Health-Related Quality of Life as Assessed by the EQ-5D-5L Predicts Outcomes of Patients Treated with Azacitidine—A Prospective Cohort Study by the AGMT

Simple Summary The EuroQol 5-Dimension 5-level (EQ-5D-5L) questionnaire is a globally used and multiply validated tool to assess health-related quality of life (HRQoL), but data on its use for patients with myeloid neoplasias is scarce. The aim of this prospective cohort study was to alleviate this knowledge gap. Our data show in a homogenous population of azacitidine-treated patients for the first time that (1) myeloid patients have significantly worse HRQoL than a population norm (i.e., a representative sample of the German general adult population) from a similar geographic region, matched by age, sex and number of comorbidities; (2) The EQ-5D-5L questionnaire response provides added prognostic value to the International Prognostic Scoring System (IPSS) and the revised IPSS (R-IPSS), which are longstanding gold standards of prognostication in these diseases; (3) the multivariate-adjusted significant predictive value of the EQ-5D-5L response parameters on patient outcomes including response to azacitidine, time to next treatment and overall survival; (4) longitudinal assessment of the EQ-5D-5L response/clinical parameter pairs revealed significant additional, independent associations. Abstract In this prospective study (NCT01595295), 272 patients treated with azacitidine completed 1456 EuroQol 5-Dimension (EQ-5D) questionnaires. Linear mixed-effect modelling was used to incorporate longitudinal data. When compared with a matched reference population, myeloid patients reported more pronounced restrictions in usual activities (+28%, p < 0.0001), anxiety/depression (+21%, p < 0.0001), selfcare (+18%, p < 0.0001) and mobility (+15%, p < 0.0001), as well as lower mean EQ-5D-5L indices (0.81 vs. 0.88, p < 0.0001), and lower self-rated health on the EuroQol Visual Analogue Scale (EQ-VAS) (64 vs. 72%, p < 0.0001). After multivariate-adjustment, (i) the EQ-5D-5L index assessed at azacitidine start the predicted time with clinical benefit (TCB) (9.6 vs. 6.6 months; p = 0.0258; HR = 1.43), time to next treatment (TTNT) (12.8 vs. 9.8 months; p = 0.0332; HR = 1.42) and overall survival (OS) (17.9 vs. 12.9 months; p = 0.0143; HR = 1.52); (ii) Level Sum Score (LSS) predicted azacitidine response (p = 0.0160; OR = 0.451) and the EQ-5D-5L index showed a trend (p = 0.0627; OR = 0.522); (iii) up to 1432 longitudinally assessed EQ-5D-5L response/clinical parameter pairs revealed significant associations of EQ-5D-5L response parameters with haemoglobin level, transfusion dependence and hematologic improvement. Significant increases of the likelihood ratios were observed after addition of LSS, EQ-VAS or EQ-5D-5L-index to the International Prognostic Scoring System (IPSS) or the revised IPSS (R-IPSS), indicating that they provide added value to these scores.


Introduction
Azacitidine is the first treatment to be associated with improved overall survival (OS), and to be approved by both US and European regulatory authorities for the treatment of patient subgroups with myeloid neoplasms. In patients with myelodysplastic syndromes (MDS) [1,2] and chronic myelomonocytic leukaemia (CMML) [3], it remains the only approved disease-modifying therapeutic substance, whereas several new drugs have recently been approved for certain patient subgroups with acute myeloid leukaemia (AML).
Globally, there has been a distinctive shift towards taking patient perspectives into account when making (regulatory) healthcare and treatment decisions. Traditional clinical ways of measuring health and the effects of treatment are thus increasingly being accompanied by patient-reported outcome measures. In the broad field of the latter, the generic EuroQol 5-Dimension (EQ-5D) questionnaire is multiply validated and globally has been the most used tool in many areas of medicine, including oncology, for over three decades.
Although regulatory agencies offered guidance for the use of patient-reported outcome measures to support labelling claims as early as 2005 [4][5][6][7], the European LeukemiaNet pointed to the importance of assessing HRQoL in the clinical management of patients with MDS in 2013 [8], and the EQ-5D has been the preferred measure of HRQoL for the UK National Institute for Health and Care Excellence since 2008, published reports on HRQoL data in MDS, CMML and AML are scarce. There are 4 publications in MDS, and 10 in AML, 13 of which only report on the mean EQ-VAS and/or the mean or median EQ-5D-5L index value. Only one report assessed the impact of the EQ-5D-5L index on a time-to-event endpoint [9], and only one publication provided details on non-composite results [10], both in patients with lower-risk MDS (Table 1). Publications correlating EQ-5D-5L measures with treatment outcomes in general, and with azacitidine-related outcomes in particular, are lacking to date. The only detailed EQ-5D-5L data on this topic stem from this report. EQ-5D-5L index, LSS and EQ-VAS were significantly associated with overall survival and the likelihood to respond to azacitidine in univariate analysis; EQ-5D-5L index was significantly associated with overall survival, time with clinical benefit and time to next treatment in multivariate-adjusted analyses. LSS was significantly associated with the likelihood to respond to azacitidine in multivariate analysis.

