Early detection of acute graft-versus-host disease by wireless capsule endoscopy and probe-based confocal laser endomicroscopy: results of a pilot study

Introduction

Graft-versus-host disease (GVHD) is the major source of morbidity and mortality after allogeneic haematopoietic stem cell transplantation (allo-HSCT).¹ The diagnosis of acute GVHD is based upon analysis of clinical and laboratory criteria in the appropriate clinical context after excluding other causes.²

While severe acute gastrointestinal GVHD (GI-GVHD) is likely the most serious and most potentially fatal form of GVHD, gastrointestinal symptoms are rather common in allo-HSCT patients, but not specific for GI-GVHD. Therefore, endoscopic examination with biopsy sampling of the mucosa is usually performed to confirm the diagnosis and exclude differential diagnoses. However, this strategy is hampered by some major limitations. Indeed, GI-GVHD is characterized by a patchy distribution of lesions, which can either affect a short segment or involve the whole gastrointestinal tract. Therefore, available endoscopic examinations such as oesophagogastroduodenoscopy (OGD) and colonoscopy have a limited diagnostic yield since they can only explore a short segment of the small bowel, an area which is frequently involved in GI-GVHD.^3-7 Moreover, usual endoscopic examinations are relatively invasive in such fragile patients. Also, one should bear in mind that endoscopic abnormalities are usually found only in a minority of cases (16–32%)⁸ and are usually nonspecific. Thus, additional biopsies are necessary but are deemed to be invasive, especially due to the high risk of bleeding in case of thrombocytopenia.

In the last decade, several novel imaging modalities have been developed to offer extensive and noninvasive explorations of the small bowel as well as alternatives to conventional histology during endoscopic procedures. One method is wireless capsule endoscopy (WCE), which allows a noninvasive examination of the entire small bowel. WCE has been successfully used for the diagnosis of anaemia and occult bleeding, inflammatory bowel disease, and recognition of intestinal tumours.⁹ Focusing on patients treated with allo-HSCT, some studies have demonstrated the ability of WCE to detect intestinal lesions suspected to be related to GI-GVHD.⁶, ^10-12 Moreover, these studies reported an excellent negative predictive value to rule out the diagnosis of GI-GVHD.^{6, 10} Another interesting method is confocal laser endomicroscopy, a high-resolution imaging modality allowing access to in vivo histology at the subcellular level during ongoing endoscopy. Confocal laser endomicroscopy, either using an endoscope-based or a probe-based technology, aims at decreasing the number of standard biopsies and their associated risks, by providing real-time in situ microscopy.¹³

Thus far, only a small single pilot study assessed integrated confocal laser endomicroscopy (iCLE) performances in GI-GVHD patients showing good sensitivity and specificity as compared with conventional histology.¹⁴ However, none of the currently available studies using WCE, iCLE, or probe-based confocal laser endomicroscopy (pCLE) attempted to predict the risk of acute GI-GVHD at an early stage in asymptomatic patients.⁶, ^10-12, ¹⁴

Therefore, this pilot prospective study aimed to evaluate the diagnostic value of both WCE and pCLE to detect early lesions suggestive of acute GI-GVHD and to determine whether pCLE and WCE could detect early lesions of GI-GVHD independently of the presence or absence of symptoms and to assess the diagnostic yield of each method separately and in combination.

Methods

Endoscopic examination

Given that the aim of this study was to determine whether pCLE and WCE could detect early lesions of GI-GVHD before the onset of symptoms and whether GVHD occurs at a median of 21 days after allo-HSCT,¹⁶ we chose to perform endoscopic examinations as soon as possible after haematopoietic recovery in patients free of GVHD. Consequently, per protocol, endoscopic examinations were performed between day 21 and day 28 following allo-HSCT. No patient was on nonsteroidal anti-inflammatory drugs (NSAID) at time of endoscopic examinations. Patients answered questionnaires using a 10-cm visual analogue scale about their tolerance of the procedure, with 10 being perfect tolerance and 0 suggesting worse acceptability.

Conventional endoscopy

Endoscopic procedures consisted in a standard upper GI endoscopy (OGD) and a rectosigmoidoscopy, which were performed during a single endoscopic session. Examinations were carried out by a single experienced endoscopist (EC), without general anesthaesia. Patients received 100 mg hydroxyzine (Renaudin, Itxassou, France) orally 1 h prior to the examinations in order to improve the tolerance of the procedures. OGD and rectosigmoidoscopy were performed using a standard gastroscope (EG530) and a standard colonoscope (EC530), respectively (Fujinon, Japan).

