Using modern causal inference methods to enhance the generalisability of randomised controlled trial evidence by leveraging observational data: the case of Deferasirox® for managing iron overload in transfusion dependent lower-risk myelodysplastic syndrome (LR-MDS) patients

Using modern causal inference methods to enhance the generalisability of randomised controlled trial evidence by leveraging observational data: the case of Deferasirox® for managing iron overload in transfusion dependent lower-risk myelodysplastic syndrome (LR-MDS) patients

Myelodysplastic syndromes (MDS) are a group of malignant blood disorders in which the bone marrow fails to produce healthy blood cells. About two-thirds of MDS patients present with lower-risk MDS (LR-MDS) which is associated with impaired health-related quality of life, risk of transformation to acute leukaemia, and reduced survival compared to the general population. Despite several new treatment options for patients with LR-MDS, red blood cell transfusions (RBCT) remain the cornerstone of supportive care in this patient group. Those who receive many RBCT are at risk of iron accumulation in their body (i.e. iron overload), which can lead to heart and liver (at times fatal) complications. Iron chelation therapy mitigates the negative impact of iron overload. Several small observational studies have reported beneficial effects of iron chelation therapy (ICT) on survival and other outcomes.Well-designed randomised controlled trials (RCTs) are the preferred (and main) source of evidence to inform clinical and policy decision making, as they are considered to provide the most reliable estimate of the effect of an intervention. RCTs do not always yield information about the effects of the intervention on a broader target population, which is often the population of interest in routine clinical practice. Observational studies may recruit larger and more representative samples of the target population. The concern with observational studies though is that - while being potentially more generalisable (i.e. high external validity) to the target population - generated evidence may carry the risk of being confounded (i.e. low internal validity).Recently, the European MDS (EUMDS) Registry published a large, well-designed observational study on people representative of the general LR-MDS population and found that those receiving ICT survived longer than those not treated with ICT. A recent RCT (the TELESTO trial) reported better event-free survival in people with LR-MDS who received ICT but their sample included individuals that were healthier and younger than those commonly seen in routine clinical practice. Furthermore, given the short follow-up period of the TELESTO trial and its relatively small sample size, little information could be produced about long-term survival or long-term health-related quality of life in this sample.Clinicians and payers wanting to use evidence from these two studies to make treatment and funding decisions for transfusion-dependent LR-MDS patients face a great deal of uncertainty. Should they assume the results of TELESTO are generalisable to the broader LR-MDS population seen in routine clinical practice? Should they assume that 'event-free survival' results will ultimately translate into longer 'long-term overall' survival for this population? Should clinicians and payers trust instead the evidence derived from the EUMDS analysis, which was conducted in a population of LR-MDS individuals arguably more representative of those treated in routine clinical practice? But what about the potential risk of confounding in its estimate of overall survival benefits for ICT, being an observational study? Our study combines experimental (i.e. TELESTO) and real-world evidence from the EUMDS registry to estimate the treatment effect of ICT in transfusion-dependent LR-MDS patients in terms of differential (long-term) overall survival and quality-adjusted life years. We will show how to use these estimates as inputs in an economic model designed to assess the value for money of health technologies where the evidence base includes both experimental and observational data. We will use modern causal inference methods to integrate the two data sources in our analysis to leverage their internal and external validity, respectively. This methodological work will contribute to expanding the set of statistical and health economics tools that clinicians and funders can use to maximise the value of existing evidence.

[{ "PostingID": 20601, "Title": "NOVARTIS-CICL670A2302", "Description": "A multi-center, randomized, double-blind, placebo-controlled clinical trial of deferasirox in patients with myelodysplastic syndromes (low/int-1 risk) and transfusional iron overload" }]