Clinical Outcomes of Acute Myeloid Leukemia Patients Harboring the RUNX1 Mutation: Is It Still an Unfavorable Prognosis? A Cohort Study and Meta-Analysis

Simple Summary Acute myeloid leukemia (AML) with mutated RUNX1 (RUNX1mut) has an adverse prognosis based on the 2022 European LeukemiaNet risk stratification. However, the WHO classifications of 2022 removed RUNX1 mutations from the unique entity because of various prognoses and treatment outcomes. Intriguingly, the overall survival (OS) and relapse-free survival (RFS) outcomes were similar in patients who had de novo AML with intermediate-risk cytogenetics with and without RUNX1mut. Our study endorsed an unfavorable prognosis of this entity. Abstract Acute myeloid leukemia (AML) with mutated RUNX1 (RUNX1mut) is considered to have an unfavorable prognosis. However, recent studies have reported comparable survival outcomes with wild-type RUNX1 (RUNX1wt). To assess the clinical outcomes of AML with and without RUNX1mut, we performed a prospective cohort study and systematic review and meta-analysis. The study enrolled 135 patients (27 with RUNX1mut; 108 with RUNX1wt). There were no significant differences in the median OS and RFS of the RUNX1mut and RUNX1wt groups (9.1 vs. 12.2 months; p = 0.268 and 7.8 vs. 14.6 months; p = 0.481, respectively). A subgroup analysis of de novo AML patients with intermediate-risk cytogenetics showed similar outcomes. Our meta-analysis pooled data from 23 studies and our study. The complete remission rate was significantly lower in the RUNX1mut group (pooled odds ratio: 0.42). The OS, RFS, and event-free survival rates also favored the RUNX1wt group (pooled risk ratios: 1.36, 1.37, and 1.37, respectively). A subgroup analysis of de novo AML patients with intermediate-risk cytogenetics demonstrated nearly identical OS and RFS outcomes. This study confirms that patients with AML and RUNX1mut had poor prognoses. Nonetheless, in de novo AML with intermediate-risk cytogenetics, the survival outcomes of both groups were comparable.


Introduction
Acute myeloid leukemia (AML) is a hematologic malignancy that results from impaired proliferation and differentiation of hematopoietic stem cells, leading to an accumulation of abnormal blast cells in bone marrow [1]. AML's high heterogeneity is reflected in different disease prognoses, which are influenced by host-and disease-related factors [1][2][3]. For instance, old age, multiple comorbidities, and secondary AML are unfavorable predictive factors and the main obstacles for candidates of intensive therapy [1].

Prospective Cohort Study
The study was conducted on newly diagnosed AML patients with and without RUNX1 mut between January 2019 and April 2022. The patients attended Siriraj Hospital, Mahidol University, Thailand, an academic university hospital and the country's largest acute leukemia referral center. Patients were enrolled if they were over 18 years old and had a diagnosis of AML requiring treatment and follow-up. We excluded patients diagnosed with acute promyelocytic leukemia or those who had not undergone molecular testing before treatment.
The Siriraj Institutional Review Board approved this research, which followed the Declaration of Helsinki guidelines and all subsequent amendments. All patients gave informed consent for biobanking and the analyses. The study was registered at the Thai Clinical Trial Registry (TCTR20220921005).
The primary outcome of this study was the OS rate. The secondary outcomes were the complete remission (CR) and relapse-free survival (RFS) rates, and significant factors associated with OS and RFS.

Statistical Analysis for the Cohort Study
All data analyses were designed a priori and performed using IBM SPSS Statistics for Windows, version 20.0 (Armonk, NY, USA: IBM Corp.). The sample size calculation was based on the OS rate from our previous pilot study. Details of the calculation are demonstrated in Supplementary Data S1. Demographic and baseline characteristics were summarized descriptively using 2 categories (RUNX1 mut and RUNX1 wt AML). Continuous variables are presented as medians and interquartile ranges or means ± standard deviations, depending on the data type. The Mann-Whitney U test was employed to compare continuous data. Categorical data are presented as numbers and percentages and compared using Fisher's exact or Chi-squared tests. A log-rank test was used to compare the factors correlated with OS and RFS, and the results are presented as a Kaplan-Meier survival curve. Cox proportional hazards analysis was used to compare the predictors of survival outcomes in the univariate and multivariate analyses. The independent variables with significance in the univariate analysis were entered into the multivariate model. The results are expressed as hazard ratios (HRs) and 95% confidence intervals (CIs). Statistical significance was determined as a probability (p) value of <0.05.

