Abstract
Background: Despite adjuvant radiotherapy, survival outcomes remain poor in patients with salivary gland malignancies who have multiple poor prognostic factors. This study aimed to determine which patients may benefit from treatment intensification. Patients and Methods: Patients who underwent curative resection with or without adjuvant radiotherapy between 2002 and 2014 were identified and a retrospective chart review was performed. Overall survival (OS) and disease-free survival (DFS) were the main outcomes measured. Results: A total of 95 patients met the inclusion criteria. The median follow-up was 46.8 months. The median age was 60 years. Radiotherapy was given to 78 patients. Multivariate analysis revealed that male sex and perineural invasion significantly reduced overall and disease-free survival. Distant metastases comprised of 67% of recurrences and 33% were locoregional. Conclusion: Adjuvant chemoradiotherapy should be considered for patients with tumors with perineural invasion, especially in males with high-risk histopathology or high-grade, late-stage disease. To our knowledge, this is the first study to assess the impact of pack-year smoking history on survival outcomes.
Salivary gland malignancies (SGMs) are a rare group of tumors that comprise of approximately 6% to 8% of head and neck cancers in the United States, with a trend of increasing incidence over the past 40 years (1). Moreover, they are an exceptionally heterogeneous group of tumors with 24 different types of SGMs recognized by the World Health Organization (WHO) (2).
Surgical resection is the pillar of curative treatment for all types of SGM and is considered sufficient treatment in those with tumors with good prognostic features such as low grade, clear resection margins, stage I-II disease, and indolent histology. Postoperative radiotherapy (PORT) significantly improves local control and overall survival (OS) in those with tumors with poor prognostic factors such as high grade, stage III-IVB, close or positive surgical margins, perineural invasion and aggressive histology (3-8). Despite PORT, survival outcomes remain poor in SGM with the poor prognostic factors listed above, as well as facial nerve dysfunction, lymphovascular invasion, and increased age at diagnosis (9-14).
Randomized, prospective trials have shown that postoperative chemoradiotherapy (POCRT) for squamous cell carcinoma of the head and neck improved locoregional control, disease-free survival (DFS) and OS in those with tumors with extracapsular extension and positive margins (15-17). No randomized, prospective data comparing PORT to POCRT exists for SGMs; however, Radiation Therapy Oncology Group (RTOG) trial 1008 is currently comparing PORT against POCRT in resected major and minor SGM with the following features: high-risk histology, high or intermediate grade, stage III-IVB disease, or stage I-II disease with positive or close margins. The existing data comparing PORT to POCRT for SGMs are sparse and consist of small, retrospective studies that employed a variety of chemotherapy regimens.
Herein, we report on a retrospective analysis of our Institution's experience treating SGMs and identify poor prognostic factors in our patient population. This study aimed to recognize tumor types that may benefit from POCRT.
Patients and Methods
Study population. Following Institutional Review Board approval (IRB#: 4798), we identified all patients with non-metastatic major or minor salivary gland tumors who underwent curative resection between 2002 and 2014. Exclusion criteria were: age <18 years, neoadjuvant treatment received, and squamous cell carcinoma with suspected skin primary. All information from this population was obtained through retrospective chart review.
Histopathology and staging classification. Tumors were classified based on WHO pathology subtypes and were further stratified into high-risk and low-risk subgroups based on the combination of histological grade and pathology (2). High-risk histopathological subtypes were: adenoid cystic carcinoma, squamous cell carcinoma, salivary duct carcinoma, high-grade mucoepidermoid, high-grade adenocarcinoma, and high-grade carcinoma ex-pleomorphic adenoma (2). Tumors were also grouped by stage into either early (stage I-II) or late (stage III-IVB) stage.
Treatment and follow-up. All patients underwent curative resection prior to receiving radiation therapy if adjuvant therapy was indicated. Neck dissection was performed therapeutically for those with node-positive disease and electively for those with node-negative disease with high-risk features (high-grade tumor, T3 or T4 tumor, or facial nerve paralysis). Postoperative radiotherapy using intensity-modulated radiation therapy or 3D conformational radiotherapy was given if one of the following characteristics was present: positive surgical margins, T2 tumor size with poor differentiation, T3 or T4 disease, node-positive disease, perineural invasion of large nerve. Target doses were between 50-66 Gy depending on pathological characteristics.
