Introduction

Optical colonoscopy is widely used colorectal cancer (CRC) or polyp screening tool and is recommended as the CRC test before treatment. A complete examination of the colon and rectum is favorable for any screening program, however, incomplete colonoscopy occurs in approximately 10% of cases [1,2,3,4]. Previous studies reported that various patient-related factors including age, sex, high or low body mass index, body habitus, bowel preparation, a history of abdominal surgery, colonic loops or angulation, and diverticular disease were associated with incomplete colonoscopy [2, 5,6,7], and it can be also caused by luminal narrowing of CRC. As a result of incomplete colonoscopy for the patients with CRC, they may miss a chance to diagnose synchronous tumors which prevalence of 1–7% proximal to the stenosing cancer, which may result in secondary surgery [8,9,10,11,12,13,14,15,16].

Some authors examined the factors related to incomplete colonoscopy using computed tomography colonoscopy (CTC), and concluded that larger diameter of sigmoid colon, colonic elongation, tortuosity, and advanced diverticular disease were associated with incomplete colonoscopy [17,18,19]. However, it is unclear whether the CTC features of local tumors influence on the incomplete colonoscopy for the patients with CRC.

The purpose of this study was to examine the relation between CTC features of CRC and incomplete colonoscopy.

Materials and methods

Study participants

This retrospective study was approved by our institutional review board, and informed consent was waived. From April 2012 to March 2015, 117 consecutive patients with advanced CRC (histologically proven adenocarcinoma) evaluated by optical colonoscopy and CTC before treatment at our department were enrolled in this study. For the purpose of this study, cases with cecal lesion (n = 6) and with poor CTC quality (intestinal collapse, n = 3) were excluded. There was no case with incomplete colonoscopy due to other reasons such as anatomic factors, adhesion after preoperative status or diverticular disease. There was one case of rectal cancer pathologically proven to invade to bladder after surgery (stage of T4b). All resected CRC were staged according to 8th edition of the UICC TNM Classification. Finally, a total of 108 patients were analyzed in this study. The patients’ clinicopathological characteristics are presented in Table 1.

Table 1 Clinicopathological characteristics
Full size table

CT colonography procedures

For each patient, CTC was performed on the same day and within one hour after a colonoscopy using a bowel preparation method utilizing polyethylene glycol. Before the CTC, a single balloon tube was inserted into the rectum by the transanal route, and colonic insufflation with carbon dioxide using a CO2 injector (PROTOCO2L; Bracco, Princeton, NJ, USA) was performed for each patient. The CTC was performed using a 320-slice MDCT (Aquilion One; Canon Medical Systems) with the following parameters: 120 kV, 100–300 mA, beam collimation 1 mm, slice thickness 1 mm, reconstruction interval 1 mm, pitch of 65. All patients underwent both supine and prone scans. The CT scans were performed with a bolus triggered technique; 2 mL/kg of nonionic contrast material (Iopamiron 370; Bayer Health Care, Osaka, Japan) was injected by an automated power injector with three scanning (arterial, portal and delayed phase).

CT colonography analysis

The CTC data sets (mainly of the patients with supine position but prone for the rectal lesion) were loaded onto a workstation (Synapse Vincent, Fujifilm Medical, Tokyo, Japan). All CTC examinations were reviewed by two board-certified radiologists who had 14- and 7-years’ experience of interpreting CTC in consensus. The CRC lesions were identified using the colonoscopy records of the same day as a reference. The readers measured a cross-sectional area of stenosis and tumor on multiplanar reconstruction (MPR). A section of the narrowest cancer canal was chosen as a representative. The cross-sectional area (mm2) of stenosis and tumor on the same section were measured by manual tracing. The readers measured a tumor length (cm), circumferential tumor extent rate (CER, %), and distance from anus (cm) on virtual colon dissection (VCD). A viewing mode of VCD is the three-dimensional (3D) model of the colon virtually unrolled, sliced open, and displayed as a flat 3D rendering of the mucosal surface, similar to a gross pathologic specimen [20]. A maximum length of the tumor on VCD was defined as the tumor length. For the measurement of the CER, the reviewers determined representative section line (normally center) within the tumor, then measured maximal diameters of the tumor and normal mucosa. They calculated the CER according to the following formula: maximal diameter of tumor/maximal diameter of normal mucosa × 100. The absolute value of the maximal tumor diameter was not available for the analyses because of bowel distortion on VCD. The distance from anus was automatically measured when a cursor was placed on the anal side of the tumor on VCD (Fig. 1).

Fig. 1
figure 1

Shema of measurement on VCD

Full size image

Statistical analysis

We divided the study cohort into two groups with complete (n = 74) and incomplete (n = 34) colonoscopy according to descriptions on the endoscopic records and then compared CTC features between the two groups. The CTC features included cross-sectional area of stenosis and tumor on MPR, and tumor length, CER, number of 100% CER cases, and distance from anus on VCD. Continuous and categoric variables were examined by student t test or Wilcoxon rank sum test (for nonparametric variables) and the χ2 test or Fisher's exact test. We also performed a receiver operating characteristic (ROC) analysis according to the results above and calculated the sensitivity, specificity and area under the curve (AUC) to assess a diagnostic performance of those CTC parameters above to predict incomplete colonoscopy. Differences with p values < 0.05 were accepted as significant. Statistical analyses were performed using SPSS 21.0 for Windows software.