Data Sources and Searches
Three investigators (T.R., W.O., and T.S.) independently searched for published articles indexed in the MEDLINE, Embase, and Cochrane Library databases from inception to April 2022. The search strategy used the terms "acute myeloid leukemia" and "molecular". Supplementary Data S2 details the strategy for each database. The study was performed following the PRISMA (Preferred Reporting Items for Systematic Reviews and Metaanalysis) statement (Supplementary Data S3) [20].

Selection Criteria and Data Extraction
To qualify for the meta-analysis, studies had to be either randomized controlled studies or cohort studies (prospective or retrospective) and have 2 groups of patients. The first group was RUNX1 mut patients, and the other group was patients with RUNX1 wt . Selected studies needed to report at least 1 of these outcomes: CR, OS, RFS, or EFS rates. Subgroup analyses included patients with de novo AML and those with intermediate-risk stratification, according to the 2017 ELN recommendations [21]. Three investigators (T.R., W.O., and T.S.) independently selected relevant articles and performed data extractions. Disagreements or questions regarding the eligibility of individual articles were discussed between the 3 investigators until a consensus was reached. Two investigators (T.R. and W.O.) subsequently examined the baseline characteristic data and the outcomes of all included studies, with the extracted data cross-checked to avoid inaccuracies.

Quality Assessment
Quality assessment and risk of bias were assessed using Risk of Bias in Nonrandomized Studies of Interventions (ROBINS-I) [22].

Statistical Analysis for the Meta-Analysis
Review Manager 5.4 software from the Cochrane Collaboration (London, UK) was used to analyze the data. The effect was estimated and combined with 95% CIs using the Mantel-Haenszel method [23]. Cochran's Q statistic was calculated, and the statistical heterogeneity among the studies was estimated using the I 2 statistic. The 4 levels of heterogeneity were based on the value of I 2 as follows: (1) insignificant heterogeneity (I 2 of 0-25%); (2) low heterogeneity (I 2 of 26-50%); (3) moderate heterogeneity (I 2 of 51-75%); and (4) high heterogeneity (I 2 of 76-100%) [24]. A random-effects model was applied based on the assumption that there was heterogeneity in each study due to individual patient characteristics, treatment differences, and disease risk stratification [24]. p values less than 0.05 were considered statistically significant. Funnel plots and Egger regression were used to detect publication biases [25]. The study protocol was registered with the International Prospective Register of Systematic Review (PROSPERO) network (CRD42022327857).

Terminology
Our patients' diagnoses and risk stratification were based on cytogenetic and molecular findings according to the 2017 ELN classifications [21]. Secondary AML in this cohort consisted of cases of AML that had an antecedent hematological disorder, prior chemotherapy, or radiation therapy [26]. CR was defined as bone marrow blasts below 5%, the absence of extramedullary blasts, blasts with extramedullary disease with an absolute neutrophil count exceeding 1000/µL, and a platelet count of more than 100,000/µL [21]. The duration between diagnosis and either death or the last follow-up was defined as OS. The duration from the CR date to relapse or death from any cause gave the RFS. The EFS duration was measured from the date of diagnosis to treatment failure, disease relapse, or death from any cause [27].

Figure 1. Kaplan-Meier curves of overall survival (A) and relapse-free survival (B). Panel A shows
Kaplan-Meier curves for overall survival corresponding to the presence or absence of the RUNX1 mutation. The median overall survival of AML patients with RUNX1 mut was 9.1 months, while it was 12.9 months in patients with RUNX1 wt (p = 0.268). Panel B shows Kaplan-Meier curves for relapse-free survival as assessed by the investigator. The median relapse-free survival was 7.8 months in the RUNX1 mut group and 14.6 months in the RUNX1 wt group (p = 0.481).
Patients aged more than 60 years and the presence of poor-risk cytogenetics, the FLT3-ITD mutation, or the SRSF2 mutation were significantly associated with poorer outcomes in the RFS analysis. However, RUNX1 mutation had no impact on RFS (HR: 1.37; 95% CI: 0.41-4.62; p = 0.604; Table 2). Some factors influencing RFS outcomes might not have reached statistical significance due to the limited number of patients.
Funnel plots of the CR, OS, RFS, and EFS outcomes of the AML patients with RUNX1 mut and RUNX1 wt were relatively symmetrical and showed no publication bias (Supplementary Data S6). We did not conduct any other funnel plot symmetrical tests since such tests are not recommended when the standard errors of the intervention effect es- timates are approximate [48]. For other biases, ROBINS-I was used for the assessments (Supplementary Data S7).

