Introduction

Graft-versus-host disease (GVHD) is a leading cause of morbidity and mortality in patients who undergo allogeneic hematopoietic stem cell transplantation (HSCT) [1,2,3]. Despite prophylactic immunosuppressive therapies, GVHD develops in ≈50% of allogeneic HSCT recipients. Acute GVHD (aGVHD) mainly affects skin, gastrointestinal (GI) tract, and liver, whereas chronic GVHD (cGVHD) commonly presents in the mouth, genitalia, liver, lungs, skin, GI tract, joints, and muscles [4, 5].

Corticosteroids are the current standard of care for initial GVHD therapy [4, 6,7,8,9]. Approximately 30–60% of patients with GVHD become refractory to treatment with corticosteroids [1, 10, 11]. Among patients with steroid-refractory (SR)-aGVHD, the 6-month survival rate is approximately 50% [8], whereas overall survival for patients with SR-cGVHD is approximately 72% at 3 years [12, 13]. Systemic corticosteroids have known severe treatment-related toxicities that limit long-term use [14].

For GVHD after failure of at least one line of systemic therapy, 3 therapies have been approved by the US Food and Drug Administration (FDA) since 2017: ibrutinib and belumosudil are approved for treatment of cGVHD only, and the selective Janus kinase 1/2 (JAK1/JAK2) inhibitor ruxolitinib is approved for treatment of both aGVHD and cGVHD [15]. Per the prescribing information, the GVHD indications for ruxolitinib are SR-aGVHD in adult and pediatric patients ≥12 years and cGVHD after failure of 1–2 lines of systemic therapy in adult and pediatric patients ≥12 years, and NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®) list ruxolitinib as a category 1 systemic agent for both SR-aGVHD and SR-cGVHD [16].

Ruxolitinib has been studied in randomized, open-label phase 3 studies in both aGVHD (REACH2) and cGVHD (REACH3). REACH2 compared efficacy and safety of ruxolitinib with investigators’ choice of best available therapy (BAT) in 309 patients with SR-aGVHD [14]. Overall response rate (ORR) was significantly higher at Day 28 for ruxolitinib compared with BAT (62% vs 39%, respectively; P < 0.001) [14]. Median failure-free survival was also significantly longer in patients treated with ruxolitinib compared with those treated with BAT (5.0 vs 1.0 month, respectively) [14]. REACH3 compared efficacy and safety of ruxolitinib with investigators’ choice of BAT in 329 patients with SR-cGVHD [1]. ORR and best overall response (BOR) up to Week 24 were significantly higher for ruxolitinib versus BAT (ORR, 49.7% vs 25.6%, P < 0.001; BOR, 76.4% vs 60.4%, P = 0.001) [1]. Patients who received ruxolitinib had significantly longer median failure-free survival versus those who received BAT (REACH3, >18.6 vs 5.7 months; P < 0.001; REACH3 3-year follow-up, 38.4 vs 5.7 months); this finding is consistent with a significantly higher response (24.2% vs 11.0%; P = 0.001) on the modified Lee Symptom Scale at 24 weeks, a predictor of longer overall survival and nonrelapse mortality [1, 17].

The REACH2/3 safety profiles were consistent with the well-established long-term safety profile of ruxolitinib [18]. Severe grade ≥3 adverse event rates were similar for the 2 treatments (REACH2: ruxolitinib [78%], BAT [78%]; REACH3: ruxolitinib [57%], BAT [57.6%]). Ruxolitinib treatment has been associated with cytopenias, and the most common grade ≥3 adverse events in both REACH trials were thrombocytopenia (REACH2, ruxolitinib [27%], BAT [15%]; REACH3, ruxolitinib [15.2%], BAT [10.1%]) and anemia (REACH2, ruxolitinib [22%], BAT [19%]; REACH3, ruxolitinib [12.7%], BAT [7.6%]) [1, 14]. Anemia was also the most common grade ≥3 event in the 3-year safety analysis of REACH3 (ruxolitinib [17.6%], BAT [9.5%]), and thrombocytopenia occurred at rates ≥5% higher for ruxolitinib than BAT [19].