Wireless capsule endoscopy

The WCE system consists of three major components: an ingestible videocapsule, a data recorder (together with a recorder belt and a sensor array), and a computer workstation. The small bowel capsule measures 11 × 26 mm and acquires video images from one end of the device (2 frames/s). The capsule transmits the acquired images via a digital, radio frequency, communication channel to the data recorder unit strapped to the patient's waist. The data recorder is an external receiving/recording unit that receives the data transmitted by the capsule. Upon completion of the examination, the physician transfers the accumulated data from the data recorder to the computer workstation for processing and interpretation. For the purpose of this study, we used standard procedures as previously described.¹⁷ For bowel preparation, patients had a 12-h overnight fast and drank 1 l of a polyethylene glycol/electrolyte solution 2 h prior to swallowing the capsule. Then, they were connected to the recording device and swallowed the WCE (Given Imaging, Yoqneam, Israel). The capsule recorded images at a rate of 2 frames/s for a period of 8 h. WCE images were interpreted by a single experienced endoscopist (VL) who knew that the patient had received allo-HSCT, but was blinded to other clinical and biological data. Visual grading of GI-GVHD potential lesions was scored according to the Brand criteria¹⁸ as follows (Figure 1): grade 0, normal; grade 1, mild: loss of vascular marking, and/or mild focal erythema; grade 2, moderate: moderate/severe and/or diffuse erythema, or nodularity; grades 3 and 4, severe: grade 3 showing erosions and/or friable mucosa and grade 4 showing desquamation, ulceration, and denuded mucosa.

Probe-based confocal laser endomicroscopy

Probe-based confocal laser endomicroscopy is a new endoscopic imaging technique, which adapts optical microscopy to ‘live’ digestive endoscopy. The system (Cellvizio; Mauna Kea Technologies, Paris, France) was installed on an independent rod dedicated to endomicroscopy. It consists of a laser unit, an ultra high-definition endomicroscopic miniprobe, and a computer with a specific software. The laser wave length was set up at 488 nm with low power (15 mW). The endomicroscopic miniprobe is a flexible microscope of 2.6 mm in diameter and 3 m in length. Images were gathered at a rate of 12 images/s. Lateral resolution was 1.4 µm and the field of view was 240 µm. pCLE was performed during the same sessions that OGD and rectosigmoidoscopy. Predetermined areas (i.e. duodenum, sigmoid, and rectum) were systematically explored following conventional endoscopic examination and prior to taking biopsies for conventional routine histology. The pCLE procedure was performed as follows: firstly, 5 ml of 10% fluorescein sodium (Novartis Pharma, Paris, France) was injected intravenously at the beginning of the OGD procedure, and served as a contrast agent to visualize crypt and vessels architecture.¹⁹ Secondly, the confocal mini probe was threaded through the operative channel of the endoscope. With its distal end resting gently on the tissue, it produced images of the digestive mucosa at a subcellular level. All pCLE images were recorded for post-procedure interpretation using specific software (Cellviewer; Mauna Kea Technologies).pCLE images were interpreted by the same endoscopist who performed OGD, rectosimgoidoscopy, and pCLE procedures (EC), independently of other examinations, immediately following acquisition. The definition of normal duodenum and colon by pCLE was based upon the previously described classifications;²⁰ in contrast, GI-GVHD was defined by the presence of one or more abnormal criteria, either in the duodenum or in the colon. These criteria were as follows (Figure 2, Supplementary Movie File, available online): (a) abnormal microvessel network; (b) mild increase in fluorescein intensity in the lamina propria; (c) distorted villi (duodenum) or crypts (colon); (d) major increase in fluorescein intensity in the lamina propria; and (e) major architectural changes defined by the destruction of villi (duodenum) or crypts (colon). The severity of GI-GHVD was graded as follows: mild when one criterion a or b was present, as moderate either when criterion c was present or a combination of criteria a and b were met, and as severe when criteria d or e were present.

