Management of Myelodysplastic Syndromes in High-Risk Patients - Episode 4

High-Risk MDS: Prognostic and Predictive Factors


Mikkael A. Sekeres, MD, explores prognostic and predictive factors inherent to high-risk MDS, including age and molecular abnormalities.

Yazan Madanat, MD: This brings us to talking about prognostic and predictive factors that you think might be relevant for our patients with higher-risk disease.

Mikkael A. Sekeres, MD: It almost seems silly to say, but this is one of my favorite questions I ask residents and fellows: What’s the greatest risk factor for survival with MDS [myelodysplastic syndromes]? Because it’s so obvious, people usually skip over it, but No. 1 is age. If you have somebody who’s presenting with MDS at the age of 98, regardless of whether that person has MDS, they probably don’t have a long predicted overall survival compared with somebody who’s presenting with MDS at age 55, which is relatively uncommon.

Secondly, we think about how the MDS arose. Therapy-related MDS is often called out as a poor-risk disease, but it’s a complicated issue. First, it’s complicated because how do we define therapy-related MDS? It’s a clinical diagnosis that somebody received something years earlier that could cause DNA damage, and now that person has MDS. Whether that agent they received, be it chemotherapy or radiation therapy, actually led to the MDS is not known biologically. What’s known is they had something in the past—let’s say it’s lymphoma or breast cancer—received treatment for that with radiation therapy or chemotherapy, and then developed MDS 7 years later.

The reason this has become so complicated is because we don’t know if that person actually had a CHIP [clonal hematopoiesis of indeterminate potential] abnormality prior to the diagnosis of that solid tumor or lymphoma, and then received chemotherapy and radiation therapy, and if that CHIP abnormality would have evolved naturally, or if it was nudged along by the radiation therapy and chemotherapy that person subsequently received. Then we get into the issue of adjusting for the poor-risk features that are sometimes associated with therapy-related disease. If somebody has a TP53 abnormality or chromosome 7 abnormality, regardless of whether that’s therapy-related MDS or de novo MDS, that person has a poor prognosis.

When we do these adjustments, it turns out that based on a study published out of Massachusetts General Hospital about a decade ago, folks who were exposed to radiation therapy for previous cancers, when adjusted for risk factors for MDS, didn’t have much worse of an outcome than patients with de novo MDS who had some of those same karyotypic abnormalities and other risk factors. Still, how I try to summarize therapy-related disease, because it’s complicated, is that if someone has what we think is true therapy-related disease, where previous chemotherapy or radiation therapy either started the clock, moving a genetic landslide that leads to MDS, or advanced a clock that had already started toward MDS, they tend to have worse genetic markers and therefore they tend to have worse outcome.

What this leads to is the third major risk factor that you alluded to, Yazan, the molecular makeup of folks who have MDS. We’re getting much more sophisticated about this. We know that there are certain abnormalities like TP53, ASXL1, EZH2, RAS mutations that portend a worse prognosis. How we incorporate these into the IPSS [International Prognostic Scoring System] and revised IPSS is also very complicated. Aziz Nazha, MD, whom you mentioned before, published a paper incorporating molecular abnormalities into the IPSS and showed that it provides additional accuracy over the clinical parameters alone. He also has a paper coming out looking at machine learning and artificial intelligence, where it takes it up a notch in looking at genetic and clinical factors. The machine learning approach to looking at prognosis is both more accurate than the IPSS or revised IPSS, even with molecular abnormalities incorporated, and it’s dynamic, meaning it can be used at day 0 of a person’s diagnosis, or 1, 2, or 3 years later, and maintain its accuracy. Also, it seems to be agnostic to therapy. That will help us risk stratify a lot better, particularly incorporating molecular abnormalities.

Transcript Edited for Clarity