Survival and Swallowing Function after Primary Radiotherapy versus Transoral Robotic Surgery for Human Papillomavirus-Associated Oropharyngeal Squamous Cell Carcinoma

The prevalence of human papillomavirus (HPV)-related oropharyngeal cancer is increasing worldwide [1]. Treatment outcomes for patients with early-stage HPV-associated oropharyngeal squamous cell carcinoma (OPSCC) are favorable, irrespective of primary treatment modality [2, 3]. Current standard-of-care options for upfront treatment include radiotherapy (RT) with or without chemotherapy or transoral robotic surgery (TORS) followed by risk-adapted adjuvant therapy [4‒7]. Prospective data comparing the two modalities are limited, but a randomized phase II study found improved swallowing function with RT compared with TORS and no significant difference in progression-free survival (PFS) and overall survival (OS) [8]. Though these data represent a substantial contribution to the evidence comparing these two excellent oncological paradigms, the results of this small study may not be entirely generalizable to routine practice. Variations in TORS techniques and expertise may impact the dysphagia outcomes noted in this study [9]. Additionally, a large proportion (51%) of patients had clinical N2 disease (AJCC 7th Ed.: single node larger than 3 cm, multiple ipsilateral nodes, or bilateral/contralateral nodal involvement), which may impact the rates of posttreatment dysphagia [10]. Both treatment approaches are associated with swallowing defects during and after treatment that may require gastrostomy tube (GT) nutritional support [3, 11‒13]. This is particularly an issue if triple-modality therapy (surgery followed by adjuvant chemoradiotherapy [CRT]) is indicated, since acute toxicity is increased [12, 14, 15]. In-depth analyses of real-world functional swallowing outcomes after primary RT or TORS are needed to determine if these disparities in swallowing function are indeed related to primary treatment modality or clinicopathological factors that dictate treatment intensity within the current standard of care. A better understanding of these factors may facilitate optimal patient selection and inform treatment, particularly in the setting of treatment deescalation for HPV-associated OPSCC. In this study, we aim to evaluate oncological outcomes as well as functional swallowing outcomes in patients with early-stage HPV-associated OPSCC.

Patients with AJCC 8th Ed. stage I-II HPV-associated OPSCC diagnosed between 2012 and 2018 were identified. Patients must have been treated with either primary RT (with or without concurrent chemotherapy) or TORS (with or without adjuvant therapy) within our health system and must have had HPV-positive disease defined by p16 immunohistochemistry or by detection of high-risk HPV though HPV DNA polymerase chain reaction testing. Exclusion criteria included clinical T3-4 or N3 (AJCC 7th Ed. T3-4b, N3) disease, recurrent or second primary head and neck cancer, reirradiation, treatment with induction or adjuvant chemotherapy, primary treatment at an outside facility, radiographic extranodal extension (ENE), incomplete course of prescribed RT, inadequate clinical follow-up (less than 3 months), and metastatic disease at diagnosis. In total, 395 patients were identified, 228 excluded, and 167 included in the final analysis (see online suppl. Fig. 1 for CONSORT Diagram; for all online suppl. material, see https://doi.org/10.1159/000531995).

Patients were evaluated within a multidisciplinary head and neck oncology group consisting of otolaryngologists/head and neck cancer surgeons, radiation oncologists, medical oncologists, neuroradiologists, dentists, speech-language pathologists, dieticians, and oncology nurse navigators. Staging imaging consisted of either CT imaging of the neck and chest or whole-body ¹⁸F-FDG PET/CT. Upfront treatment was generally guided by multidisciplinary tumor board recommendations considering tumor and nodal disease burden, surgical resectability, patient age/comorbidity, and preference. Surgical management included TORS with ipsilateral or bilateral neck dissection by the discretion of the operating surgeon. The primary tumor was resected with a goal of obtaining negative deep and circumferential mucosal margins [16]. Tracheotomy was not performed routinely. Adjuvant therapy was generally recommended after a multidisciplinary discussion and indicated as follows: adjuvant RT in the case of pT3 or higher tumor, AJCC 7th Ed. pN2 or greater, perineural invasion, lymphovascular invasion, and close margin (<2 mm); and adjuvant CRT in the case of positive margin or ENE (of any extent). Patients that received adjuvant RT at our institution (n = 22; 41 patients were referred to other facilities for adjuvant RT) were treated as previously described; briefly, 60–66 Gy in 30–33 fractions was prescribed to the tumor bed and involved neck (66 Gy to areas of positive margin or ENE) with 52.8–59.4 prescribed to low-/intermediate-risk elective regions using a simultaneous integrated boost technique [17]. In the RT group, definitive RT was delivered using 70 Gy in 35 fractions to the high-risk clinical target volume surrounding GTV, 63 Gy in 35 fractions to the intermediate-risk target volume including involved and first-echelon nodal levels, and 56 Gy in 35 fractions to the low-risk elective neck using simultaneous integrated boost. RT alone was utilized for patients with limited disease (T1-2N0M0); concurrent CRT was recommended for patients with more locally advanced disease (T3-4N1-2M0).

