A phase 1 study of blinatumomab/lenalidomide in relapsed/refractory B-cell lymphoma: toxicity, efficacy, and correlative analysis

Key Points

• •Blinatumomab and lenalidomide were well tolerated, and produced encouraging responses and durable remissions in heavily pretreated B-NHL.
• •Correlative analysis provides a better understanding of the mechanism of action and identification of predictive biomarkers.

Visual Abstract

Abstract

Despite a recent increase in therapeutic options, patients with relapsed/refractory B-cell non-Hodgkin lymphoma (R/R B-NHL) eventually require novel therapies. We conducted a phase 1 trial of blinatumomab and lenalidomide in R/R B-NHL. Three dose levels representing 2 schedules were explored. The primary end points were adverse events (AEs) and determining the maximum tolerated dose (MTD)/recommended phase 2 dose (RP2D). Thirty-five patients were enrolled, and 34 patients initiated treatment with a median number of prior regimens of 3 (range, 2-8). There were no dose-limiting toxicities (DLTs) in the first 2 dose levels. Dose level 3, 20 mg of lenalidomide daily on days 1 to 21 and days 29 to 49 of a 56-day induction cycle plus blinatumomab 9 μg/d continuous IV infusion (CIVI) on days 1 to 7, 28 μg/d CIVI on days 8 to 14, and 112 μg/d CIVI on days 15 to 56 was determined to be the MTD/RP2D. The most common grade ≥2 AE was neurotoxicity in 11 of 34 patients (32%), with 4 of 16 patients (25%) at the RP2D. At the RP2D, there was 1 DLT, a patient with grade 2 tremor and word-finding difficulty. For all patients completing induction, the overall response rate was 80% (95% confidence interval, 56-94) with a complete response rate of 70%, and 8 of 34 patients (24%) had durable remissions lasting >2 years. GranB⁺ CD56bright CD16dim CD11b⁺ natural killer cells and memory regulatory T cells in the peripheral blood at baseline were predictive of response. Concomitant administration of lenalidomide appeared to reduce blinatumomab-mediated T-cell exhaustion. In conclusion, encouraging activity was seen with blinatumomab and lenalidomide in heavily pretreated R/R B-NHL (NCI Protocol no. 9924).

Introduction

Despite the high response rates with frontline chemoimmunotherapy in relapsed/refractory B-cell non-Hodgkin lymphoma (R/R B-NHL), and the development of novel targeted therapeutics, many patients will still relapse and die from their disease.^1,2 Even with the expanding armamentarium in B-NHL, for many patients the backbone of salvage treatment includes chemotherapy. Given the median age of B-NHL diagnosis is 67 years,³ and that patients may have significant comorbidities, chemotherapy-free approaches may be advantageous.

Blinatumomab is a bispecific antibody (BSA) that recruits and activates CD3⁺ cytotoxic T cells, redirecting them to CD19-expressing B-NHL cells, ultimately targeting CD19⁺ tumor cells for destruction.⁴ It is approved by the US Food and Drug Administration for the treatment of acute lymphoblastic leukemia. High-dose blinatumomab can induce responses in R/R B-NHL, with an overall response rate (ORR) of 8 of 23 (34.8%), and a complete response rate (CRR) of 4 of 23 (17.4%)⁵; however, the median progression-free survival (PFS) is only 3.7 months (95% confidence interval [CI], 1.4-7.7).⁵ Clinical benefit may be limited by the inability to recruit competent cytotoxic T cells or inevitable T-cell exhaustion.^6,7 Bispecific-mediated T-cell activation and exhaustion has been of particular interest given its role in ineffective T-cell responses either acutely or from subsequent chimeric antigen receptor (CAR) T-cell–mediated therapies.⁸ We hypothesized that immunomodulatory agents would enhance recruitment and activation of competent T and natural killer (NK) cells, reduce T-cell exhaustion, and enhance the efficacy of blinatumomab.

Lenalidomide exerts direct antiproliferative effects on lymphoma cells while possessing immunomodulatory properties that upregulate T- and NK-cell mediated responses against lymphoma.^2,9,10 Given these properties, lenalidomide combined with other antibody-based therapies and bispecifics, such as rituximab and mosunetuzumab, have demonstrated remarkable activity in a wide range of R/R B-NHL subtypes even in patients that are rituximab refractory.^2,11,12 This enhanced activity is believed to be through increased antibody-dependent cellular cytotoxicity via enhanced T-cell and NK-cell recruitment and activation.¹³

We hypothesized that lenalidomide would augment the antilymphoma activity of blinatumomab via amplification of host immune effector cells interacting with the BSA against B-NHL. We report the results of a phase 1 study examining the combination of higher dose blinatumomab and lenalidomide in R/R B-NHL. In addition, we present correlative immunophenotypic data as candidate predictive biomarkers that describe potential mechanisms of activity and may support further studies of BSA combinations in B-NHL.

