Helicobacter pylori Detection and Antimicrobial Susceptibility Testing

Biopsy specimens. (i) Specimen collection.The best specimens to culture H. pylori are biopsy samples obtained during endoscopy. Care must be taken to insure that the patients did not receive antibiotics or antisecretory drugs, especially proton pump inhibitors (PPI). Although PPI have no direct antimicrobial effect at the concentration present in the gastric mucosa (360), they indirectly interfere with H. pylori distribution in the stomach by changing the pH of its bacterial niche, leading to its disappearance in the antrum. The recommendation is not to consume these drugs 2 weeks prior to endoscopy. Obviously, the impact also depends on the dose and length of treatment; a single dose will not be as detrimental as a substantial acid suppression. Other ulcer drugs, e.g., sucralfate, have not shown any effect.

The potential contamination of biopsy specimens via endoscopes was a key issue in the past but now seems to have been resolved, at least in developed countries. Indeed, due to the discovery of prions and viruses, i.e., human immunodeficiency virus and hepatitis C virus (HCV), potentially transmitted by endoscopes, considerable attention has been given to the cleaning and disinfecting of endoscopes. Medical societies have even issued recommendations to use disposable forceps in some cases. A consequence for microbiologists is the elimination of gastric biopsy contaminants, e.g., Pseudomonas species and other environmental bacteria and sometimes H. pylori itself, which can be transferred from one patient to another. Contamination cases have been documented or suspected in the past, based on postendoscopy acute achlorhydria (534). Contamination of the stomach with oropharyngeal flora, and with bacteria from the intestine in the case of bacterial overgrowth, obviously remains. Pretreatment of the biopsy specimens by washing in saline could improve the recovery of H. pylori (239).

The number of biopsies which is necessary to diagnose H. pylori infection is a subject of controversy. A single biopsy specimen taken from the antrum (2 cm from the pylorus) gives good sensitivity but is not sufficient for a reliable diagnosis. Indeed, H. pylori may have a patchy distribution, and the more biopsy specimens analyzed, the higher the chance of organism detection (27). The recommendation is, therefore, to take two biopsy specimens from the antrum as well as two specimens each from the anterior and posterior corpus. There are some rare cases where the infection lies only in the corpus, but usually, H. pylori is present in all sites. After consumption of antisecretory drugs, the corpus may be the only site which remains positive.

Biopsy specimens for culture must be taken before specimens for histological examination, the latter being introduced in a fixative, otherwise there is a risk of transferring small amounts of fixative to the container for biopsy specimens to be used for culture.

(ii) Transport of biopsy specimens.A key point is the transport of the biopsy specimens from the endoscopic suite to the laboratory. Problems at this stage are surely the cause of many culture failures and have led to the negative opinion which some gastroenterologists have of culture. H. pylori is a fragile organism. It must be protected from desiccation and contact with oxygen and room temperature. It is mandatory not to expose the biopsy specimens to air and to place them either in a saline solution for short-term transport (4 h maximum) (368) or in a transport medium, usually consisting of semisolid agar, maintained at 4°C. A commercially available medium, Portagerm pylori (bioMérieux, Marcy l'Etoile, France), is effective for this purpose (192, 214). Culture may be delayed by 24 h if such transport media are used, allowing biopsies to be sent by mail in a cool transport container (Sarstedt, Nümbrecht, Germany) (192, 214). If these transport conditions cannot be used, it is better to freeze the biopsy specimens at −70°C or in liquid nitrogen in a dry tube and transport them frozen to the laboratory. Storage at 4°C in a medium containing 20% glycerol also led to H. pylori recovery in 81% of the biopsy specimens tested (203).

The proposal of direct biopsy specimen plating in the endoscopy suite is complicated, since biopsy specimen grinding and a special atmosphere for plate transportation are needed. This procedure is not used.

(iii) Microscopic examination of smears.A standard bacteriology test includes a microscopic examination of a smear prepared from the biopsy specimen or touch cytology (95). Indeed, gastric brushings (43, 59, 77) followed by examination with phase-contrast microscopy in the endoscopy suite allow a quick diagnosis (452) and are an alternative to the urease test (93) (Fig. 1).

• Open in new tab
• Download powerpoint

FIG. 1.

Helicobacter pylori observed on a gastric biopsy smear after acridine orange staining. Magnification, ×1,000.

Staining can also be performed, using Gram stain (166, 390), rapid Giemsa (598), or the fluorescent acridine orange stain (504). This method has a sensitivity in the range of 80%.

Besides its ability to provide a rapid result, this method allows visualization of inflammatory cells and observation of Helicobacter heilmannii, which can be identified by its typical morphology including 6 to 8 spirals (96).

(iv) Grinding of biopsy specimens.In most instances, the bacteria are not distributed homogeneously in the biopsy specimens. Therefore, if they are only streaked on plates, several contiguous organisms will lead to one colony, whereas if the same bacteria were dispersed, several colonies would appear.

Comparison of culture performed with or without grinding showed a higher number of colonies after grinding, although the number of positive specimens did not change (182). For this reason, grinding of the biopsy specimen is mandatory. The recommendation is to use an electrical/mechanical grinder and a small volume of broth. Care must be taken to thoroughly wash and sterilize the probes. Another possibility is to perform manual grinding with disposable material. This solution avoids the risk of possible DNA contamination when performing molecular techniques.

(v) Culture on plates.Due to (i) the difficulty of growing H. pylori in broth and (ii) the usual abundance of bacteria, no enrichment has been proposed, only a direct plating is performed.

The media components include an agar base, growth supplements, and selective supplements. Most agar bases are satisfactory for growing H. pylori, e.g., brain heart agar, Columbia agar, and Wilkins Chalgren agar. Concerning the growth supplement, it is mandatory to add blood or serum, which includes numerous nutrients (vitamins and oligoelements, etc.) which enhance H. pylori growth. The proportion of blood or serum can be 5, 7, or preferably, 10%. Red blood cells can be lysed for these growth substances to be more readily available. Animal blood, e.g., sheep and horse blood, can be added, but human blood, to be used after appropriate testing, seems to confer a slight advantage (570). Other growth supplements such as egg yolk (569), charcoal (177), starch (46), bovine serum albumin, and catalase (208) are used. Cyclodextrins, which are cyclic oligosaccharides produced from starch by enzymatic treatment retaining the same properties as starch, are employed (423). More recently, ferrous sulfate, sodium pyruvate, and swine mucin have been proposed to enhance H. pylori growth (235).

Cellini et al. proposed a blood-free medium supplemented with isovitalex (2%) and hemin (10 mg/liter). They also added urea (20 g/liter) and a pH indicator (phenol red) to identify the urease-positive colonies (64). However, our experience does not favor the incorporation of a high urea concentration because the bacteria can be inhibited by the high pH induced. Indeed, pH is an important variable to consider; H. pylori grows best at a slightly acidic pH (5 to 6), in agreement with its ecological niche, the mucus layer, where a pH gradient exists (243). Another supplement which may be helpful to readily identify H. pylori colonies is 2,3,5-triphenyltetrazolium chloride (40 mg/liter). This compound is reduced by H. pylori to insoluble red formazan complexes, resulting in easily distinguished pigmented golden colonies (458).

Selective supplements are also quite important because of the presence of contaminating flora as mentioned before: buccal flora consisting mainly of gram-positive cocci, intestinal flora in the case of duodenal reflux, and bacterial overgrowth. Furthermore, Candida species may colonize ulcer craters.

Different selective supplements containing antimicrobial compounds have been proposed: vancomycin or teicoplanin to inhibit gram-positive cocci; polymyxin, nalidixic acid, colistin, trimethoprim, or cefsulodin to inhibit gram-negative rods; and nystatin or amphotericin B to inhibit fungi. The Dent supplement, a modification of Skirrow's formula in which cefsulodin replaces polymyxin and amphotericin B is added, is commercially available (105).

Several studies performed in the early days of H. pylori detection showed the importance of using both a nonselective medium and a selective medium (273, 446) or even two different selective media (13, 514).

A critical point is to use fresh media (less than a week old) which is kept in closed boxes at 4°C to maintain humidity and avoid light exposure.

Helicobacters are microaerophilic and capnophilic bacteria. When their growth was studied according to the oxygen concentration, optimal growth occurred with a pO₂ of 2 to 10 KPa, whereas no growth occurred at a pO₂ of air (243). In contrast to these findings, Xia et al. claimed that most of their H. pylori strains grew under aerobic conditions but with reduced cell counts and smaller colonies (585). This discrepancy can be explained by the fact that, at low bacterial concentrations, H. pylori is microaerophilic, while at high bacterial concentrations, it can grow aerobically (50). Several systems can be used to achieve a microaerobic atmosphere, from the most sophisticated systems, such as a microaerobic cabinet or an incubator with an adjustable gas level, to jars in which the adequate atmosphere is created with an automatic apparatus (Anoxomat, MART Microbiology BV, Lichtenwoorde, The Netherlands) or with H₂-CO₂-generating packs. Besides their cost, the gas-generating packs take several hours to produce the optimal atmosphere. The atmosphere in jars will vary according to the quantity of bacteria consuming oxygen; therefore, the gas pack should be changed every other day. While H. pylori growth is possible in a candle jar (136), it takes a longer time and results in small colonies; therefore, we do not recommend this solution.

The optimal culture temperature is 37°C, testifying to the adaptation of this bacterium to humans. Certain strains can, nevertheless, grow at 30°C or 42°C (359). For primary culture under optimal conditions, colonies may appear after 3 days and are at their optimum on day 4. However, in the case of negative culture, a 7- to 10-day incubation is recommended to ensure that the result is negative; if only a few organisms are present, this time lapse may be necessary to visualize the colonies (544).

In contrast, subcultures only take 2 to 3 days. When few colonies are present, the recommendation is to subculture by plating the colonies on a small area of the agar plate. It is important to remember that once H. pylori reaches its growth plateau, it becomes coccoidal and loses its viability, most likely due to a lack of adequate nutriments.

Culture from biopsy specimens has the potential of leading to a high sensitivity, given that only one bacterium can multiply and provide billions of bacteria. However, both strict transport conditions and careful handling in the laboratory are necessary. Culture is extremely dependent on desiccation, contact with air, and temperature, as previously stated, and sensitivity reaches 95% at best. In contrast, its specificity is 100% because, once the bacterium has grown, all phenotypic and genotypic identification tests can be performed.

Quantitative culture of viable bacteria present in gastric mucosa can be performed. However, given the variability in the biopsy specimen size (1 to 10 mg), it is recommended to weigh the biopsy specimen. The bacterial load can be extremely variable, ranging from 5 to 10⁵ CFU/mg, in one study (24). Reproducibility is higher in the antrum versus the corpus. A high H. pylori density has been associated with bacterial virulence determinants (cagA, vacA, s1m1), gastric inflammation, and duodenal ulceration, suggesting a central role in pathogenesis (21). Given that quantitative culture is expensive, time-consuming, and not very precise, a semiquantitative evaluation is most often performed.

(vi) Broth culture.Similar media, supplements, and conditions used for agar have been applied to broth culture (492), while brain heart infusion may be preferable for studies on physiology and metabolism (315). A chemically defined broth medium consisting of serum-free Ham's F-12 medium has also been described previously (515). The growth atmosphere is important, and two recommendations have been proposed in this respect: (i) constantly check the tubes in a microaerobic atmosphere (393) or (ii) prepare a biphasic medium with a thin broth layer in a cell culture flask incubated in a microaerobic atmosphere (491). Resulting bacteria are more likely to have a typical morphology and be motile. They can also aggregate. The use of gas-permeable cell culture flasks may improve H. pylori growth (490).

To obtain an important bacterial cell mass, subculture in successively increasing volumes from 5 ml to 500 ml is recommended (492). An alternative is to use fermenters.

(vii) Phenotypic identification.The growth of small, circular, smooth colonies observed after 3 to 4 days on the selective media plated with gastric biopsy specimens is an important criterion for H. pylori identification. No hemolytic activity is readily observed but may appear after a few days at 4°C.

Microscopic examination of the cultured bacteria may show a morphology different from the bacteria present in the biopsy specimen, i.e., bacilli which are neither spiral shaped nor motile, but straight or curved. Indeed, spiral shape and motility are directly related to the viscosity of the medium (207).

The identification of cultured bacteria consists essentially of testing for the presence of certain enzymes: cytochrome oxidase, catalase, and urease (312) and eventually γ-glutamyl transpeptidase, leucine aminopeptidase, and alkaline phosphatase. Aminopeptidases and esterases (C₄ to C₁₂) are also present (359).

The family Helicobacteraceae is comprised of the genera Helicobacter and Wolinella. Helicobacteraceae and Campylobacteraceae are in the Epsilonproteobacteria, class nov. according to the latest edition of Bergey's Manual (157).

Cytochrome oxidase is present in all members of the Epsilonproteobacteria. It is usually detected with special reagents on a disk or a strip. Catalase is also present in all helicobacters and most members of the Campylobacteraceae and is detected by introducing a loopful of bacteria into a drop of hydrogen peroxide and observing a very abundant production of bubbles. Nevertheless, catalase-negative mutants of H. pylori have been described previously (571). Urease is definitely the most important enzyme for identification. To survive in its particular ecological niche, H. pylori produces large amounts of this enzyme to buffer the acidic medium and creates a microenvironment (342). Mobley et al. reported that in H. pylori, 6% of the total protein content was urease (381). When a loopful of H. pylori is put in contact with a few drops of urease medium, the color change occurs instantaneously regardless of the formulation. Other diagnostic tests are indeed either strictly based on urease, like the rapid urease test and urea breath test, or partially based, like serology and PCR, which may target urease genes.

The ApiCampy strip (bioMerieux, Marcy l'Etoile, France) allows identification of H. pylori via positive urease, γ-glutamyl transpeptidase, and alkaline phosphatase and negative nitrate reductase and hippuricase, all located in the first part of the strip, which explores preformed enzymes. The second part of the strip, which explores organic compounds as the only carbon source, cannot be used because growth of the organism cannot be supported by the minimal medium used.

