Introduction

Graft-versus-host disease (GVHD) occurs when immune cells from a donor graft recognize the transplant recipient as foreign, causing an immune reaction in the recipient. It is a complication of hematopoietic stem cell transplantation (HSCT) that frequently affects the gastrointestinal (GI) tract. This can present as a wide range of symptoms including nausea, vomiting, anorexia, abdominal pain, rectal bleeding, and/or diarrhea.

The diagnosis of GVHD requires pathologic examination of tissue obtained from endoscopy; however, the optimal endoscopies and sites of biopsies are unclear. Furthermore, GVHD patients are typically very ill and may have difficulty tolerating complete endoscopic evaluation (e.g., colonoscopy (CLN)). Preferred biopsy sites may also depend on the experience of the HSCT center [1]. An informal review of practice among gastroenterologists at Vancouver General Hospital reveals differing endoscopies (e.g., esophagogastroduodenoscopy (EGD) and flexible sigmoidoscopy (FS), CLN) and biopsy sites (e.g., random, multiple colonic biopsies) in evaluating GVHD.

Previous studies have suggested that rectal biopsies alone may be sufficient in establishing the diagnosis of GVHD in both adults [2, 3] and children [4–6]; however, a negative rectal biopsies may not rule out the diagnosis [7] and may require further evaluation of the lower GI tract including the distal colon [8, 9] and terminal ileum [10]. There appears to be little advantage in obtaining more proximal colon biopsies [4, 9, 11] yet CLNs are frequently used in the evaluation of GVHD. With respect to EGD, gastric biopsies appear to be the highest yield [4, 12, 13], but more recent studies suggest that the duodenum may be an important biopsy site if endoscopy is normal [14, 15].

The main objectives of this study are to compare the endoscopies and biopsy sites of biopsy-proven GVHD (GVHD +) to a sample of GVHD-negative (GVHD−) cases (patients who were brought to the endoscopy suite specifically to rule out GVHD but yielded negative biopsies) and to determine the ideal biopsy sites in diagnosing GVHD.

Materials and Methods

All cases of biopsy-proven GI GVHD (GVHD+) in the past 2 years (December 2012 to December 2014) at a tertiary center (Vancouver General Hospital, Vancouver, BC, Canada) were obtained from a pathology database. The search term “graft versus host” without the terms “skin, cornea, liver, bone, pancreas, vascular, buccal, femoral, kidney, tubular, renal, corneal, aorta, marrow, karyotype, knee, valve, and grafted” was utilized (n = 46). To be classified as a GVHD+ patient, at least one biopsy site was positive for GVHD of any grade as defined below (Grade ≥ 1). For comparison, 46 consecutive patients were selected who had undergone endoscopy to rule out GVHD but ultimately had negative biopsies at all sites sampled (GVHD−). Not all sites in the GI tract were sampled in the GVHD− population.

The site of biopsies as well as the severity of GVHD was documented from these pathology reports. Severity of GVHD was determined by pathological criteria. Grade 1 was defined as a single cell of apoptosis in the crypt epithelium. Grade 2 was defined as destruction of a single crypt or gland with apoptotic crypt abscesses. Grade 3 was defined as focal mucosal necrosis with ulceration. Grade 4 was defined as diffuse mucosal necrosis with denudation. Grade 0–1, 1–2, 2–3, or 3–4 noted on pathology report was rounded up and classified as Grade 1, 2, 3, and 4, respectively. A patient’s highest grade of GVHD of any biopsy was recorded.

A chart review was completed to identify demographic information including age, gender, and reason for HSCT as well as clinical information including symptoms and types of endoscopies completed. The related endoscopy report was analyzed to determine endoscopic findings and biopsy sites.

Comparisons between the GVHD+ and GVHD− groups were calculated by Fisher’s exact test unless otherwise indicated. Statistical significance was set at p < 0.05. Sensitivities at various biopsy sites with each type of endoscopy were calculated (i.e., number of biopsies with the diagnosis of GVHD divided by the total number of biopsies at that site). Since there were no cases of GVHD diagnosed subsequently after negative endoscopic evaluations, specificities of biopsy sites could not be determined. Agreement between GVHD results at different biopsy sites was determined with kappa statistical analysis. On upper endoscopies, agreement between esophageal, gastric, and duodenal biopsies were compared. On lower endoscopies, agreement between right-sided biopsies (proximal to the splenic flexure—i.e., ileal, cecal, ascending colon, and transverse colon biopsies) was compared to left-sided biopsies (distal to the splenic flexure—i.e., descending colon, sigmoid, and rectal biopsies). “Random colon” biopsy refers to a single biopsy in which the exact site is not mentioned in the endoscopy report. Statistical analyses were completed using GraphPad Prism (GraphPad, La Jolla, California) and Stata 11.2 (StataCorp, College State, Texas). This study was approved by the Ethics Review Board of the University of British Columbia.

