We conducted a cross-sectional study of patients with leukemia in European, African, Asian, North American, and South American countries between 18 September 2021 and 07 January 2022. This study was conducted in accordance with the Declaration of Helsinki and International Society for Pharmacoepidemiology (ISPE) guidelines for Good Clinical and Pharmacoepidemiology Practice (GPP). The study protocol and informed consent form were reviewed and approved by each participating patient support group in their respective country. The ALAN Network based in Switzerland, was responsible for coordination and the contract research organization, IQVIA (UK & Ireland Healthcare and Government Division) was responsible for data management and analysis for the study. Patients were provided with written information explaining the study design and were asked to provide consent electronically before participating in the study. Adequate protections were taken to maintain the confidentiality of their responses. Data were managed in compliance with the General Data Protection Regulation and any regulations regarding management of personal data required by participants’ respective country of residence.

The STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) Checklist (24) was implemented in the design and reporting of the study ( Supplementary Table 1 ).

The study questionnaire was developed based on an existing patient pathway experience questionnaire (14). The ALAN Network also agreed to include a number of questions relating to the COVID-19 pandemic, as well as the validated established Hematological Malignancy Patient Reported Outcomes (HM-PRO) instrument (21–23). The HM-PRO contains items specific to QoL aspects impacted by hematological malignancies; these include physical, social, and emotional health, eating and drinking, symptoms, and side effects (21–23). The HM-PRO consists of two parts, assessing impact (Part A) and signs and symptoms (Part B) of hematological malignancies. Both scales have linear scoring systems ranging from 0 to 100, with higher scores representing greater (negative) impact on QoL and symptom burden. The HM-PRO recall period for Part A is “at the moment” (i.e., at present, today) and for part B is the last three days. A minimal clinically important difference (MCID) for HM-PRO in patients with hematological malignancies is 6.2 according to standard error of measurement (25). Overall, the study assessment tool consisted of 200 questions (some with sub questions, across 16 sections, including demographic information). This analysis reports on Part A of the patient questionnaire.

The data included socio-demographics, disease and treatment profile, and QoL. The results reported in this manuscript are those relevant to the study aims: to assess whether age and years living with the disease, are correlated to the HM-PRO scores of patients with a) chronic and b) acute leukemia.

The HM-PRO, as part of its original development in English language underwent extensive psychometric testing including content validity (21–23). Content validity is identified as the most important step of psychometric testing of a newly developed PRO instrument by FDA in their PRO Guidelines documents (26). The 10 translations of the HM-PRO were carried out by MAPI Research Trust following the established standard approach outlined in the ISPOR PRO instrument cross-cultural adaptation document (27). This process encompassed two forward and backward translations, consolidation at the end of each step and then arbitration by the developers of the original HM-PRO, followed by content validity (cognitive debriefing) involving patient with hematological malignancy. On finalization of cross-cultural adaption of the HM-PRO to the 10 languages, respective translation certificates were issued by the MAPI Research Trust.

This was a global study; participants were recruited through national patient support groups (i.e. members of ALAN, CLLAN, and CMLAN Networks), via email, social media, and newsletters. The questionnaires were made available in ten languages: Chinese, English, French, German, Hebrew, Italian, Korean, Portuguese (Brazilian), Russian, and Spanish. These official certified translations were used in all 76 countries. The study was of 16 weeks duration (18 September 2021 and 07 January 2022) and the patients were asked to only complete the sections that were relevant to their leukemia subtype. It was determined that a sample size of at least 350 patients would be sufficient to address the study objective based on 5% margin of error and a 95% confidence interval.

For all questions (with the exception of those asked in the form of “tick all that apply”) the percentage responses were calculated after excluding those respondents who did not answer that particular question.

The overall number of evaluable responses for questions which were presented in the form of “tick all that apply” was determined by the number of respondents eligible to respond. Missing responses were those where any eligible respondents chose not to select any options. Additionally, where applicable, scores have been recalculated to exclude non-specific responses (such as don’t know/can’t remember), or responses indicating that the question was not applicable to the participant’s circumstances. Where responses were missing, percentages are based on the number of patients who answered that question. Age was calculated by subtracting birth year from 2021; years living with the disease (disease duration) was calculated by subtracting year of diagnosis from 2021-.

