Alimentary Pharmacology and Therapeutics

Staging of intestinal- and diffuse-type gastric cancers with the OLGA and OLGIM staging systems S.-J. Cho*, I. J. Choi*, M.-C. Kook*, B.-H. Nam†, C. G. Kim*, J. Y. Lee*, K. W. Ryu* & Y.-W. Kim*

*Center for Gastric Cancer, National Cancer Center, Goyang, Korea. † Center for Clinical Trials, National Cancer Center, Goyang, Korea.

Correspondence to: Dr I. J. Choi, Center for Gastric Cancer, National Cancer Center, 323 Ilsan-ro, Ilsandong-gu, Goyang, Gyeonggi 410-769, Korea. E-mail: [email protected]

Publication data Submitted 23 July 2013 First decision 4 August 2013 Resubmitted 10 September 2013 Accepted 10 September 2013 EV Pub Online 6 October 2013

SUMMARY Background Operative link on gastritis assessment (OLGA) and Operative link on gastric intestinal metaplasia assessment (OLGIM) staging systems have been proposed for gastric cancer (GC) risk estimation. Aim To validate the OLGA and OLGIM staging systems in a region with high risk of GC. Methods This retrospective study included 474 GC patients and age- and sexmatched health screening control persons in a cancer centre hospital. We classified gastritis patterns according to the OLGA and OLGIM systems using the histological database that a pathologist prospectively evaluated using the updated Sydney system. GC risk according to the OLGA and OLGIM stages was evaluated using logistic regression analysis. Results More GC patients had OLGA stages III–IV (46.2%) than controls (26.6%, P < 0.001), particularly among patients with intestinal-type GCs (62.2%) compared with diffuse-type GCs (30.9%). OLGA stages III and IV were significantly associated with increased risk of GC [odds ratios (ORs), 2.09; P = 0.008 and 2.04; P = 0.014 respectively] in multivariate analysis. The association was more significant for intestinal-type (ORs, 4.76; P = 0.001 and 4.19; P = 0.002 respectively), but not diffuse-type GC. OLGIM stages from I to IV were significantly associated with increased risk of both intestinal-type (ORs, 3.64, 5.15, 7.89 and 13.20 respectively) and diffuse-type GC (ORs, 1.84, 2.59, 5.08 and 6.32 respectively) with a significantly increasing trend. Conclusion As high OLGA and OLGIM stages are independent risk factors for gastric cancer, the staging systems may be useful for risk assessment in high-risk regions, especially for intestinal-type gastric cancer. Aliment Pharmacol Ther 2013; 38: 1292–1302

1292

ª 2013 John Wiley & Sons Ltd doi:10.1111/apt.12515

OLGA and OLGIM staging systems and gastric cancer INTRODUCTION Chronic inflammation of the gastric mucosa can trigger a cascade of genotypic and phenotypic derangements that may eventually result in gastric cancer (GC).1 The presence of atrophic gastritis, intestinal metaplasia (IM) and dysplasia of the gastric mucosa are important risk factors for intestinal-type GC.1 Surveillance of patients with these lesions may therefore detect GC early and improve prognosis.2 However, diagnoses of atrophic gastritis, IM and dysplasia are often disregarded in clinical practice.2 The updated Sydney system recommends several biopsy specimens from the antrum, corpus and incisura angularis for assessment of gastritis, including atrophic gastritis and IM.3 Because this method rarely helps to estimate GC risk directly, Operative link on gastritis assessment (OLGA) staging system was proposed for clinical purposes to simplify the assessment of GC risk using gastric atrophy.4 The gastritis staging integrates the atrophy score using the updated Sydney system (obtained by biopsy) and atrophy topography (antral and oxyntic/corpus). Several studies have shown the association of low OLGA stage (0, I, II) with low risk of GC and high OLGA stage (III, IV) with increased risk of GC.5–7 The recently proposed Operative link on gastric intestinal metaplasia assessment (OLGIM) staging system8 basically adopts the OLGA frame, but replaces the atrophy score with IM score, which was suggested to have low inter-observer variation.8 The OLGIM and OLGA staging systems must be validated in predicting risk of GC, and several studies have approached this issue. However, these were limited to small numbers of cancer cases6, 7 in regions with low incidence of GC.5, 6, 9 Gastric cancers have different clinicopathological characteristics according to Lauren’s system,10 which histopathologically classifies them as intestinal- and diffuse-type carcinomas. Helicobacter pylori is a causal factor in the atrophic gastritis-IM-intestinal-type GC sequence.1 The prevalence of H. pylori infection seems to be greater in intestinal-type than in diffuse-type GCs.11, 12 However, there have been no studies validating the OLGA and OLGIM systems according to the Lauren’s types. In the present study, we validated the OLGA and OLGIM staging systems by comparing patients with GC and control subjects in Korea, where GC incidence is high.13 We also evaluated differences in the OLGA and OLGIM staging systems according to histopathological types of GC.

