Gastric biopsies : The gap between evidence-based medicine and daily practice in the management of gastric Helicobacter pylori infection

Department of Pathology, University Health Network, Toronto, Ontario Correspondence: Dr Hala El-Zimaity, Toronto General Hospital, 200 Elizabeth Street, Toronto, Ontario M5G 2C4. Telephone 416-340-4551, fax 416-340-5175, e-mail hala.el-zimaity@uhn.ca Received for publication January 20, 2013. Accepted July 10, 2013 Helicobacter pylori infection remains clinically relevant, particularly in centres that serve the general population with a high mean age and a high number of immigrants (1-3). The infection plays an active role in many diseases, including peptic ulcer disease, gastric mucosaassociated lymphoid tissue (MALT) lymphoma and adenocarcinoma, dyspepsia, iron-deficiency anemia, idiopathic thromobocytopenic purpura and, in some patients, coronary artery disease (4-18). Active infection should be eradicated in patients with uninvestigated dyspepsia, active peptic ulcer disease, a high risk of gastric cancer and gastric MALT lymphoma (18,19). Importantly, to reduce the risk of gastric atrophy, active H pylori infection should be investigated and eradicated before administering proton pump inhibitors (PPIs) to patients with gastroesophageal reflux disease (20). Although several tests are available, many consider histopathological diagnosis of H pylori infection to be the ‘gold standard’ (21,22). Obtaining corpus biopsies in addition to antral biopsies has been shown to increase diagnostic accuracy (22,23); however, because PPIs decrease H pylori density, distribution and shape (24,25), their use renders the bacteria more difficult to detect. Therefore, the diagnostic accuracy of histopathology is dependent on which regions are sampled, pathological interpretation (26) and whether the patient is taking PPIs. The American Gastroenterological Association (AGA) and American College of Gastroenterology (ACG) recommend discontinuing PPIs two weeks before endoscopy, and taking biopsies from both the body and antrum (9,18). The present study evaluated both oriGinal arTicle

daily endoscopy practice and its influence on the histopathological diagnosis of H pylori infection.We aimed to establish whether gastroenterologists regularly sample the gastric antrum and body; the frequency of PPI use at endoscopy; and how these endoscopic practices influenced pathological interpretation.In patients who were H pylori positive, we also evaluated the effect of biopsy site on diagnosis.The present study addressed the influence of everyday practice on diagnostic efforts.

Selection of biopsy specimens
The pathology files at Toronto General Hospital (Toronto, Ontario) were reviewed for biopsy specimens submitted specifically for H pylori diagnosis between 2005 and 2010.The anatomical sites from where the biopsies were taken were noted to establish whether AGA guidelines had been followed for sampling the body and antrum.
Biopsies from 150 patients, in which both the antrum and corpus had been sampled at endoscopy, were randomly chosen for further clinical and pathological analysis.In this group of 150 patients, it was established whether both regions were correctly sampled (by analysis of the histology of the gastric mucosa -transitional mucosa was grouped with antral mucosa), and the density of H pylori within each region.The electronic patient record system was used to evaluate whether the patients were taking PPIs at the time of endoscopy.
To investigate the sampling pattern when the endoscopy report indicated that only one region was sampled, the histology pattern in 200 consecutive specimens that satisfied that criterion was reviewed.

Histological and immunohistochemical staining
Hematoxylin and eosin sections were retrieved from the archives and two 4 μm sections were cut from each block.These were stained with dual silver/periodic acid-Schiff (27) and anti-H pylori rabbit polyclonal antibody (760-2645, Ventana Medical Systems Inc, USA; according to manufacturer's instructions), respectively.Each set of three slides was separately randomized and coded for use within the study.

Histopathological examination of the slides
The slides were examined independently by four expert gastrointestinal pathologists and one trainee histopathologist.Observers were blinded to the clinical information of the patients, and had no knowledge of their own or their colleagues' interpretation of the other slides.
A visual analogue scale graded from 0 (absent) to 5 (high) was used to score H pylori, acute inflammation and chronic inflammation on each slide (28).A score of 1 indicated that only one or two bacteria were observed in the entire biopsy specimen, while a score of 5 indicated that the surface was covered with bacteria and bacterial aggregates.A biopsy specimen was considered to be positive when all observers agreed that bacteria with the characteristic morphological features of H pylori were present in at least one of the special stain (dual silver/periodic acid-Schiff-or immunohistochemistry-stained) slides.Slides exhibiting discrepancy among observers were resolved by joint re-examination to reach consensus interpretation.

data analysis
All scores were entered into a database and analyzed using Stata/SE version 11.2 (StataCorp, USA).The Fisher's exact test or, when appropriate, the χ 2 test (both two-tailed), were used for comparison of proportions; P<0.05 was considered to be statistically significant.

