Original Article Blinatumomab Compared With Standard of Care for the Treatment of Adult Patients With Relapsed/Refractory Philadelphia Chromosome–Positive B-Precursor Acute Lymphoblastic Leukemia 1 2 3 Alessandro Rambaldi, MD ; Josep-Maria Ribera, MD, PhD ; Hagop M. Kantarjian, MD ; Hervé Dombret, MD4; Oliver G. Ottmann, MD ; Anthony S. Stein, MD ; Catherine A. Tuglus, PhD ; Xiaoyue Zhao, PhD7; 5 6 7 Christopher Kim, PhD, MPH7; and Giovanni Martinelli, MD8 BACKGROUND: A single-arm, phase 2 trial demonstrated the efficacy and safety of blinatumomab, a bispecific T-cell–engaging anti- body construct, in patients with relapsed/refractory (r/r) Philadelphia chromosome–positive (Ph+) acute lymphoblastic leukemia (ALL), a rare hematologic malignancy with limited treatment options. This study compared outcomes with blinatumomab with those of a historical control treated with the standard of care (SOC). METHODS: The blinatumomab trial enrolled adult patients with Ph+ ALL who were r/r to at least 1 second-generation tyrosine kinase inhibitor (n = 45). Propensity score analysis (PSA) was used to compare outcomes with blinatumomab with those of an external cohort of similar patients receiving SOC chemotherapy (n = 55). The PSA miti- gated confounding variables between studies by adjusting for imbalances in the age at diagnosis and start of treatment, sex, duration from diagnosis to most recent treatment, prior allogeneic hematopoietic stem cell transplantation, prior salvage therapy, and number of salvage therapies. Bayesian data augmentation was applied to improve power to 80% with data from a phase 3 blinatumomab study in r/r Philadelphia chromosome–negative ALL. RESULTS: In the PSA, the rate of complete remission or complete remission with partial hematologic recovery was 36% for blinatumomab and 25% for SOC, and this resulted in an odds ratio of 1.54 (95% confidence interval [CI], 0.61-3.89) or 1.70 (95% credible interval [CrI], 0.94-2.94) with Bayesian data augmentation. Overall survival favored blinatumomab over SOC, with a hazard ratio of 0.81 (95% CI, 0.57-1.14) or 0.77 (95% CrI, 0.61-0.96) with Bayesian data augmentation. CONCLUSIONS: These results further support blinatumomab as a treatment option for patients with r/r Ph+ ALL. Cancer 2020;126:304-310. © 2019 The Authors. Cancer published by Wiley Periodicals, Inc. on behalf of American Cancer Society. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. KEYWORDS: blinatumomab, Philadelphia chromosome–positive acute lymphoblastic leukemia, propensity score analysis, remission, standard of care, survival. INTRODUCTION The development of BCR-ABL1 protein–specific tyrosine kinase inhibitors (TKIs) has significantly improved outcomes in Philadelphia chromosome–positive (Ph+) acute lymphoblastic leukemia (ALL).1-3 The standard of care (SOC) for de novo Ph+ ALL is induction with conventional or attenuated chemotherapy in combination with a TKI.2,3 Most patients achieve complete remission (CR) and proceed to allogeneic hematopoietic stem cell transplantation (allo-HSCT).4 However, relapse can occur and is commonly associated with TKI-resistant mutations in the kinase domain of the BCR-ABL1 oncogene.5 There is no definitive evidence of a sustained response or long-term survival with TKIs after a relapse, with overall survival (OS) ranging from approximately 4 to 6 months.4,6,7 Compounding these challenges, Ph+ ALL is rare,8 and this limits most clinical trials evaluating new treatments to single-arm studies.2,3 Blinatumomab is a bispecific T-cell–engaging antibody construct that binds simultaneously to CD3-positive cytotoxic T cells and CD19-positive B cells and allows endogenous T cells to recognize and eliminate CD19-positive ALL blasts.9 Corresponding author: Alessandro Rambaldi, MD, Pope John XXIII Hospital, University of Studies of Milan, Piazza OMS 1, 24127 Bergamo, Italy; alessandro.rambaldi@ unimi.it 1 Department of Oncology and Hematology, University of Milan, Bergamo, Italy; 2 Clinical Hematology Service, Catalan Institute of Oncology, University Hospital Germans Trias i Pujol, Barcelona, Spain; 3 Department of Leukemia, University of Texas MD Anderson Cancer Center, Houston, Texas; 4 Department of Hematology, Hopital Saint-Louis, Paris Diderot University, Paris, France; 5 Division of Cancer and Genetics, School of Medicine, Cardiff University, Cardiff, United Kingdom; 6 Gehr Family Center for