Purpose:

In locally advanced p16+ oropharyngeal squamous cell carcinoma (OPSCC), (i) to investigate kinetics of human papillomavirus (HPV) circulating tumor DNA (ctDNA) and association with tumor progression after chemoradiation, and (ii) to compare the predictive value of ctDNA to imaging biomarkers of MRI and FDG-PET.

Experimental Design:

Serial blood samples were collected from patients with AJCC8 stage III OPSCC (n = 34) enrolled on a randomized trial: pretreatment; during chemoradiation at weeks 2, 4, and 7; and posttreatment. All patients also had dynamic-contrast-enhanced and diffusion-weighted MRI, as well as FDG-PET scans pre-chemoradiation and week 2 during chemoradiation. ctDNA values were analyzed for prediction of freedom from progression (FFP), and correlations with aggressive tumor subvolumes with low blood volume (TVLBV) and low apparent diffusion coefficient (TVLADC), and metabolic tumor volume (MTV) using Cox proportional hazards model and Spearman rank correlation.

Results:

Low pretreatment ctDNA and an early increase in ctDNA at week 2 compared with baseline were significantly associated with superior FFP (P < 0.02 and P < 0.05, respectively). At week 4 or 7, neither ctDNA counts nor clearance were significantly predictive of progression (P = 0.8). Pretreatment ctDNA values were significantly correlated with nodal TVLBV, TVLADC, and MTV pre-chemoradiation (P < 0.03), while the ctDNA values at week 2 were correlated with these imaging metrics in primary tumor. Multivariate analysis showed that ctDNA and the imaging metrics performed comparably to predict FFP.

Conclusions:

Early ctDNA kinetics during definitive chemoradiation may predict therapy response in stage III OPSCC.

Translational Relevance

The incidence of human papillomavirus (HPV)-associated oropharyngeal squamous cell carcinoma (OPSCC) is increasing. While patients with early-stage disease have high cure rates, patients will often present with locally advanced cT4 or N3 disease which is associated with a lower progression-free survival rate of 50%–70%. Real-time biomarkers such as HPV circulating tumor DNA (ctDNA) and/or imaging markers early in treatment could identify which of these poorer-prognosis patients may benefit from alternative or intensified therapy. In a prospective clinical trial cohort of patients with AJCC8 stage III p16+ OPSCC, we discovered that both HPV16/18 ctDNA and MRI/PET imaging markers independently predict tumor response to definitive chemoradiation as early as 2 weeks into therapy. These data support advancing these early biomarkers into a validation trial, which if successful, would offer the ability to detect response early in treatment and identify patients who may benefit from alternative therapies.

Over the past several decades, the incidence of oropharyngeal squamous cell carcinoma (OPSCC) has steadily increased in the United States (1, 2), which is attributable to increases in high-risk human papillomavirus (HPV)-driven disease (3, 4). In fact, because of the continually low rates of HPV vaccination as well as the time between HPV infection and the development of cancer (5), studies suggest that the overall incidence of HPV-positive (HPV+) OPSCC will continue to increase through the year 2030 (2, 6). Importantly, while patients with early-stage disease often have good clinical outcomes, patients with locoregionally advanced disease, such as the AJCC8 stage III disease that we study here, have worse overall survival. Novel clinical and personalized paradigms will be required to improve outcomes for these advanced-stage patients.

The prognosis of patients with head and neck cancers has traditionally been estimated on the basis of clinical stage with standard-of-care chemoradiation consisting of 70 Gy in 2-Gy fractions with concurrent cisplatin, and some studies have already started evaluating the use of biomarkers to improve patient selection strategies. For example, in HPV-related/p16+ OPSCC, imaging parameters such as pretreatment tumor volume (7), overt radiographic extracapular extension (8), and pretreatment and mid-treatment FDG-PET metrics including metabolic tumor volume (9–11) have been found to improve the accuracy of modeling for patterns of failure in addition to clinical characteristics. Furthermore, biologic markers also show promise to assess treatment response and provide complementary information with imaging (12, 13). HPV circulating tumor DNA (ctDNA) represents a minimally invasive biomarker with the potential to provide assessment of treatment response early in the treatment course. Emerging data in HPV+ recurrent or metastatic OPSCC show that serum HPV ctDNA levels correlate with total disease burden (14), and preliminary studies in the treatment deescalation setting have suggested that a favorable profile of high serum HPV ctDNA pretreatment or early in treatment followed by clearance by week 4 of chemoradiation may be associated with favorable outcomes (15). Importantly, while this 4-week timepoint may be useful for identifying patients with poor outcome, we believe that predicting outcome within 2 weeks of therapy initiation would enable providers to adapt and individualize therapy. Furthermore, work by our group and others has shown that repeated HPV ctDNA detection posttreatment is associated with high sensitivity and specificity for cancer recurrence (16, 17).

Given these recent advances in biomarkers for HPV+ OPSCC, we sought to evaluate the utility of both early 2-week imaging and HPV ctDNA for implementation in future adaptive medicine paradigms for patients with advanced-stage disease. If clinically validated, imaging and/or biologic markers could be used to personalize radiation dose or select patients for additional systemic therapy. Therefore, in this study, we sought to evaluate the predictive value of HPV ctDNA in the context of MRI and FDG-PET imaging pretreatment and mid-treatment biomarkers in patients with AJCC8 stage III p16+ OPSCC accrued to a prospective phase II randomized trial.

Patients

AJCC8 stage III p16+ OPSCC comprised a subset of patients enrolled on an Institutional Review Board (IRB)-approved randomized trial for definitive chemoradiation with concurrent weekly platinum (NCT02031250). The trial randomized p16-negative head and neck cancer as well as high-risk patients with p16+ OPSCC to standard chemoradiation (arm A) or dynamic-contrast-enhanced (DCE)-MRI–directed radiation boost to 80 Gy (arm B). The study was approved by the IRBs of all participating sites and was conducted in accordance with the Declaration of Helsinki, the Belmont Report, and U.S. Common Rule. Written informed consent was required from all participating patients prior to enrollment. Starting in 2016 with trial patient #40, we began to collect blood samples at pretreatment (baseline); during chemoradiation at weeks 2, 4, and 7; and then in follow-up at 3, 6, and 12 months post-radiation for planned exploratory HPV ctDNA analyses. All patients also had DCE and diffusion-weighted (DW) MRI, and FDG-PET scans pre-chemoradiation and during chemoradiation at week 2 (10 fractions of radiation). Patients were randomized to standard radiation to 70 Gy (arm A) or DCE-MRI–directed radiation boost to 80 Gy (arm B). The high-risk tumor subvolumes boosted consisted of low blood volume (LBV; poorly perfused subvolume) defined from DCE-MRI and low apparent diffusion coefficient (LADC; high cellular subvolume) defined from DW MRI.

Tumor volumes and imaging acquisition and analysis

MRI and PET scan acquisition and image processes have been described previously (18, 19). In brief, T1-weighted DCE volumetric MRI was acquired on a 3T scanner (Skyra, Siemens Healthineers) using a three-dimensional gradient echo sequence to cover head and neck anatomy from the base of skull to the top of shoulders to obtain 60 dynamic image volumes with a temporal resolution of approximately 3 seconds and a voxel size of approximately 2 × 2 × 2 mm. The extended Tofts model was used to quantify the fractional plasma volume that was converted to blood volume (BV). The DW images with b values of 50 and 800 mm2/second were acquired using a RESOLVE sequence to reduce geometrics distortion, obtain spatial resolution of 2 × 2 × 4.8 mm, and cover primary and nodal tumors. ADC maps were calculated using the two b-value images.