Among tolerance tests utilized to differentiate Campylobacter species, H. pylori can grow with 2,3,5-triphenyltetrazolium chloride (0.4 and 1 mg/liter), sodium selenite (0.1%), and glycine (1%), but not with glucose (8%) and sodium chloride (3.5%). H. pylori is susceptible to cephalothin but resistant to nalidixic acid, with some exceptions (359).

When the bacteria are isolated from gastric biopsy specimens, phenotypic tests are sufficient for a precise identification. However, this is not the case when the bacteria are isolated from other specimens, e.g., stools, saliva, and the environment. Because other known or unknown bacteria sharing the same characteristics may be present, it is mandatory to confirm the identification with molecular methods. If PCR is used, several reactions targeting at least two different genes must be positive simultaneously. Sequencing of the urease gene may be worth performing because data are available for a number of Helicobacter species. Another spiral bacterium which we and others (478) identified in the stomach by culture is the urease-negative Campylobacter jejuni subsp. doylei. It was isolated for the first time from the stomach in 1985 (247).

Culture is an important step to perform further antimicrobial susceptibility testing (see below) and strain typing when necessary (49).

(viii) Question of coccoidal forms.As previously mentioned, when nutrients are lacking, H. pylori loses its spiral shape and becomes progressively coccoidal. Most people believe that these forms are both nonculturable and nonviable (278). However, others claim that some of them may be viable but nonculturable and constitute a resistant form of the bacterium (30).

The diagnostic implications are as follows: (i) a discrepancy is possible between the results from culture and from molecular tests as long as the DNA is not degraded and (ii) it is therefore important to subculture the colonies as soon as they reach their optimal size because afterwards they may die.

(ix) Strain maintenance. H. pylori is difficult to maintain. Colonies can survive on plates for a week provided they are kept in a microaerobic atmosphere at 4°C. For a long-term conservation, bacteria must be frozen at a low temperature (−70°C freezer or liquid nitrogen). Different broth media have been used, always with a cryoprotective agent, such as glycerol, either in cryotubes or on beads. The freezing-thawing process is always deadly for the bacteria, and only a small proportion survives. For this reason, it is mandatory to use bacteria in their exponential growth phase, as they are more likely to survive. Frozen H. pylori specimens can be maintained for decades at −70°C. Freezing at −20°C is insufficient, and lyophilization appears to be difficult. Loss of viability is noted particularly during the dehydration phase (428). Storage of lyophilized vials at 4°C could help bacterial survival (506).

Other culture specimens. (i) Gastric juice.Gastric juice reflects the whole stomach and is easier to obtain than biopsy specimens but still necessitates an invasive procedure, i.e., introduction of a nasogastric tube. H. pylori is found in gastric juice due to the turnover of gastric mucosa, but the sensitivity is much lower than culture from biopsy specimens, with the best results for the former being in the range of 60% (149, 555). It does not seem crucial to buffer the collected liquid.

A less invasive procedure has been proposed by Perez-Trallero et al. involving Enterotest (HDC, San José, CA), also named the string test (443). A gelatin capsule fixed to a nylon fiber is ingested by the patient. After 1 h, the fiber is removed, with the capsule having been digested. The distal part of the fiber can be used for H. pylori detection by culture but also by urease test or PCR. This technique was applied by others but with sensitivity rates ranging from 38% to 97% (304, 482, 524). Leodolter et al. proposed to use the string test after the urea breath test as a diagnostic package for the identification of treatment failure and antibiotic resistance without the necessity of upper gastrointestinal endoscopy. The sensitivity of culture from the string test was 87% compared to culture from gastric biopsy specimens (303).

To study the possible transmission of H. pylori during vomiting, Parsonnet et al. obtained vomitus after induction of emesis in 16 volunteer subjects. They were able to grow H. pylori from all of the samples, often in high quantities (434). In contrast, spontaneous vomitus allowed the isolation of H. pylori in only one of four positive cases (306). The delay before performing the analysis on unbuffered gastric juice is critical: the survival rate was 62% after 2 h versus 15% after 24 h (154).

Gastric juice is well adapted for performance of quantitative culture and estimation of the potential for transmission of this organism (153, 593). In the study of induced vomitus from infected subjects, H. pylori grew in high quantities (>10³ CFU/ml), while low quantities (50 to 500 CFU/ml) were found in postemesis saliva and cathartic stools (2,000 to 5,000 CFU/ml) in most of the subjects (434).

(ii) Blood. H. pylori may reach the blood flow from the gastric mucosa in the case of ulceration and hemorrhage. However, only one isolate was found in a blood culture from a patient with a gastric lymphoma. The strain grew after 5 days in a Bactec NR730 aerobic bottle (403). Indeed, regular blood culture systems may not be optimal for H. pylori growth. Conversely, blood culture systems used for fastidious organisms, e.g., brucella broth, led to satisfactory growth when experimentally inoculated (255).

(iii) Liver.Numerous attempts have been made to isolate H. pylori or other helicobacters in the liver and bile ducts in different hepatic diseases. Only one strain has been grown from a patient with Wilson disease in Brazil (103).

Gastric culture of other helicobacters.The other Helicobacter species possibly present in the human stomach is Helicobacter heilmannii, previously known as Gastrospirillum hominis. This bacterium is commonly found in the stomachs of pets (dogs and cats) and pigs. It rarely colonizes humans and is considered a zoonosis (365).

This organism is nonculturable on standard media; it can essentially be cultured in vivo by introducing it into the mouse stomach (295). We succeeded in this endeavor three times in a clinical context.

In 1996, Andersen et al. were the first to grow H. heilmannii from a 23-year-old patient with dyspepsia on the same plate medium used in their laboratory to grow H. pylori, and growth occurred within 5 days in a standard microaerobic atmosphere. From primary culture, only 10 to 15% of the bacteria exhibited a typical morphology contrary to the majority after subculture (10). This is indeed the only positive culture on a plate for this bacterium which was later identified as Helicobacter bizzozeronii (233).

In summary, culture is still a valuable tool. It is considered the reference method when comparing the accuracy of noninvasive techniques. Its specificity is maximal, since the availability of colonies allows identification with all techniques, including sequencing of key genes (urease, vacA cytotoxin, cag pathogenicity island [PAI] open reading frames). Its sensitivity may be inferior, due to the stringent transport conditions required. Theoretically, its sensitivity should be the best because the presence of one organism in the inoculum is enough to give a positive result. However, in studies evaluating noninvasive tests, it is now the rule to consider a specimen negative by culture as H. pylori positive if both histology and urease tests are positive. Furthermore, culture is still the ideal method for antimicrobial susceptibility testing and typing, although molecular methods can also be used now directly on specimens for this purpose.

Specimen collection and transport.The usual recommendation derived from the Sydney system (109, 453) is to obtain 2 biopsy specimens from the antrum and 2 specimens from the corpus. While the rationale is essentially to type the gastritis which may be present, H. pylori detection benefits from the high number of specimens tested. As indicated before, sensitivity increases with the number of biopsies (27). Bacteria are usually present at both sites even if the lesions occur essentially in the antrum. When topographic studies of H. pylori distribution and gastritis were performed, the best site suitable for diagnosis was the lesser curvature of the midantrum, while for the corpus, there was a discrepancy between greater and lesser curvatures (159, 375, 484).

Biopsy specimens are immediately introduced into a fixative of 10% formaldehyde. Consequently, there is no transport problem, since the material can be sent at room temperature by regular mail and is no longer infectious. This fixative maintains the morphology of the bacteria, and most stains can be used. However, storage in formaldehyde is limited because, after a week, the diagnosis becomes difficult (142). The Bouin fixative must be avoided because it alters the bacterial morphology.

Preparation of histological slides.Ideally, an orientation of the biopsy specimens should be made before paraffin embedding to have sections which show the surface epithelium where bacteria are essentially located. The use of filter paper before fixation must be discouraged because the absorption of gastric mucus by the filter paper decreases sensitivity (590). It is usually necessary to cut three thin sections (3 to 5 μm) at different levels. The accuracy of this technique is very much dependent on the quality of the histological preparation, which must allow observation of the surface epithelium and the crypts.

There is no stain specific for H. pylori. The routine hematoxylin-eosin stain is not well suited for H. pylori detection because of the weak contrast between the bacteria and the mucus. There are several special stains (Table 1) which allow better visualization than the standard hematoxylin-eosin stain. This is essentially due to the acidophile characteristic of these stains which allows the staining of H. pylori DNA and not of the mucus where the bacteria are present, leading to a good contrast. The conclusion of the presence of H. pylori can be made easily, since it is seldom that bacteria other than H. pylori with similar morphology occupy this ecological niche.

The Warthin Starry stain allows an excellent visualization of the bacteria, but the procedure is difficult to carry out. This technique is time-consuming and costly and requires extemporaneous preparation of the reagents to be used. The Giemsa stain is probably one of the most popular stains because of its simplicity and good contrast (134, 269). A special stain has been proposed for H. pylori: the Genta stain with which both the bacteria and gastritis lesions can be observed (160).

Slide examination.The typical morphology of H. pylori in biopsy specimens is a comma or S-shaped bacillus (2.5 to 4 μm long and 0.5 to 1 μm thick). The bacteria are observed on the epithelial surface at high magnification. They adhere to the gastric mucus cells or are free in the mucus and may eventually be present in the intercellular spaces; very few bacteria, if any, are intracellular. Some organisms may be found in the canaliculi of the fundic parietal cells. Others may be present on metaplasic antral cells outside of the stomach, especially in the duodenal bulb (584) or in the esophagus. They do not colonize the intestinal metaplasia but can adhere to areas of incomplete intestinal metaplasia which consists of a hybrid epithelium with both gastric and intestinal features (426).

There are limits to this technique. First, it is necessary to obtain good quality biopsy specimens, which is not always possible, and from a series of patients, it is common to have biopsy specimens where few epithelial surfaces can be observed. Second, when there is a low number of bacteria present, and in the event that they do not have a typical morphology, it is difficult to draw a conclusion.

The same conditions as previously described for culture (i.e., unsuccessful eradication treatment, consumption of PPI or antibiotics) may also jeopardize the accuracy of histological diagnosis (237). These conditions may trigger the evolution of the bacteria from a typical morphology to a coccoidal form which does not allow a diagnosis to be made. A high proportion of coccoidal forms was noted after gastrectomy in one study (65).

The proportion of cases with a chronic active gastritis but no bacteria present on the slides is currently increasing. The main reason is probably an uncontrolled usage of PPIs. Indeed, omeprazole administered as a monotherapy for 4 weeks decreases the bacterial density in both the antrum and the corpus (185) but at a differing magnitudes from one patient to another. This effect progressively disappears a few weeks after stopping the PPIs. Similar phenomena can be observed with H₂ antagonists but to a limited extent. Antibiotics used as a monotherapy also have a dramatic effect on the bacterial density even if they rarely lead to H. pylori eradication. Another entity, namely focally enhanced gastritis, seems to offer a picture similar to H. pylori gastritis. This gastritis is frequently observed in patients with Crohn's disease and, according to Pusztaszeri and Genta, is an indication to look for this disease (456).

Finally, the histological diagnosis of H. pylori infection is very much dependent on the expertise of the pathologists and on the time devoted to the diagnosis. A comparison of results obtained from experienced pathologists performing routine diagnosis highlights this problem (328, 382). The interobserver reproducibility was tested in 3 series (11, 74, 483) and measured by the Kappa test. A value of >0.75 corresponds to an excellent reproducibility, values of 0.4 to 0.75 correspond to an acceptable reproducibility, and a value of <0.4 is insufficient. The Kappa value was 0.69 and 0.74 in 2 series (11, 483), while in the third it varied from 0.39 to 0.82 according to observer pairs (74). In this last study, the intraobserver value was 0.65 to 0.88. These data cast doubt on the reliability of this diagnosis. Other studies have confirmed this problem (125, 238).

When doubt occurs, pathologists may use the presence of active gastritis as a surrogate, since this condition is almost pathognomonic of H. pylori infection. Immunostaining with a specific H. pylori antibody (Dako, Glostrup, Denmark) can also be used. In all of the studies performed, immunohistochemistry had the highest sensitivity and specificity. It allows a better interobserver agreement than routine histology and can also be performed with an autoimmunostainer (17, 110, 238, 526). In situ hybridization has been developed using a 16S rRNA gene probe labeled with digoxigenin (246) or with fluorescein (fluorescence in situ hybridization [FISH]). The latter method is now commercially available and will be described in detail later, since it can also be used for antimicrobial susceptibility testing.

Histological results must be reported according to guidelines drawn up in 1990, known as the Sydney system (453). The presence of bacteria in the corpus and in the antrum is expressed semiquantitatively on a scale of 0 to 3. In addition, the histological characteristics of the gastric mucosa (inflammation, activity, atrophy, intestinal metaplasia) are also reported. The recent update of the Sydney system proposes inclusion of biopsy specimens from the incisura, an area where premalignant lesions are commonly found (109).

Furthermore, histology allows the diagnosis of the nonculturable H. heilmannii, which is seldomly present and easily identified by its typical morphology: it is longer than H. pylori with 6 to 8 spirals, nonadherent to epithelial cells, and commonly localized in aggregates in the crypts' lumen (Fig. 2). It also induces gastritis, which is sometimes transient, and may lead to gastric ulcer and gastric MALT lymphoma (96, 216, 394, 505). The FISH method can also be used to detect this bacterium (528).

• Open in new tab
• Download powerpoint

FIG. 2.

Helicobacter heilmannii observed on a gastric biopsy histological preparation after hematoxylin-eosin staining. Magnification, ×1,000.

In summary, histological detection can reach a sensitivity of 95% under optimal conditions, with the limits being the quality of the material and the pathologist's expertise. The presence of inflammation (lymphocytes) and especially inflammatory activity (polymorphs) is an obvious indication for an in-depth search for helicobacter organisms. The specificity is also in the range of 95%. The presence of other bacteria on the mucosa may be a cause of false-positive results if there are few bacteria with atypical morphology. However, the pathologist can use immunohistochemistry or FISH for confirmation.