Methods

Study design and patient population

We conducted a multicenter phase 1 trial through the National Cancer Institute Experimental Therapeutics Clinical Trials Network using blinatumomab and lenalidomide in adult patients with histologically confirmed CD19⁺ R/R B-NHL after ≥2 prior lines of therapy. A full list of inclusion and exclusion criteria is included in the protocol (supplemental Appendix 1).

Study procedures

The trial was registered at (ClinicalTrials.gov identifier: NCT02568553). The trial consisted of a dose escalation phase followed by a dose expansion phase at the maximum tolerated dose (MTD)/recommended phase 2 dose (RP2D) of lenalidomide. All patients received blinatumomab 9 μg/d continuous IV infusion (CIVI) on days 1 to 7, 28 μg/d CIVI on days 8 to 14, and 112 μg/d CIVI on days 15 to 56. Patients on the dose level 1 cohort received lenalidomide 10 mg daily on days 29 to 49, and on dose level 2 received 20 mg daily on days 29 to 49. After noting an excess number of patients with early progression prior to day 29, the protocol was amended to include a third dose level, where lenalidomide 20 mg was given concurrently with blinatumomab on days 1 to 21 and days 29 to 49 (supplemental Figure 1). The induction cycle was 56 days long, and was followed by a 28-day rest period. Responding patients received 6 cycles of consolidation with blinatumomab 112 μg/d CIVI on days 1 to 7, and lenalidomide at the dose used in induction (20 mg at the MTD) orally daily on days 1 to 21, followed by a maximum of 26 cycles (2 years) of maintenance with lenalidomide at the same planned dose used during induction. Treatment continued up to 2 years, or until unacceptable toxicity or disease progression. Safety rules during escalation and expansion were based on preventing an excessive number of dose-limiting toxicities (DLTs). A full description of DLT definitions is listed in the protocol (supplemental Appendix 1).

Patients were admitted for and were premedicated with dexamethasone 20 mg orally 6 to 12 hours prior to initiation, and 10 mg IV 30 minutes prior to initiation of blinatumomab. Given that patients who completed induction had higher ORRs (16 of 20 [80%], including 14 of 20 [70%] complete response [CRs]), the protocol was subsequently amended to allow for continued use of dexamethasone during blinatumomab infusion at the discretion of the treating investigator, and was recommended in the case of steroid-responsive central nervous system-related adverse events (AEs). AEs were graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events version 4.0 until 31 March 2018, after which Common Terminology Criteria for Adverse Events version 5.0 was used.

Response assessment was performed using computerized tomography or positron emission tomography/computerized tomography every 8 weeks during induction and consolidation, and every 16 weeks during maintenance without centralized review using the Lugano 2014 criteria.¹⁴

Clinical outcome analysis

The primary end points were AEs and determining the MTD/RP2D during induction as per the phase 1 queue modified 3+3 design.¹⁵ Preplanned secondary end points were ORR and CRR. For a preliminary estimate of the antitumor activity of blinatumomab and lenalidomide, the protocol required ≥12 patients who started treatment with lenalidomide at the MTD. Exploratory end points were PFS and overall survival (OS), which were calculated by the Kaplan-Meier method. Baseline characteristics were summarized using descriptive statistics.

Correlative analysis

Means or medians, as appropriate, were plotted along with CIs, ranges or interquartile ranges; boxplots were also plotted. Samples for analysis of plasma cytokine levels and subset populations of T, B, and NK cells were collected at baseline and serially (CONSORT diagram, supplemental Figure 2). To quantitatively evaluate the patterns over time, a mixed effects linear regression model was constructed using immune cells subsets and geometric mean fluorescence intensity as dependent variables. Time (as a categorical variable) and response to therapy were set as fixed effects, as was the interaction term, and patient was set as random effect. To control for multiple testing, the Benjamini-Hochberg false discovery rate correction was applied. Exploratory and descriptive analyses were also planned to identify patterns that would merit further study with the goal of identifying a potential predictive biomarker. Additional methods are detailed in supplemental Appendix 2, supplemental Methods.

