Diagnostic yield of gastric biopsy specimens when screening for preneoplastic lesions☆☆☆★

Materials and methods 

Asymptomatic volunteers were recruited through radio announcements of a prevention study of gastric cancer in the highlands of Chiapas, Mexico. People included in the study were asymptomatic volunteers >40 years old. After receiving informed written consent from respondents, we selected people at high risk for atrophic gastritis and/or intestinal metaplasia by testing serum for antibodies to the H. pylori cytotoxin-associated gene A protein (cag-A).15 A cohort of participants were randomly assigned to 1 of 2 treatment groups: a 3-drug H. pylori eradication regimen or placebo. In each of these groups, 3 endoscopies were performed, the first before starting medications, the second 4 weeks after completing the intervention, and the last 1 year after treatment. Endoscopies were performed by 2 trained physicians; at each endoscopy, 7 biopsy specimens for histologic examination were obtained in a specific order: (1) antrum (A), (2) corpus (C), (3) A, (4) A, (5) A, (6) C, and (7) C.

Biopsy specimens were formalin-fixed, paraffin-embedded, and stained with hematoxylin and eosin (H&E). Separately, 2 pathologists studied each gastric specimen, recording the mucosal type as corpus if the biopsy specimen included glands with parietal, chief, and mucous cells and as antrum if it included mucous glands only. Each pathologist also graded the amount of H. pylori, acute and chronic gastritis, atrophy, and intestinal metaplasia using the visual analogue scale of the Sydney classification.12 Quantification of H. pylori using special stains was necessary only in cases where bacteria were scant or could not be identified by H&E staining. A final conjoint diagnosis encompassing the different histologic features for a case was reached when the pathologists reviewed discrepant diagnoses and reached a consensus. Interobserver variability of the different parameters was assessed in the first 150 cases as described previously.16

Pathology data of 733 endoscopies were used for this study. Individual pathology reviews were used to define the endoscopies for which the mapping sequence was correctly followed. A database of the histologic parameters for each biopsy of each endoscopy was then created. From this database, the number of cases with H. pylori, atrophy, metaplasia, and dysplasia were determined after study of the first 3 biopsy specimens (A, C, A) and 5 biopsy specimens using the first 3 specimens plus biopsy 4 (A) and biopsy 6 (C). The 3- and 5-specimen diagnoses (individual reviews) were compared with the conjoint final diagnoses (hereafter referred to as 7-biopsy specimen diagnoses). Using 7-biopsy specimen diagnoses as reference and the results of 3 and 5 biopsy specimens as tests, sensitivity, specificity, predictive values, and the accuracy (sum of true positive and true negative cases, over the total number of cases that followed the mapping sequence) were obtained for H. pylori, atrophy, metaplasia, and dysplasia. The term “preneoplastic lesions” included diagnoses of atrophy, intestinal metaplasia, and dysplasia. Finally, we analyzed the final diagnoses for the cases as they were reached with 3, 5, and 7 biopsy specimens. Final diagnoses were defined as the combined coexisting histopathologic features for a case at each number of biopsy specimens; for example, an endoscopy at 3 specimens may have H. pylori, atrophy, and intestinal metaplasia.

Results 

We examined 3,969 biopsy specimens from 733 endoscopies. Histologically, 2,262 (57%) of the specimens were from the corpus (compared with 43% predicted by the mapping scheme), 1,350 (34%) from the antrum (compared with 57% predicted), and 357 (9%) from an indeterminate site (too superficial). Thus endoscopists frequently misidentified corpus and antrum. The correct mapping sequence was followed in 239 (33%) endoscopies.

Table 1 summarizes the diagnoses, sensitivity, negative predictive value, and accuracy for H. pylori, atrophy, metaplasia, and dysplasia using 7, 5, and 3 biopsy specimens.

Table 1. Diagnoses, sensitivity, negative predictive value, and accuracy for 3, 5, and 7 biopsy specimens

		Seven biopsy specimens	Five biopsy specimens				Three biopsy specimens
	Pathology	Diagnoses	Diagnoses	Sensitivity	Negative PV	Accuracy	Diagnoses	Sensitivity	Negative PV	Accuracy
	H. pylori	205	205	100	100	100	203	99	94	99
	Atrophy	152	146	96	93	97	125	82	76	88
	Metaplasia	135	129	95	94	97	111	82	81	89
	Dysplasia	22	21	95	99	99	18	81	98	98

Abbreviation: PV, predictive value.

Table 2 shows the results of the final diagnoses after review of 3, 5, and 7 biopsy specimens for each endoscopy.

Table 2. Number and percentage of cases with final diagnoses after review of 7, 5, and 3 biopsy specimens

	Pathology	Seven specimens number	Five specimens number (%)	Three specimens number (%)
	No histopathology	8	8 (100)	8 (100)
	Gastritis alone	3	3 (100)	3 (100)
	Hp gastritis	38	38 (100)	38 (100)
	Atrophy	6	6 (100)	6 (100)
	Hp with atrophy	46	45 (98)	36 (78)
	IM	4	4 (100)	4 (100)
	Hp with IM	32	29 (91)	20 (62)
	Atrophy & IM	10	9 (90)	7 (70)
	Hp, atrophy & IM	70	63 (90)	50 (71)
	Dysplasia*	3	1 (33)	1 (33)
	Hp and dysplasia	19	18 (95)	14 (74)
	Total	239	224 (94)	187 (78)
	*Dysplasia was usually accompanied by either atrophy and/or intestinal metaplasia.