Results

Cross-sectional area of tumor and stenosis were significantly different between the two groups (p = 0.001 and < 0.001) according to the Wilcoxon rank sum test. The cross-sectional area of tumor and stenosis of complete colonoscopy group were significantly smaller and larger than those of incomplete colonoscopy group. CER and number of 100% CER cases were also significantly different between the two groups (p < 0.001). Median CER of complete and incomplete colonoscopy group was 45.0% and 100.0%. Accordingly, number of 100% CER cases was lower in complete colonoscopy group (20/74) than in incomplete colonoscopy group (32/34). Tumor length and distance from anus were not significantly different between the two groups (Table 2).

Table 2 CTC features of CRC with complete and incomplete TCS
Full size table

In the ROC analysis, when the optimal cut-off; was set at 457 mm2 for cross-sectional area of tumor, 168 mm2 for cross-sectional area of stenosis and 100% for CER, the sensitivities were 93.9, 96.9 and 94.1%, and the specificities were 45.8%, 79.2% and 73.0, respectively. The cross-sectional area of stenosis showed AUC of 0.916, which was the best to predict incomplete colonoscopy among the three variables (Table 3, Fig. 2). Example of the images of incomplete colonoscopy are shown in Fig. 3.

Table 3 Results of ROC analysis
Full size table
Fig. 2
figure 2

ROC curves for diagnosing incomplete colonoscopy (a cross-sectional area of tumor, b cross-sectional area of stenosis, c CER)

Full size image
Fig. 3
figure 3

A case of incomplete colonoscopy in a 75-year-old man with a sigmoid colon cancer. a Air enema image shows severe stenosis forming “apple-core sign” in the sigmoid colon (arrow). b MPR image shows complete circumferential thickening of colonic wall. A cross-sectional area of tumor and stenosis was 1443 mm2 (yellow line) and 52 mm2 (red line). c VCD image shows the tumor with 100% CER (arrows)

Full size image

Discussion

Our study results showed that larger cross-sectional area of tumor and smaller cross-sectional area of stenosis on MPR were significantly related to incomplete colonoscopy. According to that result, it is speculated that CRC would grow increasing its volume and advancing bowel stenosis. It could finally cause incomplete colonoscopy for the patient with CRC. Park et al. performed CT volumetry for CRC and concluded that tumor volume of CRC showed an incremental trend with T-stage [21]. Yang et al. analyzed the clinicopathologic characteristics of obstructive CRC; the number of cases increased with the increasing level of T category [22].

We examined relationship between incomplete colonoscopy and VCD features in this study and found that incomplete colonoscopy significantly correlated with CER but not with tumor length. Incomplete colonoscopy group included only two cases of CER of less than 100%. According to that results, CRC may extend with circumferential spread rather than extend along with long axis direction of colonic lumen, finally result in incomplete colonoscopy.

Various image display techniques are used to interpret CTC including a two-dimensional (2D) axial review, 2D MPR review, and various 3D display options. VCD, one of the 3D displays of CTC, has advantages over other display options providing overview of the entire mucosal surface of bowel with high objectivity and reproducibility. Advanced CRC was easily detected in this study, and the parameters including tumor length, CER and distance to anus were also easily estimated. Accordingly, 100% CER may be one of the useful CTC parameters of incomplete colonoscopy that can be applicable in clinical setting.

There have been several papers that examined CTC features of cases with incomplete colonoscopy for screening colonoscopy. Hanson et al. reported that anatomic features associated with failure to reach the cecum at optical colonoscopy included colonic elongation, tortuosity, and advanced diverticular disease [18]. Eickhoff et al. examined CTC of normal colon anatomy focusing on length, number of flexures and tortuosity and assessed frequency and type of looping. Their results showed that increased colonic length, tortuosity and redundancy of the colon could be associated with an increased risk of looping and/or incomplete colonoscopy [17]. Lee et al. reported that larger colonic diameter of sigmoid colon corresponds to incomplete colonoscopy according to their research using CTC focusing on the colonic diameter [19]. According to the ROC results of our study, local CTC features including cross-sectional area of tumor and stenosis and CER were found to be parameters which predict incomplete colonoscopy for the patients with CRC. The cross-sectional area of stenosis showed the highest AUC among them.

Our study has several limitations. First, this was a retrospective analysis with a relatively small number of patients. Second, we did not analyze other parameters of colon such as tortuosity nor colon diameter. Furthermore, we did not take distortion caused by VCD into consideration. It may cause a small bias to the results. Third, the imaging analyses performed by two radiologists in consensus may not be fully objective nor reproductive. However, it would not have great influence on the result.

Conclusion

Our study showed that local CTC features including larger cross-sectional area of tumor, smaller cross-sectional area of stenosis on MPR and 100% CER on VCD were significantly associated with incomplete colonoscopy for the patients with CRC. These features may be helpful in predicting scope passage for CRC lesion found by CTC prior to colonoscopy.