INTER-OBSERVER AND INTRA-OBSERVER AGREEMENT IN THE ANALYSIS OF OESOPHAGEAL PH-IMPEDANCE TRACINGS.

AIM Information resulting from oesophageal 24h pH-impedance monitoring (pH-MII) may have important consequences for patients’ management. Aims of this study were 1) to evaluate agreement for number of reflux episodes and symptom/reflux association indexes in MII-pH tracings analysis between and within three experienced observers working in different European Centers, 2) to evaluate the predictivity of a shorter manual analysis targeted to the two-minutes window periods before each symptomatic episode with regards to the traditional. MATERIAL AND METHODS Forty consecutive pH-MII tracings of patients (23 women, median age 55 years) referred for suspected oesophageal (n=24) or extra-oesophageal (n=16) GORD symptoms in two Northern Italian Centers were included (20 tracings/Center). After automatic analysis (MMS, Enschede, The Nederlands), tracings were duplicated and anonymized. Three expert observers blindly performed the traditional manual analysis on the 40 duplicated tracings, erasing or adding reflux episodes according to their judgment based on the Oporto Criteria. The first manual analysis was performed in a sequential order, the second one in a randomly assigned order. Subsequently a short manual analysis was also blindly performed. Values of both symptom association indexes (Symptom Index=S.I. and Symptom Association Probability=S.A.P.) were transformed into binary response (i.e. positive or negative). Inter- and intra-observer agreement was calculated with Cohen’s Coefficient (k) and/or with percentage of agreement. RESULTS Inter-observer agreement on acid reflux was excellent between the three observers for both S.I. and S.A.P. (K=1.0; 100%), whereas considering non-acid reflux it was high for S.I. (95%, 92.5% and 95% of concordance for observer 1,2 and 3 respectively) and moderate for S.A.P. (K=0.35; 85% between observer 1 and 2, K=0.36; 82.5% between observer 1 and 3, K=0.23; 87.5% between observer 2 and 3). Intra-observer agreement on acid reflux was good to excellent for both S.I. (K=0.77; 95% for observer 1 and 2 and K=1.0; 100% for observer 3), and S.A.P. (K=0.80; 95% , K=0.72; 92.5% and K=1.0; 100% for observer 1, 2 and 3 respectively). Intra-observer agreement on non-acid reflux was high for S.I. (97.5%, 95% and 97.5% of concordance for observer 1, 2 and 3 respectively) and moderate for S.A.P. (K=0.48; 85% and K=0.54; 90% for observer 1 and 3 and good for observer 2, K=0.79; 2.5%). Considering all the three observers the short analysis was predictive of the traditional one for both S.I., 88% and 98% of the tracings for acid and non-acid reflux, respectively and S.A.P., 99% and 97% of the tracings for acid and non-acid reflux, respectively. CONCLUSIONS We found a good inter- and intra-observer agreement for symptom/reflux association indexes, especially when considering acid reflux. Moreover we demonstrated that short analysis was highly predictive of the traditional manual analysis, suggesting that the short one could be used in routine clinical practice with considerable savings.