Study population and variables
We performed a retrospective, single-center study at Dongguk University Ilsan Hospital, Republic of Korea. Patients who underwent SBCE (PillCam SB3, GIVEN Imaging Ltd., Yoqneam, Israel) between January 1, 2018, and December 31, 2020, were included in the study. Patients who underwent SBCE other than Pillcam SB3, incomplete (unable to identify the cecum) SBCE studies, and those with inaccurate or missed data were excluded.
Eighty-five patients who met the inclusion criteria were finally enrolled in the study. The following variables were analyzed based on electronic medical records and endoscopic images: age, gender, American Society of Anesthesiologists (ASA) classification, whether colonoscopy was performed before SBCE, reasons for SBCE, time from taking the last dose of purgative to colonoscopy and SBCE, whether prokinetics were used, gastric and small bowel transit time. This study was approved by the Institutional Review Board of Dongguk University Ilsan Hospital (DUIH 2021-08-036-001). All methods in this study were carried out in accordance with relevant guidelines and regulations. And, informed consent was obtained from all subjects or, if subjects are under 18, from a parent and/or legal guardian.
 Endoscopy procedure and purgative administration time (Figure 4)
(1) Small bowel capsule endoscopy alone
Patients undergoing only SBCE were instructed to fast for 12 hours overnight, start taking the purgative 3 hours before the SBCE, and finish taking it 1 hour before the SBCE. All patients were instructed to take a purgative consisting of half dose (total 1 liter with clear liquid) of polyethylen glycol (PEG) 3350 + ascorbic acid (CoolprepTM, Taejoon Pharm Co., Ltd, Seoul, Korea).
(2) Small bowel capsule endoscopy after colonoscopy
In this case, SBCE was performed about 1 hour after the end of the colonoscopy. For the colonoscopy, the colon was prepared by full dose (total 2 liter with clear liquid) of CoolprepTM using either the split-dose or the same-day regimen. In the case of colonoscopies before noon, the patients were instructed to take the first dose at 8 pm the day before and the last dose at 6 am on the day of colonoscopy as in the split-dose regimen. In the case of colonoscopies in the afternoon, the patients were instructed to take the first dose at 6 am and the last dose at 9 am on the day of colonoscopy as in the same-day regimen. The patients were instructed to complete taking the purgative 2 hours before the colonoscopy. Midazolam and pethidine were administered for sedation colonoscopy. If the colonoscopy was completed and the indications for SBCE were applicable, antidotes such as flumazenil and naloxone were administered and the patient was observed for one hour. And then, capsule endoscope was swallowed under the supervision of a medical staff. A prokinetic agent was administered if necessary when slow gastric transit time was expected in relation to sedative administration.
 Artificial intelligence (AI) algorithm for calculating bowel cleanliness score
In this study, a convolutional neural network (CNN) algorithm for calculating bowel cleanliness scores based on InceptionResnetV2 was used.11 Each SBCE image was calculated according to a 5-step scoring method. A cleanliness score of 1 was calculated when the mucosal visualization was less than 25% and a score of 5 when the mucosal visualization was 90% or more. In the previous study, when the cut-off cleanliness score value was 3.25, the area under the curve (AUC) of 0.977 for adequate and inadequate bowel preparation was confirmed. Therefore, the same cut-off score (3.25) was applied in this study as well.
Two expert endoscopists (Oh, D.J., Nam, J.H.) identified the images of duodenal and cecal transition in each SBCE case. Afterwards, only small bowel images were extracted and a small bowel cleanliness score was automatically calculated by the AI algorithm that measures the score of each image (Figure 5).
 Outcomes and statistical analyses
The primary outcome was the difference in small bowel cleanliness scores between the SBCE immediate after colonoscopy and SBCE alone. The secondary outcome was to identify the factors related to poor small bowel cleanliness scores. Student’s t-test and Chi-square analysis were performed on the variables. Analysis of variance (ANOVA) were performed to analyze the differences among the three or more groups. Factors for poor preparation were analyzed by logistic regression with odds ratios (ORs) and 95% confidence intervals (CIs). Statistical significance was set at a p value of < 0.05 in both univariate and multivariate analyses. Statistical analysis was carried out by IBM SPSS Statistics v25.