This trial was approved by the institutional review board, and conducted according to the principles of the Declaration of Helsinki and International Conference on Harmonization Good Clinical Practice guidelines. All patients provided written informed consent prior to enrollment.

Results

Participant characteristics

Thirty-five patients enrolled and 34 patients initiated protocol therapy. This report is based on the 34 patients who initiated treatment. Seven patients were treated at dose level 1, 11 at dose level 2, and 16 at dose level 3 of lenalidomide (see patient characteristics in Table 1). The median number of prior regimens was 3 (range, 2-8). Diffuse large B-cell lymphoma was the most common histology (n = 15), followed by mantle cell lymphoma (n = 6). Seven (21%), 9 (26%), and 3 (9%) patients previously received CAR T-cell therapy, autologous hematopoietic cell transplant, and lenalidomide, respectively.

Table 1.

Baseline characteristics at time of enrollment

Variable	All patients, N = 34 (%)	Dose level 1, N = 7 (%)	Dose level 2, N = 11 (%)	Dose level 3, N = 16 (%)
Median age, y	61 (22-84)	53 (30-65)	66 (30-84)	62 (22-73)
Gender
Male	23 (68)	5 (71)	6 (55)	12 (75)
Female	11 (32)	2 (29)	5 (45)	4 (25)
Race/ethnicity
Non-Hispanic White	24 (71)	7 (100)	7 (64)	10 (62)
Hispanic White	5 (15)	0	3 (27)	2 (13)
African American	2 (6)	0	0	2 (13)
Asian	2 (6)	0	0	2 (13)
Unknown	1 (3)	0	1 (9)	0
ECOG performance
0	18 (53)	3 (43)	7 (64)	8 (50)
1	12 (35)	4 (57)	1 (9)	7 (44)
2	4 (12)	0	3 (27)	1 (6)
Diagnosis
DLBCL	15 (44)	4 (57)	2 (18)	9 (56)
MCL	6 (18)	2 (29)	1 (9)	3 (19)
Unspecified B-NHL	5 (15)	0	2 (18)	3 (19)
FL	4 (12)	0	3 (27)	1 (6)
BL	1 (3)	0	1 (9)	0
MZL	1 (3)	0	1 (9)	0
PMBL	1 (3)	1 (14)	0	0
SLL	1 (3)	0	1 (9)	0
Primary site of disease
Lymph node	22 (65)	3 (43)	8 (73)	11 (69)
Bone marrow	3 (9)	0	0	3 (19)
Liver	2 (6)	0	1 (9)	1 (6)
Breast	1 (3)	0	1 (9)	0
Eye	1 (3)	1 (14)	0	0
Leg	1 (3)	0	0	1 (6)
Nasopharynx	1 (3)	0	1 (9)	0
Peritoneum	1 (3)	1 (14)	0	0
Rectum	1 (3)	1 (14)	0	0
Spleen	1 (3)	1 (14)	0	0
Median prior treatments	3 (2-8)	3 (2-3)	3 (2-4)	4 (2-8)

BL, Burkitt lymphoma; DLBCL, diffuse large B-cell lymphoma; ECOG, Eastern Cooperative Oncology Group; FL, follicular lymphoma; MCL, mantle cell lymphoma; MZL, marginal zone lymphoma; PMBL, primary mediastinal B-cell lymphoma; SLL, small lymphocytic lymphoma.

Safety

Table 2 summarizes the DLTs observed during study. The MTD/RP2D was dose level 3 (concurrent blinatumomab and lenalidomide, see the schema in supplemental Figure 1) where 16 patients initiated treatment. Of the 16 patients, 5 were not evaluable for DLT, 3 were taken off early for progressive disease (PD) without a DLT-qualifying AE but not adequately treated (≥90% of the planned dose of each drug) to be evaluable for DLT per protocol and were replaced, 1 was taken off treatment prior to completing the requisite amount of treatment to be evaluable for a DLT due to an unrelated Strongyloides infection, and 1 patient did not receive the mandatory steroid prophylaxis (patient 21 in supplemental Table 1 on neurotoxicity) and was replaced as the patient was not treated per protocol. See the consort diagram (supplemental Figure 2) for the number of patients treated and evaluable for DLT at the 3 dose levels. A summary of treatment-related AEs at the RP2D is found in Table 3 (grade ≥2). The patient with grade 4 sepsis was determined to have a Strongyloides parasitic infection, and the only other grade 4 events were hematological (5 absolute neutrophil count, 2 lymphocyte count decreased). As noted in supplemental Table 1 on neurotoxicity, during induction there were 3 patients with grade 3 neurotoxicity (including 1 patient that did not receive steroid prophylaxis as specified by protocol), and 5 patients with grade 2 neurotoxicity at the RP2D. One additional patient experienced grade 3 confusion on the RP2D during consolidation. The single DLT was a patient with a grade 2 neurotoxicity that required the cessation of therapy.

