Photo Credit: Undefined

Using NGS in prognostic assessments enhances myelofibrosis prognosis and survival estimates, representing a step toward personalized medicine in myelofibrosis.

Integrating next generation sequencing (NGS) data into prognostic assessments, where available, shows promise in improving clinical decision-making for patients with myelofibrosis (MF), according to research published in HemaSphere.

MF is a chronic myeloproliferative neoplasm characterized by the aberrant activation of the JAK-STAT pathway, often caused by mutations in the JAK2, CALR, and MPL genes. These genetic abnormalities and additional mutations affecting epigenetic regulators and spliceosome components significantly shape the disease’s clinical features. MF can manifest as primary (PMF) or secondary (SMF) after the progression of polycythemia vera or essential thrombocythemia.

Although the median overall survival (OS) for patients with MF is around 6 years, the clinical course varies widely, making accurate prognostic assessment crucial for determining the most appropriate treatment, including the potential need for allogeneic hematopoietic cell transplantation, the only curative option.

Traditional prognostic models for MF have been instrumental in stratifying patients into risk categories, but they have limitations. Many models are specific to particular MF subtypes, require karyotypic analysis, or rely on NGS data, which may not be widely accessible.

To address these challenges, a study comprising 1617 patients with MF across 60 Spanish institutions led to the development of the AIPSS-MF (Artificial Intelligence Prognostic Scoring System for Myelofibrosis), a machine learning-based model that uses eight clinical variables. This model has demonstrated superior predictive accuracy for OS and leukemia-free survival (LFS) compared to established prognostic tools like the International Prognostic Scoring System (IPSS) for primary MF and Myelofibrosis Secondary to PV and ET-Prognostic Model. Notably, AIPSS-MF is based on clinical data, making it suitable for use in diverse healthcare settings.

However, the study’s initial limitation was the lack of molecular data on additional somatic mutations, which could further enhance the model’s prognostic accuracy.

Surveying the Genomic Landscape

Juan C. Hernández-Boluda, MD, PhD, and colleagues conducted a subsequent study involving 581 patients with MF who also had available NGS data from the GEMFIN database. DNA samples were analyzed using targeted NGS, focusing on 20 genes consistently covered across different NGS panels. The researchers employed random survival forest models to predict OS and LFS, emphasizing the variant allele frequency (VAF) of mutations, which provides a more detailed view of the genomic landscape.

The study found that a machine learning model based on VAF outperformed models that considered only the presence/absence of mutations or the total mutation count per gene. This VAF-based model, named the NGS model for overall survival, highlighted the importance of specific genes such as TP53, SRSF2, and EZH2 in predicting patient outcomes. For LFS, a similar model focusing on 20 genes, including TP53, EZH2, and IDH1, demonstrated strong predictive accuracy. Incorporating molecular data from specific mutations in the CALR gene and the U2AF1 Q157 mutation further improved the model’s performance.

When the NGS model was combined with the clinical AIPSS-MF score, the resulting AIPSSmol-MFSurv model showed modest improvements in predictive accuracy for OS. This combined approach proved superior to traditional scores like IPSS and Mutation and Karyotype-Enhanced International Prognostic Scoring System for Primary Myelofibrosis in adults 70 and younger, particularly in younger patients and those with PMF.

Similarly, integrating the NGS model with the AIPSS-MF score for LFS prediction (AIPSSmol-MFLeuk) enhanced the model’s accuracy, especially in predicting outcomes for patients with myelodepletive MF.

Moving Closer to Personalized Medicine

The researchers’ findings underscore the significant prognostic role of mutations in TP53, spliceosome components, and the RAS pathway while questioning the relevance of ASXL1 mutations, aligning with recent research. The AIPSS-MF model consistently outperformed traditional scores in predicting OS, and the integration of molecular data further refined LFS predictions, highlighting the potential of NGS data in personalized MF management.

To facilitate clinical application, the researchers developed an online calculator that integrates these findings, offering a user-friendly tool for personalized prognostic assessment.

This represents a significant step toward personalized medicine in MF, enabling more accurate and individualized treatment decisions, particularly regarding the timing of transplantation in younger patients. This innovative approach holds promise for improving outcomes and tailoring treatment strategies to the specific needs of patients with MF.

Key Takeaways

• Traditional prognostic models for myelofibrosis have been used to stratify patients into risk categories, but they have limitations.
• A next-generation sequencing model for overall survival highlighted the role of specific genes, including TP53, SRSF2, and EZH2, in predicting outcomes.
• The findings show the prognostic role of mutations in TP53, spliceosome components, and the RAS pathway and question the relevance of ASXL1 mutations.
• Together, the results represent a step toward personalized medicine in myelofibrosis.