Long-term survival in patients with metastatic head and neck squamous cell carcinoma treated with metastasis-directed therapy

Background Our objective was to evaluate the outcomes of metastatic head and neck squamous cell carcinoma (HNSCC) by disease burden with an emphasis on metastasis-directed therapy (MDT) in patients with limited metastatic disease burden. Methods In total, 186 patients who developed metastatic disease after definitive therapy for HNSCC were included. Clinically and radiographically apparent metastases were enumerated. Kaplan–Meier methods were used to estimate survival. Cox regression was used to assess the association between clinical variables. Results Patients with a single metastasis had a 5-year overall survival (OS) of 35% (95% CI 16–54%) in contrast to patients with multiple metastases with a 5-year OS of 4% (95% CI 2–9%). Thirty patients (16.1%) underwent MDT. On multivariable analysis, oral cavity or sinonasal primary (HR 2.22 95% CI 1.16–4.25, p = 0.015; HR 4.88, 95% CI 1.10–21.70, p = 0.037, respectively) were associated with higher risk of death, whereas receipt of MDT (HR 0.36, 95% CI 0.17–0.74, p = 0.006) was associated with lower hazard of death. Median subsequent metastasis-free survival and 5-year survival after MDT (n = 30) were estimated at 26.4 months (95% CI: 9.8–54.0) and 31%, (95% CI: 15–48%). Conclusions HNSCC patients with limited metastatic disease may derive significant benefit from MDT. Prospective trials evaluating MDT in HNSCC are warranted.

BACKGROUND Distant metastatic disease occurs in~15% of HNSCC patients after initial definitive management. [1][2][3] Patients can present in various states from a single site of metastasis and controlled local disease to widely disseminated metastases with or without local recurrence. Despite the intuition that these patients may have very different clinical trajectories, there is little guidance on whether management of HNSCC patients should be tailored based on metastatic disease burden.
Mounting scientific and clinical evidence has accumulated since initial proposal of the oligometastatic state by Hellman and Weichselbaum in 1995. 4 There has been a dramatic increase in the understanding of the highly complex process of metastasis initiation and metastatic outgrowth. 5 The increased understanding of the complexity and variability of metastatic biology has been accompanied by increasing reports of success using MDT to treat patients with oligometastatic disease.
Recently, prospective studies examining the role of local therapy in managing metastatic disease for modification of disease outcome has clarified a long-running debate about the role of local therapy in patients with limited metastatic disease burden. 6 Randomised phase II trials have demonstrated meaningful benefits of metastasis-directed therapy (MDT): androgen deprivation therapy-free survival in oligometastatic prostate cancer, 7 overall survival in patients treated with consolidative radiotherapy after chemotherapy for oligometastatic non-small cell lung cancer 8 and overall survival in multiple histologies treated with ablative radiation therapy versus physician choice systemic therapy. 9 However, as none of these studies focused on patients with HNSCC, we reviewed the outcomes of patients treated at our centre for locally advanced HNSCC who subsequently developed metastatic disease and analysed the outcomes of patients with limited metastatic disease who underwent MDT.

Patients
Patients treated at Memorial Sloan Kettering Cancer Center with either definitive or adjuvant local radiotherapy for HNSCC including oropharynx (OPC), oral cavity (OC), larynx, hypopharynx, nasopharynx (NPC), sinonasal and unknown primary between 1989 and 2014 were eligible for inclusion under an Institutional Review Board approved retrospective protocol. All patients who developed metastatic disease after the completion of their initial www.nature.com/bjc definitive treatment were included in this analysis. Patients with metastatic disease at the time of definitive local therapy for their HNSCC were excluded. Patients were considered metastatic based on pathology (biopsy or resection) or based on radiology alone in cases where biopsy was not performed, and imaging strongly suggested metastatic disease. In cases of lung metastases, available histopathological information such as concordance of HPV and or p16 status with primary disease, absence of in situ component and pathologist's impression were used to determine whether the lung disease more likely represented metastasis versus a new primary lung tumour. Non-concordant HPV/ p16 status, in situ disease and pathologist conclusion that disease was more likely new primary than metastatic disease were not included. The number of sites and total number of clinically and radiographically apparent metastatic disease at distant metastatic diagnosis were enumerated.
Follow-up after initial local therapy After completion of radiotherapy for local disease, patients were evaluated every 2 to 3 months for the first 2 years following treatment, and subsequently every 4 to 6 months. Follow-up visits consisted of a physical exam and flexible fibreoptic endoscopy. Three months after treatment, PET/CT, CT or MRI of the neck were performed. Afterwards, imaging studies were performed as clinically indicated.

