Inotuzumab Ozogamicin for Relapsed/Refractory Acute Lymphoblastic Leukemia in the INO-VATE Trial: CD22 Pharmacodynamics, Efficacy, and Safety by Baseline CD22

Acute lymphoblastic leukemia (ALL) is a rare disease; the global incident cases of ALL were 108,000 and 52,000 deaths were caused by ALL in 2017 (1). Multiagent chemotherapy and (for selected patients) allogeneic hematopoietic stem cell transplant (HSCT) comprise the backbone of therapy for ALL, although new therapies for ALL are emerging (2); approximately 50% of adults with B-cell ALL can achieve long-term remission (3).

Cluster of differentiation-22 (CD22) is a B-cell–associated transmembrane glycoprotein with an immunoglobulin domain that binds sialic acid. CD22 is expressed in ≥90% of leukemic lymphoblasts in >90% of patients with B-cell ALL (4–7). Screening for CD22 is recommended as part of the immunophenotyping panel for diagnosis of B-cell ALL (8) and is used for assessment of minimal/measurable residual disease (MRD). CD22 is an attractive target for use in antibody–drug conjugate (ADC) therapies to treat B-cell malignancies (8–11).

Inotuzumab ozogamicin, an anti-CD22 antibody conjugated to calicheamicin, demonstrated significantly greater efficacy versus standard of care (SC) in patients with relapsed/refractory (R/R) B-cell ALL in the INO-VATE trial (12, 13). The rate of complete remission (CR) and CR with incomplete hematologic recovery (CRi) per blinded independent endpoint adjudication committee (EAC) in the first 218 patients randomized was 80.7% [95% confidence interval (CI), 72%–88%] among patients who received inotuzumab ozogamicin versus 29.4% (95% CI, 21%–39%) in the SC group (P < 0.001; ref. 12). Patients treated with inotuzumab ozogamicin achieved a significantly higher MRD negativity rate, progression-free survival (PFS), and 2-year overall survival (OS) in an ad hoc analysis (12, 13).

As a CD22-directed ADC, the cytotoxic drug, calicheamicin, in inotuzumab ozogamicin is delivered to cells expressing CD22 and induces DNA damage and apoptosis (5, 10, 11). Nonclinical findings from a limited number of cell lines showed no apparent relationship between CD22 expression and sensitivity to inotuzumab ozogamicin in vitro (5). Using data obtained from the INO-VATE study, we further assessed the efficacy and safety of inotuzumab ozogamicin versus SC in patients with baseline leukemic blast CD22 positivity ≥90% versus <90% as assessed by a central laboratory. In addition, we performed a quartile analysis to assess potential relationships between CD22 receptor density [as assessed by molecules of equivalent soluble fluorochrome (MESF) at the central laboratory] and efficacy and safety. On the basis of the results of these analyses, a similar quartile analysis was undertaken to assess the relationship between CD22 positivity assessed locally and efficacy and safety.

The gene encoding histone-lysine N-methyltransferase 2A (KMT2A), formerly named the mixed lineage leukemia gene, has been found to exert broad, positive effects on gene transcription. Translocation/rearrangements of the KMT2A gene, including the canonical t(4;11) translocation, are associated with poor prognosis in B-cell ALL (14, 15). Low CD22 positivity (22%–82%) was reported in some patients with R/R ALL who had 11q23 (KMT2A) rearrangement (7). Therefore, the potential relationship between CD22 expression assessed centrally and KMT2A status was explored in this subgroup analysis of data from the INO-VATE study.

Downregulation or loss of surface antigens (e.g., CD19) targeted by various therapeutic modalities, including chimeric antigen receptor T-cell (CART) therapy and bispecific T-cell engagers, has been implicated in acquired resistance to those therapies (16). In pediatric patients with ALL, low baseline CD22 expression has been identified as a potential biomarker of poor response to inotuzumab ozogamicin treatment (17). Thus, analyzing CD22 expression levels (assessed centrally) by response status in patients with R/R ALL treated with inotuzumab ozogamicin or SC (particularly those who responded to treatment and subsequently relapsed) may provide information to facilitate understanding of whether or not relapse might be associated with downregulation or loss of CD22 expression.

Study design, patient population, and treatment groups from the multicenter, global, open-label, randomized (1:1) phase III trial INO-VATE (NCT01564784) have been described in detail previously (12, 13). Patients (age ≥18 years) with R/R CD22-positive B-cell ALL (≥5% marrow blasts assessed by morphology) who were due to receive first or second salvage therapy were randomized 1:1 to receive inotuzumab ozogamicin or SC. CD22 expression was determined using peripheral blood or bone marrow aspirate by local laboratories at screening. The definition of CD22-positive status was amended (June 24, 2013, Protocol Amendment 2) from ≥20% to any percentage of leukemic blasts expressing CD22. Central laboratory results could be considered for eligibility if CD22-negative status had been reported by local laboratories. Flow cytometry or IHC was allowed for CD22 detection.

INO-VATE trial was carried out between August 2012 and January 2017. Final data (last-patient-last-visit: January 4, 2017) were used for this analysis. The study was approved by the independent ethics committee and/or institutional review board at each study center, and conducted in compliance with the Declaration of Helsinki and all International Conference on Harmonisation Good Clinical Practice guidelines. All local regulatory requirements were followed. Written informed consent was obtained from all participating patients.

