Refractory Graft-Versus-Host Disease–Free, Relapse-Free Survival as an Accurate and Easy-to-Calculate Endpoint to Assess the Long-Term Transplant Success

Accepted Manuscript Title: Refractory Graft-Versus-Host Disease-Free, Relapse-Free Survival as an Accurate and Easy-to-Calculate Endpoint to Assess the Long-Term Transplant Success Author: Koji Kawamura, Hideki Nakasone, Saiko Kurosawa, Kazuki Yoshimura, Yukiko Misaki, Ayumi Gomyo, Jin Hayakawa, Masaharu Tamaki, Yu Akahoshi, Machiko Kusuda, Kazuaki Kameda, Hidenori Wada, Yuko Ishihara, Miki Sato, Kiriko Terasako-Saito, Misato Kikuchi, Shun-ichi Kimura, Aki Tanihara, Shinichi Kako, Heiwa Kanamori, Takehiko Mori, Satoshi Takahashi, Shuichi Taniguchi, Yoshiko Atsuta, Yoshinobu Kanda PII: DOI: Reference:

S1083-8791(18)30082-X https://doi.org/10.1016/j.bbmt.2018.02.004 YBBMT 55031

To appear in:

Biology of Blood and Marrow Transplantation

Received date: Accepted date:

7-12-2017 7-2-2018

Please cite this article as: Koji Kawamura, Hideki Nakasone, Saiko Kurosawa, Kazuki Yoshimura, Yukiko Misaki, Ayumi Gomyo, Jin Hayakawa, Masaharu Tamaki, Yu Akahoshi, Machiko Kusuda, Kazuaki Kameda, Hidenori Wada, Yuko Ishihara, Miki Sato, Kiriko TerasakoSaito, Misato Kikuchi, Shun-ichi Kimura, Aki Tanihara, Shinichi Kako, Heiwa Kanamori, Takehiko Mori, Satoshi Takahashi, Shuichi Taniguchi, Yoshiko Atsuta, Yoshinobu Kanda, Refractory Graft-Versus-Host Disease-Free, Relapse-Free Survival as an Accurate and Easy-toCalculate Endpoint to Assess the Long-Term Transplant Success, Biology of Blood and Marrow Transplantation (2018), https://doi.org/10.1016/j.bbmt.2018.02.004. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Refractory graft-versus-host disease-free, relapse-free survival as an accurate and easy-to-calculate endpoint to assess the long-term transplant success

Koji Kawamura1, Hideki Nakasone1, Saiko Kurosawa2, Kazuki Yoshimura1, Yukiko Misaki1, Ayumi Gomyo1, Jin Hayakawa1, Masaharu Tamaki1, Yu Akahoshi1, Machiko Kusuda1, Kazuaki Kameda1, Hidenori Wada1, Yuko Ishihara1, Miki Sato1, Kiriko Terasako-Saito1, Misato Kikuchi1, Shun-ichi Kimura1, Aki Tanihara1, Shinichi Kako1, Heiwa Kanamori3, Takehiko Mori4, Satoshi Takahashi5, Shuichi Taniguchi6, Yoshiko Atsuta7,8, Yoshinobu Kanda1,9

1

Division of Hematology, Saitama Medical Center, Jichi Medical University, Saitama, Japan;

2

Department of Hematopoietic Stem Cell Transplantation, National Cancer Center Hospital, Tokyo,

Japan; 3Department of Hematology, Kanagawa Cancer Center, Yokohama, Japan; 4Division of Hematology, Department of Medicine, Keio University School of Medicine, Tokyo, Japan; 5

Department of Hematology and Oncology, The Institute of Medical Science, The University of

Tokyo, Tokyo, Japan; 6Department of Hematology, Toranomon Hospital, Tokyo, Japan; 7Japanese Data Center for Hematopoietic Cell Transplantation, Nagoya, Japan; 8Department of Healthcare Administration, Nagoya University Graduate School of Medicine, Nagoya, Japan; 9Division of Hematology, Department of Medicine, Jichi Medical University, Shimotsuke, Japan.

