Digestion. 2013;87(3):163-9.

A screening trial of Helicobacter pylori-specific antigen tests in saliva to identify an oral infection.

Kuo Ching Yee, M.H.Wei, Hsian Ching Yee, Karin D.E. Everett, Hsian Pei Yee and Noriko Hazeki-Taylor

Research Division, Ameritek USA, Everett, Washington 98208, USA. kcyee@ameritek.org



OBJECTIVE: Helicobacter pylori infection places a heavy burden on medical and economic resources. Standard diagnosis requires the presence of established H. pylori gastric disease.

STUDY DESIGN AND SETTING: A multicenter screening trial assessing 2 immunochromatographic H. pylori antigen oral tests was carried out with 201 participants. The analysis also included a urea breath test (UBT), a Campylobacter-like organism test, silver stain, culture, serology, and stool tests.

RESULTS: The participants were grouped into UBT positive (UBT+) and UBT negative (UBT-) people, using conventional methods with congruent clusters based on p values from McNemar’s paired χ2 analysis and 95% CI estimates. Both oral tests were also positive in 82% of the seropositive UBT- people. However, oral antigen and seroprevalence divided UBT- people into 2 statistically separate CI subgroups: the UBT- symptomatic (highly positive) group and the UBT- asymptomatic (mostly negative) group. 90.5% of all people whose oral tests were both negative were also UBT-.

CONCLUSIONS: Saliva H. pylori antigen is an important indicator in UBT- asymptomatic patients. Currently, its clinical significance remains uncertain, but saliva may be a reservoir from where H. pylori is transmitted to the stomach. In symptomatic patients, it is strongly associated with stomach infection.

Copyright © 2013 S. Karger AG, Basel.

PMID: 23615458



What is already known about this subject:

  • An article published in Lancet by Warren J.R. and Marshall B. indicated that H. pylori contributes to ulcers, gastritis, and gastric adenocarcinoma. During 2005, they shared the Nobel Prize in Physiology and Medicine for this discovery. However, despite their discovery, during the past several years the severity of H. pylori infection has not significantly diminished. After successful eradication, the annual H. pylori recurrence rate is approximately 13% due to oral H. pylori infection.
  • Established clinical diagnostic techniques do not identify an oral etiologic basis to H. pylori prior to gastric infection.
  • During the past 20 years, there has been disagreement as to whether oral infection of H. pylori exists or not with no definite conclusion.

What are the new findings:

  • In medical practice, patients with negative results on the UBT suggest that their stomach infection of H. pylori is cured. In fact, patient can present negative UBT results and yet exhibit H. pylori infection due to oral infection.
  • The present paper provides evidence that H. pylori oral infection is nonetheless present. In Asia, more than 90% of population suffered with oral H. pylori infection but with no one can tell.

How might it impact on clinical practice in the foreseeable future?

  • Oral H. pylori antigen in people with no evidence of gastric disease would suggest that eventual gastric infection may be preventable by surveillance, dental care, and antibiotic treatment.



Helicobacter pylori infection places a heavy burden on medical and economic resources in developing nations. Standard diagnosis requires the presence of established H. pylori gastric disease or fecal shedding. Thus, rapid and convenient identification and treatment of people at risk for developing infection is not possible. This study aims to determine whether evidence of H. pylori antigen in the mouth can be used to identify people at risk for disease.


A multicenter screening trial of two rapid immunochromatographic H. pylori oral antigen tests (flagellin and urease) was carried out using saliva from 201 men and women (ages 11–82). They were also tested by conventional H. pylori methods: urea breath test (UBT), CLO, silver stain, culture, serology, and stool tests.


All test results were increasingly positive with increasing evidence of disease. Conventional methods grouped into UBT+ and UBT– congruent clusters (primarily positive results or primarily negative results, respectively), based on p-values from McNemar’s paired chi-square analysis and 95% confidence interval (CI) estimates. Oral antigen and serology tests also formed a highly positive UBT+ cluster. However, oral antigen and seroprevalence in UBT– people clustered together into two statistically separate CI subgroups: UBT– symptomatic (highly positive) and UBT– asymptomatic (mostly negative). Both oral tests were positive in three of every four seropositive UBT– people. 90∙5% of all people whose oral tests were both negative were also UBT–.


These data suggest that oral H. pylori antigen may be a significant indicator of disease risk. The absence of antigen would signal that costly screening was unnecessary. Oral H. pylori antigen in people with no evidence of gastric disease would suggest that eventual gastric infection may be preventable by surveillance, dental care, and antibiotic treatment.

Sorting Data into Coherent Groups

Agreement between UBT and culture data is often used as a Gold Standard for H. pylori gastric infection. The symptomatic people in this study were tested by both methods. These data showed near-perfect agreement of UBT and culture (98∙2%, only two of 110 individuals differed). Thus, either method was an appropriate reference standard for gastric H. pylori infection in this study. As a general unsorted population, 110 of the 201 individuals overall were symptomatic (54∙7%) and 115 of the 201 were UBT+ (57∙2%). Sorting results for each symptomatic group into UBT+ and UBT– subgroups resulted in Exact 95% CI clusters that were congruent within each subgroup (Figures 1 and 2). For UBT– individuals, the ranges of 95% CIs of gastric and stool tests were ≤27∙4% (culture = 0–12%). The H. pylori-specific oral and serum antibody tests for UBT– individuals, however, had 95% CI ranges that fell into two distinct clusters: low ranges for asymptomatic individuals and high ranges for symptomatic individuals (Figure 2). Within every group and subgroup, McNemar’s p-value calculation showed that oral H. pylori test results were always significantly similar to H. pylori-specific serum antibody results (p>0∙0783).

