Indirect imaging of the small intestine by radiologic investigations has remained dominant in clinical practice due to its wide availability, acceptable tolerance, and the ability to cover the entire gastrointestinal system, all of which are recognized as potential limitations of conventional endoscopy. The status quo changed with the introduction of video capsule endoscopy (PillCam, formerly M2A Capsule, Given Imaging, Yoqneam, Israel) in August 2000 [1]. Segments of the small bowel that could not be reached by standard endoscopy or push enteroscopy could now be visualized with the capsule endoscope. While capsule endoscopy has been compared with small bowel follow through, there is limited data comparing capsule endoscopy and more advanced radiological techniques [2–5]. We are aware of only two European studies from the same institution comparing capsule endoscopy and CT enteroclysis [6, 7]. No study has compared capsule endoscopy with double contrast barium–air enteroclysis. We wished to retrospectively compare the yield of the newer radiologic techniques with capsule endoscopy in a heterogeneous group of patients being investigated for small bowel disease.

Materials and methods

Patients

The study was approved by our institutional review board, with waiver of informed consent. From the electronic records of the gastroenterology and radiology departments we retrospectively identified all patients who had undergone both capsule endoscopy and an enteroclysis examination between January 2002 and December 2003. We excluded patients who had surgery (n = 4) or anti-inflammatory treatment (n = 3) in between the two studies. We identified 65 patients that met these criteria. The enteroclysis techniques were either fluoroscopic (n = 35) or CT enteroclysis (n = 30). The fluoroscopic double contrast enteroclysis as performed with combination of barium with methylcellulose (n = 17) or with carbon dioxide (n = 18). Unlike many prior studies, we included patients in whom the capsule endoscopy was equivocal or nondiagnostic (n = 6). We compared the diagnostic yield of the capsule study and each of the radiologic tests per indication and the type of lesion found.

Technique

Capsule endoscopy

Capsule endoscopy was performed using a swallowable capsule (PillCam SB, Given Imaging, Yoqneam, Israel). The capsule contains a metal dioxide silicon chip camera, a focusing lens, four white light emitting diodes, two silver oxide batteries, and a UHF band radio transmitter. Patients fasted for 12 h prior to the study. The capsule was swallowed with water. Two images were acquired per second with resolution of 256 × 256, magnification of 1:8, 1–30 mm depth of view, and 140 degrees field of view. Images were transmitted over the life of the battery, typically up to 8 h. Before swallowing the capsule, an array of sensors was taped to the patient’s abdominal wall in a predetermined pattern. The sensors were connected to a belt pack that contains a power cell, receiver, processor, and a hard drive that stores the images. After ingestion of capsule, patients were asked to refrain from taking medication for 2 h and to fast for 4 h. At the end of 8 h after capsule ingestion, the data received on the hard drive were downloaded to a work station (RAPID workstation, Given Imaging). Interpretation took between 30 and 90 min [8] and speed of image review varied between 1 and 50 per second.

Fluoroscopic enteroclysis

The details of the fluoroscopic enteroclysis techniques have been reported previously [9]. For fluoroscopic enteroclysis examinations we used biphasic contrast with barium, followed by either methylcellulose or carbon dioxide. For carbon dioxide enteroclysis we used 550 mL of barium and between 1 to 2 L of carbon dioxide until distal small bowel showed double contrast appearance on fluoroscopy.

All patients in this study had conscious sedation prior to enteroclysis studies. The patients were monitored throughout the procedure by a registered nurse. Fentanyl (Fentanyl citrate, Abbott Laboratories) was given intravenously at a dose of 25–50 μg. A titrated dose of midazolam (Versed, Roche Laboratories) was given in small aliquots to a total of 15 mg. In addition, metoclopramide (Reglan, A.H. Robbins Co., Richmond, VA) 10 mg was given intravenously as a pro-peristaltic agent. A higher dose of metoclopramide (up to 20 mg) was given in those already on narcotic analgesia or anti-cholinergic mediation.

CT enteroclysis

All examinations were performed on a multidetector row CT with four or 16 channels of detectors (Mx8000 or IDT, Philips Medical Systems, Cleveland OH). Section (slice) widths of 2 and 3.2 mm with reconstruction interval of 1 and 1.3 mm were used for the 16- and 4-slice scanners, respectively. CT enteroclysis images were reviewed on a workstation in both the axial and reformatted coronal and sagittal planes. Neutral enteral contrast CT enteroclysis (n = 10) was performed using water. Intravenous contrast was given with this technique. Positive enteral contrast CT (n = 20) was performed using 12% diatrizoate meglumine (Gastrografin, Bracco, Princeton, NJ). No intravenous contrast was given with positive enteral contrast CT enteroclysis. Site of tube placement, volumes, and rate of enteral and intravenous contrast infusion were as per previous report [9].

