[18F]FDG-PET accurately identifies pathological response early upon neoadjuvant immune checkpoint blockade in head and neck squamous cell carcinoma

Abstract

Purpose: To investigate the utility of [18F]FDG-PET as an imaging biomarker for pathological response early upon neoadjuvant immune checkpoint blockade (ICB) in patients with head and neck squamous cell carcinoma (HNSCC) before surgery.

Methods: In the IMCISION trial (NCT03003637), 32 patients with stage II‒IVb HNSCC were treated with neoadjuvant nivolumab with (n = 26) or without (n = 6) ipilimumab (weeks 1 and 3) before surgery (week 5). [18F]FDG-PET/CT scans were acquired at baseline and shortly before surgery in 21 patients. Images were analysed for SUVmax, SUVmean, metabolic tumour volume (MTV), and total lesion glycolysis (TLG). Major and partial pathological responses (MPR and PPR, respectively) to immunotherapy were identified based on the residual viable tumour in the resected primary tumour specimen (≤ 10% and 11-50%, respectively). Pathological response in lymph node metastases was assessed separately. Response for the 2 [18F]FDG-PET-analysable patients who did not undergo surgery was determined clinically and per MR-RECIST v.1.1. A patient with a primary tumour MPR, PPR, or primary tumour MR-RECIST-based response upon immunotherapy was called a responder.

Results: Median ΔSUVmax, ΔSUVmean, ΔMTV, and ΔTLG decreased in the 8 responders and were significantly lower compared to the 13 non-responders (P = 0.05, P = 0.002, P < 0.001, and P < 0.001). A ΔMTV or ΔTLG of at least - 12.5% detected a primary tumour response with 95% accuracy, compared to 86% for the EORTC criteria. None of the patients with a ΔTLG of - 12.5% or more at the primary tumour site developed a relapse (median FU 23.0 months since surgery). Lymph node metastases with a PPR or MPR (5 metastases in 3 patients) showed a significant decrease in SUVmax (median - 3.1, P = 0.04). However, a SUVmax increase (median + 2.1) was observed in 27 lymph nodes (in 11 patients), while only 13 lymph nodes (48%) contained metastases in the corresponding neck dissection specimen.

Conclusions: Primary tumour response assessment using [18F]FDG-PET-based ΔMTV and ΔTLG accurately identifies pathological responses early upon neoadjuvant ICB in HNSCC, outperforming the EORTC criteria, although pseudoprogression is seen in neck lymph nodes. [18F]FDG-PET could, upon validation, select HNSCC patients for response-driven treatment adaptation in future trials.

Trial registration: https://www.

Clinicaltrials: gov/ , NCT03003637, December 28, 2016.

Keywords: Head and neck squamous cell carcinoma; Metabolic response assessment; Neoadjuvant immune checkpoint blockade; [18F]FDG-PET.

Conflict of interest statement

Joris L. Vos, Laura A. Smit, Abrahim Al-Mamgani, and Wouter V. Vogel declare no competing interests. Charlotte L. Zuur reports receiving institutional research financial support from BMS to fund the present trial. Jan Paul de Boer reports receiving institutional research funding from Merck KGaA and institutional honoraria for an advisory role for MSD, both outside the submitted work. Michiel W.M. van den Brekel reports receiving institutional research funding from ATOS Medical, outside the submitted work. John B.A.G. Haanen reports, all outside the submitted work: institutional honoraria for advisory roles for AIMM, Amgen, BioNTech, BMS, GSK, Ipsen, MSD, Merck Serono, Molecular Partners, Neogene Therapeutics, Novartis, Pfizer, Roche/Genentech, Sanofi, Seattle Genetics, Third Rock Ventures, Vaximm; stock option ownership of Neogene Therapeutics; Institutional research funding from Amgen, BioNTech, BMS, MSD, Novartis.

© 2021. The Author(s).

Figures

Fig. 1

Flow chart of trial treatments, timelines, and patients included in IMCISION and their [18F]FDG-PET-based metabolic treatment response. Patients were treated with neoadjuvant nivolumab or nivolumab and ipilimumab (week 1), followed by nivolumab (week 3). Surgery was performed in week 5 or, ultimately, week 6. An evaluable baseline and on-treatment scan were obtained in 21 patients, of whom 6 had a major pathological response (MPR), 1 a partial pathological response (PPR), and 12 no pathological response (NPR) at their primary tumour site. Two patients did not undergo complete resection of their primary tumour and were classified according to MR-RECIST v.1.1, which was in agreement with physical examination in both cases: 1 RECIST-PR (responder) and 1 RECIST-PD (non-responder)