Table 2.

Dose levels, DLT evaluability, and responses per dose level

Dose level based on lenalidomide∗	No. of patients treated	No. of patients excluded from course 1 DLT evaluation	No. of patients with DLTs	DLT description	Best responses during therapy
DL1: Len 10 mg, days 29-49	7	4	0	—	CR-3 PR-0 PD-3 NA-1
DL2: Len 20 mg, days 29-49	11	6	0	—	CR-4 PR-2 SD-1 PD-3 NA-1
DL3: Len 20 mg, days 1-21, 29-49	16	5	1	Grade 2 neurotoxicity stopping therapy	CR-7 PR-0 SD-1 PD-6 NA-2

Len, lenalidomide; N/A, not applicable; PR, partial response; SD, stable disease.

^∗For induction, all patients received blinatumomab 9 μg/d CIVI on days 1 to 7, 28 μg/d CIVI on days 8 to 14, and 112 μg/d CIVI on days 15 to 56. The induction cycle was 56 days long, and was followed by a 28-day rest period. Responding patients received 6 cycles of consolidation with blinatumomab 112 μg/d CIVI on days 1 to 7, and lenalidomide at the dose used in induction days 1 to 21 on a 28-day cycle.

Table 3.

List of grade 2 and 3 treatment-related AEs in the 16 patients treated at the RP2D (dose level 3)

AE	Grade 2	Grade 3	Grade 4
Neutrophil count decreased	2 (13%)	2 (13%)	5 (31%)
Lymphocyte count decreased	1 (6%)	1 (6%)	2 (13%)
Sepsis		1 (6%)	1 (6%)
Neurotoxicity NOS	2 (13%)	2 (13%)
Tremor		2 (13%)
Fatigue		2 (13%)
Anemia	2 (13%)	1 (6%)
White blood cell decreased	2 (13%)	1 (6%)
Dysarthria	1 (6%)	1 (6%)
Lung infection		1 (6%)
Port site infection		1 (6%)
Deep venous thrombosis		1 (6%)
Ataxia		1 (6%)
Confusion		1 (6%)
Gait disturbance		1 (6%)
Febrile neutropenia		1 (6%)
Aminotransferase increased		1 (6%)
Fever	5 (31%)
Vomiting	2 (13%)
Encephalopathy	2 (13%)
Mucositis oral	1 (6%)
Rash maculo-papular	1 (6%)
Emotional lability	1 (6%)
Pseudomonas wound infection	1 (6%)
Systemic inflammatory response syndrome	1 (6%)
Diarrhea	1 (6%)
Headache	1 (6%)
Vertigo	1 (6%)
Dysgeusia	1 (6%)
Myalgia	1 (6%)
Platelet count decreased	1 (6%)
Platelet count decreased	1 (6%)

Excludes single occurrence of grade 2 hypertension, hypokalemia, lipase increase, reflux disease, and pain.

NOS, not otherwise specified.

Grade ≥2 treatment-related AEs in the 34 patients can be seen in supplemental Table 2 (Toxicity for all dose levels). One of 7 patients treated on dose level 1 (14.3%) experienced grade 3 neurotoxicity (confusion), and 6 of 11 (54.5%) treated on dose level 2 experienced grade 3 neurotoxicity (3 confusion, 1 confusion with encephalopathy, 1 tremor, 1 dysphasia), and as noted previously 4 of 16 treated at dose level 3 (25.0%) experienced grade 3 neurotoxicity as previously noted, including 1 patient during consolidation.

Efficacy

The median follow-up for the entire cohort was 32.3 months (95% CI, 16.4-43.7). A breakdown of responses and DLTs per dose level is depicted in Table 2. In all patients treated, the ORR was 47% (16/34) and the CRR was 41% (14/34). In the 7 patients that received prior CAR T-cell therapy (all at the RP2D), 2 patients (29%), both of whom had a diffuse large B-cell lymphoma diagnosis, achieved a response (both CR), and all 7 patients had confirmed CD19 expression. In an attempt to continue treatment and prevent neurotoxicity, 20 patients were treated with concomitant steroids during induction, beyond premedication, and had an ORR of 50% suggesting that steroids did not have a significant effect on efficacy (a subset analysis, not for comparison purposes).

