In a study published in the Journal of Clinical Oncology, researchers reported that a newly developed personalized prediction model comprising clinical and genomic data performed better than established predictive models in patients with myelodysplastic syndromes (MDS). The model predicted both survival and leukemia transformation possibilities unique to each patient, suggesting that it could “be used as a stand-alone model or in conjunction with the IPSS/IPSS-R scoring systems to improve their accuracy,” the researchers, led by Aziz Nazha, MD, from Cleveland Clinic, wrote.
The study included a training cohort that consisted of 1,471 patients with MDS who were assessed at Cleveland Clinic and the Munich Leukemia Laboratory in Germany between 2001 and 2018. Dr. Nazha and colleagues used the training cohort to build the model with a random survival algorithm. The model was later validated in an independent group of patients with MDS who were treated at Moffitt Cancer Center in Tampa, Florida, between 2005 and 2018. Researchers performed targeted deep sequencing on 38 genes commonly reported in genomic panels that have shown clinical significance in MDS.
In the patients with MDS who were included in the training cohort, the median age was 71 years (range = 19-99 years). Based on the revised International Prognostic Scoring System (IPSS-R) scores, patients had the following cytogenetic risk:
- very good: 4% (n=65)
- good: 72% (n=1,060)
- intermediate: 13% (n=193)
- poor: 4% (n=60)
- very poor: 6% (n=93)
In addition, risk stratification per the IPSS-R revealed that 51% of patients (n=749) had very low or low risk MDS, 23% had intermediate risk (n=336), and 19% had high or very high risk (n=182). Risk was not evaluable for 7% of patients.
The commonly mutated genes in the overall training cohort were SF3B1, TET2, and ASXL1. Factors identified as having an effect on overall and leukemia-free survival in the algorithm included chromosomal karyotype, platelet, hemoglobin levels, bone marrow blast percentage, age, other clinical variables, and mutation number.
During a median follow-up of 43.6 months, the median overall survival (OS) was 32.2 months (range = 0.03-221.8). A total of seven genes were associated with OS in an analysis adjusted for age and IPSS-R score.
A multivariate analysis that included age, IPSS-R risk categories, and significant mutations found that the predictive value of some mutations varied depending on the variables included in the prognostic model. For instance, mutations in ASXL1, EZH2, KRAS, NRAS, RAD21, SF3B1, and TP53 were identified as significant in the univariate analysis were included in the multivariate analysis. In contrast, mutations in ASXl1, CBL, EZH2, NRAS, RA21, SF3B1, TET2, and TP53 were considered significant when the researchers added 24 genes that were mutated in 30 or more patients.
IPSS-R cytogenetic risk groups were considered the most important variable for survival, while blast percentage was considered the most important variable predicting transformation to acute myeloid leukemia, the authors reported.
The researchers also noted that the model was “dynamic” and allowed them to upstage and downstage patients into more appropriate risk categories. For example, patients who had a survival probability of less than 50% at 18 months were considered “higher-risk” after application of the model. Additionally, the new model identified 306 patients (20%) in the training group who were also higher-risk, including 148 patients who were considered lower-risk per IPSS and 107 who were considered lower risk per IPSS-R. Seventy-three patients (16%) in the validation cohort were classified as higher-risk with the new model, including 28 and 33 who were classified as lower risk per IPSS and IPSS-R, respectively.
The investigators explained that while the model included age as an important prognostic factor, younger patients with MDS often experience greater loss of life years compared with older patients. “Thus, including age in prognostic scores may lead to less intensive treatment in younger patients since they have a better OS,” the researchers wrote.
The authors report no relevant conflicts of interest.
Reference
Nazha A, Komrokji R, Meggendorfer M, et al. Personalized prediction model to risk stratify patients with myelodysplastic syndromes [published online ahead of print, 2021 Aug 18]. J Clin Oncol. doi: 10.1200/JCO.20.02810.