Leukemia Research, City of Hope, Duarte, California; 7 Amgen, Inc, Thousand Oaks, California; 8 Scientific Institute of Romagna for the Study and Treatment of Cancer, Meldola, Italy See editorial on pages 253-5, this issue. Medical writing and editorial assistance, which was supported by Amgen, Inc, was provided by Karen O’Leary, PhD, and Michael Raffin (Fishawack Communications, Inc). Additional supporting information may be found in the online version of this article. DOI: 10.1002/cncr.32558, Received: February 28, 2019; Revised: June 24, 2019; Accepted: July 11, 2019, Published online October 18, 2019 in Wiley Online Library ­(wileyonlinelibrary.com) 304 Cancer   January 15, 2020 Blinatumomab in Relapsed/Refractory Ph+ ALL/Rambaldi et al Prior studies have established the efficacy and safety of salvage therapy, and data were collected from diagnosis blinatumomab in relapsed/refractory (r/r) Philadelphia until the date of death or last follow-up. Investigators chromosome–negative (Ph–) ALL.10 Both Ph– and Ph+ received approval from an institutional review board or B-precursor leukemic cells express CD19; therefore, bli- ethics committee of participating centers. natumomab was assessed in a single-arm, phase 2 study of patients with r/r Ph+ ALL who had received a sec- Blinatumomab Ph+ ALL study ond-generation TKI.11 Of the 45 patients enrolled, 36% The blinatumomab study was an open-label, single- achieved CR or complete remission with partial hemato- arm, multicenter, phase 2 clinical trial of blinatumomab logic recovery (CRh). The median OS was 7.1 months. in adults with r/r Ph+ ALL (ClinicalTrials.gov iden- To assess the relevance of the blinatumomab study tifier NCT02000427). The study was conducted at results within the wider context of available treatment 19 centers in Europe and the United States. Details of this options, we compared the treatment outcomes with those study have been previously reported.11 Patients with Ph+ of an external control population. For rare diseases with- B-precursor ALL who were 18 years old or older were out a satisfactory SOC, regulatory agencies support the eligible for enrollment provided that they were r/r to at use of external controls as a method for demonstrating least 1 second-generation TKI, had >5% bone marrow the efficacy of new treatments.12 A problem with this blasts, and had an Eastern Cooperative Oncology Group approach is the substantial variability among patients performance status of 2 or lower. Exclusion criteria in- in the external control cohort. Propensity score analysis cluded allo-HSCT within the 12 weeks before the start of (PSA) provides a better balance between patients receiv- blinatumomab, active acute or chronic (grade 2-4) graft- ing the treatment of interest and the external control with versus-host disease, systemic treatment of graft-versus- respect to relevant baseline factors, and it enables a less host disease within 2 weeks of starting blinatumomab, a biased comparison of outcomes. history or presence of clinically relevant central nervous Here we report the results of a PSA comparing ef- system pathology (including central nervous system ALL), ficacy data from the phase 2 blinatumomab study and isolated extramedullary disease, and a history of malig- those of an external population: patients with r/r Ph+ nancy other than ALL within 5 years. Blinatumomab was B-precursor ALL who had received SOC after the failure administered as a continuous intravenous infusion at a of or resistance to treatment with second-generation TKIs. dose of 9 µg/d in week 1 of cycle 1 and at a dose of 28 µg/d thereafter. For each treatment cycle, blinatumomab was administered for 4 weeks, which was followed by 2 weeks MATERIALS AND METHODS off treatment. Patients who achieved CR/CRh could External SOC receive up to 3 additional cycles of treatment. The base- The external SOC cohort was identified and developed line period for patients began in January 2014, and the from existing clinical databases at centers in Italy (Pope study ended in May 2015. All patients provided informed John XXIII Hospital [Bergamo] and Sant’Orsola Policlinic consent, and the study was approved by the institutional [Bologna]) and Spain (Josep Carreras Research Institute, review boards of participating centers. Hospital Germans Trias i Pujol, Catalan Oncology Institute [Barcelona]). To align with the eligibility criteria of