Real‐world complication burden and disease management paradigms in transfusion‐related β‐thalassaemia in Greece: Results from ULYSSES, an epidemiological, multicentre, retrospective cross‐sectional study

Abstract Patients with transfusion‐dependent beta (β)‐thalassaemia experience a broad range of complications. ULYSSES, an epidemiological, multicentre, retrospective cross‐sectional study, aimed to assess the prevalence and severity of treatment and disease complications, capture disease management and identify predictors of complications in patients with transfusion‐dependent β‐thalassaemia, treated in routine settings in Greece. Eligible patients were adults diagnosed with β‐thalassaemia ≥12 months before enrolment and having received ≥6 red blood cell (RBC) units (excluding elective surgery) with no transfusion‐free period ≥35 days in the 24 weeks before enrolment. Primary data were collected at a single visit and through chart review. Between Oct 21, 2019, and Jun 15, 2020, 201 eligible patients [median (interquartile range, IQR) age 45.7 (40.2–50.5) years; 75.6% > 40 years old; 64.2% female] were enrolled, a mean (standard deviation) of 42.9 (7.8) years after diagnosis. Median (IQR) age at diagnosis and RBC transfusion initiation were 0.8 (0.4–2.8) and 1.3 (1.0–5.0) years, respectively. From diagnosis to enrolment, patients had developed a median of six (range: 1–55) complications; 19.6% were grade ≥3. The most represented complications were endocrine/metabolic/nutrition disorders (91.5%), surgical/medical procedures (67.7%) and blood/lymphatic system disorders (64.7%). Real‐world data generated by ULYSSES underscore the substantial complication burden of transfusion‐dependent β‐thalassaemia patients, routinely managed in Greece.


K E Y W O R D S
complications, epidemiological, real-world, transfusion-related β-thalassaemia

INTRODUCTION
Beta (β)-thalassaemia is a hereditary blood disorder characterised by a deficiency of functional β-globin chains [1][2][3][4]. Based on severity, β-thalassaemia has been traditionally classified as major, intermedia or minor, although a newer approach distinguishes patients into transfusion-dependent, requiring lifelong regular blood transfusion for survival, and non-transfusion-dependent [5,6]. The prevalence of βthalassaemia is high in Mediterranean countries and is estimated at about 25 per 100,000 population in Greece [7].
The pathophysiology of β-thalassaemia is largely explained by the accumulation of excessive α-globin chains due to their imbalance over β chains, which impairs the maturation and viability of the mature erythroid cells and their precursors, resulting in anaemia and related complications [2]. Management of β-thalassaemia mainly aims at alleviating these states, with conventional therapies including red blood cell (RBC) transfusions, iron chelation therapy (ICT) and splenectomy, though the latter is gradually becoming obsolete [8,9]. Such interventions have contributed to the substantial prolongation of patient life expectancy [10]. However, treatment approaches can also result in complications. The increased cumulative exposure to various treatment options that results from lifespan expansion of the thalassaemic patient, in combination with comorbidities related to advanced age, lead to an evolving spectrum of complications, with substantial impact on the affected individual and a burden to the healthcare system [11,12].
The present study aimed to generate data on the patterns and determinants of disease-and treatment-related complications in a representative sample of patients with transfusion-dependent β-thalassaemia treated in a routine care setting in Greece.

Study objectives and outcomes
The primary aim of the study was to determine the prevalence and severity of disease-and treatment-related complications in a represen- Study data were collected via web-based electronic data capture (use of electronic case report form) fulfilling all data privacy rules.

Statistical approach
The sample size calculation was based on the study's primary endpoint. A sample of 200 patients was considered adequate, as it offered a maximum margin of error below 7.0% for the maximum indetermination of any complication frequency in the overall population (i.e., the estimation of a frequency of 50%, where the margin of error is largest). This represents a scientifically acceptable level of precision, and was thus sufficient for the primary endpoint, without the requirement to include the entire Greek registry of these patients. In order to control for and minimise selection bias, physicians were requested to consecutively enrol the first patients attending their clinic that met the eligibility criteria. Information bias, due to the involvement of several laboratories rather than a central laboratory, was minimised with the implementation of source data verification and quality assurance measures.
The normality of the distribution of continuous variables was examined using the Shapiro-Wilk test, with data presented as mean (standard deviation, SD) when following a normal distribution and median (interquartile range, IQR) when not. No imputation for missing data was implemented, except for partial dates. In the context of the primary objective, the number and proportion of patients with at least one disease-and/or treatment-related complication were calculated along with the relevant 95% Wald confidence intervals (CIs). The hospitalisation rate per patient-year along with 95% CI during the 48-week period prior to enrolment was estimated using a Poisson regression model.
The association of the presence of disease-and/or treatment-related complications was examined using univariable and multivariable logistic regression analysis. The multivariate logistic regression model was derived using a stepwise procedure based on the minimisation of Akaike's information criterion, while, due to separability issues, Firth's penalised likelihood approach of logistic regression was used. All statistical tests were two-sided and at a 0.05 significance level. Statistical analysis and sample size determination were performed using SAS v9.4 (SAS Institute, Cary, NC).

Patient disposition and characteristics
A total of 208 patients were enrolled in the study; seven patients were excluded from the analysis, since they did not fulfil all eligibility criteria, resulting in 201 patients ultimately analysed.