Because hematologic events are a manifestation of GVHD pathogenesis, ruxolitinib treatment for aGVHD/cGVHD should be optimized for patients with cytopenias. Significant symptomatology and rapid organ progression may be associated with aGVHD/cGVHD, requiring prompt recognition of SR disease and initiation of second-line therapy. Timing of treatment initiation might affect GVHD treatment outcomes and is an important factor in GVHD management [20]. The post hoc analyses of the REACH2 and REACH3 trials presented here evaluated the impact of timing of treatment initiation after SR diagnosis and effect of concomitant cytopenias on treatment response in patients with SR-aGVHD and SR-cGVHD.

Methods

REACH2 and REACH3 clinical studies

REACH2 (NCT02913261) and REACH3 (NCT03112603) phase 3 clinical studies compared efficacy and safety of ruxolitinib with investigator-selected BAT in patients with GVHD [1, 14]. Patients from REACH2 had grade II–IV SR-aGVHD. Patients from REACH3 had moderate or severe SR or steroid-dependent cGVHD, collectively referred to as patients with SR-cGVHD. Both studies were multicenter, randomized, open-label, and enrolled patients ≥12 years old who had received an allogeneic HSCT.

Patients were randomized in a 1:1 ratio to 10 mg ruxolitinib twice daily or investigator-selected BAT. All patients continued to receive corticosteroids with/without calcineurin inhibitors. The REACH trials were designed/conducted in accordance with the International Council for Harmonisation Guideline for Good Clinical Practice, Declaration of Helsinki, and local regulatory requirements. Study protocols were approved by institutional review boards at each site. Patients or their guardians provided written informed consent.

Patient subgroups for treatment initiation timing analysis

To analyze the effect of treatment initiation times on outcomes, patients were divided into subgroups based on time from SR-GVHD diagnosis to treatment initiation. SR-aGVHD subgroups were early (0–3 days), late (4–6 days), and very late ( ≥ 7 days). SR-cGVHD subgroups were early ( ≤ 14 days), late (15–28 days), and very late ( > 28 days). For some analyses, late and very-late subgroups were combined. Subgroups were selected to be congruent with and applicable to current clinical practice for patients with aGVHD and cGVHD [21, 22].

Patient subgroups for cytopenia analysis

Cytopenia analysis was done only with patients with SR-aGVHD, based on subgroups defined by thresholds for platelet (PLT), hemoglobin (Hb), white blood cell count (WBC), or absolute neutrophil count (ANC) measured between baseline and Week 4. Low blood count thresholds were defined as: <30 × 109 cells/L (PLT); <8 g/dL (Hb); <5 × 109 cells/L (WBC); <1 × 109 cells/L (ANC). Thresholds were selected to be clinically relevant for guiding treatment decision-making and supportive care interventions for patients with GVHD [23].

Endpoints and assessments

Post hoc analyses evaluated (1) ORR, defined as the percentage of patients who had complete or partial responses (CR/PR, per REACH2/3 [1, 14]) at Day 28 (with SR-aGVHD) or Week 24 (with SR-cGVHD); (2) BOR, defined as the percentage of patients with CR or PR at any time; and (3) duration of response (DOR), defined for SR-aGVHD as time from first response until aGVHD progression or new systemic therapy and for SR-cGVHD as the time from first documented CR/PR among patients with a response at Week 24 until cGVHD progression, death, or additional systemic cGVHD therapies. For DOR, competing events for patients with SR-aGVHD were onset of cGVHD or death without progression of aGVHD; no competing events were defined for SR-cGVHD DOR.

In REACH2, patients with SR-aGVHD who received ≥1 dose of trial treatment were evaluated for the impact of cytopenias on treatment responses and laboratory values. Durable response was defined as the percentage of patients in each SR-aGVHD treatment group who maintained a Day 28 response at Day 56.

Statistical analysis

For comparative analyses between ruxolitinib treatment versus BAT or between ruxolitinib-only treatment initiation groups, odds ratios and 95% Wald CIs for ORR (or BOR) were calculated using the stratified Cochran-Mantel-Haenszel test. For REACH2, DOR was plotted by cumulative incidence curve, and median and quartiles were estimated using the Kaplan-Meier method. For REACH3, DOR was analyzed using the Kaplan-Meier method.