the phase Efficacy Endpoints 2 blinatumomab trial, patients with r/r Ph+ ALL included Efficacy endpoints for the PSA included OS and CR/CRh. in the external SOC cohort were 18 years old or older, were For time-to-event analyses, patients were followed from r/r to at least 1 second-generation TKI (dasatinib, nilotinib, the start date of blinatumomab or SOC therapy to the bosutinib, or ponatinib), and had >5% bone marrow blasts. event or were censored at the time they were lost to Patients were excluded if they had a history of malignancy follow-up or alive. CR was defined as ≤5% bone mar- other than ALL within 5 years of initiating salvage SOC, row blasts, with a platelet count >100,000/μL, an central nervous system or extramedullary disease, or prior absolute neutrophil count >1000/μL, and no evidence of therapy with blinatumomab. There were no restrictions on extramedullary disease. CRh was defined as ≤5% bone qualifying salvage therapy. marrow blasts, with a platelet count >50,000/μL and an Data collection began in August 2017 and ended in absolute neutrophil count >500/μL. The response was January 2018. Fifty-five patients met all eligibility criteria determined within the first 2 treatment cycles in the bli- and were included in the current analysis (see Supporting natumomab study (approximately 70 days) but varied for Fig. 1). The baseline period started from the initial di- the SOC cohort with the treatment (the median time to a agnosis of ALL and ended at the start of the qualifying response was 48 days). Cancer   January 15, 2020 305 Original Article Propensity Score Analysis ratio of CR/CRh from the phase 3 trial of blinatumomab The PSA was planned and prespecified before endpoint versus SOC in patients with r/r Ph– B-cell precursor analyses were conducted. The PSA created a balance ALL.10,20 For Bayesian models, point estimates and 95% between the blinatumomab and external SOC cohorts credible intervals (CrIs) were estimated with summary with respect to available baseline covariates that deter- statistics and the relative highest posterior density inter- mined both the propensity for a patient to be treated val of the posterior distributions for model parameters of (with blinatumomab) and a patient’s prognosis.13-15 The interest. Bayesian models used enough “borrowing” from baseline covariates included the age at diagnosis and the phase 3 trial to achieve a power of 80%. Potential bias treatment, sex, time from diagnosis to most recent treat- was assessed by the completion of sensitivity analyses with ment (months), prior allo-HSCT status (yes or no), prior prespecified lower levels of borrowing (ie, power levels of salvage therapy status (yes or no), and number of prior 70% and 75%). Statistical programming was conducted salvage therapies (0, 1, 2, 3, or ≥4). An estimated propen- in SAS 9.4 (SAS Institute, Cary, North Carolina). sity score (ie, the predicted probability of participating in the blinatumomab phase 2 trial) was assigned to each RESULTS patient on the basis of the selected covariates. The balance Patient characteristics are presented in Table 1. All but of covariates between patients in the blinatumomab trial 1 patient in each of the cohorts were enrolled on the basis and patients in the external cohort was determined by the of r/r Ph+ ALL to a second-generation TKI. One patient calculation of standardized differences in each covariate in the external SOC cohort was intolerant to a second- before and after propensity score adjustments and box generation TKI and had failed or was intolerant to plot overlap in propensity scores. imatinib, whereas 1 patient in the blinatumomab cohort In the estimation of treatment effects, propensity was resistant to imatinib but had not received a second- scores were used to adjust for differences between patients generation TKI (protocol deviation). in the blinatumomab and external SOC cohorts via in- The study populations were generally similar with verse probability of treatment weighting (IPTW) meth- respect to sex and age, but differences were noted for geo- ods.16 The objective was to estimate the average treatment graphic region and prior treatments. The proportion of effect (ATE) from moving the entire population from an patients with no prior salvage therapy was higher in the untreated status to a treated status.17 Sensitivity analyses blinatumomab cohort (13% vs 31%), as were the