Impact of Relapse after Allogeneic Hematopoietic Stem Cell Transplant on Subsequent Outcome in Patients with Chronic Myeloid Leukemia: A Better Understanding through Multi-State Modeling.

Donor lymphocyte infusions and imatinib have been the mainstay of treatment for chronic myeloid leukemia (CML) patients relapsing after allogeneic hematopoietic stem cell transplantation (alloSCT) in recent years. Improvements in transplant related mortality and a reduction of lethal infections have been documented. No comprehensive data are yet available about the impact of relapse (REL) itself on survival and a possible improvement over calendar time.

We analyzed 13418 patients with CML transplanted between 1980 and 2003 (60% male, median age 36 years [range 0–71]) from an allogeneic donor (65% HLA identical sibling, 1% twin, 7% other related, and 27% unrelated). Of those, 2558 (19%) were reported with REL.

We investigated the impact of REL on death rate in 4 periods: A [1980–92], B [1993–96], C [1997–99] and D [2000–03], adjusting for other factors. We applied a series of Cox models where REL was used as a time-depending covariate and where we used an integrated approach in the framework of multi-state models, with REL representing an intermediate event in a patient’s post-transplantation history. We used the EBMT risk score as a propensity score to adjust for intrinsic differences in survival over calendar time and Calendar time as categorical factor to assess trends.

REL increased death rate over the whole observation period. However, there was a significant improvement over the years: death hazard ratio (HR) decreased from 4.5 to 3.2, 2.7 and 2.4 for the periods A, B, C and D, respectively. The impact of REL on death rate was also influenced by the EBMT risk score [HR 1, 1.9, and 3.5 for low, intermediate, and high risk score, respectively]. We investigated, with the same methodology used above, the impact of REL on death rate in a subset of 1977 patients transplanted after 1992 and with known information on the type of relapse at its first evidence (molecular/cytogenetic [M-REL] or hematological [H-REL]). The M-REL/H-REL ratio increases over the years (1.1, 1.2, and 2.5 in periods B, C, and D, respectively). Results confirmed that the impact of REL on death rate decreases over the years, but hardly and not significantly for H-REL (HRs around 6 in periods B, C, D) and more pronounced and significantly for the M-REL (HRs 0.68, 0.52, and 0.12 in periods B, C, and D, respectively). It is notable that H-REL was associated with higher death rate, M-REL with lower.

We have shown that:

• REL after alloSCT for CML is associated with a higher risk for subsequent death;

• the effect of REL on death rate is decreasing in more recent years;

• the effect of REL on death rate depends on the type of REL (H-REL or M-REL) and the EBMT risk score;

• an integrated approach like multi-state modeling is flexible and allows clinically relevant interpretations.

Our findings suggest that the effect of REL on death rate is decreasing in more recent years, mainly because of the increasing proportion of patients in whom REL is first detected at pre-clinical level (ie. molecular/cytogenetic relapse). Our results will form the basis for future description and characterization of patient groups in comparative retrospective analyses and prospective trials.