Abstract
The value of allogeneic hematopoietic stem cell transplantation (alloHSCT) in the treatment of multiple myeloma (MM) patients continues to be debated. Previously, we reported a donor versus no donor (DvND) analysis of MM patients, who were enrolled in the prospective upfront HOVON50 trial (Lokhorst, Blood 2012). By intention to treat, donor availability did not appear to be associated with better outcome. However, as a considerable number of MM patients with a matched related donor (MRD) were not transplanted, a DvND analysis may underestimate the effect of alloHSCT. Therefore, we performed a re-analysis using alloHSCT as a time dependent variable; this method has recently gained more attention since it may more closely evaluate the real impact of alloHSCT. In addition, a landmark (LM) analysis was performed and subsequently the 3 methods were compared.
Briefly, the HOVON50 study included adult patients up to the age of 65 years with newly diagnosed Salmon and Durie stage II or III MM, who were randomized to receive induction therapy with vincristine, adriamycine, dexamethasone or thalidomide, adriamycine, dexamethasone. Subsequently, patients received high dose melphalan (HDM) and autoHSCT followed by maintenance therapy. Patients with a MRD were intended to receive alloHSCT following reduced intensity conditioning (RIC). 260 MM patients underwent HDM in the HOVON50 trial, including 122 patients with a MRD of whom 99 patients actually proceeded to alloHSCT. Methods of analysis were compared with respect to overall survival (OS), progression free survival (PFS), non-relapse mortality (NRM), and the cumulative incidence of progression (CIP). To evaluate the impact of alloHSCT (or donor, for the DvND method) the hazard ratio (HR) and 95% confidence interval (95%CI) using a multivariate Cox model with adjustment for age, sex, disease status at the time of HDM, Salmon and Durie classification, LDH, adverse cytogenetics and treatment arm were assessed. The maintenance group constituted the reference group, while the no-donor group constituted the reference in the DvND comparison.
The DvND comparison did not show differences between patients with (n=122) or without (n=138) a MRD as from the time of HDM; median follow-up was 93 months. By time dependent analysis, all patients started in the control maintenance group and only those patients, who were actually transplanted, switched to the alloHSCT-group at the exact date of transplantation. For a LM analysis, all patients, who were alive and disease-free at 6 months after HDM (n = 239) were included and outcome was evaluated, according to whether they received (n = 91) or did not receive (n = 148) an alloHSCT. By time dependent analysis, it appeared that alloHSCT significantly improved PFS, while the DvND did not reveal a PFS benefit (Table). The 6-months LM analysis also revealed an advantage in PFS for patients who underwent alloHSCT. None of the 3 methods revealed an advantage in OS, although HR's following LM analysis and the time dependent analysis were < 1.0 (respectively 0.82 and 0.89, not significant (ns)), while the DvND analysis yielded a HR of 1.22 (ns). Moreover, the graft versus myeloma (GvM) effect in terms of reducing the CIP appeared stronger after LM and time dependent analysis as compared to the DvND analysis (respectively 0.46 (p=0.001) vs 0.49 (p<0.001) vs 0.60 (p=0.008).
Collectively, our results show that the efficacy of alloHSCT is underestimated by a DvND analysis, while both the time dependent analysis and a landmark analysis may more reliably and quantitatively reflect the allogeneic GvM effect of RIC alloHSCT as part of first line treatment in MM. Exclusion of the bias of the guarantee-time and the inclusion of all patients with either a MRD or MUD, actually being transplanted further constitute advantages of the time dependent approach as compared to a LM analysis. We conclude that the value of alloHSCT in MM should be reassessed using a time dependent analysis as preferred method in future studies.
Abstract 1209. Table Donor vs No donor Time dependent Landmark HR 95%CI p HR 95%CI p HR 95%CI p OS 1.22 [0.81-1.86] 0.345 0.89 [0.57-1.39] 0.601 0.82 [0.51-1.33] 0.422 PFS 0.80 [0.56-1.13] 0.201 0.69 [0.47-0.99] 0.047 0.63 [0.43-0.93] 0.019 CIP 0.60 [0.41-0.88] 0.008 0.49 [0.33-0.75] 0.001 0.46 [0.30-0.71] < 0.001 NRM 6.15 [2.09-18.08] 0.001 6.79 [2.27-20.27] 0.001 3.91 [1.50-10.21] 0.005
Disclosures
No relevant conflicts of interest to declare.
Nonpremixed Counterflow Flames: Scaling Rules for Batch Simulations