Results

Baseline characteristics of patients with SR-aGVHD

Demographics and clinical characteristics of patients with SR-aGVHD are shown by subgroups in Table 1. Of 309 patients, 154 (49.8%) patients were allocated to the ruxolitinib arm and categorized as receiving treatment early (n = 112 [72.7%]), late-and-very-late, (n = 42 [27.3%; late, n = 18, very late, n = 24]), or very late (n = 24 [15.6%]). The remaining 155 (50.2%) patients were randomized to BAT and categorized as receiving treatment early (n = 115 [74.2%]), late-and-very-late (n = 40 [25.8%; late, n = 18, very late, n = 22]), or very late (n = 22 [14.2%]). Mean time from diagnosis of initial aGVHD grade ≥II to randomization for treatment (ruxolitinib or BAT) of SR-aGVHD was 22.4, 37.5, and 42.0 days for the early, late-and-very-late, and very-late subgroups, respectively. Median ages were similar across subgroups, and most patients were men (Table 1).

Table 1 Baseline characteristics of patients with SR-aGVHD.
Full size table

At randomization, most patients had grade II or III SR-aGVHD, with more patients having grade III than grade II SR-aGVHD across all treatment initiation subgroups (34.8%, 30.5%, and 30.4% had grade II SR-aGVHD in the early, late-and-very-late, and very-late subgroups, respectively, vs 42.3%, 48.8%, and 47.8% with grade III SR-aGVHD). Skin and lower GI tract organs were most frequently involved; (56.8%, 46.3%, and 43.5% had skin involvement in the early, late-and-very-late, and very-late subgroups, respectively; 67.0%, 72.0%, and 65.2% had lower GI tract involvement).

Patients with SR-aGVHD who received ≥1 dose of trial treatment (N = 302) were included in the safety analysis, which evaluated the impact of cytopenias (ruxolitinib, n = 152 [50.3%]; BAT, n = 150 [49.7%]).

Baseline characteristics of patients with SR-cGVHD

Demographics and clinical characteristics of the 329 patients with SR-cGVHD are shown in Table 2. Of these patients, 165 (50.2%) were randomized to ruxolitinib and categorized as receiving treatment early (n = 88 [53.3%]), late-and-very-late (n = 77 [46.7%; late, n = 30, very late, n = 47]), and very late (n = 47 [28.5%]). The remaining 164 (49.8%) patients were randomized to BAT and categorized as receiving treatment early (n = 85 [51.8%]), late-and-very-late (n = 79 [48.2%; late, n = 33, very late, n = 46]), and very late (n = 46 [28.0%]). Mean time from diagnosis of cGVHD to randomization of ruxolitinib or BAT was 175.7, 290.1, and 323.2 days for early, late-and-very-late, and very-late subgroups, respectively. Median ages for the early, late-and-very-late, and very-late subgroups were all 49.0 years. Most patients were men (59.5%, 62.8%, and 64.5% for the early, late-and-very-late, and very-late subgroups, respectively).

Table 2 Baseline characteristics of patients with SR-cGVHD.
Full size table

Approximately half of patients with cGVHD (n = 180/329 [54.7%]) had experienced prior aGVHD. A numerically greater percentage of patients in the early treatment group had prior aGVHD compared with the other 2 groups (57.2%, 51.9%, and 48.4% of patients in the early, late-and-very-late, and very-late subgroups, respectively). At time of study entry, all but 1 patient (in the very-late subgroup) had moderate or severe SR-cGVHD.

