Ibrutinib for therapy of steroid-refractory chronic graft-versus-host disease: a multicenter real-world analysis

Chronic graft-versus-host disease (cGVHD) is a common late immune-mediated disorder after allogeneic hematopoietic cell transplantation (HCT).¹ Standard initial therapy with prednisone often fails to provide durable complete resolution,^2,3 thus next-line systemic immune suppressive agents (ISs) are commonly required. Multiple agents are used for steroid-refractory cGVHD (SR-cGVHD), and 3 of these currently have the US Food and Drug Administration (FDA) approval (ibrutinib, ruxolitinib, and belumosudil).^4-6 Durable success in SR-cGVHD therapy is limited, however.^7,8 In secondary cGVHD therapy, failure-free survival (FFS; a composite end point inclusive of death, relapse, and requirement of an additional line of IS therapy) has been reported to be 56% by 6 months and 45% by 12 months.⁹ Thus, advances are needed to improve outcomes.

Prior work supports the importance of cell signaling mediators (Bruton tyrosine kinase in B cells and interleukin-2 inducible T-cell kinase in T cells) in the immune response that drives cGVHD. In the preclinical setting, targeted disruption of Bruton tyrosine kinase and T-cell kinase ameliorated cGVHD.¹⁰ In a multicenter phase 2 clinical trial, ibrutinib was tested in patients with cGVHD who had failed at least 1 prior line of IS therapy.⁵ The trial enrolled 42 patients, and they were treated with ibrutinib 420 mg daily until cGVHD progression. With a median follow-up of 13.9 months, the National Institutes of Health (NIH) best overall response was 67%, and most responders demonstrated response for at least 20 weeks. Responses were seen across involved organ sites, and steroid dose reduction was observed. Adverse events included diarrhea, fatigue, muscle spasm, nausea, and bruising. These results led to an FDA indication for ibrutinib in the treatment of cGVHD failing at least 1 prior line of therapy in August 2017.¹¹ A subsequent report described longer-term outcomes,¹² a pediatric trial has been conducted,¹³ and a single-center retrospective study has been completed.¹⁴ 

Questions remain regarding the efficacy, safety, and use of ibrutinib in SR-cGVHD, and no large-scale study has been conducted since approval. We hypothesized that a multicenter real-world observational study would provide information regarding ibrutinib treatment response, variation in response according to baseline cGVHD features, durability of treatment success, and treatment-emergent toxicities leading to ibrutinib dose reduction or discontinuation.

We assembled a nationally representative collaboration of 19 HCT centers in the United States. From each center, investigators identified HCT recipients treated with ibrutinib for SR-cGVHD after failure of at least 1 prior line of therapy. The eligible time frame was from 2 August 2017 (the FDA approval date for ibrutinib for cGVHD) to 17 August 2020 (to provide at least 6 months of follow-up time for patients at the time of initial protocol development). Relapsed malignancy after HCT was excluded, given the potential confounding on GVHD treatment practices and immune suppression (IS) taper/discontinuation.¹⁵ Otherwise, no limitations were placed on patient baseline characteristics, cGVHD organ involvement, or prior lines of IS therapy for cGVHD.

Investigators from each site retrospectively abstracted the data elements using standardized data collection guidelines and securely reported to the coordinating center (Moffitt) under an institutional Review Board (IRB) approved protocol. Variables collected in 18 of 19 total sites (n = 217 patients total) included patient characteristics, HCT features, donor type, graft source, GVHD prophylaxis, cGVHD features,^16,17 serial cGVHD organ scores at and after ibrutinib start, changes in prednisone dose, new IS therapy after ibrutinib, reasons for ibrutinib dose reduction or stop, and survival outcomes. Extensive data on treatment-emergent adverse events were also collected. The final study site (n = 53 patients) submitted data on the above except for cGVHD organ scores after ibrutinib start (thus did not contribute to response assessment, duration of response, or analysis on the association of baseline variables with response), ibrutinib reduction/discontinuation, or treatment-emergent adverse events (TEAE). Full details on data collection are listed in supplemental Methods.