Results

Table 1 demonstrates the baseline characteristics of the GVHD+ (n = 46) and GVHD− (n = 46) groups. The median age was 55 years old (interquartile range (IQR): 45–60 years old) in the GVHD + group and 48 years old (IQR: 36–60 years old) in the GVHD− group (p = 0.18). There were no significant differences in gender. The most common reason for the HSCT was acute myeloid leukemia (AML) or non-Hodgkin’s lymphoma (NHL) (22 % each) in the GVHD+ group, while in the GVHD− group, the most common reason was AML. About 50 % of the patients were less than 100 days from their HSCT in the GVHD+ group versus 39 % in the GVHD− group. The most common symptoms were diarrhea and nausea and vomiting (N/V) at 54 and 11 %, respectively, in the GVHD+ group. Results were similar in the GVHD− group. Diarrhea (54 %) was more common in the GVHD+ group (p = 0.02), while epigastric pain was more common in the GVHD− group (p = 0.03). As a result, not surprisingly, more CLNs were performed in the GVHD+ population (p = 0.02) and more EGDs were performed in the GVHD− population (p < 0.01). FSs, however, were performed equally in both the GVHD+ and GVHD− groups. On endoscopy, erythema was more commonly reported in the GVHD+ population but was not statistically significant (p = 0.07).

Table 1 Demographic information of GVHD+ and GVHD− groups
Full size table

Table 2 shows the distribution of GVHD grade among the GVHD+ population. Over 70 % of the GVHD+ population had the highest grade of GVHD at any biopsy site of 1 or 2. When the two most common findings noted on endoscopy reports were compared (i.e., normal and erythema) in Table 3, there was a trend of higher rates of erythema with higher grades of GVHD in the GVHD+ population.

Table 2 Distribution of the highest grade of GVHD on biopsy site of the GVHD+ population
Full size table
Table 3 Common findings noted on endoscopy reports compared to grade of GVHD
Full size table

Table 4 demonstrates the distribution of biopsies with any EGD (including combined with FS or CLN). In the GVHD+ group (n = 22), 10 patients underwent an EGD alone, 10 patients had both an EGD and FS, and 2 patients had both an EGD and CLN. Duodenum was sampled in all those who had undergone an EGD alone. In the GVHD− group (n = 32), 27 patients underwent an EGD alone, 3 patients had both an EGD and FS, and 2 patients had both and EGD and CLN. Duodenum was nearly sampled in all cases in those who had undergone an EGD alone (93 %). The distribution of biopsy sites showed no differences except esophageal biopsies were more commonly obtained in the GVHD− group (p = 0.04).

Table 4 Distribution of biopsy sites with upper and lower endoscopies
Full size table

With respect to lower endoscopies, in the GVHD+ group (n = 24), a FS was performed in nine patients, and a CLN was performed in 15 patients (Table 2). The most commonly sampled site on FS was rectum (78 %), while on CLN, descending colon and ascending colon were commonly sampled (both at 73 %) in the GVHD+ group. In the GVHD− group (n = 14), a FS was performed in nine patients, and a CLN was performed in five patients. The most commonly sampled site on FS was sigmoid (56 %); on CLN, the descending colon and ascending colon were sampled in all cases. When the GVHD+ and GVHD− groups were compared, the distribution of biopsy sites was similar.

For EGDs, including those performed in combination with FS or CLN, the sensitivities were highest for duodenal biopsies, ranging from 86 to 100 %. When different sites were compared for concordance, there was only 53 % agreement between gastric and duodenal biopsies (kappa statistic 0.01, p = 0.46). There was only one case of GVHD not detected by duodenal biopsy but found on a gastric biopsy.

For FSs and CLNs, the sensitivities among all sites were similar. There was 85 % agreement among biopsies obtained from the lower GI tract (kappa statistic 0.58, p = 0.01). There were no cases in which GVHD was diagnosed on a right-sided colon biopsy without a positive biopsy in the left side of the colon.

When the results of all endoscopies are combined (Table 5), in the upper GI tract, duodenal biopsies produced the highest yield in diagnosing GVHD (sensitivity of 89 %). In the lower GI tract, aside from the transverse colon and terminal ileum with minimal sampling in our cohort, the sensitivities were similar ranging from 69 to 88 %. The sensitivity of the lower GI tract was lower compared to the upper GI tract.

Table 5 Sensitivities of GVHD at different sites of all endoscopies combined
Full size table

Discussion

In this cohort of biopsy-proven GI GVHD, we found that for EGDs, duodenal biopsies produced the highest yield in diagnosing GVHD when compared to other sites of the upper GI tract. On lower endoscopies, sensitivities were similar in the right and left sides of the colon, suggesting that a FS is sufficient in evaluating GVHD in suspected patients with diarrhea.

