In this subgroup analysis of REACH3 study, 37 Japanese patients were randomized to receive either ruxolitinib (n = 22) or BAT (n = 15) (Fig. 1). At data cut-off (May 08, 2020), study treatment was ongoing in 11 (50%) patients in ruxolitinib group vs 6 (40%) patients in BAT group; 11 (50%) and 9 (60%) patients in the ruxolitinib and BAT groups discontinued the randomized treatment. Reasons for treatment discontinuation (ruxolitinib vs BAT) included adverse events (36.4% vs 20%), lack of efficacy (4.5% vs 6.7%), disease relapse (4.5% vs 6.7%), and death (4.5% vs 0%). Of the 15 patients in the BAT group, 2 (13.3%) participants crossed over to ruxolitinib on or after week 24.

Patient disposition. *Ongoing treatment and/or assessments during the main treatment period at the data cut-off date May 08, 2020. AE adverse events, BAT best available therapy, MMF mycophenolate mofetil

Demographics and baseline characteristics of the patients were similar among both the treatment groups (Table 1). Among the Japanese patients who participated in REACH3 study, median age was 48.5 years (range, 15–68) and 40 years (range, 16 − 66) in ruxolitinib and BAT groups, respectively. Patients who had severe cGvHD at baseline were 75.7% that was numerically higher than the overall population in REACH3 study (56.8%). The stem cell sources were bone marrow (45.5%, 66.7%), peripheral blood (50.0%, 20.0%), and single cord blood (4.5%, 13.3%) in the ruxolitinib vs BAT groups, respectively. The proportions of HSCT from HLA-matched donors were 52.2% in the ruxolitinib group and 37.5% in the BAT group. Median time from the onset of cGvHD to randomization was 28.9 weeks (5.9–93.0) in ruxolitinib and 24 weeks (2.6–46.0) in BAT groups. The most common reason for diagnosis of SR/D cGvHD was lack of response or disease progression after prednisone treatment in 36.4% of patients in the ruxolitinib group and 13.3% of patients in the BAT group and overall, 37.8% of patients had steroid dependency. Prior medications and organ baseline scores are presented in Supplementary Table 1. The most common organ involvement in Japanese patients was lung (68.2% vs 66.7%) that was numerically higher than the overall population (44.8% vs 40.9%) in ruxolitinib versus BAT groups. In the overall population, skin was the most common organ involved (73.3% vs 68.9%) while this was lower in Japanese population (50.0% vs 60.0%).

The most common initial BAT selected in the study is presented in Supplementary Table 2. The underlying diseases, malignant in majority of patients, primarily included acute myeloid leukemia (63.6%, 40%) and myelodysplastic syndrome (22.7%, 20.0%) in the ruxolitinib vs BAT groups, respectively (Supplementary Table 3).

PK parameters (Cmax, Tmax, and AUClast) were obtained in 1 adolescent and 2 adult Japanese patients in the ruxolitinib group on each sampling day of day 1 and day 15. The ranges of individual values of Cmax, Tmax, and AUClast were 152–312 ng/mL, 0.483–0.667 h, and 595–1400 ng∙h/mL on day 1, and 281–451 ng/mL, 0.483–2.00 h, and 849–2190 ng∙h/mL on day 15, respectively. Trough concentrations in Japanese patients in the ruxolitinib group were obtained postdose up to week 24. The individual data ranged from 10.9 to 196 ng/mL (n = 2 to 17). The Japanese PK data were comparable to the individual range observed in non-Japanese patients.

The primary endpoint of ORR at week 24 was numerically higher with ruxolitinib than with BAT (50.0% [11/22] vs 20.0% [3/15]; OR, 4.13; 95% CI, 0.90–18.9; Fig. 2a) and ORR at week 12 was 45.5% [10/22] in ruxolitinib group relative to 33.3% (5/15 in BAT group [OR, 1.78; 95% CI, 0.45–7.15]). The number of responders at week 24 in the BAT group was 2 patients for MMF and 1 each for imatinib. The BOR, up to week 24, was observed in 68.2% (15/22) of patients treated with ruxolitinib vs 46.7% (7/15) of patients treated with BAT (OR, 2.69; 95% CI, 0.66–10.9; Fig. 2b). Median (range) time to first response for BOR was 4.1 (2.1–15.3) weeks in ruxolitinib and 8.0 (2.1–11.1) weeks in BAT groups.

a Overall response rate (ORR) and b best overall response (BOR) up to week 24. a Overall response rate at week 24. ORR is defined as proportion of patients who achieved CR or PR according to 2014 NIH consensus criteria. b Best overall response rate up to week 24. BOR is defined as proportion of patients who achieved CR or PR at any time point up to week 24 or the start of additional systemic therapy for cGvHD. BAT best available therapy, BOR best overall response, CR complete response, n number of patients who are at the corresponding category, OR odds ratio, ORR overall response rate, PR partial response

Two patients in BAT crossed over to ruxolitinib, both achieved PR by crossover week 24. Although interpretation requires caution because of very limited number of patients, response at week 24 was observed in majority of patients independently from organs involved. Patients with skin (45.5% vs 11.1%), mouth (50.0% vs 0%), joints and facia (60.0% vs 33.3%) involvement were the most remarkable responses at week 24 that were observed in comparing ruxolitinib with BAT group (Table 2). Median FFS were 18.6 months in ruxolitinib and 3.7 months in BAT, respectively (HR, 0.34; 95% CI, 0.14–0.85). Kaplan–Meier (KM) estimate of FFS at 6 months was 72.73% in ruxolitinib (95% CI, 49.10–86.71) and 33.33% in BAT (95% CI, 12.15–56.40) (Supplementary Fig. 1). The median OS in ruxolitinib group was 21.9 (95% CI, 3.2, not evaluable [NE]) months and NE (6.7, NE) in the BAT group.