Evaluation of overall sampling pattern
From 2005 to 2010, gastric biopsies from 10,268 patients were specifically examined for H pylori infection in the pathology department of Toronto General Hospital.Specimens received from 6143 patients (60%) had biopsies taken from one region only, 4827 (47%) patients had biopsies taken from the antrum only and 1316 (13%) patients had biopsies taken from the body only.Thus, biopsies from both the antrum and corpus were taken only in 40% of the patients at endoscopy.

Evaluation of sampling accuracy
Biopsies in which the endoscopy report indicated that two regions were sampled: In the 150 biopsy sets randomly chosen from patients in whom the endoscopy report indicated that biopsies had been taken from both the antrum and body, 701 individual biopsies (265 [38%] antrum, 393 [56%] body and 43 [6%] cardia) were reviewed.Each patient had a minimum of two biopsies available for examination (mean 6, median 4, range two to 12 biopsies).
Although the endoscopy report indicated that antral and body biopsies were obtained from each of these patients, histopathological review confirmed that biopsies originated from both regions in only 85 (57%) patients.In the remaining biopsies, the gastric region sampled was body only in 26 patients (17%), antrum only in 16 patients (11%), antrum and cardia in eight patients (5%), and body and cardia in 15 patients (10%) (Table 1).Biopsies in which the endoscopy report indicated that only one region was sampled: In the 200 consecutive specimens the endocopy report indicated antral biopsies were obtained from 178 patients (89%) and oxyntic biopsies were obtained from 22 patients (11%).
When the endoscopy report indicated antral biopsies were obtained, histology review confirmed that the biopsies were from the antrum in only 119 (66.85%).In the remaining specimens, the gastric region sampled was antrum and body in 47 (26.40%), antrum and duodenum in one (0.56%), body in 10 (5.62%) and transitional mucosa in one (0.56%).
Therefore, an extrapolation suggested that that 47% of patients in the entire cohort had biopsies from one region of the stomach only (30% [one-half of specimens in which endoscopy indicated only one region was sampled] and 17% of patients in which the endoscopy report indicated both regions were sampled but incorrect sampling was determined pathologically).The largest source of error appeared to be that endoscopists did not take biopsies far enough distally to ensure that the gastric antrum was actually sampled in the biopsies.

Evaluation of histopathological sensitivity
From the histological re-assessment, 56 (37%) patients were positive for H pylori and 94 (63%) were negative.At routine signout, six of 56 (10.7%) patients had incorrectly been interpreted as H pylori negative.The biopsies from patients with a false-negative diagnosis either did not have any active inflammation or only had mild active inflammation.The average number of biopsies taken in patients with a false negative was three (range one to seven).In comparison, the mean number of biopsies taken in patients with a true-positive diagnosis was five (range two to 11).
Of 94 patients negative for H pylori on reassessment, two (2%) had been interpreted as positive for the infection at routine signout.Biopsies from these patients had mild active inflammation only.Active inflammation was focal and, in one case, was limited to an area with intestinal metaplasia.The sensitivity, specificity, negative predictive value and positive predictive value of histopathological analysis for H pylori in the present study were 89.29%, 97.87%, 93.88% and 96.15%, respectively.

The influence of sampling pattern (regions examined) on diagnostic accuracy
The antrum and body had been correctly sampled in 34 of 56 (61%) patients positive for H pylori infection.Bacteria were identified in the antrum only in five (15%) patients, and in the body only in seven (21%) patients, as shown in Figure 1.This confirms that obtaining biopsies from both the antrum and the body increases the probability of correctly diagnosing active infection.
In patients who received a true-positive diagnosis, the antrum and body were correctly sampled in 33 of 50 (66%) cases.In contrast, the antrum and body were correctly sampled in only one of six (17%) patients who received a false-negative diagnosis (P=0.0194), as shown in Table 2 and Figure 2.
Overall, sampling bias was apparent in many of the biopsy sets and was a major contributor to a false-negative diagnosis.H pylori density was generally low in biopsies with a false-negative reading (mean score of 1 [scale 0 to 5]).In these biopsy sets, only a few bacteria were present per biopsy, or the bacteria were distributed only focally within each fragment or were present in only a few of the available biopsy fragments (sampling bias for the infection).