Other histological methods.Electron microscopy can be used to detect H. pylori on the mucosa (145). However, this technique is too time-consuming and difficult to employ routinely.

Biopsy specimens frozen with a cryostat before an extemporaneous examination have been used as a rapid method, but their interpretation is difficult (244).

Principle.The discovery that H. pylori was a strong urease producer was made by Langenberg et al. early in the bacterium's history (290). It was first applied to direct rapid diagnosis by McNulty and Wise (356).

To adapt to its special ecological niche where the concentration of urea diffusing from blood to the gastric mucosa is low, H. pylori produces large amounts of urease. H. pylori urease also has the highest specific activity (36 ± 28 μmol/min/mg of protein) among bacterial ureases (381).

The other urease-positive bacteria present in the gastric mucosa, i.e., streptococci and staphylococci, produce a lower amount of urease, which does not interfere in a short-time detection (<2 h), rendering the method specific to H. pylori.

When a biopsy specimen containing H. pylori is introduced into a urea-rich medium, the urease breaks the urea down into carbon dioxide and ammonia. The ammonium ion increases the pH, and a pH indicator, e.g., phenol red, changes color, in this case from yellow to red or violet.

The different urease media commonly utilized in bacteriology, i.e., Christensen medium and urea-indole medium, can be used, but specific media are preferable.

Different ways to improve the sensitivity of the tests include: (i) increasing the urea concentration from 2 to 10% (41), (ii) incubating at a higher temperature (37°C) (287, 595), and (iii) suppression of the buffer (15), while the size of the biopsy does not appear to be important (286, 594). Unbuffered media have a limited shelf life (<5 days at 4°C) (406).

The first-generation commercial kits were agar based, e.g., the CLO test (Kimberley-Clark, Neenah, WI). The new generation kits are strip-based tests with two areas separated by a microporous membrane, one where the urease hydrolyzes urea and the other where NH₃ is trapped and changes the pH (PyloriTek, Serim, Elkhart, IN).

The agar-based tests exhibit a good sensitivity only after 24 h (90 to 95%), compared to 70 to 80% after 1 h. Therefore, they cannot be classified as rapid tests after such a long delay. Furthermore, other urease-positive bacteria from the mouth can decrease the specificity when reading is performed after 24 h (489).

H. heilmannii, which may be present in the stomach, as previously indicated, is also a urease-positive bacterium but may not give a positive result, as the bacterial load is often limited. Therefore, the distinction between H. heilmannii and H. pylori is made by histological examination as previously indicated.

Attempts to obtain a rapid quantification of H. pylori based on the urease produced have been performed using an ammonia electrode measuring the ammonia liberated from samples placed in a urea solution (51).

Factors influencing results of the urease test.A limit to the urease test is the bacterial load necessary to obtain a sufficient sensitivity. A semiquantitative evaluation of the bacteria present by histology clearly showed that false-negative urease tests corresponded to the lowest histological scores for H. pylori (38, 522, 597). It seems that at least 10⁵ bacteria are necessary for a valid result. This amount may not be present 4 weeks after the failure of eradication therapy, which makes this test less advisable for posteradication follow-up. Treatment with PPI may also jeopardize the result. By changing the milieu where the bacteria are present, especially the antrum, PPI renders it inhospitable and the bacterial load decreases. In addition, PPI themselves may have antiurease properties (533). Another reason for a false-negative test is the presence of intestinal metaplasia, which also corresponds to an inhospitable environment for H. pylori.

Per endoscopy, stains such as methylene blue may negate the urease test (198) so the specimens for this test must be taken before employing these stains.

The great progress in this area was the introduction of a strip-based test (PyloriTek) in 1995. In the first study, Rogge et al. compared this new test to the CLO test using histological detection as the reference. PyloriTek showed a 99% sensitivity and 95% specificity after 2 h, which is superior to those of the CLO test (468).

These results have been confirmed since then in several studies (288, 455, 592, 596). Indeed, the sensitivity of the final reading is not significantly different from that of the CLO test, but the last reading can be done after 1 h versus 24 h for the CLO test. For routine use, most endoscopists read the CLO test earlier than recommended, which leads to a marked decrease (20%) in sensitivity (288, 454).

Other specimens for the urease test.The use of the urease test on specimens other than gastric biopsy specimens, e.g., oral specimens, must definitely be discouraged because many other urease-positive bacteria (Staphylococcus spp., Streptococcus spp., etc.) can be present and give false-positive results.

Biopsy specimens. (i) Standard PCR for the detection of H. pylori.The choice of a target is important in designing the primers which must be specific for H. pylori but conserved in all strains of the species. It is therefore necessary to know the DNA sequence of the target in as many strains of H. pylori and as many strains of other bacterial species as possible. The first targets used were the genes of the urease operon: ureA (78) and glmM, formerly named ureC (107), or the 16S rRNA gene (222, 226) (Table 2). While highly conserved in bacteria, the 16S rRNA gene exhibits sequences which are specific for different species. However, the specificity of this PCR has been challenged because the 109-bp amplicon obtained with primers targeting the 16S rRNA gene reacted with human tissue samples (73). Other genes with unknown function, e.g., the gene encoding a 26-kDa protein identified as tsaA (201), or random sequences were used (540).

Among the drawbacks of PCR are (i) the possible existence of Taq polymerase inhibitors which can decrease the sensitivity of the reaction (520) and (ii) the possible contamination of the specimen by exogenous H. pylori DNA, which alters the specificity. Fortunately, it is uncommon to find Taq polymerase inhibitors in gastric biopsy specimens, so the procedure of DNA extraction can be simplified. The simplest technique consists of grinding the tissue to release the bacteria and bacterial lysis by boiling at 100°C in a lysis buffer.

Theoretically, PCR can detect only one copy of the target DNA when tested in water, but the amplification efficacy is not as good when biological material is concerned. Most studies have shown that standard PCR sensitivity is similar to that of culture for pretreatment diagnosis (31, 132, 297, 386, 540, 553). Several methods have been proposed to improve sensitivity.

To improve DNA extraction, phenol extraction or the use of special extraction kits (QIAGEN, Valencia, CA) to eliminate PCR inhibitors is superior to simple boiling (563). An internal control may help (520) but does not seem to be mandatory.

The use of a nested or seminested PCR was suggested. However, the use of nested PCR increases the risk of airborne contamination by amplicons and must be discouraged (131). Performing a second PCR with the same primer has also been proposed (502).

To increase the sensitivity of amplicon detection, a probe hybridization method can be used instead of the standard detection of the amplicons based on their size by electrophoresis (123, 567). New methods in liquid phase (DNA-enzyme immunoassay) (282, 385) and the reverse dot blot line probe assay (LiPA) (552) have been proposed. In addition, this approach definitely increases the PCR's specificity.

The use of reverse transcription-PCR (RT-PCR) has been reported by several authors (123, 126, 418, 438). Given that it is based on mRNA, it determines the viability of the bacteria present, but no improvement in sensitivity has been shown.

A significant improvement in sensitivity occurred with the introduction of real-time PCR protocols. As is always the case when a new method is more sensitive than the reference method, it is necessary to prove that the results are not falsely positive. One possibility is to conduct a follow-up of the patient over a few months to confirm the reality of the infection.

Specificity is not usually a problem when dealing with gastric biopsy specimens. However, the possibility of false positives arises when endoscopes (470) or grinding apparati in the lab are not correctly cleaned. It is recommended to use strict cleaning procedures in the endoscopy suite and disposable material in the lab to avoid such contamination. An important risk of contamination occurs when the amplicons are analyzed by electrophoresis. For this reason, there are strict rules stating that three independent rooms must be used for performing the different steps of a PCR.

An advantage of PCR is that DNA does not require strict transport conditions, and it has been performed on urease tests sent by mail (108, 314, 316).

(ii) Standard PCR for detection of pathogenic factors of H. pylori.An important application of standard PCR is the detection of specific pathogenic factors of H. pylori. There are two main pathogenic factors: the cag PAI and the polymorphism of the vacA gene. Other genes involved in adherence (babA2, sabA) or in pathogenicity (oipA, dupA, iceA) can also be detected by PCR.

Numerous studies have shown that strains harboring the cag PAI are associated with more severe diseases, especially peptic ulcer disease (84, 281) and gastric adenocarcinoma (36), as well as precancerous lesions (130, 450) and extradigestive diseases (148, 448). In contrast, no association was found with gastric MALT lymphoma except for the vacA m2 allele (100, 296), and a negative association exists with gastroesophageal reflux (319). However, these associations have not been found everywhere. In fact, in Asia, most of the strains are cag PAI positive, and therefore, this pathogenic association cannot be used in case control studies (497, 501).

The main gene of the cag PAI is cagA. It was the first to be used as a marker for the presence of cag PAI (281). We now know that CagA proteins induce morphological alterations of the gastric epithelial cell, while other genes of the cag PAI lead to production of interleukin 8 (86, 588) via the intracellular Nod1 receptor and the nuclear factor κB pathway. The chemokine production is recognized as being responsible for an increased gastric inflammation and subsequent disease development.

While detection of the cag PAI has limited consequences on an individual basis, it has interesting implications in epidemiological and pathogenic studies. In addition to the disease association already mentioned, cagA-positive strains are easier to eradicate (44, 548).

It is possible to look for antibodies directed against the CagA protein or to detect the cagA gene. The first studies were performed on H. pylori strains by PCR. The variability in cagA sequences favors the successive use of two PCRs (281) or a Southern hybridization (501) to avoid false-negative results. The same PCR can be applied directly on DNA isolated from gastric biopsy specimens and gives similar results with bacterial colonies (67, 283, 433). The former is the preferred technique because it gives a picture of what exists in vivo (194). Material from a positive CLO test as well as from formalin-fixed embedded tissue may allow cagA detection (349, 487).

The presence of both cagA-positive and cagA-negative isolates in the same patient has been described previously (94, 137, 523). Comparison of the isolates by molecular typing showed that they were indeed the same strain (523), indicating the possible deletion of part of or the entire cag PAI. By using in situ hybridization with biotylinated probes for an H. pylori common gene and an oligonucleotide specific for cagA, it was possible to identify cagA-negative bacteria in the mucous gel of the apical epithelial surface, whereas cagA-positive bacteria colonized the immediate vicinity of epithelial cells or the intercellular spaces (57).

H. pylori also produces a vacuolating cytotoxin, VacA, which has been associated with the more severe diseases, e.g., peptic ulcer disease (144) and gastric adenocarcinoma (371). The gene encoding this cytotoxin is present in all strains but exhibits a mosaicism in the terminal (s) and median (m) regions. There are several alleles corresponding to various amounts of toxin produced: s1m1 corresponds to the highest production, followed by s1m2, while strains with the s2m2 allele do not produce any toxin (19). A PCR has been proposed to identify the genotype (20).

When both cagA and vacA detection is performed, a strong association exists between the presence of cagA and vacA s1, corresponding to strains with the highest production of cytotoxin (23, 202, 217, 475, 547, 589) as well as more severe pathology and disease.

A test has been designed to detect H. pylori and its main virulence factors: the INNO-LiPA (Innogenetics, Ghent, Belgium). This assay is comprised of a strip containing multiple specific probes for the vacA s region, the vacA m region and cagA. Some of the corresponding H. pylori genes are amplified with a biotin-labeled PCR primer and subsequently analyzed by a single-step reverse hybridization on the strip (552). Other technical developments include a multiplex PCR for vacA and cagA genotypes (67) and real-time PCR melting curve analysis using SYBR green I dye (477).

(iii) Real-time PCR.The new real-time PCR technique is a breakthrough in the diagnosis of H. pylori because it allows not only a quick and precise detection of H. pylori but also its quantification and the detection of point mutations associated with antibiotic resistance.

The principle consists of following the increase in amplicons formed in real time. This can be done by using an intercalating agent, e.g., SYBR green I dye, or the fluorescence resonance energy transfer (FRET) principle (576).

Primers specific for the target gene are designed, as are a biprobe on the amplicon, the acceptor probe (sensor probe) 5′ labeled with LC-Red 640, which hybridizes where the mutation site is located, and a donor probe (anchor probe) 3′ labeled with fluorescein, which hybridizes 3 to 5 nucleotides upstream. When the anchor probe fluorophor is excited, it transfers the energy to the sensor probe fluorophore, which emits a signal. To detect point mutations or a polymorphism, a melting curve analysis is performed.

Since the essential application is not limited to H. pylori detection, real-time PCR will be reviewed with susceptibility testing.

(iv) PCR on fixed biopsy specimens.Performing PCR on formaldehyde-fixed paraffin-embedded material is a very attractive possibility because (i) it would allow the study of archive material and (ii) for fresh biopsies, there would be no maintenance problem. Unfortunately, fixed specimens are far inferior to frozen material. The best results are obtained with PCRs producing short amplicons because DNA can be broken by fixatives.

(v) PCR for quantification.PCR is also used for H. pylori quantitation first by using a competitive PCR, followed by the development of a real-time PCR which has considerably facilitated the process.

Competitive PCR is based on the coamplification of an internal standard (PCR mimic) and a target DNA sequence of different sizes with the same set of primers. Various amounts of PCR mimic are added to the target, and they compete for primer binding and amplification. A visual comparison allows the detection of bands with the same intensity indicating the amount of target DNA present (389). A colorimetric detection is also possible (430). The quantity of H. pylori in gastric mucus correlated with other invasive tests as well as with the urea breath test (UBT) in a study by Furuta et al. (152). They also used this method for patient posteradication follow-up with a high predictive value (153).

Real-time PCR was first used to quantify H. pylori DNA in gastric biopsy specimens in 2002. He et al. carried out a noncompetitive PCR on a LightCycler apparatus and FRET technology (209). The detection limit was 10³ target copies (ureC) and there was a linear variation from 10³ to 10⁹ CFU/ml. While this study brings forth interesting data, the presence of target DNA in supposedly H. pylori-negative biopsy specimens casts a doubt on the specificity. The authors did not use surrogate markers of infection or follow up in these discrepant cases. The same method was used by Lascols et al. but with another target, 132 bp of the 23S rRNA gene. A linear correlation was also found with as few as 300 CFU/ml up to 3 × 10⁸ CFU/ml. The correlation with other invasive tests used semiquantitatively (histology, culture) was excellent (r, 0.86). Sensitivity and specificity of the technique were 97% and 94.6%, respectively (292). Another technological tool, the TaqMan, was used on paraffin sections of the biopsy samples (264). There was a good correlation with the UBT values.