Bone marrow aspirates were collected at screening (i.e., baseline); days 16–28 of cycles 1, 2, and 3, then every one to two cycles or as clinically indicated; and at end of treatment (EOT, which was within 4 weeks from the last dose of study drug). Bone marrow aspirates would not be expected more than once per cycle, and could be collected less frequently. Samples were analyzed by a central laboratory (Navigate BioPharma) using multiparametric flow cytometry. CD22 expression was detected using a CD22 mouse anti-human mAb (clone RFB-4; PE-conjugate, Life Technologies) as detailed in Jabbour and colleagues' study (18) and CD22 expression in leukemic blasts was quantified as the percentage of leukemic blasts that were CD22 positive and as MESF (i.e., standardized fluorescence intensity units; ref. 19). To quantify the percentage of CD22 expression on leukemic blasts by central laboratory, the leukemic blasts (identification based on leukemia-associated immunophenotypes or aberrant antigen expression) were overlaid on a histogram plot and CD22 positivity was measured using fluorescence minus one as a negative control. For more assay details see Supplementary Materials and Methods.

A peripheral blood sample was used if a patient had an inadequate bone marrow aspirate at screening. If multiple CD22 results existed in a specific cycle for a patient, the latest CD22 result in the cycle was reported.

KMT2A FISH analysis was performed centrally (Navigate BioPharma) using a “break-apart” probe mixture that hybridizes to the KMT2A locus within chromosomal region 11q32. Break-apart signals were scored as KMT2A translocations/rearrangements. Other KMT2A abnormalities, such as changes in 11q copy number, were also detected. KMT2A translocation t(4;11) was also assessed locally by karyotyping.

Details of the efficacy and safety endpoints have been described previously (12). The two primary endpoints in INO-VATE were OS and remission (CR/CRi), assessed by an EAC. The key secondary efficacy endpoints included MRD rate, duration of remission (DoR), and duration of CR, PFS, and patient-reported outcomes.

For this analysis, two parallel analyses were carried out in the intent-to-treat (ITT) population: one by CD22 positivity detected on ≥90% versus <90% of blasts, and one by quartile of MESF. The cut-off point of 90% CD22 positivity was specifically chosen after reviewing the CD22 data from the central laboratory because it divided the patients into two subgroups with sufficient numbers of patients in each group for further analysis. Given the predominance of patients with high percentages of blasts expressing CD22, which was consistent with previously reported data that most patients with B-cell ALL had CD22 expression in ≥90% leukemic lymphoblasts (4, 5, 20, 21), it was not feasible to select a number below 90% as the cut-off point. Similarly, a number more than 90% was not practical as it would not provide enough differentiation. Efficacy endpoints included the rate of CR/CRi and MRD negativity rates among responders. CR was defined as a disappearance of leukemia as indicated by <5% marrow blasts and the absence of peripheral blood leukemic blasts, with recovery of hematopoiesis defined by absolute neutrophil count (ANC) ≥ 1,000/μL, platelets ≥ 100,000/μL, and resolution of any extramedullary disease. CRi was defined as CR except with ANC < 1,000/μL and/or platelets < 100,000/μL. Remission included CR or CRi. Flow cytometry carried out at a central laboratory was used to assess MRD negativity. CD9, CD10, CD13, CD19, CD20, CD33, CD34, CD38, CD45, CD58, CD66c, and CD123 were detected simultaneously to maximize discrimination between normal and abnormal cells of B-cell lineage and similar maturational stage (Supplementary Materials and Methods). In this analysis, MRD negativity was defined as MRD below the minimum detectable MRD threshold, which was 0.01%. MRD status was assessed from MRD tests conducted from post-baseline to EOT +7 days. PFS was defined as time from randomization to earliest date of any of the following events: death from any cause, progressive disease (objective progression, relapse from CR/CRi, and treatment discontinuation due to global deterioration of health status), or starting new induction therapy or posttherapy HSCT without achieving CR/CRi. OS was defined as the time from randomization to date of death from any cause. Patients for whom the date of death could not be verified were censored at date of last contact. DoR was time to progression or death from date of first remission in responders. The final statistical analysis plan of INO-VATE (December 15, 2014) included CR/CRi subgroup analysis by baseline CD22 levels. All other analyses were post-hoc.

Safety was assessed in all randomized patients who received ≥1 dose of study drug. Treatment-emergent AEs (TEAE) included AEs from cycle 1, day 1 (C1D1) to within 42 days of last dose and all treatment-related AEs thereafter. All sinusoidal obstruction syndrome (SOS), formerly known as veno-occlusive disease (VOD), events within 2 years of randomization, regardless of causes, were included. MedDRA (v19.1) coding was applied. AEs were graded according to the NCI Common Terminology Criteria for Adverse Events Version 3.0.

On the basis of the results of these correlative analyses using central laboratory CD22 data, similar analyses were conducted to assess the potential relationship(s) between CD22 measured locally and selected efficacy (response and survival) and safety endpoints.

Determination of sample size has been described in detail previously (12). CR/CRi was summarized by percentage of leukemic blasts that were CD22 positive at baseline and baseline CD22 expression level (measured as MESF) as quartiles. MESF quartile 1 (Q1) had the lowest and Q4 had the highest CD22 expression levels. Differences between treatment groups were tested using χ² test or Fisher exact test (if any cell count was <5). The OS, PFS, and DoR of patients in the inotuzumab ozogamicin group were compared with corresponding measures from the SC group using the stratified log-rank test at a one-sided 0.0125 significance level. The HR and corresponding 97.5% two-sided CI were calculated using stratified Cox proportional hazard regression (same stratification factors as for randomization). The median OS and PFS were estimated using the Kaplan–Meier method with two-sided 95% CI, as described by Brookmeyer and Crowley (22). Similar analyses were performed using baseline local CD22 positivity as quartiles considering the broad distribution and range of CD22 expression assessed locally.

