AIM: To examine the predictive factors of capsule endoscopy (CE) completion rate (CECR) including the effect of inpatient and outpatient status.

METHODS: We identified 355 consecutive patients who completed CE at Rush University Medical Center between March 2003 and October 2005. Subjects for CE had either nothing by mouth or clear liquids for the afternoon and evening of the day before the procedure. CE exams were reviewed by two physicians who were unaware of the study hypotheses. After retrospective analysis, 21 cases were excluded due to capsule malfunction, prior gastric surgery, endoscopic capsule placement or insufficient data. Of the remaining 334 exams [264 out-patient (OP), 70 in-patient (IP)], CE indications, findings, location of the patients [IP vs OP and intensive care unit (ICU) vs general medical floor (GMF)] and gastrointestinal transit times were analyzed. Statistical analysis was completed using SPSS version 17 (Chicago, IL). Chi-square, t test or fisher exact-tests were used as appropriate. Multivariate logistic regression analysis was used to identify variables associated with incomplete CE exams.

RESULTS: The mean age for the entire study population was 54.7 years. Sixty-one percent of the study population was female, and gender was not different between IPs vs OPs (P = 0.07). The overall incomplete CECR was 14% in our study. Overt obscure gastrointestinal bleeding (OGB) was a significantly more common indication for the IP CE (P = 0.0001), while abdominal pain and assessment of IBD were more frequent indications for the OP CE exams (P = 0.002 and P = 0.01, respectively). Occult OGB was the most common indication and arteriovenous malformations were the most common finding both in the IPs and OPs. The capsule did not enter the small bowel (SB) in 6/70 IPs and 8/264 OPs (P = 0.04). The capsule never reached the cecum in 31.4% (22/70) of IP vs 9.5% (25/ 264) of OP examinations (P < 0.001). The mean gastric transit time (GTT) was delayed in IPs compared to OPs, 98.5 ± 139.5 min vs 60.4 ± 92.6 min (P = 0.008). Minimal SB transit time was significantly prolonged in the IP compared to the OP setting [IP = 275.1 ± 111.6 min vs OP = 244.0 ± 104.3 min (P = 0.037)]. CECR was also significantly higher in the subgroup of patients with OGB who had OP vs IP exams (95% vs 80% respectively, P = 0.001). The proportion of patients with incomplete exams was higher in the ICU (n = 7/13, 54%) as compared to the GMF (n = 15/57, 26%) (P = 0.05). There was only a single permanent SB retention case which was secondary to a previously unknown SB stricture, and the remaining incomplete SB exams were due to slow transit. Medications which affect gastrointestinal system motility were tested both individually and also in aggregate in univariate analysis in hospitalized patients (ICU and GMF) and were not predictive of incomplete capsule passage (P > 0.05). Patient location (IP vs OP) and GTT were independent predictors of incomplete CE exams (P < 0.001 and P = 0.008, respectively).

CONCLUSION: Incomplete CE is a multifactorial problem. Patient location and related factors such as severity of illness and sedentary status may contribute to incomplete exams.

In recent years, capsule endoscopy (CE) has emerged as a sophisticated and widely used tool to evaluate small bowel pathology. It has applications in Crohn’s disease, celiac disease, Peutz-Jeghers syndrome, familial adenomatous polyposis, and hereditary polyposis syndromes[1-4]. CE is also particularly useful for the investigation of obscure gastrointestinal bleeding (OGB)[5,6], especially if upper and lower endoscopies fail to reveal the underlying pathology and a small bowel source is suspected. Locating the source and treating OGB has long been a challenge for physicians. Identification of the bleeding site is the key step not only to diagnose but also to provide effective treatment. During a brisk overt hemorrhage, the rapidity at which the bleeding site is found is critical. CE is therefore utilized often in patients hospitalized for bleeding and anemia. The hospital setting, however, may not be ideal in obtaining optimal results with CE. In-patients tend to be sedentary and are more prone to delayed transit or ileus, which may prolong gastric and small bowel transit times. Therefore, we hypothesized that in-patient (IP) CE is more often incomplete as compared to out-patient (OP) CE. The two previously published European studies indirectly examining this question have compared IP vs OP CE completion rates (CECR)[7,8], but both studies have limitations: both used a prokinetic agent and/or bowel preparation to increase CECR and neither of these interventions is part of the standard CE protocol at United States institutions. Furthermore, the patient numbers were small and the studies did not examine the effects of inpatient location (i.e., an intensive care unit (ICU) vs a general medical floor (GMF) setting). In this study, we aimed at overcoming these limitations by analyzing all consecutive IPs and OPs, studying a larger IP group, doing a sub-group analysis to compare the ICU and GMF settings, without the use of prokinetic and bowel preparation agents.

