Comparison of Conditioning Regimens for Patients Undergoing Allogeneic Hematopoietic Stem Cell Transplant: A Systematic Review and Meta-Analysis

Abstract

Background Allogeneic hematopoietic stem cell transplantation (HSCT) has been used as a treatment for hematologic and lymphoid cancers since the 1960s, and continued improvements have been realized to refine HSCT as the best curative option for many patients. Healthy donor hematopoietic stem and progenitor cells can differentiate into all lineages of the lympho-hematopoietic system, giving rise to all the specialized cells of the blood and immune system. Prior to infusing donor cells, a patient’s own hematopoietic cells are weakened or destroyed and the host immune system suppressed and/or eliminated to facilitate donor engraftment through the use of a series of chemotherapeutic drugs and/or irradiation procedures called a conditioning regimen. A variety of conditioning regimens have been developed over the years to achieve complete destruction of the hematopoietic and immune systems (myeloablative regimens) or their partial destruction (non-myeloablative or reduced-intensity regimens). Myeloablative regimens are associated with high non-relapse mortality in the first 100 days after transplantation due to organ toxicity and/or complications of immunosuppression (e.g., serious infection) prior to hematopoietic engraftment and immune reconstitution from the donor cells. However, myeloablative regimens are associated with a lower risk of relapse of the underlying cancer due to greater cytoreduction of the malignant cells. In non-myeloablative regimens and reduced intensity regimens, the lower doses have less impact on residual malignant cells but remain highly immunosuppressive, facilitating a more gradual takeover of the hematopoietic and immune systems by donor cells. While this approach reduces the risk of serious organ toxicity and early death due to transplant-related complications and allows older patients with co-morbidities to undergo HSCT, patients are still heavily immune-suppressed and remain at high risk of infections, rejection, and graft-versus host disease. The curative potential of non-myeloablative and reduced intensity conditioning therapies relies more heavily on the ability of the new graft-derived immune system to eliminate any residual primary cancerous cells not eradicated by the conditioning regimen—the “graft-vs-tumour effect”—which remains challenging to fully characterize and predict. Thus, while reducing the risk of death due to non-relapse causes in the pre-engraftment period, these regimens may be associated with an increased risk of relapse of the underlying disease. The ideal conditioning regimen would demonstrate a balance in the risks of both non-relapse mortality and relapse of the primary cancer. Comparing the effectiveness of conditioning regimens is ideally addressed by randomized controlled trials (RCTs) given the many factors that can influence outcomes following HSCT. A systematic search of RCTs addressing conditioning regimens in HSCT will allow us to identify optimal strategies to improve patient outcomes, identify standard treatment arms that should be used in future studies, provide insight on clinical outcomes that should be reported, and may permit network meta-analysis to infer potential comparisons between treatments that have not been directly compared before. Objective addressed in this review: To compare the benefits (e.g., reduction in mortality and relapse) and harms (e.g., increased risk of HSCT-related conditions) of competing regimens used to condition patients prior to undergoing HSCT, and to establish a hierarchy of intervention strategies according to their efficacy and safety. Overview of Research Approach Using data from RCTs, networks of evidence of conditioning regimens were developed. For regimens that have never been directly compared in head-to-head trials, their effects can be compared using network meta-analysis to derive comparisons between therapies and to rank all regimens according to their 3

