Predictors of Graft-Versus-Host Disease with Post-Transplant Cyclophosphamide

Post-transplant cyclophosphamide (PTCy) graft-versus-host disease (GvHD) prophylaxis has shifted the epidemiology of GvHD by reducing the rate of grade III-IV acute (aGvHD) and chronic GvHD (cGVHD). Use of PTCy has recently expanded beyond the haploidentical (haplo) transplant (SCT) setting, which necessitates reassessment of the conventional risk factors for GvHD in the current era. Here, we performed a single-center retrospective study to evaluate predictors of aGvHD and cGvHD after allogeneic SCT (allo-SCT) with PTCy/tacrolimus ± mycophenolate mofetil (MMF) GvHD prophylaxis.

Eligible patients were those that engrafted after first peripheral blood (PB) or bone marrow (BM) allo-SCT for AML/MDS between 2018-2023. Donors were HLA-matched related (MRD), 10/10 -matched unrelated (MUD), haplo, or 9/10 -mismatched unrelated (MMUD). Evaluated risk factors included donor and recipient age (years), sex, and CMV serostatus, hematopoietic stem cell comorbidity index (HCT-CI), diagnosis, disease risk index, donor type, stem cell source, conditioning (myeloablative vs other), and GvHD prophylaxis ( ± MMF). All analyses accounted for competing risks. Multivariate analysis (MVA) was indicated when more than one predictor was significant (p<0.05) in univariate analysis (UVA). Given the differential impact of PTCy on grades II versus III-IV aGvHD, these outcomes were evaluated separately.

A total of 948 consecutive patients met the eligibility criteria, including 237, 463, 182, and 66 recipients of SCT from a MRD, MUD, haplo, and MMUD, respectively. Median recipient age was 60, 63, 57, and 60, respectively. Donor and transplant -but not disease-characteristics differed across donor types. Donor age was higher in MRD (median: 57 (range 6-79), p<0.01) compared with MUD (29 (18-73)), haplo (34 (12-69)), and MMUD (31 (18-58)). Graft source was BM in 71% of haplo (p<0.01) versus 2% in MRD, 14% in MUD, and 18% in MMUD. Male donors were more (p<0.01) likely in haplo (66%) and MUD (70%) versus MRD (47%) or MMUD (48%) SCT. CMV positive/positive donor/recipient pairs were predominant in MRD (55%) and haplo (46%), and negative/positive pairs in MUD (42%) and MMUD (38%) SCT. Conditioning was myeloablative in 75%, 42%, 65%, and 68% in MRD, haplo, MUD and MMUD, and MMF was used in 51%, 98%, 68%, and 88%, respectively. Median follow-up was 24 (0.6-70) months.

Overall, day 180 cumulative incidence (CI) of grades II and III-IV aGvHD was 28% and 9%. Two-year CI of cGvHD was 15%. In UVA, HLA disparity impacted aGvHD but not cGvHD. The rate of grade III-IV aGvHD was higher (HR=1.7, p 0.01) with mismatched [haplo (12%) and MMUD (16%)] versus matched [MRD (7%) and MUD (8%)] SCT. A similar trend was seen for grade II aGvHD, but the difference was significant only for MRD (23%, HR=0.7, p 0.04) versus MUD (28%), haplo (32%), or MMUD (33%) SCT. Two-year CI of cGvHD was comparable across donor types: 18%, 14%, 12%, and 15% in MRD, MUD, haplo, and MMUD SCT, respectively.

Since the distribution of some donor and SCT characteristics was skewed by donor type, we evaluated the remaining GvHD risk factors separately for each donor type. Significant predictors of grade III-IV aGvHD in UVA were CMV mismatched (positive/negative or negative/positive) donor/recipient status (HR=2.9, p 0.03) in MRD; recipient age ≥45 (p 0.003) in MUD; donor age >30 (HR=14, p 0.01) in haplo; and donor age >40 (HR=3.6, p 0.04) in MMUD SCT. MVA was not indicated for grade III-IV aGvHD. Significant predictors of grade II aGvHD in UVA were HCT-CI >2 (HR=1.9, p 0.03) and SCT from a male donor to a male recipient (HR=1.9, p=0.02) for MRD; donor age >40 (HR=1.7, p 0.01) and CMV seropositive donor and/or recipient (HR=2.2, p 0.003) in MUD; SCT from a female donor to a male recipient (HR=2.2, p 0.01) in haplo, and donor age >30 (HR=2.7, p 0.04) in MMUD. Predictors remained significant in MVA for the MRD and MUD groups. Significant predictors of cGvHD in UVA and MVA were recipient age ≥60 (HR=3.2, p 0.02) and use of PB (HR=3.2, p 0.01) in the haplo group. No significant predictors were identified for MRD or MUD SCT. Only 6 patients were diagnosed with cGvHD in the MMUD group.

Our results show that for allo-SCT with PTCy prophylaxis, risk factors for GvHD vary across donor types and for grade II aGvHD, grade III-IV aGvHD, and cGVHD. These variations should be considered in donor selection algorithms and signal potential varying biological impacts of PTCy on GvHD depending on the SCT setting.