Accurate staging of non-metastatic colon cancer with CT: the importance of training and practice for experienced radiologists and analysis of incorrectly staged cases.

PURPOSE: To investigate whether locoregional staging of colon cancer by experienced radiologists can be improved by training and feedback to minimize the risk of over-staging into the context of patient selection for neoadjuvant therapy and to identify potential pitfalls of CT staging by characterizing pathologic traits of tumors that remain challenging for radiologists. METHODS: Forty-five cases of stage I-III colon cancer were included in this retrospective study. Five experienced radiologists evaluated the CTs; 5 baseline scans followed by 4 sequential batches of 10 scans. All radiologists were trained after baseline scoring and 2 radiologists received feedback. The learning curve, diagnostic performance, reader confidence, and reading time were evaluated with pathologic staging as reference. Pathology reports and H&E slides of challenging cases were reviewed to identify potential pitfalls. RESULTS: Diagnostic performance in distinguishing T1-2 vs. T3-4 improved significantly after training and with increasing number of reviewed cases. Inaccurate staging was more frequently related to under-staging rather than over-staging. Risk of over-staging was minimized to 7% in batch 3-4. N-staging remained unreliable with an overall accuracy of 61%. Pathologic review identified two tumor characteristics causing under-staging for T-stage in 5/7 cases: (1) very limited invasive part beyond the muscularis propria and (2) mucinous composition of the invading part. CONCLUSION: The high accuracy and specificity of T-staging reached in our study indicate that sufficient training and practice of experienced radiologists can ensure high validity for CT staging in colon cancer to safely use neoadjuvant therapy without significant risk of over-treatment, while N-staging remained unreliable.

Defining disease status in gastroenteropancreatic neuroendocrine tumors: Choi-criteria or RECIST?

PURPOSE: Adequate monitoring of changes in tumor load is fundamental for the assessment of the course of disease and response to treatment. There is an ongoing debate on the utility of RECIST v1.1 in gastroenteropancreatic neuroendocrine tumors (GEP-NETs). METHODS: In this retrospective real-life cohort study, Choi-criteria were compared with RECIST v1.1. The agreement between both criteria and the association with survival endpoints were evaluated. RESULTS: Seventy-five patients were included with a median follow-up of 35 months (range 8-53). Median progression-free survival (mPFS) according to RECIST v1.1 was 15 months (range 2-50) compared to 14 months (range 2-50) in Choi. According to RECIST, 33 (44%) patients were classified as having stable disease (SD), 40 (53%) as progressive disease (PD) and two (3%) patients as partial response (PR), compared to 9 (12%) patients classified as SD, 50 (67%) as PD and 16 (21%) as PR according to Choi-criteria. Overall concordance between the criteria was moderate (Cohen's Kappa = 0.408, p < 0.001) and agreement varied between 57 and 69% at each consecutive scan (p < 0.001). Survival analysis showed significant differences in overall survival (OS) for RECIST v1.1 categories PD and non-PD (log-rank p = 0.02), however, in Choi no significant differences in OS were found (p = 0.27). CONCLUSION: RECIST v1.1 had a better clinical utility and prognostic value compared to Choi-criteria. Still, RECIST were also not sufficient to adequately predict OS. This outlines the need for new tools that provides accurate information on the disease course and treatment response to support precise prognostication in patients with GEP-NETs.

Differentiating normal myelination from hypoxic-ischemic encephalopathy on T1-weighted MR Images: a new approach.

BACKGROUND AND PURPOSE: Hypoxic-ischemic cerebral changes can be difficult to distinguish from normal myelination on T1-weighted images. We hypothesized that comparing signal intensity (SI) of brain structures on T1-weighted images enables differentiation of myelination from hypoxic-ischemic brain damage. MATERIALS AND METHODS: T1-weighted images, obtained in 57 infants aged 1-104 days and born after a gestational age of 35 weeks or older, were retrospectively evaluated. Subjects were assigned to a patient (n = 23, with perinatal hypoxic-ischemic encephalopathy [HIE] stage 2/3) or a control group (n = 34). In each subject, an SI score was assigned to 19 brain structures on the basis of pairwise comparisons with the other 18 structures. In both groups, mean total SI scores were calculated for the 19 structures. Independent samples t tests assessed whether the mean total score of a structure differed significantly between the 2 groups. Logistic regression assessed which comparison was best to distinguish between the groups and to predict the presence of hypoxic-ischemic injury. RESULTS: In patients, mean total SI scores for posterolateral putamen (PP) and peri-Rolandic cortex (PC) were significantly higher (P = .000 for both). Mean total SI scores of the posterior limb of internal capsule (PLIC) and the corona radiata (CR) were significantly lower in patients (P = .000 and 0.005, respectively). Two comparisons (PLIC versus CR, PP versus PC) were best to distinguish patients and controls and to predict absence or presence of HIE (P < .0001). CONCLUSION: SI changes due to hypoxia-ischemia can be differentiated from normal myelination by comparing SI of 4 brain structures on T1-weighted images.