KC Yee-fig1Figure 1. 95% Confidence Intervals for gastric and stool tests. Light bars = Symptomatic people tested by CLO, silver stain, and culture; bold bars = Asymptomatic people tested with the stool flagellin test. UBT+ mean positive rates (triangles): CLO = 90∙1%; silver stain = 86∙4%; culture = 97∙5%; stool = 94∙1%. UBT– mean positive rates (triangles): CLO = 0∙0%; silver stain = 10∙3%; culture = 0∙0%; stool = 14∙0%.


KC Yee-fig2Figure 2. 95% Confidence Intervals for serum and oral tests. Light bars = Symptomatic people; bold bars = Asymptomatic people; UBT = urea breath test; serum = serum antibody test. UBT+ mean positive rates (triangles): serum = 96∙3% and 82∙4%; oral urease = 100∙0% and 85∙3%; oral flagellin = 92∙6% and 76∙5%. UBT– mean positive rates (triangles): serum = 75∙9% and 29∙8%; oral urease = 75∙9% and 43∙9%; oral flagellin = 72∙4% and 36∙8%.

Among UBT+ individuals the 95% CIs for H. pylori-specific salivary urease, salivary flagellin, and serum antibody were all above 58∙7% (the symptomatic subgroup spanned 84∙6–100%) (Figure 2). Although the mean positive rates for oral vs. serum tests did not differ significantly (p>0∙0832) in the UBT+ symptomatic subgroup, the oral flagellin vs. oral urease test results were significantly different (p=0∙0143).

Comparative Significance Testing of Oral Data

The consistent similarities of oral and serum data were further examined by paired statistical difference comparisons with other tests (Table 2). In both UBT– subgroups, serum and both oral test results were significantly more often positive than stool, CLO, silver stain, and culture (p<0∙0003). In the UBT+ asymptomatic subgroup, only stool vs. oral flagellin results differed significantly, the stool results being 7-fold higher (94∙1%) than in UBT– asymptomatic people (14∙0%). In the UBT+ symptomatic subgroup, oral test results were not significantly different from culture (p=0∙1573). The oral urease antigen test was significantly more positive than the oral flagellin test, CLO, or silver stain. Serum results were significantly more often positive than silver stain.

Table 2. Significant test differences compared to serum and oral H. pylori tests in four subgroups. Note that the stool test was carried out only in asymptomatic people. CLO, silver stain, and culture were carried out only in symptomatic people. B = specific serum antibody test; C = culture; CLO = CLO; OF = oral flagellin test; SS = silver stain; SF = stool flagellin test; U = urease oral antigen test.

KC Yee-fig3

Predictive value of oral tests and serum

Population statistics do not show individual-to-individual variation within a population. So in Table 3 we used data for each individual to document of serum and oral test differences and to calculate the degree to which this predicted an individual’s UBT status. The analysis showed that people with positive oral and serum tests had a 3-to-1 chance of also being UBT+. When only the serum antibody test was positive, the predictive ratio was 2∙7-to-1. Remarkably, this also meant that 39 serum-positive people were UBT– and that 41 people who were positive for both oral tests were UBT–. Regardless of UBT status, 127 individuals who were antigen positive by both oral tests were also serum antibody positive (14 were not serum positive).

Table 3. Individual-level presentation of oral and serum test results. (A) The number of individuals who were positive or negative for the indicated tests is recorded, and the ratios of UBT status within the indicated test(s) results are calculated. (B) Data were also summarized as the percent of UBT group total. For example, the number of people in the UBT– group was 57 + 29 = 86 people. Serum positive people in this group were 17 + 22 = 39. So, 39/86 = 45∙3% of individuals in the UBT– group were serum positive.

KC Yee-fig4

Whereas 82 people in this study were positive for every test they were given (Table 1), it was equally important to recognize people who were negative. Overall, 34 people were negative for every test, so negative predictive ratios were examined for oral tests and serum (Table 3). The strongest predictive ratio was when both oral tests were negative (these people had a 9∙5-to-1 chance of being UBT–). Oral test results among the 86 UBT– people grouped into two categories of almost identical size: 38 people were negative for both oral tests and 41 were positive for both oral tests. This indicated that a significant proportion of people without gastric disease carried H. pylori in the mouth.


Performance and reliability of the oral antigen tests

The Table 3 analysis showed that overall a high percentage of UBT+ individuals were also positive for serum, oral tests, or both (>82%). Very low percentages of UBT+ people were negative for serum or oral tests (≤7∙8%). These observations indicated that both the serum and oral antigen tests performed reliably by detecting evidence of H. pylori in infected individuals. In contrast, all Table 3 percentages for UBT– individuals were quite similar (mean = 45∙2%). Examination of UBT– symptomatic and asymptomatic subgroups, however, revealed that this was because the percentages were the sum of high numbers of test-positive symptomatic people and high numbers of test-negative asymptomatic people. Normalized data indicated that both antigen tests were strong indicators of the presence of H. pylori antigen in the mouth (Figure 3).

Well defined negative results are essential for meaningful testing. Table 3 and Figure 3 showed that 44∙2% of all 201 people were negative for both oral antigen tests and that these were distributed primarily to UBT– asymptomatic individuals. The incidence of both tests being negative was high in UBT– people and declined stepwise to zero as the evidence of disease increased (Figure 3). Just 3∙5% of UBT+ people were negative for both oral tests and these were only in the asymptomatic subgroup.

KC Yee-fig3-5Figure 3. Normalized oral test results for individual people. Every person was either ++ = both oral tests positive; +/- = only one oral test positive; or – – = both oral tests negative. Each subgroup has been normalized to represent the same size group as UBT+ symptomatic (81 people).

Multiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier SchönmannMultiselect Ultimate Query Plugin by InoPlugs Web Design Vienna | Webdesign Wien and Juwelier Schönmann