Baseline Patient Characteristics
A total of 8022 patients were included: 1093 RUNX1 mut AML patients and 6929 RUNX1 wt AML patients. The age of the participants varied markedly in both groups (from 11 to 92 years). Most patients (92.2%) had de novo AML, while 7.8% had secondary AML. According to the 2017 ELN risk classifications, the participants comprised 646 patients with favorable risk, 3861 patients with intermediate risk, and 1404 patients with poor risk. However, there were 2111 patients whose prognostic details were unavailable. Three common molecular mutations were the NPM1 mutation, FLT3-ITD mutation, and DNMT3A mutation. Most patients received standard induction chemotherapy, including the 3 + 7 regimen and the high-dose cytarabine regimen. The characteristics of the studies and patients are summarized in Table 3.     Abbreviations: AML-acute myelogenous leukemia; ECOG-Eastern Cooperative Oncology Group; F-female; HMA-hypomethylating agent; HSCT-hematopoietic stem cell transplantation; NA-not available; No.-number; PRO-prospective cohort study; RET-retrospective cohort study; RUNX1 mut -RUNX1 mutation; RUNX1 wt -RUNX1 wild type. TAD-9: thioguanine 100 mg/m 2 twice daily on days 3-9, cytarabine 100 mg/m 2 continuous infusion on days 1 and 2 and 100 mg/m 2 twice daily on days 3-8, and daunorubicin 60 mg/m 2 on days 3-5; HAM: cytarabine 3 g/m 2 twice daily on days 1-3 and mitoxantrone 10 mg/m 2 on days 3-5.
This was despite the CR rate being significantly lower for the RUNX1 mut patients (RR: 0.36;

Discussion
Our prospective study found higher proportions of AML patients with advanced age, secondary AML, and intermediate-risk cytogenetics among those with RUNX1 mut than among those with RUNX1 wt . These findings are consistent with the results of previous investigations [9,13,15,29,33]. Regarding molecular mutations, the ASLX1 mutation was found to be a significant co-mutation in the RUNX1 mut arm. Our data show that FLT3-ITD and TP53 mutations correspond to worse prognostic factors for OS and RFS, in parallel with the risk stratification from the 2017 ELN recommendations [21]. Currently, the FLT3-ITD mutation is deemed an intermediate risk by the 2022 ELN recommendations [5]. However, our investigation found that this mutation was associated with poor outcomes. Unfortunately, the patients in our study with the FLT3-ITD mutation could not access an FLT3 inhibitor for their treatment because they could not afford the medication. Furthermore, among the few patients with allogeneic stem cell transplantations in the present investigation's cohort, the OS and RFS of the patients with RUNX1 mut were comparable to those with RUNX1 wt . We hypothesize that our results might represent the sole impact of RUNX1 mut because less than 20% of the AML patients in our cohort had secondary AML, which was an independent risk factor for lower response rates, OS, and EFS [49]. Unlike AML with mutated TP53, which is a novel subtype with a homogeneously dismal prognosis, AML patients with RUNX1 mut seemed to have a heterogeneous prognosis, depending on the host and disease factors [50][51][52][53][54][55].
A systematic review and meta-analysis conducted in 2018 revealed that OS and disease-free survival were almost twice as poor in patients with RUNX1 mut ; however, only a few studies were analyzed [12]. In addition, the meta-analysis did not conduct a subgroup analysis of AML patients without poor risk factors, such as adverse cytogenetic risk and

