Camrelizumab Plus Gemcitabine, Vinorelbine, and Pegylated Liposomal Doxorubicin in Relapsed/Refractory Primary Mediastinal B-Cell Lymphoma: A Single-Arm, Open-Label, Phase II Trial

Primary mediastinal B-cell lymphoma (PMBCL) is a distinct pathologic entity of diffuse large B-cell lymphoma (DLBCL) with unique clinical, pathologic, and genetic features. It occurs predominantly in women in their third or fourth decade of life, but this gender preponderance is only seen in Caucasians. Ethnicity is also a prognostic factor, and other ethnic groups fare worse 5-year overall survival (OS) than Caucasians. Although the outcome of the recommended first-line rituximab-containing therapy is favorable, 10%–30% patients are considered to be relapsed/refractory PMBCL (rrPMBCL) with limited treatment options and poor prognosis (1, 2). There is currently no optimal therapy for rrPMBCL because of the limited data available for efficacy assessment of salvage regimens (1). Commonly used salvage therapies for rrPMBCL comprise high-dose chemotherapy followed by autologous stem cell transplantation (ASCT), as in rrDLBCL (3, 4). However, response rates to salvage therapies are unacceptably low (0%–25%) and typically few patients proceed to ASCT in this context. Because of this poor response, the 2-year OS was only 15% in rrPMBCL, and 2-year progression-free survival (PFS) for refractory patients was 0% (5). Chimeric antigen receptor-19 (CAR-19) T-cell therapy was approved for treatment of rrPMBCL, but the possibility of severe toxicity due to high cytokine levels, such as cytokine release syndrome and neurotoxicity were the major challenges with this pioneering therapy. Aside from palliation, new salvage therapeutic strategies are needed (2, 6).

Despite its classification as a subtype of DLBCL, PMBCL shares many common characteristics with classical Hodgkin lymphoma, including the loci coamplification of programmed death-1 (PD-1) ligands PD-L1/PDL-2 and JAK2 in chromosome 9p24.1. The overexpression of PD-L1/PDL-2 and constitutive activation of JAK/STAT signaling exert positive feedback. Furthermore, comparative genomic analyses revealed that molecular feature of PMBCL with significant somatic copy-number alterations and structural variants would increase sensitivity to PD-1 blockade. Anti-PD-1 monotherapy with pembrolizumab demonstrated a 41% objective response rate (ORR) with 13% complete response (CR) in patients with rrPMBCL, and was approved by the FDA for patients with refractory PMBCL or relapsed following two or more lines of prior therapy (7, 8). However, owing to the delayed tumor degradation effect of anti-PD-1 monotherapy, this treatment option is impractical for those with bulky aggressive lesions or life-threatening tumor mass (2, 6).

The GVD chemotherapy combination of gemcitabine, vinorelbine, and pegylated liposomal doxorubicin (PLD) has proved to be effective and well-tolerated in patients with rrDLBCL and classical Hodgkin lymphoma, which shares many features with PMBCL (9–11). As an uncommon mild salvage chemotherapy, utilizing GVD would introduce less cross-resistant drugs for patients with rrPMBCL (9). Furthermore, gemcitabine is a myeloid-derived suppressor cells (MDSC)-depleting chemotherapeutic agent and can reduce immunosuppression in the tumor microenvironment (12). Meanwhile, PLD is a chemotherapeutic drug and an immunomodulatory agent that induces immunogenic cell death (13, 14). These effects could augment the antitumor immunity of the anti-PD-1 antibody (15, 16). The combination of PD-1 blockade with GVD chemotherapy is a reasonable choice for patients who undergo rapid clinical deterioration.

Camrelizumab (SHR1210) is a fully humanized immunoglobulin G4/κ anti-PD-1 mAb that blocks the interaction between PD-1 and its ligands. Camrelizumab has been favorably assessed in patients with recurrent or metastatic nasopharyngeal and esophageal carcinoma and relapsed/refractory classical Hodgkin lymphoma (15, 17, 18). Here, we report the efficacy and safety results of camrelizumab plus GVD chemotherapy in a phase II study in patients with bulky aggressive rrPMBCL.