1 Page 1 of 23

Correspondence: Yoshinobu Kanda, M.D., Ph.D. Division of Hematology, Saitama Medical Center, Jichi Medical University 1-847, Amanuma-cho, Omiya-ku, Saitama-city, Saitama 330-8503, Japan TEL: +81-48-647-2111, FAX: +81-48-644-8167, E-mail: [email protected]

Running Head: Refractory GRFS as a simple and accurate endpoint Overall word count for the manuscript: 2681 Abstract word count: 240 Numbers of tables and figures: 2 tables and 4 figures Number of references: 15 Keywords: allogeneic hematopoietic stem cell transplantation; acute GVHD; chronic GVHD; quality of life; endpoint.

Highlights 

We developed a new endpoint, refractory graft-versus-host disease-free, relapse-free survival (rGRFS).



We validated current GRFS (cGRFS) using two distinct methods.



cGRFS and rGRFS more accurately reflect transplant success than conventional GRFS.

2 Page 2 of 23



The curves of cGRFS and rGRFS overlapped after the first two years of post-transplant follow-up.



rGRFS can be easily calculated and analyzed with widely-used statistical approaches.

Abstract The aim of this study was to develop a new composite endpoint that accurately reflects the long-term success of allogeneic hematopoietic stem cell transplantation (allo-HCT), since the conventional graft-versus-host disease-free, relapse-free survival (GRFS) overestimates the impact of GVHD. First, we validated current GRFS (cGRFS), which recently was proposed as a more accurate endpoint of long-term transplant success. cGRFS was defined as survival without disease relapse/progression or active chronic GVHD at a given time after allo-HCT, calculated using two distinct methods; a linear combination of a Kaplan-Meier estimates approach and a multistate modelling approach. Next, we developed a new composite endpoint, refractory GRFS (rGRFS). rGRFS was calculated similarly to conventional GRFS treating grade III-IV acute GVHD, chronic GVHD requiring systemic treatment, and disease relapse/progression as events, except that GVHD that resolved and did not require systemic treatment at the last evaluation was excluded as an event in rGRFS. The two cGRFS curves obtained using two different approaches were superimposed and both were superior to that of conventional GRFS, reflecting the proportion of patients with resolved chronic GVHD. Finally, the curves of cGRFS and rGRFS overlapped after 3 Page 3 of 23

the first two years of post-transplant follow-up. These results suggest that cGRFS and rGRFS more accurately reflect transplant success than conventional GRFS. Especially, rGRFS can be more easily calculated than cGRFS and analyzed with widely used statistical approaches, whereas cGRFS more accurately represents the burden of GVHD-related morbidity in the first two years after transplantation.

4 Page 4 of 23

Introduction Allogeneic hematopoietic stem cell transplantation (allo-HCT) has been established as a curative therapy for high-risk hematologic malignancies. Although prevention of disease relapse and transplant-related mortality (TRM) is absolutely imperative to achieve successful transplantation, quality of life (QOL) after allo-HCT should also be taken into account. Unfortunately, no existing endpoint accurately reflects all of these points, since QOL data are usually not readily available in transplant databases. Recently, graft-versus-host disease-free, relapse-free survival (GRFS), defined as the absence of grade III-IV acute graft-versus-host disease (GVHD), chronic GVHD requiring systemic treatment, relapse, or death, was proposed as a novel composite endpoint for clinical trials evaluating GVHD prophylaxis after allo-HCT [1]. GRFS is now widely used to evaluate the success of allo-HCT, since chronic GVHD is associated with impaired QOL [2-5]. However, this conventional GRFS treats GVHD as a fixed failure event, similar to relapse or death, even though GVHD may be resolved by treatment. Therefore, conventional GRFS apparently overestimates the impact of GVHD on the outcome of allo-HCT. Very recently, Solomon et al. proposed a new dynamic composite endpoint, termed current GRFS (cGRFS), defined as survival without relapse and active moderate to severe chronic GVHD at the time of most recent assessment [6]. They showed that nearly half of allo-HCT recipients were considered to be cured without active GVHD when calculated by cGRFS, whereas only one-fourth of the recipients achieved success when evaluated by conventional GRFS. Therefore, cGRFS might be a better 5 Page 5 of 23