The gross tumor volume (GTV) of each primary or nodal tumor was contoured on post-Gd T1-weighted images by the treating attending radiation oncologist and reviewed by trial principal investigator (PI; M. Mierzwa). The BV and ADC in each GTV were thresholded at <7.64 mL/(100 g/min) and 1,200 mm2/second to obtain the subvolumes of LBV and LADC, respectively. Metabolic tumor volume at 50% of maximum standardized uptake value (SUV; MTV50) of each tumor was calculated after registering to post-Gd T1-weighted images. The mean BV and mean ADC in each GTV and the mean SUV in each MTV50 were calculated.

Absolute counts of ctDNA pre-chemoradiation and relative counts during and post chemoradiation were correlated to previously demonstrated predictive image biomarkers such as tumor subvolumes of LBV and LADC and MTV50 as well as mean BV, mean ADC, and mean SUV of FDG in tumor volumes (20).

Plasma HPV16/18 ctDNA analysis

Plasma cell-free DNA (cfDNA) isolation from Streck tubes was carried out per manufacturer recommendations from a 1 mL aliquot of plasma, which was processed using QIAamp MinElute ccfDNA Mini Kit (Qiagen) and eluted in 40 μL buffer. Genomic DNA from cell lines UM-SCC-104 (HPV16 positive) and UM-SCC-105 (HPV18 positive) was extracted using the Promega Wizard DNA Purification Kit protocol (Promega A1120; ref. 21). Droplet digital PCR was then performed for HPV ctDNA analysis using our analytically validated assay as described previously (22) and detailed in the Supplementary Materials and Methods. Similarly, an HPV18 E6 assay was also utilized with forward primer 5′-GGAGACACATTGGAAAAACTAACTAACAC-3′, reverse primer 5′-CTGCTGGATTCAACGGTTTCTG-3′, and probe 5′-FAM-AATAAGGTGCCTGCGGTGCCAGAAA-MGBNFQ-3′. A ddPCR assay for the RPP30 control gene was used to assess sample quality (with HEX probe, Assay ID: dHsaCP2500350, Bio-Rad). We have previously tested over 200,000 diploid genome equivalents of human genomic DNA as a non-HPV template and did not observe any positive reactions (22) supporting the high specificity of this assay.

Targeted DNA sequencing and host mutation calling

Blocks or sections were requested from patients with evaluable ctDNA for next-generation sequencing (NGS)-based tumor analysis. Regions with >60% tumor content as identified by our head and neck pathologist (J.B. McHugh) were identified for DNA isolation as described previously (23, 24) using tumor and adjacent normal tissue cores and the Qiagen Allprep DNA/RNA FFPE kit (Qiagen). In the case of UM041 and UM060, whole blood was available and used in place of adjacent normal as reference material using protocols as described previously (25). Isolated DNA was quantified using a Qubit as described previously (26). Of the 28 patients positive for HPV ctDNA in the study, 12 had excess biopsy material available in our archives and all 12 had sufficient genetic material for targeted NGS analysis. Targeted capture sequencing on DNA that passed our quality control standards was performed by the University of Michigan Advanced Genomics Core for using the DNA Thruplex kit for library preparation (Takara Biosciences). Using a custom-designed probe panel from Nextera, targeted capture was performed with high-density probes covering the HPV genome, and also included probes for head and neck carcinoma–related genes, known actionable gene and several common cancer-related genes (total 226 gene targets total) as described in Heft Neal and colleagues (27). Following library preparation and capture, the samples were sequenced on an Illumina NOVASeq6000 using a 300 cycle run and FastQ files were archived. Somatic alteration analysis was performed as in Heft Neal and colleagues and is fully detailed in online supporting material. Briefly, Varscan v2.4.1 was used to call variants from these mpileup files using the somatic mode of the variant caller. Variant calls were annotated using Goldex Helix Varseq v2.1.0. This was followed by filtering the variants in the introns and intergenic regions. Variants with a minimum of 5 reads supporting the alternate allele in the tumor samples were considered as potential positives.

HPV16 mutation and INDEL analysis

The quality trimmed reads were aligned to the HPV16 genome using bowtie2/2.4.1. Samtools v1.9 was used to sort and index the mapped reads and picard-tools v2.8.1 was used to mark duplicates. Next, pileup files were created from the sorted and deduplicated BAM files using the “mpileup” function of samtools. Variants were called from these pileup files using the germline mode of Varscan v2.4.1. “mpileup2snp” was used to call SNPs whereas “mpileup2indel” was used to call insertions and deletions (INDEL).

HPV16 integration analysis

HPV integrations were called by SearcHPV (28), an HPV integration caller that we recently developed for targeted capturing data. The downstream analysis was performed with R 3.6.1 and Python.

Statistical analysis

Predictive values of ctDNA for the freedom from progression (FFP), including local, regional, and distant progression were analyzed using Kaplan–Meier method and Cox proportional hazards models with adjustment to account for the effect of radiation boost on locoregional control. Particularly, the absolute values pre-chemoradiation, the early increase or decrease at week 2 as a binary variable, and the relative counts or clearance at week 4 were tested. The time for progression or death was calculated from the start of radiation to the time of progression, death, or last follow-up. In the analysis of FFP, death and last follow-up were censored. Correlations were evaluated between absolute/relative ctDNA vales at different timepoints with previously established imaging biomarkers including GTV, subvolumes of LBV and LADC, MTV50, mean BV, mean ADC, and mean SUV in primary and nodal tumors pre-chemoradiation and at week 2 (20); these were tested by Spearman rank correlation with false discovery control (FDC). Finally, the absolute and relative ctDNA values that predicted FFP in univariate analysis were tested with image predictive biomarkers for complimentary effects for prediction of FFP using multivariable Cox models. P < 0.05 was considered as significant.

Data were generated by the authors and available on request. FastQ files related to this study have been made available through the Sequencing Read Archives, study ID: PRJNA771858 (https://www.ncbi.nlm.nih.gov/Traces/study/?acc=PRJNA771858).

Ethics approval and consent to participate

A prospective cohort was consented to a University of Michigan IRB-approved clinical trial (UMCC2013.062).

Availability of data and material

All ctDNA data are available by request from J.C. Brenner and imaging data are available from Y. Cao. FastQ files related to this study have been made available through the Sequencing Read Archives, study ID: PRJNA771858 (https://www.ncbi.nlm.nih.gov/Traces/study/?acc=PRJNA771858).

Patient characteristics

A total of 131 blood samples were analyzed on 34 patients with stage III AJCC8 p16+ OPSCC (Supplementary Table S1). Six of 34 (17.6%) patients had nondetectable HPV DNA at baseline, leaving 28 patients with analyzable ctDNA. Of note, the 6 patients with nondetectable baseline ctDNA all tested positive for high-risk HPV by our clinical ISH assay. Patient characteristics are given in Table 1. All patients in the analysis had stage III AJCC8 p16+ OPSCC. Among the patients with detectable baseline ctDNA, median age was 63 years old and 92% were male. Seventeen patients had 70 Gy total radiation dose and 11 had radiation boost to the high-risk tumor subvolumes of LBV and LADC of 80 Gy. The median primary tumor volume pre-chemoradiation was 36.6 cm3 (range: 9.1–154 cm3), which was greater than the median total nodal tumor volume of 15.7 cm3 (range: 0–274 cm3).

