Pathologic Features May Predict the Clinical Outcome of Hepatic Graft Versus Host Disease (H-GVHD).

Liver histology is regarded as inadequate for grading the severity of H-GVHD. Thus, liver biopsies (LB) are only performed to rule out other causes of liver injury when the results may lead to a radical therapeutic change. In this study we reviewed the pathologic features in LB from 33 consecutive evaluable patients with at least one LB, confirmed diagnosis of H-GVHD and no other concomitant causes of liver dysfunction. We asked whether the pathologic features in the first LB may predict their clinical outcome. Underlying diseases were AML (9), ALL (8), CML (7), MM (4), MDS (3), NHL (1) and AA (1). H-GVHD developed after BM (15) or PB (11) allo-HSCT, or DLI (7). Twenty-eight donors were related and 5 unrelated, 30 HLA-matched and 3 mismatched. All patients received myeloablative TBI-based (31) or chemotherapy only (2) conditioning regimens, and GVHD prophylaxis with cyclosporine-methotrexate (21) or T-cell depletion (12). The onset was acute in 22 (median day 35, range 5–86) and chronic in 11 (median day 139, range 103–545). The median overall survival (OS) from the onset of liver dysfunction was 6.2 months, with a non-relapse mortality (NRM) of 73.2% at 1 year (CI95=50.2–96.2). The causes of NRM were infection (8), liver failure (3), hemorrhage (3) and others (4). The first LB was performed at a median of 14 days (4–68) from the onset of abnormal liver function tests. The commonest pathologic finding in this series was bile duct damage, but ductopenia was uncommon (4). Neither the presence nor the degree of bile duct damage associated with the clinical outcome. Lobular inflammation (LI) was present in 25 cases (grade 1: 16; grade 2: 6; grade 3: 3). Higher degree of LI in the first LB associated with a reduced NRM (P=.001) and improved OS (P<.001). Patients with LI grades 2/3 had higher median OS than those with grades 0/1 (not-reached vs 4.3 months; P=.012), and were more likely to respond to primary treatment (RPT) with steroids (P=.044), the most important clinical prognostic factor for OS in our series (HR=4.3; P=.001). Presentation of H-GVHD as hepatitis, with markedly increased aminotransferases levels, has been reported, in particular after DLI. In our group, there was no association between DLI-induced H-GVHD and LI, or between LI and aminotrasferases peak or biopsy-time levels. No study to date has shown that LI in the LB associated with clinical outcome in H-GVHD. In liver transplantation however, a histologically documented lobular hepatitic phase has been shown to anticipate ductopenia in chronic rejection, and associate with good response to treatment. In addition to LI, hepatocyte ballooning (HB) was associated with NRM (P=.003) and OS (P=.018) in our series. Patients with HB grades 0/1 had higher median OS than those with grades 2/3 (6.5 months vs 1.3 months; P=.004). HB did not associate with RPT. In multivariate analysis the effects of LI and HB on NRM (HR=5.1, P=.033; and HR=5.5, P=. 018, respectively) and OS (HR=4.0, P=.032; and HR=4.2, P=.037, respectively) remained significant. All other pathologic features examined had no association with patient clinical outcome. Our results suggest that LI and HB in the first diagnostic LB from patients with H-GVHD may predict the probability of RPT, NRM and OS. These results should be considered in the indications of LB and the selection of candidate patients with H-GVHD for new experimental therapies and in the design of future trials. Prospective validation of our findings is under way.