Validation of International Working Group response criteria in higher‐risk myelodysplastic syndromes: A report on behalf of the MDS Clinical Research Consortium

Abstract The utility of the International Working Group (IWG) 2006 response criteria for myelodysplastic syndromes (MDS) as a surrogate endpoint for outcomes is unclear. We assessed the validity of the IWG 2006 response criteria in a large cohort of higher‐risk MDS patients (pts) treated at centers from the MDS Clinical Research Consortium. The best overall response rate (ORR) by IWG 2006 criteria to first‐line therapy among 597 evaluable pts was 38% and include complete response (CR) 16%, marrow CR (mCR) 2%, partial response (PR) 10%, hematological improvement (HI) 10%, stable disease (SD) 33%, and progressive disease (PD) 24%. CR was associated with a better overall survival (OS) compared to all other response groups (P < 0.001). Among 470 pts treated with hypomethylating agent (HMA) as first‐line therapy, the overall Response Rate, defined as HI or better was 39%. The median OS from time of best response was 21 mo, 8 mo, 14 mo, 12 mo, 13 mo, and 8 mo for CR, mCR, PR, HI, SD, and PD, respectively (P < 0.001). We validated those results in a separate cohort of 539 higher‐risk MDS pts treated at Moffitt Cancer Center who received first‐line HMA therapy, particularly addressing the value of mCR and mCR+HI. mCR alone without HI, SD, and PD outcomes were inferior to CR, PR, mCR+HI, and HI. In conclusion, CR by IWG 2006 response criteria can be used as a surrogate endpoint for OS in higher‐risk MDS pts. Any response associated with restoration of effective hematopoiesis is associated with better outcome.


| INTRODUCTION
The primary goal for treatment of higher-risk myelodysplastic syndromes (MDS) patients is to improve overall survival (OS) and delay evolution to acute myeloid leukemia (AML). 1 The criteria to assess response after MDS therapy were originally proposed by an International Working Group (IWG) of experts based on available data and consensus opinion in 2000 and were subsequently modified in 2006 (Table 1). 2,3 The IWG 2006 response criteria are widely used in clinical trials to evaluate the efficacy of MDS treatments in a defined, systematic way. However, the IWG 2006 response criteria have not been accepted by health regulatory agencies as surrogate endpoints that clearly translate to clinically meaningful benefits, such as improved OS and delayed AML transformation. Moreover, in clinical practice, the use of IWG 2006 response criteria to determine treatment efficacy and patient benefit is not widespread. In an ad hoc landmark analysis of the AZA-001 study in which higher-risk MDS patients treated with azacitidine were compared to those treated with conventional care regimens, patients who achieved 2006 IWG-defined responses of hematological improvement (HI), complete response (CR), or partial response (PR) demonstrated improved OS. 4,5 Increasingly, the response criterion marrow CR (mCR) has also been included in clinical trial measures of overall response rates. Whether this is associated with clinically meaningful endpoints has not been determined.
The utility of IWG 2006 response criteria outside of the context of clinical trials and their association with outcomes is unclear. Here, we assess the validity of the IWG 2006 response criteria, including mCR and SD, in a large cohort of higher-risk MDS patients treated with multiple drugs at centers from the MDS Clinical Research Consortium (MDSCRC) and validated in a large separate cohort at Moffitt Cancer Center (MCC).

| METHODS
Adult patients (>18 years) with higher-risk MDS (ie, Intermediate-2  or High Risk by International Prognostic Scoring System 6 [IPSS]) with diagnoses confirmed per World Health Organization criteria 7 and who had received treatment and for whom details of response and outcome were available were included from MDSCRC centers. Follow-up bone marrow assessments were obtained within 4-6 months of therapy initiation to assess response. Karyotyping was based on 20 metaphases with at least two cells expressing an abnormality required to define a clone. Patients were also reclassified per revised IPSS (IPSS-R). 8 The best response to treatment was categorized per the

Category
Response Criteria (must last at least 4 weeks) Complete Remission • Bone marrow: ≤5% myeloblasts with normal maturation of all cell lines • Persistent dysplasia will be noted • Hgb: ≥11 g/dL, platelets: ≥100 × 10 9 /L, neutrophils: ≥1.0 × 10 9 /L, blasts: 0% The primary endpoint was median OS and secondary endpoint was AML transformation, both measured from time start of therapy. For the MDSCRC cohort survival data were not censored for allogeneic transplant, the MCC patients selected cohort did not undergo transplant. The majority of front-line treatments were the hypomethylating agents azacitidine or decitabine, alone or in combination with other drugs. Descriptive statistics were used for baseline characteristics. The Kaplan-Meier method was used to estimate OS and a log-rank analysis was used to compare response categories. Cox regression analysis was used for multivariable analyses.. A two-sided alpha <.05 defined significance.