Previous studies have suggested that gastric biopsies are ideal sites for diagnosing GVHD [4, 12, 13], but more recent studies have demonstrated otherwise [14, 15]. Nomura and colleagues examined 115 patients with biopsy-proven GVHD [14]. They found that gastric mucosal exfoliation, albeit rare, had a specificity and positive predictive value of 100 % for diagnosing GVHD. In the duodenum, erosions and ulcers had a specificity of 90 and 93 %, respectively. Often, endoscopically, the duodenum was normal, but GVHD was diagnosed nonetheless. Similarly, in a recent study, Ma et al. [15] showed that the highest prevalence of GVHD was in the duodenum/jejunum at 81 % compared to the stomach at 61 % among 110 cases of biopsy-proven GVHD. Our study corroborated these recent studies, supporting the importance of biopsies in the duodenum. Furthermore, our study showed discordance when gastric and duodenal biopsies were compared, which may suggest that GI GVHD may not be an evenly distributed disease in the upper GI tract. The mechanism for why the duodenum was a better site of biopsy for diagnosing GVHD is unclear and warrants further evaluation.

Our study showed similar sensitivities in diagnosing GVHD throughout the lower GI tract with good agreement among different parts of the colon. There were no cases of GVHD diagnosed in the right side of the colon without a positive biopsy in the left side of the colon. Previous studies have shown conflicting results regarding the best biopsy sites of the lower GI tract to yield a diagnosis of GVHD. Some studies have suggested that rectal biopsies are the best location of sampling in adults [2, 3] and children [4–6]. Snover and colleagues, however, have shown that a negative rectal biopsy may not rule out the diagnosis of GVHD [7]. In their series of 24 patients, if only rectal biopsies had been used, 5 of 13 cases of GVHD would have been missed. In another study of 24 patients, Thompson et al. found that a distal colon biopsy produced the best diagnostic yield at 82 %, but more invasive procedures, namely CLN with ileal biopsy or EGD combined with FS, increased the yield to about 94 % [8]. Similarly, in a study of 175 patients, Kreisel demonstrated that CLNs with terminal ileal biopsies produced the highest diagnostic yield [10].

In contrast, in a study of 12 patients with GVHD, there were no remarkable differences in endoscopic findings from the terminal ileum to the rectum in two patients who were able to tolerate a CLN [9]. For the remaining 10 patients who underwent an incomplete CLN, endoscopic findings were noted in the distal colon. The authors suggested that limited visualization by CLN (e.g., FS) might be adequate in diagnosing GVHD. In comparison with our study, however, this was a smaller cohort with higher grades of GVHD. In addition, only two patients were able to complete a CLN.

A more recent study from Japan also confirmed the poor yield of ileal biopsies [11]. Among 11 patients with confirmed GI GVHD, eight patients had positive biopsies in both the ileum and rectum. In no cases was there only a positive result in the ileum without involvement of the rectum. These investigators also found no correlation of GVHD diagnosis to the severity of disease. This study, however, did not examine any right-sided biopsies as we have in our cohort.

Given that patients with HSCT are often ill and unable to tolerate colonic preparation for CLN, limited endoscopic evaluation in diagnosing GVHD appears to be an appropriate strategy. Based on the results of our current study and the findings of these previous studies, a FS would be a reasonable initial choice of investigation for diagnosing GVHD with diarrhea. In our cohort, CLNs were commonly performed, which may suggest that this test is inappropriately utilized and unnecessary in assessing GVHD.

A strength of our study is the relatively large sample size of 46 patients with GVHD when compared to previous studies. We were able to compare a GVHD+ group to a GVHD− group in various parts of the GI tract with correlation of the diagnosis at multiple sites.

A limitation of our study includes the retrospective collection of data. Endoscopic findings (e.g., mucosal denudation previously reported in the literature) and reasons for completing endoscopic evaluation (e.g., N/V) were occasionally ambiguous and inferred from reports. Because of this inconsistent endoscopic reporting, correlating endoscopic and histologic results was difficult in our study, and we could only provide an analysis of normal findings or erythema noted on these endoscopic reports. These findings are not classic findings of GVHD and may relate to the relative lower grades of GVHD in our study compared to previous studies and possibly underreporting of typical endoscopic findings of GVHD. Similarly, the reasoning for why certain sites were chosen for biopsy was unknown. For example, esophageal biopsies on EGDs were more commonly obtained in the GVHD− group when compared to the GVHD+ (p = 0.04, Table 4) despite equal presence of esophagitis in both groups (p = 1.00, Table 1).

This study was also limited to one tertiary center and would require further larger studies involving multiple centers to provide stronger conclusions. Furthermore, various regimens of immunosuppression used in the patients of our cohort may vary from different centers, which may potentially influence the distribution of GVHD.

Future directions would include implementing a standardized protocol in the evaluation of patients with suspected GVHD, outlining appropriate endoscopies and their respective biopsy sites. A comparison could be made to see whether the yield of GVHD disease is improved with this new protocol. Ultimately, guidelines regarding GVHD assessment may be developed. Another important area of improvement is accurate reporting of endoscopic findings, clinical assessment, and reasoning for endoscopies, which were lacking in previous endoscopy reports in this study.

In conclusion, in our cohort of GI GVHD, we found that on EGDs, duodenal biopsies produced the highest yield in diagnosing GVHD. For lower endoscopies, sensitivities were similar throughout, suggesting that a FS is adequate in diagnosing GVHD.