A categorical assessment of potential correlations between central laboratory CD22 positivity and KMT2A abnormalities with a focus on translocations/rearrangements was summarized using patient data.

Subgroup analyses were not adjusted for multiple testing.

Upon request, and subject to certain criteria, conditions, and exceptions (see https://www.pfizer.com/science/clinical-trials/trial-data-and-results for more information), Pfizer will provide access to individual deidentified participant data from Pfizer-sponsored global interventional clinical studies conducted for medicines, vaccines, and medical devices (1) for indications that have been approved in the United States and/or European Union or (2) in programs that have been terminated (i.e., development for all indications has been discontinued). Pfizer will also consider requests for the protocol, data dictionary, and statistical analysis plan. Data may be requested from Pfizer trials 24 months after study completion. The deidentified participant data will be made available to researchers whose proposals meet the research criteria and other conditions, and for which an exception does not apply, via a secure portal. To gain access, data requestors must enter into a data access agreement with Pfizer.

A total of 326 patients were randomized (inotuzumab ozogamicin, n = 164 and SC, n = 162) and were included in the ITT analysis population, and baseline characteristics were generally balanced between treatment groups (12). For the ITT population, the inotuzumab ozogamicin group, and the SC group, the median leukemic blast CD22 positivity at baseline based on central laboratory measures was 98% (range, 11%–100%), respectively. Most patients in both treatment groups had high (≥90%) leukemic blast CD22 positivity (inotuzumab ozogamicin, 65% and SC, 57%), with only a small fraction exhibiting leukemic blast CD22 positivity <70% [inotuzumab ozogamicin, 3% (5/164) and SC, 11% (18/162)]. Most patients [77.6% (52/67)] in MESF Q1 had CD22 positivity <90%, while the majority in the higher quartiles [Q2, Q3, Q4, and Q2–Q4 combined (Q2–Q4)] had CD22 positivity ≥90% (Table 1).

For the 164 patients in the inotuzumab ozogamicin group, 143 (87.2%) had no dose reduction and 91 (55.5%) had no dose delay. The mean actual overall dose was 4.32 mg/m² [median (range) 4.22 (0.78–9.59) mg/m²] and the mean actual dose intensity was 1.54 mg/m²/cycle [1.58 (0.77–2.06)]. Similar exposure to inotuzumab ozogamicin was found across CD22 MESF quartiles. In the overall study population, a small group of patients (21/164 patients) in the inotuzumab ozogamicin arm had dose reduction. The AEs leading to inotuzumab ozogamicin dose reduction were neutropenia (2/30, 6.7%), alanine aminotransferase increased (1/30, 3.3%), and platelet count decreased (1/30, 3.3%) in Q1; thrombocytopenia (1/38, 2.6%) in Q2; and aspartate aminotransferase increased (1/33, 3%) in Q4. The most common AEs leading to inotuzumab ozogamicin dose delays were hematologic (neutropenia, thrombocytopenia, and febrile neutropenia) for the overall study population. In Q1, the most common AEs leading to inotuzumab ozogamicin dose delays were thrombocytopenia (6/30, 20%) and neutropenia (6/30, 20%). In Q2, the most common AEs leading to inotuzumab ozogamicin dose delays were neutropenia (7/38, 18.4%), febrile neutropenia (4/38, 10.5%), and thrombocytopenia (3/38, 7.9%; Supplementary Table S2). No unexpected toxicity was found as reason for dose delays and dose reduction. Across CD22 positivity quartiles as assessed by local laboratories, the exposure to inotuzumab ozogamicin was similar and the proportions of patients with dose reduction or dose delay were comparable (Supplementary Table S1). Furthermore, analysis of population pharmacokinetics of inotuzumab ozogamicin showed that for patients with R/R B-cell ALL, CD22 expression (baseline percentage of leukemic blasts expressing CD22 in peripheral blood) was not retained as a covariate of inotuzumab ozogamicin distribution and elimination in the final model (23). Therefore, in this analysis, inotuzumab ozogamicin exposure was not assessed as a function of CD22 positivity as assessed by the central laboratory.

To explore changes in CD22 expression over time, the percentage of leukemic blasts that was CD22 positive in the bone marrow of patients (either in CR or not, but with samples subjected to MRD analysis) who remained MRD positive was assessed (Fig. 1A). A decrease in CD22 positivity from 98.4% at baseline to 45.2% at EOT was seen in the inotuzumab ozogamicin group; no consistent or substantial change in CD22 positivity over time was found in the SC group. Similar patterns were found in levels of CD22 expression as assessed by MESF (Fig. 1B).

The rate of CR/CRi (status determined by investigators) was significantly higher with inotuzumab ozogamicin versus SC in patients with both high and low leukemic blast CD22 positivity (Fig. 2A; Supplementary Table S3). In patients with higher (≥90%) CD22 positivity, the rate (95% CI) of CR/CRi was 78.5% (69.5%–85.9%) for inotuzumab ozogamicin versus 35.5% (25.8%–46.1%) for SC, and the rate difference was 43% (97.5% CI, 28.8%–57.3%; one-sided P < 0.0001). In patients with lower (<90%) CD22 positivity, CR/CRi rate of inotuzumab ozogamicin versus SC was 65.7% (47.8%–80.9%) versus 30.6% (16.3%–48.1%), and the rate difference was 35.2% (97.5% CI, 10.3%–60%; one-sided P = 0.0015). Specifically, within the inotuzumab ozogamicin group, the rate of CR (indicating a response with better quality of hematopoietic recovery than CRi) reported was 42.1% (45/107) and 20% (7/35) for patients with CD22 positivity ≥90% and <90%, respectively. The rates of CR/CRi were similar across the CD22 MESF (measured by flow cytometry at the central laboratory) quartiles and were consistent with the primary endpoint results seen in the study, as well as those when analyzed by CD22 positivity per central laboratory (Fig. 2B). However, CR was more common in higher than lower CD22 quartiles for patients treated with inotuzumab ozogamicin: for Q1 (the lowest quartile), rate of CR was 45.5% (10/22) of the responding patients; while for Q4 (the highest quartile), CR was reported for 81.8% (18/22) of the responding patients (Fig. 2B).

