End of Treatment Cone-Beam Computed Tomography (CBCT) is Predictive of Radiation Response and Overall Survival in Head and Neck Cancer

Background: Image guidance in radiation oncology has resulted in signicant improvements in the accuracy and precision of radiation therapy (RT). Recently, the resolution and quality of cone beam computed tomography (CBCT) for image guidance has increased so that tumor masses and lymph nodes are readily detectable and measurable. During treatment of head and neck squamous cell carcinoma (HNSCC), on-board CBCT setup imaging is routinely obtained; however, this CBCT imaging data has not been utilized to predict patient outcomes. Here, we analyzed whether changes in CBCT measurements obtained during a course of radiation therapy correlate with responses on routine 3-month follow-up diagnostic imaging and overall survival (OS). Materials/Methods: Patients with oropharyngeal primary tumors who received radiation therapy between 2015 and 2018 were included. Anatomical measurements were collected of the primary tumor and largest nodal conglomerate (LNC) at CT simulation, end of radiation treatment (EOT CBCT), and routine 3-month post-RT imaging. At each timepoint anteroposterior (AP), mediolateral (ML) and craniocaudal (CC) measurements were obtained and used to create a 2-dimensional (2D) maximum. Results: CBCT data from 64 node positive patients were analyzed. The largest nodal 2D maximum and CC measurements on EOT CBCT showed a statistically signicant correlation with complete response on 3-month post-RT imaging (r=0.313, p=0.02 and r=0.318, p=0.02, respectively). Furthermore, patients who experienced a 30% or greater reduction in the CC dimension had improved OS (continuous HR 1.07, 95% CI 1.02-1.15, p=0.045). Conclusion: Decreased size of pathologic lymph nodes measured using CBCT setup imaging during a radiation course correlates with long term therapeutic response and overall survival of HNSCC patients. These results indicate that CBCT setup imaging may have utility as an early predictor of treatment response in oropharyngeal HNSCC and could ultimately help guide management.


Introduction
Head and Neck Squamous Cell Carcinoma is a signi cant health issue with more than 600,000 new cases of HNSCC diagnosed worldwide and more than 60,000 new cases per year in the United States [1].
Radiation therapy (RT) plays an integral role in the curative treatment of HNSCC and radiation with concurrent chemotherapy remains the standard of care for locally advanced oropharyngeal HNSCC [2]. Despite advances in RT the survival rate for recurrent HNSCC remains low and treatment is associated with signi cant long-term toxicity [3,4]. Thus, additional strategies to personalize RT or tailor concurrent or adjuvant treatment are needed.
One of the major advances in RT over the past decade is the use of advanced imaging techniques and widespread adoption of image guided radiation therapy. Image guidance is a critical tool in radiation oncology which improves the accuracy and precision of radiation delivery. Image guidance can include plain lm kV imaging prior to treatment, cone beam computed tomography (CBCT), or even magnetic resonance imaging (MRI) during treatment. The improved accuracy and precision with image guided RT also permits smaller setup error margins and decreased treatment eld sizes which can reduce toxicity. During treatment of HNSCC, CBCT imaging is commonly obtained daily or weekly for setup veri cation and to monitor for changes in patient anatomy due to tumor response or weight loss. Over the past decade, innovations including Varian's Iterative CBCT have tremendously improved the resolution and quality of cone beam computed tomography (CBCT) such that discrete masses and lymph node conglomerates are readily discernable [5][6][7]. However, this large dataset of CT information has not been routinely used to monitor treatment responses largely because of the low resolution of CBCT in the past.
Interestingly, during a 7-week course of chemoradiation some patients have rapid complete clinical responses while other patients have little to no clinical response. Previous literature has documented measurable changes in disease volume over the course of radiotherapy for HNSCC [8,9]. Sanguinetti G., et al tested a nodal response model based on nodal density on the simulation scan with categorization of patients by the degree of shrinkage or growth throughout the treatment period. The volume and density of nodal disease was correlated to response on post-treatment imaging, although there was no analysis of impact on overall survival [9]. Similarly, Rosen et al., described signi cant changes in nodal volume during radiation therapy for head and neck cancer, particularly after the fourth week of therapy, but these were not correlated to treatment response or outcomes [10]. In head and neck cancer, CBCT has been utilized to measure volumetric parotid changes with resultant prediction of chronic xerostomia [10]. Additional studies have evaluated factors including the diameter change of the primary lesion and volumetric changes in nodal size utilizing CBCT but did not correlate these ndings to oncologic outcomes [11,12]. To date, despite this readily available source of CT data, studies have not assessed whether these change in lymph node measurements can predict treatment outcomes and survival.
The current gold standard to evaluate radiation treatment responses is CT and/or positron emission tomography (PET) imaging obtained several months post-treatment. PET/CT is a suitable method of disease evaluation with a negative predictive value > 90%, although false positive rates can exceed 50% [13][14][15][16][17][18]. Furthermore, in order to allow for post-treatment in ammation to subside the PET/CT often occurs 3-4 months after completing therapy which is quite a long time after treatment and off therapy, especially in cases of residual or progressive disease.
Here, we assessed whether changes in CBCT measurements obtained during the course of radiation treatment could be used as an early surrogate marker of treatment responses. We correlated responses on CBCT imaging with responses on 3-month routine diagnostic imaging and overall survival (OS) in patients with oropharyngeal HNSCC.