Table 1.

Patient characteristics.

N34 total patients with available blood samples28 patients with analyzable ctDNA
Age (years) 
 Median (range) 64 (47–79) 63 (47–79) 
Sex 
 Female 
 Male 31 26 
T stage 
 T4 29 25 
 T3 
 T0 
N stage 
 N3 
 N2 
 N1 28 23 
 N0 
Radiation dose 
 70 Gy 18 17 
 80 Gy 16 11 
Chemotherapy 
 Cisplatin 13 10 
 Carboplatin 21 18 
Smoking status 
 <10 pack-year 14 11 
 ≥10 pack-year 20 17 
Progression 
 D/RD  3/1 
 L/LR  4/1 
N34 total patients with available blood samples28 patients with analyzable ctDNA
Age (years) 
 Median (range) 64 (47–79) 63 (47–79) 
Sex 
 Female 
 Male 31 26 
T stage 
 T4 29 25 
 T3 
 T0 
N stage 
 N3 
 N2 
 N1 28 23 
 N0 
Radiation dose 
 70 Gy 18 17 
 80 Gy 16 11 
Chemotherapy 
 Cisplatin 13 10 
 Carboplatin 21 18 
Smoking status 
 <10 pack-year 14 11 
 ≥10 pack-year 20 17 
Progression 
 D/RD  3/1 
 L/LR  4/1 

Abbreviations: D, distant; L, local; LR, locoregional; RD, regional distant.

With a mean follow-up of 28 months (range, 12–49 months), patterns of failure are as follows: among patients with nondetectable baseline ctDNA, 3 of 6 patients have had treatment failure: 2 patients with persistent locoregional disease at the 3-month PET and one distant failure (DF) also noted at 3-month post-radiation PET. Among the 28 patients with analyzable ctDNA, 4 had local failure (LF), 1 had locoregional failure (LRF), 3 had DF, and 1 had regional and distant failure. Four of the 5 with LF or LRF had no radiation boost.

Temporal profiles of plasma ctDNA

Pretreatment absolute ctDNA values demonstrated high interpatient variability with a median number of 460 copies per mL (range, 0–34,714 copies/mL). Figure 1 illustrates normalized ctDNA profiles over time. At week 2 (10.5 ± 4.3 days from the radiation day 1), all patients had a change in ctDNA values compared with baseline (Fig. 1) and their changes were sufficiently large to divide patients into two subgroups based on change at 2 weeks from pretreatment. The median increase was 152% (range, 31%–992%) and the median decrease was 66% (range, 35%–100%). In the early increase group, there was only 1 patient who had LF at 14.8 months and had no radiation boost. In the early decrease group, 7 of the 15 patients had LRF or DF.

Figure 1.

A, Time profiles of relative ctDNA values over 12 months from the start of chemoradiation therapy normalized to maximum values in the first 2 weeks. There are two distinct time profiles in the early response to chemoradiation: an early increase in the ctDNA counts (green) versus a decrease (blue). B, Kaplan–Meier curves of freedom from progression. Freedom from progression between subgroups with an early increase in ctDNA values (green dashed line) during the early course of chemoradiation versus a decrease (blue solid line) were significantly different, P < 0.02.

Figure 1.

A, Time profiles of relative ctDNA values over 12 months from the start of chemoradiation therapy normalized to maximum values in the first 2 weeks. There are two distinct time profiles in the early response to chemoradiation: an early increase in the ctDNA counts (green) versus a decrease (blue). B, Kaplan–Meier curves of freedom from progression. Freedom from progression between subgroups with an early increase in ctDNA values (green dashed line) during the early course of chemoradiation versus a decrease (blue solid line) were significantly different, P < 0.02.

Close modal

At week 4 (24.0 ± 3.9 days), ctDNA values decreased from the pretreatment or week 2 maximum except 2 patients who had an increase in the counts but no disease progression. Ten patients reached nondetectable ctDNA levels by week 4 and 3 of these patients had DF (N = 2) or LF (N = 1). Notably, 8 of these 10 patients had early decrease at week 2.

At week 7 and 3 months post-radiation, only 4 patients had non-zero counts. The time profiles of relative counts of ctDNA normalized to maximum values either at baseline or week 2 are shown in Fig. 1.

At week 7 of chemoradiation, 4 patients had detectable ctDNA: 2 went on to clear at 3 months post-radiation and remain NED (no evidence of disease), and 2 remained persistent at 3 months (1 with radiographic regional and distant progression on 3-month post-radiation PET, the other never developed radiographic disease but was placed in hospice due to medical comorbities 6 months after chemoradiation). At 3 months, 4 patients had detectable ctDNA. Two of the 4 are described above (1 with radiographic regional + distant progression at 3-month post-radiation PET, the other never developed radiographic disease but was placed in hospice due to medical comorbities 6 months after chemoradiation). The other 2 patients were non-detectable at week 7, but changed to detectable at 3 months and both had radiographic and biopsy-proven disease at 3-month post-radiation PET (1 local and 1 distant).

Predictive values of plasma ctDNA for progression

Pretreatment absolute value of ctDNA was a significant predictor for FFP [HR = 1.06; 95% confidence interval (CI), 1.01–1.12, per 1,000 copies/mL; P < 0.03] with higher values associated with tumor progression within 12 months of therapy. At week 2, patients with an increase in ctDNA compared with baseline had significantly less tumor progression [HR = 0.11; 95% CI, 0.01–0.95; P < 0.05]. After adjusting for the boost effect as a covariable in the model, pretreatment absolute value of ctDNA was not significant with P = 0.056, but the early increase in ctDNA at week 2 was still significant with P < 0.05. Furthermore, Kaplan–Meier analysis showed that the early increase versus decrease subgroups had significantly different FFP with a relative decrease at week 2 associated with increased failure (P < 0.02, χ2 = 5.67; see Fig. 1). At week 4, relative ctDNA counts normalized to pre-chemoradiation or patient's maximum were not significantly predictive of progression (HR = 0.99; 95% CI, 0.99–1.08, per 10% increase; P = 0.8). In addition, assessment of 95% ctDNA clearance at week 4 did not predict FFP (HR = 0.73; 95% CI, 0.15–3.61; P = 0.7).

Patient tumor molecular characteristics

To then evaluate HPV status in these tumors and the sequence of our ctDNA binding site, we used the HPViewer pipeline, which found an average of 691,411 reads in HPV16-positive tumors and identified one tumor each with HPV18 (UM071) and HPV33 (UM074; Supplementary Table S2). Tumor mutational status is reported in Supplementary Fig. S1A and Supplementary Tables S3 and S4. Interestingly, of the two primary and recurrence pairs in the cohort, we noted that one tumor had an amplification of HPV in the recurrence relative the primary, while HPV read counts were unchanged between the second primary and recurrence pair (Fig. 2A). We then adapted our germline alteration pipeline to call alterations in the HPV16 genome, this identified a total of 578 polymorphisms (range, 0–143 per tumor; Supplementary Table S5) and 25 total INDELs (range: 0–5 per tumor; Supplementary Table S6). Importantly, none of the viral genomes contained any SNPs or INDELs in our ctDNA primer/probe binding site, and we note that four of tumors had two SNPs each in the NavDx binding site, though it is unclear whether these would alter probe efficiency.