Patient-reported outcomes (PROs) were measured in this study. Modified Lee Symptom Scales, mLSS, is a cGvHD-specific self-administered survey where higher score indicates worse symptoms. At baseline, the median (range) mLSS score in the ruxolitinib (n = 22) was 15.48 (1.2–64.0) that was comparable with BAT group (n = 15) 16.90 (4.0–31.2). At week 24, median (range) mLSS was improved to 3.33 (0.0–40.7) in ruxolitinib group (n = 13), whereas it was 12.26 (0.0–28.9) in BAT group (n = 10) (Figs. 3, 4). Three patients in ruxolitinib group (13.6%) achieved clinically meaningful response defined as a ≥ 7-point TSS reduction from baseline at week 24 while none in BAT. Dose of steroid gradually decreased over time in both ruxolitinib and BAT groups, with a slightly faster decrease with ruxolitinib (Supplementary Fig. 2). Other PROs, FACT-BMT and EQ-5D-5L scores, suggested slight improvements of QoL in the patients treated with ruxolitinib, where higher scores suggest improvement of QoL in both scales (Supplementary Table 4). Median DOR was NE (95% CI, 2.2–NE) with ruxolitinib and 5.1 months (95% CI, 1.2–NE) with BAT. KM estimated for DOR at 6 and 12 months were 78.97% (95% CI, 47.91–92.71) and 71.79% (95% CI, 41.11–88.38) with ruxolitinib and 42.86% (95% CI, 9.78–73.44) and NE (95% CI, NE–NE) with BAT groups, respectively.

Kaplan–Meier estimate of failure-free survival (FFS) at week 24. BAT best available therapy, BID twice a day, FFS failure-free survival, RUX ruxolitinib. FFS is defined as time from the date of randomization to the earliest of recurrence of underlying disease, start of new systemic treatment for cGvHD, or death

Change from baseline in summary symptom score (mLSS) at week 24. Change from baseline calculated for patients with available data at baseline and the corresponding post baseline timepoint. *Patients with change of or addition of new systemic cGvHD treatment are counted as non-responders irrespective of the TSS value. BAT best available therapy, cGvHD chronic graft-versus-host disease, mLSS modified Lee cGvHD symptom scale, RUX ruxolitinib, TSS total symptom score

Among the Japanese participants, safety analysis was performed for 37 patients (ruxolitinib = 22, BAT = 15) who received at least one dose of study treatment. The median duration of exposure to therapy up to week 24 was similar between two arms, 25.6 weeks (range 3.0–25.6) in the ruxolitinib group and 25.6 weeks (range 0.6–25.6) in the BAT group, whereas the exposure up to data cut-off was longer in the ruxolitinib group, 44.4 weeks (range, 3.0–112.1) in the ruxolitinib group, and 29.1 weeks (range, 0.6–83.3) in the BAT group. The median dose intensity of ruxolitinib up to week 24 was 19.8 mg/day (range, 8.4–20.5) in Japanese patients which was similar to the overall population (19.6 mg/day, range 4.8–20.5).

Adverse events of any grade up to week 24 are reported in Table 3. In the ruxolitinib group, 13 (59.1%) patients reported at least one grade ≥ 3 adverse event compared with 12 (80.0%) patients in the BAT group. Grade ≥ 3 anemia (5 [22.7%] vs 1 [6.7%]) and pneumonia (5 [22.7%] vs 3 [20.0%]) were the most common adverse events reported in ruxolitinib vs BAT group (Table 3). Serious adverse events up to week 24 occurred in 10 (45.5%) patients treated with ruxolitinib and 8 (53.3%) patients treated with BAT in the study. Up to week 24, adverse events leading to dose adjustment or interruption occurred in 7 (31.8%) and 2 (13.3%) patients and adverse events leading to treatment discontinuation occurred in 9 (40.9%) and 3 (20.0%) in the ruxolitinib and BAT groups, respectively (Supplementary Table 5).

Up to week 24, infections of any type occurred in 16 (72.7%) of patients in the ruxolitinib group vs 11 (73.3%) in the BAT group, and ≥ grade 3 occurred in 7 (31.8%) vs 2 (13.3%), respectively. Viral infections were the most common (8 [36.4%] vs 8 [53.3%]), followed by bacterial (8 [36.4%] vs 7 [46.7%]), and fungal (5 [22.7%] vs 0; Table 4). At data cut-off, on-treatment death was reported in 5 patients who received ruxolitinib and 2 patients who received BAT, primarily due to complications caused by cGvHD and/or its treatment. Fatal serious adverse events related to study treatment were reported for 4 patients; 3 patients in the ruxolitinib group with pneumonia (n = 1), respiratory failure (n = 1), pneumonia bacterial (n = 1), sepsis (n = 1) where pneumonia and respiratory failure had one case reported from a same patient, and 1 patient in the BAT group with pneumonia (n = 1). All patients who died had severe cGvHD at baseline. The incidence of cancer relapse and progression was low in both arms (1 patient each).