Following completion of therapy, patients were followed up by their surgeon every 2-3 months for the first 2 years and then every 6-12 months based on surgeon's assessment of histopathological risk factors.
Statistical methods. Simple descriptive statistics were created for all covariates. Bivariate analysis was performed on continuous variables using analysis of variance. Chi-square analysis and Fisher's exact tests were performed on categorical variables. In order to evaluate OS and DFS, Kaplan–Meier survival curves and log-rank tests of homogeneity identified where the survival functions differed between the comparison groups. Three- and 5-year survival probabilities were imputed for each variable as well using a linear interpolation. A Cox proportional hazards model was then used to assess the impact of the significant variables on OS and DFS with both hazard ratio (HR) and 95% confidence interval (CI).
For all final multi-variable models, an inclusion criterion of p<0.25 on univariate analysis was used. All two-way interactions were then explored if the variable met this criterion. Finally, backwards elimination was performed, eliminating variables with the highest p-value. As variables were excluded, confounding was assessed. Whenever a variable's exclusion changed the HR by ≥20% and it had a p-value <0.30, it was retained in the model and considered a confounder.
Results
Overall, 95 patients were included in the analysis. The median follow-up was 46.8 months. The average age of the patient cohort was 57.9±16.7 years, with a median age of 60 years. A total of 78 patients (82%) received PORT. A table of all measured simple descriptive variables can be made available upon request.
Frequency of sex, smoking and stage by grade (n=95).
Patient and tumor characteristics associated with tumor grade and smoking history. Since high-grade tumors have a poor prognosis, association of tumor grade with sex, smoking history, and tumor stage was explored (Table I). Male sex (p=0.0119) and late-stage disease (p<0.0001) were significantly associated with high-grade disease. Association of any previous smoking history with lymphovascular invasion, perineural invasion, and neck dissection was also explored; however, no significant association was identified (data not shown).
Influence of patient and tumor characteristics on OS. OS at 3 and 5 years for the PORT group (n=78) was 79% (95% CI=69-89%) and 68% (95% CI=56-80%) respectively. OS at 3 and 5 years for the group treated with surgery alone (n=17) was 96% and 88%, respectively, with non-reliable CIs due to small sample size and low mortality of this group.
Concerning patient characteristics, univariate analysis revealed male sex (p=0.0033) and ≥10 pack-year smoking history (p=0.0144) significantly reduced OS. Concerning tumor characteristics, high grade (p=0.0024), N2 disease (p=0.0047), squamous cell carcinoma histology (p=0.0044), perineural invasion (p=0.0013) and neck dissection (p=0.0483) significantly reduced OS (Table II).
The variables that significantly reduced OS on multivariate analysis were male sex (HR=5.67, 95% CI=1.79-18.00; p=0.0032), and perineural invasion (HR=3.88, 95% CI=1.50-10.04; p=0.005) (Figures 1 and 2). Stage (early vs. late) did not significantly influence OS on multivariate analysis (HR=1.52, 95% CI=0.65-3.58; p=0.3018).
Univariate Cox proportional hazards results for overall survival (OS) and disease-free survival (DFS).
Influence of patient and tumor characteristics on DFS. DFS at 3 and 5 years for PORT group was 70% (95% CI=59-89%) and 58% (95% CI=46-70%), respectively. DFS at 3 and 5 years for the group treated with surgery alone was 78% (95% CI=58-98%) and 72% (95% CI=47-97%), respectively. There was no statistically significant difference in DFS between these two groups (p=0.1794).
Most results for DFS were very similar to those of OS with only minor changes in HR noted (Table II). However, significant differences were seen amongst some tumor variables. Variables that significantly influenced DFS but not OS were intermediate grade (p=0.0132), late-stage disease (p=0.0375), and histopathological risk (p=0.0306). Variables that significantly influenced OS, but not DFS were N2 disease (p=0.1182) and neck dissection (p=0.1096).
The variables that significantly reduced DFS on multivariate analysis were male sex (HR=2.29, 95% CI=1.04-5.03; p=0.0397) and perineural invasion (HR=2.95, 95% CI=1.43-6.11; p=0.0035). Stage (early vs. late) did not significantly influence DFS on multivariate analysis (HR=1.49, 95% CI=0.74-3.00; p=0.2660).
