This retrospective study was approved by the Institutional Review Board of our institution, and the requirement for informed consent was waived. Medical records were searched through a computerized search to identify patients with pathologically-proven GI involvement of RCC from January 1994 to December 2014. Of the 3637 patients diagnosed with RCC at our institution during this period, 26 patients with 30 GI lesions were identified. No patient had synchronous GI metastasis at the time of the RCC diagnosis, and eleven patients were excluded from the analysis. Six of 11 patients, each with a single GI lesion, were excluded because the primary RCC had directly invaded the GI tract; five patients, each with a single lesion, were excluded due to poor CT image quality. Thus, this study included 15 patients (11 men and 4 women; mean age, 61.1 years; range, 45–80 years) with 19 GI lesions. Specimens for pathologic diagnosis were obtained through a surgical resection of 15 GI lesions and an endoscopic biopsy of four GI lesions. The reasons for laparotomy were endoscopically uncontrollable GI bleeding (n = 9), intussusception (n = 4), GI obstruction (n = 1), and clinical necessity of excisional biopsy for diagnostic purposes (n = 1).

One radiologist reviewed the electronic medical records and recorded medical information for each of the 15 patients. The data recorded included patient age, sex, clinical presentation, hemoglobin concentration, TNM stage with histologic type and grade of RCC, concomitant distant metastasis, interval between RCC diagnosis and detection of GI involvement, treatment after primary tumor resection, exact treatment modality, and patient outcomes.

All included patients underwent contrast-enhanced CT scans. Several CT scanners were used during the 20-year follow-up period, including the Sensation 16, Somatom Definition, Somatom Definition flash, and Somatom Definition AS + scanners (Siemens Medical Systems, Erlangen, Germany) and the LightSpeed 16, LightSpeed Plus, and LightSpeed VCT scanners (GE Healthcare, Milwaukee, WI, USA). As a result of the long follow-up period, the CT protocols varied. For scans of the abdomen and pelvis CT (n = 12), intravenous contrast medium (120–150 mL of 300–370 mgI/mL non-ionic contrast [iopromide, Ultravist 300 or Ultravist 370; Bayer Healthcare, Berlin, Germany]) was administered at a rate of 2–3 mL/sec using an automatic power injector through an 18-gauge angiographic catheter into an antecubital vein. With the patient in the supine position, venous-phase images were obtained from the dome of the diaphragm to the pubic symphysis 70–80 seconds after a contrast medium injection.

The dynamic kidney CT (n = 2) images were obtained during unenhanced, corticomedullary, parenchymal, and early excretory phases. The time for the corticomedullary phase scan was determined using an automatic bolus triggering technique: scanning was started when the CT number of an region of interest in the upper abdominal aorta reached 100 Hounsfield unit (HU). The scan delays for the parenchymal and early excretory phases were 90 seconds and 180–240 seconds, respectively.

The dynamic pancreas CT (n = 1) images were obtained during the non-enhanced, arterial, and portal phases. The arterial phase images were obtained using a bolus tracking technique with a trigger enhancement threshold of 100 HU at the upper abdominal aorta. After the threshold was reached, a diagnostic delay time of 10 seconds was used for the arterial phase. The portal-phase images were obtained 72 seconds after contrast injection.

Beginning in 2008, coronal reformations of the venous phase were routinely performed for all CT scans at our hospital. Positive oral contrast medium consisting of a barium sulfate solution (1.5 Easy CT Solution; Taejoon Pharmaceutical, Seoul, Korea) was administered to cancer patients for opacification of the small intestine. The use of a positive oral contrast agent was prohibited if bowel obstruction, perforation, bleeding, inflammatory bowel disease, or biliary stone disease was suspected. Beginning in 2013, patients were instructed to drink water instead of barium sulfate as a neutral oral contrast.

The scan parameters included beam collimation at 16 × 0.75 mm, 32 × 0.6 mm, or 64 × 0.6 mm; beam pitch at 0.75–1.5; gantry rotation time of 0.5–0.8 second; field of view to fit; 120 kVp; automated dose reduction system (CARE Dose 4D; Siemens Medical Solutions, Erlangen, Germany) with tube current set to 170–250 mAs. The images were reconstructed in the axial and coronal planes, ranging from a 2.5-mm thickness at 2.5-mm intervals to a 5-mm thickness at 5-mm intervals.

The CT scans were retrospectively reviewed by two radiologists using DICOM image viewing software, with a diagnosis reached by consensus. The involved bowel sites, lesion multiplicity, and type of involvement were assessed on the CT scans, along with the morphological appearance and contrast enhancement patterns of the lesions. The type of GI involvement was categorized as distant GI metastasis or direct GI invasion from adjacent metastatic lesions. The morphological appearance of the lesions was subjectively categorized as: 1) an intraluminal polypoid mass (Fig. 1); 2) an exophytic mass (Fig. 2); or 3) bowel wall thickening. The contrast enhancement for the GI lesions was compared to that of normal adjacent segments of the bowel wall and was categorized as 1) hypo, 2) iso, or 3) hyperenhanced.