AML
Uyl-de Groot C.A. [13] 1998 NR NR AML 3L 70.6 (NR) 64 EQ-5D-5L index, LSS and EQ-VAS were significantly associated with overall survival and the likelihood to respond to azacitidine in univariate analysis; EQ-5D-5L index was significantly associated with overall survival, time with clinical benefit and time to next treatment in multivariate-adjusted analyses. LSS was significantly associated with the likelihood to respond to azacitidine in multivariate analysis.
NR indicates not reported.
In this prospective study, we compared EQ-5D-5L responses between patients with MDS, CMML and AML and a population norm (i.e., a representative sample of the German general adult population without myeloid (or other) neoplasias) from a similar geographic region, matched by age, sex and number of comorbidities. In myeloid patients treated with azacytidine within the Austrian Registry of Hypomethylating Agents, we performed more detailed analyses and assessed (1) whether EQ-5D-5L composite variables provided added value to the (R)-IPSS; (2) there might be a predictive value of EQ-5D-5L composite variables (including LSS, EQ-VAS and EQ-5D-5L index value) on the response to azacitidine and several time-to-event endpoints; and (3) performed longitudinal assessments of EQ-5D-5L response/clinical parameter pairs.

Study Design and Participants
In this prospective cohort study, data from non-selected, consecutive patients were provided by seven Austrian centres (Supplementary p. 1) participating in the Austrian Registry of Hypomethylating Agents of the Austrian Group for Medical Tumour Therapy (AGMT) Study Group (NCT01595295; ethics committee approval 415-EP/39/Feb-2009; details published previously [23,24]; Figure 1).
The EQ-5D consists of five questions (also known as dimensions (5D): mobility, selfcare, usual activities, pain/discomfort, anxiety/depression) with 5 levels (5L) of problem severity in the responses, as well as a Visual Analogue Scale (EQ-VAS) aiming to capture a respondents' rating of their 'health today' on a scale from 0-100. The composite scores Level Sum Score (LSS) and EQ-5D-5L index are explained in Supplementary p. 2. The EQ-5D questionnaires were assessed at the start of azacitidine treatment cycles. The EQ-5D-5L results of patients diagnosed with MDS, CMML or AML were compared with those of a German population norm (i.e., a representative sample of the German general adult population without myeloid (or other) neoplasias) [28]. The EQ-5D-5L German value set [29] and the reverse crosswalk tool provided by EuroQol on November 16, 2020 were used for calculation of the EQ-5D-5L indices (Supplementary p. 3).
Patients with an EQ-5D available at azacitidine treatment start were stratified according to their LSS, EQ-VAS or EQ-5D-5L index being </≥ the respective group median.