DNA extraction is an important step in these procedures. It was evaluated in a mouse stomach model. Homogenization with glass beads, followed by use of the QIAGEN DNA mini tissue kit, was found to be the most suitable method (472). The same model was used to test the reliability of real-time RT-PCR to assess the abundance of transcripts in the gastric mucosa. Comparison to infected humans confirmed the value of this tool (469).

Other specimens. (i) Gastric juice.As for culture, PCR targeting mainly ureA genes has been applied to gastric juice aspirates with a good sensitivity and specificity (254, 347, 535, 572).

Since the introduction of the string test (442), this method of gastric juice collection has been favored (112, 140, 471, 591). The cagA gene, which is highly prevalent in the Far East, has been the target in a study using gastric juice in Taiwan (564).

(ii) Blood.PCR is not performed on blood samples for the detection of H. pylori. Interestingly, Dore et al. tested serum samples from a small number of infected patients and controls with genus-specific primers (namely C97-C98), as well as with primers designed on the conserved region of the vacA gene; 65% of the H. pylori-positive patients were positive with the Helicobacter genus primers and 75% of them also resulted in the amplification of vacA sequences. Cloning and sequencing of 16S rRNA gene amplicons of 3 samples had 99% identity with H. pylori. This surprising finding indicates that H. pylori DNA might circulate in peripheral blood (116).

(iii) Specimens from other sites.PCR was used to detect H. pylori in various sites along the digestive tract. The bacterium was detected in 2 of 46 appendix specimens from patients with appendicitis (4%) (437), in 4 of 12 ethmoid specimens from patients with chronic rhinosinusitis (33%) (429), and in 7 of 30 tonsil and adenoid tissue samples (30%) (76), probably following reflux.

PCR on liver specimens also yielded positive results for H. pylori-like organisms which seem to be associated with more severe diseases (cirrhosis, hepatocellular carcinoma) versus hepatitis alone in HCV-positive patients (467). The low discriminant power of 16S rRNA gene sequencing for species identification in the genus Helicobacter limits a definitive incrimination of H. pylori (424).

Attempts to detect enterohepatic helicobacters from the intestinal mucosa of patients with inflammatory bowel disease have also allowed the detection of H. pylori, probably as a bystander rather than a pathogen. Indeed, H. pylori crosses the intestine to be shed in feces but as a nonviable organism (422, 600).

The problem of specificity of PCR becomes fundamental in specimens from sites other than the stomach. At least two PCRs based on different target genes must be positive to conclude an H. pylori infection, as indicated before (76, 467).

Gastric PCR for other helicobacters.The tightly spiral H. heilmannii organisns are comprised, in fact, of 2 types: H. heilmannii I corresponds to a bacterium from the pig (58) and is also known as “Candidatus Helicobacter suis”; H. heilmannii II corresponds to 3 species, H. bizzozeronii, Helicobacter felis, and Helicobacter salomonis, usually present in cats and dogs. A 16S rRNA gene-based PCR was developed to detect these three species as a group. A small target DNA fragment (78 bp) enabled the study of formalin-fixed paraffin-embedded material (98). A multiplex PCR based on the tRNA intergenic spacers and on the urease gene allowed a discrimination of the three species and, combined with a 16S rRNA gene PCR for “Candidatus Helicobacter suis,” covers all of the H. heilmannii types (25, 99).

A novel PCR targeting 16S rRNA gene has been designed to detect H. heilmannii-like organisms which are also pathogenic and could remain undetected by microscopic observation. Screening of 131 gastric biopsy specimens from dyspeptic patients with this assay revealed a prevalence of 2.3% in Southeast England. It was then combined with an H. pylori vacA PCR in a multiplex PCR assay allowing the detection of both pathogens (70).

The use of genus-specific primers based on the 16S rRNA gene (positive) in addition to glmM primers (negative) allowed the detection of gastric infection by Helicobacter species other than H. pylori. The 2 cases detected out of 126 were identified as Helicobacter cinaedi, an enterohepatic helicobacter, after cloning and sequencing (439).

Historical aspects.The first description of a UBT to detect gastric urease activity was published by Kornberg et al. in 1954 on a cat. After an intravenous injection of [¹⁴C]urea, they were able to measure the ¹⁴CO₂ in the cat's breath. When antibiotics were administered before [¹⁴C]urea, the amount of ¹⁴CO₂ decreased. This was the first demonstration of the bacterial origin of a gastric urease (270), now considered to be due to H. heilmannii and H. felis.

The same principle was applied in 1987 by Graham et al. using urea labeled with ¹³C, a nonradioactive isotope, to detect H. pylori in humans (186). Marshall and Surveyor then described the [¹⁴C]UBT in 1988 for the same purpose (345).

Principle.A solution of labeled urea ingested by the patient is rapidly hydrolyzed by H. pylori urease if this organism is present in the stomach; the labeled CO₂ is absorbed by the blood and exhaled in expired air. If the patient is not infected, no labeled CO₂ is produced and most of the isotope is eliminated in urine without modification.

When [¹³C]urea is used, a specimen collection is performed before and 30 min after the ingestion. The ¹³C/¹²C ratio is measured in both specimens, and the result is expressed as the difference between the two measurements. The need for a baseline value is due to the various amounts of ¹³C present in breath according to one's diet. The result is expressed as delta over baseline (DOB) per 1,000.

When [¹⁴C]urea is used, specimen collection occurs only 20 min after ingestion.

Factors influencing the result and standardization of the [¹³C]UBT.As with the other diagnostic tests, UBT results are dependent on the patient, the bacteria, and the test itself.

Some of these factors have been standardized. To maximize the contact between the labeled urea solution and the bacteria, it is recommended that gastric emptying be delayed by ingestion of a test meal after fasting. The need to fast has been challenged: a 98% agreement was observed comparing fasting and nonfasting protocols (380). However, the nonfasting protocol leads to DOB values which increase the risk of negative results when the DOB is close to the cutoff, even if it is infrequent (128). Eliminating the test meal completely has been advocated (577), but this also leads to lower DOB values and the possibility of false-negative results.

Originally, the test meal consisted of lipids (Ensure; Abbott Laboratories, Abbott Park, IL) to delay gastric emptying (18). This meal was even included in a European standard protocol (320). Later, it became clear that citric acid was a better test meal and was much more practical (113). The accuracy is the same when [¹³C]urea diluted in the citric acid solution is administered (299), and this is now the most highly recommended method. Indeed, citric acid is able to sensitize the test not only by delaying gastric emptying but by a direct effect on intragastric urease activity, possibly on UreI, a proton-gated urea channel, making urea more accessible to the intrabacterial urease (499). Similar results have been found in another study where urease activity increased sixfold between pH 7 and pH 3 (431). Whether the enhanced H. pylori urease activity in vivo is strictly pH dependent or not is still a matter of debate. Indeed, ascorbic acid does not lead to the same result as citric acid (3), which confirms that fruit juices like orange juice are not as good (300).

Another attempt to simplify the procedure concerned the omission of basal samples. However, because the basal DOB is diet dependent, this omission represents a loss in accuracy, especially posttreatment (168, 262).

The dose of labeled urea has also been standardized. It is important that the gastric urea concentration exceed the urease K_max. With 100 mg of [¹³C]urea, this concentration is 10 times higher than the K_max and, therefore, is completely saturated. The dose has since been decreased to 75 mg, which is now the standard dose for adults. There are attempts to decrease this dose by administering a [¹³C]urea tablet (199, 415) or [¹³C]urea capsule (32, 440). This approach allows an earlier collection of the breath specimens (20 min or less).

With regard to breath collection, the pooled sample collection, which consisted of collecting breath at 5-min intervals for 30 min in the same bag, was quickly abandoned for the one-point sample collection after 30 min (262, 322). The goal is to minimize urea hydrolysis which could occur by contact with oral bacteria and is observed after a few minutes. Nevertheless, the recommendation of brushing one's teeth before performing the test does not seem important. During these 30 min, the amount of ¹³CO₂ increases progressively. The patient is advised not to smoke, eat, or drink and to remain seated quietly. It may be good to turn toward one side and then the other to increase the contact between the bacteria and the urea when not many bacteria are present.

There are several methods to measure the ¹³C/¹²C ratio. The most accurate is isotope ratio mass spectrometry, which requires only a small amount of breath to obtain a result (usually collected by blowing in a tube with a straw) (42). Another method is isotope-selective nondispersive infrared spectrometry (249, 266). While the measurement is less precise, it has no consequence on the result, and the main limitation is the need to collect a large amount of breath air (500 ml) by blowing into a special balloon. Consequently, this measurement must be performed locally rather than being sent by mail. A laser-assisted ratio analysis (399) was also designed but was never developed commercially (62, 485, 543). When the reproducibility of the UBT performed locally or in a central facility was studied, no difference was noted (210). Comparison of the specimen collection made by the patient under medical supervision or at home resulted in a 5% decreased accuracy in the latter (517).

Interestingly, an apparatus (Breath ID; Oridion, Israel) has been developed for online follow-up of the ¹³C/¹²C ratio, which usually allows the detection of H. pylori-positive patients within 20 min (232). The expression of the ratio is recorded in parts per 1,000 using an international standard of calcium carbonate named pee dee belemnite as the reference (88). Basic breath air values are lower than the standard, explaining the negative values obtained. These values increase after [¹³C]urea ingestion in H. pylori-positive patients. By convention, the results are expressed as an excess of ¹³C excretion and not as a percentage of the administered dose or urease activity per minute, which facilitates the comparison between laboratories.

The cutoff used to determine H. pylori positivity is still a matter of debate. The standard method of taking the mean plus 3 standard deviations of the values obtained in H. pylori-negative subjects (as determined by invasive tests) gives a cutoff of 4.9 per 1,000. When a comparative gold standard is available, a good method to use is the receiver operating characteristic curve (373), which leads to a cutoff of 3 per 1,000. Other methods have been used, such as the cluster analysis, and a cutoff of 3 per 1,000 was determined (374). A gray zone to separate positive and negative results is not usually proposed but could be a good solution. In fact, there are a limited number of cases falling into this category. Some authors suggest using a lower cutoff for posttreatment follow-up because the negative values seem more homogeneous (3.5 per 1,000).

Given this high standardization, the lack of impact of transport conditions and the independence of the result from human judgment, the UBT is a highly accurate and reproducible test. Among the noninvasive indirect tests, UBT has the best sensitivity, approximately 95%, i.e., in the range of the best invasive tests. Only a few cases with a limited number of bacteria present are missed (173). Specificity is also excellent. However, other urease-positive bacteria present in the stomach may also give a positive result. One example is H. heilmannii, but in this rare case, an eradication treatment is also worth prescribing. Enteric bacteria, e.g., Proteus sp., can theoretically be present in the stomach in the case of hypochlorhydria and bacterial overgrowth, but they have not been isolated from adults in developed countries. The situation in children, especially those younger than 2 years and from developing countries, needs further investigation (118). False-negative results may also occur in the case of bleeding, i.e., if there is a bleeding ulcer or even after taking gastric biopsy specimens. In a study where a high concentration of citric acid (4 g) was given, false negatives were eliminated when omeprazole or pantoprazole were used but not with lansoprazole or esomeprazole (307). A limitation of UBT is the need for an expensive apparatus, especially the gas chromatography isotope ratio mass spectrometer, but numerous specimens received by mail can be analyzed simultaneously with one apparatus.

Quantitative UBT.Numerous studies tried to correlate UBT values with the quantity of H. pylori measured by an invasive method, culture, or histology. A positive correlation was always found, with correlation coefficients ranging from 0.56 to 0.74 (66, 97, 121, 249, 495, 512, 558), except in one study (0.17) (24). High UBT values have also been used as predictors of severe gastritis (350) and low eradication (285). However, in another study, the mean delta values of UBT were not significantly different between patients who later showed eradication success or failure (171).

Other methods not using breath analysis.Measuring labeled bicarbonate in blood instead of measuring ¹³CO₂ in breath was first proposed in 1993 by Moulton-Barrett et al. (397). In this assay, the ¹³C/¹²C ratio is determined in serum by measuring before and 1 h after ingestion of labeled urea. The sensitivity and specificity were lower than that of UBT in one study (259) and equivalent in two other studies, including a United States multicenter trial (69, 89).

The possibility of using another isotope, [¹⁵N]urea, and testing ¹⁵NH₄ in urine was also explored (275, 582), but it does not offer any particular advantage over ¹³C.

Particularities of the [¹⁴C]UBT.Most of the characteristics of the [¹⁴C]UBT are similar to those of the [¹³C]UBT. However, for [¹⁴C]UBT, the volume of expired CO₂ is critical for the measurement of ¹⁴CO₂. The expired air is first dehydrated by passing it through a solution of CaCl₂, and then it is trapped in an alcohol solution of hyamine. A color change indicates that a sufficient volume of CO₂ has been collected. The results are adjusted to the endogenous production of CO₂, which depends on the patient's height and weight. An alternative way to collect ¹⁴CO₂ is to use the Heliprobe breath card and analyzer (Noster System AB, Stockholm, Sweden) (215).

¹⁴C is a radioactive isotope with a half-life of 5.7 years. According to Munster et al. (398), approximately 90% of the ¹⁴C from a UBT is excreted as CO₂ in expired air or as urea in urine.

Originally, the dose administered was 10 μCi (345), but because of the potential risk, attempts were made to decrease this dose first to 5 μCi (344, 353) and then to 1 μCi (= 37 KBq) (8, 32, 344, 445, 459). To avoid oral hydrolysis, these authors proposed to administer this amount in a capsule (200, 215, 344). The ingestion of citric acid is also used to sensitize the test (462) instead of the traditional test meal.