Treatment initiation in patients with SR-aGVHD

Among patients with SR-aGVHD, Day 28 ORR was significantly higher with ruxolitinib treatment versus BAT in the early and late-and-very-late subgroups and numerically greater in the very-late treatment subgroup (Fig. 1a). A greater percentage of patients treated with ruxolitinib versus BAT achieved CR in all 3 time-to-treatment subgroups (Fig. 1a). BOR was also significantly higher with ruxolitinib treatment compared with BAT in the early and late-and-very-late subgroups and numerically greater in the very-late treatment subgroup (Fig. 1a). Median DOR was longer with ruxolitinib treatment versus BAT in the early (178.0 vs 101.0 days) and late-and-very-late treatment initiation subgroups (152.0 vs 88.0 days; Fig. 2a).

Fig. 1: Overall response rate and best overall response by treatment initiation subgroup.
Fig. 1: Overall response rate and best overall response by treatment initiation subgroup.
Full size image

a ORR: CR + PR at Day 28 (SR-aGVHD). b ORR: CR + PR at Week 24 (SR-cGVHD). Values at the top of graphs indicate OR (95% CI) for ruxolitinib vs BAT. BAT best available therapy, CR complete response, OR odds ratio, ORR overall response rate, PR partial response, SR-aGVHD steroid-refractory acute graft-versus-host disease, SR-cGVHD steroid-refractory chronic graft-versus-host disease.

Fig. 2: Duration of response in patients treated with ruxolitnib or BAT.
Fig. 2: Duration of response in patients treated with ruxolitnib or BAT.
Full size image

a DOR in patients with SR-aGVHD. b DOR in patients with SR-cGVHD. BAT best available therapy, DOR duration of response, IQR interquartile range, NE not evaluable, SR-aGVHD steroid-refractory acute graft-versus-host disease, SR-cGVHD steroid-refractory chronic graft-versus-host disease.

Responses among early, late, and very-late subgroups were compared in patients with SR-aGVHD treated with ruxolitinib. No significant differences between time-to-treatment subgroups were observed for ORR or BOR, although CR rates were numerically higher in early and late ruxolitinib initiation subgroups versus the very-late subgroup (Fig. 3a). Probability of loss of response was greater with late initiation and very late initiation of ruxolitinib versus early initiation (Fig. 3b).

Fig. 3: Time to ruxolitinib treatment initiation subgroup comparisons for responses and duration of response.
Fig. 3: Time to ruxolitinib treatment initiation subgroup comparisons for responses and duration of response.
Full size image

a ORR and BOR for patients with SR-aGVHD. b DOR for patients with SR-aGVHD. c ORR and BOR for patients with SR-cGVHD. d DOR for patients with SR-cGVHD. OR odds ratio, SR-aGVHD steroid-refractory acute graft-versus-host disease, SR-cGVHD steroid-refractory chronic graft-versus-host disease. a, c Values at the top of graphs indicate OR (95% CI) for early versus late or very late treatment initiation.

Treatment initiation in patients with SR-cGVHD

Among patients with SR-cGVHD, Week 24 ORR was significantly higher with ruxolitinib treatment versus BAT in all 3 time-to-treatment groups (Fig. 1b). BOR was significantly higher in the late-and-very-late and very-late ruxolitinib initiation subgroups with ruxolitinib treatment versus BAT, and numerically higher in the early subgroup (Fig. 1b).

Ruxolitinib treatment increased DOR compared with BAT in patients with SR-cGVHD. Median DOR ranged between 167.0–211.0 days in the BAT subgroups and was not reached in any ruxolitinib subgroup (Fig. 2b).

Responses among the early, late, and very-late subgroups were compared in patients with SR-cGVHD treated with ruxolitinib. No significant differences in ORR, BOR, or DOR were observed among ruxolitinib time-to-treatment subgroups (Fig. 3c, d).