Since the HM-PRO Part A scores were not normally distributed (right skewed), non-parametric statistical methods were used for the analysis. Two sample Wilcoxon rank-sum, or Kruskal-Wallis rank tests, were used to test for differences in scores between groups. Chi-squared (χ²) test was used to test differences between categorical variables.

The Spearman’ rank correlation coefficient (ρ) was used to determine the strength of relationships (and partial correlation) between the scores, patient age and years living with the disease (disease duration); an absolute value of 0.7 or greater indicated a strong relationship, 0.4 -0.7, a moderate relationship and between 0.2 and 0.4 a weak relationship.

The HM-PRO domain score (physical behavior score [PB SCORE], emotional behavior score [EB SCORE], social well-being score [SW SCORE], eating and drinking [ED SCORE]), and total HM-PRO [PARTA SCORE] were represented in score banding (no impact=0-7, small impact=8-25, moderate impact=26-41, very large impact = 42-74, extremely large impact=75-100) (28). The independent correlates of ordinal dependent variables (namely, PB SCORE, EB SCORE, SW SCORE, ED SCORE, and PARTA SCORE) were identified by preliminary testing the proportionality of odds associated to all covariates across response categories in multiple regression models. The proportionality assumption was tested by the likelihood-ratio (LR) test (29) considering as null hypothesis that there was no difference in the regression coefficients linking each covariate to the levels of the outcome variable (that is, no covariate has a disproportionate effect on a specific level of the dependent variable). If the null hypothesis is accepted (P value of LR test >0.05), an ordinal logistic regression was fitted for each outcome variable. In these models, the odds ratio indicates how much increases the odds of being in a higher level of the outcome variable, given that all the other variables in the model are held constant. If the null hypothesis is rejected (P value of LR test ≤0.05), different logistic models were fitted to describe the relationship between covariates and each pair of outcome categories appropriately grouped. For example, the PART - A SCORE categories are numbered 1, 2, 3, 4, and 5. The first panel of odds ratios can be interpreted as those from a binary logistic regression where the dependent variable is recoded as 1 vs 2 + 3 + 4 + 5. The second panel of odds ratio can be interpreted as those from a binary logistic regression where the dependent variable is recoded 1 + 2 vs 3 + 4 + 5, etc. Thus, in this analysis odds ratio are interpreted as in standard binary logistic models where categories of outcome variable are collapsed into two categories. Positive coefficients mean that higher values on the covariates make higher values on the dependent variable more likely. In these models, data were expressed as odds ratio (OR), standard error (SE), z value, P value, and 95% confidence intervals. Missing answers were random and were excluded for these tests. The probability of type I error was set at p <0.05. Stata 11 (29) was used for data processing and analysis.

Overall, 2,628 patients aged 16 and over completed the questionnaires and were included in the analysis.

The most common diagnosis was CLL (1,202 patients [45.7%]), while 893 (34.0%) patients had a diagnosis of CML, 310 (11.8) patients had AML, and 93 (3.5%) patients had ALL ( Table 1 ). A further 130 patients (5.0%) responded “other” for diagnosis type.

The questionnaire was completed by patients from 76 countries. The majority of patients were from the UK (1,312 patients, 49.9%), USA (147, 5.6%), Canada (110, 4.2%), Israel (70, 2.7%), Australia (69, 2.6%), France (53, 2.0%), Ukraine (60, 2.3%) and Germany (52 patients, 2.0%) ( Table 1 ). All other countries had <50 respondents combined (<2.0% of all respondents).

More patients were female (1,417, 56.2%) overall, and across all subtypes. The proportions differed significantly between the subtypes (p<0.001) with e.g., 51% of men having CLL, compared with 41.8% of women.

A total of 2,526 patients reported their age (median 62, IQR [50, 70]). Only 18 (0.73%) were aged under 35. Age differed significantly between the subtypes (p<0.001) with ALL patients on average younger, and CLL older ( Table 1 ).

Date of diagnosis was reported by 2,495 respondents. Disease duration ranged from 0 to 63 years ( Table 1 shows the breakdown by subtypes). Four respondents (all with chronic disease) gave their year of diagnosis as their year of birth, so these, together with the respondent (with CLL) who provided no year of birth but reported a (implausible) duration of 63 years, were excluded from analysis of duration of disease. Disease duration differed significantly (p<0.001) across leukemia subtypes; with medians [IQR]; 5 [1,9] for ALL; 4 [2,8] for AML; 6 [3,11] for CLL; and 6 [3,12] for CML.