Our primary objective was to compare PFS between the two treatment groups. PFS was defined as the time from diagnosis to death, recurrence, or censorship at the last clinical follow-up, whichever came first. Secondary endpoints included OS, defined as time from diagnosis to death from any cause or censorship at the last clinical follow-up, whichever came first; presence of a GT (whether or not it was being used) at 1-year posttreatment (yes vs. no); swallowing function at baseline and 1-year posttreatment scored by the Functional Oral Intake Scale (FOIS) [18]; 1-year posttreat GT dependence (defined as the FOIS score of 1–3); and percent weight loss at 1-year posttreatment compared to weight at diagnosis.

We first examined baseline characteristics of patients, stratified by the treatment group. For continuous variables, we reported medians and interquartile ranges, and medians were compared across the two groups using the Kruskal-Wallis test. We report frequencies and percentages for categorical variables and compared these variables across the two groups using the χ² (or, where appropriate, Fisher’s exact) test. Similarly, we also compared treatment-related variables and swallowing-related outcomes, as well as change the score pertaining to swallowing, in bivariate tests across the two groups, using t tests for comparisons of means and χ² (or Fisher’s exact tests if cell sizes required) for categorical variables. To examine the trivariate relation between treatment group, nodal status, and swallowing outcomes, we present descriptive summary data without statistical tests, due to small sample sizes, heterogeneous variances in swallowing outcomes by the treatment group, and the post hoc nature of our exploration. We used the Kaplan-Meier method to compare PFS and OS across the two treatment groups and evaluated proportional hazards models for the two survival outcomes where we compared the two treatment groups after adjusting for baseline characteristics that differed significantly between the two groups. In online supplementary Table 1 where we compare three subgroups at a time on swallowing-related outcomes, we report p values from global hypothesis tests across the three groups. In this table, we use p values derived from F tests (from the analysis of variance models) for comparisons of means of continuous variables, and p values derived from Fisher’s exact tests for the categorical variables. We used a two-tailed alpha of 0.05 to denote statistical significance throughout. All analyses were conducted in SAS 9.4.

A total of 167 patients were treated, 116 with primary TORS and 51 with primary RT. Median age (58–59 years) did not differ between the two groups, nor did smoking history, with approximately the same percentage in each group (51–52%) having a ≥10 pack-year smoking history (Table 1). In the TORS group, the predominant primary cancer site was the tonsil (67.2%), while in the RT group it was the base of the tongue (56.9%; p value for difference between groups, 0.009). There were significant differences in the baseline clinical stage by the AJCC 7th and 8th Eds.: there were more patients with advanced tumor and nodal stage in the RT group compared to the TORS group. Large LN (>4 cm clinically) was present with almost equal (approximately 20%) frequency in both groups; no patients in either group had radiographic ENE. The median Charlson Comorbidity Score was 1 for the TORS group and 2 for the RT group (p = 0.037). Workup was performed by either a CT neck/chest (97–98% in each group) or PET/CT (78.5% in TORS vs. 96.1% in RT, p value for difference = 0.004).

TORS was performed with a unilateral (73.3%) or bilateral (25.0%) neck dissection (Table 2). Tracheotomy was performed in 4% of patients receiving upfront TORS. Mean total estimated blood loss was 100 mL (IQR: 69–200). No patients had a grade 5 event, nor was massive bleeding observed. Fifty-three (45.7%) patients were treated with TORS alone. Adjuvant RT was delivered to 63 (54.3%) patients: RT alone in 30 (of TORS patients, 25.9%) and concurrent CRT in 33 (of TORS patients, 28.4%).