NIH 2014 Consensus Guidelines were used to compute response at fixed time points, including NIH complete response (CR)/partial response (PR) at 3, 6, 9, and 12 months after ibrutinib start in the absence of progression (PD) or new line of IS therapy.¹⁸ Overall response rate (ORR) included CR/PR. All other responses (unchanged, mixed response, and progression) were considered lack of response. Best ORR was defined by the proportion with CR/PR at any point in the absence of progression (PD) or new line of IS therapy. Baseline variables were examined for association with response. Duration of response was defined as the time from the first achievement of CR/PR to either PD or new line of IS therapy and calculated separately for fixed time point response (6-month ORR) and best ORR. FFS was defined as the time from ibrutinib start to the first of death, relapse, or initiation of new IS after ibrutinib. Overall survival (OS), nonrelapse mortality, and relapse were estimated from ibrutinib start. Steroid dose reduction was defined by change in steroid dose (mg prednisone per kg body weight) from ibrutinib start onward. Ibrutinib discontinuation was defined as the date of permanent stop, whereas dose reductions were defined by change in dose from baseline. Toxicity events for reduction/discontinuation and the above-described TEAE of interest were captured using Common Terminology criteria for adverse events (CTCAE) version 5.

Patient, disease, HCT, and cGVHD variables were summarized using descriptive statistics. Treatment response rates were summarized with associated 95% confidence intervals (CIs). Time-to-event outcomes including OS and FFS used the Kaplan-Meier method, and stratified plots were compared using log-rank test. The cumulative incidence function accounting for competing risk events were used to estimate the incidence of outcomes including relapse, nonrelapse mortality, and use of systemic IS therapy beyond ibrutinib. The association between baseline variables and response outcomes and FFS were studied in univariable and multivariable (MVA) analyses. Variables retained in the MVA model had P value of <.25. Steroid dose reduction was examined as percent decrease from baseline dose over the time points studied. Toxicities were tabulated, and proportions of subjects affected by each were summarized.

A total of 270 patients were included. The median age was 57 years (range, 4-77), and 57% were male. A total of 12 patients (4.4%) were aged <18 years. The most common diagnoses were acute myelogenous leukemia (AML) (n = 89 [34%]), non-Hodgkin lymphoma (NHL) (n = 42 [16%]), myelodysplastic syndrome (MDS) (n = 37 [14%]), and acute lymphoblastic leukemia (ALL) (n = 35 [13%]). Sixty-one percent had HCT from matched unrelated donors, 31% from matched related, and 7.7% from other donors (mismatched unrelated, haploidentical, and cord blood). Eighty-eight percent had HLA matched donors, with the remainder having mismatched. Peripheral blood mobilized stem cells (PBSC) grafts (90%) were most common, whereas bone marrow (7.8%) and cord blood (1.9%) were less common. Original GVHD prophylaxis included calcineurin inhibitor (CNI)/methotrexate (n = 129 [49%]), sirolimus (SIR)/CNI-based (n = 82 [31%]), mycophenolate mofetil (MMF) with CNI or SIR (n = 46 [17%]), post-HCT cyclophosphamide (n = 6 [2%]), or ex vivo T-cell depletion (n = 2 [1%]).

Baseline cGVHD features are presented in Table 1. The patients (n) evaluable for each end point are listed in text, tables, and in supplemental Data. Involved organs at ibrutinib start included skin (75%), eye (61%), mouth (54%), joint/fascia (47%), GI (26%), lung (27%), liver (19%), genital (7%), and other (4.4%). Among those with skin involvement and skin type reported (n = 152), 68 (45%) had deep sclerosis, and 43 (28%) had superficial sclerosis, with the remainder having nonsclerotic disease. From 73 of 270 total patients with any reported baseline lung score, 59 had data collected on the presence of bronchiolitis obliterans (BOS) and baseline forced expiratory volume in 1 second (FEV1) percent (whereas others had only NIH 0-3 lung scores provided). Of these 59 patients, 32 (54%) were verified to have BOS per study site, and 22 (69%) had FEV1 values <75% (in which median FEV1 was 59% predicted). Overall NIH severity was mild in 5.7%, moderate in 42%, and severe in 53%. Thirty-nine percent had overlap subtype. KPS was ≥80% in 72% of patients. The median prednisone at ibrutinib start was 0.21 mg/kg (range, 0-2.27). Ibrutinib was started at a median of 18.2 months after cGVHD onset, most commonly at 420 mg dose (66% of patients; reductions in others for drug interaction [58% of dose reductions] or clinician discretion [36% of dose reductions]) and largely in earlier lines of cGVHD therapy (second line, 26%; third, 30%; fourth, 21%; fifth, 10%; sixth, 10%; seventh or higher, 1%). The median time from cGVHD onset to ibrutinib start differed according to the line of therapy for ibrutinib: second line, 9.5 months; third line, 13 months; fourth line, 23.8 months; fifth line, 37.1 months; and sixth line or higher, 50.7 months. This diversity in starting dose and line of therapy reflected actual practice outside of a clinical trial and was not controlled. The most commonly used prior agents were prednisone (87%), CNI (35%), SIR (26%), ruxolitinib (24%), and extra-corporeal photopheresis (ECP) (26%). Others (rituximab 17%, MMF 15%, methotrexate 5%, and belumosudil 4%) were less common.

