Personalized Prediction Model for Outcomes After Allogeneic HSCT for Myelodysplastic Syndromes

Researchers developed a personalized prediction model, based on genes commonly mutated in myeloid malignancies, to predict survival and relapse probabilities for patients with myelodysplastic syndromes (MDS) after allogeneic hematopoietic stem cell transplantation (allo-HSCT), according to study results published in Biology of Blood and Marrow Transplantation.

Although HSCT is the only potentially curative treatment for MDS, the mortality rate following HSCT is high among this patient population; therefore, identifying patients who may or may not benefit from HSCT is an important clinical need.

The investigators identified patients with MDS who were enrolled in the Center for International Blood and Marrow Transplant Research Registry, and sequenced DNA from their peripheral blood samples to assess the genotypes of 129 genes that are commonly mutated in myeloid malignancies. They then built a predictive model using machine learning algorithms and tested its accuracy using concordance (C-) indices.

A total of 1514 patients were included in the study; the median age was 59 years (range, 0.4-77 years). The majority of patients (79%) had 1 or more mutations, with a median of 2 mutations per sample (range, 0-15) prior to transplantation. The most commonly mutated genes were ASXL1 (20%), TP53 (19%), DNMT3A (15%), and TET2 (12%). In addition to the gene mutations, a number of patient characteristics and baseline parameters were tested for inclusion in the model. The median follow-up duration for survivors was 47 months.

The machine learning algorithm identified the following variables prior to HSCT that affected overall survival (OS) in order of importance: age, TP53 mutations, white blood cell count, donor group, Karnofsky performance status, International Prognostic Scoring System-Revised (IPSS-R), diagnosis of therapy-related MDS, cytogenetics (per IPSS-R), donor age, absolute neutrophils counts, conditioning regimen, peripheral blast percentage, year of HSCT, monosomal karyotype, hemoglobin, platelets, bone marrow blast percentage, CUX1 mutations, graph-vs-host prophylaxis, PPM1D mutations, number of mutations/sample, and graft type.

Similarly, the machine learning algorithm identified the following variables prior to HSCT that affected the posttransplant risk of relapse in the order of importance: TP53 mutations, conditioning regimen, disease status, hemoglobin, IPSS-R, cytogenetics, donor age, absolute neutrophils counts, age, mutation of genes in the RAS pathway, white blood cell count, peripheral blast percentage, bone marrow blast percentage, diagnosis of therapy-related MDS, number of mutations/sample, in vivo T-cell depletion, CUX1 mutations, PPM1D mutations, TET2 mutations, platelets, and year of HSCT.

The C-index for the personalized prediction model was 0.62 for OS and 0.68 for time to relapse compared with 0.55 and 0.58, respectively, for the IPSS-R MDS model and 0.57 and 0.61, respectively, for the Center for International Blood and Marrow Transplant Research MDS model.

The authors estimated that a sample size of 5000 is needed to improve the C-indices, and that independent validation of the model is needed.

“The new model can provide survival probability at different time points that are specific (personalized) for a given patient based on the clinical and mutational variables that are listed above,” wrote the authors. “The outcomes probability at different time points may aid physicians and patients in their decision regarding HCT.”

Disclosures: Some authors have declared affiliations with or received funding from the pharmaceutical industry. Please refer to the original study for a full list of disclosures.

Reference

Nazha A, Hu Z-H, Wang T, et al. A personalized prediction model for outcomes after allogeneic hematopoietic cell transplant in myelodysplastic syndromes patients. Biol Blood Marrow Transplant. Published online August 8, 2020. doi:10.1016/j.bbmt.2020.08.003