Introduction
Response assessments defined by the European Leukemia Network (ELN) were written in the context of intensive induction chemotherapy (IC) [1]. Extending these definitions to non-IC treatments may be problematic. After achieving morphologic blast clearance, the ELN defines remission subtypes based on the degree of peripheral blood count recovery [1]. By convention, bone marrow (BM) biopsies to assess remission occur upon hematologic recovery [2], typically 4–6weeks after IC, allowing a> 3-week break from myelosuppressive therapy prior to evaluation. Because that treatment-free interval can be sufficient to recover counts in the setting of a response, suboptimal count recovery is likely due to residual disease. Consistent with this, the degree to which counts recover after IC impacts overall survival (OS) [3]. A day-28 BM biopsy to assess initial response in trials utilizing venetoclax-based combinations has been adopted [4–6]. However, in this setting, a myelosuppressive therapy, venetoclax, is administered continuously for 28days, and taken by patients up to the day of the response assessment. Despite this, according to the ELN, the same count recovery expectations as patients who received IC are applied, and the same assumptions about the superiority of count recovery at this time point are made.
Currently, the recommendation for administration of venetoclax-based regimens is to perform an end-ofcycle 1 BM biopsy around day 28. When a morphologic remission is confirmed, the recommendation is to hold therapy until count recovery, sometimes with the assistance of granulocyte colony-stimulating factors (GCSF) [7,8]. However, a strict interpretation of ELN mandates a cycle 1 response assessment on the day of the BM biopsy; there is no allowance to ‘upgrade’ a response if counts improve after this procedure. An alternative approach to assessing responses to venetoclax-based regimens, involving waiting to record the response until a 14-day period after the BM biopsy has passed, and upgrading the response from either a morphologic leukemia free state (MLFS) on the date of the biopsy to a complete remission (CR)/CR with incomplete recovery of blood counts (CRi), or a CRi on the date of the biopsy to a CR, may more accurately inform expectations regarding OS.
Materials and methods
Eight hundred and thirty-five patients with a diagnosis of non-acute promyelocytic leukemia AML treated at the University of Colorado from January 2007 to July 2020 were screened; 435 were newly-diagnosed with a diagnostic test performed at our institution and follow-up through an initial post-treatment assessment or death. Response categories were assigned based on ELN [1]. Between group and within group comparisons were made by chi-square tests for categorical data and Wilcoxon Rank Sum for comparison of medians for continuous measures. OS analyses were conducted using SAS version 9.4. Median OS curves were created using Kaplan–Meier product-limit estimates with Hall–Wellner 95% confidence bands. Mantel–Cox logrank analysis was used to compare OS times with a significance threshold of 0.05. OS was defined as the time between diagnosis and death or last follow-up. Progression-free survival (PFS) was defined as the time between diagnosis and relapse or death. No patients were lost to follow-up.
Results
101 patients treated with venetoclaxþazacitidine (ven/aza) and 138 treated with IC had a response after cycle 1. Baseline characteristics of these patients are shown in Table 1. Between group comparisons suggest that the ven/aza patients were older with more adverse disease biological factors; within group comparisons revealed no major differences for ven/aza patients, while IC patients who achieved CR were less likely to have adverse disease features. The median follow up time for ven/aza patients was 748days (95% CI: 6,561,190) and 1400days (95% CI: 12,661,803) for IC patients. For ven/aza patients, as evaluated on the day of the BM biopsy, 14 achieved CR, 52 achieved Pediatric medical device CRi (CRi platelets [platelet recovery 三100x 109/L] ¼ 2, CRi neutrophils [absolute neutrophil count 三1.0x 109/ L] ¼ 50) and 35 achieved MLFS (Table 2). Following a 14-day period off therapy after the BM biopsy, responses were re-assessed using ELN criteria [1]; 50 responses were upgraded, 6 were downgraded, 34 had no change in response category, and 11 had no follow-up counts during the 14-day window (Table 2).