Statistical considerations
Overall survival (OS) was calculated from time of distant metastases (DM) until death. Patients alive at last follow-up were censored. In patients who received MDT, subsequent metastasesfree survival (SMFS) was calculated from time of distant metastases (DM) until development of new lesions or death. Patients alive without additional metastases by last follow-up were censored. Kaplan-Meier methods were used to estimate median and annual outcomes with 95% confidence intervals (CI). The number of metastases was analysed two ways: (1) 1, 2, 3, 4, 5 + and (2) single vs. multiple. Based on the Kaplan-Meier findings, the number of metastases categorised as single vs. multiple was included in multivariable analyses.
Univariable and multivariable Cox regression were used to assess the relationship between number of metastases and other potential confounding factors with OS. Additional factors examined included gender, number of organs with metastasis (single vs. multiple), metastasis location (lung only, lymph node only, all others), age at DM, Karnofsky performance status (KPS) 10 at DM, months between initial diagnosis and DM (logarithmically transformed), original tumour location (oropharynx (OPC), oral cavity (OC), nasopharynx (NPC), sinonasal, all others), metastases location (lymph nodes only, lung only, all others), local control at the time of DM (yes vs. no), Charlson comorbidity index (CCI), chemotherapy after DM (yes vs. no) as a time-dependent factor and MDT (yes vs. no) as a time-dependent factor. Due to the high degree of missing data in HPV/p16 status (78%), this factor could not be explored in multivariable analyses. All other variables were included in multivariable analyses regardless of significance. However, as number of organs and metastasis location were essentially measuring the same factor (single versus multiple sites) and were found to be highly associated, we did not include metastasis location in the multivariable model. Furthermore, because of the overlapping definitions of number of organs with metastases with number of metastases (a patient with singular metastasis could not have multiple organs with metastases), we only included number of metastases in the multivariable model.
We assessed differences in clinical characteristics between patients who did and did not receive MDT with Fisher's Exact test and the Wilcoxon rank-sum test. In OPC patients, we analysed the relationship between HPV/P16 status and OS with Kaplan-Meier (KM) methods, both as a complete case analysis, and including missing as a covariate level.
Two-sided p-values < 0.05 were considered statistically significant. All analyses were performed with SAS 9.4 (The SAS Institute, Cary, NC).
The number of metastases, HN cancer site, histology, KPS, time between initial and DM diagnosis, systemic treatment, MDT, age, gender, CCI and local control were included in the multivariable model for 178 patients with complete data. The number of organs with metastases and metastases location were essentially measuring the same concept, so metastases location was not included in the multivariable model. In addition, due to the overlapping definitions of number of metastases with number of organs with metastases, only number of metastases was included in the multivariable model. After controlling for these covariates, the number of metastases was no longer significantly associated with OS (HR: 1.68, 95% CI: 0.84-3.40, p = 0.14). KPS, HN cancer site, time between initial and DM diagnosis and MDT remained significant predictors of OS as well; however, local control and chemotherapy receipt after DM were not significantly associated with OS in multivariable analyses (Table 3).
Factors associated with MDT We examined relevant clinical features of patients who did and did not receive MDT in Table 4. Patients with MDT had a higher KPS (median: 90, range: 70-100) compared with patients who did not have MDT (median: 80, range: 40-100, p = 0.001). In addition, a lower proportion of patients who received MDT had >1 metastases (40%, 12/30) compared with patients who did not receive MDT (96%, 149/156), p < 0.001. Seven MDT patients had two metastases, two patients had four metastases and three patients had five or more metastases. Only 17% of patients (5/30) who received MDT had more than one organ involved with metastases compared with 53% of patients (82/156) who did not receive MDT (p < 0.001).    (Fig. 4).