Disease recurrence trends of analyzed patients. Concerning disease recurrence, a total of 27 tumors recurred (28%), with 18 patients experiencing distant metastasis (67%) and nine experiencing locoregional recurrence (33%). The lungs were the most common site of distant metastasis (14 patients). Bone and brain were the other sites of metastasis, with those of bone being more common than those of brain.
Concerning locoregional recurrence, regional nodes were the most common site of recurrence (eight patients). Recurrence at the site of primary tumor was the next most common site (three patients). Since a small number of patients experienced recurrence and prognostic factors influencing DFS had already been analyzed, statistical analysis was not performed on this specific subgroup.
Kaplan–Meier curve of overall survival according to patient sex. 3-Year survival: Male vs. female: 74% [95% confidence interval (CI)=61-87%] vs. 90% (95% CI=80-100%); 5-year: 58% (95% CI=42-74%) vs. 88% (95% CI=76-99%), respectively (p=0.0014).
Kaplan–Meier curve of overall survival according to perineural invasion. 3-Year survival: invasion vs.no invasion: 69% [95% confidence interval (CI)=54-85%] vs. 93% (95% CI=85-100%); 5-year: 54% (95% CI=35-72%) vs. 88% (95% CI=77-99%), respectively (p=0.0004).
Univariate Cox proportional hazards results for overall survival (OS) and disease-free survival (DFS) adjusted for age, grade, stage and histology.
Prognostic factors when adjusting for age, grade, stage, and histology. In order to better elucidate the impact of tumor and treatment variables, we adjusted for age, grade, stage and histology. The only variable that significantly reduced OS and DFS after adjustment was perineural invasion (Table III).
Discussion
Given the rarity and heterogeneity of SGMs, no randomized, prospective trials have examined the treatment, outcomes, or prognostic factors of these types of cancers. While the inherent nature of this retrospective study limits the conclusions that can be drawn from it, the results are consistent with the growing body of literature examining the prognostic factors and outcome of SGMs (6, 12, 14, 18).
Characteristics associated with high-grade tumors. A retrospective study by Haderlein et al. demonstrated that poor tumor differentiation, regardless of histology, was the only factor that predicted significantly shorter survival outcomes (13). While high-grade did not impact our cohort's survival on multivariate analysis, it was significantly associated with male sex and late-stage disease. No literature exists as to why male sex is associated with high-grade tumors and further research should examine the biological and social factors that are likely at play. High-grade disease is a characteristic of aggressive cancer that is more likely to present as late-stage, which is the mostly likely explanation for this association.
Patient variables that influence survival outcomes. Male sex and smoking history of 10 pack-years or more significantly reduced OS and DFS on univariate analysis, with only male sex remaining significant on multivariate analysis. Previous studies have demonstrated that male sex is a significant negative prognostic factor when examining survival outcomes (4, 19). The association between male sex and high-grade tumor found in our cohort partially explains these univariate analysis results. No explanation exists in the literature for the multivariate analysis findings, and its etiology is likely due to a multitude of biological and lifestyle/environmental factors.
Previous studies have not demonstrated a significant impact of smoking history on survival outcomes in patients with SGM, but pre-diagnosis smoking has been shown to have a significant negative impact on survival in patients with head and neck cancer, especially males (20). Notably, as far as we are aware of, this study is the first to quantify smoking history in patients with SGMs and assess the impact of pack-year smoking history on survival outcomes.
Surprisingly, age ≥65 years had no significant impact on survival in our cohort. In previous studies, ages greater than between 50 and 61 have all been associated with significantly reduced survival outcomes (7, 9, 10, 14, 18).
Tumor variables that influence survival outcomes. After multivariate analysis and adjusting for age, stage, grade and histology, only perineural invasion had a significant impact on OS and DFS. Both perineural and lymphovascular invasion have been shown to be negative prognostic factors for locoregional control and survival in SGM (12, 14, 18). There is more evidence demonstrating the significant impact of perineural invasion on locoregional control and survival than there is for the impact of lymphovascular invasion (5, 11, 18). Our results are consistent with the literature regarding perineural invasion and suggest that it is the tumor variable that portends the worst prognosis.