pro- explored the use of stabilized inverse probability of treat- portions with prior treatment with 3 or more TKIs (16% ment weighting (sIPTW), which accounts for potential vs 38%) and prior allo-HSCT (33% vs 44%). Dasatinib instability caused by very large weights,18 and the average was the most common prior TKI in both cohorts (89% vs treatment effect of treated weights (ATT).19 Covariates 87%). Prior treatment with imatinib was more common with a standardized difference >0.20 after IPTW adjust- in the external SOC cohort (87% vs 56%), whereas prior ment were added to statistical models as covariates. treatment with ponatinib was more common in the blina- CR/CRh rates were analyzed with a logistic regres- tumomab cohort (13% vs 51%). sion model with a single-treatment indicator covariate and Qualifying salvage therapies in the external SOC propensity score–based weights to adjust for differences cohort included chemotherapy (22%), chemotherapy between the blinatumomab and external SOC cohorts. plus a TKI (29%), and a TKI alone (31%; see Supporting The model’s coefficient for treatment effect was used to Table 1). Common chemotherapy agents included mer- obtain an odds ratio, and a robust variance estimation captopurine, vincristine, cytarabine, cyclophosphamide, (applied with a generalized estimating equation) was used and mitoxantrone. Generally, chemotherapy included to construct 95% confidence intervals (CIs) to evaluate combination regimens such as high-dose cytarabine and the probability of CR/CRh. Similarly, OS was analyzed via mitoxantrone (HAM), mitoxantrone, etoposide, and a Cox proportional hazards model with a single-treatment cytarabine (MEC), and cyclophosphamide, vincristine, indicator covariate and with propensity score–based doxorubicin, and dexamethasone (hyper-CVAD). Other IPTW or sIPTW weights to adjust for differences. treatments (18%) included an aurora kinase inhibitor, Because of the small sample sizes, the PSA had a sta- bortezomib, and donor leukocyte infusion, used alone tistical power of 65% to detect an assumed hazard ratio or as part of a combination chemotherapy regimen, of 0.75 favoring blinatumomab treatment. To increase and salvage allo-HSCT. Use of corticosteroids was com- power, Bayesian data augmentation was applied to end- mon. Fifteen patients (27%) in the external SOC cohort point analyses using distributions of OS and the odds achieved CR/CRh with their qualifying salvage therapy, 306 Cancer   January 15, 2020 Blinatumomab in Relapsed/Refractory Ph+ ALL/Rambaldi et al TABLE 1. Baseline Characteristics TABLE 2. Treatment Outcomes External Blinatumomab External Blinatumomab Characteristic SOC (n = 55) Study (n = 45) SOC (n = 55) Study (n = 45) Age, median (range), ya 53 (20-82) 55 (23-78) Response to treatment, No. (%)a Age category, No. (%)a Overall complete remission 15 (27) 16 (36) 18-34 y 9 (16) 5 (11) Complete remission 14 (25) 14 (31) 35-54 y 22 (40) 17 (38) Complete remission with 1 (2) 2 (4) ≥55 y 24 (44) 23 (51) partial hematologic recovery Sex, No. (%) Complete remission with i 1 (2) 2 (4) Male 28 (51) 24 (53) ncomplete hematologic recovery Female 27 (49) 21 (47) Blast-free hypoplastic or NA 3 (7) Geographic region/country, No. (%) aplastic bone marrow United States 0 11 (24) Partial remission 1 (2) 2 (4) European Union 55 (100)b 34 (76) No response NA 12 (27) Lines of prior salvage treatment, Refractory/progressive disease/ 28 (51) 4 (9) No. (%) early death 0 7 (13) 14 (31) Unknown/missing 10 (18) 6 (13) 1 31 (56) 12 (27) Proceeded to allo-HSCT, No. (%) 8 (15) 4 (9) ≥2 17 (31) 19 (42) Overall survival, median 6.0 (4.4-9.2)b 7.1 (5.6 to NE) No. of prior TKI treatments, No. (%) (95% CI), mo 1 6 (11) 7 (16) 2 41 (75) 21 (47) Abbreviations: allo-HSCT, allogeneic hematopoietic stem cell transplantation; ≥3 8 (15) 17 (38) CI, confidence interval; NA, not available; NE, not estimable; SOC, standard Prior TKIs, No. (%)a of care. a Imatinib 48 (87) 25 (56)c Response within the first 2 cycles of treatment for the blinatumomab study. b Dasatinib 49 (89) 39 (87) Overall survival data were available for 51 patients (4 were missing the treat- Ponatinib 7 (13) 23 (51) ment start or last follow-up date). Nilotinib 10 (18) 16 (36) Multiple TKIs 49 (89) 38 (84) Prior allo-HSCT, No. (%) TABLE 3. Summary of CR/CRh Analysis With and Yes 18 (33) 20 (44) Without Bayesian Data Augmentation and With No 37 (67) 25 (56) IPTW-ATE Adjustments Abbreviations: allo-HSCT, allogeneic hematopoietic stem cell transplantation; External SOC Blinatumomab SOC, standard of care; TKI, tyrosine kinase inhibitor. Endpoint (n = 55) Study (n = 45) a Before the start of qualifying salvage therapy for the external SOC cohort. b Spain (n = 14) and Italy (n = 41). Non-Bayesian data OR, 1.54 (95% CI, 0.61-3.89); P = .26 c One patient had acute lymphoblastic leukemia resistant to imatinib and was augmentation (65% power) never exposed to a second-generation or later TKI. CR/CRh, % (95% CI) 26 (16-40) 36 (22-52) Bayesian data augmentation OR, 1.70 (95% CrI, 0.94-2.94); P = .076 (80% power) with 14 (25%) achieving CR and 1 (2%) achieving CRh CR/CRh, % (95% CrI) 25 (17-34) 36 (28-46) (Table 2). For the 51 patients for whom OS data were Abbreviations: ATE, average treatment effect; CI, confidence interval; CR, available, the median OS was 6.0 months (95% CI, complete remission; CRh, complete remission with partial hematologic re- 4.4-9.2 months; Supporting Fig. 2). covery; CrI, credible interval; IPTW, inverse probability of treatment weighting; OR, odds ratio; SOC, standard of care. The primary analysis of the blinatumomab study has been previously reported.11 The CR/CRh rate was 36% after 2 cycles, with 14 patients (31%) achieving CR allo-HSCT. Because the difference remained >0.20, the and 2 patients (4%) achieving CRh. The median OS was propensity score models incorporated an IPTW-ATE 7.1 months (95% CI, 5.6 to not estimable). adjustment with prior allo-HSCT as a covariate. The Bayesian-augmented (80% power) odds ratio Propensity Score Analysis estimate for CR/CRh was 1.70 (95% CrI, 0.94-2.94) and All propensity scores for the external SOC control were favored blinatumomab over the external SOC (Table 3). contained within the 95% range of the propensity scores Corresponding CR/CRh rate estimates for the blinatu- for blinatumomab, and this indicated that most patients momab and external SOC cohorts were 36% (95% CrI, in the external SOC would have been eligible to re- 28%-46%) and 25% (95% CrI, 17%-34%), respectively. ceive blinatumomab treatment (Supporting Fig. 3). Two The non-Bayesian (65% power) odds ratio was 1.54 covariates had a >0.20 standardized difference between (95% CI, 0.61-3.89). the cohorts: prior allo-HSCT and no prior salvage ther- The Bayesian-augmented (80% power) hazard ratio apy (Supporting Table 2). After adjustments with IPTW, comparing the OS of blinatumomab with the OS of the the standardized difference became 0 for no prior salvage external SOC was 0.77 (95% CrI, 0.61-0.96), and this therapy and was reduced from –0.33 to –0.23 for prior suggested a statistically significant 23% reduction in the Cancer   January 15, 2020 307 Original Article TABLE 4. Summary of OS Analysis With and blinatumomab in comparison with the external SOC, a Without Bayesian Data Augmentation and With numerical benefit that did not reach statistical significance. IPTW-ATE Adjustments The Bayesian-augmented analysis of OS showed a statisti- Blinatumomab cally significant 23% decrease in the hazard of death with Endpoint External SOC Study blinatumomab treatment in comparison with SOC, and Non-Bayesian data augmen- HR, 0.81 (95% CI, 0.57-1.14); P = .20 sensitivity analyses were consistent with these findings. For tation (65% power) OS probability, % (95% CI) future salvage treatment strategies, these observations will 3 mo 79 (70-89) 83 (74-93) be of great importance, particularly among patients for 6 mo 9 mo 52 (40-68) 39 (27-57) 59 (47-74) 47 (35-64) whom allo-HSCT is planned only in second remission.21,22 12 mo 32 (20-50) 40 (28-57) Although safety data for the external SOC were not Bayesian data augmentation HR, 0.77 (95% CrI, 0.61-0.96); P = .031 available for comparison, treatment toxicity is a relevant (80% power) OS probability, % (95% CrI) concern. During the phase 2 blinatumomab study, all 3 mo 79 (77-81) 83 (82-85) patients experienced at least 1 treatment-emergent adverse 6 mo 51 (47-55) 60 (57-63) 9 mo 39 (34-43) 48 (44-52) event (AE), and 82% experienced a grade 3 or higher AE, 12 mo 31 (26-35) 41 (37-44) but these were generally manageable because only 7% Abbreviations: ATE, average treatment effect; CI, confidence interval; CrI, of the patients discontinued treatment on account of an credible interval; HR, hazard ratio; IPTW, inverse probability of treatment AE.11 The most