This study was conducted in compliance with ethical guidelines and was approved by Rush University Institutional Review Board. All CE examinations done at Rush University Medical Center between March, 2003 and October, 2005 were identified retrospectively. A total of 355 CE exams, which were performed using a Given PillCam^TM small bowel (SB) capsule, were reviewed using the Given Rapid^® Access software. Patients with capsule malfunction (n = 6), prior gastric surgery (n = 3), endoscopic capsule placement (n = 3) and studies lacking sufficient data for analysis were excluded (n = 9; 8 OP, 1 IP). In the remaining 334 exams (264 OP, 70 IP), patient characteristics such as age, gender, gastric retention of the capsule, completeness of the examination (whether or not the capsule reached the cecum), gastric and small bowel transit times, indications and findings were recorded.

Subjects for capsule endoscopy had either nothing by mouth (except their medications) or clear liquids (excluding any red colored foods) for the afternoon and evening of the day before the procedure. The CE exam was delayed if subjects used iron supplementation within the past 5 d prior to the procedure.

CE exams were reviewed by two physicians. For the majority of cases, gastroenterologists who read the CE exams did not have any information regarding patient location. The hypotheses of the study were conceived after the exams were read and gastroenterologists were unaware of the study hypotheses.

Overt gastrointestinal bleed (GIB) was defined as having melena and/or hematochezia with a hemoglobin < 10 g/dL. Occult obscure GIB was defined as subjects who had a positive fecal occult blood test or complaints of intermittent black bowel movements, and who had both colonoscopy and upper endoscopy without any source of bleeding identified. Accurate SB transit time (SBTT) were obviously not available in those subjects in whom the capsule stopped recording before reaching the colon. In this circumstance the minimum SBTT, which reflects the minimum time the capsule spent in the small bowel, was calculated using the formula, 480 min - the gastric transit time (GTT) (min) = minimum SBTT. Four hundred and eighty minutes was used as this has the generally accepted battery life of the Given PillCam™ SB capsule at the time of the CE exams.

Statistical analysis was completed using SPSS version 17 (Chicago, IL). Chi-square, t test or fisher exact-test was used as appropriate. Multivariate logistic regression analysis was used to identify variables associated with incomplete CE exams. A multivariate logistic regression model was constructed using forward stepwise method with capsule reaching the cecum or not as the dependent variable using the univariate variables that were significantly predictive of incomplete capsule passage. In addition, differences in bowel transit time between the groups were compared by a survival model (Kaplan-Meier) and (Log-Rank test).

Our study represents a first look at the CECR in IPs in the United States, and demonstrates that both GTT and SBTT are significantly prolonged in IP CE exams compared to OP exams. In a multivariate regression model, we show that patient location (IP vs OP) and GTT are independent predictors of incomplete CE, suggesting factors other than GTT must be contributing to the lower CECR in inpatients. Here we demonstrate that increased SBTT is another important contributor for low CECR in addition to GTT. We further show that the CECR is lowest in the ICU setting.

CECR in patients with OGB has been investigated previously. In these studies, CECR are variable and lower than what we note in our study. In a prospective European study of only mild to moderate obscure overt GI bleeders, incomplete IP CEs were 21.6%, but in 75% of the incomplete exams there was some localization of the bleed site[9]. However, this study excluded all severe bleeders and inpatients with other indications. In another study of 48 inpatient obscure overt GI bleeders from Europe, incomplete CE rates were 27%[10]. In a recent Canadian study of 535 patients, overt GI bleed was found to be a major risk factor for incomplete CE (P = 0.002, OR = 2.7) and the overall incomplete CE rate was 29.5%[11]. However, in this study, only 31 subjects were inpatients. Therefore, hospitalization status did not reach statistical significance despite a large trend (P = 0.054). In studies that analyzed CECR in patients with GIB, CECR in relation to patient location was not examined and direct comparisons to outpatients were not complete. Using subgroup analysis of our patients with GIB, we found a lower IP CECR compared to these prior studies (95% in OP vs 80% in IP). More interestingly, regardless of the GIB diagnosis, incomplete CE rate was also significantly lower in the OP setting compared to the IP setting (P = 0.001).