Surprisingly, N2 disease, which has been shown to be a poor prognostic factor, did not have a significant impact on survival on multivariate analysis of our cohort (9, 11, 12, 14). The lack of significant impact on DFS in univariate analysis could be due to the improvement of locoregional control using PORT in node-positive disease (3, 5). The significant decrease in OS for N2 disease on univariate analysis is likely due to an association with other poor prognostic variables, such as high grade, and short disease course following metastasis. However, patients with late-stage tumors and high-risk histopathology had significantly worse DFS despite receiving PORT, which suggests that patients with high-risk histopathology or high-grade, late-stage disease may benefit from the addition of adjuvant chemotherapy if seen with other poor prognostic factors, especially perineural invasion.
Surprisingly, neither positive surgical margins nor any category of negative margins significantly affected survival outcomes. Previous studies have shown positive or close margins to be a negative prognostic factor for survival and locoregional control outcomes (5, 6, 12, 14). Such a lack of significance for positive margins in our cohort was unexpected, but this finding could be due to the small size of our cohort and having only 23 cases of positive margins to analyze.
Additional factors that have been shown to significantly reduce survival outcomes and locoregional control are skin invasion, facial nerve paresis, bone invasion, and latency ≥8 months from symptom onset to diagnosis (4, 14, 18). While our study did not examine these variables, patients with tumors with these characteristics may benefit from concomitant chemotherapy as they are associated with poor outcomes despite PORT.
Potential for POCRT in patients with high-risk SGM. A limited number of retrospective studies have compared PORT to POCRT with conflicting results due to variable study designs. The most convincing data for POCRT is a case-controlled, retrospective study by Tenvetyanon et al. comparing OS and toxicity between two groups of 12 patients with late-stage disease and close or positive margins. POCRT significantly improved 3-year OS (88% vs. 44%, p=0.05), but no significant difference existed for locoregional control (21). The POCRT group experienced more grade 3 toxicities and was more likely to receive intensity-modulated radiation therapy (21). POCRT consisted of platinum-based chemotherapy and a median radiation dose of 63 Gy.
Most studies that have compared POCRT against PORT have not demonstrated significant improvement of survival outcomes when utilizing POCRT in treating SGMs (19, 22-25). Some of these studies even demonstrated significantly worse survival outcomes and locoregional control in the POCRT group due to more adverse prognostic factors commonly found in tumors treated with POCRT (19, 24). POCRT groups were more likely to have late-stage and high-grade disease, high-risk histology, positive margins, and perineural invasion (19, 22, 24). Moreover, chemotherapy regimens varied between these studies but all were platinum-based regimens.
Our results identify patients that are in need of more aggressive adjuvant treatment. The fact that the majority of treatment failures in our cohort were distant metastases highlights the need for systemic treatment in select patients. Our results further support the data of Terhaard et al. that identified tumors with T3-4 size, N2-N3 disease, or perineural invasion being at high risk for distant metastases (18). The majority of failures in that series were also distant metastases.
Lastly, RTOG 1008 is not evaluating perineural invasion or squamous cell histology – two poor prognostic factors we identified. Specifically, our results suggest that perineural invasion is the tumor variable with the worst prognosis and patients with such tumors may benefit from POCRT. Future research should include perineural invasion as a defining feature of high-risk disease and examine the efficacy of POCRT in patients with such tumors. Furthermore, targeted therapies and immunotherapies should continue to be evaluated in patients with SGMs, specifically salivary duct carcinoma and adenoid cystic carcinoma, as our knowledge of the molecular biology of these tumors continues to expand.
Conclusion
Survival outcomes were significantly worse in patients with tumors with perineural invasion. Moreover, perineural invasion was found to be a poor prognostic factor regardless of age, histology, stage and grade. Additionally, male sex portended significantly reduced survival. Notably, as far as we are aware of, this study is the first to quantify smoking history and assess the impact of pack-year smoking history on survival outcomes inpatients with SGMs, with a smoking history of 10 pack-years or more being significant on univariate analysis. Given that distant metastasis is the predominant form of recurrence of SGM, the addition of adjuvant chemotherapy should be considered for patients with tumors with perineural invasion, especially males with high-risk histopathology or high-grade, late-stage disease.
Acknowledgements
The Authors thank all of their colleagues who helped care for these patients over this 12-year period and those who continue to do so.
Footnotes
This article is freely accessible online.
- Received June 8, 2017.
- Revision received July 4, 2017.
- Accepted July 6, 2017.
- Copyright© 2017, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