Figure 2.

Molecular characteristics of tumors in the cohort. A, HPViewer was used to quantify normalized HPV read counts in each sample; primary and recurrence pairs are shown. B, Plot shows distribution of all identified HPV to host genome integration sites identified in the cohort. Genomic coordinates are shown along the outside as indicated. C and D, Plot shows comparison of HPV integration sites identified in the primary (inner circle) and metastatic (outer circle) pair.

Figure 2.

Molecular characteristics of tumors in the cohort. A, HPViewer was used to quantify normalized HPV read counts in each sample; primary and recurrence pairs are shown. B, Plot shows distribution of all identified HPV to host genome integration sites identified in the cohort. Genomic coordinates are shown along the outside as indicated. C and D, Plot shows comparison of HPV integration sites identified in the primary (inner circle) and metastatic (outer circle) pair.

Close modal

To further characterize the impact of HPV on these genomes, we then used an optimized viral integration assembler that we recently developed to identify HPV16 integration events for the 12 formalin-fixed paraffin-embedded (FFPE) patients that had sufficient genetic material for targeted NGS analysis. We characterized 335 integration events in total for all types of tissue. Among them 234 integrations were called from 11 primary tumor samples, 99 were called from three recurrent tumor samples, and 2 were called from 12 adjacent normal tissue samples (Fig. 2B; Supplementary Table S7). About 53% integration sites (177/335) fell into known genes. Four integrations from one primary tumor sample clustered within/near SRCRB4D. We further investigated more on the 2 patients that had both primary tumor and recurrent tumor sequenced (Fig. 2C and D). By comparing the loci of integrations in primary and recurrent tumor, we found 1 patient had four integration sites that occurred identically or 500 bp nearby each other in both primary and recurrent tumors. Given the large number of integration sites detected in these tumors, we assessed the local homology at the breakpoints between host and viral sequences to evaluate possible mechanisms of DNA repair-mediated integration, which suggested that the integration events occurred through a microhomology-mediated DNA repair pathway (Supplementary Fig. S1B). Finally, we did not observe correlations of HPV integration site number with baseline ctDNA values or treatment-related ctDNA kinetics (data not shown).

Correlation with clinical factors and predictive imaging biomarkers

Clinical variables were uniform in this cohort of the patients with AJCC8 stage III p16+ OPSCC. A total of 89% of patients with analyzable ctDNA had cT4 tumor, and therefore smoking status was the only clinical variable included in models to avoid overfitting. The ctDNA values at baseline, week 2, and week 4 during radiation were not correlated with smoking status as a binary variable (<10 pack-years vs. ≥10 pack-years) with P > 0.1. At week 7, the ctDNA clearance was correlated with smoking status (P = 0.05), in which 93.8% of non-smokers and 62.5% of smokers had ctDNA clearance. However, the smoking status did not predict FFP in this cohort of patients (P > 0.9).

Pretreatment absolute values of ctDNA were significantly correlated with nodal GTV, nodal subvolumes of LBV and LADC, and nodal MTV50 pre-chemoradiation and at week 2 (P < 0.05 with FDC) except the subvolume of LADC at week 2. We saw no correlation between pretreatment ctDNA and any of the tumor volumes or subvolumes in primary tumors at either pre-chemoradiation or week 2 (P > 0.3 with FDC), suggesting that the more cellularly dense nodal burden may dominate the shedding of tumor DNA into plasma before the initiation of treatment.

After starting chemoradiation, the ctDNA values at week 2 were significantly and negatively correlated with primary GTVs pre-chemoradiation and at week 2 and primary MTV50 pre-chemoradiation (P < 0.03 with FDC). We saw no correlation with other imaging metrics, including any nodal metrics (P > 0.1 without FDC), suggesting that the contribution of primary tumors to shedding DNA to plasma during the early course of chemoradiation is smaller for larger primary GTVs, possibly due to more necrosis and/or less response to treatment. Interestingly, in the first 2 weeks of treatment, we saw similar response rates between nodal and primary tumors by CT, DCE-MRI, and FDG-PET (11).

At week 4, the relative counts of ctDNA to pre-chemoradiation were significantly and negatively correlated with mean BV and mean SUV in the primary GTVs pre-chemoradiation (P < 0.05 with FDC). We saw no other significant correlations with other image metrics, including all nodal imaging metrics (see Table 2).

Table 2.

Correlations of ctDNA (copies/mL) values (vertical left column) with image metrics (horizontal rows).

Pre-CRT absolute ctDNA valuesTotal nGTVTotal nSubV LBVTotal nSubV LADCTotal nMTV50Total nGTVTotal nSubV LBVTotal nSubV LADCTotal nMTV50
Pre-CRTPre-CRTPre-CRTPre-CRTWeek 2Week 2Week 2Week 2
rs 0.54 0.46 0.49 0.58 0.49 0.40 0.34 0.53 
P wo FDC 0.004 0.02 0.01 0.004 0.01 0.04 0.09 0.009 
P w FDC 0.016a 0.024a 0.016a 0.016a 0.016a 0.046a 0.09 0.016a 
Pre-CRT absolute ctDNA valuesTotal nGTVTotal nSubV LBVTotal nSubV LADCTotal nMTV50Total nGTVTotal nSubV LBVTotal nSubV LADCTotal nMTV50
Pre-CRTPre-CRTPre-CRTPre-CRTWeek 2Week 2Week 2Week 2
rs 0.54 0.46 0.49 0.58 0.49 0.40 0.34 0.53 
P wo FDC 0.004 0.02 0.01 0.004 0.01 0.04 0.09 0.009 
P w FDC 0.016a 0.024a 0.016a 0.016a 0.016a 0.046a 0.09 0.016a 
Week 2 relative ctDNA valuespGTVpSubV LBVpSubV LADCpMTV50pGTVpSubV LBVpSubV LADCpMTV50
Pre-CRTPre-CRTPre-CRTPre-CRTWeek 2Week 2Week 2Week 2
rs −0.56   −0.58 −0.54   −0.47 
P wo FDC 0.007 0.9 0.6 0.009 0.009 0.7 0.3 0.04 
P w FDC 0.024a 0.9 0.8 0.024a 0.024a 0.8 0.5 0.07 
Week 2 relative ctDNA valuespGTVpSubV LBVpSubV LADCpMTV50pGTVpSubV LBVpSubV LADCpMTV50
Pre-CRTPre-CRTPre-CRTPre-CRTWeek 2Week 2Week 2Week 2
rs −0.56   −0.58 −0.54   −0.47 
P wo FDC 0.007 0.9 0.6 0.009 0.009 0.7 0.3 0.04 
P w FDC 0.024a 0.9 0.8 0.024a 0.024a 0.8 0.5 0.07 
Week 4pMean BVpMean ADCpMean SUVpMean BVpMean ADCpMean SUV
relative ctDNA valuesPre-CRTPre-CRTPre-CRTWeek 2Week 2Week 2
rs −0.49 −0.37 −0.52  −0.42 −0.36 −0.28  
P wo FDC 0.02 0.08 0.02  0.04 0.08 0.2  
P w FDC 0.048a 0.1 0.048a  0.08 0.1 0.2  
Week 4pMean BVpMean ADCpMean SUVpMean BVpMean ADCpMean SUV
relative ctDNA valuesPre-CRTPre-CRTPre-CRTWeek 2Week 2Week 2
rs −0.49 −0.37 −0.52  −0.42 −0.36 −0.28  
P wo FDC 0.02 0.08 0.02  0.04 0.08 0.2  
P w FDC 0.048a 0.1 0.048a  0.08 0.1 0.2  

Note: Note that pre-chemoradiation absolute ctDNA values are highly correlated with several total nodal imaging metrics while week 2 and week 4 ctDNA values are correlated with several primary tumor imaging metrics.