Cytopenias in patients with SR-aGVHD

The ORRs for patients with SR-aGVHD were compared among patients stratified by low versus not-low WBC, ANC, PLT, or Hb count. ORRs at Day 28 among patients treated with ruxolitinib were similar between low and not-low WBC, ANC, PLT, or Hb subgroups individually (eg, low PLT, ORR 59.2% [95% CI 46.8–70.7] vs not-low PLT, ORR 66.7% [95% CI 55.3–76.8]). For the period of time that ORR was measured (until Day 28), the most frequent dose reduction was from 20 mg/d to 10 mg/d (44.4% of 135 patients with any cytopenia, vs 17.6% of 17 patients with no cytopenias; Supplementary Fig. 1), although the median dose of ruxolitinib (any cytopenia, 20.0 mg/d; no cytopenia, 20.0 mg/d; Q1–Q3 of median doses across all 152 patients with or without cytopenias, 17.85–20.0 mg/d) was approximately equal to the 20 mg/d starting dose in all subgroups regardless of cytopenia status. The presence of any cytopenia (low WBC, ANC, PLT, or Hb) did not diminish favorable response rates among patients treated with ruxolitinib (cytopenia subgroup, ORR 66.7% [95% CI 58.0–74.5] vs no cytopenia subgroup, ORR 35.3% [95% CI 14.2–61.7]; Fig. 4a). Response rate was numerically, but not significantly, higher for BAT than ruxolitinib in patients with no cytopenias (n = 6/17 [35.3%] treated with ruxolitinib; n = 7/15 [46.7%] treated with BAT). However, response rates were numerically or significantly higher in patients treated with ruxolitinib versus BAT for all cytopenia subgroups and for patients overall (Fig. 4a).

Fig. 4: Effect of cytopenias on responses in patients with SR-aGVHD treated with ruxolitinib or BAT.
Fig. 4: Effect of cytopenias on responses in patients with SR-aGVHD treated with ruxolitinib or BAT.
Full size image

a ORR at Day 28. b Durable response at Day 56, defined as the proportion of all patients who achieved a CR or PR at Day 28 and maintained a CR or PR at Day 56. c Median duration of response. ANC absolute neutrophil count, BAT best available therapy, CR complete response, Hb hemoglobin, IQR interquartile range, NE not evaluable, NR not reached, OR odds ratio, ORR overall response rate, PLT platelet count, PR partial response, SR-aGVHD steroid-refractory acute graft-versus-host disease, WBC white blood cell count. * Cytopenias were defined as low blood counts between baseline and Week 4; low blood cell counts were defined as: WBC, <5 × 109 cells/L; ANC, <1 × 109 cells/L; PLT, <30 × 109 cells/L; Hb, <8 g/dL.

Durable responses were observed across all subgroups (Fig. 4b). For the subgroup of patients with any cytopenia, median DOR was numerically longer for patients treated with ruxolitinib versus BAT (Fig. 4c). Mean PLT counts and Hb concentrations decreased over the first 4–12 weeks of treatment, then recovered, and were mostly similar between ruxolitinib and BAT treatment groups through Week 24 (Fig. 5). Mean WBC and ANC counts similarly decreased through Week 8 with both ruxolitinib and BAT and were generally lower in the ruxolitinib versus BAT group through Week 24 (Fig. 5).

Fig. 5
Fig. 5
Full size image

Hematologic laboratory values in patients with SR-aGVHD. ANC absolute neutrophil count, BAT best available therapy, Hb hemoglobin, PLT platelet count, SR-aGVHD steroid-refractory acute graft-versus-host disease, WBC white blood cell count.

Discussion

Treatment initiation in SR-aGVHD

Consistent with the overall results of the phase 3 REACH2 trial, these post hoc analyses found that Day 28 ORRs, BORs, CR rates, and DOR were higher for patients treated with ruxolitinib versus BAT [14]. From these analyses, early and late treatment groups had numerically higher CR rates than the very-late treatment group (36.6%, 33.3%, and 25.0%, respectively). These results are consistent with a retrospective analysis that found higher Day 28 CR rates among patients with SR-aGVHD treated with ruxolitinib within 7 days (n = 29/42 [69%] ≤7 days [median, 3 days] vs n = 5/20 [25%] >7 days, P = 0.001) [24]. For patients in REACH2, the lower GI tract was the most commonly affected organ. In a separate study on GI SR-aGVHD, the CR rate was similar to these post hoc analyses (35.1% for 77 patients with GI SR-aGVHD, 57 of whom were treated with ruxolitinib as second-line therapy) [25]. Overall, these findings suggest early treatment with ruxolitinib may provide the most benefit for patients with SR-aGVHD.