common grade 3 or higher AEs included weighting; OS, overall survival; SOC, standard of care. febrile neutropenia (27%), thrombocytopenia (22%), and anemia (16%). One fatal AE (septic shock) was con- risk of death associated with blinatumomab in compar- sidered treatment-related by the investigator. Overall, bli- ison with the external SOC. The non-Bayesian (65% natumomab was tolerable with manageable AEs. power) hazard ratio was 0.81 (95% CI, 0.57-1.14; Table 4 Given the benefit-to-risk profile of blinatumomab and Fig. 1). in the phase 2 trial and across clinical trials in ALL,10,11 Sensitivity analyses of less borrowing for Bayesian future studies are looking to pair blinatumomab with TKIs data augmentation were consistent with these analyses because the combination may provide additional bene­fit (see Supporting Table 3 [CR/CRh] and Supporting Fig. 4 to patients with Ph+ ALL.23 There is also evidence in [OS]), as were ATT sensitivity analyses (Supporting Ph– ALL to support the use of blinatumomab in patients Tables 4 [CR/CRh] and 5 and Supporting Figs. 4 and 5 at earlier stages of treatment, including patients who have [OS]) and sIPTW analyses (Supporting Fig. 6 [OS]). achieved CR/CRh with induction therapy but still have minimal residual disease (MRD).10,24 In the phase 2 trial, DISCUSSION 18 of 45 patients with r/r Ph+ ALL who received blina- In the single-arm, phase 2 blinatumomab trial, adult tumomab achieved an MRD response, with the median patients with r/r Ph+ ALL receiving blinatumomab OS not reached for MRD responders versus 3.9 months achieved a CR/CRh rate of 36% with a median OS of for MRD nonresponders.11 MRD response data were not 7.1 months.11 These results suggested an improvement in available for all patients in the external SOC cohort, so a treatment outcomes with blinatumomab in comparison comparison was not possible. with historical studies but were limited by the single-arm PSA has become an established method to support trial design.4,6,7 In the current analysis, PSA places these the development of novel treatments for rare malignan- efficacy results into the context of available treatment cies.12,25 However, there are limitations. Although PSA options. By aligning the eligibility criteria of the external mitigates the impact of known confounders and bias, it SOC with those of the blinatumomab study, we selected is not a replacement for randomization. In the propen- a similar patient population for comparison. Both patient sity score model, one can consider only known covariates populations were heavily pretreated and balanced for that are measured in both studies. The use of PSA cannot most baseline covariates. In the external SOC group, the address imbalances in unmeasured/unknown covariates CR/CRh rate was 26%, and the median OS was 6.0 months; or postbaseline variables. For the external SOC cohort, we this was consistent with historical Ph+ ALL studies.4,6,7 did not have data for some of the eligibility criteria that PSA, adjusted for imbalances in prognostic covariates defined the blinatumomab study population (eg, prior and Bayesian data augmentation, was applied to improve graft-versus-host disease, duration of remission with prior statistical power. Bayesian-augmented PSA demonstrated allo-HSCT, and Eastern Cooperative Oncology Group a 70% increase in the odds of achieving remission with performance status) or data for other important prognostic 308 Cancer   January 15, 2020 Blinatumomab in Relapsed/Refractory Ph+ ALL/Rambaldi et al A 1RQ%D\HVLDQGDWDDXJPHQWDWLRQ SRZHU   2YHUDOOVXUYLYDO    +5  &,± 3                0RQWKV B %D\HVLDQGDWDDXJPHQWDWLRQ SRZHU   2YHUDOOVXUYLYDO    +5  &UO± 3                0RQWKV %OLQDWXPRPDE ([WHUQDO62& Figure 1. Cox proportional hazards model estimates of survival by treatment (A) with and (B) without Bayesian data augmentation (80% power). IPTW-ATE adjustments were made. Survival estimates were calculated with the proportion of prior hematopoietic stem cell transplantation: 0.327 for the control group and 0.4 for the blinatumomab group. ATE indicates average treatment effect; CI, confidence interval; CrI, credible interval; HR, hazard ratio; IPTW, inverse probability of treatment weighting; SOC, standard of care. factors (eg, kinase domain mutations). Furthermore, post- and chemotherapies were standard treatment options. treatment allo-HSCT was more frequent with blinatum- Selecting for