Abbreviations: CRT, chemoradiation; FDC, false discovery control; GTV, gross tumor volume; LADC, low apparent diffusion coefficient; LBV, low blood volume; MTV50, metabolic tumor volume of FDG at 50% of maximum SUV; “p”, primary tumor; rs, correlation coefficient by Spearman rank correlation; “Total n”, total nodal tumor volume; SubV, subvolume of tumor.

aP value with FDC < 0.05 was considered significant.

Additive values of plasma ctDNA and imaging biomarkers for prediction of tumor progression

As reported previously, large primary tumor subvolumes with LBV and LADC, and large MTV50 were associated with tumor progression (20). In this cohort of patients, we analyzed association of FFP with total subvolume of LBV or LADC, or total MTV50 of all tumors, including both primary and nodal tumors. Univariate Cox models showed that (i) the patients with large total subvolumes of LBV pre-chemoradiation had inferior FFP compared with those with small subvolumes [dichotomized by the median value of 14 cm3 of the group, with HR = 5.37 (95% CI, 1.10–26.20) and P < 0.04]; (ii) the patients with large total subvolumes of LADC pre-chemoradiation had inferior FFP compared with those with small subvolumes [dichotomized by the median value of 16 cm3 of the group, with HR = 12.20 (95% CI, 1.51–98.9) and P < 0.02]; and (iii) the patients with large total MTV50 pre-chemoradiation and at week 2 had inferior FFP compared with those with the small MTV50, with HR = 1.06 (95% CI, 1.01–1.10) per 1 cm3 and P < 0.008 and HR = 1.12 (95% CI, 1.05–1.19) per 1 cm3 and P < 0.0006, respectively. After adjusting for the effect of radiation boost, the total subvolume of LBV pre-chemoradiation, the total subvolume of LADC pre-chemoradiation, and the total MTV50 pre-chemoradiation and at week 2 were still significant (P < 0.03, < 0.02, < 0.03, and < 0.002, respectively). Kaplan–Meier curves of these predictive imaging metrics are shown in Fig. 3. In this cohort, total GTV was not significantly associated with tumor progression.

Figure 3.

Examples of predictive imaging metrics from DCE and DW MRI and FDG-PET (A–C) and corresponding Kaplan–Meier curves of freedom from progression (D–F). A, Examples of patients with small (left) and large (right) low blood volume subvolume (TVLBV) on DCE-MRI imaging; D, FFP between subgroups with a lower TVLBV in total tumor volumes pre-chemoradiation (dashed line) versus a higher (solid line) were significantly different, P < 0.02. B, Examples of patients with small (left) and large (right) low apparent diffusion coefficient tumor subvolume (TVLADC) on DW-MRI; E, FFP between subgroups with a lower TVLADC in total tumor volume (dashed line) versus a higher (solid line) were significantly different, P < 0.003. C, Examples of patients with small (left) and large (right) MTV50% on FDG-PET; F, FFP between subgroups with a lower MTV50 (dashed line) of total tumor volume at week 2 versus a higher (solid line) were significantly different, P < 0.005.

Figure 3.

Examples of predictive imaging metrics from DCE and DW MRI and FDG-PET (A–C) and corresponding Kaplan–Meier curves of freedom from progression (D–F). A, Examples of patients with small (left) and large (right) low blood volume subvolume (TVLBV) on DCE-MRI imaging; D, FFP between subgroups with a lower TVLBV in total tumor volumes pre-chemoradiation (dashed line) versus a higher (solid line) were significantly different, P < 0.02. B, Examples of patients with small (left) and large (right) low apparent diffusion coefficient tumor subvolume (TVLADC) on DW-MRI; E, FFP between subgroups with a lower TVLADC in total tumor volume (dashed line) versus a higher (solid line) were significantly different, P < 0.003. C, Examples of patients with small (left) and large (right) MTV50% on FDG-PET; F, FFP between subgroups with a lower MTV50 (dashed line) of total tumor volume at week 2 versus a higher (solid line) were significantly different, P < 0.005.

Close modal

When the ctDNA baseline values or the early increase versus decrease at week 2 of treatment was paired with one of the three predictive imaging biomarkers above and tested for relative prediction of FFP in stepwise multivariable models, only total MTV50 pre-chemoradiation and at 2 weeks were significant with P < 0.03, suggesting that total MTV50 was a stronger predictor for progression than ctDNA absolute value at baseline or 2 weeks in this cohort (see Table 3).

Table 3.

Multivariate Cox models.

ModelctDNA parameterImaging parameter
baseline ctDNA value (per 1,000 copies) total SubV LBV pre (binary) 
 HR = 1.04 (0.98–1.10), P = 0.2 HR = 3.22 (0.57–18.27), P = 0.2 
baseline ctDNA value (per 1,000 copies) total subV LADC pre (binary) 
 HR = 1.03 (0.97–1.08), P = 0.3 HR = 1.00, P = 0.96 
baseline ctDNA value (per 1,000 copies) total MTV50 pre (per cm3
 HR = 0.94 (0.84–1.05), P = 0.3 HR = 1.11 (1.01–1.23), P = 0.03a 
baseline ctDNA value (per 1,000 copies) total MTV50 week 2 (per cm3
 HR = 0.94 (0.84–1.05), P = 0.3 HR = 1.20 (1.03–1.40), P = 0.02a 
increase vs. decrease in ctDNA values at week 2 total SubV LBV pre (binary) 
 HR = 0.14 (0.02–1.20), P = 0.07 HR = 3.28 (0.63–17.07), P = 0.2 
increase vs. decrease in ctDNA values at week 2 total subV LADC pre (binary) 
 HR = 1.00, P = 0.96 HR = 1.00, P = 0.96 
increase vs. decrease in ctDNA values at week 2 total MTV50 pre (per cm3
 HR = 0.21 (0.02–2.27), P = 0.2 HR = 1.04 (0.99–1.10), P = 0.1 
increase vs. decrease in ctDNA values at week 2 total MTV50 week 2 (per cm3
 HR = 0.21 (0.02–2.27), P = 0.2 HR = 1.10 (1.02–2.19), P = 0.02a 
ModelctDNA parameterImaging parameter
baseline ctDNA value (per 1,000 copies) total SubV LBV pre (binary) 
 HR = 1.04 (0.98–1.10), P = 0.2 HR = 3.22 (0.57–18.27), P = 0.2 
baseline ctDNA value (per 1,000 copies) total subV LADC pre (binary) 
 HR = 1.03 (0.97–1.08), P = 0.3 HR = 1.00, P = 0.96 
baseline ctDNA value (per 1,000 copies) total MTV50 pre (per cm3
 HR = 0.94 (0.84–1.05), P = 0.3 HR = 1.11 (1.01–1.23), P = 0.03a 
baseline ctDNA value (per 1,000 copies) total MTV50 week 2 (per cm3
 HR = 0.94 (0.84–1.05), P = 0.3 HR = 1.20 (1.03–1.40), P = 0.02a 
increase vs. decrease in ctDNA values at week 2 total SubV LBV pre (binary) 
 HR = 0.14 (0.02–1.20), P = 0.07 HR = 3.28 (0.63–17.07), P = 0.2 
increase vs. decrease in ctDNA values at week 2 total subV LADC pre (binary) 
 HR = 1.00, P = 0.96 HR = 1.00, P = 0.96 
increase vs. decrease in ctDNA values at week 2 total MTV50 pre (per cm3
 HR = 0.21 (0.02–2.27), P = 0.2 HR = 1.04 (0.99–1.10), P = 0.1 
increase vs. decrease in ctDNA values at week 2 total MTV50 week 2 (per cm3
 HR = 0.21 (0.02–2.27), P = 0.2 HR = 1.10 (1.02–2.19), P = 0.02a 