Data analysis

Patient demographic information and the original interpretation of the enteroclysis examination were retrieved from hospital and/or radiology information systems. Pertinent surgical and pathological reports were also documented. For each patient, the enteroclysis examinations were interpreted by one of two gastrointestinal radiologists with over 25 years post-fellowship experience. Capsule endoscopy data was interpreted by one of six board certified and fellowship trained gastroenterologists with a minimum of 2 years experience in capsule endoscopy interpretation. Both groups of clinicians interpreted the studies in light of available clinical and imaging information. Findings were considered doubtful if the only findings were focal erythema, luminal blood without focal lesion, or up to three superficial ulcers. These findings were not included in determining diagnostic yield.

Findings that were documented were categorized into vascular lesions such as angioectasia, inflammatory lesions such as Crohn disease, tumor, strictures (such as from adhesions), and a miscellaneous group that included celiac disease. Differences in yield of the capsule and radiological study were analyzed using Fisher’s exact test (2-tailed). Statistical analysis was performed using Matlab Statistics 5.2 (MathWorks Inc., Natick, MA).

Results

The main indications for the tests were obscure gastrointestinal bleeding (n = 37), suspected Crohn disease (n = 17), chronic unexplained abdominal pain (n = 8), suspected small bowel tumor (n = 2), and suspicion for celiac disease (n = 1). The mean age was 54.2 years (range 9–88 years); 45 patients were female. Double contrast fluoroscopic enteroclysis studies were performed in 35 patients using methylcellulose (n = 17) or carbon dioxide (n = 18). CT enteroclysis techniques were performed in 30 patients with either positive (n = 20) or neutral (n = 10) enteral contrast. CT enteroclysis examinations were grouped together. Enteroclysis occurred before capsule endoscopy in 55 of 65 patients (85%). The time interval between the capsule and enteroclysis studies was less than 3 months in 54 patients and between 3 and 6 months in 11 patients. All surgical and pathological gold standard data was available within 3 months of the studies.

The capsule did not reach the cecum in six patients (9%) prior to the end of battery life (typically 8 h). In one patient, the capsule was retained permanently and was removed surgically. In four patients, the capsule views were suboptimal due to excess luminal fluid.

Diagnostic yield per indication

The diagnostic yields of capsule endoscopy, methylcellulose enteroclysis, carbon dioxide enteroclysis, and CT enteroclysis are given in Table 1. There was no statistical difference in the diagnostic yields of CT enteroclysis (9/30) and capsule endoscopy (8/30) (P = 0.78). Capsule endoscopy had a significantly higher diagnostic yield (8/17) compared to methylcellulose barium enteroclysis (1/17) (P = 0.02). There was no significant difference in yield of positive finding with carbon dioxide barium enteroclysis (12/18) compared to capsule endoscopy (10/18) (P = 0.73). Carbon dioxide enteroclysis had a high yield in showing small ulcers in patients with obscure gastrointestinal yield (Fig. 1).

Table 1 Diagnostic yields of enteroclysis studies and capsule endoscopy per indication
Full size table
Fig. 1
figure 1

A 68-year-old female with obscure gastrointestinal bleeding, with history of nonsteroidal anti-inflammatory drug NSAID use. (A) Enteroclysis with barium and carbon dioxide showed linear mucosal abnormalities (arrowheads) interpreted as ulcers. (B) Capsule endoscopy performed 2 days later demonstrated linear ulcers (arrow) with raised margins. NSAID enteropathy was diagnosed and bleeding resolved when NSAID usage ceased.

Full size image

CT enteroclysis showed numerous adhesions in three patients with chronic abdominal pain following prior surgery. Capsule endoscopy was negative in these patients. For other indications, both CT enteroclysis and barium carbon dioxide enteroclysis were equivalent to capsule endoscopy. However, methylcellulose had poor diagnostic yields for both obscure gastrointestinal bleeding and suspected Crohn disease.

Diagnostic yield per etiology of lesions

The diagnostic yields per etiology are given in Table 2. The numbers in each subgroup were too small to make satisfactory statistical comparisons. Nevertheless, it appears that methylcellulose enteroclysis was suboptimal compared to capsule endoscopy in assessing both vascular and inflammatory lesions.

Table 2 Diagnostic yields of enteroclysis studies and capsule endoscopy per etiology
Full size table

CT enteroclysis and carbon dioxide enteroclysis were not as good as capsule endoscopy in diagnosing vascular lesions, particularly angioectasia. However, these radiologic tests were as good as or better than capsule endoscopy in assessing small inflammatory ulcers from Crohn disease or nonsteroidal anti-inflammatory drug (NSAID) use. This finding was in part due to equivocal capsule studies, or the capsule not reaching the cecum before the end of battery life. In addition, although the capsule endoscopy had a high sensitivity, its specificity was low (Fig. 2).

