A predictive nomogram improves diagnostic accuracy and inter-observer agreement
PUBLISHED: 2015-11-27  284 total views, 1 today

Junjie Peng1, Tong Tong2, Linghui Xu2, Ying Ding3, Renjie Wang1, Xiaoji Ma1, Sanjun Cai1

1Department of Colorectal Surgery, Fudan University Shanghai Cancer Center, 2Department of Radiology, Fudan University Shanghai Cancer Center, 3Department of Biostatistics, University of Pittsburgh


Objective:The purpose of this study is to develop a predictive nomogram to improve the diagnostic accuracy and inter-observer agreement of pre-therapeutic lymph nodes metastases in patients with rectal cancer, by incorporating pre-therapeutic clinicopathological and MRI imaging variables. Method: All patients, who underwent preoperative high resolution MRI scanning and then followed by immediate surgeries for resection of the primary tumors, were selected in our series. A total of 411 patients with rectal cancer, locating within 12cm from anal verge, were retrospectively collected at Fudan University Shanghai Cancer Center between January 2005 and December 2014. To develop a nomogram, 288 patients were assigned to the training group; and the other 123 patients were assigned to the validation group. The variables assessed by high-resolution MRI included primary tumor and local or regional lymph nodes, including cT category, CRM involvement, tumor location within the bowel circle, and tumor location related to peritoneal reflex. The assessment of lymph nodes included the number of detected nodes, size of the nodes, irregularity of nodes' border and uniformity of signal intensity within the nodes. All MRI images were read by two independent observers. Each observer was required to make a diagnosis of N category (N+/N-). The inter-observer agreement was measured between two observers according to their original diagnosis of N category. Logistic regression models were performed to develop a predictive model for the status of lymph node metastasis. Result: The diagnostic accuracy of MRI-assessed cT classification was 81.3%; 12.2% of the patients were over-staged and 6.5% of the patients were under-staged. The diagnostic accuracy of MRI-assessed cN classification (on the overall data) was 68%; 14.2% of the patients were over-staged and 17.8% of the patients were under-staged. We studied the inter-observer agreement of MRI-assessed cT/cN classification. For cT classification, the two radiologists disagreed in 3.8% of the patients (11 cases), with a kappa value of 0.905. However, 35.1% of the patients (101 cases) had disagreed diagnosis for the cN classification by two radiologists, with a kappa value of 0.295. All the disagreed cases were in the group of detected lymph nodes with size % of the cases were disagreed in patients with lymph nodes sizing ≤5mm and 39.4% of the cases were disagreed in patients with detected lymph nodes sizing>5mm- (Figure 1).A nomogram for lymph node metastasis was successfully developed, with an AUC of 0.79 in the independent validation data. The predictors included in the nomogram are cT classification, CRM involvement, preoperative CEA, tumor grade and MRI-assessed lymph node classification (Figure 2). Conclusion: By incorporating important clinicpathological variables and MRI imaging features, our nomogram improved the diagnostic accuracy and minimized the inter-observer agreement in diagnosing lymph nodes metastases in rectal cancers.


Key Words: Rectal cancer  Lymph node metastases  Nomogram

Copyright © 1998 - 2018 Chinese Society of Clinical Oncology(CSCO). All Rights Reserved

京公网安备 11010502031031号

Contact Us



Phone:86(10)67726451 (Beijing)

86(25)84547290 (Nanjing)