New Classification for Myeloproliferative Neoplasms

Determining prognoses for patients with myeloproliferative neoplasms can be a challenging task because some patients survive for decades, whereas others live for a few months. An analysis of mutations offers a new genomic classification of myeloproliferative neoplasms and may lead to more accurate prognoses and personalized treatment.

The current classification system for myeloproliferative neoplasms relies on clinical and laboratory findings to sort patients into three main categories. Those with polycythemia vera produce too many red blood cells, whereas patients with essential thrombocythemia have an excess of platelets. In contrast, patients with myelofibrosis, the most severe type of myeloproliferative neoplasm, develop fibrosis in the bone marrow.

Many patients with myeloproliferative neoplasms have a normal life span, but in about 10% to 20% of cases, the disease worsens, evolving into acute myeloid leukemia (AML). Current algorithms for identifying patients whose disease is likely to progress incorporate some genomic information, but their accuracy is limited.

To improve these predictions, Peter Campbell, PhD, of the Wellcome Trust Sanger Institute in the UK, and colleagues analyzed mutations in 2,035 patients with myeloproliferative neoplasms. For 1,887 patients, the researchers sequenced 69 myeloid cancer genes, including JAK2, CALR, and MPL. The scientists performed whole-exome sequencing on samples from the remaining 148 patients.

Using the results, the team divided myeloproliferative neoplasms into eight categories. One group contains the people who have TP53 mutations and the worst overall survival odds. This group constitutes between 1% and 5% of the patients in the traditional categories and includes patients from all three of them, although patients with myelofibrosis are the most common.

At the other end of the survival spectrum is the subtype that lacks any confirmed driver mutations. Of the 192 patients in this category in the study, only two had disease that progressed to AML after a median follow-up of 8 years. Most patients in this group were diagnosed with essential thrombocythemia, but some had myelofibrosis.

The researchers also created a multivariate statistical model that incorporated 63 clinical and genomic variables to forecast disease outcomes. They then tested the model by forecasting prognoses for all study participants. The model's predictions of individual risks of disease progression and death correctly ranked the observed outcomes for pairs of patients 76% to 86% of the time, considerably better than existing prognostic schemes. The models performed equally well on data from an external cohort of 515 patients.

Campbell says that the study shows that “you can predict what will happen to a patient sitting in front of you in the clinic.” Benefits for patients could include refined prognoses and better treatment approaches. For instance, doctors might more accurately predict which patients have benign disease that won't require treatment and which ones have poor survival odds and might be candidates for clinical trials.

“It's a pivotal study,” says Aaron Gerds, MD, of the Cleveland Clinic Taussig Cancer Institute in Ohio, who wasn't connected to the research. “This is a first large attempt looking at genomes first and foremost to try to reclassify the disease,” he says. Before physicians can begin using this approach to make decisions about their patients' treatment, he cautions that researchers need to confirm the results in an independent study—and show that applying the approach in prospective clinical trials leads to increased survival. –Mitch Leslie