The ¹⁴C measurement is performed on a sample collected after 20 min with a scintillation counter, an apparatus widely available in many settings. The cutoff value to define a positive sample is 2% of the administered dose. The reproducibility of [¹⁴C]UBT within 1 week is excellent (508).

Due to the radioactive nature of the isotope, this test cannot be used on children and pregnant women and is forbidden in some countries, but it is a cheap and simple method. A variant of the test where the radioactivity is measured in urine has also been proposed (436).

Culture from stools.As everything present in the stomach will be found in the stools, there is no doubt that H. pylori is eliminated via this route. However, because the bacterium is susceptible to biliary salts, especially deoxycholic acid (204), and to vital competition from other bacteria, it may take the form of a nonculturable bacterium.

Indeed, the ability to culture H. pylori from feces may depend on the transit time. The first successful culture was carried out by Thomas et al. using feces from malnourished children from developing countries with a short transit time (519). By inducing diarrhea with sodium phosphate in H. pylori-positive adult volunteers, Parsonnet et al. were able to grow the bacterium in only 50% of the samples (434). Dore et al. also reported the isolation of H. pylori from three H. pylori-positive patients among 12 explored, by adding cholestyramin, a bile-sequestrating agent, before plating on the culture medium (115). These poor results, obtained despite the special conditions provided, cast doubt on the validity of the results of others who claimed the recovery of H. pylori from adult patients (256). However, more recently, Liang and Redlinger reported successful isolation of H. pylori in feces of four subjects, highlighting the importance of the antibiotic concentration of the media and the ability to recognize H. pylori colonies after 5 to 7 days as well as the timely collection of the specimens (311).

Indeed, our own experience and that of Namavar et al. (401) indicate that bacteria similar to H. pylori may be present in feces, and therefore, it is necessary to carry out a molecular identification before drawing a conclusion. Under no circumstances can H. pylori culture from stools be used as a routine diagnostic method.

PCR on stool samples.In contrast to gastric biopsy specimens, where H. pylori is usually the only living bacterium, stool samples are very complex material for which the pre-PCR steps are quite important.

The first attempts to detect H. pylori DNA in stools using PCR were made by Mapstone et al. in 1993 (337). These results were not reproduced by Van Zwet et al. (553). It became clear that small amounts of target DNA and the presence of substances inhibiting the polymerase reaction were responsible for these failures. Monteiro et al. explored the nature of these inhibitors. They were identified as essentially complex polysaccharides of vegetable origin (384).

Very long, complex, and fastidious methods have been proposed to eliminate these compounds (332). A 48-h vegetable-free diet decreases their amount (388). This observation is in line with superior results obtained with infants who did not have a diversified diet. However, numerous attempts have been made to obtain pure DNA. An immunomagnetic separation (magnetic beads coated with H. pylori antibodies) was proposed and was found to be experimentally suitable for young cultures but not for older cultures, probably due to the different antigenicity of the coccoidal forms (127). Polysaccharide concentrations were also significantly reduced using macroporous polypropylene filters (61) and two commercial kits: QIAamp tissue kit and Xtraction DNA extraction kit (QIAGEN) (14). Monteiro et al. also developed a PCR combining H. pylori extraction by immunomagnetic separation and an agarose-embedded DNA preparation (387). This method, which was developed to recover intact DNA from pulsed-field gel electrophoresis, provides clean, high quality template DNA. It requires a small number of steps, reducing the risk of DNA loss or contamination and, furthermore, avoids the use of nested PCR and its risk of contamination (131). The sensitivity of this method with ureA primers was nevertheless only 81%, with a specificity of 100%. The same proportion of positive samples was obtained by gene capture with a biotinylated oligonucleotide probe targeting the H. pylori 16S rRNA gene (326). After incubation, it is harvested on paramagnetic polystyrene beads coated with streptavidin. The recent availability of a DNA extraction kit devoted solely to stools (QIAamp DNA stool kit; QIAGEN) seems to be progress. For detailed information on extraction protocols, see the review by Kabir (241). However, the use of a very specific H. pylori target is imperative because other helicobacters may be present in stools, e.g., H. cinaedi, Helicobacter pullorum, and H. canis.

PCR also allows the detection of pathogenicity factors. Sicinschi et al. used an extraction protocol delivering sufficient amounts of DNA, radiolabeled primers, and 80 amplification cycles for the detection of cagA and vacA genotyping (502). They also compared their genotyping procedure in stool samples to that of biopsy specimens (unfixed and paraffin embedded) and claimed that the stool protocol offered more-accurate results. In another study, cagA was also detected directly in stool specimens (476).

Of even greater interest is the ability to test for antimicrobial susceptibility. A simple DNA extraction method, followed by a seminested PCR and a restriction fragment length polymophism (RFLP) analysis, was used in Italy (146). It is, however, surprising that the two clarithromycin-resistant cases found did not exhibit the usual 23S rRNA gene mutations but had T2717C mutations.

The use of biprobe real-time PCR followed by melting curve analysis is indeed a more promising approach, as described by Schabereiter-Gurtner et al. They obtained a complete agreement for diagnosis, but for specimens with a mixed population of susceptible and resistant strains, only the susceptible type was detected (486). A summary of the publications concerning PCR in stools is presented in Table 3.

The development of a commercial test to detect H. pylori and its eventual clarithromycin resistance in stools by PCR will undoubtedly provide an accurate and convenient diagnostic method (457). It is, however, too early to provide figures for sensitivity and specificity.

Detection of H. pylori antigen in stools.The first report of successful detection of H. pylori antigens in stools was made in 1997. Kozak et al. reported an enzyme-linked immunosorbent assay (ELISA) performed on stools, using polyclonal anti-H. pylori antibodies coated on microwells to capture H. pylori antigen and peroxidase-conjugated polyclonal antibodies to detect the immune complex (272). This test was named H. pylori stool antigen test (HpSA) (Meridian Diagnostics, Cincinnati, OH). Its evaluation on stools obtained from 100 adult dyspeptic patients indicated 88.8% sensitivity and 94.5% specificity (332). A large multicenter study was then carried out in Europe in 11 endoscopy suites. Five hundred patients were included, and sensitivity and specificity were 94.1% and 91.8%, respectively, versus 95.3% and 97.7% for UBT (537). Numerous studies were then performed, and two systematic reviews confirmed the value of HpSA as a pretreatment diagnostic tool (174, 539). The interest of this test for assessing H. pylori eradication was also confirmed in 162 patients issued from the multicenter study (538) as well as in others summarized in the systematic reviews (172, 539).

However, recent studies may not be as convincing. Perri et al. compared UBT and HpSA on 458 patients at 4 to 6 weeks posttherapy. They found discrepant or indeterminate results in 8% of the cases. The patients with unclear results underwent endoscopy with biopsy sampling for several invasive tests. On the basis of these further diagnostic methods, the authors concluded that HpSA was less accurate than the UBT (444). In the studies of Bilardi et al. and Matsuda et al., HpSA was also shown to be inferior to UBT, especially with regard to specificity (33, 346).

A new generation of antigen stool tests emerged with the Femtolab (Connex), also known as Amplified-IDEA-HpStAR (Dako, Glostrup, Denmark). This new kit was compared to HpSA on stool specimens from 48 pediatric patients (330) and 53 adults (4) and exhibited a comparable accuracy. This kit uses monoclonal antibodies instead of polyclonal antibodies, which ensure a constant quality of reagents and, therefore, an excellent reproducibility rather than intertest variations. The optical densities obtained are also significantly higher than those of HpSA and are far from the cutoff value (330).

Another progress is the use of monoclonal antibodies in an office-based stool test, ImmunoCard STAT HpSA (Meridian) (Table 4), but its accuracy is lower than the standard laboratory ELISA (174).

However, the stool antigen tests have some limitations. The impact of bowel movements has not been studied in detail. In principle, a short transit time should favor elimination of unaltered antigens (unless there is an important dilution factor), while constipation should lead to degradation of the antigens. Preservation of the specimen has not been considered important but should be studied. Indeed, in experimentally spiked specimens, the lowest concentration could not be detected after only a 24-h delay (388). Treatment with the mucolytic agent N-acetylcysteine has been shown to decrease both the test's sensitivity and specificity (104). As previously stated, determining a cutoff to define a positive sample is crucial, especially for HpSA, and adjustments may have to be made. While these tests are quite specific, it is possible that rare Helicobacter species present in stools (enteropathic helicobacters) may also be detected.

Serodiagnosis. H. pylori infection almost constantly induces a specific systemic immune response which may reflect the antibodies produced at the gastric mucosal level, while only 2% of patients fail to seroconvert (276). This response was used for the diagnosis of this infection immediately following the discovery of H. pylori (236, 343). The immune response varies according to the antigens present in the infecting strains and to the host.

(i) Techniques for serodiagnosis.A complement fixation test was the first to be used in 1984 with an 85% sensitivity (236). A commercial complement fixation kit had an insufficient sensitivity (71 to 81%), while the specificity exceeded 90% (183).

A passive hemagglutination test using an H. pylori sonicate appeared to have a better sensitivity (343). A latex agglutination was also proposed (220, 568), but the essential technique used for more than 20 years is the standard ELISA and its derivatives, such as rapid immunoenzymatic assays and immunoblotting.

(ii) Antigens used for serodiagnosis.The performance of an ELISA is largely dependent on the nature of the antigens used. They must be highly immunogenic, common to most H. pylori strains and absent from other bacteria, and easy to prepare and purify, and they must bind to microwells and be stable upon storage.

The first antigens to be used were whole-cell sonicates, due to the fact that H. pylori is quite a unique bacterium in humans. They gave results with high backgrounds and limited sensitivity and specificity.

A partial purification to obtain surface antigens can be achieved by glycine acid extraction. ELISAs with these complex antigens had a sensitivity of 95% with a specificity of 90% according to Newell and Rathbone (404).

The presence of cross-reacting antigens, especially flagellar proteins, between H. pylori and campylobacters led to the development of more purified antigen preparations. Attempts were made with purified urease, for example, but the results were not good enough. For this reason, most of the commercially available ELISAs use a mixture of specific antigens for which the exact composition is patented.

A comparison of several kits testing the same sera in several laboratories in Europe indicated that Pyloriset EIA-G (Orion Diagnostics, Espoo, Finland) had the best accuracy (139). Serological tests to detect CagA antibodies have also been designed based on recombinant proteins. Interestingly, when they were used on sera of gastric adenocarcinoma patients, more cases were detected than with the standard H. pylori ELISA, probably because CagA is a highly immunogenic molecule, whose corresponding antibodies persist for longer periods (120).

(iii) Antibodies detected. H. pylori infection is a chronic condition and immunoglobulin G (IgG) (subclasses 1 and 4) is the predominant immunoglobulin class, even in children (376). IgG are present at the mucosal level and detected in virtually all blood samples. IgM are rarely observed, merely because acute H. pylori infections are seldom available for study. In the experimental infection carried out by Morris et al., an initial IgM response was observed (395). IgA are also elevated in the majority of infected cases but not in all. Therefore, as the relevance of IgM and IgA is limited, commercial kits are primarily designed to detect IgG. It is well known that IgG levels begin to decrease after eradication of H. pylori, but this fall in titer may take months or years before returning to baseline values (271).

(iv) Factors affecting the ELISA result.Since H. pylori infection is a mucosal infection, the antibody response may be modest, so it is important to have the best antigen and an appropriate cutoff value, but the sensitivity of ELISA serology can be excellent.

As previously stated, H. pylori is a very diverse bacterial species, with different clusters found, especially between the Far East and the West (218, 405, 461, 550). Diverse antibody profiles can be seen when different strains are studied by immunoblotting (351). Some authors have emphasized the need to use local strains as the source of antigen rather than foreign strains to obtain the best test performance. This may be especially true in Asia (223, 412). However, others insist on the importance of common antigens present between the strains (9) and recommend pooling antigens from strains representing the different groups (225, 340).

It has often been stated that the cutoff should be validated locally on a pool of reference sera obtained locally rather than using the cutoff values proposed by the manufacturer (234, 318). Indeed, this statement should be revised because some commercial kits propose cutoffs which can be used universally, at least in the West.

False-negative results may occur following a new infection before the antibody level is sufficiently elevated. The specificity of serology is satisfactory (90%). In developed countries, the risk of cross-reactions with other bacteria is limited. In developing countries, the specificity of the test may be altered by the presence of concomitant infections, especially with campylobacters or other related bacteria; furthermore, the host immune response may be lower due to malnutrition. Such cases definitely justify the reevaluation of the tests, and for this purpose, making an receiver operating characteristic curve is the best alternative (527).

The very slow decrease in antibodies after eradication (25% in titer within 6 months or more) is also a cause of a false-positive result. For this reason, the tests detecting active infection are preferable.

Indeed, the negative opinion on the specificity of serology is mainly based on correlations made with poorly performed gold standard methods. Recent data have shown that serology is the best method in difficult situations where bacterial density may be low due to gastric atrophy or due to previous treatment with PPI or antibiotics, as is now frequently the case (334).

(v) Development of point-of-care tests.In contrast to laboratory tests, these tests can be used by the doctor in his/her office. They are also called “office tests,” “near-patient tests,” or “doctor tests”. Their advantage is to be simple, cheap, and able to deliver a quick result. They are essentially based on the diffusion of antibodies from a drop of serum or whole blood obtained by finger puncture through a membrane and an immunoenzymatic reaction. The first test was proposed in 1994 (Quickvue, Quidel, San Diego, CA) with a very promising performance (sensitivity, 92%; specificity, 88%).

Unfortunately, more than 30 studies were later performed with this test as well as others (Helisal, FlexSure, Hp Check, Stat Simple, Accustat, BM test, and Pyloriset Screen, etc.), with sensitivity and specificity values rarely reaching 90% and a mean of 82%. Furthermore, in most of these studies, the tests were not performed under the conditions which should have been applied, i.e., general practitioners carrying out a few tests per month, rather than dedicated laboratory personnel carrying out all of the tests at once, with the latter favoring better performances of the tests. When we compared one of these tests (Pyloriset Screen; Orion, Espoo, Finland) performed by general practitioners to a laboratory ELISA, we nevertheless found a 95% accuracy, which is comparable to the evaluation made in comparison to invasive tests (417).

