Background: Both simple proportions and statistical tests are utilised for symptom-reflux association. We systematically compared three such tests in a clinical setting.
Aim: To compare the three commonly used symptom reflux association tests in a large cohort of patients undergoing ambulatory pH monitoring for the evaluation of oesophageal symptoms.
Methods: Ambulatory pH data from 772 symptomatic subjects (49.1 ± 0.5 years; 479 F) tested off therapy were assessed for acid exposure time (AET, elevated when pH <4 for ≥4%), symptom index (SI, ≥50% when positive), and symptom association probability (SAP) and Ghillebert probability estimate (GPE, P < 0.05 when positive). Test concordance and discordance were individually assessed; discordance between statistical tests was minor if one had P < 0.1 while the other was positive. Logistic regression determined independent predictors of test discordance.
Results: The SAP, GPE and SI were positive in 42.7%, 39.3% and 33.9% respectively. GPE performed extremely well compared to SAP (sensitivity 0.95, specificity 0.91), with major discordance in only 2.8%. Positive concordance was significantly higher when AET was abnormal. GPE underestimated symptom association compared to SAP, whereas SAP was subject to symptom over-counting in 33.3% of discordant cases. GPE-SAP discordance was associated with higher AET (7.5% vs. 5.1%) and more symptoms (19.3 vs. 10.7, P > 0.001 for each comparison with concordant tests); both remained significant on logistic regression analysis (P ≤ 0.003). SI was discordant with SAP when symptoms were extremely frequent (median 19, IQR 10-32) or limited (median 1, IQR 1-2), and concordant when median 6 symptoms (IQR 3-12) were recorded.
Conclusions: The GPE can be used interchangeably with SAP in symptom reflux association. SI has uncertain value with very high and very low symptom counts.
© 2012 Blackwell Publishing Ltd.