MRD negativity rates among responding patients were higher with inotuzumab ozogamicin versus SC (Fig. 2C; Supplementary Fig. S1). In patients with higher (≥90%) versus lower (<90%) CD22 positivity, the rate (95% CI) was 82.1% (72.3%–89.6%) with inotuzumab ozogamicin versus 30.3% (15.6%–48.7%) with SC (one-sided P < 0.0001) and 60.9% (38.5%–80.3%) with inotuzumab ozogamicin versus 45.5% (16.7%–76.6%) with SC (one-sided P = 0.1985), respectively. In the small subgroup of patients with baseline CD22 positivity <70% (11%–68%) who received inotuzumab ozogamicin (n = 5), three patients achieved CR/CRi, including two patients who achieved MRD-negative status (Fig. 2D). More patients achieved MRD-negative status with inotuzumab ozogamicin versus SC across all CD22 MESF quartiles, and the rate differences in the higher quartiles (Q2–Q4) were significant (Fig. 2E).

DoR was significantly longer with inotuzumab ozogamicin versus SC in patients with CD22 positivity ≥90% who achieved CR/CRi (HR, 0.424; 97.5% CI, 0.250–0.719; one-sided P < 0.0001; median 5.4 vs. 3.1 months; Fig. 3A; Supplementary Fig. S3A). No significant difference in DoR between treatment arms was observed in patients with CD22 positivity <90% who achieved CR/CRi (HR, 1.280; 97.5% CI, 0.502–3.262; one-sided P = 0.7236; median 3.9 vs. 5.8 months; Fig. 3A; Supplementary Fig. S3B). When analyzed by CD22 MESF quartiles, DoR showed no significant difference in Q1 between the treatment groups, a nonsignificant trend (HR, 0.678 and 0.424, respectively, with 95% CI including 1) was seen, suggesting longer DoR with inotuzumab ozogamicin versus SC in Q2 and Q4, and DoR was significantly longer in Q3 and in Q2 to Q4 with inotuzumab ozogamicin versus SC (Fig. 3A).

PFS was significantly longer with inotuzumab ozogamicin versus SC overall (Fig. 3B; Supplementary Fig. S2). In patients with CD22 positivity ≥90%, the HR was 0.378 (97.5% CI, 0.260–0.551). A similar, albeit nonsignificant trend was found in patients with CD22 positivity <90% (HR, 0.647; 97.5% CI, 0.363–1.152). Longer PFS for inotuzumab ozogamicin versus SC was seen across all four CD22 MESF quartiles, and the difference was significant, except for Q1 (Fig. 3B; Supplementary Fig. S2).

OS was significantly longer with inotuzumab ozogamicin versus SC in patients with CD22 positivity ≥90% [HR, 0.551; 97.5% CI, 0.384–0.790; P < 0.0001; median 8.4 (95% CI, 6.5–11.8) vs. 5.3 (95% CI, 4.5–7.4) months], while no significant difference in OS was observed in patients with CD22 positivity <90% [HR, 1.294; 97.5% CI, 0.720–2.326; P = 0.8388; median 5.7 (95% CI, 4.5–8.0) vs. 7.7 (95% CI, 2.3–14.2) months; Figs. 3C and 4A]. Differences in OS between patients with CD22 positivity ≥90% and <90% were analyzed using stepwise Cox regression modeling for each treatment arm. For the inotuzumab ozogamicin arm, the HR (95% CI) was 0.537 (0.355–0.813) and the two-sided P value was 0.0033; for the SC arm, the HR (95% CI) was 1.175 (0.766–1.803) and the two-sided P value was 0.4589. No significant difference was observed between the treatment groups in MESF Q1. Q2 showed a nonsignificant trend for longer OS with inotuzumab ozogamicin versus SC (HR, 0.701; 97.5% CI, 0.383–1.286), and OS was significantly longer in Q3 and Q4 with inotuzumab ozogamicin versus SC (Fig. 3C). For combined Q2 to Q4, OS, as for DoR and PFS, was significantly longer with inotuzumab ozogamicin versus SC; the respective HRs and medians were similar compared with patients who had CD22 positivity ≥90% (Fig. 3). The Kaplan–Meier curves of OS in Q1 were similar between treatment groups (Fig. 4B), but the difference between the treatment groups became more pronounced in the higher quartiles. Difference between the treatment groups was significant only in those with CD22 positivity ≥90%, but not in the <90% group (Fig. 4A). The Kaplan–Meier curves of PFS and DoR showed similar trends (Supplementary Figs. S2 and S3).

Efficacy outcomes were also analyzed by CD22 positivity quartiles per local laboratory assessments, Q1 having the lowest and Q4 having the highest CD22 positivity. In general, the results showed improvement in measures of efficacy for inotuzumab ozogamicin over SC that was comparable across all four CD22 positivity quartiles. CR/CRi rates showed no difference between quartiles in the inotuzumab ozogamicin arm (P = 0.5906) or SC arm (P = 0.2061), and were significantly higher with inotuzumab ozogamicin versus SC for all quartiles. Within quartiles, MRD negativity rates in responders were also higher with inotuzumab ozogamicin versus SC, with the differences within the lower quartiles (Q1–Q3) being significant (Supplementary Table S4; Supplementary Fig. S1B). For DoR, PFS, and OS, there was a suggestion of greater benefit for patients in higher CD22 positivity quartiles treated with inotuzumab ozogamicin (Fig. 4D; Supplementary Table S4, Supplementary Figs. S2 and S3).

