Introduction Increasing duration of pH studies improves consistency of GERD diagnosis but clinical utility of the method is not established. Aim: (1) to identify measurements from prolonged pH studies that discriminate healthy volunteers (HVs) and GERD patients (2) to compare prediction of PPI response from prolonged and standard measurement.
Methods HVs and patients with reflux symptoms entered a prospective trial. Quality of life (RAND-36) and symptom severity (Eraflux) was assessed on and off PPI and after 2 weeks ×2/day PPI. Endoscopy recorded mucosal disease. Wireless pH system (Bravo®, Given Imaging) measured acid reflux and symptoms up to 4 days. Receiver Operating Curve (ROC) assessed prediction of PPI response. For each prediction 80% of patients were randomly selected as training set, remaining 20% constituted test set. This was repeated 200 times producing average ROC with SEs. Area under Curve (AUC) quantified quality of prediction.
Results Complete data were available from 25/33 HVs (18F, age 20–56) and 70/108 patients (31F, age 18–77), >320 days in total. Oesophagitis was present in 9 HVs (32%: Grade A) and 26 patients (33%: Grade A=19, B=2, C-D=5). Acid exposure time was elevated (AET >5.6%) in 3 (12%) HVs and 35 (50%) patients. Eraflux off-PPI was >25 (consistent with GERD) in 60/63 patients and fell by mean 7 (95% CI 5 to 10) on PPI, 46% reported positive response (>3 fall). Diagnosis: Endoscopy, AET and reflux-symptom association analysis (Symptom Index (SI)) did not discriminate health/disease; but reflux-associated symptoms/day (nRS/Day) covered different ranges for HV and patients. Logistic regression with bootstrap validation identified that ≥3 RS/day corresponded to ∼50% probability that participant was a patient.
PPI response: Clinical parameters and AET did not predict outcome. SI (9.2 vs 30.2, p=0.0023) and nRS/Day (1 vs 2.6, p=0.012) were higher in responders. RAND-36 scores for poor physical role and pain were higher in non-responders (p∼0.1). SI ROC had an AUC of 0.73 (CI 0.51 to 0.92). SI >25 was the optimal cut-off for identifying PPI responders (Abstract PTU-197 figure 1). Prediction quality from 24 h studies was lower (AUC 0.69) and CIs for all parameters were wider with lower CI.
Conclusion Diagnostic consistency for all parameters increases with study duration. A simple count of nRS/Day best discriminates HVs from patients on pH studies. SI >25 provides single best prediction of PPI response; but quality of predictions was modest in this population with low PPI response.
Competing interests None declared.