At the RP2D, both the ORR and CRR were 44% (7 out of 16 patients). At the RP2D, the median OS and PFS were not reported (NR; 95% CI, 3.6-NR) and 5.6 months (95% CI, 1.8-24.2; Figure 1). For all patients completing induction, the ORR was 80% (95% CI, 56-94) with a CRR of 70%, and 8 (24%) had durable remissions lasting >2 years. The median OS and PFS for the entire 34 patient cohort were 56.6 (95% CI, 6.2 to NR) and 4.1 months (95% CI, 1.8-20.0), respectively (supplemental Figure 3A). The median PFS and duration of response for patients completing induction were 52.0 (95% CI, 10.7 to NR) and 20.0 months (95% CI, 4.4 to NR), respectively (supplemental Figure 3B-C). The swimmer plot (Figure 2) shows that 8 patients (24%) maintained a CR for >2 years (follicular lymphoma, n = 2; small lymphocytic lymphoma, n = 1; mantle cell lymphoma, n = 2; diffuse large B-cell lymphoma, n = 2; B-cell lymphoma not otherwise specified, n = 1).

Figure 1.

Kaplan-Meier estimates of PFS and OS for the 16 patients treated at the RP2D. NR, not reported.

Figure 2.

Swimmer plot. Durable responses were observed on all doses. N/A, not applicable; PR, partial response; SD, stable disease.

Correlative analysis

Blood samples for correlative analyses from 34 patients were drawn at baseline (pretreatment), and on days 15, 29, and 57. This allowed us to measure the relative abundance of the different lymphocyte subsets (eg, NK cells, NK T cells, T cells, and B cells) by flow cytometry. Data for analysis were available for 31 patients at baseline (supplemental Figure 2). To evaluate the effects of lenalidomide and blinatumomab on immune subpopulations, the median percentages of each subpopulation (± interquartile ranges) were graphed against time, and mixed effects linear regression models were constructed as described in the Methods. We have previously reported that in patients with refractory B-cell lymphoma receiving immunotherapy with either rituximab plus ipilimumab¹⁶ or rituximab plus lenalidomide,¹⁷ that B-cell percentage (of total lymphocytes),¹⁶ and peripheral B-cell depletion on day 15 could identify patients likely to have higher and more durable responses to therapy. In the current study, B cells were calculated as the percentage of live CD3⁻ lymphocytes that expressed HLA-DR and CD19 (lymphocyte gating strategy is shown in supplemental Figure 4). Median B cells (± interquartile ranges) were graphed as a line graph to visualize B-cell depletion from baseline to day 15. This analysis revealed that there was no significant difference in B-cell percentage at baseline when patients achieving a CR were compared with those with PD and stable disease (P = .94, Student t test). However, similar to our previous studies,^16,17 on day 15 complete responders had a greater reduction in peripheral B-cell numbers (P ≤ .022 CR; Figure 3A). Although complete responders had an overall reduction in total B cells on day 15, patients that achieved a CR were hallmarked by a paradoxical increase in immunoglobulin D positive (pos) programmed cell death protein 1 (PD-1) pos B cells compared with nonresponders. To evaluate this increase as a means to predict response to therapy, receiver operator characteristic (ROC) curves were constructed, and the areas under the curves (AUC) were calculated. This analysis revealed that an increase in immunoglobulin D pos PD-1 pos B cells could classify complete responders from patients with stable or progressive disease (AUC = 0.75 ± 0.03; supplemental Figure 5C).

Figure 3.

Immunophenotypic subsets predict response to therapy. Responders (blue) vs nonresponders (red). Responder groups consisted of only those who achieved a CR. Nonresponder group consisted of those with stable disease or progressive disease. Displayed are median values with error bars indicating 75th and 25th percentiles. (A) Immunoglobulin D (IgD⁺) PD-1⁺ B-cell frequencies. Line graph with median values indicates that the frequency of IgD⁺ PD-1⁺ B cells increases in the responder group at day 15 and day 29. (B) Line graph shows that responders have a higher frequency of CD56^Bright CD16^Dim NK cells over all time points. (C) Line graph shows that nonresponders have a higher frequency of CD335⁺ NK T cells across all time points. (D) Line graph shows that responders have a higher frequency of memory Tregs across all time points. P values are false discovery rate-corrected for multiple comparisons.