Fig. 2

SUVmax waterfall plot and two cases illustrating the EORTC criteria’s relative inaccuracy to determine pathologic response at the primary tumour site early upon neoadjuvant ICB in HNSCC. a Waterfall plot presenting the percentage change in primary tumour SUVmax from baseline to on-treatment per individual patient. Green bars represent pathological responders, red bars non-responders. Patients in whom response was assessed per MR-RECIST are marked with an asterisk. Dotted grey lines at 25% and − 25% represent the EORTC criteria for progressive metabolic disease (PMD) and partial metabolic response (PMR), respectively; patients in between both lines had stable metabolic disease SMD. Two patients visually had a complete metabolic response (marked CMR), though the SUVmax did not become 0. Patients marked with b and c are further illustrated under b and c, respectively. b A patient with clinically rT2N0 carcinoma of the left cheek mucosa demonstrated primary tumour PMD after 2 cycles of nivolumab + ipilimumab, with a SUVmax increase from 5.4 to 11.7 (117%). Evaluation of the surgically resected specimen (right panel) revealed a major (near-complete) pathological response, with some viable residual tumour (‘T’) surrounded by a dense infiltrate of immune cells (‘IC’). c A patient with cT4aN2b HNSCC of the floor of the mouth shows a SUVmax increase from 15.9 to 17.5 (10%). Correlative histopathology shows a partial pathological response: 69% of the histologically identifiable tumour bed is taken up by keratinous debris (KD) under apparent clearance of multinucleated giant cells and foamy macrophages (arrows)

Fig. 3

Individual primary tumour [18F]FDG-PET-based metabolic parameters at baseline (Pre) and on-treatment (On). a‒d Change in primary tumour SUVmax (a), SUVmean (b), metabolic tumour volume (MTV, c), and total lesion glycolysis (TLG, d) from baseline to on-treatment. Patients with a response at their primary tumour site are shown in green, and patients without a primary tumour ICB response are shown in red. The 2 patients with a complete metabolic response are included with values ‘0’ for on-treatment SUVmean, MTV, and TLG. P-values were calculated using a Wilcoxon signed-rank test. Primary tumour SUVmean, MTV, and TLG could not be determined in the patient who had primary tumour metabolic pseudoprogression based on SUVmax (illustrated in Fig. 2b); this patient was excluded from b‒d. Please note that y-axes of c and d were interrupted to facilitate visualization

Fig. 4

Kaplan–Meier survival estimates of patients with or without a total lesion glycolysis (TLG)–based primary tumour metabolic response. a Time to progression (TTP) since surgery of patients with a ≥ 12.5% decrease in TLG at primary tumour site (green) from baseline to on-treatment and patients without a ≥ 12.5% decrease (red). Only patients who underwent surgery are included here. b Overall survival since the start of ICB, for the same TLG-based metabolic response groups. The 2 patients who did not undergo surgery are included here. The deceased patients with a TLG-based metabolic response all died disease-free of causes unrelated to HNSCC. P-values were calculated using a log-rank test

Fig. 5

Metabolic responses, progression, and pseudoprogression in cervical lymph nodes. a Waterfall plot showing the absolute change in SUVmax from baseline to on-treatment for pathologically proven lymph node metastases in the neck dissection specimens. Bar colour indicates ICB response (green) or non-response (red). Bars marked b and ei are further detailed under b and ei, respectively. b A patient with cT4aN2c HNSCC of the left alveolar process of the mandible showed PMD at the primary tumour site (SUVmax + 35%, SUVmean + 35%, MTV + 144%, and TLG + 230%). Two ipsilateral level 2 lymph nodes (arrows) showed a SUVmax decrease from 8.8 to 5.7 (− 35%) and 6.6 to 5.6 (− 15%). Correlative keratin 14-stained pathology slides revealed one node with disturbed architecture but little viable tumour (12 × image), corresponding to an MPR. The other level 2 lymph node showed a PPR (not shown). c Waterfall plot showing the absolute change in SUVmax from baseline to on-treatment for pathologically proven tumour-negative lymph nodes. Bars marked with d and eii are further detailed under d and eii, respectively. d A patient with a SUVmax increase from 3.4 to 5.3 (56%) in a left (contralateral) level 1b lymph node after neoadjuvant ICB (arrows). Correlative H&E slide of the left level 1b neck dissection specimen revealed no lymph node metastases. This patient’s primary tumour showed a partial pathological response (shown in Fig. 2c). e Level 3 transversal [18F]FDG-PET and keratin 14-stained pathology images of the same patient shown under b. Two level 3 nodes are detected: one left (ipsilateral, marked ei) with an SUVmax increase from 4.0 to 8.5, and one right (contralateral, marked eii) with an SUVmax increase from 4.1 to 8.8. Correlative keratin 14-stained pathology slides showed a metastasis in level 3 left without evidence of ICB response (ei), while none of the resected right level 3 nodes contained tumour (eii)