Results

Discussion

This pilot study demonstrates the feasibility, tolerance, and safety of a combined minimally invasive endoscopic approach in patients at risk of GI-GVHD after allo-HSCT. Both WCE and pCLE were able to detect early mucosal lesions in asymptomatic allo-HSCT patients, albeit with different diagnostic performances. Despite the relatively small number of included patients, pCLE was likely to be sensitive while WCE had good specificity. Moreover, the results obtained with these novel technologies were better correlated with the classical modified Gluckberg severity score compared to the results obtained with conventional histology.

Results from the current pilot prospective analysis provide new insights with respect to several aspects: (i) it combined two novel minimally invasive imaging modalities (i.e. WCE and pCLE), while previous studies assessed the diagnostic yield of WCE or pCLE separately; (ii) WCE and pCLE were performed early after allo-HSCT with a prediction rather than a diagnostic goal; and (iii) the majority of the current patients (87%) were asymptomatic at time of examination (in contrast to previous studies, which enrolled patients with diarrhoea or abdominal pain). However, the size of the current study was limited and did not allow for depicting reliable and final conclusions as for sensitivity, specificity, PPV, and NPV. These calculations were only performed for exploratory purposes. WCE had a sensitivity of only 50% for the detection of early lesions of GI-GVHD. Although WCE usually allows visualization of the entire small bowel, this lack of sensitivity is not surprising given the early timing of WCE and the subtle mucosal abnormalities suggestive of GI-GVHD contrasting with more severe lesions reported in symptomatic patients.^{6, 10, 12} Nevertheless, WCE appeared to be more sensitive than standard histology to detect early stages/asymptomatic acute GI-GVHD. One major issue with WCE is that lesions found in GVHD patients are quite nonspecific. Indeed, erythema, denudation, and small ulcers have been reported in patients receiving NSAID and even in some healthy subjects.²² In addition, it may be difficult to discriminate between lesions of GI-GVHD and lesions induced by opportunistic infections.²³ However, in the specific context of allo-HSCT, abnormalities detected by WCE could very accurately predict GI-GVHD while their absence cannot rule out this possibility.

In contrast to WCE, pCLE appeared as the most sensitive technique to detect early acute GI-GVHD lesions. Its performance is also better than standard histology. Conversely, pCLE showed a relatively poor specificity for the diagnosis of acute GI-GVHD. Only a single previous study, Bojarski et al.,¹⁴ reported both high sensitivity and specificity for CLE as compared to standard histology; however, there were differences from our study on three important points. The first is that Bojarski et al. performed their study in symptomatic patients, with diarrhoea being the predominant symptom, while no patient had GI symptoms at time of examination in our study. The second difference concerns the technique used for endomicroscopy, not only in terms of technology but also the type of contrast agent used and access route. Indeed, these authors used iCLE onto a colonic mucosa prestained with both intravenously administered fluorescein combined with topical application of acriflavin. Compared to our technique, the technique used by Bojarski et al. might increase the number of subcellular details since iCLE is slightly superior to pCLE in terms of lateral resolution and acriflavin allows for studying the nuclei. However, our study focused on the vascular compartment and dynamic changes seen with videosequences of the mucosa obtained with pCLE. In addition, we examined both the duodenum and the colon (including rectum) while Bojarski et al. only performed colonoscopies. However, whether biopsies obtained from the small intestine are superior to the ones obtained from the colon to diagnose acute GI-GVHD remains to be determined. The third difference is the gold standard that was used to establish the diagnostic value of CLE. In the study from Bojarski et al., iCLE was compared to standard histology while we compared pCLE to the Glücksberg scoring system. Our choice was driven by the fact that the Glücksberg scoring system is a common clinico-biological assessment widely used in routine practice by the allo-HSCT community to diagnose acute GVHD and manage patients. Also, we thought to establish the different correlations using the Glücksberg overall grade, because it is likely that intestinal lesions may reflect the general immune alloreactive status of the patient.