Note: The ctDNA values and the imaging metrics that were significant for FFP in univariate Cox models were tested in the multivariate Cox models as a pair due to the small number of patients.

Abbreviations: LADC, low apparent diffusion coefficent; LBV, low blood volume; MTV, metabolic tumor volume; subV, subvolume.

aSignificant P value.

In the current study, we found that the baseline ctDNA values and its early kinetics within the first 2 weeks of treatment were predictive of tumor recurrence. Particularly, higher baseline HPV ctDNA and a decreased ctDNA at week 2 of chemoradiation relative to pretreatment levels were associated with high risk for treatment failure/tumor progression within 12 months post-radiation. Clearance of ctDNA at weeks 4 or 7 of therapy were not predictive of progression. The positive correlations between pretreatment ctDNA values and nodal tumor volumes and the subvolumes identified by blood volume and diffusion MRI and FDG-PET biomarkers suggest that ctDNA shedding pretreatment may be highly dependent on nodal metastases. The negative correlations between the relative values of ctDNA at week 2 of chemoradiation and the GTV and MTV of primary tumors suggest that the primary tumor response to therapy may play a larger role in ctDNA shedding early in treatment. The multivariate analysis suggests that the ctDNA and the imaging biomarkers might not provide complimentary predictive value for tumor progression but could be utilized differently as ctDNA for prediction of overall tumor progression and imaging biomarkers for specific progression patterns (e.g., local vs. distant progression) as well as for the patients with undetectable ctDNA.

Recent studies in p16+ HPV-related OPSCC have begun to suggest that early ctDNA release could reflect tumor cell death and predict a favorable treatment outcome (29, 30). A previous study of HPV ctDNA in the deescalation setting of p16+ OPSCC found lower plasma levels (≤200 copies/mL) in higher-risk patients such as T4 (n = 7) or >10 pack-years (n = 26). They identified a favorable risk profile of high pretreatment ctDNA with >95% clearance by day 28 (15). Possibly due to low events, the investigators were not able to determine whether the profile was independent of T-, N-category, or smoking status. Here, notably, our uniform stage III patient group, largely consisting of T4 tumors, had a high failure rate with 11 of 34 (32%) overall and 9 of 28 (29%) in ctDNA analyzable patients, compared with only 7 failures (10%) occurring in 67 patients in the Chera and colleagues experience. We saw that early ctDNA kinetics were predictive of outcomes and did not see any predictive value of ctDNA clearance at week 4. This suggests that a formal clinical trial should be performed to evaluate the potential clinical utility of evaluating ctDNA clearance at different timepoints. We furthermore saw that high baseline ctDNA was associated with tumor progression. However, similar to their findings, we did see that baseline HPV ctDNA levels correlated with nodal tumor burden. We note that the analytic details of our assay have been published previously (22), and sensitivity of different assays in the literature may be slightly different, and as such the copies/mL between different assays may not be directly comparable.

We saw small numbers of patients with detectable ctDNA at week 7 of chemoradiation or 3 months post-radiation. Larger studies are urgently needed, but our limited data suggest that persistent ctDNA alone may not be sufficient to drive adjuvant trials as 50% (2/4) cleared by 3 months and are NED. Detectable ctDNA at 3 months post-radiation was associated with high rate of progression somewhere, but in three of four cases here the disease was also radiographically apparent at 3 months, thus ctDNA did not provide an earlier detection.

A previous study suggested that early small volume tumors may not have correlations between baseline ctDNA levels and gross tumor volume (31). Our granular data in high-risk patients where we calculated primary tumor versus nodal GTV and MRI/PET volume metrics suggest that baseline ctDNA levels correlate with pretreatment nodal volume. During treatment, we saw that imaging characteristics in the primary tumor correlated with ctDNA values, while nodal imaging metrics no longer correlated. At a molecular level, the literature suggests early shedding may happen within hours of radiation or chemotherapy with early shedding as a result of mitotic catastrophe and later shedding due to apoptosis and necrosis (32, 33). Furthermore, tumor vascularity, perfusion, and hypoxia are some of the confounding factors likely affecting the rate kinetics of cfDNA release (34). Thus, the details of treatment and the relative timing for blood draw with radiation or chemotherapy may impact results. Further clinical studies may incorporate this into their blood sample acquisition protocol.

We have previously characterized the analytic details of our HPV16 ctDNA assay, and shown excellent performance with a limit of detection of less than five molecules of ctDNA (22), but found in this study that 6 of 34 p16+ patients had undetectable ctDNA at baseline. Unfortunately, a limitation of our study was that FFPE material sufficient for sequencing was only available on 12 of the patients preventing an in-depth analysis of the impact of tumor HPV genetics (e.g., to know whether non-HPV16/18 types were present in the 6 undetectable patients) and assay performance. However, within this dataset, we did observe HPV sequence variation within the tumors, suggesting that minor sequence variation could impact the efficiency of single probe-based ctDNA assays (and which needs to be experimentally explored in the future). Furthermore, we found a range of HPV copy numbers in the tumors, and do not discount the possibility that low copy number is one of the contributing factors to the inability to detect ctDNA at baseline. We note that we used only 1 mL of plasma for our assay, but other studies are using larger plasma volumes, which may account for some of the discordance between the various assays that are emerging. Despite the limitations of HPV analysis, however, we were able to characterize the mutational status of the tumors, including of two primary and recurrence pairs. We show that several of the tumors that failed therapy had mutations in DNA repair pathway genes, suggesting that in the future, genetic profiling of these tumors may be warranted as an additional potential biomarker capable of predicting response.

Additional limitations of our study include small sample size as the collection of blood for HPV ctDNA analysis was added to our trial in 2016. However, the number of failures here in our uniform high-risk population is comparable or higher than other reports of more variable populations seen in the literature to date. As ours and others published to date include small numbers of patients, larger studies are urgently needed to define the value of ctDNA kinetics in prediction of outcomes in p16+ OPSCC.