The open-label, single-arm, phase II trial was approved and monitored by the Institutional Review Board of the Chinese PLA General Hospital (Beijing, China, S2016-127-01) and was conducted in accordance with the ethical guidelines of the Declaration of Helsinki and the International Conference on Harmonisation guidelines for Good Clinical Practice. All authors vouch that the study protocol was strictly followed and for the accuracy and completeness of the data. Written informed consent was obtained from each enrolled patient. This trial is registered with ClinicalTrials.gov, number NCT02961101 (for multiple tumors) and NCT03346642 (for PMBCL).

Eligible patients were adults (aged 18 years or older) with central, pathologically confirmed PMBCL, who experienced refractory [defined as no CR to first-line therapy and lack of response (CR and partial response (PR)) to any consequent treatment] or relapse (defined as recurrence within 6 months after first-line therapy) to at least one rituximab-containing chemotherapy regimen. Eligible patients must have at least one mediastinal mass with the greatest diameter at least 5 cm, with/without extrathoracic disease; an Eastern Cooperative Oncology Group performance status 0–2; and adequate organ function. The primary exclusion criteria included a history of allogeneic hematopoietic stem cell transplantation or organ transplantation, central nervous system involvement, prior ASCT, anti-PD-1, or CAR T therapy within 3 months, and previous radiotherapy within 6 months. A complete list of the inclusion and exclusion criteria is provided in the Supplementary Data.

All patients received intravenous chemotherapy (day 1) consisting of gemcitabine (1,000 mg/m²), vinorelbine (30 mg), and PLD (20 mg/m²) plus 200 mg camrelizumab (day 2) every 3 weeks. If chemotherapy-related adverse events (AE) ≥grade 3 occurred, the next dose was withheld until the toxicity was managed with supportive care. Treatment would also be withheld for ≥grade 2 immune-related toxicities until relief of AEs or discontinuation of treatment. Treatment continued until progressive disease, development of unacceptable toxicity, or withdrawal of consent. The combination treatment continued until a second confirmed CR was achieved or patients had received up to 12 doses followed by a maintenance therapy. In the maintenance phase, 200 mg of camrelizumab was administered intravenously every 3 weeks for four cycles followed by every 6 weeks up to 1 year.

Tumor assessments were performed at the baseline, every 6 weeks by CT (to assess object response), and every 12 weeks by ¹⁸F-fluorodeoxyglucose–PET–CT (FDG-PET/CT, to confirm CR). All CT and FDG-PET/CT scans were reviewed and scored by the same radiologist and nuclear radiologist. All AEs were monitored and documented throughout the study and 3 months after the last dose according to the Common Terminology Criteria for Adverse Events (CTCAE), version 5.0.

Tumor samples were obtained as core needle biopsies at the baseline if the clinical condition allowed. The biopsy samples were used for pathologic and Opal multiplex immunofluorescence tissue analysis. Peripheral blood samples were collected for T-cell activity assessment by flow cytometry before each of the first five doses. Serum levels of cytokine and chemokine were measured with the use of LEGENDplex Immunoassays (BioLegend).

The primary endpoint was the ORR, defined as those with radiologically confirmed CR or PR according to the Lugano 2014 classification (19). Response was determined on the basis of the best overall response up to the data cut-off date for each patient. The primary safety endpoint was the incidence of treatment-related AEs in patients who received at least one dose of study treatment. Safety was assessed by quantifying the toxicities and grades experienced by subjects per CTCAE v5.0. Secondary endpoints included the CR rate (PET negative per Deauville score, 1–3), PFS (time from first dose to first documentation of disease progression or death from any causes), duration of response (DOR; time from first documented response until radiological disease progression or death), time to response (TTR; time from first dose to first evaluated response), and OS (time from first dose to death or date last known to be alive). Exploratory efficacy endpoints included the effect of PD-L1 expression, peripheral T-cell activity, and serum cytokine and chemokine level.

