Stueck AE, Schiano TD, Fiel MI. Development of a novel histologic diagnostic algorithm for hepatic graft-versus-host disease.
Mod Pathol 2018;
31:442-451. [PMID:
29192646 DOI:
10.1038/modpathol.2017.151]
[Citation(s) in RCA: 6] [Impact Index Per Article: 1.0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 07/08/2017] [Revised: 09/21/2017] [Accepted: 09/22/2017] [Indexed: 12/16/2022]
Abstract
The histopathologic diagnosis of hepatic graft-versus-host disease post bone marrow and stem cell transplantation can be challenging, but timely and unambiguous diagnosis is essential for appropriate patient management. To address this diagnostic dilemma, we identified histologic features specific for hepatic graft-versus-host disease and developed a diagnostic algorithm. Two hepatopathologists blindly evaluated 40 liver biopsies from patients with clinically and biologically confirmed graft-versus-host disease, as well as 44 controls, for percent bile duct loss, bile duct damage, intraepithelial lymphocytes, ductular reaction, acidophilic bodies/10 high power fields (HPF), cholestasis, portal and lobular inflammation, and endotheliitis. Compared with controls, graft-versus-host disease cases had significantly more bile duct loss (P<0.0001), bile duct damage (P=0.0002), cholestasis (P<0.0001), and acidophilic bodies/10 HPF (P=0.0006), as well as significantly less ductular reaction (P<0.0001). Significance was maintained with a drug-induced liver injury-only control group. No histologic differences were noted in acute versus chronic graft-versus-host disease, nor cholestatic versus hepatitic types. An algorithm to predict likelihood of graft-versus-host disease was developed, with a three-tiered scoring system: 1-2 not, 3-4 probable, and 5-8 unequivocal graft-versus-host disease. This algorithm had a sensitivity of 93%, specificity of 93%, and accuracy of 92%. We identified histologic features with specificity for hepatic graft-versus-host disease and developed a simple algorithm for pathologists to predict its likelihood, distinguishing this critical diagnosis promptly from mimickers having vastly different treatments and prognoses.
Collapse