relative effects on the outcomes of interest (e.g., the risks of mortality, relapse, and the development of several HSCT-related conditions such as graft-versus-host disease (GVHD), veno-occlusive disease (VOD), and bronchiolitis obliterans). This approach allows us to understand the strength of the evidence supporting the various conditioning regimens. This is helpful to generate recommendations for the optimal management of HSCT patients and for the selection of the most appropriate control arm in future RCTs to leverage the existing foundation of evidence. Systematic Review Methods The databases Medline, PubMed, Embase, and the Cochrane Register of Controlled Trials were searched for randomized controlled trials of patients undergoing HSCT. Studies were included if patients underwent allogeneic HSCT in the treatment of hematologic neoplasias or benign disease and were randomly allocated to receive a conditioning regimen. Outcomes of interest included overall mortality, non-relapse mortality, relapse of underlying disease, risk of acute and chronic GVHD, and specific harms. We conducted Bayesian network meta-analyses to compare conditioning regimens for outcomes of interest, where feasible. All outcomes were analysed as binary endpoints, with summary comparisons between regimens reported as odds ratios with 95% credible intervals. For outcomes for which network meta-analysis was not possible, detailed narrative summaries have been provided. RESULTS Eighteen trials assessed 18 unique conditioning regimens in 2,361 total patients; represented regimens are listed in Table 1. Three trials were not included in our network meta-analyses because either (1) they were conducted strictly on aplastic anaemia patients (2 studies, n = 213) or (2) the conditioning regimens evaluated were not described in detail (i.e., “standard conditioning” with or without total lymphoid irradiation; 1 study, n = 235). A total of 1,913 patients in 15 trials were available for inclusion in network meta-analyses. Overall, there was substantial variability in patient populations with respect to age, underlying hematologic disease, disease risk of relapse/mortality, and transplant donor status (i.e., related vs. unrelated, matched vs. unmatched). Trial publication dates ranged from 1988–2015. Networks of treatments were often disconnected, thereby limiting the ability to compare all interventions encountered. ____________________________________________________________________________________ Interventions compared in trials of conditioning regimens for HSCT A) Available for network meta-analyses: • CY+TBI • BU+CY • BU+FLUD • ATG+BU+FLUD • FLUD+TBI • ATG+TLI • TBI • ETP+TBI • MELPH+TBI

B) Disconnected from networks and summarized narratively: • ALZ+MELPH+FLUD • ATG+THIO+MELPH+CY • mBU+CY • mBU+FLUD

C) Not included in network meta-analysis: • CY (aplastic anaemia patients only) • CY+ATG+FLUD (aplastic anaemia patients only) • CY+ATG (aplastic anaemia patients only) • “Standard conditioning regimen” • “Standard conditioning regimen”+TLI Abbreviations: ALZ = alemtuzumab; ATG = anti-thymocyte globulin; BU = busulfan; CY = cyclophosphamide; ETN = etanercept; FLUD = fludarabine; mBU+CY = BU+CY+ hydroxyurea+semustine+cytosine arabinoside; mBU+FLUD = BU+FLUD+hydroxyurea+ semustine+cytosine arabinoside; MELPH = melphalan; TBI = total body irradiation; TLI = total lymphoid irradiation ____________________________________________________________________________________ Results of network meta-analyses and narrative summaries In the networks analysed, most comparisons between conditioning regimens were informed only by indirect evidence (i.e. head-to-head trials were not available), and many of the direct comparisons were based on single studies with small numbers of patients. Thus, a sparse evidence base and considerable between-study heterogeneity in patient populations complicated targeting of findings from meta-analyses. Separate network meta-analyses for overall mortality were conducted at the follow-up times of 100 days, and 1, 2, 4, and 5 years. Regarding disease relapse, only studies following patients for a median of 2–3 years were included in the network. CY+TBI was considered the standard conditioning regimen for comparison purposes. Clinical interpretation of findings were as follows: • Overall mortality: 5 conditioning regimens were included in network meta-analyses at all follow-up times: CY+TBI, BU+CY, BU+FLUD, ETP+TBI, and MELPH+TBI. No statistically significant differences between regimens were identified at 100 days. However, CY+TBI demonstrated significantly reduced overall mortality at 1 year compared to BU+CY, and at 2 and 4 years, when compared to either BU+CY or BU+FLUD. Fewer patients contributed to overall survival outcomes estimated at 5 years of follow-up and no significant differences were found between regimens.

• Non-relapse mortality (NRM): Network meta-analyses could not be conducted due to reporting limitations on the competing risk of relapse. Narrative summaries demonstrated significantly reduced NRM at 1, 2, and 5 years for BU+FLUD compared to BU+CY in one study (Rambaldi, 2014), but not in 2 other studies that compared the same regimens (Lee, 2013 and Liu, 2013). At 1 year of follow-up, there was a trend toward FLUD+TBI offering reduced NRM compared to CY+TBI in one study (Bornhauser, 2011). Additionally, early and late (7 year) cumulative incidence of NRM was significantly reduced for CY+TBI when compared to BU+CY; however, when adjusted for early vs. advanced disease and donor age > or < 30 years in a multivariable model, no significant difference was found between the regimens (Ringden, 1999). • Relapse: Four regimens provided data for a network meta-analysis of disease relapse at 2–3 years post-transplant (CY+TBI, BU+CY, BU+FLUD, and FLUD+TBI). There were no significant differences between regimens with respect to their effects on the risk of relapse.