Alterations in immune cell populations with cytotoxic activity

Various immune cell populations have been reported to play roles in anticancer adaptive immune responses. These include interferon-gamma (IFN-γ)–expressing cytotoxic (CD8⁺) T cells, NK cells,^17,18 and NK T cells. Thus, we assess if the lenalidomide and blinatumomab combination therapy induced alterations in any of these immune cell populations, and if this correlated with response.

Results revealed that in addition to the decline in peripheral B cells, lenalidomide and blinatumomab combination therapy increased the percentage of GranB-expressing NK cells (P = 8.7 × 10⁻³; supplemental Figure 6A). The amount of GranB expressed per individual NK cell subpopulation was assessed by measuring GranB geometric mean fluorescence intensity by flow cytometry. This revealed that GranB expression increased over time in various NK cell subpopulations following initiation of therapy (supplemental Table 3), most notably in CD56Dim CD16Bright NK cells (P = 1.4 × 10⁻³; supplemental Figure 6B). In contrast to GranB-expressing NK cells, IFN-γ–secreting CD8⁺ T cells and NK T cells did not significantly increase over time (supplemental Table 4).

Baseline percentage GranB-expressing NK cells can predict response to therapy

Previous immunotherapy trials in R/R NHL have shown that a baseline elevation in GranB-expressing NK cells predicts a CR to therapy. We thus sought to assess the utility of the different NK cell subpopulations to classify response to therapy. NK gating strategy is shown in supplemental Figure 7. Results demonstrate that CD56Bright CD16Dim GranB-expressing NK cells were significantly elevated at baseline and across all time points in patients who had a CR to therapy (P = 2.8 × 10⁻⁵; Figure 3B). To evaluate the ability of CD56Bright CD16Dim GranB-expressing NK cells to classify complete responders, ROC curves were constructed and AUCs were calculated that revealed that the percentage of CD56Bright CD16Dim CD11bpos GranB-expressing NK cells could predict response to therapy (supplemental Figure 8B; AUC = 0.90 ± 0.06). Other GranB-expressing NK subpopulations were also significantly elevated in patients with a CR (supplemental Table 4).

Additional baseline predictors of response to therapy

Interleukin-10 (IL-10) is an inhibitory cytokine that, among other things, inhibits the differentiation of IFN-γ–secreting T cells. In contrast to IL-10, IFN-γ is a proinflammatory cytokine linked to cancer survival. Supplemental Figure 9 demonstrates that the ratio of IL-10 to IFN-γ–expressing CD8⁺ T cells predicts response to therapy. A mixed effects model demonstrated that complete responders tended to have a higher ratio of IFN-γ to IL-10 expressing CD8⁺ T cells, and ROC analysis revealed that this ratio at baseline had utility as a predictive classifier (supplemental Figure 9C; AUC = 0.79 ± 0.03).

The percentage of immune cells expressing inhibitory receptors was also assessed for ability to predict a response to therapy. This analysis revealed that the percentage of CD335 pos NK T cells was significantly lower in complete responders when compared with those with stable or progressive disease (P = 2.4 × 10⁻³), and the percentage of CD335⁺ NK T cells was able to predict response to therapy (AUC = 0.78 ± 0.03; supplemental Figure 8C). Responders also have few CD56Dim CD16Dim NK cells expressing the inhibitory receptor NKG2A (supplemental Figure 10).

In a previous study, patients with R/R NHL that responded to immunotherapy had a high percentage of memory regulatory T cell (Treg) within the total Treg population. Thus, in our current study we measured the percentage of memory Tregs over time. This analysis revealed that across all time points, Tregs of complete responders were predominantly of the memory phenotype (P = 2.0 × 10⁻⁴; Figure 3D). ROC analysis revealed that the percentage of memory Tregs (memory Tregs/total Tregs) could predict response to therapy, that is separate complete responders from patients with stable or progressive disease, P = 1.4 × 10⁻² (ROC AUC = 0.88 ± 0.02 at baseline; supplemental Figure 8D).

T-cell exhaustion

There has been recent interest in the development of T-cell exhaustion and how this effects T-cell–mediated therapeutic efficacy. There has been particular interest in bispecific-mediated T-cell exhaustion and how this may affect subsequent response to CAR T-cell–based therapy.⁸ We assessed T-cell exhaustion markers in patients treated with the combination of blinatumomab and lenalidomide, and compared sequential (DL1-2) vs concurrent (DL3). We found no significant changes in T-cell subpopulations coexpressing PD-1 and T-cell immunoglobulin and mucin-domain containing-3 (supplemental Figure 11). There was a difference at baseline between sequential and concurrent treatment in lymphocyte activation gene-3 (LAG-3) coexpression, and this persisted over time. The persistence of low LAG-3 in the concurrent treatment suggests limited T-cell exhaustion, but requires further evaluation in more balanced baseline groups. The effects were observed in both CD4⁺ and CD8⁺ T cells (Figure 4A); however, the effect was more significant in CD8 T cells at all time points, especially at day 29 (Figure 4B-C; P = .087 vs P < .001 day 29 in CD4 vs CD8, respectively).