DISCUSSION
We found that patients who were selected for MDT for metastatic disease from LA-HNSCC had a significantly higher OS than patients who did not receive MDT. Patients with a single metastasis had greater OS and were most commonly selected for MDT, but the results from our analyses suggested that even patients with multiple metastases could derive clinically meaningful benefit from MDT. Below we discuss the relevant literature and highlight a number of points are necessary to consider when interpreting the findings of this report. Several other reports on the outcome of pulmonary metastastectomy for head and neck cancers have been published. The largest study from the Metastatic Lung Tumour Study Group of Japan registry database reported 114 patients with HSNCC who underwent pulmonary metastastectomy. They reported a 5-year OS of 26.5%, with OC primary, lymph node metastasis and male sex associated with worse survival on multivariable analysis. 11 An earlier report from Memorial Sloan Kettering Cancer Center was the largest institution-level report, with 41 patients undergoing pulmonary resection for HNSCC. 12 Similar to this study, HNSCC patients with MDT achieved long-term survival in~20% of cases, and long-term survival was achievable in patients with solitary and multiple metastases. Whereas that report evaluated other cancers of the head and neck including glandular cancers, this study focuses specifically on HNSCC in a more modern era of MDT techniques with more modern systemic therapy. Another study evaluated patients who developed pulmonary metastases after primary head and neck cancer therapy with no evidence of extrapulmonary metastases was carried out and attempted to compare the outcomes of surgical resection to best alternative chemotherapy or supportive care. 13 They reported a three-year OS of 68% in the surgical group of 24 patients compared with 15% in the non-surgical group of 45 patients. Of note they included non-SCC histologies (17%). While receipt of surgery was significant on multivariate analysis, the selection bias inherent to who undergoes MDT is challenging to completely control for with available clinical characteristics. As with our study, MDT patients were more likely to have a solitary metastasis. These and other previous studies evaluating the outcomes of oligometastasis in head and neck malignancies has recently been the subject of an excellent  14 In our study, patients with a single metastasis had significantly better OS when compared with two or more metastases in univariable analyses, and we did not find a significant difference in outcomes between patients with 2, 3, 4 or 5 + metastases. Patients with a single metastasis have been hypothesised to be the most likely to exhibit favourable metastatic biology which alone may explain superior survival compared to patients with more than one metastasis. 4,15,16 However, in our multivariable model controlling for receipt of MDT and the number of metastases, solitary versus multiple metastases was not associated with OS. This may be due to the fact that the majority of patients (18/25) with a solitary metastasis underwent MDT. However, patients with 2-4 metastases receiving MDT had a median SMFS of 11.2 months, and long-term disease-free survival was attained in some patients. Notably, MDT remained significantly associated with OS on multivariable analysis. As such, we propose that although our study lacks sufficient power to confirm statistically, well-selected patients with more than one metastasis may still benefit from MDT, and that the exceptional OS for patients with a solitary metastasis is in part related to their propensity to receive MDT. Other features associated with improved OS on multivariable analysis in our study include longer time between cancer diagnosis and DM and higher KPS at DM. Longer time to metastatic diagnosis has been reported in other studies to be associated with improved survival, 12,13,17,18 however, we are unaware of additional studies controlling for performance status. Patients with a longer time between initial local therapy and development of metastatic disease, particularly those who develop oligometastatic disease are likely to be those with the most favourable metastatic biology and are the patients who should be most strongly considered for MDT and considered as factors when designing trials for oligometastatic disease.
The degree to which definitive local therapy of metastatic disease impacts survival is difficult to address in a retrospective population due to obvious biases in patient selection for MDT. The majority of our patients with a single metastasis had MDT, which weakens a comparison between those with a single metastasis treated definitively and those who were not. However, our observed 10-year SMFS of 31% leads to the proposition that MDT represents an opportunity for cure of metastatic disease in well-selected patients with HNSCC, a prospect not expected with systemic therapy or supportive care. The converse finding that many patients develop subsequent metastases or die rapidly, 33% at 1 year and 54% at 3 years, highlights the fact that these patients who are clinically similar and deemed appropriate for MDT may have radically different outcomes after MDT. This likely reflects differing underlying metastatic biology and means to differentiate between these clinically oligometastatic patients with vastly different post-MDT outcomes are needed as we seek to personalise cancer care.
The majority of our patients were treated with surgery, which is an excellent modality for managing oligometastatic disease in certain locations such as the lung and has the advantage of permitting tissue diagnosis at the time of resection. The improvement in image-guided radiotherapy and increasing clinician comfort with ablative radiation therapy in managing both primary disease in the lung 19 and metastases 20,21 offers another approach to managing oligometastatic disease which is highly effective, non-invasive, well tolerated and not limited by sites which are less easily resected yet frequently sites of metastasis such as bone. Indeed, the highest level evidence available for MDT to date come from the SABR-COMET 9 and Gomez 8 Phase II trials. Importantly, these studies have both recently reported an OS benefit to MDT with ablative radiation.
The limitations of this study warrant serious consideration in the interpretation of these findings. There is great variability in our patient sample in terms of their initial disease status, disease status at metastasis, competing comorbidities and physician and personal preferences, all of which influence the selection of therapies in the setting of metastasis. We have attempted to control for these factors as rigorously as possible including validated indices of comorbidity such as CCI and KPS, but we do not propose that all bias is eliminated. Many patients received systemic therapies, and the study period runs from prior to cetuximab, which has been shown to have a survival benefit in metastatic HNSCC, 22 into the advent of immunotherapy, which has subsequently become a component of the standard of care for platinum-refractory disease. 23 The vast heterogeneity in systemic therapies, duration of treatment, and number of therapies prevents rigorous incorporation of the impact of these treatments into the current study.

CONCLUSIONS
The outcome of metastatic disease in HSNCC appears to be highly variable. The observation that some patients with apparently oligometastatic disease receiving MDT undergo rapid disease progression while others are cured highlights a spectrum of metastatic biology that has yet to be meaningfully defined. MDT offers an opportunity for long-term disease-free survival for select 1