Data Source and Patient Selection
We identi ed 64 patients with squamous cell carcinoma of the oropharynx diagnosed between 2016 and 2018 with available follow up and treatment outcomes who received curative-intent radiation therapy at the University of California, San Diego. Patients must have completed de nitive therapy with radiation with or without chemotherapy. Patients were excluded if they received post-operative radiotherapy. Patients were considered post-operative if they had received any of the following: subtotal resection, gross total resection, or neck dissection (ipsilateral or bilateral). Additionally, patients were excluded if the course of radiotherapy was prolonged by greater than one week. The EOT CBCT measurements were reported as a percent change in size relative to CT Sim. The Post-RT measurements were reported as response based on RECIST 1.1 criteria.
Correlation of percent change in size seen on EOT CBCT to Post-RT response was evaluated utilizing Pearson correlation. The Post-RT response was dichotomized to complete response (CR, 'Responders') or less than CR ('Non-responders'), to include partial response (PR), stable disease (SD) or progressive disease (PD) thus creating a point biserial correlation coe cient. Logistic regression was utilized to derive the relationship between the percent change in size on EOT CBCT and response on Post-RT imaging. OS was examined using the KM method. Univariate survival analysis (UVA) was performed with the log-rank test and unadjusted Cox proportional hazards models to estimate hazard ratios (HR), with HR > 1 corresponding to worse OS and progression free survival (PFS).

Results
A total of 64 patients were included with a median follow up of 1.6 years (Range 0.7-3.7) and patient and treatment characteristics are presented in Table 1. The median patient age at diagnosis was 59.5 years (Range 39-87). Forty-two patients (65.6%) had stage I-II disease. The vast majority of patients were p16-positive, n = 58 (90.7%). A total of 57 patients received systemic therapy with regimens including cisplatin (n = 47), cisplatin plus pembrolizumab, cetuximab and carboplatin/paclitaxel. A total of 8 patients (12.5%) experienced a locoregional failure and 4 patients (6.3%) experienced distant metastasis including liver, lung and bone. Average time to failure was 0.56 years.

Characteristics of Complete Responders and Non-Complete Responders
On post-therapy imaging including PET/CT or CT Soft Tissue Neck, a total of 40 patients had experienced a CR, while 19 patients experienced PR and 5 patients had SD. There were no statistically signi cant differences in any patient characteristics between Complete responders and Non-Complete responders.
Of the patients who experienced a CR, 75% had stage I-II disease as compared to 50% in non-responders (p = 0. 19). Approximately 90% of both responders and non-responders were p16 positive. Representative examples of CR, PR and SD are shown in Fig. 1B, and patient characteristics according to treatment response are shown in Table 2.