A subgroup of patients who responded to treatment and subsequently relapsed (inotuzumab ozogamicin, 24 and SC, 14) and who did not respond to treatment (inotuzumab ozogamicin, 43 and SC, 112) were included in this analysis. At baseline, CD22 positivity was evaluable for 36 of the responders (inotuzumab ozogamicin, 22 and SC, 14) and 120 (inotuzumab ozogamicin, 35 and SC, 85) of the nonresponders. For the responders, the majority of patients in both treatment groups had high (≥90%) CD22 positivity at baseline [inotuzumab ozogamicin, 15/24 (62.5%) and SC, 13/14 (92.9%)]; very few patients had CD22 positivity <70% [inotuzumab ozogamicin, 2/24 (8.3%) and SC, 1/14 (7.1%); Supplementary Table S5]. Among patients who responded to inotuzumab ozogamicin treatment, at EOT/relapse, a decrease in CD22 positivity was apparent compared with baseline and the majority of patients (9/10) had CD22 positivity <90% (Supplementary Table S5; Supplementary Fig. S4). As expected, the majority of evaluable patients who received SC and achieved CR/CRi exhibited CD22 positivity that remained ≥90% at EOT/relapse (Supplementary Table S5; Supplementary Fig. S4). When CD22 expression was quantified as MESF, a similar trend was evident (Supplementary Table S5; Supplementary Fig. S4).

Approximately half of the nonresponders had high (≥90%) CD22 positivity at baseline [inotuzumab ozogamicin, 23/43 (53.5%) and SC, 60/112 (53.6%)], while a small fraction of patients had CD22 positivity <70% [inotuzumab ozogamicin, 2/43 (4.7%) and SC, 12/112 (10.7%)]. All evaluable patients had detectable CD22 positivity at EOT; however, the number of subjects was low. Among inotuzumab ozogamicin–treated patients who failed to respond to treatment, a decrease in CD22 expression (MESF) was apparent at EOT compared with baseline (Supplementary Table S5). The majority of evaluable subjects who received SC and failed to respond to treatment had similar CD22 expression at baseline and at EOT.

In addition, data from 24 patients who responded to inotuzumab ozogamicin and subsequently relapsed in a phase I/II study (NCT01363297; ref. 24) were analyzed. The findings were similar, except that four patients who had 85% to 98% CD22 expression at baseline exhibited CD22-negative leukemic blasts at EOT/relapse (Supplementary Table S5; Supplementary Fig. S5).

Not all patients had samples evaluable for KMT2A status by FISH. In ITT patients evaluable for KMT2A status, prevalence of KMT2A translocations/rearrangements, including t(4;11) translocation, was significantly higher in those whose leukemic blast CD22 positivity was <90% versus ≥90% (16.9% vs. 3.5%; P = 0.0005; Supplementary Table S6). Consistent with this, CD22 MESF was lower in patients with KMT2A translocations/rearrangements, including t(4;11) translocations, than in patients with normal cytogenetics (median MESF 1,289, 762, and 3,587, respectively; with P values of 0.003 and 0.009 for ANOVA F-test pairwise comparisons of these KMT2A subgroups vs. normal; Supplementary Table S7). In patients who had KMT2A rearrangements, OS did not differ statistically for inotuzumab ozogamicin versus SC independent of CD22 positivity (≥90% or <90%); unstratified HR (97.5% CI) was 1.852 (0.260–13.198; P = 0.7624) and 1.494 (0.304–7.341; P = 0.7151), respectively (Fig. 4C; Supplementary Table S8).

A total of 307 patients were included in the safety population. The most frequent grade ≥3 AEs were hematologic for both inotuzumab ozogamicin and SC groups, while hepatic AEs of any grade were more common in the inotuzumab ozogamicin group (50.6% vs. 36.4% in the SC group; ref. 13). Among patients in the inotuzumab ozogamicin group, no apparent difference was seen in the incidence of most grade ≥3 AEs by baseline CD22 positivity as assessed by central laboratory (Table 2). For patients with CD22 positivity ≥90% versus <90%, the rates of grade 3 to 4 neutropenia, thrombocytopenia, and febrile neutropenia were 45.8% versus 51.4%, 32.7% versus 62.9%, and 27.1% versus 31.4%, respectively. For patients defined by CD22 MESF quartiles, there was a trend showing numerically more cytopenias in the lower quartiles (Q1 and Q2) compared with the higher quartiles (Q3 and Q4), but no clear trend was seen across the quartiles (Table 2). For patients in Q1 versus Q4, the rates of grade 3 to 4 neutropenia, thrombocytopenia, and febrile neutropenia were 56.7% versus 39.4%, 53.3% versus 27.3%, and 36.7% versus 21.2%, respectively.

Hepatic AEs were summarized on the basis of baseline CD22 positivity and CD22 MESF quartile (Supplementary Fig. S6). The rates of grade ≥3 hepatic AEs and hyperbilirubinemia were 19.6% and 5.6% for patients with CD22 positivity ≥90%; for patients with CD22 positivity <90%, the rates were 8.6% and 2.9%, respectively. No apparent relationship was observed between hepatic AEs (including hyperbilirubinemia and SOS/VOD) and CD22 MESF quartiles.