Another problem is the difficulty in reading the results in 10% of the cases, and the patients do not express a strong preference for a finger puncture versus venipuncture (510).

The final consequence is that the point-of-care tests have not been recommended in the Consensus Conferences (334).

(vi) Immunoblot analysis and detection of CagA antibodies.Immunoblot analysis has proved to be very attractive for the diagnosis of many infectious diseases. While its sensitivity may be equal or inferior to standard ELISA, the resulting antibody profile makes it more specific.

The antigens separated by electrophoresis can correspond to the whole bacterium (351) or to special extracts like cell surface proteins (407). In spite of these potential advantages and the widespread availability of blotting equipment in clinical laboratories, Lepper et al. (305) reported that immunoblot for H. pylori antibodies is not common practice and that this was possibly due to discrepant recommendations for interpretation. Indeed, most of the publications concerning this subject are merely descriptive (9, 90, 138, 195, 261, 298, 560), and only four proposed interpretation criteria (219, 407, 527, 559). Lepper et al. compared these criteria on their sera collection. Depending on the aim of performing immunoblotting, the chosen criteria were not always the same. Indeed, the immunoblot is most likely used as a second-step technique to identify false-positive cases detected by ELISA, in which case, the criteria proposed by Nillson et al. (407) and Trautmann et al. (527) leading to 100% specificity must be used. In contrast, when immunoblotting is used for preendoscopy screening, the criteria of Lepper et al., which emphasize sensitivity, are more suitable (305).

A commercial immunoblot test is now available, which is an important advance toward standardization, Helicoblot 2.1 (Genelabs, Singapore) (305, 383, 432).

Interestingly, attempts have been made to correlate the presence of an immune response to a specific antigen with a specific disease. The main antigen in this respect is CagA. The importance of the immune response to CagA as a marker of more severe diseases, i.e., peptic ulcer disease (83, 87), gastric adenocarcinoma (36), and atrophy (29), was indeed discovered by immunoblotting.

However, because the presence of CagA antibodies is not a good predictor of these diseases on an individual basis, its detection has not been included in their management (334). Nevertheless, ELISAs which are specific for CagA antibodies are now on the market.

Serological response to other antigens have tentatively been associated with particular diseases (22, 277, 557).

Detection of H. pylori antibodies in urine.Specific H. pylori IgG antibodies are eliminated in urine but at very low concentrations. Alemohammad et al. presented the first data in 1993 using both ELISA and immunoblotting (6) with good accuracy.

Commercial tests have been developed in Japan: a standard ELISA (Urinelisa; Otsuka, Tokyo, Japan) and a rapid point-of-care test based on immunochromatography (Rapirun; Otsuka Diagnostic, Tokyo, Japan) (252). They were evaluated in 14 pretreatment studies, 11 on adults and 3 on children, including 10 in Japan. Sensitivity appeared satisfactory, with no difference between Urinelisa and Rapirun nor between children and adults, but the specificity was not as good (Table 5). Furthermore, most of the studies were performed in Asia, so other evaluations are warranted in the West where H. pylori strains appear to be different.

There is a good correlation between the nature of antibodies present in urine and in blood when tested by immunoblotting (248) but not between the antibody titers (251), which are dependent on the urine concentration. The accuracy of the test is not dependent on the pH nor the presence of bacteria in urine. In contrast, a high total IgG may induce false-positive results and a low total IgG may induce false-negative results (251, 496); therefore, patients with urine analysis abnormalities should be excluded. Another critical point is the need to perform the analysis on fresh urine samples and not on frozen samples. Freezing and thawing specimens may lead to protein precipitation, and the test's sensitivity was indeed lower on frozen urine (2, 6). However, once the urine is collected with a preservative, it can be sent by mail and analyzed a few days later. An apparent low specificity was noted in the study of Kato et al. when they compared Urinelisa to biopsy-based tests which was not confirmed because 10 of the 12 false positives were indeed found to be positive by immunoblotting (248).

Detection of H. pylori IgG in urine is definitely an easy, rapid, inexpensive, and absolutely noninvasive way to diagnose H. pylori infection, and the availability of a point-of-care test exhibiting the same performances is definitely attractive. However, more studies are needed before recommending them in the West.

Detection of H. pylori antibodies in saliva.Salivary antibodies are secreted during the immune response to infectious agents. Surprisingly, detection of specific IgA to H. pylori could not distinguish between infected or noninfected individuals (324, 574), but detection of IgG could (324). Following the princeps study by Patel et al. (435), 15 studies have been published (Table 6), six of them using the commercial kit Helisal (Cortecs Diagnostics, Deeside, United Kingdom) for salivary testing and the Oralfluid ELISA (Enteric Products, Stony Brook, NY). Four of the studies were conducted with children. Specimen collection is easy to perform by spitting saliva into a tube. However, another method has been proposed using a special device (OraSure; Epitope, Beaverton, OR), a swab to be rubbed on the gums to obtain a gingival transudate, enriched in IgG (333). Sensitivities and specificities obtained in these studies tend to be low, rarely reaching 90%.

The best results seemingly occur when the serum titer is high, as in patients with duodenal ulcer (156) or in children older than 5 years (165), despite a good correlation found in one of the first studies between titers obtained in sera and saliva (324). The specificity of the test can be improved by immunoblotting (179, 184, 341).

Sensitivity and specificity.The characteristics dependent on the test are sensitivity and specificity and, when combined, determine the accuracy. In contrast, the positive and negative predictive values, which are important for clinical practice, are very dependent on the prevalence of the infection in the community or on the group of patients considered, as has been highlighted in a number of articles, especially one by Shiotani and Graham for H. pylori (498). Another approach which is now more commonly used to evaluate test results is the positive or negative likelihood ratio. The likelihood ratio is the likelihood that a given test result would be expected in a patient with the target disorder compared to the likelihood that the same result would be expected in a patient without the target disorder.

Numerous studies have focused on evaluating the sensitivity, specificity, accuracy, and positive and negative predictive values of the tests used to detect H. pylori. They were all faced with the problem of the appropriate gold standard, since none of the tests fulfill this role. To evaluate noninvasive tests, invasive tests are usually considered the reference. As a general rule, H. pylori cases are deemed positive when (i) culture is positive or (ii) in the event of a negative culture, both histology and urease tests are positive. This appears to be a good compromise. In a recent study, culture was positive in 88% of the cases and histology plus urease test were positive in 12% (362). Sensitivity can also vary if the test is performed prior to or 4 to 6 weeks after an H. pylori eradication treatment because, in the latter instance, even if the bacteria are still present, there is most likely a much smaller bacterial load.

There are indeed few studies where all of the methods have been performed and, furthermore, under optimal conditions. The results of two studies, one from The Netherlands comparing essentially invasive tests, UBT, and serology (516), and one from our laboratory comparing noninvasive tests to a gold standard based on invasive tests (386) are presented in Table 7.

Availability of the tests.The availability of the different tests described varies considerably from country to country. Histology is probably the most widely available method and, because of the fixative, avoids the problem of transport. It is indeed difficult to find a laboratory performing H. pylori culture, and this is further complicated by the strict transport conditions required in sending samples. While urease tests are easy to perform, their use may be limited due to a lack of medical reimbursement in some countries. Molecular methods are only carried out in specialized centers; however, the diffusion of a commercial real-time PCR assay allowing H. pylori stool detection should contribute to their increased utilization. UBT can now be performed in most countries worldwide, and the easy transport of breath specimens allows a wider accessibility to this assay. The stool antigen ELISAs could theoretically be performed in all laboratories, but the cost of a kit designed for numerous assays may discourage some who have a limited use to buy it. To the contrary, serology is widely available but the detection of salivary and urinary antibodies is not.

Rapidity in obtaining results.The delay between sampling and receiving the results varies between a few minutes and 2 weeks. If a microscope is available in the endoscopy suite, the direct examination of a slide is the quickest method. A stain can be performed in a few minutes. All of the immunoenzymatic tests (point-of-care tests) to detect antibodies in serum and urine or antigens in stools can be performed within 10 to 15 min in the endoscopy suite. The strip-based urease test can be read within the hour. The UBT requires 30 min to collect the specimens and a few minutes to measure the ¹³C/¹²C ratio. If an apparatus is available in the vicinity of the patient's room, it also allows a 1-h diagnosis, but this is rarely the case; normally, the specimen must be sent to a central laboratory, often outside the city, which delays the result by at least 2 days. All laboratory ELISAs for antibody detection or antigen detection can be performed within 2 h. However, for organizational reasons, series of tests are usually performed, causing a certain delay. The same is true for molecular methods, although a real-time PCR protocol can be carried out in 2 h, including both the extraction procedure and detection of mutations associated with antimicrobial resistance. Histology can be performed within a day. Only culture requires from 3 to 10 days depending on the growth. Further susceptibility testing will increase the delay by 3 to 4 days.

Possibility of quantitative tests.None of the tests currently available determine the number of bacteria present in the stomach; only an estimation can be obtained.

Culture, histology, PCR, and UBT provide an approach to quantification, but in routine practice, a semiquantitative evaluation with either histology or UBT is the method employed.

Possibility to detect pathogenic properties.It clearly appears that some H. pylori strains are more pathogenic than others, and although routine detection of pathogenic properties associated with severe disease evolution is not yet recommended (334), tests with this potential are definitely of interest for epidemiological studies.

Culture is obviously the best method in this respect because it allows a complete screening of pathogenic properties either by a phenotypic method (VacA) or molecular methods (cag PAI genes, vacA alleles, babA2, dupA, iceA, and sabA, etc.) (296).

Molecular methods can be applied directly to biopsy specimens with the same sensitivity as culture, and the advantage is to get an accurate picture of the strains present, which is not the case after culture, since some strains may grow better than others (67, 194, 283, 433).

The development of FISH has allowed the detection of pathogenic properties on histological preparations. However, probes are not commercially available, in contrast to kits for susceptibility testing. Real-time PCR can also be performed on histological material.

Among the noninvasive tests, serology is the only method which allows the detection of H. pylori pathogenic properties, and it applies essentially to the CagA antigen, which is highly immunogenic. CagA serology is even more sensitive than H. pylori serology because the antibodies may remain for longer periods after H. pylori eradication. This detection can be performed by ELISA or immunoblotting, as previously described.

Globality of the test.Invasive tests do not provide a global picture of H. pylori in the stomach because they only explore a small portion of the gastric mucosal surface (85). They are subject to sampling error (27). In contrast, noninvasive tests provide a global picture of H. pylori in the stomach, and this may explain the discrepancies between techniques (245, 265). Indeed, the presence of atrophy and intestinal metaplasia corresponds to an inhospitable environment for H. pylori. H. pylori may only subsist in small areas of the stomach. Serology is the best detection method in such cases, with the bacterial load sometimes being inferior to the cutoff value of UBT and stool antigen tests.

Obtaining gastric juice can be a means of obtaining a global result using direct tests. PCR is then the preferred method because the sensitivity of a bacterial culture is lower than that of gastric biopsy.

Cost of the test.The cost of each test is extremely variable from one country to another. It depends first on the cost of endoscopy, which is a prerequisite for invasive tests, and second on the availability of some apparati which may or may not be present in a given setting.

When endoscopy must be performed, urease test and smear examination are the cheapest tests. Histology and culture require a specific environment and manpower and therefore will always be costly. Molecular tests which are easier to automatize will undoubtedly see their cost decreased in the future.

Among the noninvasive tests, serology is usually the cheapest. UBT and stool antigen tests are often in the price range of histology and culture without the cost of an endoscopy.

Added value of certain tests.In addition to the characteristics previously mentioned, some tests may have a unique property of importance in the choice.

Endoscopy is necessary to obtain biopsy specimens for direct testing, but this examination may provide important information for the patient.

Histology allows an evaluation of the status of gastric mucosa, including the presence of atrophy, intestinal metaplasia, lymphoid follicles, dysplasia, and carcinoma. Observation of activity, which is a surrogate marker for H. pylori infection, and disappearance of the activity after eradication can be noted. H. heilmannii, which may eventually be present, can be visualized by this technique.

Culture allows antimicrobial susceptibility testing, which is a key issue in patient management (357). It offers the possibility of applying all typing methods to differentiate between recrudescence and reinfection when a relapse occurs.

Molecular methods performed directly on the biopsy specimens can also answer these questions. They include PCR-RFLP for typing and detection of 23S rRNA gene mutations and real-time PCR.

When performed 4 to 6 weeks after the end of the treatment, the UBT and stool antigen test can detect H. pylori eradication noninvasively, which is in the patient's interest.

Before treatment.The approach recommended in guidelines for dyspeptic patients less than 45 years old without alarming symptoms is the so-called “test and treat strategy,” i.e., testing the presence of H. pylori with a noninvasive test before prescribing an eradication treatment. The value of this approach has been proven in several European countries and may be due to a high proportion of patients with peptic ulcer among the dyspeptic patients treated (352). The cost effectiveness has been evaluated and is of value up to a certain threshold of H. pylori prevalence; it is also dependent on medical costs. Patient satisfaction has, however, been questioned (293). The low incidence of gastric carcinoma at a young age is indeed the reason for the relatively high age threshold; however, it may be increased or decreased depending upon the population and particular situation at hand.

The noninvasive test recommended as the first choice is UBT because of its high accuracy, convenience, and availability, and the second choice is the stool antigen test. In contrast to previous guidelines, the Maastricht III 2005 Consensus Report acknowledges that certain highly accurate serology kits can also be used (334). Indeed, serology is the only test for which PPI consumption has a limited impact, and patients taking PPI before consulting a doctor has become a common situation. Laboratory serology may also be the only accurate test for patients with atrophy or bleeding ulcers. Other antibody tests using urine or saliva and whole-blood doctor tests are not recommended.