A total of 214 patients were eligible for 6-month response assessment (from the parent 270, a total of 53 had no serial organ scoring available for response assessment from 1 study site, and 3 others from the remaining sites had no organ scores for response assessment). Of these 214, NIH response at 6 months (fixed response) was CR in 4, PR in 55, mixed response in 17, no change in 27, progression in 28, and not evaluable (due to missing organ scores) in 83. Thus, the 6-month NIH ORR (CR/PR) among 131 evaluable patients was 45% (42% PR and 3% CR; Figure 1). The median duration of response was 15 months (range, 1-46). Comparisons of those with vs without adequate serial organ scores to compute 6-month response are presented in supplemental Data; some differences were noted, including, for those without 6-month response data, longer days to ibrutinib start, greater overlap subtype, greater deep sclerosis, and greater overall NIH severe cGVHD at baseline. The fixed 6-month ORR per line of therapy was 59% (second line), 43% (third line), 48% (fourth line), 28% (fifth line), and 36% (sixth or greater line). Most cGVHD features (overlap status, platelets, KPS, prednisone dose, NIH overall severity, time from cGVHD, and prior lines of therapy) did not have a significant association with 6-month CR/PR, except liver involvement (MVA odds ratio, 5.49; 95% CI, 2.3-14.2; P < .001) that was associated with a greater likelihood of response (Table 2). The univariate 6-month ORR per baseline features is presented in supplemental Data. Assuming all nonevaluable patients were nonresponders, the 6-month ORR would be 28%. In contrast, from the 217 eligible for best ORR assessment, responses included CR in 11, PR in 101, lack of response in 89, and not evaluable (missing scores) in 16. Best ORR per 201 evaluable was 56%, whereas best ORR for all subjects (assuming nonevaluable were failures) was 52%.

With a median follow-up time for survivors of 30.5 months (range, 0.2-65.5), FFS was 59% (95% CI, 53-65) at 6 months and 41% (95% CI, 36-48) at 12 months (Figure 2A). As expected, most of failure events were from initiating next line of systemic IS therapy (Figure 2B). FFS stratified by 6-month ORR status is in Figure 2C, and FFS per univariate analysis of baseline features is presented in supplemental Data. The 12-month FFS per line of therapy was 73% (95% CI, 60-90) for second line, 53% (95% CI, 39-73) for third line, 44% (95% CI, 28-68) for fourth line, 56% (95% CI, 37-84) for fifth line, 62% (95% CI, 40-95) for sixth or greater line. Systemic IS agents given after ibrutinib included ruxolitinib in 86 (32%), ECP in 38 (14%), belumosudil in 27 (10%), axatilimab in 11 (4.1%), methotrexate in 10 (3.7%), MMF in 8 (3%), rituximab in 8 (3%), SIR in 7 (2.6%), interleukin-2 in 4 (1.5%), cyclophosphamide in 4 (1.5%), total lymphoid irradiation (TLI) in 3 (1.1%), prednisone in 3 (1.1%), bortezomib in 2 (0.7%), and tacrolimus in 2 (0.7%); then 1 each (0.4%) for baricitinib, etanercept, cyclosporine, ultraviolet B (UVB), prednisone/others, tocilizumab, and imatinib. On final MVA (Table 2), increased age (hazard ratio [HR], 1.01, 95% CI, 1.0-1.02; P = .033), higher baseline prednisone dose (HR, 1.92; 95% CI, 1.09-3.38; P = .032), and lung involvement (HR, 1.58; 95% CI, 1.1-2.28; P = .016) were associated with worse FFS. OS at 6 and 12 months were 93% (90%-97%) and 87% (83%-91%), respectively. Causes of death were relapse/progression (25% of total deaths), cGVHD (23%), infection (23%), respiratory failure (12%), multiorgan failure (5.8%), secondary malignancy (5.8%), GI hemorrhage (1.9%), myocardial infarction (1.9%), and postprocedural hemorrhage (1.9%). Ibrutinib dose reductions and discontinuation are listed in Table 3. TEAE are summarized in supplemental Data. Steroid dose reduction from baseline is also presented in supplemental Data.

In a subgroup analysis, we aimed to replicate findings from the prior ibrutinib phase 2 trial that led to FDA approval. We restricted our patient sample to those with evidence of erythematous skin rash and/or oral involvement and up to 3 prior lines of therapy and determined the best ORR to be 60% (95% CI, 43-75), which was similar to the reported best ORR in the original trial of 67%.

