Early circulating tumor DNA changes predict outcomes in head and neck cancer patients under re-radiotherapy

Abstract

Local recurrence after radiotherapy is common in locally advanced head and neck cancer (HNC) patients. Re-irradiation can improve local disease control, but disease progression remains frequent. Hence, predictive biomarkers are needed to adapt treatment intensity to the patient's individual risk. We quantified circulating tumor DNA (ctDNA) in sequential plasma samples and correlated ctDNA levels with disease outcome. Ninety four longitudinal plasma samples from 16 locally advanced HNC patients and 57 healthy donors were collected at re-radiotherapy baseline, after 5 and 10 radiation fractions, at irradiation end, and at routine follow-up visits. Plasma DNA was subjected to low coverage whole genome sequencing for copy number variation (CNV) profiling to quantify ctDNA burden. CNV-based ctDNA burden was detected in 8/16 patients and 25/94 plasma samples. Ten additional ctDNA-positive samples were identified by tracking patient-specific CNVs found in earlier sequential plasma samples. ctDNA-positivity after 5 and 10 radiation fractions (both: log-rank, p = .050) as well as at the end of irradiation correlated with short progression-free survival (log-rank, p = .006). Moreover, a pronounced decrease of ctDNA toward re-radiotherapy termination was associated with worse treatment outcome (log-rank, p = .005). Dynamic ctDNA tracking in serial plasma beyond re-radiotherapy reflected treatment response and imminent disease progression. In five patients, molecular progression was detected prior to tumor progression based on clinical imaging. Our findings emphasize that quantifying ctDNA during re-radiotherapy may contribute to disease monitoring and personalization of adjuvant treatment, follow-up intervals, and dose prescription.

Keywords: circulating tumor DNA; copy number variations; head and neck cancer; predictive biomarker; re‐radiotherapy.

FIGURE 1

(A) Box plot comparing CPA scores between healthy donors and head and neck cancer (HNC) patients. The maximum CPA score across healthy donors served as ctDNA detectability threshold (dotted line; CPA = 1.044). ctDNA‐positive and ‐negative samples are colored black and gray, respectively. ctDNA‐positive and total sample number per group is given in brackets. Box plots represent median, upper and lower quartile with Tukey Whiskers. (B) Summary of recurrent CNVs in HNC patients with detectable ctDNA (n = 8). The y‐axis represents the frequency of a detected copy number state at the chromosomal coordinate given on the x‐axis. All longitudinal plasma samples were considered, however, recurrently detected CNVs in two or more samples of the same patient were only counted once. The CNV frequency (i.e., the number of occurrences in the eight patients assessed) of amplifications and deletions is given in red and green, respectively. Areas shaded in gray represent the q‐arm of the respective chromosome. Genes associated with HNC tumorigenesis are labeled.

FIGURE 2

Head and neck cancer patient cohort overview highlighting the detectability of ctDNA copy number variation (CNV) analysis from low‐coverage whole genome sequencing. Green and blue colors indicate detectable CNVs by de novo CNV‐calling (CPA score) and using information from previous serial plasma samples of the same patient (ctCPA score), respectively. Missing squares represent unavailable plasma samples. The number of ctDNA‐positive and total number of samples per time point is given in brackets.

FIGURE 3

Association between ctDNA levels and progression‐free survival after re‐radiotherapy. Progression‐free survival (PFS) according to ctDNA detectability, as assessed via ctCPA scores, in plasma specimens collected after five fractions (A), 10 fractions (B) and at the end of re‐radiotherapy (re‐RT; C). (D) Association between PFS and CPA score change from samples taken after five re‐RT fractions to re‐RT end. Relative changes were calculated using CPA scores or ctCPA scores (if available). A decrease of ≥30% was used for partitioning of patients. Groups were compared using the two‐sided log‐rank test.

FIGURE 4

Representative patients (A–D) illustrating the utility of the ctCPA score for disease monitoring in HNC patients under re‐radiotherapy (re‐RT). Administered radio‐ and systemic therapies are represented by shaded areas. Time points of disease progression (PD) and PD location (if available) are denoted by vertical lines. Horizontal, dotted lines indicate the patient‐specific detectability threshold of the ctCPA score (i.e., maximum score across 57 healthy donors). Filled dots highlight samples with detectable ctCPA scores. Early signs of PD in months (i.e., increasing ctCPA scores prior to radiographic tumor progression) are highlighted in gray.