Discussion
Our prospective study found higher proportions of AML patients with advanced age, secondary AML, and intermediate-risk cytogenetics among those with RUNX1 mut than among those with RUNX1 wt . These findings are consistent with the results of previous investigations [9,13,15,29,33]. Regarding molecular mutations, the ASLX1 mutation was found to be a significant co-mutation in the RUNX1 mut arm. Our data show that FLT3-ITD and TP53 mutations correspond to worse prognostic factors for OS and RFS, in parallel with the risk stratification from the 2017 ELN recommendations [21]. Currently, the FLT3-ITD mutation is deemed an intermediate risk by the 2022 ELN recommendations [5]. However, our investigation found that this mutation was associated with poor outcomes. Unfortunately, the patients in our study with the FLT3-ITD mutation could not access an FLT3 inhibitor for their treatment because they could not afford the medication. Furthermore, among the few patients with allogeneic stem cell transplantations in the present investigation's cohort, the OS and RFS of the patients with RUNX1 mut were comparable to those with RUNX1 wt . We hypothesize that our results might represent the sole impact of RUNX1 mut because less than 20% of the AML patients in our cohort had secondary AML, which was an independent risk factor for lower response rates, OS, and EFS [49]. Unlike AML with mutated TP53, which is a novel subtype with a homogeneously dismal prognosis, AML patients with RUNX1 mut seemed to have a heterogeneous prognosis, depending on the host and disease factors [50][51][52][53][54][55].
A systematic review and meta-analysis conducted in 2018 revealed that OS and disease-free survival were almost twice as poor in patients with RUNX1 mut ; however, only a few studies were analyzed [12]. In addition, the meta-analysis did not conduct a subgroup analysis of AML patients without poor risk factors, such as adverse cytogenetic risk and secondary AML. Since that meta-analysis, several studies have been published. Our systematic review and meta-analysis gathered all existing studies; 4 to 245 RUNX1 mut AML patients were in each study cohort, with a median of 26 cases per cohort. The number of RUNX1 mut AML patients in the present investigation was as high as the median (26) of the included studies. Most of the included studies were conducted on Western populations. To better understand the prognosis of AML with RUNX1 mut in non-high-risk groups, we conducted the present meta-analysis using the included studies combined with our cohort study's results to better determine real-world outcomes.
As expected, RUNX1 mut is a contributing factor to unpleasant prognoses. The de novo RUNX1 mut AML patients with intermediate-risk cytogenetics had comparable OS and RFS to the RUNX1 wt group, despite the RUNX1 mut AML patients having a lower CR rate. As stated in the 2017 and 2022 ELN AML risk stratifications, patients with RUNX1 mut should undergo allergenic stem cell transplantation [5,8]. Integrated, measurable residual disease (MRD) monitoring was recommended by the European LeukemiaNet MRD Working Party of 2021 as a surrogate marker for guiding the treatment of AML patients [56]. Patients with intermediate-risk cytogenetics will be considered for consolidation chemotherapy or autologous stem cell transplant if their MRD status is negative after two cycles of chemotherapy [56]. Venditti et al. demonstrated outcomes that supported the use of autologous stem cell transplantation in patients with favorable risk and intermediate risk with MRD negativity [57]. Concerning our study outcomes, autologous stem cell transplantation might be the treatment paradigm for RUNX1 mut de novo AML patients with intermediate-risk cytogenetics who achieve negative MRD. This approach would be beneficial in countries where the availability of allogeneic stem cell transplantations is limited.
Our study had some limitations. First, the impact of co-genetic mutations could not be explored either in the prospective study or the meta-analysis. A previous study showed the significant impact of co-mutations on prognosis [4]. Unfortunately, several studies in our meta-analysis did not detail mutations other than RUNX1 mut nor the impact of co-mutations. Second, some baseline characteristics, such as induction treatment regimens, MRD monitoring, and hematopoietic stem cell transplantation, were missing from observational studies. Third, limited studies were available for our subgroup analyses. Fourth, the subgroup analysis of the outcomes of patients who received either targeted therapy or allogeneic stem cell transplantation could not be evaluated in our meta-analysis due to a lack of data. Additionally, our prospective study included a small number of patients with relatively short follow-up periods. Lastly, we can see from Figure 1 that the hazards in patients with RUNX1 wt and RUNX1 mut were not proportionate. This finding opposes our assumption and may have led to an overestimated HR size.

Conclusions
Owing to the high proportion of secondary AML and elderly patients among AML patients with RUNX1 mut , our study affirmed the poor prognosis of this mutation. However, the survival outcomes of de novo AML patients with intermediate-risk cytogenetics in the RUNX1 wt and RUNX1 mut groups might be similar. Informed Consent Statement: Informed consent was obtained from all patients before being included in the prospective study of our cohort.

Data Availability Statement:
The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.