Searching for clarity in therapy-related myelodysplastic syndrome/acute myeloid leukemia prognostication

Therapy-related myelodysplastic syndrome (t-MDS) and therapy-related acute myeloid leukemia (t-AML), collectively defined as therapy-related myeloid neoplasms by the 2008 World Health Organization (WHO) “classification of tumours of haematopoietic and lymphoid tissues” [1], are almost always associated with a poor prognosis relative to de novo disease. Intuitively, this is not unexpected, since therapy-related disease is derived from aberrant myeloid cells which have survived and proliferated despite prior cytotoxic exposure. However, other than a broad designation of “poor outcome,” accurate prognostication in t-MDS/t-AML remains elusive, and the contribution of disease- and patient-specific characteristics would benefit from further clarification.

In an attempt to address the problem of prognostication in therapy-related myeloid disease, in this issue of Leukaemia and Lymphoma Bacher and colleagues [2] report data from a series of 508 patients with t-MDS/ t-AML, including 277 with outcome data. Within the study, patients with t-MDS (defined as < 20% blasts detected in the bone marrow at diagnosis) had significantly better overall survival than patients with t-AML (≥ 20% blasts at diagnosis) (43.8 vs. 18.2 months). Singh and colleagues [3] also previously reported a significantly better overall survival for patients with t-MDS, although the difference between patient groups was relatively modest by comparison (8.5 months vs. 6.5 months for t-MDS and t-AML, respectively). Taken together, these data suggest that stratification by blast count, based on the 2008 WHO definitions of t-MDS and t-AML, provides prognostically relevant information, and that co-classification of these neoplasms as a single entity, although perhaps relevant from an etiological perspective, may be too simplistic for the purposes of prognostication. Consistent with this notion, Bacher and colleagues [2] also report that blast count was significantly prognostic when included as a continuous variable in survival analysis, suggesting prognostic value of this marker within, as well as between, the t-MDS and t-AML subgroups. If confirmed, then morphological subclassification, which is primarily based on blast count in the marrow and the extent of morphological dysplasia, could also have prognostic relevance in t-MDS, as it does in de novo MDS [4].

Cytogenetics is also highly prognostic in de novo MDS and AML. Numerous studies have documented the higher frequency of adverse karyotypic alterations in both t-MDS and t-AML relative to de novo disease [3,5–7], and also that cytogenetics retain prognostic significance in t-MDS and t-AML [3,7]. The data reported by Bacher and colleagues [2] support these conclusions, and further suggest that somatic point mutations prognostic in de novo disease are similarly prognostic in therapy-related disease, although the relative frequencies might differ. A higher frequency of adverse cytogenetics undoubtedly contributes to poor outcome in the t-MDS/t-AML patient group as a population; however, therapy-related disease still has a worse outcome even within the same cytogenetic group [8–10]. This observation points toward the existence of unidentified mechanisms independent of cytogenetics that also contribute to poor survival, and perhaps justifies the development of prognostic indices specifically tailored to therapy-induced disease which incorporate these mechanisms as they are elucidated.

It is not surprising that disease-specific features prognostic in de novo disease, such as blast count and cytogenetics, are also prognostic in therapy-related disease. Unfortunately, numerous factors have conspired to limit progress in t-MDS/t-AML prognostication. The relative rarity of t-MDS/t-AML has historically necessitated the accrual of cases diagnosed over extended time periods (often several decades) to establish even modestly sized cohorts. As such, progressive improvements in therapy and supportive care have made it difficult to accurately evaluate the prognostic value of disease-specific features, such as blast count. In contrast, the patient cohort in the study of Bacher et al. [2] benefits from a relatively short recruitment period (2005–2011) made via a single diagnostic laboratory. Another barrier to accurate prognostication is considerable patient heterogeneity (including site of first cancer, age at diagnosis of first cancer, type of therapy for first cancer, age at diagnosis of therapy-related myeloid malignancy and co-morbidities), requiring the development of large case-series in order to support sufficiently powered analyses. Furthermore, “contamination” of therapy-related studies with patients whose myeloid disease is in fact a second de novo malignancy and unrelated to previous therapy also weakens the ability to accurately define the prognostic value of disease-specific markers. As such, the need for large studies with carefully defined inclusion criteria is particularly apparent if prognostication is to be clarified for therapy-related disease.

What remains clear, unfortunately, is that t-MDS/t-AML has a generally dismal outcome, with stem cell transplant currently representing the best chance of long-term survival [11]. Regardless of the means by which it is achieved, accurate prognostication in t-MDS and t-AML will provide a solid foundation upon which to better plan and determine response to therapy, and is urgently needed.

Potential conflict of interest:

Disclosure forms provided by the authors are available with the full text of this article at www.informahealthcare.com/lal.