-Author name in bold denotes the presenting author
-Asterisk * with author name denotes a Non-ASH member
Clinically Relevant Abstract denotes an abstract that is clinically relevant.

PhD Trainee denotes that this is a recommended PHD Trainee Session.

Ticketed Session denotes that this is a ticketed session.

4558 Validating the MPN Personalized Risk Calculator in 475 Patients with Polycythemia Vera

Program: Oral and Poster Abstracts
Session: 634. Myeloproliferative Syndromes: Clinical and Epidemiological: Poster III
Hematology Disease Topics & Pathways:
Research, Clinical Research, Real-world evidence
Monday, December 9, 2024, 6:00 PM-8:00 PM

Katie Erdos1*, Neville Lee Jr.1*, Yushu Shi, PhD2*, Ghaith Abu-Zeinah, MD1 and Joseph M. Scandura, MD, PhD1

1Richard T. Silver, MD Myeloproliferative Neoplasms Center, Weill Cornell Medicine, New York, NY
2Department of Population Health Sciences, Weill Cornell Medicine, New York, NY

Background: Risk stratification algorithms for polycythemia vera (PV) were developed to predict thrombosis risk using immutable factors of age and thrombotic history. While current PV care standards have greatly reduced thrombotic complications, the long-term risks of progression and mortality persist. Therefore, dynamic prognostic models for PV are needed to better predict outcomes including overall, myelofibrosis-free and leukemia-free survival (OS/MFS/LFS). Grinfeld et. al1 developed the myeloproliferative neoplasms (MPN) Personalized Risk Calculator (MPN-PRC) to stratify risk using genomic profiles and select laboratory and clinical factors. This model has not been further validated. We used data from our richly annotated cohort of patients (pts) with PV to assess the performance of the MPN-PRC in predicting OS/MFS/LFS.

Methods: Model parameters including age, sex, thrombotic history, splenomegaly, white blood cell count (WBC), platelet count (PLT), hemoglobin (HGB), karyotype and genomic data were collected from our MPN research data repository, as previously described2. Pt data was uploaded to the online risk calculator3, and the outcome predictions were compiled. We computed truncated c-indexes with confidence intervals (CI) based on 1000 bootstraps for each type of outcome, treating events from other outcomes as censoring4,5. Kaplan-Meier methods were used to estimate OS, MFS, LFS and event-free survival (EFS) with PV events defined as MF progression, leukemic transformation or death, as done for the MPN-PRC. Cox proportional hazard models were used for multivariable analysis (MVA) of MF progression and mortality risk.

Results: Prognosis was predicted for 475 PV pts. Median age was 56 years (yrs), 49% were female, and 15% had a history of thrombosis. Median follow-up duration was 11 yrs. Complete genomic data and karyotype was available in 225 (47%) and 232 (49%) pts, respectively. At genomic sampling, median blood counts were the following: WBC 10 x103/µL; PLT 427 x103/µL; and HGB 14.5 g/dL.

Median OS was 27.2 yrs. The model effectively predicted OS across all considered timepoints with computed c-indexes of ≥0.8 and a significant bootstrap 95% CI (5-yr 0.84, CI 0.77-0.90; 10-yr 0.85, CI 0.80-0.91; 20-yr 0.81, CI 0.59-0.98). As expected, age was a strong predictor of OS at all time points. In addition, TP53 mutation was associated with lower 5 yr and 10 yr OS (p=0.005, p=0.019) in univariate models but was not significant in MVA including age. A higher 20-yr mortality was associated with older age (p<0.001), PHF6 mutation (p=0.013), chromosome 5q deletion (p=0.011) and splenomegaly (p=0.045), but only age and PHF6 status remained significant in MVA.

Median MFS was 28.6 yrs, with 92 (19%) pts progressing to MF. MFS prediction was excellent at the 5-yr timepoint (c-index 0.91, CI 0.89-0.94), but poor thereafter (c-index at 10-yr 0.49, CI 0.32-0.62; at 20-yr 0.56, CI 0.34-0.81, not significant). Pts who progressed to MF within 5 yrs were older (71 vs 56, p=0.085) and had lower PLT (273 vs 432 x103/µL, p=0.035). We identified no significant imbalance in genomic profiles between those that progressed early vs those that did not. Median LFS was not reached and prediction of LFS was weak (c-index 5-yr 0.68, CI 0.48-0.88; 10-yr 0.72, CI 0.53-0.89; 20-yr 0.65, CI 0.58-0.93). This poor performance may be due to the long 10-yr LFS of 99% in our cohort. Actual median EFS was 20 yrs whereas the model predicted 24 yr median EFS. Including all events occurring within the 25 yr period considered by the MPN-PRC, correlation between predicted and actual EFS was poor (correlation coefficient 0.43).

Conclusion: The MPN-PRC performed well in predicting OS and early onset MF but did not reliably predict long-term MFS or LFS. Predictions were primarily driven by age, with genomic profiles adding limited value. The model overestimated median EFS in our PV cohort, perhaps owing to the MPN-PRC model having been trained on a cohort predominated by pts with ET. These findings confirm the model's utility in clinical practice for short-term, personalized OS and MFS prognosis for pts with PV, while indicating a need for refinement in long-term predictions to aid clinical decision-making.

References

1. Grinfeld et al. NEJM, 2018.

2. Abu-Zeinah et al. Leukemia, 2021.

3. Grinfeld et al. https://www.sanger.ac.uk/tool/progmod/progmod/

4. Uno et al. Stat Med, 2011.

5. Shi, Y. https://github.com/YushuShi/correctedC

Disclosures: Scandura: Constellation: Consultancy, Membership on an entity's Board of Directors or advisory committees; Karyopharm: Membership on an entity's Board of Directors or advisory committees; Incyte: Membership on an entity's Board of Directors or advisory committees; Protagonist Therapeutics: Membership on an entity's Board of Directors or advisory committees; Morphic: Consultancy; Medpacto: Research Funding; Calico: Consultancy; SDP Oncology: Membership on an entity's Board of Directors or advisory committees.

*signifies non-member of ASH