Diagnostic performance of fecal Helicobacter pylori antigen test in Uganda

Over 50% of people in developing countries have their Helicobacter pylori (H. pylori) infection partly due to the ubiquity of factors such as lack of clean water supply, poor sanitation, and overcrowding that are conducive to oral-fecal disease transmission [1]. H. pylori is a class I carcinogen [2] for gastric cancer which ranks as the second leading cause of cancer-related deaths and the fourth most common cancer worldwide [3].

Though the risk of gastric cancer is unknown in developing countries, for Japanese men (a low prevalence group), H. pylori infection confers up to 17% lifetime risk of gastric cancer (compared to 1% for uninfected) [4]. Despite the availability of H. pylori eradication therapy, host genetics, environmental factors, and bacterial virulence limit the effectiveness of therapy [5]. H. pylori strains that disproportionately express the Vacuolating cytotoxin (VacAs1) and Cytotoxin-associated antigen A (CagA) genes predispose to severe disease [6, 7]. Moreover, the ability to type the infecting H. pylori strain to guide clinical care in endemic geographical regions is limited thus more virulent strains accrue from prolonged exposure to ineffective therapy.

Polymerase chain reaction (PCR) based methods for the detection of H. pylori have high sensitivity (> 95%) and specificity (˃95%) [8] and provide additional evidence of antibiotic-use-induced coccoid forms that other H. pylori detection methods do not [9, 10]. Amplification of at least two or more target genes increases the specificity of H. pylori diagnosis and reduces the false-positive rates [11]. Insofar as H. pylori antigen tests that utilize fecal specimens have been available for decades, the diagnostic performance of this easy-to-use and the non-invasive test has not been evaluated in H. pylori endemic settings.

A robust and cheap test is important for diagnostic programs for populations at the highest risk of H. pylori infection and may improve early diagnosis thereby reducing gastric cancer incidence, prevalence, and mortality. We aimed to estimate the diagnostic performance characteristics of a point-of-care qualitative stool antigen test using the presence/absence of nucleotide sequences of two H. pylori virulent genes as the reference standard in southwestern Uganda.

We conducted a diagnostic study between March 2018 and April 2019 at Mbarara Regional Referral Hospital (MRRH) in southwestern Uganda. We consecutively enrolled adults (aged 18 years or greater) referred to the endoscopy unit for first-ever esophagogastroduodenoscopy (EGD). Those screened for enrollment had to have dyspepsia as an indication of EGD and no prior clinical diagnosis of any upper gastrointestinal cancer. The following sequential exclusions were applied; symptoms of gastrointestinal bleeding within seven days, and self-report use of any antibiotics, H₂-receptor antagonists, proton-pump inhibitors, or non-steroidal anti-inflammatory (NSAIDs) medications within a fortnight before referral.

After exclusions, data from 150 participants with dyspepsia were analyzed. The mean age was 55 (standard deviation 16⋅8) years, age ranged from 19 to 90 years, and 78 (52%) were females. Fifty-nine (39⋅3%) tested positive for the stool antigen test. Of the 67 (44⋅7%) participants categorized as positive for 16S rRNA (i.e., with nucleotide sequences of H. pylori-associated genes in gastric tissue), 22 (32⋅8%) had the VacAs1 gene nucleotide sequences, 14 (20⋅9%) had the CagA nucleotide sequences, and 31 (46⋅3%) expressed both VacAs1 and CagA gene nucleotide sequences.

Fifty-seven were positive for SAT and 16 sRNA nucleotide sequences thus the sensitivity of 0⋅85 (95%CI 0⋅76, 0⋅94) and positive predictive value of 0⋅97 (95%CI 0⋅92, 1⋅01). Men had higher sensitivity and positive predictive value than women. Seventy-nine were negative for SAT and 16S rRNA nucleotide sequences thus the specificity of 0⋅98 (95%CI 0⋅94, 1⋅01) and the negative predictive value of 0⋅89 (95%CI 0⋅83, 0⋅95). The specificity and negative predictive value were similar for men and women (Table 1). About 12 individuals with dyspepsia need to be tested with SAT to correctly diagnose 10 individuals as would be positive with 16S rRNA.