measurement of transplant success than conventional GRFS, although statistical method for calculating cGRFS is complex. In this study, we aimed to validate and to further explore cGRFS, and develop a more practical easy-to-calculate endpoint, termed refractory GRFS (rGRFS), which was expected to provide a similar estimate as cGRFS without a complex statistical analysis.

6 Page 6 of 23

Methods Study design and definitions The major objective of this study was to develop a new composite endpoint that more accurately reflects transplant success. First, we calculated and compared cGRFS using two distinct methods, as described in the Statistical analysis section. In this estimation, we did not include grade III-IV acute GVHD as an event, since acute GVHD status was considered to transition to either resolved GVHD, active chronic GVHD, or death. However, for greater certainty, we compared the estimates obtained with this approach to those obtained while treating grade III-IV acute GVHD as an event (cGRFS2). Next, we compared cGRFS with conventional GRFS and disease-free survival (DFS). Finally, we developed rGRFS, which is an easy-to-calculate composite endpoint, and compared rGRFS with cGRFS. cGRFS was defined as survival without active chronic GVHD requiring systemic treatment or disease relapse/progression at any time after transplantation. cGRFS2 was similar to cGRFS except that grade III-IV acute GVHD was included as an event. In the calculation of rGRFS, grade III-IV acute GVHD, chronic GVHD requiring systemic treatment, and disease relapse/progression were regarded as events as in conventional GRFS, but GVHD that resolved and did not require systemic treatment at the last evaluation was not considered an event in rGRFS. Although grade III-IV acute GVHD that did not resolve at the last evaluation was regarded as an event, it was treated as the transition to chronic GVHD at day 100 in case that it evolved into chronic GVHD. Active acute GVHD was defined as grade III-IV acute GVHD that was not resolved 7 Page 7 of 23

despite treatment, and active chronic GVHD was defined as chronic GVHD that was not resolved despite systemic treatment, or for which immunosuppressive drugs were continuously administered. Resolved chronic GVHD was defined as the absence of any clinically active symptoms without immunosuppressive drugs. Acute and chronic GVHD were graded according to previously published criteria [7, 8], but the National Institutes of Health consensus criteria was not utilized in chronic GVHD grading [9]. DFS was defined as survival without disease progression or relapse. Acute leukemia in the first or second remission, chronic myeloid leukemia in first and second chronic phase, myelodysplastic syndrome, Hodgkin and non-Hodgkin lymphoma in complete or partial chemotherapy-sensitive remission were defined as low-risk disease, whereas others were defined as high-risk disease. Patients for conventional, current, and refractory GRFS analyses Three hundred fifteen patients who underwent their first allo-HCT between 2007 and 2016 at our institution were included in this study. Clinical data for these patients were obtained from our database and the clinical charts. This study was approved by the Institutional Review Board of Saitama Medical Center, Jichi Medical University. Statistical analysis The probabilities of GRFS, DFS, and rGRFS were estimated according to the Kaplan-Meier method. cGRFS was estimated by the following two statistical approaches, based on a multistate model including 10 health states (cGRFS) or 14 health states (cGRFS2) that a 8 Page 8 of 23

