Redefining Prognosis in Patients with KIT D816V Positive Advanced Systemic Mastocytosis on KIT Inhibitors

Introduction: In advanced systemic mastocytosis (AdvSM), established risk scores such as the Mutation-Adjusted Risk Score (MARS) or the International Prognostic Scoring System (IPSM) have been developed in parallel to clinical studies involving the meanwhile approved KIT inhibitors midostaurin and avapritinib. Consequently, MARS and IPSM have not been validated to account for the prognostic impact of KIT inhibition in the real-world setting. We therefore sought (i) to assess the prognostic superiority of KIT inhibitor-treated patients compared to KIT inhibitor-naïve patients and (ii) to devise a risk stratification algorithm specifically designed for patients undergoing treatment with KIT inhibitors.

Methods: In 229 patients with KIT D816V positive AdvSM, inverse probability of treatment weighting (IPTW)-weighted Cox proportional hazard models, adjusted for unbalanced covariates post-weighting, were employed to compare outcomes between KIT inhibitor-treated (n=129, 56%; midostaurin, n=129; second-line avapritinib, n=23) and KIT inhibitor-naïve (n=100, 44%) patients. Clinical and genetic variables at diagnosis were assessed upon their association with overall survival (OS, time from diagnosis to death). The risk score was further validated for leukemia-free survival (LFS, time from diagnosis to leukemic transformation into secondary mast cell leukemia [sMCL] and/or secondary acute myeloid leukemia [sAML]). Smoothing splines were utilized to model the functional risk representation of continuous variables, with linear functions proving effective for these parameters. Continuous encoding of variables yielded significantly better performance compared to categorical encoding informed by maximally selected rank statistics. Cumulative incidence functions were used to estimate the rates of leukemic transformation into sMCL/sAML, with death without progression treated as a competing risk for these estimates.

Results: The weighted Cox analysis revealed a significantly improved OS in patients treated with KIT inhibitors (hazard ratio [HR] [95% CI]: 0.645 [0.422, 0.979], P=0.0396). In multivariable analysis, adjusted for WHO-HAEM5 diagnosis, six risk factors were adversely associated with OS among patients receiving KIT inhibitors: (i) higher age (years), (ii) lower hemoglobin levels (g/dL), (iii) higher monocyte levels (x10/nL), (iv) increased number of mutations in the high-risk panel SRSF2/ASXL1/RUNX1, (v) cytogenetic abnormalities (excluding -Y), and (vi) absence of skin involvement. By aggregating the HR-coefficient-adjusted risk factors, four prognostically different risk categories were derived: low-risk (n=33, 25%, -2.0395 to 0.0460, median OS 12.3 years), intermediate-risk 1 (n=33, 25%, 0.0461 to 0.8734, median OS 5.1 years), intermediate-risk 2 (n=32, 25%, 0.8735 to 1.7357, median OS 3.1 years) and high-risk (n=32, 25%, 1.7358 to 5.7265, median OS 1.1 years). The proposed risk-score revealed superior discriminatory performance (Harrell’s C-index 0.775) compared to the WHO-HAEM5 classification (0.621) and the established scoring systems MARS (0.663) and IPSM (0.678). Leukemic transformation into sMCL/sAML was observed in 6%, 12%, and 20% of patients after 1, 3, and 8 years, respectively. The respective risk groups were also predictive for LFS (median 12.3 vs. 4.6 vs. 1.8 vs. 1.0 years).

Conclusions: After balancing for prognostically relevant disease variables, KIT inhibitor-treated patients showed significant superior OS. The proposed risk-score offers a contemporary four-tier risk categorization for OS and LFS in KIT D816V positive AdvSM patients on treatment with KIT inhibitors and may serve as a valuable tool for clinicians when considering intensifying treatment, e.g. with allogeneic hematopoietic cell transplantation.