With the exclusions applied, overall, there were 2,552 evaluable HM-PRO scores, with a median score of 22.3. Scores differed significantly between leukemia subtypes (p<0.001) Patients with ALL reported the worst outcomes across all four leukemia subtypes ( Figure 1 ). Additionally, women (median=24.6) experienced significantly (p<0.001) higher impact on their QoL than men (median=19.7).

Of the 2,552 respondents, 2,494 reported their age (median=62 years) and 2,432 reported a valid year of diagnosis (median=6 years). Overall, 2,383 patients with an HM-PRO score had reported their age and had a valid year of diagnosis. Both age (ρ= –0.24, p<0.001) and years living with the disease (ρ= –0.14, p<0.001) were significantly negatively correlated (weakly) with HM-PRO scores. Investigation of partial correlation coefficients, between the scores with both age and disease duration, indicated that disease duration was independently associated with the HM-PRO scores, although this association was even weaker when HM-PRO scores were adjusted for age (age: ρ= –0.21, p<0.001), years living with the disease (ρ= –0.08, p<0.001).

The independent correlates of PB SCORE, EB SCORE, SW SCORE, ED SCORE, and PARTA SCORE are detailed in Tables 2 – 6 . In ordinal logistic regression analyses, age and sex were both significantly related to EB SCORE, SW SCORE, and ED SCORE ( Tables 3 - 5 ) but not to PB SCORE ( Table 2 ). Age was coherently associated in an inverse fashion to EB SCORE, SW SCORE, and ED SCORE whereas female sex was related directly to EB SCORE and ED SCORE and inversely to SW SCORE. Overall, patients who achieved a higher level of education had a lower odds ratio of being in a higher category of PB SCORE, EB SCORE, and ED SCORE when compared to those with no formal qualification ( Tables 2 , 3 , 5 ). No association was found between scholarity and SW SCORE ( Table 4 ). Among disease types, patients affected by CLL had a lower odds ratio of being in a higher category of PB SCORE, EB SCORE, SW SCORE, and ED SCORE ( Tables 2 - 4 ) when compared to those with acute lymphoblastic leukemia and this was also true for patients with chronic myeloid leukemia although this latter did not attain the statistical significance for the ED SCORE ( Table 5 ). The duration of disease was significantly and inversely related to EB SCORE ( Table 3 ) and ED SCORE ( Table 5 ) but not to PB SCORE ( Table 2 ) and SW SCORE ( Table 4 ). Never treated patients and patients on ongoing treatment had higher odds ratios of being in a higher category of EB SCORE ( Table 3 ) and ED SCORE ( Table 5 ), respectively, when compared to those previously treated. The associations between age, sex, level of education, disease type, disease duration, and treatment with PARTA SCORE are reported in Table 6 (see Methods-Statistical Analysis for more details).

In an additional sub analysis, we evaluated whether the acute or chronic leukemia subtypes and type of treatment could have an impact on HM-PRO domain scores. When considering patients with acute leukemias compared to patients with chronic leukemias, we observed a difference in all HM-PRO domains which reached the MCID, with the exception of EB-SCORE ( Table 7 ). We next examined whether different treatment types could impact upon patient QoL. No clinically significant differences were observed across all HM-PRO domain scores when patients were stratified by oral vs intravenous treatment or in those receiving Bruton’s tyrosine kinase (BTK) inhibitors vs other treatment (only in patients affected by CLL). However, in patients who underwent allogenic transplantation, we observed marked statistically significant (and clinically important) higher scores (worse QoL) in all HM-PRO domains compared to patients who did not undergo allogenic transplantation (with the exception of EB SCORE) ( Table 8 ). Furthermore, when categorizing patients according to time since allogenic transplantation, patients <1 year after allogenic transplant were observed to have higher domain scores (i.e. lower QoL) compared to 1-5 years and >5 years after transplant, with higher scores seen across SW, ED and PART A SCORE domains ( Figure 2 ). This was not observed when comparing QoL scores in acute leukemia patients who received and did not receive allogenic transplantation.