RT was delivered to all 51 patients in the RT group and 63 (54.3%) patients in the TORS group (p < 0.001). Median RT dose was 60 Gy (range: 60–66) in the TORS group, and all patients in the RT group received 70 Gy (p < 0.001). All but 1 patient in the RT group received concurrent chemotherapy. RT, if given, was primarily delivered to the bilateral neck using intensity-modulated RT (Table 2). Salvage neck dissection was performed in 4 (7.8%) patients in the primary RT group.

With a median follow-up of 40.6 months in the total sample (95% CI: 33.1–44.6 months), there were no observed differences in OS or PFS between the treatment groups. 3-year OS in the TORS group was 90.4% compared to 81.0% in the RT group (log-rank p = 0.497, Fig. 1). After adjustment for age, stage, and Charlson Comorbidity Index, the adjusted HR for the RT group (vs. TORS) for OS was 1.13 (95% CI: 0.46–2.74). The 3-year PFS was 87.6% versus 74.8%, respectively (log-rank p = 0.173, Fig. 2). Adjusted HR (RT vs. TORS) for PFS was 1.57 (95% CI: 0.74–3.31).

GT presence, dependence, weight change, and FOIS outcomes are detailed by the treatment group in Table 3. One-year GT prevalence (whether or not it was being used) was higher in the RT group, as was GT dependence (FOIS 1–3). Overall, 102 (88%) patients in the TORS group and 46 (90%) patients in the RT group were evaluable for 1-year swallowing function by FOIS. The mean baseline FOIS was 6.9 in the TORS group versus 6.7 in the RT group (p = 0.06). At 1 year, this had decreased to 6.4 and 5.6, respectively. Grouped categorically, higher rates of 1-year FOIS scores of 1–3 and 4–5 and lower rates of FOIS score 7 were observed in the RT group. The distribution of change scores in the FOIS scale was significantly different by the treatment group, with the RT group more likely to experience a negative (worsening) change in the FOIS score (Table 3).

We wished to explore this latter finding in more depth to examine possible confounding of the treatment group-FOIS change relation by the nodal status. Because of the heterogeneous variances in the FOIS change score in the 2 treatment groups, our ANOVA modeling results reported here must be interpreted cautiously and are meant to convey mostly exploratory/hypothesis-generating findings in this context. In exploratory ANOVA modeling of this change score where we examined the joint impacts of tumor stage, nodal stage, and treatment group, only nodal stage was significantly associated with the FOIS change score. Figure 3 shows mean FOIS change scores by the treatment group and the nodal status (in the model, and in the table, we grouped N0 and N1 to avoid very small cell sizes). In both treatment groups, the N2 group experienced significantly greater worsening of their FOIS scores than the N0/N1 group. Because N2 stage is far more prevalent (over 5 times so, per Table 1) in the RT group compared to the TORS group, it is likely that the bivariate association between the treatment group and the FOIS change score reported in Table 3 is mostly due to confounding by nodal stage.

In order to determine the functional swallowing impact of ultimate treatment intensity incorporating adjuvant therapy after TORS, we further examined these outcomes in a subgroup analysis by the totality of therapy delivered: TORS alone, TORS + RT, TORS + CRT, or primary RT (nearly all of which were treated with CRT). Swallowing outcomes among the four groups are described in online supplementary Table 1. When comparing the three primary TORS groups (TORS alone, TORS + RT, and TORS + CRT), there was no significant difference between GT prevalence and dependence, but percent weight loss and FOIS at 1 year significantly differed between groups, favoring the TORS alone group. The mean (SD) absolute difference in FOIS at 1 year was 0 (0.5), −0.6 (0.8), and −1.1 (1.5) for the TORS alone, TORS + RT, and TORS + CRT groups, respectively (p < 0.001, online suppl. Fig. 2). When we examined the groups treated with TORS followed by adjuvant therapy (either RT or CRT) with the primary RT group, the GT prevalence was significantly higher in the primary RT group, but no other differences in swallowing measures was observed (online suppl. Table 1). The 1-year mean (SD) absolute difference in FOIS in the primary RT group was −1.1 (1.8), and there was no significant difference in 1-year mean change in FOIS when comparing the TORS + RT, TORS + CRT, and primary RT groups with each other (p = 0.386, online suppl. Fig. 2).

