SPSS version 16 for Windows (SPSS, Chicago, IL, USA).

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.