AUTHOR=Wang Song , Ye Fei , Sheng Yuan , Yu Wenyong , Liu Yingling , Liu Dehua , Zhang Kaiguang 

TITLE=Development and Validation of Nomograms to Predict Operative Link for Gastritis Assessment Any-Stage and Stages III–IV in the Chinese High-Risk Gastric Cancer Population

JOURNAL=Frontiers in Medicine

VOLUME=Volume 8 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2021.724566

DOI=10.3389/fmed.2021.724566

ISSN=2296-858X

ABSTRACT=Purpose: It is very essential to diagnose gastric atrophy in the area with high prevalence of gastric cancer. Operative link for gastritis assessment (OLGA) was developed to detect the severity of gastric atrophy. The aim of this study was to develop and validate nomograms for predicting OLGA any-stage and stage III-IV in the Chinese high-risk gastric cancer population.
Methods We retrospectively analyzed 7945 participants obtained by a multicenter cross-sectional study. We randomly selected 55% individuals (4370 participants, training cohort) to analyze and generate the prediction models and validated the models on the remaining individuals (3575 participants, validation cohort). A multivariate logistic regression model was used to select variables in the training cohort. The corresponding nomograms were developed to predict OLGA any-stage and stage III-IV respectively. The area under the receiver operating characteristic curves and the GiViTI calibration belts were used to estimate the discrimination and calibration of the prediction models.
Results There were 1226 (28.05%) participants in the training sample and 970 (27.13%) in the validation sample who were diagnosed with gastric atrophy. The nomogram predicting OLGA any-stage had an area under the curve (AUC) of 0.610 for the training sample and 0.615 for the validation sample, with favorable calibrations in the overall population. Similarly, the nomogram predicting OLGA stage III–IV had an AUC of 0.702 and 0.714 for the training and validation samples, respectively, with favorable calibrations in the overall population. Conclusions The prediction model can early identify the occurence of gastric atrophy and the severity stage of gastric atrophy to some extent.