specific qualifying salvage therapies may omab than external SOC (15% vs 9%). Other limitations have introduced additional bias. include the small sample sizes of the cohorts, response In conclusion, the results from the PSA reported assessment by centralized review (blinatumomab study) here suggest that blinatumomab improves treatment out- versus investigator review (external SOC), and geographic comes in patients with r/r Ph+ ALL in comparison with and chronologic differences between the study cohorts. external SOC. These data further support blinatumomab The blinatumomab study was conducted in the United as a treatment option for patients with r/r Ph+ ALL. States and Europe, whereas the external SOC included patients enrolled at centers in Italy and Spain, with some FUNDING SUPPORT patients treated 9 years before the initiation of the bli- This study was funded by Amgen, Inc. The University of Texas MD Anderson Cancer Center is supported by the National Institutes of Health natumomab study. Although differences in clinical prac- (grant P30 CA016672). tice could be present between these cohorts (eg, the use of newer TKIs such as ponatinib and nilotinib), general practice patterns for ALL over time and between regions CONFLICT OF INTEREST DISCLOSURES Alessandro Rambaldi reports consultancy for and travel support from were not dramatically different. Treatment with TKIs Amgen, Pfizer, Roche, Celgene, Novartis, Italfarmaco, Gilead, and Omeros. Cancer   January 15, 2020 309 Original Article Josep-Maria Ribera reports consultancy for, travel support from, and grants four consecutive risk-adapted trials by the PETHEMA study group. from Amgen, Pfizer, Celgene, Novartis, Jazz Pharmaceuticals, Shire, and Haematologica. 2010;95:589-596. Takeda. Hagop M. Kantarjian reports honoraria and/or research fund- 7. Spyridonidis A, Labopin M, Schmid C, et al. Outcomes and prog- ing from Agios, Daiichi-Sankyo, Cyclacel, Jazz Pharma, Novartis, Astex, nostic factors of adults with acute lymphoblastic leukemia who AbbVie, Orsenix, Immunogen, Actinium, Delta-Fly Pharma, ARIAD relapse after allogeneic hematopoietic cell transplantation. An analysis Pharmaceuticals, Bristol-Meyers Squibb, Pfizer, Takeda, and Amgen. Hervé on behalf of the Acute Leukemia Working Party of EBMT. Leukemia. Dombret reports honoraria and/or research funding from Amgen, Agios, 2012;26:1211-1217. Ceylad, Seattle Genetics, Celgene, Sunesis, Roche, Pfizer, Ambit-Daiichi 8. Faderl S, O’Brien S, Pui CH, et al. Adult acute lymphoblastic leukemia: Sankyo, Shire-Baxalta, Ariad-Incyte, Karyopharm, AbbVie, Novartis, Kite, concepts and strategies. Cancer. 2010;116:1165-1176. Otsuka, Celator-Jazz, Astellas, Menarini, Cellectis, Janssen, ImmunoGen, 9. Wolach O, Stone RM. Blinatumomab for the treatment of Philadelphia and Servier. Oliver G. Ottmann reports consultancy for Amgen, Novartis, chromosome–negative, precursor B-cell acute lymphoblastic leukemia. Incyte, Takeda, Celgene, Fusion Pharma, Pfizer, and Roche; research sup- Clin Cancer Res. 2015;21:4262-4269. port from Incyte and Celgene; and travel support from Amgen, Novartis, 10. Kantarjian H, Stein A, Gokbuget N, et al. Blinatumomab versus che- and Incyte. Anthony S. Stein reports work on speakers bureaus for Amgen, motherapy for advanced acute lymphoblastic leukemia. N Engl J Med. Stemline, and Celgene. Catherine A. Tuglus reports employment by and 2017;376:836-847. stock holdings in Amgen. Xiaoyue Zhao reports employment by Amgen. 11. Martinelli G, Boissel N, Chevallier P, et al. Complete hematologic Christopher Kim reports employment by and stock holdings in Amgen. and molecular response in adult patients with relapsed/refractory Giovanni Martinelli reports work as an advisor to Amgen, Ariad/Incyte, Philadelphia chromosome–positive B-precursor acute lymphoblas- Pfizer, Roche, Celgene, Janssen, and Jazz Pharmaceuticals; work on speakers tic leukemia following treatment with blinatumomab: results from bureaus for Novartis, Pfizer, and Celgene; personal fees from AbbVie; and a phase II, single-arm, multicenter study. J Clin Oncol. 2017;35: travel compensation from Daiichi Sankyo, Roche, and Shire. 1795-1802. 