Statistical Analyses
Only observed values were analysed. Baseline and treatment-related factors were compared using the χ 2 test for categorical variables and the Wilcoxon test for continuous variables. Patient subgroups were compared using the log-rank test. All p-values and 95% CIs are two-sided. The threshold for statistical significance was 0.05. Time-to-event endpoints were analysed using the Kaplan-Meier method.
Proceeding in analogy to Efficace et al. [30] who demonstrated that self-reported fatigue provided added value to the IPSS and R-IPSS in patients with MDS, the likelihood ratio (LHR) test was used to determine whether EQ-5D-5L response parameters provided added value to the IPSS or R-IPSS.
The prognostic information provided by LSS, EQ-VAS and EQ-5D-5L index, with regards to whether a patient is likely to respond to azacitidine or not was assessed by univariate and multivariate-adjusted logistic regression analyses. Cox regression models for time-to-event endpoints were applied.
To identify variables that might be associated with patient-reported outcomes, linear mixed-effect modelling was utilised, with patient identity as the grouping variable. p-values were visualised using heatmaps.
Sensitivity analyses were performed to check the general conclusions by assessing different endpoints (response subtypes, OS, TCB, TTNT), and assessing both continuous and dichotomised variables. The definition of outcomes and further statistical details are given in Supplementary pp. 4-7. March 2022. Database lock (last patient in) was on 13 December 2020. The inclusion criteria were (1) the diagnosis of MDS, CMML or AML, which was independently and centrally verified on the basis of submitted data; (2) treatment with azacitidine; (3) inclusion in the Austrian Registry of Hypomethylating agents; (4) the presence of a written informed consent for all patients alive at the time of data entry; (5) age ≥ 18 years; (6) the completion of at least one EQ-5D questionnaire. No data from patients <18 years were received. No patients fulfilling these criteria were excluded from the analyses. A total of 6 of 1456 (0.4%) of EQ-5D questionnaires were excluded (empty questionnaire). Permissions to use the German version of EQ-5D questionnaires was obtained from EuroQol. All data for this study were collected prospectively. This study has been reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. To ensure uniformity, composite variables based on provided data were allocated for each individual patient at the start of azacitidine treatment, including diagnosis of MDS, CMML or AML according to the WHO 2016 diagnostic criteria [24], cytogenetic risk group according to the International Prognostic Scoring System (IPSS) [25] and the revised IPSS (R-IPSS) [26] and the IPSS and R-IPSS risk categories themselves. (2) treatment with azacitidine; (3) inclusion in the Austrian Registry of Hypomethylating agents; (4) the presence of a written informed consent for all patients alive at the time of data entry; (5) age ≥ 18 years; (6) the completion of at least one EQ-5D questionnaire. No data from patients <18 years were received. No patients fulfilling these criteria were excluded from the analyses. A total of 6 of 1456 (0.4%) of EQ-5D questionnaires were excluded (empty questionnaire). Permissions to use the German version of EQ-5D questionnaires was obtained from EuroQol. All data for this study were collected prospectively. This study has been reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. To ensure uniformity, composite variables based on provided data were allocated for each individual patient at the start of azacitidine treatment, including diagnosis of MDS, CMML or AML according to the WHO 2016 diagnostic criteria [25], cytogenetic risk group according to the International Prognostic Scoring System (IPSS) [26] and the revised IPSS (R-IPSS) [27] and the IPSS and R-IPSS risk categories themselves.
Assign Data Management and Biostatistics GmbH performed statistical analyses with SAS ® 9.4. The Life & Medical Sciences Institute, University of Bonn performed statistical analyses including mixed-effect linear modelling with Python 3.8.12.

Myeloid Patient Characteristics
Data from 272 patients diagnosed with MDS, CMML or AML who were treated with azacitidine between 21 May 2007 and 21 December 2020 were prospectively analysed ( Figure 1). Of these, 205 had filled out an EQ-5D at azacitidine treatment start ( Figure 1). This subset was used for time-to-event endpoint analyses.

Patients Treated with Azacitidine Reveal Profound Impairments in HRQoL
Supplementary pp. 14-15 show the most frequent response patterns for questionnaires filed out at azacitidine treatment start and for all EQ-5D questionnaires. Supplementary p. 16 gives an overview of the EQ-5D responses by patient group, response status and number of azacitidine treatment cycles. The mean number of filled-out EQ-5D questionnaires per patient was 5.4 (SD 6.2), the median number was 3.0 (IQR 1.0-3.0).

Patients Treated with Azacitidine Reveal Profound Impairments in HRQoL
Supplementary pp. 14-15 show the most frequent response patterns for questionnaires filed out at azacitidine treatment start and for all EQ-5D questionnaires. Supplementary p. 16 gives an overview of the EQ-5D responses by patient group, response status and number of azacitidine treatment cycles. The mean number of filled-out EQ-5D questionnaires per patient was 5.4 (SD 6.2), the median number was 3.0 (IQR 1.0-3.0).

