Rapid molecular response kinetics to anti-PD-1–based first-line treatment of Hodgkin lymphoma in the GHSG NIVAHL trial

TO THE EDITOR:

Immune checkpoint blockade (ICB) targeting anti–programmed death protein 1 (anti-PD-1) induces high response rates in patients with previously untreated and relapsed/refractory classic Hodgkin lymphoma (HL).1, 2, 3 In the first-line setting, ICB was applied in either sequential or concomitant combination with doxorubicin, vinblastine, and dacarbazine (AVD) chemotherapy, followed by optional radiotherapy in most trials.^2,3 Moreover, ICB single-agent efficacy in relapsed/refractory HL⁴ and early (near) complete reduction of metabolic tumor volume in treatment-naïve patients with HL⁵ suggest that chemotherapy and radiotherapy could be further reduced in patients with rapid and deep response to ICB. However, it remains unclear how to identify these patients and guide a more individualized anti-PD1–based first-line HL treatment.⁶

Recently, we and others have developed genetic classifications of HL based on circulating tumor (ct)DNA sequencing, in which subtypes are characterized by distinct genetic alterations and tumor microenvironment (TME) composition.7, 8, 9 These classifications distinguish subtypes characterized by inflammation and immune escape phenomena from subtypes frequently associated with somatic mutations in key oncogenic driver genes of HL.7, 8, 9 Specifically, Inflammatory immune escape HL is characterized by high-level amplifications of 9p24.1, harboring PD-L1/PD-L2, and a TME enriched with potentially immunosuppressive regulatory T cells.⁸ Virally-driven HL is frequently associated with Epstein-Barr virus (EBV), human herpesvirus 6, and TME features supporting latent viral infection and reduced local antitumor immunity.⁸ Therefore, it is reasonable to assume that Inflammatory immune escape HL and Virally-driven HL are more susceptible to ICB than Oncogene-driven HL, a subtype characterized by oncogenic alterations that showed favorable outcomes after treatment with chemotherapy in the German Hodgkin Study Group (GHSG) HD21 trial.⁸

Here, we present a hypothesis-generating pilot study assessing genetic HL subtypes and minimal residual disease (MRD) kinetics during anti-PD-1–based first-line treatment in samples obtained from the GHSG NIVAHL trial (NCT03004833).^2,10

As previously described, patients were randomly assigned to receive either a sequential or concomitant combination of nivolumab (N) with AVD followed by 30 Gy involved-site radiotherapy (IS-RT). In the concomitant arm A, patients received 4× N-AVD, whereas in the sequential arm B, patients received 4× N, followed by 2× N-AVD and 2× AVD.^2,10 Peripheral blood samples were prospectively collected as part of an ancillary substudy at predefined time points: at baseline (before treatment initiation); ∼1 week after the first treatment dose, 1× N-AVD in arm A or 1× N in arm B; and at each of the 3 scheduled response assessments using Fluorodeoxyglucose F18 (¹⁸F-FDG) positron emission tomography (PET). In line with our primary research questions (very early MRD detection after 1 infusion of N in arm B; MRD status across genetic subtypes of HL), material from arm B was prioritized for this study. After exclusion of patients without written informed consent for participation in the ancillary substudy and patients without sufficient suitable biomaterial, a cohort of 43 patients with a baseline sample and at least 1 follow-up sample was available (total samples, n = 146; baseline, n = 43; follow-up, n = 103; Figure 1). Sequencing data from 79 of 146 samples were available from previous studies,^8,11 whereas the remaining samples underwent sequencing as part of this study.

Figure 1.

Patient and Sample Distribution. Flowchart depicting sample numbers (dark gray boxes) and the treatment applied before the respective sampling time point (light gray boxes) per treatment arm. Abbreviations: N = nivolumab, AVD: doxorubicine, vinblastine, dacarbazine, IS-RT: involved-site radiotherapy.

Sample collection, extraction of cell-free DNA, quality control, library preparation, sequencing, bioinformatic analysis, MRD detection, and assignment of HL subtypes were performed as previously described.^8,11

In brief, cell-free DNA was extracted from 2 to 4 mL peripheral blood and quantified considering size distribution (i.e., size window of interest, 50 to 700 base pairs [bp]) using the TapeStation 2200 System (Agilent, Santa Clara, CA). Libraries were prepared with the SureSelect XT HS2 V2 automated target enrichment (Agilent), using duplex molecular identifiers for error suppression and the Agilent Bravo platform (Agilent).