PPi use and diagnostic accuracy
Seventy-one patients (47%) were taking PPIs at the time of endoscopy, as shown in Table 3.Of the patients with a false-negative diagnosis, four of six (67%) were taking PPIs at the time of biopsy.In contrast, of the 50 patients correctly diagnosed positive for the infection (ie, true positives), 22 (44%) were taking PPIs at the time of endoscopy, as shown in Figure 3  Furthermore, many of these patients exhibited chronic active inflammation.This suggests that PPI use is an important contributor to the inability to detect H pylori in patients with an otherwise typical H pylori pattern of chronic active inflammation.

diSCuSSion
Despite national and international guidelines (18,21) for the clinical management of H pylori infection, H pylori bacteria can be -and frequently are -missed on endoscopy.Factors that may cause a falsenegative result include sampling the wrong area, taking an inadequate number of biopsies and/or continued use of PPIs at the time of endoscopy.All of the above factors are encountered in daily practice, including in the present study.As a result, each year, the incorrect diagnosis is made and many patients are inappropriately managed.
An important question is the probability of an incorrect diagnosis in patients undergoing endoscopy with subsequent histological  assessment of the biopsy.In the present study, 5% (one in 20) were misclassified during routine practice.In the false-negative group, the microorganisms were either rarely present, with only one or two bacteria per slide, or the distribution of microorganisms was in only a few biopsies per set, which may have been restricted to one region (antrum or body).In the absence of significant active inflammation, chronic inflammation was interpreted to be negative for H pylori, irrespective of severity.These results are consistent with previous studies that show H pylori may be present in biopsy specimens that are histologically close to normal, especially in the gastric body (23,29,30).H pylori is more likely to be missed if the classical histological picture is not present on the hematoxylin and eosin sections (31).This is particularly exaggerated with PPI use, for which bacterial density is low and typically more proximal (21,22) (Figure 4).The present study confirms that obtaining biopsies from both the antrum and the body increases the probability of diagnosing active infection (32).These data are consistent with previous data in which a combination of a biopsy from angulus incisura and one from the greater curvature of the corpus is needed to correctly identify all treatment failures (23).To prevent missing a true-positive result when intestinal metaplasia is present at the incisura, it is optimal that corpus biopsies, as well as biopsies from the antrum closer to the pylorus than the incisura, are taken in addition to a biopsy from the incisura angularis (33).