An important limitation is that the confocal classification used in this study has not been externally validated but has been adapted from a previous international classification describing CLE patterns in inflammatory bowel diseases.²⁰ Indeed, previous studies assessing CLE in GVHD patients used acriflavin in order to identify the apoptotic cells associated with GVHD. However, acriflavin is not approved for in vivo use in France and in many countries, thus could not be applied in our patients. Instead, we chose to focus our study on other parameters such as the vascular compartment and dynamic changes seen with videosequences of the mucosa obtained with pCLE. Indeed strong data suggest that neovascularization is of major importance in the pathophysiology of acute GVHD.²³ Also, most pCLE classifications, such as the Miami classification²⁰ used for the diagnosis of other precancerous or inflammatory lesions, did not try to superimpose pCLE and standard histology criteria but rather developed novel diagnostic criteria such as extracellular intensity of fluorescein, as in our study. However, in this study we did not use specific software tools, which might be helpful to perform quantitative analysis of the brightness of the lamina propria in order to provide more objective analysis. Last but not least, discordance between the Glücksberg scoring system and standard histology has been reported,⁴ emphasizing the need for novel staging criteria. In our study, it is interesting to note that there are discrepancies between the pCLE and standard histology results (Table 5). For instance, standard histology was not able to predict GVHD onset in three grade I patients without clinical sign of GI-GVHD, whereas they were correctly identified by pCLE, suggesting that these patients probably had a very early GI-GVHD. However, one must acknowledge that the so-called ‘screening’ biopsies as performed in the current study are not standardized so far, and previous histological criteria were developed for diagnosis of established GVHD but not for prediction as aimed here. Another limitation is that we did not directly compare physical and optical biopsies, since it is almost impossible to target exactly the same spot due to the small field of view of pCLE. However, physical biopsies were performed in the same macroscopic area, within approximately 5–10 mm range from the pCLE probe apposition.

Another potential explanation for the discordance between pCLE and standard histology is that pCLE screens larger areas of mucosa than standard histology and combines functional and morphological criteria rather than morphological criteria alone. Indeed, data suggest that neovascularization is a major prognostic factor in the setting of acute GVHD.²⁴ Regarding the discriminating ability between grades 0–I and grades II–IV, which is a critical clinical issue in terms of acute GVHD management decision making, it is noteworthy that four out of five patients with grades II–IV GVHD were accurately diagnosed by WCE combined with pCLE.

Obviously, standardization, and validation testing are still needed. However, these results suggest that a new diagnostic algorithm consisting in performing WCE first might be proposed. In case of abnormalities and after excluding other causes such as concomitant intake of NSAIDs, diagnosis of GI-GVHD could be accepted with a high degree of confidence and management adapted accordingly. If WCE is normal then we suggest performing pCLE without biopsy to further adapt the treatment strategy. In our series, such strategy allowed for an accurate staging in all but one case, while adding histology did not improve the diagnostic yield. Early detection of acute GVHD lesions prior to symptom onset is crucial. Acute GVHD remains the major source of morbidity and mortality after allo-HSCT.¹ So far the diagnoses rely almost entirely on the presence of clinical symptoms and should be confirmed by biopsy of the involved target organ.² Early detection of GVHD prior to symptom onset has important therapeutic consequences including more stringent monitoring and/or preventive care. If several acute GVHD biomarkers are promising, they remain to be validated.²⁵ WCE and especially pCLE represent a promising alternative to these biomarkers for early detection of acute GI-GVHD.

Finally, the current data established the safety and tolerance of these new endoscopic approaches in such fragile patients. Both WCE and pCLE were well tolerated and no significant complications occurred, especially following fluorescein administration, even if pCLE added 5–10 minutes to the standard endoscopic examination. As a matter of fact, larger controlled and multicentre series will be needed to establish the routine use of these novel endoscopic procedures. Also, such studies should pay a special attention to interpretation of early mucosal/vascular lesions detected by pCLE, which can be challenging as our current study did not attempt to assess the interobserver agreement of pCLE criteria. Despite these limitations, we believe that the current pilot findings can pave the way for a new diagnostic algorithm combining WCE first and pCLE in case of a negative WCE result in order to detect early lesions of acute GI-GVHD, allowing for improved therapeutic decisions in high-risk allo-HSCT patients.

Funding

This work was supported by the Région Pays de Loire, the Association pour la Recherche sur le Cancer (ARC; grant no. 3175, to MM), the Fondation de France, the Fondation contre la Leucémie, the Agence de Biomédecine, the Association Cent pour Sang la Vie, the Association Laurette Fuguain, and the IRGHET. Our transplant programmes are supported by several grants from the French National Cancer Institute (PHRC, INCa, to MM). FM was supported by educational grants from the Association for Training, Education and Research in Hematology, Immunology and Transplantation (ATERHIT).