The parenchymal-phase images of the kidney CT scans and portal phase images of pancreas CT scans were evaluated. The enhancement patterns were characterized as homogeneous or heterogeneous, and the images were also evaluated for the presence of bowel complications, lymphadenopathy, ascites and metastasis to other intra-abdominal organs. The size criterion for lymphadenopathy was 10 mm in the shortest axis (22).

Descriptive statistics were used to generate means and standard deviations for continuous and categorical data. The Kaplan-Meier method was used to estimate the overall patient survival following a diagnosis of the GI involvement. All statistical analyses were performed using IBM SPSS Statistics for Windows, Version 21.0 (IBM Corp, Armonk, NY, USA).

The clinical characteristics of the 15 included patients are summarized in Table 1. The most common clinical presentation was GI bleeding, 10 patients (66.7%), consisting of melena or hematochezia. Of the remaining five patients, two had abdominal pain, two patients had no symptoms, and one experienced weight loss. The mean hemoglobin concentration was 9.4 ± 2.6 g/dL (range, 3.7–13.3 g/dL). All patients had undergone curative nephrectomy, and adjuvant treatment modalities included chemotherapy alone (n = 9), radiotherapy alone (n = 1), and both chemotherapy and radiotherapy (n = 1). Four patients had not undergone adjuvant therapy. The mean time between the radical nephrectomy and detection of GI involvement was 30.4 ± 37.4 months (range, 0–113.4 months; median, 31.4 months).

Fourteen patients (93.3%) had concomitant metastases at the time of GI-involved RCC diagnosis. The mean number of concomitant metastatic sites per patient was 3.1 ± 1.5 (range, 0–6). The sites of simultaneous intra-abdominal metastases included the peritoneum, pancreas, liver, adrenal gland, contralateral kidney, spleen, and regional lymph nodes. The sites of simultaneous extra-abdominal metastases were the lung, brain, and musculoskeletal system. Simultaneous metastasis was most frequently detected in the lung (60%), followed by the peritoneum (53.3%).

The RCC stage at initial presentation was stage IV in five patients, stage III in six, stage II in two, and stage I in one. The staging information was not available in one patient who underwent nephrectomy at another hospital.

The most common histologic type of primary RCC was clear cell carcinoma (n = 14, 93.3%). The histologic type of one lesion could not be determined because the nephrectomy was performed at another hospital. The most common Fuhrman histologic grade was 4/4 (n = 8, 53.3%), followed by 2/4 (n = 4, 26.7%), and 3/4 (n = 3, 20.0%). Histologically, one lesion was an unclassified adenocarcinoma of unknown histologic grade. The pathologic findings of the primary RCC are summarized in Table 2.

The average follow-up period after detection of GI involvement was 20.1 ± 23.5 months (range, 0–76.6 months). Four patients (26.7%) were lost to follow-up, and four (26.7%) died after a mean follow-up period of 18.0 ± 27.2 months (range, 0–65.0 months). Of these four patients, three died due to disseminated metastases with consequent multi-organ failure, and one died due to duodenal ulcer bleeding unrelated to metastasis. The overall median survival from the detection of GI involvement was 65.0 months (95% confidence interval [CI], 0–149.8 months).

Nineteen lesions were detected in the CT scans of the 15 patients, including seven (36.8%) in the ileum, six (31.6%) in the duodenum, and three (15.8%) in the colon and stomach each. Two patients (13.3%) had multiple lesions (two and four lesions). In 12 patients (80%), the type of GI involvement was distant metastasis after curative nephrectomy, whereas the other three patients (20%) experienced direct GI invasion from an adjacent metastatic lesion. The morphology of the GI lesions varied in the CT scans, with 12 (63.2%) manifesting as intraluminal polypoid masses, six (36.8%) as exophytic masses, and one (5.3%) as wall thickening. In the two patients with more than one lesion, the individual lesions all had the same morphologic features. The average size of all lesions except wall thickening was 30.4 ± 37.4 mm, 36.5 ± 17.0 mm for exophytic lesions and 27.7 ± 8.7 mm for intraluminal polypoid lesions.

The lesions exhibited variable contrast enhancement, with 15 (78.9%) showing predominant hyperenhancement, 12 (63.2%) showing heterogeneous enhancement patterns (63.2%), and seven (36.8%) having homogeneous enhancement patterns.

The complications included bowel obstruction (n = 1) (Fig. 3), intussusception (n = 1) (Fig. 4), bile duct dilatation (n = 1), and combined bile duct/pancreatic duct dilatation (n = 1). Bile duct and pancreatic duct dilatation were associated with lesions at the duodenal second portion and the ampulla of Vater. Intussusception was associated with four lesions in one patient, all of which consisted of an intraluminal polypoid. None of the patients had regional lymphadenopathy. Ascites were found in three patients (20%), with two showing no evidence of malignancy. A summary of the CT findings is provided in Table 3.