In separate exploratory analyses, we also focused on those specifically with BOS or cutaneous sclerosis, given the high morbidity and limited treatment response expected in these subgroups. For the BOS subgroup (defined by baseline lung involvement by lung 0-3 scoring and not by confirmed FEV1 reduction, given the limited availability of these data), the 6-month NIH ORR was 31%, and FFS at 6 and 12 months were 51% (95% CI, 36-73) and 33% (95% CI, 20-56), respectively. For the cutaneous sclerosis group (defined by baseline skin involvement and confirmed presence of either/both superficial or deep sclerotic changes), the 6-month NIH ORR was 38%, and FFS at 6 and 12 months were 63% (52%-76%) and 39% (29%-54%) respectively.

This multicenter real-world analysis provides large-scale coverage of safety and efficacy of ibrutinib in SR-cGVHD. The study illuminates actual use and practice patterns after its FDA approval in 2017. Although these are retrospective data, these provide valuable information in a much larger nationally representative context, with more diverse representation of involved cGVHD organ sites, lines of prior therapy, and specific types of prior therapy compared with the original phase 2 trial.

This analysis supports that ibrutinib has activity in SR-cGVHD with a total number of patients included on par with patients treated in other more recent major trials.^4,6 For comprehensive coverage, we examined several well-accepted measures of treatment benefit in cGVHD trials, including fixed 6-month NIH ORR (CR/PR), best ORR, and FFS. Our efficacy estimates reported here approach but do not fully recapitulate those reported in the original ibrutinib phase 2 trial (likely related to nonoverlapping eligibility/study populations),⁵ whereas ORR reported in the more recent pediatric trial appears higher (an expected difference in light of largely nonoverlapping age distribution).¹³ Additionally, responders achieved durable disease control, as previously reported.¹² As a point of reference, the observed 6-month fixed ORR of 45% is similar to that reported with ruxolitinib in second line in the prior REACH3 trial (49.7% ORR, comprising 43% PR and 6.7% CR).⁶ We explored potential predictors of treatment response and identified that liver involvement was most strongly predictive of 6-month ORR. Additionally, ORR appeared to worsen in those with greater prior lines of systemic therapy for cGVHD, also an expected finding. Of note, varied starting doses of ibrutinib were used (for drug interactions or clinician discretion), and this variation may have also affected the observed responses. We acknowledge that retrospective collection of serial NIH cGVHD organ scores is imperfect, and participating centers had varying levels of consistent capture of these scores prospectively at the point of care, with this concern further compounded by missing data further limiting the evaluable patient population in our study. The degree of missingness increased over time after ibrutinib start; hence, the evaluation of best ORR (in which most declared best ORR within 1-3 months after starting ibrutinib) was much less affected by this concern.

In contrast to the above limitations, we note the unambiguous nature of the FFS outcome, which is more robust to challenges in retrospective studies. The FFS observed here was heavily driven by the start of new lines of systemic IS therapy as expected, whereas death and relapse were uncommon. We note that the observed FFS is nearly identical to a prior large analysis of FFS in second-line cGVHD therapy,⁹ in which second-line therapy comprised primarily of agents such as MMF, tacrolimus, sirolimus, or ECP. We acknowledge FFS reported after other newer agents (eg, ruxolitinib and belumosudil) appears higher. However, we also acknowledge that this retrospective study and these more recent prospective trials are not directly comparable for many reasons, including the less stringent patient selection in this real-world retrospective analysis. This analysis suggests patients treated with ibrutinib for SR-cGVHD can enjoy prolonged FFS, and 6-month NIH ORR stratifies patients into distinct subsequent FFS trajectories. Intuitively, those with 6-month ORR had significantly improved subsequent FFS. We also identified that older age, lung involvement, and higher baseline prednisone dose were associated with worse FFS. These factors may help shape patient selection to identify those with greater likelihood of long-term benefit from the therapy. Additionally, we note that FFS of this agent in an even more contemporary population may have lower FFS, given the larger range of available cGVHD treatment options (hence, greater opportunity for treatment change).

The study also provides important safety data in a large patient population after FDA approval. In total, the observed safety profile largely recapitulates that seen in the prior trials, as well as the overall compendium of safety data considering ibrutinib-treated patients across various indications.^19-27 Dose reduction events were less common and largely driven by myalgia, GI toxicity, bleeding/bruising, cytopenias, and fatigue. However, drug discontinuation events were quite common and mostly due to either progressive cGVHD or treatment toxicity. The toxicities leading to drug discontinuation were largely overlapping with the above dose reduction reasons, yet it also included infection and cutaneous toxicities (rash and nodules). Of note, the observed rate of cardiac toxicities (eg, cardiac arrhythmia or decreased cardiac ejection fraction) leading to discontinuation were low (including when considering the included age range) and on par with expected rates from prior analyses.