To enable genotyping and ultrasensitive MRD detection, we used LymphoVista HL, a technically validated ctDNA sequencing assay.¹² LymphoVista HL targets 305 regions of interest across ∼100 kbp of the human genome (hg38; supplemental Table 1).

After additional quality control using the TapeStation 2200 System (Agilent), pools of libraries were generated and sequenced on an Illumina NovaSeq6000 sequencing system (Illumina, San Diego, CA) using S1 to S4 flow cells (2× 100-bp read length; 5 Gb per sample). Basic data processing, somatic single-base substitution calling, and assessment of MRD were performed using an in-house bioinformatic pipeline as previously described.⁸

MRD status was categorized as positive or negative based on being above or below the limit of detection as previously described.^8,11 Remission status by PET was categorized according to the Lugano classification, with Deauville Score (DS) 1 to 3 considered PET negative and DS 4 to 5 considered PET positive.¹³ Descriptive statistical analyses and visualizations were performed using R (version 4.1.2).

Patient genotypes and phenotypes resembled previously identified hallmarks of each subtype⁸ (supplemental Table 2). Patients with Virally-driven HL were frequently associated with EBV positivity, as assessed by immunohistochemistry or in situ hybridization (33.3%), and had lower tumor mutational burden (median, 1.5 mutations per Mb; range, 0.9-139), whereas patients with Oncogene-driven HL had higher tumor mutational burden (median, 9.5 mutations per Mb; range, 2.3-25.4). Patients with Oncogene-driven HL depicted higher baseline metabolic tumor volume (median, 133 mL; range, 17.5-437) than Inflammatory immune escape HL (median, 96.7 mL; range, 11-581) and Virally-driven HL (median, 43.5 mL; range, 18.7-75.6). Interestingly, ctDNA levels showed a contrary trend, indicating that tumors with lower tumor volume might still shed high loads of ctDNA, a finding that was also reported by Alig et al.⁷

As expected, the proportion of patients with negative MRD increased with cumulative treatment exposure (Figure 2A).

Figure 2.

MRD detection during and after anti-PD-1–based first-line treatment. (A) Bar plots depicting the percentage of patients with negative and positive MRD status assessed by ctDNA sequencing at each time point. Week 1 represents a sample obtained after the first infusion of N or N-AVD; RA1 represents a sample obtained at the first response assessment after 4x N or 2x N-AVD, respectively; RA2, represents a sample obtained at the second response assessment after the end of systemic treatment; and EOT represents a sample obtained at the end of treatment after radiotherapy. (B) Bar plots depicting the percentage of patients with negative and positive MRD status assessed by ctDNA sequencing after just 1 infusion of N + AVD (arm A) or the first infusion of single-agent N (arm B). (C) Bar plots depicting the percentage of patients from arm B with negative and positive MRD status assessed by ctDNA sequencing after just 1 infusion of N according to biologic subtypes of HL. (D) Bar plots depicting the percentage of patients treated in both arms with negative and positive PET as well as negative and positive MRD status, as assessed by ctDNA sequencing at the end of systemic treatment, before initiation of 30 Gy IS-RT. Abbreviations: MRD, minimum residual disease; N, nivolumab; EOT, end of treatment; RA1, first response assessment; RA2, second response assessment; HL, classic Hodgkin lymphoma; PET, positron emission tomography.

To date, most studies^7,8,11,14 assessed MRD after multiple cycles of chemotherapy, but data on very early response kinetics, especially in the context of ICB, remain limited. Within 1 week (median, 7 days; range, 3-16) after treatment initiation, 10 of 23 patients (43.5%) had negative MRD (Figure 2A). All 4 patients treated in arm A achieved MRD negativity after 1× N-AVD. A total of 19 patients were treated in arm B, thus receiving only 1× N before MRD assessment. Strikingly, 6 of 19 patients (31.6%) achieved MRD negativity, suggesting that 1× N is sufficient to eradicate measurable lymphoma in selected patients (Figure 2B). Of note, 3 of 10 patients (30%) with negative MRD vs only 1 of 13 patients (7.7%) with positive MRD had EBV-positive HL. Patients with Inflammatory immune escape HL or Virally-driven HL were more likely to achieve negative MRD after just 1× N compared with Oncogene-driven HL (50% vs 23.1%; Figure 2C). These observations align well with the recent report of more pronounced benefit with N-AVD in EBV-positive vs EBV-negative patients with advanced-stage HL.¹⁵

At the end of systemic treatment (i.e., after cumulative 4× N-AVD in both arms), 8 of 24 patients (33.3%) had residual metabolic activity as assessed by PET. In contrast, 0 of 24 patients (0%) had detectable MRD (Figure 2D). During the available follow-up,² none of the 8 PET-positive patients experienced relapse, progression, or death. These results suggest that MRD allows for the detection of complete molecular response in patients with potentially false-positive PET, a well-acknowledged challenge after ICB.