Among patients with posttreatment HSCT, there was no apparent relationship between the post-HSCT incidence of SOS/VOD and baseline CD22 positivity or CD22 MESF quartile. For patients with CD22 positivity ≥90% and <90%, the rate of post-HSCT SOS/VOD was 26.7% (16/60) and 16.7% (2/12). For patients in Q1 to Q4, the rate was 26.7% (4/15), 22.2% (4/18), 36.4% (8/22), and 11.8% (2/17), respectively (Supplementary Fig. S6).

Safety outcomes were similar when analyzed by CD22 positivity quartile per local laboratory. No relationship was found between AEs and CD22 positivity assessed by local laboratory. In the inotuzumab ozogamicin arm, cytopenias were the most common grade ≥3 AEs with similar rates across the quartiles; rates of grade ≥3 SOS/VOD were 13.2%, 7.9%, 7.3%, and 17.1% in Q1 to Q4, respectively; three grade 5 SOS/VOD events occurred in Q1, two in Q4, and zero in Q2 and Q3 (Supplementary Table S9).

This post-hoc analysis using data from the INO-VATE study showed that at least 90% of blasts were CD22 positive in the majority (∼60%) of patients using central laboratory testing. While responses and MRD negativity were observed at all levels of CD22 expression, improved outcomes in selected endpoints (OS and DoR) were observed only in patients with CD22 positivity ≥90% or at higher CD22 MESF quartiles. A similar association was seen with response rates (CR/CRi) and OS based on CD22 positivity quartiles per local laboratory.

MRD is an important prognostic factor, as well as an indicator for treatment intensity and duration; it is also used for timing HSCT (9, 25). Various cell surface markers associated with B-cell ALL are commonly used in MRD assessments by flow cytometry. As targeted immunotherapies are becoming available, the expression of such markers that are also the antigens targeted (e.g., CD19 and CD22) may change for patients who receive these treatment options, and different flow cytometry strategies are being designed (26, 27). In our study, we found that among patients who remained MRD positive, there was a decrease over time in the percentage of leukemic blasts that were CD22 positive in the inotuzumab ozogamicin group, which was consistent with the mechanism of action of inotuzumab ozogamicin. Notably, there seemed to be an increase in MESF in MRD–positive patients for the SC group in cycles 1 and 2 with stabilization in subsequent cycles (Fig. 1B). Considering the very small number of evaluable SC group patients at these timepoints, this potential increase in MESF, particularly for the MESF peak of 6,048 at cycle 3 when there were only three individual values of 2,390, 6,048, and 10,091 (data not shown), likely reflects variability associated with low sample numbers. It is noteworthy that for these MRD–positive patients in the SC group, the median MESF at EOT (3,546.5, n = 22) was very close to the baseline median (2,963, n = 109), which is consistent with SC having a CD22-independent mechanism of action.

In the INO-VATE trial, significantly more patients in the inotuzumab ozogamicin group achieved CR/CRi compared with those who received standard therapy (12). This analysis further shows that the rate of CR/CRi was significantly higher with inotuzumab ozogamicin versus SC in patients with both higher and lower leukemic blast CD22 positivity. MRD status assessment showed that a higher proportion of patients within the inotuzumab ozogamicin group who had higher (≥90%) versus lower (<90%) CD22 positivity achieved MRD-negative status (82% vs. 61%), indicating a better quality of response to inotuzumab ozogamicin in patients with high CD22 positivity. However, patients in the lowest CD22 positivity category (i.e., CD22 positivity < 70%) still achieved responses, two patients achieved MRD negativity. Although the number of patients was low, these outcomes indicate the potential benefit of inotuzumab ozogamicin in this patient subpopulation with ALL with relatively low CD22 expression, with no CD22 expression threshold identified below which clinical benefit was not potentially achievable.

Inotuzumab ozogamicin–induced cell death is due to calicheamicin-induced apoptosis. Results of an in vitro study in primary B-cell precursor ALL showed that although CD22 expression is required for inotuzumab ozogamicin binding and internalization, only inotuzumab ozogamicin internalization and the cell's sensitivity to calicheamicin are strongly correlated with the efficacy of inotuzumab ozogamicin, while basal and renewed CD22 expression are not (5). Furthermore, low CD22 expression level is sufficient to achieve high enough intracellular calicheamicin levels to induce cell death (5). On the basis of these preclinical findings, it was speculated that the degree of CD22 positivity should not be a limitation for patients with B-cell ALL to receive inotuzumab ozogamicin. Data from this analysis have provided important information to support this premise. Overall, it appears that patients can respond to inotuzumab ozogamicin treatment and achieve MRD negativity independent of CD22 expression based on the percentage of CD22-positive blasts. However, these data indicate that, when inotuzumab ozogamicin is given as a single agent, the quality of the responses and outcomes may be improved in patients with higher CD22 expression, and also those without t(4;11) translocation benefit most. inotuzumab ozogamicin has been frequently used in combination with chemotherapy, such as mini-hyperfractionated cyclophosphamide, vincristine, and dexamethasone, to treat R/R ALL (28, 29) or newly diagnosed ALL (30). Further investigation is needed to understand whether CD22 expression is a predictive marker of response to these combination regimens.