The relationship between CECR and hospitalization was addressed more directly in two other previous European studies. When all indications were taken into account, incomplete CE rate was higher in hospitalized patients, 34%-35%[7,8]. Our results in the United States confirm these findings in a larger dataset. GTT was elevated in the IP setting in both studies, similar to our findings. In the former studies, there were additional interventions to improve CE quality: per protocol, some patients received bowel preparation[8], while others received both bowel preparation and prokinetic agents[7]. Bowel preparation and the use of prokinetic agents are not routine in the United States. None of these prior studies specifically investigated the effect of IP location in CECR, and in one of these studies, patients underwent a 23 h hospitalization solely to perform the CE, which is not done in the United States. In addition lower IP CECR shown in these studies were attributed to decreased gastric motility; whereas we now show that the minimum SBTTs are also significantly prolonged in patients who had IP CE, suggesting that factors such as generalized dysmotility of gastrointestinal tract may be at play.

The overall CECR has been reported to be 81.1% and 83.5% in two recent meta-analyses[12,13]. The first meta-analysis compared CECR in patients who were prepped with purgatives vs clear liquid diet prior to the CE and showed that there was no significant difference in CECR, although the purgative use resulted in better small bowel visualization quality and increased diagnostic yield[13]. The second study reported an overall CECR of 83.5% in patients with OGIB, CD, and neoplastic lesions[12]. The overall CECR in our study at 86% reflects a slightly higher value than reported in these two meta-analyses. Nevertheless, even if we were to assume that the CECR in our OP group was as low as 81%-83% (similar to the two meta-analyses), this would still be markedly numerically higher than our observed CECR of 68% in the IPs, and should not alter our conclusions. Our CECR may perhaps be reflective of the careful case selection at a tertiary urban medical center, in order to avoid retention. Alternative explanations could include a larger number of OPs compared to IPs in our study. Notably, the percentage of OPs and IPs were not delineated in the meta-analyses. While the latter studies included in the meta-analyses all come from the English literature, they could be conducted in various countries, where CE may not be routinely done as an outpatient. Additionally even within the United States, the capture area of each medical center may be different and may require some subjects to be IPs especially if the patient catchment area is wide and mostly rural. Our findings of lower CECR rates in the ICU setting might suggest that the severity of the acute illness may also play important roles in CECR. However, possible confounding factors such as medication use and certain comorbidities such as diabetes, neuropathy, and hypotension were not predictive of incomplete CE in hospitalized patients. This suggests that multiple factors may have a combined effect in the outcomes observed in hospitalized patients. While our hospitalized patients were older than our outpatients and while our ICU patients were older than our GMF patients, age was not an independent factor that explained CECR differences, which have previously been demonstrated[10,14].

Our study has several strengths including a larger IP sample size and a sub-group analysis on patient location and GIB. In our study, no confounding prokinetic agents or bowel preparations were used, and the hospitalization status was exclusively based on the severity of patients’ clinical condition reflective of the current practice patterns in the United States. The major limitation of our study is its retrospective design; therefore, further prospective studies in inpatients are necessary to confirm our observations. Additionally, we took consecutive patients who had CE, and did not do a matched case-control design. This resulted in more subjects in our OP group compared to IP group - due to increasing use of CE in the OP setting. Therefore, it is possible that having less number of IPs may have resulted in potential unidentified confounders. As such, larger studies on IPs are needed to confirm our findings. Another potential limitation of our study is that, the readers of CE examinations were not blinded to the patient’s clinical data. However, the readers were blinded to the hypotheses, making an outcome bias less likely to impact the study results. The lack of data regarding medication use, diabetes, and neuropathy in the outpatient group, lack of a generalized quality measure of the views obtained during CE exams, and the single center design are other limitations of our study.

Several interventions have been undertaken to improve CECR. The results of the studies which investigated the effect of bowel preparation on CECR were not consistent with each other[15-22], but improved image quality has been reported. Previous studies also investigated the role of endoscopic placement of capsule and/or real-time viewing and showed that these approaches can improve CECR[23-26]. Use of 12 h CEs could also be used to improve CECR[27]. Future studies are needed to examine the effects of such interventions to improve CECR among inpatients.

In conclusion, inpatients have a lower than usual CECR, which is more pronounced if admitted to the ICU. While gastric retention is a reason in some patients for the lower CECR, there seems to be generalized dysmotility in inpatients with incomplete CE exams. Other factors embodied in the IP setting such as sedentary status and acuity of the illness could be partially contributing. A large prospective study is needed to further delineate the most important factors that contribute to incomplete exams in the inpatient setting and determine the impact of incomplete CE on patient outcomes.