Comparison of HRQoL with a Reference Population Matched by Age, Sex and Number of Comorbidities
We compared HRQoL of the myeloid cohort with that of a German population norm (i.e., a representative sample of the German general adult population without myeloid (or other) Figure 2. EQ-5D-5L responses available at azacitidine treatment start (n = 205), stratified by ECOG-PS. Table 2. Prevalence of problems in patients with myeloid neoplasias (assessed by EQ-5D-5L at azacitidine treatment start (n = 205) 1 ) by disease-related and patient-related parameters.   1 EQ-5D in cycle 1 or 2 with a non-missing value for the respective parameter (hence patient numbers may vary slightly for each parameter analysed). 2 Problems were defined as answer options 2, 3, 4 or 5 for EQ-5D-5L and answer options 2 or 3 for EQ-5D-3L. 3 Represents the numerical sum of all EQ-5D responses. 4 The EQ-5D-5L index is measured on a scale from 0 to 1, whereby 0 indicates death and 1 perfect health. 5 Baseline parameters and EQ-5D-5L results were compared using the Chi-squared test (based on non-missing observations) for EQ-5D-5L problems (=2,3,4,5) vs. EQ-5D-5L no-problems (=1). 6 Baseline parameters and EQ-5D-5L results were compared using the Wilcoxon rank-sum test (also called Mann-Whitney U-test or Mann-Whitney-Wilcoxon Test) for Level Sum Score, EQ-5D-5L index value and EQ-VAS. Font color is red for all significant p-values <0.05.

Comparison of HRQoL with a Reference Population Matched by Age, Sex and Number of Comorbidities
We compared HRQoL of the myeloid cohort with that of a German population norm (i.e., a representative sample of the German general adult population without myeloid (or other) neoplasias) [28] with a similar ethnical and socioeconomic background (Figure 1). Myeloid patients reported more pronounced restrictions in mobility (51 vs. 35%, p < 0.0001), selfcare (25 vs. 7%, p < 0.0001), usual activities (56 vs. 28%, p < 0.0001) and anxiety/depression (+15%, p < 0.0001), as well as lower mean EQ-5D-5L indices (0.81 vs. 0.88, p < 0.0001) and lower self-rated health on EQ-VAS (64 vs. 72%, p < 0.0001) than the German population norm (Table 3). These significant differences could also be observed after stratification by age group, sex or number of comorbidities (Table 3).

IPSS and R-IPSS Prognosticate OS and TTNT
Myeloid patients with lower-risk IPSS had significantly longer unadjusted survival than patients with higher-risk IPSS (

EQ-5D-5L Composite Scores at Azacitidine Start Provide Added Value to the (R)-IPSS
For the endpoint OS, significant increases of the likelihood ratio (LHR) were observed after addition of (i) the LSS to the IPSS (LHR increased from 7.32 to 10.69; p = 0.0048) or the R-IPSS (LHR increased from 5.37 to 9.05; p = 0.0108); (ii) the EQ-VAS to the IPSS (LHR increased from 7.32 to 11.56; p = 0.0031) or the R-IPSS (LHR increased from 5.37 to 10.28; p = 0.0058); (iii) the EQ-5D-5L index to the IPSS (LHR increased from 7.32 to 13.02; p = 0.0015) or the R-IPSS (LHR increased from 5.37 to 13.48; p = 0.0012), indicating that they provided added value to the IPSS and R-IPSS (Table 4, grey shaded columns).
For the endpoint TTNT, significant increases of the LHR were observed after addition of (i) the LSS to the R-IPSS (LHR increased from 4.31 to 7.74; p = 0.0209); (ii) the EQ-VAS to the R-IPSS (LHR increased from 4.31 to 6.85; p = 0.0327); (iii) the EQ-5D-5L index to the IPSS (LHR increased from 3.60 to 6.38; p = 0.0411) or the R-IPSS (LHR increased from 4.31 to 6.55; p = 0.0378), indicating that they provided added value to the IPSS and R-IPSS (Table 4, grey shaded columns).