As limitations, we note the following. First, there are limitations to retrospective data collection, most notably the capture and accuracy of serial NIH cGVHD organ scores. We have addressed this through the above methods, in which we considered multiple efficacy measures (ORR, best ORR, and FFS) and reported a range of possible response outcomes considering fixed ORR in only evaluable patients vs treating nonevaluable patients as nonresponders. Given this, we have also not emphasized conclusions on organ-specific responses; however, we note (supplemental Data) some evidence for improved best ORR in mouth, GI, and liver organ sites. These data provide some initial estimate of efficacy in these subgroups, yet further study is clearly needed. Second, we note that some patient characteristics were better represented than others, and we notably have limited pediatric representation in this study. We note that a dedicated pediatric trial has now been completed since FDA approval, hence filling this evidence gap directly. Third, we note that this study does not provide definitive evidence-based guidance on where to position ibrutinib in the sequence of available cGVHD treatments, and ibrutinib use in this study covered a broad range of prior lines of therapy. Outside of following the established FDA approvals for each currently approved agent, uncertainty remains regarding practical application of agents to individual patients to optimize treatment response. Acknowledging ruxolitinib is now considered standard in second-line cGVHD therapy⁶ and belumosudil for third-line onward⁴ (with recent approval of axatilimab)²⁸ and the total evidence for ibrutinib (ie, prior phase 2 trial and current analysis), we suggest that the use of ibrutinib in earlier lines of therapy among patients with liver/GI/mouth/erythematous skin involvement would likely represent the best use of this agent to optimize response.

In total, this real-world analysis provides practice patterns for ibrutinib in SR-cGVHD after its FDA approval and provides important efficacy and safety data to guide ongoing use.

Contribution: All authors have meaningfully contributed to study design, data collection, analysis, interpretation, and writing of this manuscript, including approval of the final version of this manuscript.

Conflict-of-interest disclosure: J.P. reports consulting and advisory board membership fees from Syndax, CTI BioPharma, Amgen, Regeneron, and Incyte; and clinical trial support from Novartis, Amgen, Takeda, Janssen, Johnson and Johnson, Pharmacyclics, CTI BioPharma, and Bristol Myers Squibb. M.E.D.F. reports research funding from Pharmacyclics, Novartis, and Incyte. Z.D. reports research support from Incyte Corp, REGiMMUNE Corp, and Taiho Oncology, Inc; consulting fees from Sanofi, Incyte Corp, Inhibrx, PharmaBiome AG, Ono Pharmaceutical, REGiMMUNE Corp, MaaT Pharma, and Forte Biosciences Inc. I.P. reports advisory board membership fees from Incyte, Sanofi, and Syndax. M.Q. reports consultancy fees from Novartis and Vertex. A.S. reports research funding from Merck and Novartis; research product support from Clasado, DSM/iHealth, and Blue Spark Technologies; and consulting fees from TargaZyme, Acrotech, Geron, and Janssen. H.R. serves as the medical monitor for the National Marrow Donor Program CD33 CAR-T trial; and reports advisory board fees from Medexus Pharmaceuticals. W.H. reports advisory board membership fees from Nkarta, Sanofi, Incyte, Rigel, and MaaT Pharma; consultancy fees from ACI Group and Therakos/Mallinckrodt; data and safety monitoring board membership fees from Angiocrine; and adjudication committee fees from CSL Behring. M.H. reports research funding from ADC Therapeutics, Spectrum Pharmaceuticals, and Astellas; consultancy fees from ADC Therapeutics, CRISPR, Bristol Myers Squibb, Kite, AbbVie, Caribou, Genmab, Autolus, and Forte Biosciences; and speaker's bureau fees from ADC Therapeutics, AstraZeneca, BeiGene, Kite, DMC Inc, Genentech, Myeloid Therapeutics, and CRISPR. N.F. reports advisory board and speaker fees from Incyte; DSMB membership fees from Chronic GVHD Consortium; and medical monitor duties Blood and Marrow Transplant Clinical Trials Network. C.K. reports consultancy fees from Incyte and Horizon Therapeutics; and advisory board fees from Incyte. E.N. reports consultancy fees from Medexus Pharmaceuticals and Novartis. The remaining authors declare no competing financial interests.

Correspondence: Joseph Pidala, Blood and Marrow Transplantation and Cellular Immunotherapy. H. Lee Moffitt Cancer Center and Research Institute, 12902 Magnolia Dr, Tampa, FL 33612; email: Joseph.pidala@moffitt.org.