Competing risk model of relapse versus non-relapse mortality in myelofibrosis patients undergoing allogeneic hematopoietic cell transplantation

Abstract

Abstract Background Allogeneic hematopoietic cell transplantation (allo-HCT) remains the only curative option for myelofibrosis (MF), but its success is limited by relapse and non-relapse mortality (NRM), which are competing events. Accurate assessment requires competing risks models. We applied cause-specific Cox regression to evaluate predictors of relapse and NRM using clinical, donor, and molecular factors, including early post-transplant chimerism. Methods We retrospectively analyzed 192 MF patients undergoing allo-HCT at City of Hope (Jan 2004–Mar 2025; cutoff Apr 25 2025). Risk was stratified per MIPSS70. Donor types included MRD, MUD, MMUD, and haploidentical; Single-locus mismatched related donors were grouped with MRD given their near-complete HLA match. Chimerism was assessed using STR-PCR, transitioning to qPCR in later years; mixed chimerism was defined as <97.5% donor, consistent with thresholds applied in qPCR-based transplant studies. Chimerism was measured on unsorted bone marrow cells, reflecting total donor engraftment rather than lineage-specific compartments; earlier time points (e.g., Day 30) were not available for analysis. High-risk mutations were defined as the presence of ASXL1, TP53, or RAS-pathway (NRAS/KRAS/CBL) mutations, consistent with transplant-specific studies demonstrating their adverse prognostic impact. Cause-specific Cox models were fit for relapse and NRM, treating competing events as censoring. Models were restricted to 139 patients with evaluable Day 100 chimerism; patients who died before Day 100 were excluded. Results Median follow-up among survivors was 5 years. The cohort's baseline characteristics were: median age 62 (38–75), 59% male, primary MF 55%, secondary MF 45%, and MIPSS70v2.0 risk distribution (Low 24%, Intermediate 25%, High 49%, Very High 4%). High-risk mutations were present in 41% (182 tested); driver mutations included JAK2 (61%), CALR (19%), MPL (4%), and triple-negative (16%). Conditioning was reduced-intensity in 98% and myeloablative in 2%. The 2-year cumulative incidences were 28% relapse and 34% NRM. Fine-Gray analysis showed D+100 mixed chimerism strongly associated with relapse. While D+100 mixed chimerism likely reflects incipient relapse, it remains a clinically relevant marker associated with high relapse risk (2-year incidence 52.9 % vs 11.5%; p<0.0001) and may guide early intervention. High-risk mutations were analyzed but not significantly associated with relapse or NRM. In multivariable Cox models, mixed chimerism at D+100 remained an independent predictor of relapse (HR 7.27; 95% CI 3.03–17.42), while MMUD donors were not significantly associated with relapse. No covariates—including donor type, TP53, or high-risk mutations—were significant for NRM. Year 1 mixed chimerism trended toward higher relapse (HR 3.19; 95% CI 0.42–24.16) but lacked power. Landmark analysis at 1 year demonstrated that patients with complete donor chimerism had superior RFS, although the small number of mixed cases precluded statistical significance. The survival curves remained clearly separated across follow-up (Figure 1B). GVHD rates were as follows: acute GVHD (grade II–IV): 43%, chronic GVHD: 36%. Cytogenetic abnormalities at relapse were documented in 19/54 patients (35%), with 74% of those retaining their pre-transplant mutations. However, relapse mutation typing was limited to pre-transplant genotypes. Conclusion This competing risks analysis confirms that Day 100 mixed chimerism (<97.5%) is a powerful predictor of relapse in both Fine-Gray and multivariable Cox models. Year 1 mixed chimerism showed a similar but non-significant trend. These findings underscore the prognostic value of early donor chimerism assessment and support its use in risk-adapted post-transplant monitoring. Patients with mixed chimerism <97.5% at Day 100 may benefit from closer surveillance and consideration of preemptive interventions.