Molecular Data Helps Better Predict Survival for Patients with Myelodysplastic Syndrome

Overall survival rates for patients with myelodysplastic syndrome (MDS) range from months to decades. Although several prognostic scoring systems have been developed to risk-stratify MDS patients, survival varies even within distinct categories, which may lead to over- or under-treatment.

Researchers at Cleveland Clinic recently hypothesized that these discrepancies may be due to analytic approaches or lack of incorporation of molecular data. To test their theory, they built a personalized predication model that incorporates individual patient genomic and clinical data and uses a machine-learning algorithm

“Determining prognosis in oncology is one of the most important parts of our job. All oncology patients want to know how long they are going to live,” says Aziz Nazha, MD, Hematology and Medical Oncology, Cleveland Clinic. “Too often we find a significant gap between what we predict, based on existing prediction models and what actually happened to our patients.”

The model that Dr. Nazha and colleagues created outperformed current prognostic scoring systems, such as the Revised International Prognostic Scoring System (IPSS-R), in predicting an MDS patient’s risk of mortality and transformation to acute myeloid leukemia (AML), a more aggressive type of bone marrow cancer.

Dr. Nazha presented the results of the model at the 2018 American Society of Hematology Annual Meeting (ASH). Also at the meeting, he also presented on a different model, one that predicts MDS patient outcomes after allogeneic hematopoietic stem cell (HCT) transplant.

Next step = a prediction website

To test the prediction model, the investigators used patient data from Cleveland Clinic and Munich Leukemia Laboratory. Patients’ clinical and mutational variables were entered into a web application that runs the personalized prediction model and provides overall survival and AML transformation probabilities at different time points that are specific for a patient. The researchers validated results by using a separate collection of patient data from Moffitt Cancer Center.

In comparison with using medical records, the new model predicted a patient’s likelihood of surviving for a given period of time 74 percent of the time, compared to IPSS-R accuracy of 67 percent. The model accurately predicted a patient’s risk of AML 80 percent of the time, compared to IPSS-R accuracy of 73 percent.

Dr. Nazha’s next step is to build a website where clinicians can input a patient’s clinical and genetic characteristics and get back the patient’s probability of surviving at different time points such as six, 12 and 18 months.

“Understanding a patient’s prognosis allows us to more appropriately develop a treatment plan and counsel patients,” Dr. Nazha says. “We are optimistic an improved prediction will lead to more personalized care.”

Who benefits from HCT?

For the HCT outcomes model, Dr. Nazha and colleagues used data from MDS patients enrolled in the Center for International Blood and Marrow Transplant Research Registry and built a web application where patients’ genomic and clinical data are computed to predict survival probability after HCT at 6, 12 and 24 months.

HCT remains the only potentially curative option for MDS. Previous research has shown genetic alterations have an impact on outcomes after allogeneic HCT in patients with MDS.

The model created by Dr. Nazha and colleagues identified several clinical and molecular variables that impacted overall survival and the hazard of relapse.

“Because of the risks of transplant-related mortality and relapse, identifying patients who may or may not benefit from HCT is clinically important,” says Dr. Nazha. “Understanding survival probability at different points in time may aid physicians and patients in their approach, prior to HCT.”