For patients with HPV-associated OPSCC, excellent prognosis and younger age at diagnosis are resulting in a population of cancer survivors with great longevity and, thus, propensity for late treatment-related sequelae. In this cohort, it is therefore critical to minimize long-term toxicities. Dysphagia is a major quality of life issue for patients treated for head and neck cancer treated either surgically or nonsurgically [19, 20]. This potential long-term radiation damage has a profound impact on social functioning and mental health [21]. In order to reduce or limit these sequelae while maintaining excellent oncological outcomes, multiple paradigms have been explored with regard to treatment deescalation including cetuximab-RT [22, 23], reduced-dose RT/CRT [24‒26], induction chemotherapy followed by reduced-dose RT [27, 28], postoperative RT with either reduced dose or RT volume [29‒31], and TORS followed by the deintensified adjuvant therapy [32]. A focus on patient and disease-related drivers of dysphagia are important to consider when reducing treatment intensity, as this may guide deescalation efforts.

Patients in this study were treated with TORS followed by conventional adjuvant therapy. The surgical group was more likely to have a tonsil primary site (vs. base of tongue) as well as lower T and N stage; of the 24 patients with N2 (AJCC 8th Ed.) disease, 17 of them were in the RT group. The rate of unilateral RT was low (approximately 5%) and slightly lower than the 12.1% reported in a recent clinical trial of lower stage patients [24]. PFS and OS were comparable, consistent with prior reports [3, 8]. In contrast to the ORATOR study [8], a very low percentage of patients in this study that were treated with upfront TORS received a tracheotomy with very low volumes of estimated blood loss noted.

Crude measures of swallowing function such as 1-year GT presence, 1-year GT dependence, on-treatment weight loss, and FOIS at 1 year were better in the TORS group, in contrast to the ORATOR study [8]. This may be related to selection bias inherent within this retrospective cohort, with a clearly significant distribution of more advanced tumor and nodal stage in the RT group. However, this may also be related to surgical variation in practice and techniques [9]. GT rates may also be biased by the institutional practice during the study period to place prophylactic GT in patients planned for CRT, particularly those planned for bilateral neck treatment (the vast majority of the RT group). Our findings that swallowing outcomes were significantly worse in the patients treated with more intensive adjuvant therapies were also not consistent with the findings of the ORATOR trial and may also be explained by differences in surgical and RT techniques, patient selection, and the differing methods utilized to measure swallowing function. Our study was unable to compare primary RT versus primary CRT due to 50 of 51 patients being treated with CRT in that group. These findings seem to suggest that a lower intensity of surgical management is associated with better 1-year functional swallowing outcomes in our cohort, and that swallowing outcomes in patients treated with more intense adjuvant therapies (i.e., adjuvant CRT) approximate those seen in patients treated with primary CRT.

In this study, we identify clinical nodal status as a factor that is associated with both treatment groups at baseline and swallowing function, identified by a decline from baseline to 1-year measurement of the FOIS score. Careful interpretation of these results indicates that clinical nodal status rather than upfront treatment paradigm is most closely associated with long-term swallowing function. This is likely related to the extent and intensity of treatment for more advanced disease, as evidenced by subgroup analyses identifying significant changes in swallowing function with increasing intensity of treatment in the primary TORS group. This may be primarily related to the RT volume and dose required to treat the neck in a patient with clinical bilateral node involvement – patients with bilateral nodal involvement are treated with adjuvant RT doses to the bilateral neck, as they are not candidates for unilateral RT [33]. They may also require higher doses in the presence of either pathological ENE or positive surgical margins [34]. This may result in higher doses to the dysphagia/aspiration-related structures (DARS) such as the pharyngeal constrictor muscles, larynx, and cricopharyngeus muscle and result in worse swallowing outcomes [10]. In this situation, RT volume reduction strategies that decrease the amount of neck treated may reduce the overall treatment impact on long-term swallowing. Omission of elective neck irradiation in the setting of a pathologically negative (pN0) dissected neck was associated with high rates of regional control (5-year rate: 94.5%) and favorable quality of life outcomes in a small, single-institutional series [30]. Unilateral elective nodal irradiation to patients with unilateral nodal drainage confirmed by SPECT/CT has demonstrated feasibility and low rates of severe dysphagia, GT feeding, and xerostomia [35].