The Opal multiplex immunofluorescence staining was performed in 17 of 24 baseline biopsy samples. The other samples were unavailable due to inadequate sampling or nonrepresentative tissue.

Slides loaded with 17 samples were deparaffinized in xylene and rehydrated in a series of graded alcohols. Antigen retrievals were performed in citrate buffer (pH 6) with a microwave (Sharp, R-331ZX) for 20 minutes at 95°C followed by a 20-minute cool down at room temperature. After the quenching of endogenous peroxidase in 3% H₂O₂, slides were incubated with blocking reagent (ZSGB-BIO, ZLI-9022) for 30 minutes at room temperature. Antigens were then successively detected using the Opal protocol. Briefly, each primary antibody was incubated for 2 hours in a humidified chamber at 37°C, followed by detection using the horseradish peroxidase (HRP)-conjugated secondary antibody (Polink-1 HRP Polymer Detection Kit, GBI Labs) and TSA-fluorophores (Opal 7-color IHC Kit, NEL797001KT, 1:100, 20–60 seconds, PerkinElmer), after which the primary and secondary antibodies were thoroughly eliminated by heating the slides in citrate buffer (pH 6.0) for 10 minutes at 95°C using a microwave. In a serial fashion, each antigen was labeled by distinct fluorophores. Nuclei were subsequently visualized with DAPI (1:2,000, ST540), and the slides were coverslipped using ProLong Gold Antifade Mountant (P36934, Thermo Fisher Scientific). Using this Opal protocol, CD20 (ab9475, 1:200, Abcam, Opal690), PD-L1 (2130WPI1A02A, WuXi Diagnostics, Opal620), PD-1 (43248, 1:100, CST, Danvers, Opal520), and CD3 (ab16669, 1:2,000, Abcam, N430) were sequentially detected within sample sections.

For each patient sample, 5–10 fields of view were acquired at 20× resolution as multispectral images by using the Vectra Polaris (PerkinElmer), depending on tumor size. All the immune cell populations from each panel were characterized and quantified using the cell segmentation and phenotype cell tool by the inform Advanced Image Analysis Software (PerkinElmer) version 2.4. Malignant cells were identified by their histomorphologic features and positive staining for CD20 and validated by professional pathologists. Fifty CD20⁺ cells per patient were counted, and the percentages of positive PD-L1 expression in CD20⁺ malignant cells were measured (1, >0% to <30%; 2, ≥30% to <70%; or 3, ≥70% to ≤100%).

The peripheral blood was collected in sodium heparin anticoagulant tubes at protocol-specific timepoints. Flow cytometric analyses of peripheral T cells were performed on a BD FACSCalibur Flow Cytometer (BD Biosciences). For the intracellular cytokine expression, blood cells were stimulated with T-cell stimulation cocktail (including PMA, ionomycin, and transport inhibitors, 00-4975-93, eBioscience) for 4-hour incubation. The cells were stained with anti-CD3-PerCP (347344, BD Biosciences), anti-HLA-DR (559866, BD Biosciences), anti-IFNγ (554700, BD Biosciences), and isotype-matched antibodies (all purchased by BD Biosciences). Stained cells were detected by flow cytometry to collect a minimum of 10,000 CD3⁺ lymphocytes, and data were analyzed by FlowJo v10.4.

All the blood samples were collected in nonheparinized tubes and allowed to clot at room temperature for 2 hours and then centrifuged at 1,500 g for 10 minutes. The serum samples were absorbed from the coagulated blood and stored at −80°C. The serum cytokines and chemokines were analyzed by LEGENDplex Bead-based Immunoassays (BioLegend) according to the manufacturer's instructions. The human CD8/NK Panel (740267, BioLegend) was used to simultaneously quantify 13 serum cytokines/chemokines, including IL2, 4, 6, 10, 17A, IFNγ, TNFα, soluble Fas (sFas), soluble FasL (sFasL), granzyme A, granzyme B, perforin, and granulysin. The Human Inflammation Panel (740808, BioLegend) was used to analyze 13 human serum inflammatory cytokines/chemokines, including IL1β, IFNα2, IFNγ, TNFα, CCL2, IL6, IL8, IL10, IL12p70, IL17A, IL18, IL23, and IL33. Data acquisition was performed on a BD FACSCalibur Flow Cytometer (BD Biosciences) and analyzed with the LEGENDplex Data Analysis Software (BioLegend).