Treatment initiation in SR-cGVHD

Efficacy outcomes (Week 24 ORRs, BORs, and DORs) were also higher in patients treated with ruxolitinib versus those treated with BAT in both the post hoc REACH3 analyses presented here and the phase 3 REACH3 trial [1]. BOR rates were comparable in these post hoc analyses and in a single-center retrospective study of patients with refractory severe cGVHD (n = 68/88 [77.3%] for early ruxolitinib treatment initiation versus n = 17/23 [74%] treated with ruxolitinib and extracorporeal photopheresis) [26].

In this post hoc analysis of REACH3, timing of ruxolitinib treatment initiation did not affect response rates in patients with SR-cGVHD (Week 24 ORRs and DOR rates were similar among early, late, and very-late treatment initiation groups). However, among patients with SR-cGVHD with a treatment response, median DOR was 5.49–6.93 months for patients treated with BAT and was not reached in any of the ruxolitinib treatment initiation time subgroups, suggesting a more durable response with ruxolitinib regardless of treatment initiation time.

Cytopenias in patients with SR-aGVHD

Cytopenias are common adverse events in patients with GVHD, regardless of treatment, and are associated with worse outcomes [1, 14, 27,28,29]. Certain circumstances introduce additional confounding variables; for example, ganciclovir treatment for cytomegalovirus reactivation has been associated with neutropenia and thrombocytopenia [30]. Additionally, GVHD pathogenesis may also contribute to the frequency of cytopenias: aGVHD/cGVHD are independent risk factors for development of thrombocytopenia after HSCT [31]. Thus, understanding the impact of cytopenias on efficacy for both ruxolitinib and BAT is important.

The post hoc analysis of REACH2 indicated that the presence of low ANC, WBC, PLT counts, or Hb concentrations did not affect the favorable efficacy outcomes (ORR/DOR) in the treatment of SR-aGVHD. Patients were able to maintain the starting dose of ruxolitinib during the period in which Day 28 response rates were measured. Occurrence of dose reductions had minimal impact on patients’ response to ruxolitinib regardless of cytopenia status. Ruxolitinib was associated with statistically higher Day 28 response rates than BAT for most low and not-low blood count groups and for the full patient population overall, with durable responses observed at Day 56.

Ruxolitinib and BAT were associated with similar reductions in blood counts early in treatment. These findings highlight the potential myelosuppressive effects of aGVHD development related to the hyperinflammatory state and increased production of cytokines versus the impact of treatment on blood counts. These results are consistent with those reported in exploratory analyses of pooled data from the COMFORT studies, in which ruxolitinib-associated anemia, which generally occurs during early therapy, did not predict shortened survival [32, 33].

Study limitations

Limitations of this study include low patient numbers in each of the non-prespecified subgroups, the lack of blinding in the REACH2 and REACH3 studies, and the hypothesis-generating nature of the post hoc analyses.

Conclusions

These post hoc analyses of the REACH2 and REACH3 trials suggest that ruxolitinib may provide more robust clinical benefit compared with BAT for patients with SR-aGVHD or SR-cGVHD, regardless of treatment initiation time after diagnosis of SR disease. In patients with SR-aGVHD, response rates were higher with early ruxolitinib treatment, reinforcing the importance of initiation of early treatment in SR-aGVHD. However, studies with a larger patient population are needed to confirm this observation. In addition, the presence of cytopenias did not affect the beneficial efficacy outcomes of ruxolitinib versus BAT in patients with SR-aGVHD. Furthermore, these patients were able to maintain dose intensity of ruxolitinib throughout the 28-day period of the primary endpoint. Timing of ruxolitinib treatment initiation did not affect response rates or DOR in patients with SR-cGVHD.

These findings suggest that physicians treating HSCT recipients should consider initiating ruxolitinib treatment within 3 days of SR-aGVHD diagnosis to optimize response rates and durability of response. Maintaining dose intensity during ruxolitinib treatment is feasible, appears to provide better efficacy versus BAT in SR-aGVHD or SR-cGVHD, and is manageable with appropriate supportive care to achieve desired treatment efficacy outcomes.