Background: HaploSCT from unmanipulated graft is a practiced option for high risk acute leukemia (AL) patients (pts) who lack a matched related or unrelated donor (Piemontese, Leukemia 2015). Since several haploidentical donors are available for most pts, selection based on optimal HLA mismatching could be relevant, but evidence on this issue is scarce.

We aimed to evaluate the impact of donor-recipient HLA mismatches (HLAmm) on unshared haplotype on HaploSCT outcomes, hypothesizing that the number of HLAmm and the different HLA loci involved could be associated with specific transplant-related risks.

Methods: We retrospectively analyzed 490 pts with de novo acute myeloid leukemia (AML, n=346, 71%) and acute lymphoblastic leukemia (ALL, n=144, 29%) reported to EBMT, receiving a first unmanipulated peripheral blood (PB) or bone marrow (BM) HaploSCT between 2007 and 2014. Donor-recipient pairs were typed at a molecular low (n=323, 66%) or high (n=163, 34%) resolution level at HLA-A, -B, -C and -DRB1 loci. HLAmm were defined at antigen level.

Results: Median follow-up was 17 (1.3-84) months; median age at transplant was 44 (18-78) years (y). Three hundred-ten pts (63%) were in complete remission (34% CR1, 29% CR2), 180 pts in advanced phase (CR3 and above, relapse or primary induction failure). Conditioning regimen was reduced intensity (RIC) in 48% of pts and myeloablative in 52%. Source of stem cells was PB for 61% of them, BM for 39%. In vivo T cell depletion (ATG or Campath) was used in 33% of pts and 56% of pts received high-dose post-transplant Cyclophosphamide (PTCy) for GVHD prophylaxis. Median donors' age was 38 (12-71) y. CMV donor/host status was neg/neg for 11% of pts, pos/neg for 6%, neg/pos for 19%, pos/pos for 62%.

Of the 490 donor-recipient pairs, 55% had 4 HLAmm on unshared haplotype, 30% had 3 HLAmm, 12% had 2 HLAmm, 3% had 1 HLAmm. Among pairs with less than 4 HLAmm, 17% were matched at locus A, 12% at locus B, 19% at locus C, 16% at locus DRB1.

Overall, 92% of pts achieved neutrophil engraftment at a median of 18 (12-63) days (d). At 100 d, cumulative incidence (CI) of grade ≥2 acute GvHD (aGvHD) was 33±4%. The 2-y CI of chronic GvHD (cGvHD), relapse (RI) and non-relapse mortality (NRM) were 30±4%, 37±5% and 29±4%, respectively. The probabilities of leukemia-free survival (LFS) and overall survival (OS) at 2 y were 34±5% and 39±5% respectively.

In multivariate analysis, increasing number of HLAmm on unshared haplotype was not an independent prognostic factor for aGvHD (HR 0.8, p 0.2), cGvHD (HR 1.3, p 0.3), NRM (HR 1.3, p 0.3), RI (HR 1.1, p 0.6), LFS (HR 1.2, p 0.3) and OS (HR 1.3, p 0.2).

Focusing on single HLA loci, match at HLA B locus on unshared haplotype was associated with lower 2y NRM in univariate analysis (17±8% in matched pairs, 31±11% in mismatched pairs, p 0.02), and this was confirmed in multivariate analysis (HR 0.5, p 0.04) accounting for all known risk factors (AL type, disease status, conditioning regimen intensity, host/donor age and sex, GvHD prevention, CMV status, number of HaploSCT performed in each center).

Moreover, donor-recipient pairs matched at HLA DRB1 locus on unshared haplotype had lower rates of 100d CI of grade ≥2 aGvHD (20±8% in matched pairs, 34±4% in mismatched pairs, p 0.01), and this proved to be an independent factor in multivariate analysis (HR 0.5, p 0.02).

Among other relevant factors, diagnosis of AML was associated with higher probabilities of LFS and OS (HR=0.6, p<10e-3 and HR=0.5, p<10e-4 respectively) and lower RI and NRM (HR=0.6, p=0.001 and HR=0.7, p=0.02 respectively). Advanced disease was a risk factor for LFS and OS (HR=3.4, p<10e-14 and HR=3.5, p<10e-12 respectively) as for RI and NRM (HR=6.1, p<10e-14 and HR=1.7, p=0.01 respectively). Significant factors for aGvHD were: PB as stem cell source (HR 1.9, p<10e-3), CMV status neg/neg (HR 0.5, p 0.02) and RIC (HR 0.7, p 0.02). PTCy and RIC regimen were the only independent significant factors for cGvHD (HR=0.6, p=0.01 and HR=0.6, p=0.002 respectively).

Conclusions: In our series of 490 AL pts, number of HLAmm on unshared haplotype did not influence outcomes of unmanipulated HaploSCT. However, a match at HLA-B and HLA-DRB1 loci on unshared haplotype was associated with lower NRM and lower risk of grade 2-4 aGvHD, respectively.

If further validated, our data could drive a more accurate selection of haploidentical donors in non T-cell depleted HaploSCT, accounting for potential biological implications of HLA locus-specific differences.


Tischer:Sanofi-Aventis: Other: advisory board. Nagler:Novaratis Pharmaceuticals Corporation: Consultancy, Honoraria, Research Funding. Mohty:Janssen: Honoraria; Celgene: Honoraria.

Author notes


Asterisk with author names denotes non-ASH members.