Sample size for the primary efficacy endpoint was estimated using A'Hern single-stage phase II design. The null hypothesis of a proportion of patients with an objective response of 40% or lower is based on historic data from a previously conducted study of anti-PD-1 monotherapy in rrPMBCL (7). With 25 patients evaluable for the primary endpoint, there was a two-sided type I error of 0.05 and 90% power to detect a 70% or higher ORR. Twenty-seven patients were enrolled to obtain a planned sample size of 25 patients if 8% of enrolled patients would not be treated.

Patients underwent follow-up for analysis until March 18, 2020. All the patients who received at least one dose of the combined treatment were included in the safety analysis. The patients who received at least one radiological evaluation were included in the efficacy analyses. Descriptive statistics and Wilson method were used to summarize the proportion of patients with response and to calculate the 95% confidence intervals (CI), respectively. Differences among subgroups were evaluated by Fisher exact test for categorical variables and t test or Wilcoxon–Mann–Whitney U test for continuous variables. Differences between treatment cycles were detected by paired t tests. All time-to-event endpoints (PFS, DOR, TTR, and OS) were estimated by the Kaplan–Meier method, and differences between subgroups were assessed by log-rank test. All statistical analyses were performed in Stata/SE 15.1.

Between July 21, 2017, and September 5, 2018, 27 patients with rrPMBCL were enrolled. All patients received at least one evaluable posttreatment tumor CT scan. Median age was 30 years (range, 18–45), 52% were female. Most patients had advanced disease with bulky tumors (≥7.5 cm; n = 16; 59%), elevated lactate dehydrogenase (LDH; n = 22; 81%), and stage IV disease (n = 14; 52%). The characteristics of all patients are summarized in Table 1 and Supplementary Tables S1and S2.

The camrelizumab plus GVD chemotherapy regimen was the fourth- or later-line of therapy of 17 (63%) patients. The median cycle of previous rituximab-containing regimens was seven (range, 2–11), and 24 (89%) patients had primary refractory disease. Most patients were assessed as being transplant ineligible because of chemotherapy resistance, and 19 (70%) had DA-EPOCH-R regimen, 10 (37%) had previous radiotherapy, 2 (7%) had prior ASCT, and 3 (11%) had anti-CD19 CAR T treatment.

With a median follow-up of 24.8 months (range, 3.2–32.4 months), the median treatment cycle of the combined regimen was eight (range, 2–12). At data cutoff on March 18, 2020, 3 (11%) of 27 patients remained on maintenance treatment; 10 (37%) finished both the treatment phases; 9 (33%) had discontinued treatment but remained on follow-up, and 5 (19%) had died of disease progression after discontinuation of treatment (Fig. 1; Supplementary Table S3).

Among all the 27 response evaluable patients, the ORR was 74.1% (95% CI, 55.3–86.8), with 55.6% (95% CI, 37.3–72.4) patients achieving a CR. The ORR was higher in females than males (P = 0.033; Fig. 2). Overall patients receiving camrelizumab plus GVD earlier in their course fared better (P = 0.0026; Fig. 2). Bulky disease was associated with CR (P = 0.0005), but not with objective response (P = 0.183; Fig. 2). Only 1 patient converted from PR to CR during maintenance treatment. Patients with higher PD-L1 expression showed a trend to relatively more frequent CR (P = 0.056; Fig. 2). Maximal tumor regression from baseline is shown in Fig. 3 and Supplementary Fig. S1, and radiographic data pre- and posttreatment are exhibited in Supplementary Fig. S2.