In summary, our results suggest that early HPV ctDNA kinetics and imaging biomarkers within the first 2 weeks of chemoradiation are predictive of outcomes in AJCC8 stage III p16+ OPSCC. Furthermore, and in contrast to previous reports discussed above, HPV ctDNA clearance at week 4 was not correlated with outcomes and thus should not be used to predict prognosis or drive adjuvant trials in stage III p16+ OPSCC. More research on the implementation of ctDNA is urgently needed to advance this biomarker. Importantly, our data show that early timepoints warrant further investigation in a prospective clinical trial setting, as both ctDNA and imaging have great promise for the ability to substantially improve our ability to predict patient outcomes at a timepoint that would facilitate individualizing care for each patient. Indeed, the use of mid-treatment imaging and/or real-time ctDNA to predict patient outcomes may be soon approaching, and if this occurs, then we may have the exciting opportunity to design trials to further personalize clinical treatment in the frontline setting. Long term, we hope that such adaptive treatment plans come to fruition and improve the overall survival of patients while minimizing treatment morbidity.

Y. Cao reports grants from NIH during the conduct of the study, as well as grants from Siemens Heathineers outside the submitted work. C. Bhambhani reports a patent for MATERIALS AND METHODS FOR MEASURING HPV CTDNA pending. M. Heft Neal reports grants from NIH during the conduct of the study. M. Tewari reports a patent for MATERIALS AND METHODS FOR MEASURING HPV CTDNA pending. P.L. Swiecicki reports a patent for MATERIALS AND METHODS FOR MEASURING HPV CTDNA pending. M. Mierzwa reports grants from NIH U01 CA 183848 and NIH R01 CA 184153 during the conduct of the study. J.C. Brenner reports grants from NIH during the conduct of the study, as well as a patent for HPV16 ctDNA assay (63/208736) pending. No disclosures were reported by the other authors.

Y. Cao: Conceptualization, resources, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. C.T. Haring: Conceptualization, data curation, formal analysis, writing–original draft. C. Brummel: Data curation, writing–review and editing. C. Bhambhani: Resources, data curation, formal analysis, writing–original draft, writing–review and editing. M. Aryal: Resources, data curation, writing–review and editing. C. Lee: Conceptualization, data curation, writing–review and editing. M. Heft Neal: Data curation, writing–review and editing. A. Bhangale: Data curation, writing–review and editing. W. Gu: Formal analysis, writing–review and editing. K. Casper: Conceptualization, resources, funding acquisition, writing–original draft, writing–review and editing. K. Malloy: Conceptualization, writing–review and editing. Y. Sun: Formal analysis, writing–original draft, writing–review and editing. A. Shuman: Conceptualization, writing–review and editing. M.E. Prince: Conceptualization, writing–review and editing. M.E. Spector: Conceptualization, resources, writing–review and editing. S. Chinn: Investigation, writing–review and editing. J. Shah: Conceptualization, resources, writing–review and editing. C. Schonewolf: Conceptualization, resources, writing–review and editing. J.B. McHugh: Resources, data curation, writing–review and editing. R.E. Mills: Formal analysis, writing–review and editing. M. Tewari: Conceptualization, data curation, writing–original draft, writing–review and editing. F.P. Worden: Conceptualization, writing–review and editing. P.L. Swiecicki: Conceptualization, writing–review and editing. M. Mierzwa: Conceptualization, resources, data curation, formal analysis, writing–original draft, writing–review and editing. J.C. Brenner: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing.