To further explore the contribution of the degree of CD22 expression to inotuzumab ozogamicin efficacy, CD22 flow cytometry data were quantified at the central laboratory as MESF, which is a standard measuring unit used to quantify the fluorescence intensity of a cell population studied using flow cytometry (19). On the basis of the values of CD22 MESF units (i.e., a precise measurement of CD22 receptor density on leukemic blasts), patient samples were analyzed in quartiles (composed of equal or near-equal numbers of patients) for correlations with treatment response, PFS, OS, and DoR. These analyses revealed that there was a comparable improvement in response rates for inotuzumab ozogamicin over SC in all four quartiles. However, there was a suggestion of increased benefit of inotuzumab ozogamicin in the higher CD22 expression quartiles. For example, the Kaplan–Meier curve for OS in Q1 (the lowest quartile) appeared similar for both inotuzumab ozogamicin and SC groups, but the difference between treatment groups appeared more pronounced in the higher quartiles. These MESF data support the hypothesis that patients with higher CD22 expression may have better quality responses to inotuzumab ozogamicin and improved outcomes as measured by DoR, PFS, and OS.

Although MESF is not commonly measured as a routine procedure in clinical care, it may still have reference value for physicians trying to extrapolate the findings of INO-VATE to local CD22 test results. Cox proportional hazards regression analysis showed that the HR of OS for CD22 MESF Q2 to Q4 was almost the same as in CD22 positivity ≥90% (Fig. 3C). Further baseline characteristics analysis showed that patients with CD22 positivity <90% per central laboratory were largely confined to MESF Q1, while higher MESF quartiles (Q2–Q4) being largely/almost exclusively composed of patients with CD22 positivity ≥90% (Table 1). Given these apparent associations, it was not surprising to see similar trends in efficacy analysis. Considering that an “absolute” 90% or 70% CD22 positivity threshold might be measuring different CD22 receptor densities in different local laboratories, identifying a trend in outcomes across MESF quartiles may help physicians to extend the application of this trend to their local assessment.

In clinical practice, CD22 positivity is typically determined by flow cytometry in local laboratories, although, due to the potentially varying thresholds and antibody reagents used, the degree of harmonization is limited among different laboratories. In the INO-VATE study, the majority of patients in the ITT population [77.6% (253/326)] had samples for CD22 tested by both local and central laboratories; all were CD22 positive (i.e., >0%) per central laboratory and one of 253 (0.4%) patient had 0% CD22-positive leukemic blasts per local laboratory. Furthermore, the percentage of leukemic blasts that were CD22 positive was typically higher when measured by central versus local laboratory (data not shown), suggesting that the central laboratory test was more sensitive than the tests used by local laboratories. This being said, the overall trends seen when exploring potential relationships between CD22 expression and clinical outcome were similar for central and local laboratory CD22 results.

CD22 expression was explored at EOT/relapse in patients who previously responded to treatment. In the inotuzumab ozogamicin treatment group, a decrease in CD22 positivity and receptor density was evident from baseline to EOT/relapse, but emergent CD22 negativity was generally not evident. As anticipated, considering the SC mechanism of action is not CD22 directed, no such decreases in CD22 expression were evident in patients administered SC. Furthermore, a similar trend toward decreased CD22 expression at EOT was evident in patients who received inotuzumab ozogamicin but never responded. It is important to note that CD22 expression is being assessed at the level of the leukemic population, hence changes in CD22 expression likely reflect selection from among preexisting subpopulations of leukemic blasts, rather than decreased CD22 expression within individual clones. Given the general lack of emergence of CD22-negative clones, these patients would still be eligible for trials featuring CD22-directed CART constructs under development, including monospecific, or multispecific, with antigen(s) in addition to CD22. In addition, these results suggest that there is selective pressure against highly CD22-expressing blasts in both responding and nonresponding patients who receive inotuzumab ozogamicin. In addition, these assessments used the CD22 mouse anti-human mAb clone RFB-4 (PE-conjugate, Life Technologies) in a flow cytometric assay in a central laboratory (Navigate BioPharma). This antibody can also bind to CD22 in the presence of bound inotuzumab ozogamicin and was selected for posttreatment CD22 assessment to avoid the risk of interference by residual inotuzumab ozogamicin (data not shown). Therefore, the possibility of steric interference of inotuzumab ozogamicin and the diagnostic anti-CD22 antibody used in the assay could be ruled out.

Among participants of the INO-VATE study, various cytogenetic characteristics, such as Ph+ ALL and KMT2A abnormalities, were present (12). Although interpreting outcomes based on 11q23 abnormality status is challenging, a subgroup analysis (31) found that in patients with KMT2A translocations/rearrangements, response rates were similar for inotuzumab ozogamicin and SC, and no difference was seen in survival between treatment groups. Results of this subgroup analysis align with published data demonstrating KMT2A translocation t(4;11) is independently associated with poor survival in patients with ALL (3). Our analyses presented here showed that baseline CD22 positivity was not different in patients with Ph+ ALL, while differences in CD22 positivity were found between patients with and without KMT2A abnormalities (Supplementary Table S6). Results of outcome analysis support the significant association between low CD22 expression and KMT2A rearrangements as demonstrated by previous studies (7, 32). Other KMT2A abnormalities, such as KMT2A deletions (11q–) and copy-number gain (11q+/amplification), were detected using FISH. However, such abnormalities are poorly characterized at the molecular and functional level, complicating interpretation of these results.

While rates of hepatic AEs, such as hyperbilirubinemia and SOS/VOD, were higher in patients treated with inotuzumab ozogamicin compared with those treated with SC, the risk of SOS/VOD is low among patients receiving inotuzumab ozogamicin who do not undergo HSCT (12, 13, 31, 33–35). This post-hoc analysis showed no apparent relationship between the post-HSCT incidence of SOS/VOD and baseline CD22 level. For patients in the highest CD22 MESF quartile (Q4) and who received HSCT after inotuzumab ozogamicin treatment, the rate of SOS/VOD was potentially lower compared with those in the lower three quartiles; however, the patient numbers were low and no apparent trend was seen in the other quartiles. These patients also had the lowest rate of grade 3 to 4 thrombocytopenia, although reduced thrombocytopenia is also related to efficacy (36).