• Risk of acute and chronic GVHD: All 9 regimens available (Table 1) were included in the network meta-analysis for aGVHD, while only 3 regimens (CY+TBI, BU+CY, and BU+FLUD) could be included for an analysis for cGVHD. For aGVHD, TBI alone was the top-ranked regimen; however, there was no significant difference in the risk of aGVHD when TBI was compared to any regimen other than ATG+BU+FLUD and MELPH+TBI. The standard conditioning regimen, CY+TBI, was was not significantly different from any other regimen. There were no significant differences in the 3 regimens included in the cGVHD network. • Additional harms-related findings: A pairwise meta-analysis comparing CY+TBI to BU+CY for the risk of veno-occlusive disease (VOD) at 28 days post-transplant found no significant difference between the regimens. However, a network meta-analysis including CY+TBI, MELPH+TIB, BU+FLUD, and BU+CY, demonstrated that CY+TBI significantly reduced the risk of VOD compared to BU+CY at ≥100 days post-transplant. Other adverse events including specific organ toxicity and infections were described in a small number of studies, precluding meta-analysis. Based on available data, CY+TBI significantly reduced the risk of bronchiolitis obliterans when compared to BU+CY in long-term follow-up in one study. CY+TBI was associated with a higher risk of a positive blood culture at 100 days compared to BU+CY in one study; however, in another study, there was no significant difference between the two regimens in the risk of bacteremia after 2 years. The single study comparing modified regimens of BU+CY and BU+FLUD identified a significantly higher risk of severe pneumonia in the modified BU+FLUD group after 1.4 years, which ultimately halted the study. Finally, the addition of FLUD to CY+ATG in conditioning regimens for aplastic anaemia patients was associated with significantly reduced regimen related toxicity and pulmonary complications.

Recommendations and Future Studies Numerous drugs are used in a broad variety of single- and multi-agent conditioning regimens prior to HSCT. This review found that comparative evidence from randomized trials is lacking for many comparisons of these regimens, especially newer regimens, which has led to considerable practice variation between institutions. This systematic review of the evidence, incorporating network meta-analyses where possible, was conducted to address these gaps. Eighteen randomized trials of 18 conditioning regimens were identified and studied to inform analyses in this review.

The following key points for clinical practice were identified in this review: • CY+TBI is associated with improved survival compared with BU+CY and with BU+FLUD • The risks of relapse and acute and chronic GVHD were not significantly different between any regimens. • CY+TBI is associated with a reduced risk of VOD after 100 days post-transplant compared with BU+CY

Future studies of conditioning regimens should be stratified for competing factors, such as patient age and underlying disease, and donor factors. A standard arm from the existing evidence network should be included in future studies to leverage existing knowledge. Consistent outcome reporting is urgently needed in HSCT studies to improve network analysis and comparison with other studies.

Key summary messages regarding conditioning regimens for HSCT were as follows: o Inconsistency in reporting outcomes in RCTs of conditioning regimens precludes the ability to perform meta-analyses for several important transplant outcomes. o Network meta-analysis demonstrated that CY+TBI was associated with lower overall mortality at 1 year post-transplant compared to BU+CY, and at 2 and 4 years post- transplant compared to BU+CY and BU+FLUD. o Network meta-analysis did not identify statistically significant differences in any of the conditioning regimens with respect to relapse (4 regimens), acute GVHD (9 regimens), or chronic GVHD (3 regimens). o CY+TBI and BU+CY were associated with similar rates of VOD at 28 days post-transplant (2 studies); however, at ≥100 days post-transplant, CY+TBI significantly reduced the risk of VOD compared to BU+CY (4 studies) o CY+TBI was associated with reduced incidence of bronchiolitis obliterans compared to BU+CY (1 study).