This work was supported by NIH/NCI grants U01CA183848 and RO1CA184153. Institutional support for funding of the clinical trial was provided through the NCI Cancer Center Support Grant—grant number P30CA046592.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Sturgis
EM
,
Cinciripini
PM
. 
Trends in head and neck cancer incidence in relation to smoking prevalence: an emerging epidemic of human papillomavirus-associated cancers?
Cancer
2007
;
110
:
1429
35
.
2.
Pytynia
KB
,
Dahlstrom
KR
,
Sturgis
EM
. 
Epidemiology of HPV-associated oropharyngeal cancer
.
Oral Oncol
2014
;
50
:
380
6
.
3.
Marur
S
,
D'Souza
G
,
Westra
WH
,
Forastiere
AA
. 
HPV-associated head and neck cancer: a virus-related cancer epidemic
.
Lancet Oncol
2010
;
11
:
781
9
.
4.
Gillison
ML
,
Chaturvedi
AK
,
Anderson
WF
,
Fakhry
C
. 
Epidemiology of human papillomavirus-positive head and neck squamous cell carcinoma
.
J Clin Oncol
2015
;
33
:
3235
42
.
5.
Chen
MM
,
Mott
N
,
Clark
SJ
,
Harper
DM
,
Shuman
AG
,
Prince
MEP
, et al
HPV vaccination among young adults in the US
.
JAMA
2021
;
325
:
1673
4
.
6.
Chaturvedi
AK
,
Engels
EA
,
Pfeiffer
RM
,
Hernandez
BY
,
Xiao
W
,
Kim
E
, et al
Human papillomavirus and rising oropharyngeal cancer incidence in the United States
.
J Clin Oncol
2011
;
29
:
4294
301
.
7.
Håkansson
K
,
Rasmussen
JH
,
Rasmussen
GB
,
Friborg
J
,
Gerds
TA
,
Fischer
BM
, et al
A failure-type specific risk prediction tool for selection of head-and-neck cancer patients for experimental treatments
.
Oral Oncol
2017
;
74
:
77
82
.
8.
Billfalk-Kelly
A
,
Yu
E
,
Su
J
,
O'Sullivan
B
,
Waldron
J
,
Ringash
J
, et al
Radiologic extranodal extension portends worse outcome in cN+ TNM-8 stage I human papillomavirus-mediated oropharyngeal cancer
.
Int J Radiat Oncol Biol Phys
2019
;
104
:
1017
27
.
9.
Rosen
BS
,
Wilkie
JR
,
Sun
Y
,
Ibrahim
M
,
Casper
KA
,
Miller
JE
, et al
CT and FDG-PET radiologic biomarkers in p16+ oropharyngeal squamous cell carcinoma patients treated with definitive chemoradiotherapy
.
Radiother Oncol
2021
;
155
:
174
81
.
10.
Pollom
EL
,
Song
J
,
Durkee
BY
,
Aggarwal
S
,
Bui
T
,
von Eyben
R
, et al
Prognostic value of midtreatment FDG-PET in oropharyngeal cancer
.
Head Neck
2016
;
38
:
1472
8
.
11.
Mowery
YM
,
Vergalasova
I
,
Rushing
CN
,
Choudhury
KR
,
Niedzwiecki
D
,
Wu
Q
, et al
Early (18)F-FDG-PET response during radiation therapy for HPV-related oropharyngeal cancer may predict disease recurrence
.
Int J Radiat Oncol Biol Phys
2020
;
108
:
969
76
.
12.
Magbanua
MJM
,
Li
W
,
Wolf
DM
,
Yau
C
,
Hirst
GL
,
Swigart
LB
, et al
Circulating tumor DNA and magnetic resonance imaging to predict neoadjuvant chemotherapy response and recurrence risk
.
NPJ Breast Cancer
2021
;
7
:
32
.
13.
Wong
SQ
,
Raleigh
JM
,
Callahan
J
,
Vergara
IA
,
Ftouni
S
,
Hatzimihalis
A
, et al
Circulating tumor DNA analysis and functional imaging provide complementary approaches for comprehensive disease monitoring in metastatic melanoma
.
JCO Precis Oncol
2017
;
1
:
1
14
.
14.
Hanna
GJ
,
Supplee
JG
,
Kuang
Y
,
Mahmood
U
,
Lau
CJ
,
Haddad
RI
, et al
Plasma HPV cell-free DNA monitoring in advanced HPV-associated oropharyngeal cancer
.
Ann Oncol
2018
;
29
:
1980
6
.
15.
Chera
BS
,
Kumar
S
,
Beaty
BT
,
Marron
D
,
Jefferys
S
,
Green
R
, et al
Rapid clearance profile of plasma circulating tumor HPV type 16 DNA during chemoradiotherapy correlates with disease control in HPV-associated oropharyngeal cancer
.
Clin Cancer Res
2019
;
25
:
4682
90
.
16.
Chera
BS
,
Kumar
S
,
Shen
C
,
Amdur
R
,
Dagan
R
,
Green
R
, et al
Plasma circulating tumor HPV DNA for the surveillance of cancer recurrence in HPV-associated oropharyngeal cancer
.
J Clin Oncol
2020
;
38
:
1050
8
.
17.
Haring
CT
,
Brummel
C
,
Bhambhani
C
,
Jewell
B
,
Neal
MH
,
Bhangale
A
, et al
Letter to the editor: implementation of human papillomavirus circulating tumor DNA to identify recurrence during treatment de-escalation
.
Oral Oncol
2021
;
121
:
105332
.
18.
Teng
F
,
Aryal
M
,
Lee
J
,
Lee
C
,
Shen
X
,
Hawkins
PG
, et al
Adaptive boost target definition in high-risk head and neck cancer based on multi-imaging risk biomarkers
.
Int J Radiat Oncol Biol Phys
2018
;
102
:
969
77
.
19.
Aryal
MP
,
Lee
C
,
Hawkins
PG
,
Chapman
C
,
Eisbruch
A
,
Mierzwa
M
, et al
Real-time quantitative assessment of accuracy and precision of blood volume derived from DCE-MRI in individual patients during a clinical trial
.
Tomography
2019
;
5
:
61
7
.
20.
Cao
Y
,
Aryal
M
,
Li
P
,
Lee
C
,
Schipper
M
,
Hawkins
PG
, et al
Predictive values of MRI and PET derived quantitative parameters for patterns of failure in both p16+ and p16- high risk head and neck cancer
.
Front Oncol
2019
;
9
:
1118
.
21.
Brenner
JC
,
Graham
MP
,
Kumar
B
,
Saunders
LM
,
Kupfer
R
,
Lyons
RH
, et al
Genotyping of 73 UM-SCC head and neck squamous cell carcinoma cell lines
.
Head Neck
2010
;
32
:
417
26
.
22.
Haring
CT
,
Bhambhani
C
,
Brummel
C
,
Jewell
B
,
Bellile
E
,
Heft Neal
ME
, et al
Human papilloma virus circulating tumor DNA assay predicts treatment response in recurrent/metastatic head and neck squamous cell carcinoma
.
Oncotarget
2021
;
12
:
1214
29
.
23.
Tillman
BN
,
Yanik
M
,
Birkeland
AC
,
Liu
C-J
,
Hovelson
DH
,
Cani
AK
, et al
Fibroblast growth factor family aberrations as a putative driver of head and neck squamous cell carcinoma in an epidemiologically low-risk patient as defined by targeted sequencing
.
Head Neck
2016
;
38
:
E1646
52
.
24.
Birkeland
AC
,
Yanik
M
,
Tillman
BN
,
Scott
MV
,
Foltin
SK
,
Mann
JE
, et al
Identification of targetable ERBB2 aberrations in head and neck squamous cell carcinoma
.
JAMA Otolaryngol Head Neck Surg
2016
;
142
:
559
67
.
25.
Swiecicki
PL
,
Bellile
EL
,
Brummel
CV
,
Brenner
JC
,
Worden
FP
. 
Efficacy of axitinib in metastatic head and neck cancer with novel radiographic response criteria
.
Cancer
2021
;
127
:
219
28
.
26.
Birkeland
AC
,
Foltin
SK
,
Michmerhuizen
NL
,
Hoesli
RC
,
Rosko
AJ
,
Byrd
S
, et al
Correlation of Crtc1/3-Maml2 fusion status, grade and survival in mucoepidermoid carcinoma
.
Oral Oncol
2017
;
68
:
5
8
.
27.
Heft Neal
ME
,
Bhangale
AD
,
Birkeland
AC
,
McHugh
JB
,
Shuman
AG
,
Rosko
AJ
, et al
Prognostic significance of oxidation pathway mutations in recurrent laryngeal squamous cell carcinoma
.
Cancers
2020
;
12
:
3081
.
28.
Pinatti
LM
,
Gu
W
,
Wang
Y
,
Elhossiny
A
,
Bhangale
AD
,
Brummel
CV
, et al
SearcHPV: a novel approach to identify and assemble human papillomavirus-host genomic integration events in cancer
.
Cancer
2021
;
127
:
3531
40
.
29.
Dahlstrom
KR
,
Li
G
,
Hussey
CS
,
Vo
JT
,
Wei
Q
,
Zhao
C
, et al
Circulating human papillomavirus DNA as a marker for disease extent and recurrence among patients with oropharyngeal cancer
.
Cancer
2015
;
121
:
3455
64
.
30.
Cao
H
,
Banh
A
,
Kwok
S
,
Shi
X
,
Wu
S
,
Krakow
T
, et al
Quantitation of human papillomavirus DNA in plasma of oropharyngeal carcinoma patients
.
Int J Radiat Oncol Biol Phys
2012
;
82
:
e351
8
.
31.
Damerla
RR
,
Lee
NY
,
You
D
,
Soni
R
,
Shah
R
,
Reyngold
M
, et al
Detection of early human papillomavirus-associated cancers by liquid biopsy
.
JCO Precis Oncol
2019
;
3
:
PO.18.00276
.
32.
Kimple
RJ
,
Smith
MA
,
Blitzer
GC
,
Torres
AD
,
Martin
JA
,
Yang
RZ
, et al
Enhanced radiation sensitivity in HPV-positive head and neck cancer
.
Cancer Res
2013
;
73
:
4791
800
.
33.
Rostami
A
,
Lambie
M
,
Yu
CW
,
Stambolic
V
,
Waldron
JN
,
Bratman
SV
. 
Senescence, necrosis, and apoptosis govern circulating cell-free DNA release kinetics
.
Cell Rep
2020
;
31
:
107830
.
34.
Wan
JCM
,
Massie
C
,
Garcia-Corbacho
J
,
Mouliere
F
,
Brenton
JD
,
Caldas
C
, et al
Liquid biopsies come of age: towards implementation of circulating tumour DNA
.
Nat Rev Cancer
2017
;
17
:
223
38
.

Supplementary data