The median time to first objective response was 1.7 months (95% CI, 1.6–1.8), while 21 [77.8% (95% CI, 59.2–89.4)] patients had a decrease in target lesions at the first CT scan assessment (Fig. 3). Median TTRs were similar between subgroups, but significantly shorter in patients without prior radiotherapy (P = 0.006; Supplementary Fig. S3A).

The median DORs were not reached (95% CI, 8.6 to not reached), with 70% (95% CI, 45.1–85.3) of patients having response durations of ≥12 months. At data cutoff, 13 of 15 (86.7%) complete responders had been in sustained remission for 17–30 months (Fig. 3). The 12-month and estimated 24-month PFS rates were 55.6% (95% CI, 35.2–71.8) and 48.2% (95% CI, 28.7–65.2), respectively, with a median PFS of 15.4 months (95% CI, 4.6 to not reached; Supplementary Fig. S3). Median OS was not reached, while estimated 2-year OS rate was 81.5% (95% CI, 61.1–91.8; Fig. 3). A post-hoc exploratory analysis showed that male and bulky disease were associated with PFS (Supplementary Fig. S3).

Twenty-four baseline biopsy samples were available for pathologic and IHC analyses, and 17 for Opal multiplex immunofluorescence staining. Several samples were not evaluable due to inadequate sampling or nonrepresentative tissue. Patients with higher PD-L1 expression showed a trend to relatively more frequent CR, but failed to reach statistical significance (P = 0.12; Fig. 4).

The activation of peripheral T cells was determined by measuring the proportion of HLA-DR–positive T cells. The trend for increased activated T cells was evident in responders, as shown by flow cytometry during the first five doses of the combined regimen (Fig. 4). Prespecified exploratory analysis assessed the association between the outcomes and serum cytokines and chemokines. Stratification by clinical response indicated the potential biomarker roles of the baseline levels of IL10, IFNγ, and sFas and the posttreatment frequency of MDSCs in clinical response, and IL1β, IL33, and Granzyme B in CR (Fig. 4; Supplementary Fig. S4).

Any grade treatment-related AEs occurred in 25 (93%) patients, and grade 3 treatment-related AEs occurred in 9 (33%) patients (Table 2). No grade 4 or 5 AEs were reported (Table 2; Supplementary Table S4). The most frequently reported grade 3 AEs comprised hematologic toxicities of neutropenia (19%) and leukocytopenia (19%). Capillary hemangioma was only reported in 1 patient (grade 1, 4%). The grade 1–2 chemotherapy-related AEs were well-tolerated, and grade 3 AEs resolved with supportive care.

Immunotherapy-related AEs were reported in 15 (56%) patients, the most common being pruritus (30%; Supplementary Table S5). Two patients discontinued treatment because of grade 2 (n = 1) and grade 3 pneumonitis (n = 1; Supplementary Fig. S5). The other immunotherapy-related toxicities were well-tolerated and did not delay treatment.

In this single-arm, phase II trial, camrelizumab plus GVD chemotherapy exhibited a high antitumor activity in patients with heavily pretreated, rapidly clinically deteriorating rrPMBCL. The ORR and CR rate were 74.1% and 55.6%, respectively. The combined therapy demonstrated a more rapid and durable response with a 1.7-month median TTR. With a median follow-up of 24.8 months, the median DOR and OS were not reached, median PFS was 15.4 months and 13 CR patients maintained their remission. The combination of camrelizumab and GVD chemotherapy was indicated to have synergistic antitumor effect in achieving the very encouraging response rate observed.

