Temporal analysis of the relationship between gastroesophageal reflux and cardiorespiratory events in preterm infants

Résumé

Gastroesophageal reflux (GER) is a physiological mechanism that occurs in adults and children. The gastric content travels backwards into the esophagus and occasionally produces symptoms such as cough, heartburn, airways blockage, etc. That is when one can speak of GER disease, a problem considered nowadays as the most expensive digestive disease whose prevalence ranges between 10% and 20% in the western population. In newborn patients, cardiorespiratory (CR) symptoms (apnea, bradicardia or oxygen desaturation) may be temporarily related to GER. The study of symptom association relies on 24-hour synchronized intraluminal pH-impedance and CR monitoring. This method allows detecting GER and CR episodes and therefore to determine whether or not a temporal association exists. In doing so, the Symptom Index (SI), the Symptom Sensitivity Index (SSI) and the Symptom Association Probability (SAP) can be applied as a measure of symptom association. A computational Monte Carlo model was implemented to study symptom association. This model was contrasted with the analysis of the inter-reflux time in 66 patients. We concluded that both, the SI and the SSI were biased estimators whereas the SAP showed a monotonically increasing trend with monitoring time. This model was later applied to investigate a reformulated version of the Binomial Symptom Index (BSI). The value of the BSI remained below chance when no association was defined, and accounted for all the factors involved in symptom association, overcoming the limitation of the previous methods. Afterwards it was studied as a predictive indicator of anti-reflux surgery in 25 preterm neonates. The BSI showed a sensitivity of 100% and a specificity of 95.6% using a one-minute window of association. In addition, an open source computer program was developed for the analysis of symptom association. In conclusion, this work provides a computational model useful to test the performance of the different metrics in symptom association. It proves the BSI to be a reliable and an easy-to-use index with excellent results in both, in silico and clinical studies.