Review in depth 1059 Assessing the severity of atrophic gastritis Bruno Annibale and Edith Lahner Atrophic gastritis, mainly the consequence of long-stand- ing Helicobacter pylori infection, is linked to the develop- ment of gastric cancer. In the case of atrophic gastritis, severity may be mainly related to the lifetime risk of the single patient to develop gastric cancer, mostly in relation to the degree and extension of mucosal damage. As atrophic gastritis is the result of complex multifactorial interactions, the application of artificial neural networks is promising and may be useful for the identification of those patients with atrophic gastritis at higher risk for gastric malignancies. The experience of application of artificial neural networks in atrophic gastritis is still scarce. The available data suggest that these systems may contribute to identify patients with corporal metaplastic atrophic gastritis and to optimize bioptic sampling during gastro- scopy. Eur J Gastroenterol Hepatol 19:1059–1063  c 2007 Wolters Kluwer Health | Lippincott Williams & Wilkins. European Journal of Gastroenterology & Hepatology 2007, 19:1059–1063 Keywords: artificial neural networks, atrophic gastritis, gastric cancer, gastric malignancies, gastroscopy, Helicobacter pylori infection Department of Digestive and Liver Disease, University ‘La Sapienza’, Second Medical School, Ospedale Sant’Andrea, Rome, Italy Correspondence to Bruno Annibale, MD, Department of Digestive and Liver Disease, University ‘La Sapienza’, Ospedale Sant’Andrea, Via di Grottarossa 1035, 00189, Roma, Italy Tel: + 39 06 49972369; fax: + 39 06 4455292; e-mail: bruno.annibale@uniroma1.it Received 29 August 2007 Accepted 29 August 2007 Introduction Atrophic gastritis (AG) is mainly the consequence of long-standing Helicobacter pylori infection and most often a silent disorder with poor clinical signs or without specific symptoms at all, which is generally diagnosed histologi- cally on gastric biopsies obtained during routine gastro- scopy [1,2]. Albeit, AG is a substantially benign disorder for a large part of patients, it is epidemiologically and biologically linked to the development of gastric cancer (GC) [3], and it was reported that about 10% of patients with moderate–severe AG will develop gastric malignan- cies during a mean follow-up of 7.8 years [4]. The term severity may be interpreted differently by the single specialists; in fact, for pathologists, severity means generally the highest degree of histological damage, while for clinicians it rather refers to a more problematic management of the disease. In the case of AG, severity may be mainly related to the lifetime risk of the single patient to develop GC, mostly in relation to the degree and extension of mucosal damage. Indeed, the assess- ment of cancer risk in individual patients with AG is difficult, mainly because gastric carcinogenesis is modu- lated by poorly defined factors, including environment, bacterial strain, and host responses [5]. Thus, the assessment of the severity of AG may be an important challenge for the management of these patients, as its features [i.e. extension of atrophy and intestinal metaplasia (IM) hypochlorhydria] may be considered as potential surrogate markers for the increased risk for GC. Innovative statistical tools, such as artificial neural networks (ANNs) are highly flexible computerized mathematical models for understanding and predicting complex and chaotic dynamics in complex biological systems, and have been effectively used to solve nonlinear problems related to diagnostic or prognostic queries in many fields of medicine, including gastro- enterological issues [6,7]. In particular, as AG is the result of complex multifactorial interactions, the application of ANNs is promising and may be useful for the identifica- tion of those patients with AG at higher risk for gastric malignancies (severity). As far as AG is concerned, the experience of application of ANNs is still scarce. The aim of this study was to analyze features responsible for the severity of AG and to summarize available data about the diagnostic and prognostic role of ANNs in patients with this disorder. Definition of atrophic gastritis Gastric mucosal atrophy is defined as the loss of appropriate glands, which occurs when glands damaged by inflammation are replaced either by connective tissue (scarring) or by glandular structures inappropriate for location (metaplasia) [8]. Most often, as in the antral mucosa, the metaplastic transformation assumes the phenotype of the glands lined by intestinal-type epithe- lium (IM), but in the oxyntic mucosa, it may also take the form of mucin-secreting antral glands (pseudopyloric metaplasia) [8]. Visual analogue scales have been proposed as a standard histological approach for scoring atrophic-metaplastic changes in both the antral and oxyntic mucosa [9,10]. In general practice, however, the diagnosis of atrophy and IM is troublesome due to an unsatisfactory interobserver agreement among patholo- gists; thus, an international group of pathologists 0954-691X  c 2007 Wolters Kluwer Health | Lippincott Williams & Wilkins Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.