Figure 4.

LAG-3 coexpression at baseline, day 15, and day 29. (A) CD3 and LAG-3 coexpression. (B) CD8 and LAG-3 coexpression. (C) CD4 and LAG-3 coexpression. Error bars represent 25th and 75th percentiles. Individual values for each patient are displayed as box-whisker plots. P values not adjusted for multiple testing.

Discussion

BSAs, such as blinatumomab, are a new class of immunotherapeutic molecules intended for the treatment of cancer by recruiting host immune effector T cells. Blinatumomab monotherapy has previously demonstrated efficacy in patients with R/R NHL with some durable CRs.¹⁹ Newer BSAs are being developed with novel platforms, longer half-lives, and improved tolerability. In fact, 3 new BSAs have been US Food and Drug Administration approved for the treatment of R/R NHL over the past few years, and there are many more in development for NHL and other malignancies. Thus, it is important to not only assess BSA-based therapeutic combinations, but to also develop predictive classifiers capable of identifying patients most likely to respond to therapy. Here, we demonstrate encouraging efficacy with the combination of lenalidomide and blinatumomab in R/R NHL. Correlative analysis facilitated a better understanding of the mechanism of action and identification of predictive biomarkers.

Rates of neurotoxicity were similar to previously reported single-agent blinatumomab (22%) vs 25% in dose level 3 of our cohort¹⁹ with the addition of lenalidomide, all were reversible, and all but one were effectively managed with either concomitant or prophylactic steroids. At the RP2D, 44% of patients achieved CR, higher than the prior CRR of 17.4% reported with blinatumomab monotherapy at target dose.⁵ Although we recognize that our median PFS is similar to previously reported with single-agent blinatumomab, it should be noted that this study included more aggressive NHL histologies, and were likely more refractory as CAR T-cell–based therapeutics were not a standard of care option at the time the single-agent blinatumomab trial was conducted.

Patients with B-NHL, particularly those with aggressive histology, who relapse after CAR T-cell therapy have a poor prognosis, and represent an unmet medical need.²⁰ Activity with blinatumomab plus lenalidomide in prior recipients of CAR T-cell therapy was modest, with 2 of 7 achieving a CR in heavily pretreated diffuse large B-cell lymphoma.

Several other studies have examined the combination of a BSA with lenalidomide in B-NHL. Mosunetuzumab is a CD20 × CD3 BSA that as a single agent produced a CRR of 60% and 37% in R/R follicular lymphoma.^21,22 Cross-trial comparisons are problematic and not possible here given the heterogeneous histologies that included heavily pretreated aggressive lymphoma, but our observed 44% CR rate at the RP2D compares well given our patient population. A recent study in R/R diffuse large B-cell lymphoma using epcoritamab and lenalidomide reported an ORR of 67% and a CRR of 51%.²³

Comprehensive flow cytometric analysis was performed to characterize the immunophenotypic profile before, during, and after treatment. This analysis revealed that several immunophenotypic subsets predicted a response to therapy. In our previous 2 studies of rituximab-based combination therapies for R/R NHL, it was observed that peripheral blood B-cell depletion could be used as a surrogate marker for the antibody-dependent cellular cytotoxicity-mediated antilymphoma response. Consistent with these prior studies, peripheral B-cell depletion on day 15 was significant in patients having a CR.

We have previously demonstrated that an increased percentage of GranB-expressing NK cells predicts clinical response to rituximab-based combination immunotherapy.^16,17,24 Results from the current study are consistent with this previous finding. Individuals who experienced a CR had higher percentages of GranB-expressing NK cells. The percentage of immune cells expressing inhibitory receptors was also assessed for ability to predict a response to therapy. This analysis revealed that the percentage of CD335 pos NK T cells was significantly lower in complete responders when compared with those with stable or progressive disease, and the percentage of CD335⁺ NK T cells was able to predict response to therapy.

