Propensity score matching analysis comparing the efficacy and treatment outcomes of belumosudil treatment with ruxolitinib therapy as second-line or beyond in chronic GVHD patients after steroid failure

Abstract

Abstract Introduction Belumosudil (BEL), a selective Rho-associated coiled-coil kinase 2 (ROCK2) inhibitor, has been approved as 3rd line or beyond treatment option for steroid-refractory (SR) chronic graft-versus-host disease (cGvHD), while ruxolitinib (RUX), JAK1/2 inhibitor, has been an established treatment of choice as 2nd line for SR-cGvHD. Although multiple real-world studies replicated a consistent clinical efficacy of BEL with improved overall response rate (ORR) and failure-free survival (FFS), questions remain unanswered whether BEL is superior or equivalent to RUX for SR-cGvHD. Propensity-score matching (PSM) analysis is a statistical methodology that balances out a bias coming from the unbalanced distribution of patient characteristics at baseline for the variable of interest (e.g., treatment option). Thus, it could mimic RCT by comparing treatment outcomes indirectly after balancing biased covariates. The present study compared treatment outcomes between BEL-treated patients and cGvHD patients treated with RUX as a standard of care. Also, PSM was applied to control for confounding variables that could bias the results between the two groups. FFS, OS, and steroid dose reduction were evaluated as statistical endpoints. Patients and methods We retrospectively analysed treatment outcomes in a total of 329 patients, including 216 pts treated with BEL and 113 treated with RUX. Propensity score was calculated to adjust the following unbalanced clinical factors between the two groups, including GvHD severity (severe vs. mild/moderate), HCT-CI score (≥3 vs. <3), and treatment line (≥4th line vs. <4th), extracting 176 patients (88 in each group) for final analysis. Results The overall population median follow-up in survivors was 11.8 months (range 0.2–60). The BEL group showed severe cGvHD more frequently (80.1% vs. 59.3%, p<0.001), were more frequently at 4th line of treatment (75.9% vs. 59.2%, p=0.001), and had failed RUX therapy (71.8% vs. 17.7%, p<0.001) compared to the RUX group; the RUX group showed a higher HCT-CI score (27.1% vs. 16.8% ≥3, p=0.074). In the overall population, the BEL group showed a 66.8% [58.9–73.5] of 12 months' FFS rate vs 64.1% [53.6–72.8] in RUX group (p=0.669), whereas 12 months' OS rates were 92.6% [86.6–95.9] and 82.3% [73.2–88.5] (p=0.003), respectively. When comparing patients in the BEL group who were RUX naïve (n=60) vs. RUX exposed (n=155), a difference in FFS was found (80.9% vs. 60.5%, p=0.009), but none for OS (p=0.180). At month 3, 29.1% more patients in the BEL group could discontinue prednisone compared to the RUX group (45.3% vs. 16.2%, p<0.001). In the PSM subgroup, no differences were found for severe cGvHD (67.0% in both, p=1), HCT-CI ≥3 (20.5% in both, p=1), or 4th line of treatment (63.6% and 64.8%, p=1) between patients of the BEL and the RUX group. BEL group showed 68.4% [56.3–77.7] of 12 months' FFS rate vs 63.5% [51.3–73.3] for RUX group (p=0.931), whereas 12 months' OS rates were 91.7% [82.2–96.2] and 84.4% [74.0–90.9] (p=0.221), respectively. In the BEL group, no differences were found for FFS (p=0.162) or OS (p=0.264) between the RUX naïve (n=37) and RUX exposed (n=50) patients. Univariate and multivariate analysis, including treatment (i.e. BEL vs. RUX), HCT-CI ≥3, severe grade cGvHD, previous history of acute GvHD, treatment line (fourth line or higher), and previous RUX failure, did not show any difference between BEL and RUX for FFS (hazard ratio (HR) 0.978 [0.594–1.611], p=0.931) or OS (HR 0.617 [0.252–1.514], p=0.292); similarly, previous RUX failure (p=0.636 and p=0.322) and treatment as fourth line or higher (p=0.105 and p=0.854) did not show difference in FFS and OS, respectively. For FFS, an HCT-CI ≥ 3 (HR 1.885 [1.077–3.229], p=0.026) and severe grade cGvHD (HR 1.851 [1.034–3.312, p=0.038) were identified to be independent prognostic factors for FFS. Similarly, an HCT-CI ≥ 3 (HR 4.584 [1.926–10.91] p<0.001) and severe grade cGvHD (HR 8.082 [1.723–37.92], p=0.008) were identified as adverse prognostic factors for OS. In PSM subgroup, 19.6% (33.9% vs. 14.3%, p=0.007) and 18.1% (41.2% vs. 23.1%, p=0.044) more patients in BEL group could discontinue prednisone at months 3 and 6, compared to RUX group, respectively. Conclusion The current study showed no difference for FFS or OS between BEL and RUX as second-line therapy or beyond in cGvHD patients after therapy failure. However, it is noteworthy that steroid tapering could be faster in BEL compared to RUX.