patient may be in after allo-HCT (Figure 1) [6, 10]. The states in the cGRFS analysis are as follows, and cGRFS is the probability of being in state 0, 4, or 8. State 0. Alive and no events. State 1. Dead or relapse before first chronic GVHD. State 2. Alive in remission with first chronic GVHD requiring systemic treatment. State 3. Dead or relapse before resolution of first chronic GVHD. State 4. Alive in remission after resolution of first chronic GVHD and before second chronic GVHD. State 5. Dead or relapse after resolution of first chronic GVHD and before second chronic GVHD. State 6. Alive in remission with second chronic GVHD requiring systemic treatment. State 7. Dead or relapse before resolution of second chronic GVHD. State 8. Alive in remission after resolution of second chronic GVHD. State 9. Dead or relapse after resolution of second chronic GVHD. The first statistical approach was based on a linear combination of five (cGRFS) or seven (cGRFS2) Kaplan-Meier estimates. The probability of being in State 0, S1(t), can be calculated with Kaplan-Meier estimates treating any death or relapse without the development of chronic GVHD, and first chronic GVHD as events. The probability of being in State 4 can be calculated by the following formula; S2(t) – S3(t), where S2(t) (probability of being in State 0, 2, or 4) was 9 Page 9 of 23

calculated by treating any death or relapse before second chronic GVHD, and second chronic GVHD as events, and S3(t) (probability in State 0 or 2) was calculated by treating any death or relapse before resolution of first chronic GVHD, and resolution of first chronic GVHD as events. Similarly, the probability of being in State 8 was calculated by the following formula; S 4(t) – S5(t), where S4(t) reflects the probability of being in State 0, 2, 4, 6, or 8 and S5(t) reflects the probability of being in State 0, 2, 4, or 6. Finally, cGRFS can be calculated by the formula; S1(t) + [S2(t) – S3(t)] + [S4(t) – S5(t)]. In a similar way, cGRFS2 was calculated as described in the Supplemental Methods. A second approach was based on a multistate model with a Markov assumption that the transition among states depends only on the current state[11]. The empirical transition matrix was calculated along with the covariate estimator described by Anderson et al [12]. The model is based on the same 10 health states as described above (Figure 1), and cGRFS was calculated as the probability of being in State 0, 4, or 8 after allo-HCT given that the patient was in State 0 at the time of allo-HCT. Categorical variables were compared among groups with the χ2-test or Fisher's exact test, and continuous variables were compared among three groups with a one-way analysis of variance or Kruskal-Wallis test with post-hoc multiple comparison using Tukey’s honestly significant difference (HSD) test. Multivariate analysis for rGRFS was performed using the Cox proportional hazards model. The following variables were considered: the patient’s age at transplantation (<50 10 Page 10 of 23

years or ≥50 years), patient sex (male or female), disease risk (low or high), Eastern Cooperative Oncology Group Performance Status (ECOG PS) (0-1 or 2-4), Hematopoietic Cell Transplantation-Specific Comorbidity Index (HCT-CI) (0, 1-2, or ≥3) [13], donor type (matched related, mismatched related, matched unrelated, mismatched unrelated, or unrelated cord blood), conditioning regimen (myeloablative or reduced-intensity), use of in vivo T-cell depletion (yes or no), GVHD prophylaxis (cyclosporine-based or tacrolimus-based), and year of transplantation (2007-2011 or 2012-2016). All P-values were two sided and P-values <0.05 were considered statistically significant. All statistical analyses were performed with EZR (Saitama Medical Center, Jichi Medical University) [14, 15], which is a graphical user interface for R (The R Foundation for Statistical Computing, version 3.0.2, Vienna, Austria). More precisely, it is a modified version of R commander (version 2.0-3) that was designed to add statistical functions that are frequently used in biostatistics. In the calculation of cGRFS, the “currentSurvival” and “etm” packages were additionally used for a linear combination of a Kaplan-Meier estimates approach and a multistate modelling approach, respectively.