Patients without changes in response category after the 14-day hold included those who had a CR that was maintained (CRstable, N= 6), a CRi that was maintained (CRistable, N= 13), and a MLFS that was maintained (MLFSstable, N= 15) (Table 2). Previously we showed no difference in OS for ven/aza patients when best response of CR and CRi were compared [9]; here, we show no difference in OS for patients whose initial response was CRstable compared to CRistable (p= 0.7172); MLFSstable patients had an inferior OS compared to CRstable (p= 0.0362) and CRistable patients (p= 0.0122) (Figure 1(A)). Similar observations were made with respect to PFS; no differences were observed for CRstable vs CRistable patients (p= 0.7952), while MLFSstable patients had inferior PFS compared to CRstable (p= 0.0342) and CRistable (p= 0.0068) patients (Figure 1(B)).
When ven/aza patients with stable responses were compared with ven/aza patients who upgraded response categories, there were no significant differences in OS for the 32 who upgraded from CRi to CR (CRiupgrade) compared to CRstable patients (p= 0.4851) (Figure 1(C)). Similarly, no PFS differences were observed for these groups (p= 0.7818) (Figure 1(D)). A significant difference in OS and PFS was observed when comparing the 18 patients with an upgraded response from MLFS (MLFSupgrade) (Table 2) to CRstable and CRistable patients (p= 0.0064 and 0.0058 for OS and PFS, respectively) (Supplemental Figure 1A and 1B).
Similar patterns were not observed for IC patients. As evaluated on the day of the BM assessment, 99 IC patients achieved CR, 29 achieved CRi (CRi platelets=13, CRi neutrophils=16) and 10 achieved MLFS (Table 2). After a 14-day period off therapy after the biopsy, there were 27 upgraded responses (N= 23 CRiupgrade, N= 4 MLFSupgrade), 10 downgraded responses, 84 without change in response category (N= 79 CRstable, N= 4 CRistable, N= 1 MLFSstable) and 17 without follow-up blood counts during the 14-day window (Table 2). In contrast to ven/aza patients, and consistent with previous publications [3], there was an OS difference between CRstable and CRistable in IC patients (p= 0.0271, Figure 2(A)); there was no difference in PFS (p= 0.1146, Figure 2(B)), and there were not enough MLFSstable patients for comparison. While IC patients with a CRiupgrade (N= 23) had no outright significant difference in OS compared to CRstable patients (p= 0.0582) (Figure 2(C)), the difference, using a 0.10 threshold, was marginally significant. For comparison, the OS difference between ven/aza CRiupgrade and CRstable patients (p= 0.4851, Figure 1(C)) was not significant at either a 0.05 or the more lenient 0.10 threshold. MLFSupgrade patients (N= 4) had an inferior OS and PFS compared to CRstable and CRistable patients (p= 0.0005 and p= 0.0026, respectively) (Supplemental Figure 2A and 2B). PFS did not differ between IC patients with CRstable and CRiupgrade (p= 0.1897) (Figure 2(D)).
GCSF is frequently administered to both IC and ven/aza patients. Between treatment initiation and a 14-day period after the initial BM biopsy, 30/90 (29.7%) ven/aza patients were administered GCSF, compared to 19/138 (13.7%) of IC patients (Table 1). For ven/aza patients, we investigated whether the use of GCSF to upgrade a response impacted OS outcomes. For this analysis we first compared those with a CRiupgrade who received GCSF (N= 9) with CRstable patients; we found no difference in OS (p= 0.2011) or PFS (0.4227) (Figure 3(A,B)). We next compared this group of CRiupgrade patients who received GCSF with CRiupgrade patients who did not receive GCSF (N= 23); we again found no differences in OS (p= 0.1842) and PFS (p= 0.1958) (Figure 3(C,D), respectively). Similarly, no differences in OS or PFS were observed when MLFSupgrade patients who received GCSF (N= 13) were compared to MLFSupgrade patients who did not receive GCSF (N= 18) (p= 0.2815 and p= 0.9473, respectively) (Figure 3(E,F)).
Allowing 14days for count recovery to occur led to the unexpected observation that a small number of patients had decreased blood counts during this period. For ven/aza patients, 6 had downgraded responses between the BM and the end of the 14-day period, from a CR to a CRi (N= 3) (CRdowngrade), and from CRi to MLFS (N= 3) (CRidowngrade) (Table 2).