EQ-5D-5L Composite Scores at Azacitidine Start Impact Time-to-Event Endpoints
Myeloid patients with an EQ-5D available at azacitidine treatment start (n = 205) were stratified according to their LSS, EQ-VAS or EQ-5D-5L index being </≥ the respective group median. In unadjusted analyses, patients with (i) an LSS < 8.0 at azacitidine treatment start had significantly longer OS and showed a trend for longer TCB and TTNT; (ii) an EQ-VAS < 65 at azacitidine treatment start had significantly longer OS; (iii) an EQ-5D-5L index ≥0.8845 had significantly longer OS, longer TCB and longer TTNT ( Table 5, first four columns) ( Figure 3A,C,E). Table 3. Comparison of HRQoL (as assessed by first available EQ-5D-5L) 1 between myeloid patients (n = 269) and a German population norm without myeloid neoplasias (n = 5001) 2 matched by age group, sex or number of comorbidities.  EQ-VAS indicates EuroQol Visual Analogue Scale. 1 First available EQ-5D with a non-missing value for the respective parameter (hence patient numbers may vary slightly for each parameter analysed). 2 Published and unpublished data provided by Grochtdreis et al. [28]. 3 Problems were defined as answer options 2, 3, 4 or 5 for EQ-5D-5L and answer options 2 or 3 for EQ-5D-3L. 4 The prevalence of EQ-5D-5L problems (=2,3,4,5) vs. EQ-5D-5L no-problems (=1) were compared using the Chi-squared test. 5 EQ-5D-Indices and EQ-VAS were compared between the Austrian Registry of Hypomethylating Agents and the German Norm cohorts using Student's T-test. Font color is red for all significant p-values <0.05. Table 4. Prognostic value of the IPSS and R-IPSS with or without baseline Level Sum Score (LSS), EQ Visual Analogue Scale (VAS) or EQ-5D-5l index value, by time-to-event endpoint (patients with EQ-5D-5L responses available at azacitidine treatment start (n = 205)).    After multivariate adjustment (for ECOG-PS, number of comorbidities, platelet count ≤30 G/L or transfusion dependence, peripheral blood blasts, azacitidine treatment line and azacitidine dose in cycle one) patients with an EQ-5D-5L index above the group median (i.e., ≥0.8845) had significantly longer OS (17.9 (Figure 3B,D,F).

Longitudinal Assessment of EQ-5D-5L Responses and Clinical Parameters
Multivariate-adjusted mixed-effect linear models of up to 1432 longitudinally assessed EQ-5D-5L response/dichotomised clinical parameter pairs revealed significant associations for haemoglobin level, red blood cell transfusion dependence, platelet count, platelet transfusion dependence, levels of ferritin, bilirubin, albumin, cholinesterase, the occurrence of adverse events, number of days with azacitidine treatment, and haematologic improvement (HI-any, HI-E, HI-P) with at least two EQ-5D dimensions, and at least one of the EQ-5D composite variables (LSS, EQ-VAS, EQ-5D-5L index) (Figure 4, Table 7). Sensitivity analyses for continuous clinical parameters yielded similar results (Supplementary pp. 17-18).

Minimally Clinically Important Differences
Of the statistically significant associations found in the dichotomised analyses, the following exhibited an effect size equal to or larger than the minimally clinically important difference: platelet transfusion dependence (LSS), ferritin ≥1000 µg/L (LSS), albumin ≥3.4 mg/dL (LSS), adverse events grade 3-4 (LSS, EQ-5D-5L index) and cholinesterase ≥2.5 U/L (EQ-VAS). These findings were corroborated in sensitivity analyses using continuous parameters. .05 and ≤0.065) of the corresponding multivariate-adjusted mixed-effect linear models using EQ-5D-5L responses as endogenous variables (x-axis), and various clinical measurements as exogenous variables (y-axis). Multivariate adjustment was performed by admitting the following variables remaining in the final Cox model as covariates: ECOG-PS, number of comorbidities, platelet count/platelet transfusion dependence, peripheral blood blasts, azacitidine treatment line and azacitidine dose in cycle one.

Minimally Clinically Important Differences
Of the statistically significant associations found in the dichotomised analyses, the following exhibited an effect size equal to or larger than the minimally clinically important difference: platelet transfusion dependence (LSS), ferritin ≥1000 µg/L (LSS), albumin ≥3.4 mg/dL (LSS), adverse events grade 3-4 (LSS, EQ-5D-5L index) and cholinesterase ≥2.5 U/L (EQ-VAS). These findings were corroborated in sensitivity analyses using continuous parameters.