11 Page 11 of 23

Results Patient characteristics for GRFS analyses The baseline characteristics of 315 patents who were included in the following GRFS analyses are shown in Table 1. The median age at transplantation was 46 years (range, 15 to 68 years), and several diseases were included in this cohort. Donor type was 6/6 (HLA-A, -B, and -DRB1 loci) matched related donor in 21.9%, mismatched related donor in 16.2%, 6/6 matched unrelated donor in 34.3%, mismatched unrelated donor in 20.3%, and unrelated cord blood in 7.3%. GVHD prophylaxis was cyclosporine-based in 91.1% and tacrolimus-based in 8.9%, and antithymocyte globulin (ATG) or alemtuzumab was used in 8.9%. Bone marrow and peripheral blood graft were used almost exclusively in unrelated allo-HCT and related allo-HCT, respectively (95% and 97%), and therefore, the stem cell source was not included in the analyses. Current GVHD-free, relapse-free survival cGRFS was calculated by two approaches, a combination of a Kaplan-Meier estimates approach and a multistate modelling approach. We confirmed that the two cGRFS curves were approximately equal (Figure 2A). Figure 2B shows the proportions of patients in each state. The probability of being in State 2 (i.e., survival with first chronic GVHD requiring systemic treatment) reached a peak at about 1 year after allo-HCT and then gradually decreased. The proportion of patients in State 6, 7, 8, or 9 was very small, which justified omission of these states after the development of third or later chronic GVHD in the multistate model. In the following analysis, we 12 Page 12 of 23

used a combination of a Kaplan-Meier estimates approach to estimate cGRFS because of the ease of calculation. Comparison of cGRFS, conventional GRFS, and DFS cGRFS, conventional GRFS, and DFS are shown in Figure 3A. The cGRFS rates, conventional GRFS rates, and DFS rates were 33.3%, 28.9%, and 56.4% at 1 year, and 35.3%, 22.1%, and 47.4% at 3 years after allo-HCT, respectively (Table 2). The difference between DFS and cGRFS corresponds to the probability of being in State 2 in Figure 2B, and the difference between cGRFS and GRFS corresponds to the probability of being in State 4. To confirm that it was appropriate to omit the development of grade III-IV acute GVHD as an event, we compared cGRFS and cGRFS2. As shown in Figure 3B, the two curves completely overlapped after day 100. Therefore, acute GVHD was not necessary to assess the long-term success of allo-HCT. Refractory GVHD-free, relapse-free survival Finally, we evaluated an easy-to-calculate endpoint, termed rGRFS, as a substitute for cGRFS, which requires a complex calculation. rGRFS can be calculated using the ordinary Kaplan-Meier method, and the only difference between rGRFS and conventional GRFS is the exclusion of GVHD that resolved and did not require systemic treatment at the last evaluation from the events in rGRFS. As shown in Figure 4A, the two curves of rGRFS and cGRFS are superimposed, except for the first two years after allo-HCT. The differences between the 13 Page 13 of 23

probabilities of rGRFS and cGRFS were 8.0%, 0.9%, and 2.1% at 1, 3, and 5 years, respectively, which were much smaller than those between the probabilities of cGRFS and conventional GRFS (Table 2). In addition, we confirmed that the difference in the cumulative area under the curve (AUC) between rGRFS and cGRFS became stable 2 years after allo-HCT, whereas the difference between cGRFS and GRFS increased year by year (Table 2). In a multivariate analysis using Cox proportional hazard modeling, recipient age (50 years of age or older; hazard ratio (HR) 1.57, 95% confidence interval (CI) 1.14-2.17, P=0.0062), recipient sex (male; HR 1.46, 95% CI 1.03-2.05, P =0.032), and disease risk (high-risk; HR 2.57, 95% CI 1.83-3.60, P <0.0001) were independent significant prognostic factors for inferior rGRFS (Supplemental Table S1). In a subgroup analysis using these factors, the curves of rGRFS and cGRFS also mostly overlapped (Figure 4 B-D).