For those older than 45 years or those with alarming symptoms, an endoscopy is mandatory and biopsy specimens must be taken. The value of the urease test is that the treatment can be given right away, since a positive result can be obtained within an hour. However, ideally another more sensitive test should be performed, either histology or culture. The choice between the two depends essentially on the prevalence of clarithromycin resistance in the area. In the case of high prevalence (>15 to 20%), it is mandatory to test the patient's susceptibility to this drug before prescribing it (334). Age may also be considered. Generally speaking, younger patients are more likely to have been exposed to this drug, and they are also more likely to experience treatment failure (45, 170) and, therefore, need culture and susceptibility testing. Conversely, older patients are more likely to have atrophy and other mucosal abnormalities, and therefore, histology is recommended. A good alternative to culture is real-time PCR, which allows both detection of the bacterium and clarithromycin susceptibility testing within 2 h.

Posteradication follow-up.Current first-choice 7-day treatment regimens have a limited efficacy, in the range of 70% eradication (101), which can be increased slightly but significantly by increasing the length of treatment to 14 days (147). With such a high failure rate, it is mandatory to have a follow-up 4 to 6 weeks or more after the end of the treatment.

For this purpose, a noninvasive test can be performed, and the UBT has been unanimously approved. If this test is not available, stool antigen test using monoclonal antibodies can be used. Serology cannot be used unless it is possible to compare the quantitative results obtained before and 6 months after the end of treatment. A decrease of at least 25% indicates eradication. This approach is not realistic in clinical practice because serum samples are not normally stored for future use. Under certain circumstances, e.g., gastric ulcer surveillance, endoscopy must be performed to detect an eventual malignancy, and in this case, histology should be performed.

Diagnosis for children.The same diagnostic methods used for adults can be used for children. However, H. pylori infection has certain particularities in children which have implications for diagnostic testing. H. pylori infection may be established slowly, so it is possible in rare instances to find the bacteria without traces of inflammation. At endoscopy, antral nodularity is common. In young children, the antigen stimulation is limited compared to that of adults. Bacterial overgrowth may be more frequent. In contrast to adults, preneoplastic lesions are exceptional (193), even in high risk areas (465). The infection is seldomly found in infants and young children, which limits the possibilities of test comparisons on large series.

Studies have been numerous over the last 10 years, and consensus conference reports were published in 1999 in Canada (494), in 2000 in the United States (180) and Europe (117), and again in 2005 in Canada (39). They all pointed out that, when faced with a child with recurrent abdominal pain or other symptoms compatible with H. pylori infection (e.g., iron deficiency anemia, growth retardation), endoscopy should be performed and biopsy specimens obtained for H. pylori diagnosis. The reason is the need to both detect the infection and identify the cause of the symptoms. This identification is primordial despite the fact that endoscopy requires deep sedation or general anesthesia, which increases the risk and cost of the procedure, and also the fact that the probable absence of a malignancy makes noninvasive tests attractive.

However, numerous articles have reported the use of noninvasive tests. UBT has been validated (28, 53, 242, 250, 253, 362, 481, 542) and is the most highly recommended noninvasive test for children. The protocol used in adults can be followed. The dose of labeled urea can possibly be decreased to 2 mg/kg (53) or to a total of 50 mg to make it simpler (28). The citric acid meal can be used if the taste is accepted by the children, otherwise orange juice is an alternative. In contrast, a fatty meal, for example, ice cream, has a negative effect on the test results (473). In very young children, a mask must be used to collect breath air (155). A cutoff of 3.5‰ is usually recommended (28, 53), except for children younger than 2 years. Indeed, a further evaluation of this age group needs to be performed because false positives have been reported (228). Bacterial overgrowth could explain these false-positive results (370). Because of the change in endogenous CO₂ production due to physical activity, stress, or food consumption, a correction in endogenous CO₂ production has been proposed (263) but is not commonly employed.

The stool antigen test (Premier Platinum HpSA) was evaluated for the first time in a multicenter study of children in 2000 (414). The accuracy was excellent with a specific cutoff. Numerous studies have been performed since then because this diagnostic approach is especially convenient for children from whom it is common to collect stools and in whom the tendency of a short transit time is favorable. The stool antigen test using polyclonal antibodies demonstrated variable results and, in general, a less favorable accuracy than the stool antigen test using monoclonal antibodies.

Serology in children younger than 10 years has encountered the problem of a lower antibody titer than in adults, and therefore, a specific cutoff value must be chosen (81, 205, 511). In addition, laboratory serology may not be well accepted by children because, although it is minimally invasive, the child may refuse the serum sampling. Serology has the reputation of a limited accuracy, but authors agree that a nice complement to serology is immunoblotting (377, 420, 460, 466). As in adults, point-of-care tests and salivary or urine tests are not suitable in children.

As an example of a comparative study on noninvasive tests, we report the results of a large multicenter European study performed in 18 centers in 15 countries. Three hundred sixteen children aged 2 to 17 years, including 133 H. pylori positive by invasive tests, had complete test results. UBT was determined to be the most accurate test, even in the 2- to 5-year-old age group, and DOB values were far from the cutoff value (Table 8). Serology ranked second in sensitivity with an excellent specificity, indicating the importance in choosing the kit. To the contrary, the polyclonal stool antigen test showed limited sensitivity, which could be slightly improved by modifying the cutoff value. Unfortunately, we were not able to test the monoclonal antibody-based stool test in this study. Urine antibody testing confirmed its poor accuracy (362).

The problem of transient infection in children from low-prevalence areas has emerged. Based on pepsinogen testing, Nurgalieva et al. showed that a single positive noninvasive test is most likely a false-positive result (411) and must be confirmed by another test.

Diagnosis for the elderly.The same techniques are also used for elderly patients. There are also some characteristics of elderly patients which can influence the diagnostic tests. A high proportion of the elderly population develops gastric atrophy and intestinal metaplasia, which can lead to a hostile environment for H. pylori, hence, fewer bacteria and possibly a negative result. Furthermore, elderly people take a number of drugs for chronic diseases and are more likely to be hospitalized. Among the medications prescribed are (i) PPI, which have a negative impact on H. pylori, and (ii) antibiotics, which can even eradicate the bacterium. Constipation is frequent and may jeopardize the detection of H. pylori antigens in stools. Severe cognitive impairment can alter the capacity to perform the UBT. Due to immune deficiencies, the antibody production may be low.

Endoscopy is frequently performed in elderly patients because of dyspepsia, given the high rate of peptic ulcer disease and gastric cancer in this population, as well as for nonspecific symptoms such as chronic anemia and anorexia. Histology is the preferred method to evaluate the status of the mucosa. It is common to find sequelae of chronic atrophic gastritis and no H. pylori for the reasons mentioned before. Both the corpus and the antrum must be explored, and sometimes, H. pylori is found only in the corpus. Sampling error is furthermore a limit for culture and urease tests, which also leads to an underestimation of H. pylori infection in elderly patients. A study showed that urease tests led to 50% false-negative results in patients older than 60 years (1).

Noninvasive tests have also been evaluated. The UBT is easy to perform, even in elderly patients, except in those with severe cognitive impairment. Pilotto et al. showed good accuracy of UBT in comparison to invasive tests (449). Others focused on possible false positives due to buccal contamination (poor dental hygiene) or bacterial overgrowth (hypochlorhydria) (68, 284). Antecedents of gastric surgery must be sought to adapt the protocol if necessary.

Obtaining stools for the stool antigen test is often more difficult, especially in cases of constipation. In addition, a prolonged period of H. pylori stay in the colon leads to antigen degradation and a decrease in sensitivity (388).

Finally, serology can be used. In the case of atrophy, Kokkola et al. showed that positive serological tests corresponded to active H. pylori infection (265). It is the only test which remains positive after antibiotic or PPI treatment, but in this case, the results may no longer correspond to an active infection. Underevaluation may also be possible in cases of proteinocaloric malnutrition, leading to humoral and cellular immunodepression (48). As for other populations, immunoblotting tends to be more accurate than serology.

Few comparative studies have been performed on the accuracy of different tests for the elderly. Pilotto et al. compared UBT and serology using invasive tests as the gold standard. They found a high accuracy for UBT, while the serology results showed both low sensitivity (74%) and low specificity (59%) (449). In a study on frail hospitalized elderly patients comparing UBT, serology, and HpSA to invasive tests, we did not obtain such good results. Surprisingly, only about half of the patients exhibited an H. pylori infection and, if stringent criteria were applied, the number dropped to only one-fourth. In several patients, only one test was positive, and this was especially true for UBT (480). These results may be explained by the type of population tested, which was considerably older (mean age, 85 years) than that of Pilotto et al. and comprised of hospitalized and frail subjects, half of whom were receiving antibiotics or PPI.

Another study on hospitalized, frail, elderly patients (230) also showed that only half of the patients were H. pylori positive. The stool antigen test (Premier platinum HpSA) compared to three tests (histology, serology, and UBT) had a sensitivity and a specificity of 76% and 93%, respectively, with the sensitivity being even higher for those taking PPI.

Diagnosis in the context of upper gastrointestinal bleeding.Recent data have shown that the number of bleeding ulcers has not decreased these last years despite the decrease in H. pylori infection. This may be due to a higher consumption of nonsteroidal anti-inflammatory drugs, including aspirin, and anticlotting agents. The diagnosis of H. pylori infection is important, but there are limited possibilities to take biopsy specimens, and diagnostic tests are considered less accurate in this context. Urease test, culture, and histology, as well as UBT, lack sensitivity, while the polyclonal stool antigen test has a lower specificity than usual conditions. Serology should remain the preferred method to be used in this context because it is not affected by the local environment.

Gisbert and Abraira recently performed a metaanalysis of the diagnostic accuracy of currently available tests. Sixteen studies were included, totaling 1,417 patients, but not all of the tests were used comparatively in all patients. The results confirmed the low sensitivity but high specificity of invasive tests (urease test, culture, histology), the low specificity of the stool antigen test, and the excellent accuracy of the UBT, but serology showed a surprisingly low specificity (Table 9) (167).

(i) Macrolides.Macrolides act by binding to ribosomes at the level of the peptidyl transferase loop of the 23S rRNA gene. H. pylori resistance is the consequence of point mutations at two nucleotide positions, 2142 (A2142G and A2142C) and 2143 (A2143G), which lead to a conformational change and a decrease in macrolide binding (Fig. 3) (413, 556). All macrolides are concerned.

• Open in new tab
• Download powerpoint

FIG. 3.

Mutations arising in domain V of Helicobacter pylori 23S rRNA and conferring resistance to clarithromycin.

(ii) Amoxicillin.Amoxicillin acts by interfering with the peptidoglycan synthesis, especially by blocking transporters named penicillin binding proteins (PBP). The rare amoxicillin-resistant H. pylori strains harbor mutations on the pbp-1a gene. The amino acid substitution Ser-414→Arg appears to be involved (163), leading to a blockage of penicillin transport. Tolerance to amoxicillin has also been described, and the mechanism proposed was the lack of a fourth PBP, namely PBP-D (114).

(iii) Tetracyclines.Tetracyclines interfere in the protein synthesis at the ribosome level by binding to the 30S subunit. The change in a nucleotidic triplet (AGA-926 to 928→TTC), cognate of the positions 965 to 967 in E. coli, has been associated with resistance to these compounds probably because of a lack of binding to the h₁ loop, which is the binding site of tetracyclines. Tetracycline targets the two rrn 16S operons (162, 531).

Single or dual mutations at these positions lead to intermediary MICs. The need to have three mutational events can explain the rarity of tetracycline resistance (91, 161, 408). Tetracycline-resistant strains with no mutation in position 926 to 928 have also been described, and efflux is the mechanism most likely to be involved. These strains, as well as those with the mutations, exhibited a decreased tetracycline accumulation inside the cells (583).

(iv) Fluoroquinolones.Fluoroquinolones inhibit the A subunit of the DNA gyrase, encoded by the gyrA gene. Mutations in the quinolone resistance-determining region of gyrA are found in H. pylori as well as in other bacteria. The amino acid positions concerned are mainly 87 and 91 (391, 513).

(v) Rifampins.Rifampins inhibit the B subunit of the DNA-dependent RNA polymerase encoded by the rpoB gene. Mutations have been described for the rpoB gene of H. pylori at positions 524, 525, and 585, as in Mycobacterium tuberculosis and E. coli (211). Another mutation at position 149 has also been described previously (213).

(vi) Nitroimidazoles.5-Nitroimidazoles have to be reduced in the cell to alter bacterial DNA. An important gene in this respect is rdxA, an oxygen-insensitive nitroreductase. Mutations in rdxA can render the protein ineffective (224). Other proteins may also be involved in this reduction process, like the flavin oxidoreductase (frxA), while their role is more controversial (338, 367). In addition, a TolC efflux pump appears to play a role in resistance to this group of drugs (541).

(i) Phenotypic methods. (a) Agar dilution method. The agar dilution method, usually considered the reference method to compare other techniques, has been proposed by the Clinical Laboratory Standard Institute (CLSI) (formerly National Committee for Clinical Laboratory Standards) as the method to be used for H. pylori clarithromycin susceptibility testing (402). In Europe, a working group of the European Helicobacter pylori Study Group has also published guidelines which are similar to the CLSI recommendations (176). They are all presented in Table 12.

For clarithromycin, the breakpoint proposed for susceptible strains is 0.25 μg/ml, that for resistant strains is >0.5 μg/ml, and that for intermediate strains is 0.5 μg/ml. Excellent predictive values for the success of the clarithromycin-amoxicillin-PPI triple therapy were obtained with these breakpoints.

While the same standardization work has not been done for other antibiotics, e.g., amoxicillin, tetracycline, rifabutin, and levofloxacin, agar dilution can nevertheless be applied.

For amoxicillin, MICs of >0.5 μg/ml are seldom found. Strains with MICs of 0.25 to 0.5 μg/ml correspond to an intermediate susceptibility, but the clinical impact has not been evaluated.

The breakpoints commonly used for the other antibiotics are as follows: tetracycline, 2 μg/ml; rifabutin, 1 μg/ml; ciprofloxacin (often tested instead of levofloxacin), 1 μg/ml.