Initial Response to Therapy
The median percent change in CC and 2D measurements of the LNC from CT Sim to EOT CBCT was 28.8% and 56.9% respectively in Complete responders compared to 7.7% and 38.2%, respectively, in Non-Complete responders ( Table 3). The percent change in both the 2D and CC measurements from CT Sim to EOT correlated to response on Post-RT imaging with r = 0.313, p = 0.02 and r = 0.318, p = 0.02, respectively.
Logistical regression of the percent change in CC measurement from CT Sim to EOT CBCT based on treatment response yielded an OR 1.05 (95% CI, 1.01-1.11) re ecting a 5% increase in CR for every 1% change from CT sim to EOT CBCT ( Fig. 2A-C). For patients who experienced 40% (n = 8) and 30% (n = 17) reduction in CC measurement from CT Sim to EOT CBCT, the rate of CR was 87% and 82% respectively.

Molecular Pro ling
Two patients who experienced progressive disease received guardant360 testing (Guardant Health, Inc. Redwood City, CA. The rst patient had notable mutations in PIK3CA, MAP2K2, APC and BRCA1. The second patient had a more extensive mutational burden including NF1, TP53, CHEK2, NOTCH2 and CCND1.

Discussion
Across the United States radiation oncologists will treat nearly 50,000 HNSCC patients with radiation therapy each year. During treatment, weekly or daily CBCT setup imaging data is acquired for each patient and the quality of this setup imaging continues to improve. However, this setup imaging data is not currently used for prognostication or therapeutic guidance. Here we analyzed CBCT imaging data from 64 patients and determined that changes in pathologically enlarged LN correlated with objective responses and patients with a ≥ 30% decrease in pathologically enlarged LN on the last day of radiation treatment had improved overall survival.
Interestingly, a prospective institutional study from Taipei Medical University utilized PET/CT during the 5th week of de nitive chemoradiation and reported a higher local recurrence for patients with higher midtreatment SUVm value [20]. This study contributes to a growing area of research into the utilization of radiation image guidance data to assess tumor response and outcomes. Previous studies have evaluated toxicity outcomes and tumor regression during the treatment period, though without correlation to overall survival [10,11]. Interestingly, a prospective institutional study from Taipei Medical University utilized PET/CT during the 5th week of de nitive chemoradiation and reported a higher local recurrence for patients with higher mid-treatment SUVm value [20]. Additionally, several ongoing trials are evaluating de-escalation of therapy by utilizing PET or CT imaging at designated time points during the treatment window including Memorial Sloan Kettering's Major Radiation Reduction for HPV + OPSCC (NCT03323463) and the Quarterback Trial (NCT01706939) at Mount Sinai. While insightful, PET/CT imaging during or immediately after a course of radiation therapy can be problematic to interpret due to edema, mucositis, and other in ammatory changes caused by radiation itself. The current standard of care used to determine therapeutic response is PET/CT or CT at 12-16 weeks post-therapy in conjunction with physical exam. Post-RT imaging is well-correlated to treatment outcomes with ndings ultimately guiding management such as neck dissection for patients with residual FDG avid neck disease [18]. Importantly PET imaging has a high negative predictive value of 90%, which, as demonstrated in the PET-NECK trial, can spare the patients the morbidity of neck dissection without compromising treatment outcomes [21,22]. However, 3-4 months is a substantial time to wait to determine therapeutic response and earlier assessments would help inform patients and potentially guide adjuvant therapy.
When we analyzed demographics from our 64 patients with oropharyngeal HNSCC, we observed the vast majority (90%) were p16+, consistent with rising rates of HPV positivity as well as low rates of smoking in our region (Table 1). While CBCT imaging was originally designed to help setup patients using soft tissue as well as boney anatomy, improved algorithms and detection arrays now permit distinct identi cation of tumor masses and LNs. Nevertheless, for oropharyngeal HNSCC we were unable to obtain precise measurements of the primary tumor site using CBCT due to the lack of contrast between the primary tumor and surrounding normal mucosal and soft tissues. However, we were able to obtain accurate and robust measurements of LN and nodal conglomerates due to contrast with surrounding fat planes and more uniform shapes and used this for our analyses. We observed that changes in LN CC dimension at the end of treatment CBCT had the strongest correlation with complete response rate with a median change of 28.8% from CT simulation to end of treatment in patients with CR versus only 7.7% in noncomplete responders. These data indicate that patients with rapid clinical responses to radiation are more likely to have complete responses at 3-month post-treatment PET/CT.
Of tremendous interest is the molecular basis for the radiographic ndings observed, which highlights the value of molecular pro ling. For example, mutations in the NOTCH pathway have been associated with poorer prognosis in HNSCC, which is consistent with the molecular pro le of a patient in our cohort who experienced local failure after therapy [23]. Similarly, PI3K family mutations, also seen in our cohort, have been associated with up to 60% of HNSCC and confer resistance to therapies, particularly EGFR inhibitors [24]. Unlike NOTCH, however, PI3K family mutations are potentially targetable representing an actionable mutation for further treatment. From RTOG 0129, we have long understood that HPVassociated HNSCC is a distinct entity from HPV-negative malignancies [25]. Recent literature including molecular studies from the University of Michigan have further divided the HPV-positive cohort by molecular and immunologic features including immune in ltration and gene expression that establish higher risk subgroups within HPV-associated HNSCC [26]. Ultimately, the next steps for precision radiation medicine involve correlation of molecular and radiomic ndings with treatment outcomes to guide therapeutic recommendations.
There are multiple limitations of this retrospective study. The use of CBCT imaging is a novel technique and standards for measurements of tumors and LN have not been established. CBCT is also prone to artifacts due to the wide incident cone-beam eld and reconstruction and ltering algorithms required. Additionally, there are patients who have a partial response or stable disease at end of treatment that go on to have a durable complete response. Furthermore, prospective studies would be needed to con rm whether patients with decreased size of pathologic LN at the end of radiation treatment have improved overall survival. Nevertheless, we believe that these results identify end of treatment CBCT as a novel imaging tool to identify rapid responders who are likely to manifest durable complete responses with improved outcomes.
Ultimately an early predictor of treatment response would be bene cial to help to guide management such as de-escalation or treatment intensi cation. This would be especially useful in light of multiple ongoing Phase III studies using checkpoint blockade immunotherapy either concurrently during chemoradiation or adjuvantly after completion of chemoradiation [27][28][29]. For example in the Phase III Paci c study consolidative anti-PD-L1 immunotherapy was administered to patients between 1-42 days after completion of chemoradiation for lung cancer. Given that the standard post-treatment imaging for HNSCC is routinely obtained around 90 days post-treatment, this creates a signi cant delay in the initiation of a potentially bene cial therapy. Additionally, because outcomes are relatively favorable for intermediate risk HPV + HNSCC patients additional treatments such as immunotherapy after completion of chemoradiation may be needed in some patients and unnecessary in others. When we analyzed survival in our patients, we found that those patients who manifested a ≥ 30% reduction in LN CC at the CBCT performed on the last day of radiation treatment had signi cantly improved overall survival (Fig. 3). Taken together, these ndings represent one of the rst examples correlating end of treatment CBCT imaging data with response to radiation therapy and overall survival of HNSCC patients.