| RESULTS
We identified 646 treated IPSS higher-risk MDS patients from the MDSCRC. Table 2 summarizes baseline characteristics. The median age was 68 years and the majority were Caucasian. Refractory anemia with excess blasts II (RAEB-II) was the most common WHO subtype and onethird were classified as therapy-related MDS (t-MDS). By IPSS stratification, two-thirds of patients were intermediate-2 risk, whereas half of the patients were very high risk by IPSS-R. The first-line treatment included a hypomethylating agent in 470 pts (74%): either azacitidine or decitabine monotherapy, or one of these two agents in combination with another drug. One-third of the patients (29%) ultimately underwent allogeneic hematopoietic stem cell transplant.
The median duration of follow-up from diagnosis was 16.2 months (mo); 95% CI: (15.2, 17.7). The median survival  Table 3 and was significantly worse for higher-risk IPSS (P = 0.007) and IPSS-R (P < 0.001) subgroups. The best response rates by IWG 2006 criteria to first-line therapy were evaluable in 597 patients and included CR in 93 patients (16%), mCR in 10 (2%), PR in 57 (10%), HI in 60 (10%), SD in 239 (33%), and PD in 144 (24%). The median OS was 23.3 mo for those who achieved CR, 10.3 mo for mCR, 13 mo for PR, 12.5 mo for HI, 12.7 mo for SD, and 6.9 mo for PD. CR was associated with a better OS compared to all other response groups (P < 0.001) (Figure 1).
Among 580 patients evaluable for AML transformation, 310 (53%) transformed to AML. There was no difference in rate of AML transformation among response groups except in PD patients, who (as expected) transformed at a higher rate compared to others. The AML transformation rate was 52%, 46%, 43%, and 76% for CR, mCR/PR/HI, SD, and PD, respectively (P < 0.001). CR was associated with better leukemia-free survival (LFS) compared to the mCR/PR/HI/SD combined group and compared to PD. Patients with mCR/ PR/HI/SD had better LFS compared to those with PD. The LFS was 13.5, 7.8, 8.5, 10.3, 7.9, and 1.8 mo for CR, mCR, PR, HI, SD, and PD, respectively, P < 0.001.
Among 470 patients treated with HMA as first-line therapy, response was evaluable in 448 pts. The overall Response Rate (ORR), defined as HI or better (CR/PR/mCR/HI) was 39%, similar to the ORR reported in the AZA-001 and United States Intergroup (azacitidine monotherapy vs. azacitidine combined with lenalidomide or with vorinostat studies). 5,9 ( Table 4) The median OS from time of starting therapy was 21 mo for CR, 8 mo for mCR, 14 mo for PR, 12 mo for HI, 13 mo for SD, and 8 mo for PD (P < 0.001). A CR was associated with better outcome compared to all other response groups. Patients with PR, HI, and SD had better outcome compared to PD. The median LFS was 16. 3, 7.7, 8.6, 9.5, 7.9, and 2.3 mo for CR, mCR, PR, Hi, SD, and PD, respectively (P < 0.001).

| DISCUSSION
In this large retrospective study, we demonstrate that the best response by IWG 2006 criteria to first-line therapy in higher-risk MDS is associated with OS. This is particularly true for patients who achieved CR, which can be used as a surrogate for clinically meaningful outcomes. While patients who achieved SD or better response, as a combined group, had improved outcome compared to PD. Within the AZA-001 pivotal study, patients with stable disease had a survival advantage compared to those with progressive disease, but survival among those with stable disease did not differ between azacitidine or CCR. 4 We previously reported that among patients who had stable disease at 4-6 months after treatment initiation, 20% achieved a response later on, while those who achieved CR had superior OS compared to patients who remained with SD (28.1 vs. 14.4 months, respectively, P = 0.04). 10 Patients with a mCR response achieved little clinical benefit to their treatment, as OS was worse than those with SD. The validation cohort from MCC corroborated those findings in a larger number of patients with a greater representation of mCR to specifically address this response category. An OS benefit was restricted to those patients with mCR who achieved HI only. Responses associated with restoration of effective hematopoiesis (CR, PR, mCR+HI, and HI) were associated with better outcome while mCR, SD, and PD were not. While this study is limited by its retrospective nature, it seems unlikely that using mCR as a response criterion in prospective trials will lead to demonstrable meaningful benefits to patients. Myeloblast reduction without restoration of some degree of hematopoiesis has not been demonstrated to improve outcomes. The ONTIME randomized clinical trial compared rigosertib to best supportive care. There were no CR or PR in both groups, the mCR was 20% in the rigosertib arm and 14% in the best supportive care. There was no overall survival benefit observed with rigosertib. 11 The potential benefit of transient myeloblast reduction prior to allogeneic stem transplant could not be examined in this study.
There are challenges assessing mCR, HI, and SD outside of the context of clinical trials. For example, the ongoing assessment of transfusion reductions, and peripheral blood count improvement is limited by available data obtained during routine practice. The timing of bone marrow assessment after starting therapy is not standardized outside of trials. A landmark analysis at certain time points was not feasible in this retrospective study, as the clinicians and the database captured best response rather than responses at selected time points. Another limitation of our study is lack of central review or audit of response assessment. Still, the ORR reported herein aligned with two of the largest studies conducted in higher-risk MDS, as did outcomes, indicating the validity of the endpoints assessed.
In conclusion, CR by IWG 2006 response criteria can be used as a surrogate endpoint for OS in higher-risk MDS patients in randomized Phase II studies determining comparison arms of Phase III trials, and for regulatory purposes.