14 Page 14 of 23

Discussion The success of allo-HCT requires not only the absence of death and disease relapse, but also a good QOL status. Chronic GVHD is a major cause of morbidity and mortality after allo-HCT and occurs in approximately 30-70% of the recipients. In addition, it impairs QOL in long-term survivors [2-5]. To reflect all of these events in a single survival curve, GRFS is widely used in clinical studies [1]. However, conventional GRFS treats the development of GVHD as a fixed event similar to relapse or death, regardless of the treatment response. This is the major drawback of GRFS, since chronic GVHD has a variety of clinical presentations and its response to treatment also differs among patients. In fact, Fraser et al. reported that active chronic GVHD had an adverse impact on many aspects of the overall health status of allo-HCT, and resolved chronic GVHD resulted in a health status similar to that in patients without chronic GVHD [2]. Therefore, treating the development of GVHD as a fixed event must result in an overestimation of the impact of GVHD on the outcome of allo-HCT. In this study, first, we validated cGRFS, which is defined as the absence of disease relapse/progression, death, or active chronic GVHD requiring systemic treatment at a given point in time after allo-HCT, treating chronic GVHD as a reversible event. This analysis is similar to the current leukemia-free survival that treats relapse as a reversible event, since relapsed leukemia may be re-induced to remission by donor lymphocyte infusion [10, 11]. We calculated cGRFS by two distinct methods, a combination of a Kaplan-Meier estimates approach and a multistate 15 Page 15 of 23

modelling approach, and proved that the two estimates were identical. Very recently, Solomon et al. reported an investigation of cGRFS using a method that was very close to our former approach, with a small difference in the definition of resolved chronic GVHD [6]. Our study validated their findings and, in addition, confirmed that the results were robust within two distinct statistical methods and within the presence or absence of acute GVHD as an event. However, both statistical methods for calculating cGRFS are complex, and conventional statistical tests, such as the log-rank test and Cox proportional hazard modeling, cannot be applied to cGRFS. Therefore, we developed a more practical alternative endpoint, called rGRFS. rGRFS can be calculated by an ordinary time-to-event analysis, including the same events as in conventional GRFS, but avoids overestimation of the impact of GVHD by excluding resolved GVHD from the events. Although there was a small difference between the cGRFS and rGRFS curves in the early post-transplantation period, this difference disappeared approximately 2 years after allo-HCT. rGRFS can be calculated using readily available data in a transplant registry and ordinary statistical software that enables univariate and multivariate analyses. Therefore, rGRFS may be a more applicative and useful measurement of transplant success than cGRFS in various clinical studies of allo-HCT. A major shortcoming of rGRFS is the discrepancy from cGRFS within two years after allo-HCT, which reflects the proportion of patients with active chronic GVHD. Therefore, this discrepancy increases when there is a higher incidence of chronic GVHD. rGRFS should be 16 Page 16 of 23

applied to a patient population with at least two years median follow-up duration after allo-HCT. In addition, we should note that rGRFS overestimates patient success within the early period after allo-HCT. In conclusion, we propose rGRFS as a more accurate endpoint for long-term transplant success. Especially, rGRFS can be applied in various clinical studies, because it does not require complicated statistical analyses. The validation of rGRFS in a larger cohort, with QOL data if possible, is warranted to confirm its usefulness as an endpoint for the outcome of allo-HCT.

17 Page 17 of 23

Acknowledgments Financial disclosure: The authors have nothing to disclose.

Conflicts of interest statement: The authors have declared no conflicts of interest.

Authorship contribution statement: K.Kawamura designed the study, collected the patient data, performed the statistical analysis, and wrote the manuscript; H.N. and Y. K. designed the study, performed the statistical analysis, and wrote the manuscript; S.Kurosawa, H.K., T.M., S.Takahashi, S.Taniguchi, and Y.Atsuta interpreted the data; K.Y., Y.M., A.G., J.H., M.T., Y.Akahoshi, M.Kusuda, K.Kameda, H.W., Y.I., M.S., K.T-S., M.Kikuchi., S.Kimura, A.T., and S.Kako provided the patient data; and all of the authors reviewed and approved the final manuscript.