In 1 patient, MRD was first assessed at the end of treatment (i.e., after cumulative 4× N-AVD and IS-RT) and was determined to be positive. PET indicated partial remission (DS4) at this time point and converted to complete remission during follow-up without additional therapy. However, the patient remains free of relapse with the available follow-up, suggesting either false-positive MRD detection or that additional assessments of MRD would have been necessary to detect the time point of achieving delayed complete molecular remission in this case.

Taken together, our hypothesis-generating, exploratory study has 3 main findings that require further validation in future studies.

First, we show that ctDNA-based genetic subtyping is feasible in the context of treatment with ICB and confirm previously discovered hallmarks of HL subtypes.

Second, a third of patients with newly diagnosed HL appears to achieve deep molecular remission after just a single dose of anti-PD ICB. This was observed more frequently in patients with Inflammatory immune escape HL or Virally-driven HL vs Oncogene-driven HL. Pending validation in future clinical trials (e.g., ongoing GHSG INDIE trial; NCT04837859), biologic classifiers of HL and very early MRD detection might serve as tools to individualize anti-PD1 ICB incorporation into treatment regimen and reduce chemotherapy intensity (anti-PD1 ICB + AVD)³ or even omit chemotherapy (anti-PD1 ICB alone) in selected patients with HL, whereas other patients might instead benefit from more intensive chemotherapy such as brentuximab vedotin, etoposide, cyclophospamide, doxorubicine, dacarbazine and dexamethasone (BrECADD).¹⁶

Third, MRD appears to be a promising tool for clarifying PET positivity at the end of systemic treatment, as recently suggested in diffuse large B-cell lymphoma.¹⁷ In the context of ICB-based first-line treatment of HL, this could spare a relevant proportion of patients from consolidative radiotherapy.

All human participant research was performed as approved by the responsible ethics committees (Uni Koeln 16-444) and authorities and in accordance with the Declaration of Helsinki.

Conflict-of-interest disclosure: J.-M.H. reports research funding from Incyte, MorphoSys, and Novartis; honoraria or adviser or consultant fees from Incyte, Roche, Novartis, Miltenyi Biotec, Sobi, and SERB Pharmaceuticals; and travel support from Sobi, and SERB Pharmaceuticals. K.T.-G. received honoraria or adviser or consultant fees from Amgen, Bristol Myers Squibb (BMS), Novartis, Grifols, Sanofi, Sobi, Stemline, and Takeda. C.-A.V. reports honoraria from Novartis Pharma and is an adviser for Novartis Radiopharmaceuticals. B.v.T. reports adviser or consultant fees from Allogene, Amgen, BMS/Celgene, Cerus, Gilead Kite, Incyte, IQVIA, Janssen-Cilag, Lilly, Merck Sharp & Dohme (MSD), Miltenyi Biotec, Novartis, Noscendo, Pentixapharm, Pfizer, Pierre Fabre, Qualworld, Regeneron, Roche, Sobi, and Takeda; honoraria from AbbVie, AstraZeneca, BMS/Celgene, Gilead Kite, Incyte, Janssen-Cilag, Lilly, MSD, Novartis, Roche, SERB, and Takeda; research funding from Esteve (institution), MSD (institution), Novartis (institution), and Takeda (institution); travel support from AbbVie, AstraZeneca, Gilead Kite, Janssen-Cilag, Lilly, MSD, Pierre Fabre, Roche, Takeda, and Novartis; and is a member of steering committees for Regeneron (institution) and Takeda. P.J.B. reports adviser or consultant fees from Hexal, MSD (MSD), Need Inc, Stemline, and Takeda; holds stock options in Need Inc; received honoraria from AstraZeneca, BeiGene, BMS/Celgene, Eli Lilly, MSD, Need Inc, Stemline, and Takeda; and reports research funding from BeiGene (institution), BMS (institution), MSD (institution), and Takeda (institution). S.B. is shareholder and managing director of Liqomics, which commercially develops the Vista test platform; and is an adviser or consultant for Galapagos, Takeda, and Miltenyi Biomedicine. The remaining authors declare no competing financial interests.