CRdowngrade patients had a marginal difference in OS and no difference in PFS compared to CRstable patients (p= 0.0895 and 0.2069, respectively, Supplemental Figure 3A and 3B). Interestingly, CRidowngrade patients had significantly reduced OS and PFS compared to CRistable patients (p= 0.0068 and 0.0080, respectively, Supplemental Figures 3C and 3D), but these analyses are limited by small numbers. For IC patients, 10 had a CRdowngrade (4 to CRi and 6 to MLFS) (Table 2). Comparing OS and PFS of IC CRdowngrade patients with IC CRstable patients, we found no significant differences in OS or PFS (p= 0.5697 and 0.0818, respectively, although the PFS value meets the criteria for marginal significance) (Supplemental Figure 3E and 3F).
Discussion
The ELN does not recommend a specific time at which response assessments are to occur; after IC, the National Comprehensive Cancer Network AML guidelines state that this should occur ‘upon hematologic recovery’ [2]. However, in the absence of hematologic recovery, it should be noted that a bone marrow biopsy is still performed, and by convention that typically happens 3-5weeks after the administration of IC. In the setting of venetoclax-based regimens, patients have a response assessment after cycle 1, or around day 28 Capmatinib cell line of therapy. Because venetoclax is a continuously-administered therapy with myelosuppressive properties, and due to recent observations that no differences are seen in OS or duration of response between those who achieve CR compared to CRi [4,9], we questioned whether the initial response assessments, as assessed by ELN and based on the degree of count recovery, would be reflective of outcomes for patients who receive venetoclax. While a 14-day period after the initial BM biopsy is an arbitrary period of time to observe a patient and allow for upgrades to occur, some defined time period is required, and 14-day breaks between ven/aza cycles have become routine in the setting of cytopenias [8].
In our analysis, the OS and PFS of ven/aza patients did not differ if a patient’s initial response was a CR or a CRi. Furthermore, when patients upgraded their responses from a CRi early response biomarkers to a CR, if warranted based on improved blood counts during a 14-day period after the initial BM biopsy, their OS and PFS were not statistically different compared to patients who experienced a full CR at the time of the BM biopsy. This was different than the experience of IC patients, who did have different OS outcomes between initial responses of CR and CRi, and for whom a CRiupgrade had inferior OS compared to those who experienced a full CR at the time of the BM biopsy. For ven/aza patients, the use of GCSF was not a major factor that could explain the lack of a difference in outcomes for patients who experienced a CRi compared to a CR.
A baseline comparison of ven/aza patients, including those with stable responses over the 14-day postBM biopsy period and those whose responses upgraded during this period, did not show significant differences that could explain the observations reported. Interestingly, IC patients who had a CRstable were less likely to have adverse baseline biological features (Table 1). This is consistent with prior observations that outcomes are comparatively worse for IC with adverse features [10], but this does not necessarily apply to the ven/aza regimen [4].
CRi appears to be a distinct entity in ven/aza patients compared to IC patients. We report here that while CRi is comprised of a relatively equal representation of CRi platelets and CRi neutrophils for IC patients, CRi neutrophils predominate as the CRi response in ven/aza patients. This fundamental difference may account for the difference in OS observed for CRi compared to CR in these two different treatment groups.
The focus of this paper is on the impact that upgraded responses have, or do not have, on outcomes like OS; however, in the course of the analysis we also noted the larger than previously appreciated number of patients in both groups whose responses were downgraded during the 14-day post BM biopsy period. While the numbers were too small to draw definitive conclusions, the relatively poor outcomes for ven/aza CRidowngrade patients was notable and should be further analyzed in larger datasets.
In general, this retrospective study was underpowered, with relatively small numbers of patients in subgroup comparisons. Insignificant p-values for underpowered comparisons should not exclude the possibility of clinically significant differences. As a result, conclusions presented should be considered to be exploratory in nature, and validating these findings in larger, and ideally, prospective, multi-institutional studies will be important to support these conclusions.
In conclusion, based on our analysis of patients treated with ven/aza, we suggest that ELN assessments are context-dependent and may not adequately reflect OS expectations in the non-IC setting. Modifications allowing for upgraded responses, or new response criteria, for ven/aza patients should be considered.