Fig. 2
figure 2

A 21-year-old male with abdominal pain and anemia. (A) Enteroclysis with barium and carbon dioxide showed large Meckel’s diverticulum (arrow) with wide neck and small pools of barium (arrowheads) suggesting presence of ulcers. (B) Capsule endoscopy performed 3 weeks earlier demonstrated ulcers (arrows), but the wide neck of Meckel’s diverticulum (arrowhead) was misinterpreted as normal ileocecal valve and patient was considered to have active Crohn disease. Surgical pathology confirmed Meckel’s diverticulum with ulceration of Brunner’s glands.

Full size image

In detecting tumors, capsule endoscopy showed small multifocal carcinoid tumors in one patient which was missed by CT enteroclysis (Fig. 3). On the other hand, CT enteroclysis showed serosal metastases which were missed by the capsule study.

Fig. 3
figure 3

A 55-year-old female with overt gastrointestinal bleeding. Multiple subcentimeter nodules in proximal jejunum (arrowheads) were confirmed to be carcinoid at surgery. CT enteroclysis with neutral enteral and intravenous contrast (not shown) was normal.

Full size image

Discussion

There is considerable data comparing capsule endoscopy with small bowel follow through. Two studies (total n = 60) assessing patients with obscure gastrointestinal bleeding found significantly lower yield of small bowel follow through (3–20%) compared to that of capsule endoscopy (45–55%) [10, 11]. Two other studies (total n = 53) investigating obscure gastrointestinal bleeding found diagnostic yields of 55–66% with capsule endoscopy [12, 13]. In these studies, all patients had prior normal small bowel follow through. Seven studies that looked at patients with suspected or established Crohn disease found the yield of small bowel follow through (0–33%) significantly lower than that of capsule endoscopy (43–71%) [14–20]. Small bowel follow through was also not particularly helpful in excluding strictures that may potentially trap the capsule. Capsule retention occurred in a total of 20 patients in four studies, all of whom had prior normal small bowel series [21]. Thus, the role of small bowel follow through in the era of capsule endoscopy is questionable.

We wished to see if the more sensitive radiologic techniques of fluoroscopic and CT enteroclysis would have a higher yield in patients who may also have capsule endoscopy. There is limited published data on such comparisons. In common with virtually all publications on capsule endoscopy [4, 6, 10, 12–17, 22–48] we assessed the diagnostic yield. To measure the true sensitivity and specificity a gold standard is required. In most patients with suspected Crohn disease or unexplained gastrointestinal bleeding, surgical pathology or intraoperative enteroscopy, which could be used as gold standard, are inappropriate.

CT enteroclysis vs. capsule endoscopy

In our study, capsule endoscopy showed vascular lesions in three patients vs. only one with CT enteroclysis. Capsule endoscopy also showed a 7 mm primary carcinoid tumor that was missed on CT enteroclysis. These findings suggest that capsule endoscopy is superior to CT enteroclysis for detecting flat mucosal lesions. On the other hand, CT enteroclysis was superior for assessing bowel wall lesions, such as adhesions, enteric fistula, and serosal metastases. Capsule endoscopy did not detect any lesions of these types, while they were seen in four patients with CT enteroclysis.

Our study findings are mirrored by another study comparing capsule endoscopy and CT enteroclysis. In this study of 56 patients with known Crohn disease [49], 15 patients (27%) did not have capsule endoscopy because preceding CT enteroclysis showed intestinal strictures. In the remaining 41 patients, capsule endoscopy was found to be superior to CT enteroclysis in detecting proximal small bowel lesions (yields of 61% and 29%, respectively) but the two tests were equivalent in detecting distal ileal disease (yields of 58% and 49%, respectively). Many of the patients in this study had early Crohn disease as manifested by the mean symptom duration of 12 months. Looking at the study more closely, the major component of the high yield of capsule endoscopy was due to detection of minor lesions, such as apthous ulcers and erythema. In the assessment of major lesions, such as fissures, fistula, and stenosis, CT enteroclysis had a higher yield than capsule endoscopy (distal ileal lesions in 13 patients and 10 patients, respectively). Capsule endoscopy has been shown to have a poor yield in patients with chronic abdominal pain [50, 51] and our results confirm this, with the capsule study being negative in all eight patients with this indication.