As the prevalence of H. pylori increased towards 100% with virtually no false negatives on the SAT (near-perfect sensitivity), the probability of a negative 16S rRNA test associated with a negative SAT (the negative predictive value) decreased. Conversely, as the prevalence decreased toward 0%, even if the SAT had near-perfect specificity (close to 100%) with virtually no false positives, the probability of a positive 16S rRNA test associated with a positive SAT (the positive predictive value) decreased. Taken together, the SAT results are most ‘predictive’ when H. pylori prevalence is at 30% (Fig. 1). Based on the estimated specificity and sensitivity, given a 72% prevalence of H. pylori among Ugandans presenting to a hospital with dyspepsia [14], for a hypothetical population of 100 Ugandans, 66 would test positive by SAT and 16S rRNA (specificity of 0.89 and specificity of 0.98). Among those who will test positive, 99% will be positive by 16S rRNA, and 75.5% of those who will test negative will not have H. pylori by 16S rRNA (Fig. 1).

In the current study evaluating the diagnostic performance of a point-of-care qualitative stool antigen test in people with dyspepsia in rural southwestern Uganda, we found that the SAT had excellent diagnostic performance characteristics for the diagnosis of H. pylori in this disease-endemic setting. Furthermore, we found that twelve individuals with dyspepsia need to be tested with SAT to correctly detect ten individuals with H. pylori infection among people with dyspepsia.

Studies comparing SAT to immunohistological H. pylori testing in low-income settings show a low sensitivity [15, 16] perhaps because most individuals tested had bleeding peptic ulcers, a known cause of false-positive H. pylori [17]. However, using similar comparison methods of SAT with immunohistology, Erzin et al. [18] found high sensitivity and low specificity compared to this current study. Our study used the presence/absence of PCR-16S rRNA to define positive/negative H. pylori as a reference standard, whereas their study used a more stringent definition for positive H. pylori of culture alone or both the histology and the rapid urease test [19]. Our findings are consistent with those from a meta-analysis of eighty-nine studies, including 10,858 untreated adults in Europe and Asia, which found similar diagnostic performance characteristics of monoclonal stool antigen tests in pretreatment individuals [20]. However, the studies included used standard tests with lower sensitivities and specificities like rapid urease test, culture, urea breath test, IgG-antibody-based serology, histology, and immunohistostaining bar PCR-based tests [20]. In addition, the prevalence of H. pylori infection in the geographical settings of the studies included is lower than in individuals with dyspepsia in Uganda.

Though the specificity and negative predictive value in this current study were similar in both sex, males had higher sensitivity and positive predictive value than females. Similarly, Krausse et al.found a higher performance in SAT in males [21]. On the contrary, Abdelmalek et al.found no difference in the diagnostic performance of SAT in Egypt [22]. These findings may reflect potential differences in exposure mechanisms by gender roles.

This current study found the prevalence of H. pylori infection at 39.3% and 44.7% using SAT and PCR-16S rRNA, respectively. These are in keeping with other hospital-based studies in the region that found H. pylori prevalence by SAT of 33.5 – 35.7% and 32.5% by immunohistology [15].

The strength of the current study is the blinding of laboratory technicians who performed SAT and PCR to the clinical diagnosis of participants reducing the differential misclassification bias that would affect the estimates. However, the findings herewith may be extrapolated to other settings with high H. pylori burden with caution that; first, this was a single-hospital study with relatively small sample size. Second, 16S rRNA PCR analysis was done on genomic DNA isolated from gastric tissues which may have underestimated the VacAs1 and CagA prevalence in the study population due to potential PCR inhibitors.

In conclusion, the stool antigen test is a non-invasive and robust diagnostic test that may be used to improve the diagnosis of H. pylori infection in people with dyspepsia in H. pylori endemic resource-limited settings.