Graft versus host disease (GVHD) is a major cause of morbidity and mortality following allogeneic hematopoietic stem cell transplantation (HSCT). Gastrointestinal (GI) GVHD occurs in 10 to 40% of patients. Its diagnosis is based upon histological findings of per-endoscopic mucosal biopsies. The aim of our study was to compare endoscopic appearance of the mucosa and histological results and assess the outcomes of transplanted patients who underwent duodenal biopsies according histological and endoscopic findings.
We performed a retrospective analysis including allografted patients who underwent duodenal biopsies because of digestive GVHD suspicion. For each patient we recorded clinical data, endoscopic signs, and then we reviewed histological findings.
Diagnosis of GVHD was histologically defined as the presence of single-cell necrosis. The severity of disease was graded on a four-point scale from 0 (absent) to 4 (very severe), according the number of single-cell necrosis, the degree of mucosal infiltration, the presence of neutrophils and eosinophils, epithelial rupture, and vascular involvement. Endoscopic findings were described and graded as normal mucosa (0), presence of erythema or oedema with congestive mucosa (1), ulceration (2), and extensive mucosal involvement (3). Agreement between endoscopy and histology was measured by to kappa statistics. Outcomes were assessed by Kaplan-Meyer survival curves.
Between 1995 and 2008, 270 duodenal biopsies in 228 patients have been performed in patients. We focused only on the first one. Indication for HSCT was acute leukaemia or myelodysplastic syndrome in 133 patients (58.3%), NHL/HD in 35 (15.4%), CML or myeloproliferative disease in 35 (15.4%), other diagnosis in 25 (10.9%). Donor was identical sibling in 48% of cases, unrelated donor in 52% (HLA matched 21% and mismatched 31%). Stem cell source was bone marrow in 46%, peripheral blood in 42% and cord blood in 12%. Finally most of patients (72%) received a myeloablative conditioning whereas the others had reduced intensity conditioning. The median time from transplantation to biopsy was 28 days. More than two thirds of patients (77%) presented an acute GVHD occurring before day 100. Upper digestive endoscopy was performed at the onset of digestive symptoms; nausea or vomiting in 25% of cases, diarrhea in 75%.
Pathological GVHD was diagnosed in 77% of patients, graded minimal (1) in 29%, mild (2) in 25%, severe (3) in 10% and very severe (4) in 12%. Concerning endoscopic findings, normal macroscopic aspect was noted in 80% of cases (n=176), erythema or oedema was reported for 12% of patients, mucosal ulceration in 6%, and extensive mucosal involvement in 2%. There was no agreement between histological diagnosis and endoscopic findings with a kappa coefficient of -0.08, [95%CI: -0.17;-0.01], even for severe histological stages. Then 80% of patients with normal endoscopy had a diagnosis of GVHD, which was graded as histologically severe or very severe in 39%.
The median follow up was 47 months (31-68), the 4 years overall survival was 47% [IC95% 41%; 54%]. Severe (n=24) and very severe (n=27) histological GVHD was significantly associated with mortality, with respectively HR 3.23 [95%CI: 1.82; 5.72] p<0.0001 and 7.22 [95%CI: 4.08; 12.8] p<0.0001. Similarly severity of mucosal involvement at endoscopy was associated with mortality, HR 3.66 [95%CI: 2.12; 6.31] p<000.1 for endoscopy stage 2 or more. Both histological (≥3) and endoscopic (≥2) grade appeared as independent prognosis factors.
Since digestive GVHD is a pan-intestinal process, systematic duodenal biopsies in allografted patients suffering from high or low digestive symptoms permit a diagnosis of GI GVHD in more than three quarters of cases. The early realization of endoscopic examination leads most often to a normal macroscopic aspect of small bowel mucosa and there is no agreement between histological and endoscopic findings. Mucosal injury could be more likely linked to a viral involvement. Nevertheless if endoscopic findings do not lead to GVHD diagnosis, presence of ulceration and extensive involvement is predictive of mortality, as well as histological severity of GVHD.
No relevant conflicts of interest to declare.
Asterisk with author names denotes non-ASH members.