Discussion
To our knowledge, our group is the first to compare EQ-5D-5L data of patients with MDS, CMML or AML with data from a reference population from a similar ethnic, socioeconomic and geographic background. In this prospective cohort analysis, we found that patients treated with azacitidine had significantly worse HRQoL than the German population norm (i.e., a representative sample of the German general adult population) [28,29] matched by sex, age group and number of comorbidities. In contrast to observations by Stauder et al. [10] who used the EQ-5D-3L, all significant differences observed for the EQ-5D-5L index and the EQ-VAS fulfilled the definitions of the minimally clinically important difference used by that group (>0.03 on the index and >3.0 on the EQ-VAS).
The current gold standards of prognostication in patients with MDS/CMML and low blast count AML are the International Prognostic Scoring System (IPSS) [26] and the revised IPSS (R-IPSS) [27]. The clinical relevance of these scores is underscored by the fact that approval of azacitidine for MDS patients in Europe is restricted to those with higher-risk IPSS (i.e., intermediate-2 and high risk categories). To our knowledge, our data are the first to indicate that LSS, EQ-VAS and EQ-5D-5L index at azacitidine treatment start provided added value to the IPSS and R-IPSS for the endpoints OS and TTNT. Other groups have prominently shown that patient-reported outcomes (other than EQ-5D) may predict OS and/or add value to the (R)-IPSS in elderly patients with MDS [30][31][32] or AML [21]. However, these questionnaires/indices incorporate 30 [30,32], 42 [31] and 44 items [21], many of which are not routinely assessed in patients with myeloid neoplasms, thus hampering the clinical everyday utility outside of clinical trials.
Our data are the only information on the impact of HRQoL, as assessed by the EQ-5D-5L, on time-to-event endpoints of patients treated with azacitidine. After multivariate adjustment (for ECOG-PS, number of comorbidities, platelet count ≤30 G/L or transfusion dependence, peripheral blood blasts, azacitidine treatment line and azacitidine dose in cycle one) an EQ-5D-5L index <0.8845 at azacitidine start indicated a significantly shorter median survival (−5.0 months), an increased risk of death (+52%), significantly shorter azacitidine treatment duration (−3.0 months), shorter TTNT or death (−3.0 months) and a significantly higher risk of requiring a next treatment or dying (+42%). Our data further show that an LSS of ≥8 at azacitidine start indicates a significantly lower chance of responding to the drug (OR 0.45).
This is the first report on the longitudinal assessment of EQ-5D-5L responses with clinical parameters. Multivariate adjusted mixed-effect linear modelling revealed significant associations for EQ-5D-5L response parameters with clinical parameters associated with haematologic improvement, disease progression, or the occurrence of adverse events. Thus, these data show that quality of life ameliorated in responding patients and deteriorates in patients experiencing disease progression or grade 3-4 adverse events. It is difficult to interpret these findings compared with the wider literature as longitudinal analyses of HRQoL data on patients with MDS, CMML or AML are scarce, performed with questionnaires other than EQ-5D-5L and are often without multivariate adjustment. Efficace et al. found no association between ferritin levels and HRQoL as assessed by EORTC QLQC30 both at baseline and during the study period in heavily transfused patients with MDS treated with iron chelation therapy using linear mixed-effect models [33]. We observed significant associations of ferritin, bilirubin and albumin levels with problems in six of eight EQ-5D-5L dimensions/composite variables. This is the first indication that these clinical variables, two of which (hypalbuminaemia and hyperferritinaemia) have been shown to be associated with adverse prognosis in patients with MDS [34][35][36], CMML [37] or AML [38,39] correlate with HRQoL.
Little is known of the longitudinal effect of azacitidine on patients' HRQoL, but recent publications demonstrating significant improvements of EQ-VAS and/or EQ-5D-5L index in patients responding to treatment in other malignancies [40,41] highlight the contemporality and clinical relevance of the topic.
The mean (SD) EQ-VAS and EQ-5D-5L index values of our cohort were similar to those previously reported in patients with MDS/CMML or AML. Problems (slight, moderate, severe or extreme) were most commonly self-reported in the dimensions of usual activities (59%), mobility (51%), pain/discomfort (50%), anxiety/depression (49%) and selfcare (22%). Similar to the lower-risk MDS population reported by Stauder et al. [10], (i) MDS, CMML and AML patients in our cohort had the least problems in the dimension of selfcare, (ii) no correlation could be found between IPSS or R-IPSS risk group at azacitidine treatment start and EQ-5D responses, and (iii) patient-related factors such as haemoglobin <10 g/dL, red blood cell transfusion dependence, ECOG-PS ≥ 2 and high-risk HCT-CI were found to be associated with significantly more problems in several dimensions and/or significantly worse EQ-5D-5L composite variables. We could, however, not find a significant difference in EQ-5D-5L response by sex or age group.
A limitation of this study is that we cannot speculate what the HRQoL would have been without azacitidine therapy. Furthermore, this question cannot be addressed by real-world evidence or by future randomised clinical trials due to ethical reasons. A further limitation is that we do not have EQ-5D-5L questionnaires for all patients for all treatment cycles. However, to impose mandatory pre-specified required time-points for filling out EQ-5D-5L questionnaires would be against the non-interventional nature of non-interventional studies in general, and of the Austrian Registry of Hypomethylating Agents in particular. Furthermore, these results cannot, eo ipso, be generalised to other treatments of patients with MDS, CMML or AML, as we exclusively studied HRQoL of patients treated with azacitidine. In the future, we aim to analyse EQ-5D-5L responses in myeloid patients irrespective of treatment type within the Austrian Myeloid Registry (NCT04438889; Ethics committee approval was provided by the Ethikkommission für das Bundesland Salzburg (415-E/2581/Feb-2020)), which is a disease-specific (rather than a drug-specific) registry, once sufficient data have been accumulated, and are open for collaborations with other study groups in this regard.
The strengths of this study are that we report the first evidence-based data on all of the above; the prospective nature of data collection; the proven quality of our database in direct patient-level comparison with randomised phase-3 clinical trial data [23]; few missing data; calculation and validation of diagnosis, cytogenetic risk groups, and prognostic scores; response to reduce human errors; multivariate adjustment; longitudinal analyses; correction for multiple testing; and that additional sensitivity analyses confirmed the robustness of our results.