Peripheral blood Treg population is not homogeneous and can be subdivided by various cell surface markers. In our current study, we demonstrate that patients with a high fraction of memory Tregs predict a response to therapy. Interestingly, the utility of memory Treg percent as a response to therapy classifier was also observed in a completed independent trial of R/R NHL using a different immunotherapy regimen, ipilimumab plus rituximab.¹⁶

Cytokine-expressing T-cell populations were also assessed. For these studies we focused on IFN-γ. IFN-γ–expressing CD8⁺ cytotoxic T cells have significantly higher effectiveness in killing target cancer cells when compared with their nonexpressing counterparts.²⁵ In contrast, IL-10 is strongly immunosuppressive, via its ability to inhibit T helper 1 (and T helper 17) cytokines, including IFN-γ and IL-2. Although studies have demonstrated protumor and antitumor properties of IL-10,²³ in general, IL-10 is considered to be an important growth factor for B-cell lymphomas,^26,27 and a potent suppressor of antitumor immune response.²⁵ It is therefore noteworthy that patients who had a CR to lenalidomide and blinatumomab had a higher ratio of IFN-γ to IL-10–secreting T cells; and at baseline, this ratio had utility as a predictor of response.

Lastly, our 2 induction schedules allowed us to compare the effects of sequential blinatumomab (D1) and lenalidomide (D29) vs concomitant (both given D1). The persistence of low LAG-3 expression in the concurrent treatment implies lenalidomide prevents bispecific-mediated T-cell exhaustion, but this warrants further evaluation, but also reflects on the limits of this study, which includes the small sample size and histological heterogeneity, a nonrandomized design and absence of a comparator arm, lack of central radiology review and extensive number of comparisons in the correlative analysis, which controlled for the false discovery rate, but not for a family-wise error rate (eg, Bonferroni).

In summary, this clinical study in relapsed and refractory B-cell lymphoma demonstrated that the combination of lenalidomide and blinatumomab was mostly safe with manageable toxicity. The CRR was encouraging, and more likely in those with a high percentage of GranB-expressing D56bright CD16dim CD11bpos NK cells and those with a high percentage of memory Treg cells in their peripheral blood. Although based on previous studies, several of the biomarkers reproducibly predict response to therapy, and future studies will be needed to determine the reproducibility of these findings especially in selected histologies. Moreover, the concurrent administration of lenalidomide may mitigate the T-cell exhaustive effects of BSAs that may dampen concerns about subsequent or concurrent treatment with CAR T-cell therapies,^28,29 but validation studies are needed. Given recent studies demonstrating improved tolerance of blinatumomab when administered subcutaneously, there is renewed interest in its development for patients with lymphoma.^26,30 This study, particularly the correlative analysis, may serve as a basis for future clinical trials with this and newer BSA to identify subsets that could respond to the combination of BSA and immunomodulatory agents.

Conflict-of-interest disclosure: J.M.T. reports research funding from Genentech, AbbVie, Pharmacyclics, ADC Therapeutics, Pfizer, Regeneron, and Genmab. C.P. reports advisory board membership with Acrotech, Incyte, Ipsen, AstraZeneca, and Citius. M.A. reports research funding from Orcha Biotech; and speakers' bureau fees from Kite. D.M. reports consulting/advisory roles for ADC Therapeutics, Genmab, Genentech (spouse), Daiichi Sankyo (spouse), and AstraZeneca (spouse); research funding for investigator initiated trials from Genentech, Genmab, Karyopharm, and AstraZeneca (spouse); and expert testimony for AstraZeneca. E.S. reports consulting for D. E. Shaw Research; and advisory board membership for Mallinckrodt Pharmaceuticals. C.C. reports research funding from Bristol Myers Squibb (BMS). M.M. reports consultancy fees from Novartis, Seagen, CTI, ADC Therapeutics, AstraZeneca, and Synthekine; speakers’ bureau fees from Seagen and Incyte; and research funding from BMS, Incyte, BeiGene, and Genentech. W.M.B. reports consulting fees from Guidepoint Global and PPD; and advisory board membership for Pfizer, Genmab, Tempus, Kyowa-Kirin, ADC Therapeutics, and Incyte. A.D. reports consulting fees from AbbVie, ADCT, AstraZeneca, BeiGene, BMS, Genmab, Incyte, Janssen, Lilly Oncology, MEI Pharma, Merck, Nurix, Regeneron, and Roche; and ongoing research funding from AbbVie, AstraZeneca, BeiGene, Genmab, Incyte, Lilly Oncology, Merck, Nurix, and Regeneron. The remaining authors declare no competing financial interests.