Figure 3) Frequency of proton pump inhibitor (PPI) use among patients
With the decline in the incidence of H pylori infection in the Western world, the need to exclude other diseases, such as Barrett's esophagus, has taken precedence in routine clinical and pathological practice.The sampling behaviour among gastroenterologists in the present study may reflect this change, or perhaps a lack of awareness of the need to take both corpus and antrum biopsies to accurately exclude active infection.As such, the problem of false-negative diagnoses in the present study (10.7%) is likely to be an underestimate, given that the present study only re-examined biopsy sets that included both antrum and corpus from each patient.We excluded 60% of biopsy sets taken exclusively for H pylori diagnosis because the endoscopist had sampled only one region (antrum only in 47% and body only in 13%).Of the remaining patients, only 57% sampled both the antrum and body (with antrum alone sampled in 11% and body alone in 26%), as shown in Table 2. Thus, when both gastric regions were sampled, a large source of error was that biopsies were not taken far enough distally to ensure that the antrum was adequately sampled.In patients positive for the infection, inadequate sampling of both body and antrum can result in a miss rate of 15% when only antral biopsies are taken (organisms migrate proximally under a variety of circumstances, including in PPI use [21,22]), and a miss rate of 21% if only oxyntic biopsies are reviewed.
In the present study, biopsies from two patients (2%) were incorrectly interpreted to be positive for the infection (ie, false positive).Active inflammation in these biopsies was mild and focal.The apices of mucous cells may be cut in a plane such that they may resemble a curved organism (Figure 5).Also, chains of cocci may be apparent and these may occasionally be confused with H pylori.It is possible that in some instances cocci represent degenerate forms of the organisms (34), although they can also be of both oral and duodenal origin.Both patients subsequently received triple therapy for H pylori infection with no side effects; so they were not seriously affected by the false-positive diagnosis.Nonetheless, it is important to point out the risks of unnecessary antibiotic use with respect to idiosyncratic or allergic reactions that are potentially fatal (35,36).
While it is possible to blame the operator (in this case, the pathologist), one also must question the context in which the mistake was made.False-negative and false-positive diagnoses are particularly observed in patients using PPIs and in patients who recently received antibiotics (31), but also if biopsies are only taken from one part of the stomach instead of both antrum and oxyntic mucosa (36).
The AGA and ACG guidelines emphasize the need to test symptomatic patients for H pylori before prescribing PPIs (9,21).PPIs were being used by 47% of patients at the time of endoscopy, with a higher frequency among patients who received a false-negative diagnosis (67% among patients with false-negative diagnosis versus 48% among patients with a true-negative diagnosis [P=0.37]).H pylori flourishes in a pH range of 3.5 to 5 (37) and, therefore, PPI use decreases the number of bacteria present, as well as facilitating proximal migration of the bacteria (38).The lower bacterial count can increase the probability of a false-negative diagnosis given the small number of bacteria present.Therefore, PPI use may be a major contributor to the inability to detect H pylori histologically in patients who have an active infection, and gastrointestinal signs and symptoms known to be related to H pylori such as ulcer-like dyspepsia.Furthermore, many patients on PPIs experience chronic gastric inflammation (39).Because pathologists rely on inflammation as an indicator of active infection, this can increase the probability of a false-positive diagnosis.This suggests that PPI use is a major contributor to both false-negative and false-positive diagnoses.
In the Canadian population, individuals belonging to high-risk groups for H pylori infection total more than four million, based on birth origin and/or area of residence (www.cdhf.ca/digestive-disorders/statistics.shtml#hpylori).H pylori infection is a potentially modifiable risk factor in many chronic diseases (gastric and nongastric) (40)(41)(42) and it is likely that H pylori infection accounts for a substantially under-recognized global burden of disease.In addition to identifying the true positive population, diagnostic accuracy is important to exclude false positives because inappropriate H pylori eradication treatment can precipitate pseudomembranous colitis, particularly in elderly patients (43,44).In addition, patients with H pylori-negative duodenal ulcers appear to experience a significantly worse outcome when treated empirically (45).
The number and location of the gastric biopsies is important for the accurate identification of H pylori because biopsies series taken from the antrum or body would result in only a significant false-negative diagnosis.If endoscopists truly wish to maximize their yield of H pylori, they must note that the highest yield is from biopsies taken from the antrum.To obtain antral mucosa, biopsies must be taken distally enough to obtain histological antral mucosa, which needs to be quite close to the pylorus.This still results in a false-negative rate of 13%, emphasizing the need for biopsies from oxyntic mucosa as well as antral mucosa.Furthermore, if the antral mucosa is not sampled, either deliberately or accidentally, the false-negative rate for Helicobacter, if present, rises to 30%.Routine sampling for Helicobacter should, therefore, always include biopsies from both the antral and oxyntic mucosa.
Given the design of the present study, one question that could not be answered was the true false-negative rate.In the present study, it was apparent that the morphological appearance of H pylori infection (chronic active gastritis that could be maximal in either the antral or oxyntic mucosa), was accompanied by parietal cell hypertrophy, indicative of hypergastrinemia, but practically of PPI use.It would, therefore, be sensible for the endoscopist to indicate whether the patient is on, or has recently been on PPIs or antibiotics, or had recently undergone Helicobacter eradication therapy, all of which can reduce the number of organisms to low or undetectable levels.Pathologists should be aware that the presence of an inflammatory pattern compatible with H pylori, and PPI-related changes in parietal cells, may be indicative of a false-negative biopsy.In addition, pathologists should also consider lymphocytic gastritis in which the bacterial load is low, especially in the oxyntic mucosa; lymphocytic gastritis is a significantly less common cause of gastritis that responds to Helicobacter eradication therapy (46).Finally, atrophic gastritis, especially with extensive intestinal metaplasia, may also result in a false-negative set of biopsies (33).
The joint AGA and ACG guidelines are provided to direct best possible care.Despite widespread dissemination and teaching of the guidelines, they are not adhered to in everyday practice.Practicing endoscopists need to be aware that they need to take biopsies from both (distal) antrum and oxyntic mucosa (ideally greater and lesser curve of both), and both clinicians and pathologists need to be aware that PPI use appears to be an important cause of a false-negative H pylori diagnosis, especially when the pattern of inflammation present suggests that organisms should be present.

TablE 2 biopsy patterns in patients with a true-positive diagnosis compared with a false-negative diagnosis Site True positive False negative
Data presented as n (%).False-negative results were from biopsy sets with few bacteria per biopsy, or from biopsy sets where the bacteria were present in one region.The antrum and body were correctly sampled in 33 of 50 (66%) patients correctly diagnosed with the infection.In contrast, the antrum and body were correctly sampled in only one of six (17%) patients who received a false-negative diagnosis (P=0.0194)