Our findings may also be related to an increase in RT dose to DARS due to the greater proportion of base of tongue tumors in the RT group, necessitating a higher dose closer to the midline. In this case, deescalation of the primary site may be more beneficial in terms of reducing RT dose to DARS. Omission of the primary site after resection with no pathological risk factors within the primary tumor is well tolerated and is associated with a low risk of local failure [31]. These findings are worthy of future study, ideally in comparative randomized trials, to continue to optimize the therapeutic index for RT for HPV-associated OPSCC. In any case, it is important to remain vigilant in reducing the dose to the DARS, which has been found to significantly improve swallowing function in a randomized trial of patients receiving definitive RT/CRT for oropharyngeal and hypopharyngeal cancers [36, 37]. Reduction of RT dose to DARS to minimize the risk of swallowing complications is translatable to patients treated with a primarily surgical approach [38, 39]. At our institution, current practice involves routine delineation and limitation of dose to the following structures: pharyngeal constrictor muscle, nontarget pharyngeal constrictor muscle (i.e., the RTOG-NRG “Pharynx” organ at risk), glottic larynx, glottic/supraglottic larynx, and cricopharyngeus muscle [40].

This study has limitations, including its limited sample size, primarily in the N2 nodal group. Selection bias is likely present, affecting the crude measures of swallowing outcomes; we attempted to alleviate this bias by including nodal status as a predictor in an exploratory ANOVA model and by presenting results stratified by nodal status. FOIS (obtained through review of the medical record) is a useful measure of swallowing function in head and neck cancer patients but is not as sensitive and consistent as other patient-reported measures of swallowing function that would have been possible with a prospective design. A continued follow-up of such patients will be done to assess longer-term swallowing outcomes as swallowing dysfunction can be progressive with time [20].

In conclusion, treatment outcomes for patients with early-stage HPV-associated OPSCC appear comparable between primary TORS or RT. Swallowing dysfunction 1 year after treatment occurs more frequently in patients with N2 nodal status at diagnosis, which necessitates more aggressive treatment to the neck, but appears independent of upfront treatment (RT or TORS). Further study involving strategies to reduce dose and volume to the contralateral neck is warranted to improve long-term swallowing risks.

This study protocol was reviewed and approved by the Institutional Review Board of Wake Forest University Health Sciences (approval number IRB00047017); written informed consent was waived based on the nature of this retrospective review.

The authors have no conflicts of interest to declare.

The authors wish to acknowledge the support of the Wake Forest Baptist Comprehensive Cancer Center Biostatistics Shared Resource, supported by the National Cancer Institute’s Cancer Center Support Grant (award number P30CA012197). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute. This work was supported by the Wake Forest Baptist Medical Center and National Center for Advancing Translational Sciences, National Institutes of Health-funded Wake Forest Clinical and Translational Science Institute (WF CTSI) through grant award UL1TR001420.

All authors made substantial contributions to the conception or design of the work (R.T.H., B.J.L., N.M., R.F.S., J.H.Y, C.M.L., B.A.F., K.M.G., and J.D.W.), acquisition (R.T.H., N.M., R.F.S., J.H.Y., and C.M.L.), analysis or interpretation of the data (R.T.H., B.J.L., N.M., R.F.S., J.H.Y., C.M.L., B.A.F., K.M.G., and J.D.W.), and drafting and critical revision of the work (R.T.H., B.J.L., N.M., R.F.S., J.H.Y., C.M.L., B.A.F., K.M.G., and J.D.W.) and provided final approval and agree to be accountable for all aspects of the work (R.T.H., B.J.L., N.M., R.F.S., J.H.Y., C.M.L., B.A.F., K.M.G., and J.D.W.).

The data that support the findings of this study are not publicly available due to their containing information that could compromise the privacy of research participants but are available upon reasonable request. Further inquiries can be directed to the corresponding author.