Overall, we observed very poor outcomes with FluMel conditioning and PBSC grafts for haplo-HCT with PTCy. One-year OS in our cohort was 34%, driven by unexpectedly high rates of NRM due to early toxicity after transplantation. Our reported NRM rates of 21% and 34% at day + 100 and 1-year, respectively, are much higher than the rates observed in other haplo-HCT studies utilizing RIC platforms with similar patient populations. A somewhat older report using FluMel conditioning with melphalan 140 mg/m2 on day − 8 for haplo-HCT with BM grafts found NRM rates at day + 100 and 1-year to be 12% and 16%, respectively2, and a subsequent cohort5 of similarly treated patients also showed a lower NRM (19% at 1-year). Besides our patients receiving PBSC grafts, our patient characteristics were comparable to both cohorts.
The high incidence of early NRM observed in our study was strongly associated with severe CRS, with 47% of these patients dying prior to day + 100. Patients with early mortality appeared to suffer from similar clinical syndromes characterized by severe CRS and/or high fevers, capillary leak, volume overload, significant edema and anasarca, weight gain, and oliguria. This was exacerbated by IV fluids administered with PTCy, often progressing to AKI, anuria, and metabolic acidosis requiring ICU admission and renal replacement therapy (RRT); some patients suffered further clinical decompensation and death. Anecdotally, while treatment with the anti-IL6R monoclonal antibody tocilizumab would rapidly ameliorate fevers, it did not appear to have an effect on patients’ overall clinical course.
These poor outcomes in patients with severe CRS are consistent with previous reports. However, the incidence of severe CRS in this study was 40%, approximately three-times the incidence described in previous analyses8,15. Abboud et al. reported severe CRS in 12% of patients undergoing haplo-HCT with PBSC grafts8. Additionally, a single institution study of 146 patients undergoing haplo-HCT with PBSC grafts reported severe CRS in 17% patients, who achieved a 6-month OS of only 50%15, further supporting the association between severe CRS and poor clinical outcomes. Severe CRS and early NRM were also associated with AKI and RRT requirement. Despite the excessive toxicity, this regimen did not appear particularly effective, with 33% of surviving patients relapsing.
CRS is thought to be primarily mediated by activated T-cells, and PBSC grafts contain nearly 8-fold more T-cells compared to BM grafts. Multiple reports have demonstrated the increased rates of severe CRS in patients receiving PBSC grafts compared to those receiving BM grafts for haplo-HCT with PTCy17. PBSC collection through G-CSF mobilization has been shown to be associated with different T-cell subsets than BM grafts. Different allograft T-cell subset contents have been associated with risk of acute and chronic GvHD as well as risk of relapse18. It is possible that allografts with a higher dose of alloreactive activated T-cells favor the development of CRS, while those with a higher number of regulatory T-cells are less likely to develop CRS. In addition to PBSC grafts, pre-transplant active disease and HLA-DRB1 mismatching have also been identified as independent predictors of grade ≥ 3 CRS19. Our patient cohort, though, had a relatively low rate of active disease, as 75% of patients with AML were in remission at time of transplantation. Other factors such as degree of HLA mismatch, patient or donor age, donor sex, natural killer (NK)-killer immunoglobulin receptors haplotype, NK-ligand mismatch, ABO mismatch, or HCT-CI have not been associated with severe CRS in prior reports8,15,17. However, the patients included in these prior studies almost exclusively received a Flu/Cy/TBI conditioning platform. This suggests that the high rate of severe CRS observed in our cohort may be related to the unique combination of using PBSC grafts with FluMel conditioning. One recent study found a higher rate of NRM and decreased OS in patients with NHL receiving FluMel conditioning for allogeneic HCT compared to fludarabine and busulfan-based conditioning13, whereas another cohort of patients with lymphoma undergoing haplo-HCT with fludarabine and cyclophosphamide conditioning did not suffer increased NRM or reduced OS14.
Melphalan is commonly used as part of fludarabine-based RIC regimens, which have been associated with decreased relapse rates and disease-free survival in other transplant types19. Melphalan pharmacokinetic studies report up to a 10-fold interpatient variability in melphalan exposure20. Melphalan binds to proteins in the red blood cell membrane, leading to higher plasma concentrations in patients with lower hemoglobin21. Renal impairment also leads to higher concentrations as renal elimination accounts for approximately 40% of melphalan clearance. Lower hemoglobin (< 9.5 g/dL) and impaired renal function (creatinine clearance < 60 mL/min) were strongly associated with worse outcomes in patients who underwent FluMel conditioning and autologous hematopoietic cell transplantation, although the use of BM versus PBSC grafts was not specified in this report22. Outcomes in our patient cohort appear consistent with these findings. Theoretically, these patients may have increased melphalan exposure and resultant toxicity. Unpublished data from our group are also finding that MAC regimens are not associated with an increased risk for severe CRS, thus it is possible that melphalan has a unique toxicity profile where increased melphalan-exposure places patients at higher risk for severe CRS.
Melphalan has also been shown to have immunomodulatory effects. Treating tumor-bearing mice with melphalan results in a rapid burst of inflammatory cytokines and chemokines, as well as transient elimination of regulatory T-cells during the cellular recovery phase after melphalan-induced leukodepletion. This effect includes surges of IFN-γ, IL-22, IL-5, IL-18, IL-27, CCL2, CCL3, CCL7, and CXCL123, with peak levels occurring two days after melphalan administration. This IFN-γ peak would coincide with the timing of PBSC infusion in our cohort, as our patients received melphalan on either day − 2 or day − 1. The combination of a melphalan-induced proinflammatory environment with T-cell replete haplo-HCT PBSC grafts may have contributed to the unacceptably high rate of severe CRS in our patients.
Our study is limited by its small sample size and retrospective nature. Patients were moderate-to-high risk and 29% had undergone prior hematopoietic cell transplantation. We also had a heterogeneous patient population with a diverse range of hematologic malignancies.
In conclusion, our institutions no longer use standard FluMel as conditioning for haplo-HCT with PTCy with T-cell replete PBSC grafts. Alternative regimens or variations on melphalan-based regimens, such as fractionated melphalan dosing, inclusion of TBI, or inclusion of thiotepa, may improve outcomes but further studies are needed. There remains a significant unmet need to better understand the optimal conditioning approach for patients with aggressive hematologic malignancies undergoing haplo-HCT who are older or unfit for standard MAC, and this should be addressed in future randomized controlled trials.