EXTERNAL VALIDATION DEMONSTRATES HIGH PREDICTIVE VALUE OF THE LYON SCORE: A NOVEL REFLUX SCORING SYSTEM BASED ON THE LYON CONSENSUS 2.0

Background: The Lyon Consensus 2.0 identifies physiologic, borderline and pathologic reflux burden from acid exposure time (AET), reflux episodes (RE), and mean nocturnal baseline impedance (MNBI). Other findings on endoscopy, reflux monitoring, and high resolution manometry (HRM) escalate or refute likelihood of reflux. Abnormal findings in combination could increase confidence in reflux presence and predict treatment outcome.
Methods: We created a novel score (Lyon Score) to quantify reflux burden and phenotype reflux syndromes. Scores and ranges were assigned a priori to thresholds for each metric based on available research (Table). The Lyon Score was initially calculated in an existing primary cohort with typical reflux symptoms (heartburn, regurgitation, chest pain) evaluated with endoscopy, HRM and pH impedance monitoring off therapy from two centers (one each in Europe and US). Two validation cohorts, one each from Europe (Italy) and Asia (Taiwan), were also analyzed. Symptom response from reflux therapy was assessed using 50% reduction in global symptom scores from validated symptom questionnaires over follow-up. Receiver operating characteristics (ROC) analysis determined performance of the scoring system in predicting overall response in all 3 cohorts, and principal component analysis identified predictors of response within the scoring system.
Results: The primary cohort had 281 patients (median age 53 yr, 57.7% F, 71.9% with heartburn). The validation cohorts had 215 patients (European cohort, age 58 yr, 60.0% F, 100% with heartburn) and 258 patients (Asian cohort, age 47 yr, 60.2% F, 73.5% with regurgitation) (Table 2). In the primary cohort, the Lyon score had an AUC of 0.816 in predicting 50% symptom improvement (Figure, p=0.026), with an optimal Lyon score threshold of 3.75 (sensitivity 81.2%, specificity 74.1%). Only AET (AUC 0.770, p=0.029), MNBI (AUC 0.747, p=0.030) and RE (AUC 0.728, p=0.030) approached the combined score. On principal component analysis, AET, MNBI and RE all demonstrated high values within the component matrix (≥0.80 for each), suggesting that each of these primary metrics carries independent weight in driving response from antireflux therapy. The Lyon score demonstrated excellent to good performance with variance of treatment response between the heartburn-predominant European cohort (AUC 0.896, p<0.001) and the regurgitation-predominant Asian cohort (AUC 0.631, p<0.001). Proportions of patients with 50% response from antireflux therapy progressively increased across phenotypes in all three cohorts as confidence in reflux evidence increased (Table, p<0.001 across groups).
Conclusions: The novel Lyon score incorporates pH-impedance metrics to segregate reflux phenotypes, and identifies varied likelihood of symptom response from antireflux therapy based on dominant symptom. Prospective validation is warranted.