Abbreviations: Hp, H. pylori; IM, intestinal metaplasia.

Table 3 lists the grades according to the Sydney classification for atrophy, intestinal metaplasia (complete, incomplete, and mixed), and dysplasia for the cases after review of 7 biopsy specimens.

Table 3. Grading according to the sydney classification for atrophy, intestinal metaplasia, and dysplasia

	Pathology	Mild	Moderate	Marked	Total
	Atrophy	70	66	16	152
	Complete IM	16	28	44	88
	Incomplete IM	20	8	1	29
	Mixed IM	2	3	13	18
	Dysplasia	21	1	0	22

Abbreviation: IM, intestinal metaplasia.

Discussion 

The updated Sydney system recommendations have become the gold standard for obtaining and reporting gastric biopsy specimens. Thus many reports now present findings as having been evaluated by the Sydney system.12 Controversy exists, however, concerning the reliability of the system to have maximum diagnostic yield, because biopsy specimen number and sites were chosen arbitrarily.13 By studying how frequently endoscopists can reliably follow a set mapping technique, we addressed some of the practical issues related to biopsy site. By analyzing the sensitivity and accuracy using the site and number suggested by the Sydney classification (5), as well as 2 fewer (3) and 2 more (7) biopsy specimens, we assessed the question of the ideal number of specimens necessary to obtain a meaningful diagnosis of the stomach during endoscopic examination.

Endoscopic examination and specimen sampling should not only include the number of biopsy specimens taken from grossly visible lesions, but also should reflect the pathology present in the entire stomach. To this end, various mapping techniques have been proposed.3, 4, 5, 6, 7, 8, 9, 10, 11 Mapping studies showing specific locations of different gastric lesions have used multiple sections from resected stomachs and endoscopically obtained biopsy specimens. Our findings indicate that experienced endoscopists with careful training can still misidentify sites in the stomach when using a set mapping technique to obtain biopsy specimens. Because most studies consider only biopsy results from successful endoscopic mapping, it is difficult to extract the frequency of unsuccessful endoscopic mapping from the literature.5, 6, 8, 9, 10 Two reports briefly address this issue; 1 stated that complete mapping was accomplished in 70% of cases, and the other noted that for their study the number of biopsy specimens ranged from 4 to 12.7, 14 In our study, we attribute the site misidentification by endoscopists to naturally occurring peristaltic movements, problems in identifying different mucosal textures, normal anatomic variation of each patient, differences in endoscopic equipment, or improperly embedded tissues. In our experience, endoscopists most often obtained corpus biopsy specimens thinking they were in the antrum.

Studies of the distribution of H. pylori in the stomach have shown increased density of the bacteria in the cardia followed by the antrum, with bacteria rarely found in areas with intestinal metaplasia.4, 17, 18, 19 Thus at least 2 biopsy specimens, 1 from the antrum and 1 from the corpus, should be obtained to demonstrate H. pylori infection. In this study, we were able to detect 99% of H. pylori infections with 3 biopsy specimens (2 from the antrum and 1 from the corpus), a ratio similar to that found by other authors.13 When the recommended 5 biopsy specimens were obtained the sensitivity reached 100%. Thus, for detection of H. pylori infection, the number and site of specimens suggested by the Sydney system is adequate.12

The preneoplastic gastric lesions associated with H. pylori infection have shown topographical predilection for the antrum.5, 6, 10, 11 Atrophy tends to show a diffuse pattern, whereas intestinal metaplasia and dysplasia are multifocal. In this study, the sensitivity for preneoplastic lesions was 80% when 3 biopsy specimens were reviewed and 95% when 5 specimens were reviewed. In a previous study, we determined that higher grades of intestinal metaplasia in an individual specimen and increasing number of specimens with metaplasia are associated with moderate and severe grades of atrophy.20 This observation suggests that fewer biopsy specimens are required to find the lesion in patients with extensive preneoplastic conditions than in patients with less severe and/or extensive lesions. Some authors have recommended a minimum of 8 biopsy specimens; however, it is uncertain whether 8 is the ideal number because it is possible that lesions can still be missed, particularly in patients with small, localized lesions.13 In our study, 7 biopsy specimens were considered the gold standard; however, because more specimens were not obtained, we do not know whether we missed diagnoses.

Different populations vary in the incidence of gastric cancer and associated preneoplastic lesions despite similar frequencies of H. pylori infection.14 The study presented here was performed in a population with high prevalence of H. pylori infection and preneoplastic lesions.21 Our results show that most subjects had moderate and severe grades of metaplasia. Thus in this population fewer biopsy specimens may be required to obtain a diagnosis of preneoplastic lesions. If the same study were performed in a population with low prevalence of preneoplastic conditions, then a greater number of specimens may be needed to make an accurate diagnosis.

In conclusion, our study demonstrates that experienced endoscopists misidentify sites in the stomach when following a set mapping technique to obtain biopsy specimens. We observed an excellent sensitivity to identify H. pylori and gastric cancer precursor lesions with 5 biopsy specimens. Thus we agree with the recommendations of the Sydney system of location and number for obtaining gastric specimens in populations in which there is a high prevalence of preneoplastic lesions. However, further research needs to be performed in areas with a lower incidence of preneoplasitic gastric lesions.