12. Simon R, Blumenthal GM, Rothenberg ML, et al. The role of non- randomized trials in the evaluation of oncology drugs. Clin Pharmacol AUTHOR CONTRIBUTIONS Ther. 2015;97:502-507. Alessandro Rambaldi: Data acquisition, critical revision of the manuscript, 13. Levenson MS, Yue LQ. Regulatory issues of propensity score method- and approval of the final version. Josep-Maria Ribera: Data acquisition, ology application to drug and device safety studies. J Biopharm Stat. critical revision of the manuscript, and approval of the final version. Hagop 2013;23:110-121. M. Kantarjian: Data acquisition, critical revision of the manuscript, and 14. Yue LQ. Statistical and regulatory issues with the application of pro- approval of the final version. Hervé Dombret: Data acquisition, criti- pensity score analysis to nonrandomized medical device clinical studies. cal revision of the manuscript, and approval of the final version. Oliver J Biopharm Stat. 2007;17:1-13. G. Ottmann: Data acquisition, critical revision of the manuscript, and 15. D’Agostino RB Jr. Propensity score methods for bias reduction in the approval of the final version. Anthony S. Stein: Data acquisition, critical comparison of a treatment to a non-randomized control group. Stat revision of the manuscript, and approval of the final version. Catherine A. Med. 1998;17:2265-2281. Tuglus: Research design, analysis, and drafting of the manuscript. Xiaoyue 16. D’Agostino RB Jr, D’Agostino RB Sr. Estimating treatment effects Zhao: Analysis, critical revision of the manuscript, and approval of the final using observational data. JAMA. 2007;297:314-316. version. Christopher Kim: Research design, analysis, and drafting of the 17. Imbens GW. Nonparametric estimation of average treatment effects manuscript. Giovanni Martinelli: Data acquisition, critical revision of the under exogeneity: a review. Rev Econ Stat. 2004;86:4-29. manuscript, and approval of the final version. 18. Cole SR, Hernan MA. Adjusted survival curves with inverse probability weights. Comput Methods Programs Biomed. 2004;75:45-49. 19. Hirano K, Imbens G, Ridder G. Efficient estimation of average treatment effects using the estimated propensity score. Econometrica. REFERENCES 2003;71:1161-1189. 1. Brissot E, Labopin M, Beckers MM, et al. Tyrosine kinase inhibitors 20. Ibrahim J, Chen M. Power prior distributions for regression models. improve long-term outcome of allogeneic hematopoietic stem cell Stat Sci. 2000;15:46-60. transplantation for adult patients with Philadelphia chromosome posi- 21. Jabbour E, Kantarjian H, Ravandi F, et al. Combination of hyper- tive acute lymphoblastic leukemia. Haematologica. 2015;100:392-399. CVAD with ponatinib as first-line therapy for patients with Philadelphia 2. Hoelzer D, Bassan R, Dombret H, Fielding A, Ribera JM, Buske chromosome–positive acute lymphoblastic leukaemia: a single-centre, C. Acute lymphoblastic leukaemia in adult patients: ESMO clinical phase 2 study. Lancet Oncol. 2015;16:1547-1555. practice guidelines for diagnosis, treatment and follow-up. Ann Oncol. 22. Ravandi F, Jorgensen JL, Thomas DA, et al. Detection of MRD may 2016;27(suppl 5):v69-v82. predict the outcome of patients with Philadelphia chromosome– 3. National Comprehensive Cancer Network. NCCN Clinical Practice positive ALL treated with tyrosine kinase inhibitors plus chemotherapy. Guidelines in Oncology: Acute Lymphoblastic Leukemia Version Blood. 2013;122:1214-1221. I.2018. Accessed September 20, 2018. https​://www.nccn.org/ 23. Assi R, Kantarjian H, Short NJ, et al. Safety and efficacy of blinatu- 4. Fielding AK, Richards SM, Chopra R, et al. Outcome of 609 adults momab in combination with a tyrosine kinase inhibitor for the treat- after relapse of acute lymphoblastic leukemia (ALL); an MRC ment of relapsed Philadelphia chromosome–positive leukemia. Clin UKALL12/ECOG 2993 study. Blood. 2007;109:944-950. Lymphoma Myeloma Leuk. 2017;17:897-901. 5. Zabriskie MS, Eide CA, Tantravahi SK, et al. BCR-ABL1 compound 24. Gokbuget N, Dombret H, Bonifacio M, et al. Blinatumomab for mini- mutations combining key kinase domain positions confer clinical resis- mal residual disease in adults with B-cell precursor acute lymphoblastic tance to ponatinib in Ph chromosome–positive leukemia. Cancer Cell. leukemia. Blood. 2018;131:1522-1531. 2014;26:428-442. 25. Gokbuget N, Kelsh M, Chia V, et al. Blinatumomab vs historical stan- 6. Oriol A, Vives S, Hernandez-Rivas JM, et al. Outcome after dard therapy of adult relapsed/refractory acute lymphoblastic leukemia. relapse of acute lymphoblastic leukemia in adult patients included in Blood Cancer J. 2016;6:e473. 310 Cancer   January 15, 2020