18 Page 18 of 23

References [1] Holtan SG, DeFor TE, Lazaryan A, Bejanyan N, Arora M, Brunstein CG, et al. Composite end point of graft-versus-host disease-free, relapse-free survival after allogeneic hematopoietic cell transplantation. Blood. 2015;125:1333-8. [2] Fraser CJ, Bhatia S, Ness K, Carter A, Francisco L, Arora M, et al. Impact of chronic graft-versus-host disease on the health status of hematopoietic cell transplantation survivors: a report from the Bone Marrow Transplant Survivor Study. Blood. 2006;108:2867-73. [3] Lee SJ, Kim HT, Ho VT, Cutler C, Alyea EP, Soiffer RJ, et al. Quality of life associated with acute and chronic graft-versus-host disease. Bone Marrow Transplant. 2006;38:305-10. [4] Pidala J, Kurland B, Chai X, Majhail N, Weisdorf DJ, Pavletic S, et al. Patient-reported quality of life is associated with severity of chronic graft-versus-host disease as measured by NIH criteria: report on baseline data from the Chronic GVHD Consortium. Blood. 2011;117:4651-7. [5] Kurosawa S, Oshima K, Yamaguchi T, Yanagisawa A, Fukuda T, Kanamori H, et al. Quality of life after allogeneic hematopoietic cell transplantation according to affected organ and severity of chronic graft-versus-host disease. Biol Blood Marrow Transplant. 2017;23:1749-58. [6] Solomon SR, Sizemore C, Zhang X, Ridgeway M, Solh M, Morris LE, et al. Current Graft-versus-Host Disease-Free, Relapse-Free Survival: A dynamic endpoint to better define efficacy after allogenic transplant. Biol Blood Marrow Transplant. 2017;23:1208-14. [7] Przepiorka D, Weisdorf D, Martin P, Klingemann HG, Beatty P, Hows J, et al. 1994 Consensus 19 Page 19 of 23

Conference on Acute GVHD Grading. Bone Marrow Transplant. 1995;15:825-8. [8] Sullivan KM, Agura E, Anasetti C, Appelbaum F, Badger C, Bearman S, et al. Chronic graft-versus-host disease and other late complications of bone marrow transplantation. Semin Hematol. 1991;28:250-9. [9] Filipovich AH, Weisdorf D, Pavletic S, Socie G, Wingard JR, Lee SJ, et al. National Institutes of Health consensus development project on criteria for clinical trials in chronic graft-versus-host disease: I. Diagnosis and staging working group report. Biol Blood Marrow Transplant. 2005;11:945-56. [10] Klein JP, Keiding N, Shu Y, Szydlo RM, Goldman JM. Summary curves for patients transplanted for chronic myeloid leukaemia salvaged by a donor lymphocyte infusion: the current leukaemia-free survival curve. Br J Haematol. 2000;109:148-52. [11] Klein JP, Szydlo RM, Craddock C, Goldman JM. Estimation of current leukaemia-free survival following donor lymphocyte infusion therapy for patients with leukaemia who relapse after allografting: application of a multistate model. Stat Med. 2000;19:3005-16. [12] Allignol A, Schumacher M, Beyersmann J. Empirical transition matrix of multi-state models: The etm package. Journal of Statistical Software. 2011;38:1-15. [13] Sorror ML, Maris MB, Storb R, Baron F, Sandmaier BM, Maloney DG, et al. Hematopoietic cell transplantation (HCT)-specific comorbidity index: a new tool for risk assessment before allogeneic HCT. Blood. 2005;106:2912-9. 20 Page 20 of 23

[14] Kanda Y. Investigation of the freely available easy-to-use software 'EZR' for medical statistics. Bone Marrow Transplant. 2013;48:452-8. [15] Kanda J. Scripts for TRUMP data analyses. Part II (HLA-related data): statistical analyses specific for hematopoietic stem cell transplantation. Int J Hematol. 2016;103:11-9.