Methylcellulose enteroclysis vs. capsule endoscopy

Two studies assessing known or suspected Crohn disease patients showed diagnostic yields of 71–93% for capsule endoscopy compared to 26–33% for enteroclysis [17, 52]. Our results confirm the lower diagnostic yield of methylcellulose enteroclysis. In our heterogeneous group, this method of enteroclysis had the lowest yield (8%) and was significantly lower than capsule endoscopy (47%) (P < 0.05). One of the lesions missed by this technique of enteroclysis included a 2 cm carcinoid tumor that was detected on capsule study.

Barium–carbon dioxide enteroclysis vs. capsule endoscopy

Double contrast barium–carbon dioxide is not commonly used in the USA but is popular in Japan. It has been considered to be a good technique for detecting superficial mucosal pathology, such as ulcers (Fig. 1) [53–55]. Carbon dioxide may be preferred to air as the negative enteral contrast because of a lower incidence of abdominal cramping.

There are no previous studies comparing carbon dioxide enteroclysis and capsule endoscopy. In our subgroup of patients with both tests, the yield of barium–carbon dioxide enteroclysis was equivalent to capsule endoscopy in both obscure gastrointestinal bleeding and in suspected Crohn disease. Ulcers from Crohn disease and NSAIDs were seen in 10 of 18 patients evaluated with this technique. In fact, this technique had a higher, but not significantly better, overall yield than capsule endoscopy (67% vs. 56%, P = 0.73). However, barium–carbon dioxide enteroclysis is more difficult to perform than methylcellulose enteroclysis, as insertion of the optimal amount of the gas is difficult to determine. It is also an uncomfortable procedure and ideally would require patient sedation. As a result, it is unlikely to be considered as a first line test. In sites with experience in this technique, we recommend it as a second line test in those with suspected obscure gastrointestinal bleeding or Crohn disease, if capsule endoscopy is equivocal or is contraindicated. We postulate two possible reasons for the difference in sensitivity of the two fluoroscopic enteroclysis techniques. The higher density of methylcellulose may reduce the visibility of a small pool of barium in a superficial ulcer. During its infusion, the more viscous methylcellulose may wash out barium from a mucosal lesion (Fig. 4) [56].

Fig. 4
figure 4

A 38-year-old male with obscure gastrointestinal bleeding. (A) Early spot view from enteroclysis with barium and methylcellulose showed subtle filling defect in distal ileum (arrow). (B) Later image after further infusion of methylcellulose did not show the filling defect which was considered artefactual. Enteroclysis was read as normal. (C) Capsule endoscopy performed one month later demonstrated submucosal mass (arrow), confirmed to be carcinoid. It is possible that methylcellulose “washed off” the barium from the lesion.

Full size image

Limitations of capsule endoscopy include the inability to perform biopsies or therapeutic procedures, capsule retention, assessment of trivial lesions, poor localization of lesions, and cessation of battery power prior to reaching the cecum. Capsule retention has been seen from 1% of patients investigated for obscure gastrointestinal bleeding to 8% of patients with Crohn disease [21]. Thus, capsule endoscopy generally is contraindicated in patients with symptoms of small bowel obstruction, long-term nonsteroidal anti-inflammatory drug use, and small bowel surgery, though anecdotal reports have challenged some of these guidelines [57, 58]. In previous studies of capsule endoscopy, small bowel ulcers have been thought to represent Crohn disease. However, 14% of asymptomatic volunteers and another 7% of those on placebo medication show at least one such ulcer [59]. Thus, earlier studies may have overinflated the diagnostic yield of capsule endoscopy [15, 60, 61]. In addition, as noted earlier capsule endoscopy does not reliably assess the small bowel wall. On the other hand, enteroclysis is a difficult technique to implement. The required logistics include a dedicated hydraulic pump, nursing support, and, in the case of CT enteroclysis, the immediate availability of the CT scanner following enteral infusion. Nasoduodenal intubation is an otherwise unpleasant procedure which is best performed under conscious sedation.

We are aware of several limitations of the study, including its small numbers and retrospective nature. The capsule and enteroclysis studies were interpreted without blinding to other tests or clinical information. A limitation of our study is verification bias favoring capsule endoscopy, as there is no pathologic proof for most cases of angioectasia. Intraoperative enteroscopy, which may be considered the gold standard examination, is not the standard of care for most common small-bowel capsule endoscopy findings.

Our study has shown that more advanced radiologic techniques such as CT enteroclysis could complement capsule endoscopy in investigating small bowel lesions. Further studies are required to determine which indications are best investigated by enteroclysis and when. The roles of other newer radiologic tests such as CT enterography or MR enteroclysis (which does not have the risk of radiation) also need to be clarified by future studies. In particular, CT enterography deserves further research as it is easier to perform than enteroclysis, and does not require nasojejunal incubation or conscious sedation.