Conclusions
In conclusion, the current findings support the use of EQ-5D-5L instruments in future clinical trials and real-world evidence databases, in order to fully consider all factors that can be potentially associated with treatment outcomes. They also extend knowledge on the safety and efficacy of azacitidine by showing that clinical benefits such as improvement of laboratory values associated with haematologic improvement, as well as haematologic improvement itself, correlate with improved HRQoL.  Table S11. Azacitidine treatment and response characteristics; Supplemental Table S12. Most frequent response patterns of EQ-5D questionnaires at azacitidine treatment start; Supplemental Table S13. Most fre-quent response patterns of all EQ-5D questionnaires; Supplemental Table S14. Overview of EQ-5D responses by patient group and responder status; Supplemental Table S15. Multivariate-adjusted longitudinal analyses of EQ-5D-5L results and continuous parameters per azacitidine treatment cycle using mixed-effects linear models; Supplemental Figure S1. Heatmap of p-values resulting from multivariate-adjusted mixed-effect linear models of longitudinally assessed EQ-5D-5L responses and concomitantly assessed continuous clinical parameters; The Ethics Commottee Approval statement; The protocoll of the Austrian Registry of Hypomethylating Agents; The signed sponsor approval page; The informed consent of the Austrian Registry of Hypomethylating Agents. (References  are cited in the Supplementary Materials). Funding: The Austrian Group for Medical Tumor Therapy (AGMT) is the sponsor for the Austrian Registry of Hypomethylating Agents and received funding from Celgene/BMS, AbbVie, and Janssen Cilag. The AGMT is a non-for-profit organisation and an academic study group. The group performed administrative and legal management, as well as funding acquisition. No pharmaceutical company and no other funding source were involved in any way and had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. No medical writer or editor was involved. we are also open to data contributions by others. Participation requests or potential joint research proposals can be made at any timepoint to the corresponding author via email (dr.lisa.pleyer@gmail.com) and are subject to approval by the AGMT and its collaborators.