Methods: We analyzed patients with acute leukemia or myelodysplastic syndrome who underwent the first UCBT (n = 3,224) between 2000 and 2012 in Japan. The effect of acute GVHD (aGVHD) on overall mortality, relapse, and non-relapse mortality (NRM) was analyzed after adjusting for other significant variables among the engrafted patients, whereas the effect of chronic GVHD (cGVHD) was analyzed among the engrafted patients who survived without relapse for at least 100 days. The occurrence of GVHD was treated as a time-dependent covariate.
Results: The occurrence of grade 3 or 4 (G3-4) aGVHD was significantly associated with a higher risk of NRM (hazard ratio [HR] 3.08, P < 0.001) than the occurrence of G0-1 aGVHD, in both the standard- and high-risk groups. The occurrence of G2 or G3-4 aGVHD, as compared with G0-1 aGVHD, was significantly associated with a low relapse rate (G2 aGVHD: HR 0.80, P = 0.003; G3-4 aGVHD: HR 0.71, P = 0.005). These resulted in the significant association between G2 aGVHD and low mortality (HR 0.79, P < 0.001), and G3-4 aGVHD and high mortality (HR 1.70, P < 0.001), as compared with G0-1 aGVHD. The association between G3-4 aGVHD and high mortality was stronger in the standard-risk group (HR 2.46, P < 0.001) than in the high-risk group (HR 1.40, P < 0.001). The occurrence of extensive cGVHD was associated with a low relapse rate as compared with no cGVHD (HR 0.77, P = 0.046). The effect of limited cGVHD was significant only in the high-risk group (standard-risk: HR 1.07, P = 0.673; high-risk: HR 0.65, P = 0.007). The occurrence of extensive chronic GVHD was significantly associated with high NRM only in the standard-risk group (standard-risk, HR 1.46, P = 0.037; high-risk: HR 0.93, P = 0.701). These resulted in the significant association between limited cGVHD and low mortality in both groups, whereas extensive cGVHD was associated with low mortality only in the high-risk group.
Conclusions: Similar to transplantations from a matched sibling or an unrelated donor, G3-4 aGVHD should be prevented because of its associated high overall and non-relapse mortality rates, although acute GVHD was associated with low relapse rates. Extensive cGVHD should be prevented particularly in the standard-risk group because of its associated high NRM. In the high-risk group, cGVHD was associated with low relapse rates. The significant association between aGVHD or cGVHD and a low relapse rate supports the presence of GVL effect in the UCBT even in the high-risk group.