S2099 A Tale of Two Machines: Discordant SIBO Breath Test Results

Introduction: Small intestinal bacterial overgrowth (SIBO) is diagnosed using a breath test. The breath test measures parts per million (ppm) of hydrogen, methane, and other gases. The clinician then makes a diagnosis based on established guidelines of acceptable limits or changes from baseline. At our center, we observed much higher rates of intestinal methanogen overgrowth (IMO) than the national average, despite machine optimization per manufacturer instructions. A replacement breath test machine had more typical results. This study aims to quantify the diagnostic difference between the 2 machines, with the aim to understand the reliability of the most common diagnostic tool for SIBO. Methods: This is a paired design study among patients who underwent SIBO glucose breath testing at a single center between 2023 and 2025. The machines used were both the same make and model of breath test machine. They were purchased in 2007 and 2024. We calculated P-values for the agreement between the machines using McNemar’s test with continuity correction as well as Cohen’s kappa (κ). We calculated these values for results of both IMO and hydrogen predominant SIBO (H-SIBO) tests. Results: We identified 349 individuals who had undergone SIBO glucose breath testing on both the old and new machines. Of these, 201 were discordant pairs for IMO; 100% of those discordant pairs were positive on the old machine and negative on the new machine (P < 0.00001). The κ for these paired tests is 0.13. For H-SIBO, there were 14 discordant pairs; 100% of these discordant pairs are positive on the new machine and negative on the old machine (P = 0.00018) The κ of these paired tests is 0.85. Conclusion: The machine purchased in 2007 had a much higher rate of IMO positivity than the national average and had higher rates than the machine purchased in 2024 using the same samples. This is statistically significant with the κ of 0.13 representing “slight agreement” between the 2 machines. However, the disagreement on H-SIBO, while still statistically significant, has a κ of 0.85, representing “near perfect agreement.” These results are significant because older breath test machines may have lower reliability, particularly for diagnosis of IMO, even after careful and repeated optimization and pump replacement.