Following the recent development of novel therapies for B-cell ALL, new methods are being explored by the scientific community to find better disease indicators and prognostic factors. Findings from this study suggest that CD22 MESF quartiles analysis could represent a prognostic/predictive tool that might complement assessment of CD22 positivity for relapsed and/or refractory ALL. Nevertheless, the translatability of MESF to everyday clinical practice currently has limited immediate application. However, the additional analysis in our study of the relationship between patient outcomes and CD22 positivity quartiles per local laboratory showed similar trends in various efficacy endpoints and demonstrated the potential predictive utility of CD22 quartile analysis. Most importantly, these post-hoc analyses yielded similar overall results, indicating that response benefit to inotuzumab ozogamicin is evident for all quartiles. Furthermore, other parameters may also serve as predictive indicators for response to inotuzumab ozogamicin treatment. For instance, the overrepresentation of cytogenetic abnormalities (e.g., KMT2A rearrangements) in patients with <90% CD22 positivity could have clinical implications.

Overall, inotuzumab ozogamicin demonstrated a favorable benefit–risk profile for patients with relapsed or refractory B-cell precursor ALL independent of CD22 positivity. Nonetheless, there did appear to be improved outcomes with inotuzumab ozogamicin in those patients with leukemic blast CD22 positivity ≥90%. Similarly, patients with higher CD22 receptor density measured as MESF units had better quality responses, and the benefit of inotuzumab ozogamicin was more apparent in clinical outcomes.

H.M. Kantarjian reports grants and other from AbbVie, Amgen, Daiichi Sankyo, and Pfizer; grants from Ascentage, BMS, Immunogen, Jazz, and Sanofi; and other from Actinium, Adaptive Biotechnologies, Aptitude Health, Bio Ascend, Delta Fly, Jansen Global, Novartis, Biometical, and Takeda during the conduct of the study. W. Stock reports grants from Pfizer during the conduct of the study, other from Pfizer outside the submitted work, and speaking fees from Pfizer for educational conferences focused on care of ALL that have included discussion of inotuzumab ozogamicin. R.D. Cassaday reports grants and personal fees from Pfizer during the conduct of the study; grants and personal fees from Amgen and Kite/Gilead; grants outside the submitted work from Merck and Vanda Pharmaceuticals; and his spouse is employed by and owns stock in Seagen. D.J. DeAngelo reports other from AbbVie, GlycoMimetics, Novartis, and Blueprint Pharmaceuticals during the conduct of the study and personal fees from Amgen, Autolus, Agios, Blueprint, Forty-Seven, Incyte, Jazz, Kite, Novartis, Pfizer, Servier, and Takeda outside the submitted work. E. Jabbour reports grants and personal fees from Pfizer, Amgen, Takeda, AbbVie, and BMS and personal fees from Genentech outside the submitted work. S.M. O'Brien reports other from Celgene, GlaxoSmithKline, Janssen Oncology, Aptose Biosciences, Vaniam Group, AbbVie, Verastem, Vida Ventures, Autolus, Johnson & Johnson, Merck, Kite, Regeneron, Acerta, Gilead, Pharmacyclics, TG Therapeutics, Pfizer, and Sunesis outside the submitted work. M. Stelljes reports grants and personal fees from Pfizer and personal fees from Amgen during the conduct of the study. T. Wang, M.L. Paccagnella, B. Sleight, E. Vandendries, A. Neuhof, and A.D. Laird are employees of and have stock and/or other ownership interests in Pfizer. A.S. Advani reports grants and nonfinancial support from Pfizer during the conduct of the study; grants and other from Pfizer, Amgen, Seattle Genetics, and GlycoMimetics; other from Kite Pharmaceuticals; and grants from Immunogen, MacroGenics, AbbVie, and OBI Pharmaceuticals outside the submitted work. No disclosures were reported by the other authors.

H.M. Kantarjian: Conceptualization, resources, supervision, validation, investigation, writing–review and editing. W. Stock: Conceptualization, resources, supervision, validation, investigation, writing–review and editing. R.D. Cassaday: Conceptualization, resources, supervision, validation, investigation, writing–review and editing. D.J. DeAngelo: Conceptualization, resources, supervision, validation, investigation, writing–review and editing. E. Jabbour: Conceptualization, resources, supervision, validation, investigation, writing–review and editing. S.M. O'Brien: Conceptualization, resources, supervision, validation, investigation, writing–review and editing. M. Stelljes: Conceptualization, resources, supervision, validation, investigation, writing–review and editing. T. Wang: Data curation, formal analysis, validation, methodology, writing–review and editing. M.L. Paccagnella: Conceptualization, supervision, writing–review and editing. K. Nguyen: Data curation, investigation, methodology, writing–review and editing. B. Sleight: Conceptualization, supervision, writing–review and editing. E. Vandendries: Conceptualization, writing–review and editing. A. Neuhof: Conceptualization, supervision, writing–review and editing. A.D. Laird: Conceptualization, supervision, validation, methodology, writing–review and editing. A.S. Advani: Conceptualization, resources, supervision, validation, investigation, writing–review and editing.

This study was sponsored by Pfizer Inc. Medical writing support was provided by Shuang Li, PhD, of Engage Scientific Solutions and funded by Pfizer. Navigate BioPharma Services, Inc. were paid contractors to Pfizer Inc. for CD22 and MRD flow cytometry analyses and KMT2A assessment by FISH.

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.