The usual protocol for rrPMBCL is salvage high-dose chemotherapy followed by ASCT. Chemotherapy response before transplant has been regarded as the most relevant predictor of treatment success. Two-year PFS rates of chemotherapy responsive and refractory patients with rrPMBCL were 50% and 0%, respectively (5). But, the lack of clinical trials focusing on rrPMBCL make it difficult to select a specific chemotherapy regimen for this patient population. Notably, all enrolled patients in our study were highly refractory to prior salvage chemotherapies and ASCT ineligible at enrollment. Utilizing GVD regimen to introduce two novel agents, gemcitabine and vinorelbine, could reduce risk of cross-resistance. Meanwhile, PLD exhibited favorable pharmacokinetics and pharmacodynamics, and limited hematopoietic toxicity and preferential accumulation in tumors (14, 20). Notably, gemcitabine and PLD may affect the antitumor effect through depletion of immunosuppressive MDSCs and induce immunogenic cell death, respectively (12–15). Both could further boost antitumor immunity potentiated by PD-1 blockade. Thus, the combination of PD-1 blockade with GVD chemotherapy provides a rational choice for patients with this otherwise fatal condition. Given the combined effect of GVD of direct cytotoxic damage and antitumor immune response, the administration of GVD regimen lasted relatively long until a second confirmed CR or up to 12 cycles. Indeed, most patients entered the maintenance phase after eight cycles (second confirmed CR; n = 15; 55.6%). Only 3 (11.1%) of 27 patients finished 12 cycles, and 1 (3.7%) patient discontinued treatment after 10 cycles because of disease progression. The toxicity profile of GVD chemotherapy was tolerable and manageable, and the leukocytopenia ameliorated without treatment delay.

Camrelizumab, a PD-1 mAb, has proven to achieve similar response rates with other PD-1 inhibitors in relapsed/refractory classical Hodgkin lymphoma and recurrent or metastatic nasopharyngeal carcinoma (15, 17). The traditional dosing paradigm of therapeutic mAbs is usually given based on body weight. But, anti-PD-1 antibody at fixed dose has been confirmed to reach the comparable clinical efficacy if given based on body weight. The 200 mg every 3 weeks fixed dose is now selected for evaluation across pembrolizumab protocols independent of the tumor type. Meanwhile, previous study of camrelizumab demonstrated a comparable exposure following a fixed dose (200 mg) and a weight-based dosing (3 mg/kg). Pharmacodynamics of camrelizumab revealed that the fixed dose of 200 mg could realize sufficient receptor occupancy saturation (15, 18, 21). Furthermore, fixed dose has advantages of reduced dosing complexity and reduced potential of dosing errors. The 200 mg every 3 weeks fixed dose was selected for this phase II trial according to the previous reports and the manufacture's recommendation.

Given the rare and aggressive nature of rrPMBCL, prospective randomized clinical trials are challenging, and few studies have examined the efficacy of salvage regimens in this context (1, 2). Several small clinical trials of new therapeutic targets have been reported with unsatisfactory response rates (22–25). PD-1 blockade has been approved for rrPMBCL treatment, but it is not recommended as a life-saving regimen for patients with rrPMBCL with rapid clinical deterioration (7, 8). A recent phase II study evaluating nivolumab plus brentuximab vedotin in rrPMBCL demonstrated a high ORR of 73% and a CR rate of 37%, which are similar rates to those reported in our trial (26). CD30 expression in PMBCL tumor cells is relatively weak and heterogeneous, and is negative in approximately 20% of patients, suggesting this group of patients would not benefit from the combination of anti-PD-1 antibody and brentuximab vedotin (27). Moreover, brentuximab vedotin has not been approved and is unavailable in China. The patient population of Zinzani and colleagues was slightly older (median age, 35.5 years), less heavily treated (median, two prior lines of therapy) than in our study, and the majority of the patients were Caucasians (87%; ref. 26).

Our study indicated that bulky disease is correlated with CR and PFS in patients with rrPMBCL. The common definition of bulky disease in Caucasians is 10 cm or more. Considering the overall smaller body size of the Chinese individuals compared with the population, the definition of 10 cm is not appropriate and could be life-threatening. In our study, superior vena cava syndrome occurred in 15 of 16 patients (93.8%) with bulky disease (≥7.5 cm), but only 2 of 11 patients with nonbulky disease (<7.5 cm). Thus, the standard of bulky disease of aggressive B-cell lymphoma (7.5 cm) was used in our study (28).