Β-thalassaemia disease characteristics
Patients' median (IQR) age at β-thalassaemia diagnosis was 0.8 (0.4-2.8) years, with a mean (SD) of 42.9 (7.8) years elapsed from diagnosis to enrolment. The β-thalassaemia clinical phenotype, assessed by the physicians was characterized as major in 170 (84.6%) patients and intermedia in the remaining 31 (15.4%) (see Table SII). All patients referred to as thalassemia intermedia were transfusion independent at the time of the study but became transfusion dependent in older age.
In addition to the β-globin gene mutation, 18 (9.0%) patients also carried a mutation, deletion and/or multiplication in one or two α-globin genes, among which two patients carried a triplication of the α-globin gene.   A family history of β-thalassaemia major or intermedia in first-or second-degree relatives was reported by 43 (21.4%) patients.

Current and past β-thalassaemia management patterns
According to the study inclusion criteria, all patients were transfusion dependent, with the median (IQR) age at receipt of the first transfusion

Β-thalassaemia disease-and treatment-related complications
At enrolment, medical records were available for a mean percentage of 93.2% of the time since diagnosis, with the full disease-and treatmentrelated medical history available for 158 (78.6%) patients and partial relevant information for the remaining 43 (21.4%) patients.
During the period between diagnosis and enrolment, at least one complication related to β-thalassaemia itself was recorded for all patients, while at least one complication related to the treatment of β-thalassaemia was recorded for 53.2% (n = 107, 95% CI: 46.3-60.1) (see Table SIV). A total of 1443 disease-and/or treatment-related complications were recorded (median 6; range: 1-55 per patient) over this period (Figure 2A). The median number of disease-related complications per patient was five (range: , and that of treatment-related complications was one (range: 0-35).  The mean (SD) length of stay in the hospital was 5.6 (4.2) days.

Association of patient and disease characteristics with β-thalassaemia disease-and treatment-related complications at enrolment
The potential association of selected patients and disease characteristics with the presence of the most prevalent treatment-and diseaserelated complications at enrolment was investigated by univariable analysis (Table SVI)

DISCUSSION
ULYSSES provides epidemiological data on the profile and complication burden of transfusion-dependent β-thalassaemia patients, managed in routine care in Greece. Over the 48-week pre-enrolment period, the average pre-transfusion haemoglobin was 9.9 g/dL, within recommended 9.5-10.5 g/L target [13]; average serum ferritin in ULYSSES was 549.3 μg/L, median liver iron concentration (LIC) was 2.2 mg Fe/g dry weight (dw) and median elevated transverse relaxation time (T2*), which serves as a measure to assess myocardial iron deposition and the risk for cardiac complications [17], was 36.8 ms, with 92.9% of evaluable patients having normal myocardial iron levels (> 20 ms).
The most frequently reported surgical procedure in ULYSSES was splenectomy (53.2% of patients). Splenectomy rates in other cohorts range from 40% to 70%, with higher prevalence among older patients [14,22,38,39], consistent with our data. Cardiac/vascular disorders are among the most important life-threatening complications in β-thalassaemia [40] and were reported for 29% of the patients in ULYSSES. Furthermore, 9% of patients experienced heart failure, consistent with the 9% pooled prevalence estimated in a meta-analysis of data from 10,968 patients, with transfusion-dependent β-thalassaemia major from multiple continents [41].
A major concern for transfusion-dependent patients is the risk of infections, with hepatitis C being the most common. In ULYSSES, hepatitis C infection was recorded for 20.9% of the overall population.
Differences between age groups (25.7% in > 40 years versus 6.1% in 18-40 years) are presumably attributed to the improvement of blood screening protocols. These rates are within the 5.9%-28.3% range, reported in a systematic review of studies in transfusiondependent thalassaemia with enrolment in 2005 or later [22,38,39,42,43]. Three patients in the present study were diagnosed with hepatocellular carcinoma, a prevalence within the previously reported range of 0.75%-3.5% among patients with thalassaemia major [22,44,45].
Although not statistically compared, the proportion of patients with treatment-related complications was higher in the > 40 age group, than in the 18-40 age group. This was expected, since advancing age, as well as the cumulative effect of longer exposure to treatments, may contribute to the complication burden. Multivariable analysis showed that disease unrelated medical history/comorbidities, prior splenectomy and a higher number of transfusions were associated with a higher odds ratio of having experienced a treatment-related complication at enrolment. In contrast, in univariable analysis, splenectomy was associated with a lower odds ratio of having a hepatobiliary disorder, which may also partially explain why complications from this system were more common among younger patients. were not included, but as they are a minority, it is highly unlikely that their data were significantly different from the ones reported in the study.
The results of the present real-world study underscore a considerable complication burden of transfusion-dependent β-thalassaemia and the need for effective monitoring and improved therapies for these patients.

ACKNOWLEDGMENTS
The authors would like to thank the patients and families who participated in this study. The authors wish to thank Qualitis Ltd. for medical writing support, which was funded by Bristol Myers Squibb.

DATA AVAILABILITY STATEMENT
The Bristol Myers Squibb policy on data sharing may be found at: https://www.bms.com/researchers-and-partners/independentresearch/data-sharing-request-process.html.