Figure legends Figure 1. Possible transitions in a multistate model for current GRFS without treating acute GVHD as an event (cGRFS) (A) and current GRFS while treating acute GVHD as an event (cGRFS2) (B).

Figure 2. Comparison of two approaches to calculate current GRFS (A) and proportions of patients in each state (B).

Figure 3. Comparison of current GRFS (without treating acute GVHD as an event) and conventional GRFS (A), and between current GRFS with and without treating acute GVHD as an event (B).

Figure 4. Comparison of refractory GRFS and current GRFS, overall (A) and grouped according to independent risk factors (age (B), disease risk (C) and recipient sex (D)) for rGRFS.

21 Page 21 of 23

Table 1. Patient characteristics Characteristics

n=315

Median age, years (range) Sex, n (%) Male Female Disease, n (%) Acute myelogenous leukemia Acute lymphoblastic leukemia Chronic myelogenous leukemia myelodysplastic syndrome/Myeloproliferative neoplasms Non-Hodgkin’s lymphoma/ Hodgkin’s lymphoma Multiple myeloma/Plasma cell leukemia Severe aplastic anemia Others Disease risk, n (%) Low risk High risk NA ECOG PS, n (%) 0-2 3-4 HCT-CI, n (%) 0 1-2 ≥3 * Donor type, n (%) Matched related Mismatched related Matched unrelated Mismatched unrelated Unrelated cord blood Conditioning regimen, n (%) Myeloablative Reduced-intensity GVHD prophylaxis, n (%) Cyclosporine-based Tacrolimus-based Use of antithymocyte globulin/alemtuzumab, n (%) Yes No Year of transplantation, n (%) 2007-2011 2012-2016

46 (15-68) 193 (61.3%) 122 (38.7%) 131 (41.6%) 52 (16.5%) 7 (2.2%) 48 (15.2%) 43 (13.7%) 9 (2.9%) 18 (5.7%) 7 (2.2%) 170 (54.0%) 123 (39.0%) 22 (7.0%) 293 (93.0%) 22 (7.0%) 206 (65.4%) 69 (21.9%) 40 (12.7%) 69 (21.9%) 51 (16.2%) 108 (34.3%) 64 (20.3%) 23 (7.3%) 231 (73.3%) 84 (26.7%) 287 (91.1%) 28 (8.9%) 51 (16.2%) 264 (83.8%) 146 (46.4%) 169 (53.7%)

*

HLA compatibility was defined according to HLA-A, HLA-B, and HLA-DR loci. NA indicates not applicable; ECOG PS, Eastern Cooperative Oncology Group Performance Status; HCT-CI, Hematopoietic Cell Transplantation-Specific Comorbidity Index.

22 Page 22 of 23

Table 2. Point estimation and cumulative area under the curve for each estimation. Point estimation Year

DFS

rGRFS cGRFS cGRFS2

GRFS

1

0.564

0.411

0.333

0.331

0.289

2

0.512

0.368

0.346

0.349

0.242

3

0.474

0.343

0.353

0.352

0.221

4

0.464

0.339

0.357

0.357

0.216

5

0.453

0.334

0.353

0.355

0.211

Cumulative area under the curve Year

DFS

rGRFS

cGRFS cGRFS2

GRFS

1

255.3

216.4

198.6

197.7

181.9

2

449.8

357.5

320.4

318.7

277.8

3

630.6

487.6

449.3

447.5

362.7

4

801.8

612.3

579.4

577.6

442.6

5

967.6

734.2

707.1

705.1

519.7

DFS indicates disease-free survival; rGRFS, refractory graft-versus-host disease-free /relapse-free survival (GRFS); cGRFS, current GRFS.

23 Page 23 of 23