Evaluation of Novel Endpoints As Treatment Outcome in Allogeneic Hematopoietic Stem Cell Transplantation: A Comparison of "Classical" and "Refined" Definition of Graft-Versus Host Disease Free and Relapse Free Survival for Registry-Based Studies

Allogeneic hematopoietic stem cell transplantation (HSCT) is often associated with major complications resulting in poor outcome, including acute and chronic graft-versus-host disease (GVHD), relapse and death. Holtan et al. (Blood 2015) proposed GVHD free/relapse free survival (GRFS), as a composite endpoint to analyse transplant associated morbidity. This endpoint considers as events grades III-IV acute GVHD, systemic therapy requiring chronic GVHD, primary disease relapse, or death for any cause. Although it could be a good indicator in prospective trials, since it seems to characterize the survival without mortality or ongoing morbidity, it might be difficult to use in retrospective setting. A "refined" GRFS (considering grade III-IV acute GVHD, extended/severe chronic GVHD, relapse and death for any cause) was proposed by Ruggeri et al (BMT 2016) to adapt this endpoint also to registry-based studies.

To compare definitions of this composite endpoint we retrospectively analyzed "classical" and "refined" GRFS on 959 consecutive patients receiving HSCT at Federico II University in Naples (n=119) and Saint-Louis Hospital in Paris (n=840) between 2007 and 2014. Data collection is based on the European Group of Blood and Marrow Transplantation (EBMT) registry via the central data processing system "Project Manager Internet Server" (ProMISe), considering the MED-A and MED-B forms for patients registered in those two centers. A MED-C form was built for all patients who developed chronic GVHD.

Median age of patients at transplant was 41 (range 4-64 years) whereas median duration of follow-up for survivors after HSCT was 38 months (range: 0-97.7 months). Characteristics of patients were reported in table 1. Cumulative incidence (CI) at day 100 of grade II-IV and grade III-IV acute GVHD were 41.1% (95%CI 37.9-44.3%) and 15.6% (95%CI 13.4-18), respectively. CI of chronic GVHD requiring systemic treatment and diagnosed according to NIH criteria was 22.3% (95%CI 20.7 -26.3) at 1 year and 32.6 % (95%CI 29.5-36.0) at 3 years. CI of severe/extended chronic GVHD (as reported in ProMise registry) was estimated at 16.6% (95%CI 14.1-19.5) at 1 year and 22.5% (95%CI 19.4-26.0) at 3 years. CI of relapse was 18.7% (95%CI 16.3-21.5%) at 1 year, 25.1% (95%CI 22.3-28.2%) at 3 years and 26.7 (95%CI 23.7-30.0) at 5-years. Probability of survival (OS) of the whole population at 1, 3 and 5 years was respectively: 75% (95%CI 73.0-76.9), 62% (95%CI 58.0-65.8), and 57% (95%CI 55.0-58.9). Event free survival (EFS) was 64% (95%CI 48.0-56.9) at 1 year, 53% at 3 years (95%CI 49.1-56.9) and 50% at 5 years (95%CI 48.0-52.0). Global classical GRFS was 71% (95%CI, 69.0-72.9) at 1 years, 52% (95%CI 48.1-55.9) at 3 years and 35% (95%CI 31.2-39.0) at 5 years, whereas refined GRFS was 71% (95%CI, 69.0-72.9), 53% (95%CI 49.1-56.9) and 38% (95%CI 34.2-42.0) at 1, 3, and 5 years respectively. Six hundred twenty-eight subjects were censored by classical GRFS whereas 546 patients have been identified as reporting at last one event according to refined GRFS. Chronic GVHD and relapse represented the major components of classical GRFS (35.7 and 35.2% respectively) whereas relapse (40.5%) and III-IV acute GVHD (25.6%) were major constituents of refined GRFS (Figure 2). According both definitions several risk factors were identified in univariate and multivariate analysis. Particularly, transplants performed after 2010, diagnosis of acute leukemia and RIC regimens had a significant impact on survival without GVHD and relapse (both in classical and in refined GRFS) whereas age, stem cell source, disease risk group (identified as influencing factors in univariate analysis) did not impacted on GRFS in multivariate model. Transplants from non-matched unrelated donors showed a significant impact only on classical GRFS in COX regression. The multivariable c-statistics were 0.609 for classical and 0.614 for refined GRFS.

Our data support the use of both composite endpoints to describe HSCT outcomes. Classical GRFS could represent the ideal endpoint following HSCT in prospective setting, whereas refined GRFS could be used as good surrogate of GRFS in retrospective studies. However, attention need to be paid concerning the use of refined indicator because of the risk of underestimation of censored events, since it may not reflect the real impact of chronic GVHD on population in study.

View largeDownload slide

View largeDownload slide

Close modal