2.1

Patients

This retrospective study was approved by the local institutional review board with a waiver for consent due to minimal risk. A search of our institutional electronic health record identified 64 patients with pathologically proven colorectal adenocarcinoma and a pre-treatment routine abdomen/pelvis CT performed within 3 months either before or after diagnosis. For all cases, initial pathologic diagnosis was established from colonoscopic biopsy, with biopsy dates spanning August 21, 2017 through December 21, 2021. If available, subsequent surgical resection pathology reports were reviewed to determine tumor size as well as T and N staging. Primary tumors of the cecum or ileocecal valve that involved the terminal ileum were included, while tumors arising entirely from the terminal ileum were not. Patients were excluded if CT was obtained after polypectomy with follow-up biopsy or resection showing no residual carcinoma. For the negative control cases, 140 patients were identified with a routine abdomen/pelvis CT performed within 1 year after or any time before a colonoscopy without evidence of malignancy.

All CTs were reviewed using 2.5 or 5 mm axial slices with coronal and sagittal reformats. Due to the large number of different CT scanners and protocols used throughout this cohort, as well as the inclusion of exams performed at outside institutions for which this information was not available, these technical parameters were not reported.

2.2

Image analysis

Two board-certified, fellowship-trained abdominal radiologists (OH and SB), with 5 and 18 years of experience respectively, reviewed each of the CT exams while blinded to patient history and exam indication. For each exam they reported whether an abnormality suspicious for colorectal cancer was present; if so they also noted the location of the abnormality (rectum, sigmoid, descending colon, transverse colon, ascending colon, cecum, or terminal ileum), morphologic descriptor, presence of regional lymphadenopathy (Yes/No), and degree of diagnostic confidence. Morphologic descriptors, as initially reported by Johnson et al. [ ], consisted of (1) concentric bowel wall thickening, (2) asymmetric bowel wall thickening, (3) polypoid intraluminal mass, and (4) exophytic mass ( Fig. 1 ). Diagnostic confidence was reported on a 1–5 Likert scale, with 1 indicating ‘Almost Certainly Benign’, 2 – ‘Probably Benign’, 3 – ‘Could be Benign or Neoplastic’, 4 – ‘Probably Neoplastic’, and 5 – ‘Almost Certainly Neoplastic’.

Morphologic descriptors of colorectal cancer. (a) Concentric wall thickening, which may be associated with luminal narrowing. (b) Asymmetric wall thickening without a discrete mass lesion. (c) Polypoid intraluminal mass. (d) Exophytic mass, which in this case involves both the cecum and terminal ileum.

Although on surgical resection pathology some tumors were noted to involve multiple contiguous segments of colon (i.e. cecum and ascending colon), for each imaging abnormality the reviewers reported only the single segment of bowel that was most involved. If a reviewer correctly identified a case with colorectal cancer but the reported site of imaging abnormality was discordant with pathologic findings, this was classified as a false negative. Reviewers could report multiple abnormalities if they perceived two or more discrete, discontinuous sites of possible cancer; these cases were classified as true positive if any of the reported abnormalities were concordant with pathologic findings.

2.3

Statistical analysis

Reviewer performance metrics including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Inter-rater agreement was calculated with Cohen’s kappa. Continuous variables were summarized as mean ± standard deviation and normality was assessed with the Shapiro-Wilk test. Age was compared with unpaired two-tailed t-tests and tumor size was compared with the Mann-Whitney U test. Categorical variables were evaluated with Fisher exact tests. Linear regression was used to evaluate the association between cancer detection and diagnostic confidence. Statistical analysis was performed with R v4.0.5 (The R Foundation).

3.1

Patient demographics

Patient demographic data is summarized in Table 1 . Of the patients without colorectal cancer, 48% (67/140) were female compared to 47% (30/64) for those with cancer (p > 0.99). The average age of patients without cancer was 57 ± 11 years (range 31–79) compared to 60 ± 16 years (range 25–90) for patients with cancer (p = 0.10). Of the patients with cancer, 33% (21/64) underwent an abdomen/pelvis CT before colonoscopic biopsy while 67% (43/64) underwent a CT after biopsy. The average time between CT and diagnosis was 16 ± 14 days (range 2–48) for patients with CT first compared to 12 ± 14 days (range 0–80) for patients with biopsy first. For patients with cancer, 61% (39/64) had a CT with only intravenous contrast, 2% (1/64) with only oral contrast, 34% (22/64) with intravenous and oral contrast, and 3% (2/64) without contrast. For patients without cancer, 84% (117/140) had a CT with only intravenous contrast, 9% (13/140) with intravenous and oral contrast, and 7% (10/140) without contrast.

Surgical resection pathology data was available for 83% (53/64) of patients with cancer ( Table 1 ). For the 11 patients without surgical resection data, four received post-diagnosis care at an outside institution, three declined treatment, two were ineligible for surgery due to distant metastatic disease, one was lost to follow-up, and one died several days after diagnosis from an unrelated medical condition. Average maximum tumor diameter was 4.8 ± 2.8 cm (range 0.1–16). Most tumors were locally advanced, with 74% (39/53) classified as either T3 or T4, compared to 26% (14/53) as Tis, T1, or T2. Most patients had no nodal metastases, with 60% (32/53) classified as N0 compared to 40% (21/53) as either N1 or N2.

3.2

Detection performance

Reviewer 1 demonstrated a sensitivity of 64% and specificity of 86%, compared to 77% and 83%, respectively, for reviewer 2 ( Table 2 ). Inter-rater agreement was moderate (κ = 0.478). Reviewer data was combined using two different approaches: one in which a case was classified as positive only if both reviewers reported it as positive (Combined-Both) and one in which a case was classified as positive if either reviewer reported it as positive (Combined-Either). Combined-Both demonstrated a sensitivity of 58% and specificity of 96%, compared to 83% and 74% respectively for Combined-Either. For each reviewer and set of combined data, tumor T stage was significantly associated with detection rate. Stage Tis disease was detected at a rate of 0% (0/1) for both reviewers, compared to 56% (5/9) and 67% (6/9) for T2 disease for reviewer 1 and reviewer 2 respectively, and 71% (5/7) and 100% (7/7) for T4 disease. Detected cancer was significantly larger than missed cancer ( Fig. 2 ), with average detected tumor size of 5.7 ± 2.8 cm (n = 37) and missed tumor size of 2.8 ± 1.5 cm (n = 16) for reviewer 1 (p < 0.001) compared to 5.2 ± 2.8 cm (n = 44) and 2.9 ± 2.2 cm (n = 9) respectively for reviewer 2 (p = 0.01).

Table 2

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related posts:

Assessing the reliability of ChatGPT4 in the appropriateness of radiology referrals Treatment of follicular lymphoma with a focus on radiotherapy Dual energy CT for the diagnosis of gout: Evaluating the optimal Hounsfield unit setting for dual energy processing Evaluating intra and inter-observer bias in the cosmetic rating for random vs. serial assessment of breast photographs Age is a risk factor for long-term effects of chemotherapy on dental development in childhood cancer survivors Percutaneous biopsy of bowel lesions: Technique, efficacy and safety

Stay updated, free articles. Join our Telegram channel

Join