Aliment Pharmacol Ther 2013; 38: 1292-1302 ª 2013 John Wiley & Sons Ltd

METHODS Study population This cross-sectional study was conducted at the National Cancer Center Hospital (Goyang, Korea). We retrospectively identified GC patients from our database of subjects who underwent oesophagogastroduodenoscopy (EGD) between January 2006 and December 2008. We classified gastritis patterns according to the OLGA and OLGIM systems using the histological database that the pathologist prospectively evaluated with the updated Sydney system3. During the period, 1300 individuals who were diagnosed with gastric adenocarcinomas provided written informed consent for obtaining biopsy specimens to evaluate H. pylori and histological diagnoses. Among them, we excluded 610 patients for the following reasons: multiple GCs (n = 114), cardia cancer (n = 64), history of H. pylori eradication (n = 102), history of neoadjuvant chemotherapy (n = 18), uncertain H. pylori infection status (n = 17), insufficient clinical information (n = 2), unavailable biopsy samples for evaluation of glandular atrophy in the antrum (n = 170) or corpus (n = 123) due to the tumour involvement of predetermined biopsy sites. There were 690 eligible GC patients. Control subjects were enrolled from a pool of healthy adults who participated in the National Cancer Screening Program and underwent EGD for GC screening.14 Between July 2007 and April 2010, there were 678 subjects who provided informed consent. We excluded 46 patients for the following reasons: subjects who were diagnosed as having gastric adenocarcinoma (n = 3), history of previous H. pylori eradication (n = 5) or inadequate histological samples for evaluation of glandular atrophy (n = 38). The 632 remaining subjects were eligible for inclusion. A sex-matched control was selected within a 3-year age range for each GC patient, resulting in 474 pairs being selected for analysis. The study protocol was approved by the Institutional Review Board of the National Cancer Center, Korea (NCCNCS-12-675). EGD and H. pylori testing During EGD, biopsy specimens were taken to confirm diagnosis of GC and seven additional biopsy specimens were also obtained to evaluate H. pylori status and underlying histology. Four experienced endoscopists (IJ Choi, S-J Cho, CG Kim, JY Lee) performed EGD according to the following protocol. The biopsy specimens were

1293

S.-J. Cho et al. taken from the antrum lesser curvature, corpus lesser curvature (2 cm posterior aspect of the body near the midline) and corpus greater curvature. Among the biopsy sites recommended by the updated Sydney system, these were considered essential for adequate histological evaluation.15 Two biopsy specimens were obtained from each site, and one additional specimen was obtained from the corpus greater curvature for rapid urease test (Pronto Dry; Medical Instruments Corporation, Solothurn, Switzerland).16 Gastric mucosa was considered infected with H. pylori if either of the following criteria were positive: the rapid urease test or a histological evaluation using the updated Sydney system.3

Histology Biopsy specimens collected for histological examination were immediately fixed in neutral-buffered 10% formalin and embedded in paraffin. Histological sections were stained with haematoxylin–eosin and Wright–Giemsa and examined by a single pathologist (M-C Kook). Based on the updated Sydney system recommendation,3 H. pylori density, polymorphonuclear cell (polymorph) activity, chronic inflammation, glandular atrophy and IM were graded as absent, mild, moderate or marked (0–3 respectively), using a visual analogue scale. Atrophy was defined as loss of appropriate glands with or without metaplasia,3, 17 which includes shrinkage or complete disappearance of glandular units, being replaced by expanded (fibrotic) lamina propria and replacement of the native glands by metaplastic glands featuring a new commitment (intestinal and/or pseudopyloric metaplasia).18 Advanced histological change for atrophy and IM was defined as grade 2 or 3. Gastritis stage was assessed according to the OLGA staging system4, which integrates the atrophy score using the updated Sydney system (obtained by biopsy) and the atrophy topography (antral and oxyntic/corpus). We also assessed the gastritis stage based on the OLGIM staging system.8 The antrum lesser curvature was considered non-oxyntic gastric mucosa (antrum score), and corpus lesser curvature was considered oxyntic gastric mucosa (corpus score). Combining the antrum and corpus score for atrophic gastritis resulted in the OLGA gastritis stage score, and combining the IM scores resulted in the OLGIM staging score. Pathological evaluation for gastric carcinoma After pathological examination following endoscopic or surgical resection, early GC was defined as a tumour that was confined to the mucosa or submucosa regardless of 1294

lymph node involvement, and advanced GC was defined as a tumour that invaded proper muscle or beyond. If resection was not performed (i.e. in cases of metastatic GC), the stage was determined by endoscopic and computed tomographic findings. A pathologist determined tumour differentiation according to the World Health Organization criteria19 and histological type according to Lauren’s classification.10

Statistical analysis Baseline characteristics were compared using Student’s t-test for continuous variables and Pearson’s v2 test for categorical variables. Fisher’s exact test or v2 tests (both two-tailed) were used to compare proportions of histological grades. The logistic regression models (unconditional forward stepwise method) were used to estimate unadjusted and adjusted odds ratios (ORs) and 95% confidence intervals (CIs). OLGA and OLGIM stages were not included in the same model because they were closely related to each other. Demographic characteristics (age, sex, H. pylori infection, family history of GC and smoking habit) were taken into account in comparing GC patients and control subjects. In addition, the c statistic as a goodness of fit measure of a model was used to evaluate how well each staging system can discriminate cancer and control subjects at different levels of stage.20 The c values of 0.7– 0.8 were regarded as moderately good discrimination and values of 0.8 and above as excellent discrimination ability.21 A P-value