Metronidazole is a special case because most of the studies have shown a lack of inter- and intralaboratory reproducibility (176, 363). The reason is not known. It may be that the intracellular redox potential is not controlled, whereas this parameter is important for metronidazole reduction. Indeed, a preincubation of the media in an anaerobic atmosphere has been shown to increase metronidazole activity (63). The lack of correlation between the susceptibility results and H. pylori eradication is also of major concern. Strains with a high MIC can be eradicated possibly due to a variable redox potential inside the stomach. The threshold commonly used to define metronidazole resistance is >8 μg/ml.

Because of this lack of clinicobacteriological correlation, the Maastricht III 2005 Consensus Report discourages routine metronidazole susceptibility testing (334).

(b) Broth dilution method. The broth dilution method has the advantage of being adaptable to automation. However, it has rarely been used for H. pylori because of the difficulty of growing this bacterium in broth (82). Supplementation of media such as brucella broth or Mueller-Hinton broth led to satisfactory results (197, 447). A good correlation has been found with Etest, except for metronidazole.

(c) Breakpoint susceptibility testing. Breakpoint susceptibility testing is a simplified version of the agar dilution method previously cited. It consists of inoculating a streak of the strain to be tested on an agar plate containing an antibiotic concentration equal to the breakpoint concentration which defines resistance, for example, 1 μg/ml for clarithromycin, or two different concentrations (0.25 and 1 μg/ml) to classify the strains as susceptible, intermediary, or resistant.

This test is easy to perform, and theoretically excellent, but the media must be prepared in the laboratory. It has been used in comparison to the agar diffusion method (Etest and disk) for metronidazole susceptibility testing with a 94% correlation, but reproducibility was not studied (40, 546).

(d) Disk diffusion testing. The disk diffusion method is the simplest and most economic for routine susceptibility testing. However, it is generally not recommended for slow-growing bacteria.

Disk diffusion has been validated in France to detect macrolide resistance. Due to an important gap between MICs of susceptible and resistant strains, a clear separation is possible by this method. The breakpoint inhibition zone, corresponding to an MIC of >0.5 μg/ml, was 22 mm for clarithromycin and 17 mm for erythromycin. This latter antibiotic is the one recommended for macrolide susceptibility testing (191).

In contrast, discrepant results have been obtained for metronidazole. As previously indicated, it could be due to a lack of standardization of parameters usually not taken into account, e.g., the delay between the medium preparation and the performance of the test, which determines the redox potential (63). This method has not been validated for the other antibiotics, but a good correlation is usually found with the other methods.

(e) Etest. The Etest method has the advantage of being a quantitative method with a direct expression of MICs, and furthermore, it is adapted to slow-growing bacteria like H. pylori. A good correlation has been found between this method and the agar dilution method, with the exception of metronidazole, as mentioned before (176).

(f) Which phenotypic method to use? For routine testing, the minimum is to test a macrolide, and the best cost-efficacy ratio is the disk diffusion method, with an erythromycin disk (191). To know the exact MIC, the Etest or agar dilution method can be used.

Given the lack of reproducibility and the low predictive value for treatment efficacy, metronidazole is no longer recommended for routine testing (334). If necessary, the best method is agar dilution or its simplified version, i.e., the breakpoint method with 8 μg/ml of metronidazole.

The agar diffusion method using disks can be used for susceptibility testing of other drugs, tetracycline, rifabutin, and levofloxacin, because of its practical aspect and limited cost. It allows the detection of resistant strains. In the case of resistance, an Etest can be performed.

(ii) Genotypic detection of resistance.As mentioned above, H. pylori resistance is essentially due to chromosomal mutations, and for the most part, a limited number of punctual mutations are present which can be easily detected with molecular tests.

(a) Clarithromycin. Given the importance of clarithromycin and the low number of mutations concerned, numerous genotypic methods have been developed to detect this resistance. They are presented in Table 13 and were originally described in a review (427).

The two main methods used are PCR-RFLP and real-time PCR.

The PCR-RFLP method was first applied to the detection of H. pylori 23S rRNA gene mutations in 1996 (556). It is based on the fact that mutations reveal restriction sites within the amplicon obtained with primers specific to the H. pylori 23S rRNA gene. These restriction sites are recognized by the enzyme BsaI for the A2142G mutation and BsbI for the A2143G mutation, so two bands will be present on the gel if one or the other of these mutations is present. More recently, a third enzyme (BceAI) has been proposed to detect the A2142C mutation (366). This method is simple but has the disadvantages of standard PCR, especially with regard to the delay in obtaining results and the need to manipulate the amplicons obtained.

The real-time PCR method has been mentioned to detect H. pylori directly in biopsy specimens as well as point mutations.

The first application including H. pylori identification and detection of clarithromycin resistance using the FRET followed by a melting curve analysis was performed on strains in 1999. SYBR green 1, a fluorophore specific for double-stranded DNA was used as the quencher which transfers its energy to a second fluorophore, Cy5, fixed on a probe specific to H. pylori to detect the mutations inducing clarithromycin resistance (164). The same method was then applied directly to biopsy specimens (71).

Another approach using a biprobe was subsequently proposed (348, 421). In the method of Oleastro et al., a 267-bp fragment of the 23S rRNA gene was amplified by using specific primers. A sensor probe and an anchor probe hybridizing three bases upstream from the former were used. After amplification using a LightCycler, a melting curve analysis was performed. The wild-type strain's melting peak was 62°C, while that for the A2142C mutant was 58°C and those for the A2143G and A2142G mutants were 53 and 54°C, respectively, given the existence of a nucleotide mismatch between the sequence and the hybridization probe. The 1°C difference between A2143G and A2142G did not allow the separation of two genotypes because of a 1°C maximum variability between runs. This method proved to be specific and is applied directly to gastric biopsy specimens. When tested on 200 H. pylori-positive samples, a perfect agreement between the genotype and the clarithromycin phenotypic susceptibility testing was found in 96.4%.

This method represents important progress in the diagnosis of H. pylori because the result can be obtained within 2 h, and the possibility of contamination with the amplicons is limited because the reaction is performed in a closed tube. A limitation of this method for H. pylori is that few 23S rRNA gene sequences from other Helicobacter species are present in the databases. In our experience, H. heilmannii is detected with H. pylori 23S rRNA gene PCR.

The application of this method on stool samples is a breakthrough in our approach to susceptibility testing (486).

A variation of this method proposed by Lascols et al. consists of a quantitative detection of H. pylori in gastric biopsy specimens by real-time FRET PCR and, in the event of a positive result, a different biprobe is used to perform another hybridization followed by a melting curve analysis (292).

The possibility of detecting clarithromycin resistance without performing PCR also exists by FISH. This method has been applied to H. pylori and its clarithromycin resistance by Trebesius et al. (529). It consists of an rRNA-based whole-cell in situ hybridization using a set of fluorescent-labeled oligonucleotide probes. Labeling of intact single bacteria is monitored by fluorescence microscopy. This method allows the detection of H. pylori with a 16S rRNA probe labeled with the fluorochrome Cy3 (red) and detection of resistant mutants with a 23S rRNA probe labeled with fluorescein (green) simultaneously. This method proved to be sensitive and specific compared to standard methods of culture and susceptibility testing. FISH has the advantage of being independent of a nucleic acid preparation, it is not prone to inhibition like PCR, and it is quick. It also allows visualization of the bacteria, including coccoidal forms, and can be performed on paraffin-embedded biopsy samples (240). A limitation may be the observer-dependent result and sometimes difficulty in reading. It has not been compared yet to the FRET-melting curve analysis protocol.

(b) Fluoroquinolones. The FRET-melting curve analysis method was also applied to detect H. pylori resistance to fluoroquinolones. All mutations eventually present in the quinolone resistance-determining region do not translate into resistance. Nevertheless, using two biprobes, it is possible to detect the main mutations at amino acid position 87 or 91 (175).

(c) Tetracycline. A PCR-RFLP has been developed which amplifies part of the 16S rRNA gene. The amplicon is then submitted to restriction by Hinft. The strains which harbor the triple mutation leading to a high tetracycline resistance level exhibit three bands, while only two are present for susceptible strains or for strains with a low resistance level (464).

(d) Metronidazole. The important number of mutations which may arise in the rdxA gene, some being independent of H. pylori resistance to this antibiotic, did not allow the development of molecular methods for detection of resistance to metronidazole.

An original approach is the detection of the RdxA protein by immunoblotting using rabbit antisera against RdxA. It is possible to visualize a 24-kDa immunoreactive band corresponding to the RdxA band of metronidazole-susceptible strains which it is not seen in resistant strains (294). However, false positives, i.e., the presence of a band in resistant strains, occur in 10% of strains.

This promising method deserves confirmation by studying its predictive value for the eradication in treatments including metronidazole.

(e) Other antibiotics. No molecular method has yet been proposed to detect H. pylori resistance to rifabutin, probably because it is seldom encountered.

A more thorough knowledge of the pbp1 gene mutations should allow the development of a molecular test for amoxicillin.

(i) Prevalence of H. pylori resistance to antibiotics.Numerous studies have been performed to determine the prevalence of H. pylori resistance to antibiotics. However, many of them have drawbacks, especially concerning the number and the representativeness of the strains tested. Most of the studies were performed in referral centers for H. pylori treatment where patients who experienced previous treatment failures were explored, increasing the probability of finding resistant strains. These cases are not representative of the patients as a whole and, because these studies are monocentric, the number of patients may be low, leading to wide confidence intervals of the prevalence obtained.

Ideal studies involving patients who are representative of a given region by being selected at random are few. An alternative has been to analyze the prevalence data obtained from clinical trials aiming to evaluate new regimens. Since prevalence is in essence an evolving phenomenon, only the most recent articles (1999 to 2005) have been taken into account in this review (357). However, given the delay in publication, only data from the end of the last decade are essentially available.

(a) Prevalence of H. pylori resistance to clarithromycin. Resistance to clarithromycin is the prevalence which has been the most widely studied so far. It ranges from close to nil to 20 to 25%. In the United States, a prevalence of 10 to 15% was found based on strains isolated during clinical trials, regardless of the region studied (119, 369, 425). In Canada, a systematic review of the studies published before the year 2000 estimated this resistance to be less than 4% (135). A multicenter European study carried out in 1998 in 17 countries found a global prevalence of 9.9% (95% confidence interval, 8.3 to 11.7) with a significant difference between Northern Europe (9.3%) and Southern Europe (18%) (178). A similar range was found in a recently published systematic review (357). The prevalence is even higher in children (24%) because children are more likely to be treated with macrolides when they have respiratory infections (268).

The essential risk factor for clarithromycin resistance is a previous consumption of macrolides, and therefore, a trend for an increased prevalence is observed in most countries.

The prevalence of secondary clarithromycin resistance, i.e., after failure of a treatment including this drug, is extremely high, in the range of 60% (212).

(b) Prevalence of H. pylori resistance to metronidazole. As stated previously, the results concerning H. pylori resistance to metronidazole are not reproducible, and on an individual basis, their interest is limited. Nevertheless, the trend of low, medium, or high prevalence rates observed at a population level seems real.

A global resistance in the range of 20 to 40% is observed in the United States (119, 369, 425). as well as in Europe (357). It was 33.1% in the European multicenter study previously cited, with no significant difference between the North and the South (178). In developing countries, the prevalence is much higher (50 to 80%) (400, 525, 573), and it is relatively rare in Japan (9 to 12%) (441).

Again, metronidazole consumption appears to be an important risk factor for this resistance.

The prevalence of secondary metronidazole resistance is in the range of 65 to 75% (212).

(c) Prevalence of H. pylori resistance to other antibiotics. The prevalence of H. pylori resistance to amoxicillin is very low (<1%), as is the prevalence of H. pylori resistance to tetracycline, except in a few countries like South Korea (258).

The prevalence of H. pylori resistance to rifampins is virtually absent, given that these antibiotics have a limited use. In contrast, fluoroquinolones, which have shown an increasing consumption over the past few years, are leading to a higher prevalence, e.g., 20% in adults in Portugal (52). A secondary resistance rate of 9% has been found in Germany (212).

(ii) Impact on H. pylori eradication.The impact of antibiotic resistance on H. pylori eradication must take into account the complexity of the treatment used. Up to 1998, few clinical trials included susceptibility testing, and they have been reviewed previously (227, 545). We performed a systematic review of clinical trials performed between 1999 and 2003, which is reported in the following paragraphs. Clarithromycin and metronidazole are essentially concerned (357).

(a) Clinical impact of clarithromycin resistance. With regard to the PPI-clarithromycin-amoxicillin treatment, a major difference was observed between susceptible strains (87.8% eradication) and resistant strains (18.3% eradication), i.e., a 70% decrease in efficacy if the strain was resistant, with the pooled Mantel-Haenzel odds ratio being 24.5 (95% confidence interval, 17.2 to 35) (Table 14). If the treatment was PPI-clarithromycin-metronidazole for metronidazole-susceptible strains, the corresponding figures were 97% eradication for susceptible strains and 50% for resistant strains.

(b) Clinical impact of metronidazole resistance. When a PPI-clarithromycin-metronidazole regimen was used for clarithromycin-susceptible strains, the eradication rate was 97% for metronidazole-susceptible strains and 72.6% for metronidazole-resistant strains, i.e., only a 25% decrease in efficacy.

With the PPI-amoxicillin-metronidazole regimen, the corresponding figures were 89% for susceptible strains and 64% for resistant strains, i.e., again a 25% decrease in efficacy.

(c) Clinical impact of resistance to both clarithromycin and metronidazole. Very few patients harbored strains with double resistance to both clarithromycin and metronidazole in the trials performed in the past. None of the 14 patients experienced an H. pylori eradication.

(d) Clinical impact of resistance to other antibiotics. We can hypothesize that resistance to amoxicillin, tetracycline, rifabutin, and levofloxacin has an impact on treatment success. However, clinical trials performed in the past either did not include resistant strains which are still rare or did not include susceptibility testing.