The safety profile was predictable and manageable, and was similar with known toxicities of anti-PD-1 antibody and GVD chemotherapy alone. The combination did not appear to exacerbate treatment-related AEs (8, 10). The higher grade (grade 3) AEs were mainly contributed to chemotherapy. The known off target effect of camrelizumab as potent agonist to human VEGFR2 would result in the frequent AE capillary hemangioma, which has been observed in our previous trial with classical Hodgkin lymphoma (17, 29). Of note, capillary hemangioma was only reported in 1 patient in our study, consistent with the combined regimen of camrelizumab and chemotherapy in patients with nasopharyngeal carcinoma (15). The lower incidence and severity might be attributed to the inhibitory effect of chemotherapy on hyperproliferative endothelial cells. The detailed mechanism needs to be further explored.

PD-L1 expression has been identified as a biomarker with clinical benefit from checkpoint blockade; however, the absence of PD-L1 expression is not an absolute indicator of the lack of clinical response (30, 31). A potential correlation of clinical response and outcome with higher PD-L1 expression on tumor cells was exhibited in our data, although the number of patients is too small to make definite conclusions. Furthermore, unlike other lymphomas, prognostic biomarkers are largely lacking in PMBCL. We performed the exploratory profile analysis to identify potential biomarkers during this trial. Baseline expression levels of serum cytokines and chemokines, including IL10, IFNγ, and sFas, and the posttreatment frequency of MDSCs have been suggested to offer a potential predictive role. Although our exploratory biomarker analysis was insufficiently powered, subsequent clinical trials should pursue validation of our preliminary correlations.

Our trial has certain limitations. The enrolled patient number is relatively small and randomization with a placebo arm was not feasible because of the difficulty to perform PMBCL-specific studies. Because of its rarity and poor survival at relapse, a prospective clinical trial of PMBCL is challenging, and previous results were obtained from cohorts comprising diverse hematologic malignancies and so-called “basket-trials” (2, 27). In addition, biopsy samples and peripheral blood samples could not be obtained in all patients, rendering the biomarker analysis somewhat limited. Nevertheless, our results are encouraging and call for evaluation of the camrelizumab plus GVD regimen in larger rrPMBCL cohorts with longer follow-up to confirm the high response rates and predictive biomarkers identified herein.

In conclusion, in patients with heavily pretreated rrPMBCL, camrelizumab plus GVD chemotherapy achieved impressively high ORR and CR rates. No new safety signal was recorded, and AEs were expected and well managed with supportive care. Longer follow-up will be undertaken to assess the DOR in this young patient cohort. The combination of camrelizumab and GVD should be considered for patients with bulky aggressive rrPMBCL.

No potential conflicts of interest were disclosed.

Q. Mei: Conceptualization, data curation, formal analysis, supervision, funding acquisition, methodology, writing-original draft, writing-review and editing. W. Zhang: Data curation, formal analysis, investigation, methodology. Y. Liu: Resources, data curation, validation. Q. Yang: Conceptualization, formal analysis, supervision, methodology. J.E.J. Rasko: Supervision, writing-original draft, writing-review and editing. J. Nie: Data curation, investigation. J. Liu: Data curation, formal analysis, investigation. X. Li: Data curation, formal analysis, investigation. L. Dong: Data curation, formal analysis, investigation. M. Chen: Data curation, formal analysis, investigation. Y. Zhang: Data curation, formal analysis, investigation. L. Shi: Data curation, formal analysis, investigation. H. Wu: Data curation, formal analysis. W. Han: Conceptualization, supervision, funding acquisition, validation, writing-review and editing.

The authors thank the patients who participated in the study, their families, and the study nurses for their support in executing the study. This study was funded by Leading Talents Grant of Science & Technology from Beijing (Z181100006318004 to W. Han), National Natural Science Foundation of China (81773248 to Q. Mei and 31991171 and 81830002 to W. Han), National Key Research and Development Program of China (2016YFC1303501 and 2016YFC1303504 to W. Han), and Fostering Funds of Chinese PLA General Hospital for National Excellent Young Scholar Science Fund (2017-YQPY-003 to Q. Mei).

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.