Conclusion
The use of radiomics and imaging data to help guide radiation therapy and systemic management is evolving rapidly. The use of radiomics and imaging data to help guide radiation therapy and systemic management is evolving rapidly. In particular the use of on-board imaging to rapidly assess responses during treatment may be key to guiding further management and personalizing therapy. Our ndings indicate that CBCT at the end of treatment may have utility as an imaging modality for early identi cation of complete responders and strati cation of HNSCC patients with signi cantly improved overall survival. Future studies are needed to prospectively assess the ability of CBCT imaging data to determine responses and patient outcomes.  based on percent change of CC measurement from CT Sim to EOT CBCT. B) Scatter Plot demonstrating differential percent change between complete responders versus patients with partial responses or stable disease. C) Histogram demonstrating differential frequency of patients with 100% decrease at end of treatment between PR/SD versus CR. Patients were binned by each 10% change and strati ed by CR compared to PR and SD Figure 3 Reduced Lymph Node size at end of Radiation Treatment correlates with Overall Survival. A) Unadjusted Kaplan-Meier curve demonstrating overall survival for patients who experienced > 30% (red) and ≤ 30% reduction (blue) in the CC measurement from CT simulation to EOT CBCT (HR 1.07, 95% CI 1.02-1.15, p=0.045). Patients with a >30% decrease in lymph node size at end of treatment demonstrated improved overall survival.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. TABLESCBCT32021.pdf