Introduction
Dietary and lifestyle recommendations are the cornerstones of the treatment of GERD. However, the evidence of the efficacy of those recommendations is lacking. We performed a retrospective analysis of the GISTAR study1 participants evaluating the symptoms of GERD and dietary and lifestyle parameters developing a tool for predicting GERD symptoms based on the aforementioned criteria.
Aims & Methods
The aim of the study was to analyse the factors pointing to GERD symptoms through the use of data mining and decision tree classifier.
Methods: GERD diagnosis was evaluated using a self-assessment questionnaire 2. Dietary, demographic and lifestyle data were collected within the GISTAR study protocol1. 3340 40-64 year old participants from the GISTAR study were included in the analysis. 1670 of them did not have GERD symptoms. 41 attributes describing demographic, lifestyle, and dietary habits were used for these respondents to induce a decision tree classification model using Orange Data Mining software3.
Results
Fried potatoes including French fries were the most predictive factor of GERD symptoms (65.3% with symptoms; 1516 from 2321), especially for females (82.6%; 1048/1228). This was more pronounced in the group who consumed sweets once per week (87.2%, 950/1090). In males, who consumed fried potatoes, the next most important factor was the consumption of sweets (46.3%, 431/930).
For participants who didn’t consume fried potatoes (14.6%; 147/1008), 1-3 cups of green tea (37.9%; 66/174) and male gender (71.4 % 20/28) were predictors of GERD symptoms.
The accuracy of the decision tree model was: sensitivity (74.8%), precision (75.6%), specificity (75.9%) and AUC (0.841%).
Conclusion
Fried potatoes, female gender and consumption of sweets are the risk factors predicting the diagnosis of GERD. None of the lifestyle factors were displayed as significant.
References
1. Leja M, Park JY, Murillo R, et al. Multicentric randomised study of Helicobacter pylori eradication and pepsinogen testing for prevention of gastric cancer mortality: the GISTAR study. BMJ Open 2017;7:e016999. doi: 10.1136/bmjopen-2017-016999
2. Jones R, Junghard O, Dent J, Vakil N, Halling K, Wernersson B, Lind T. Development of the GerdQ, a tool for the diagnosis and management of gastro-oesophageal reflux disease in primary care. Aliment Pharmacol Ther. 2009 Nov 15;30(10):1030-8. doi: 10.1111/j.1365-2036.2009.04142.x. Epub 2009 Sep 8. PMID: 19737151.
3. Demsar J, Curk T, Erjavec A, Gorup C, Hocevar T, Milutinovic M, Mozina M, Polajnar M, Toplak M, Staric A, Stajdohar M, Umek L, Zagar L, Zbontar J, Zitnik M, Zupan B (2013) Orange: Data Mining Toolbox in Python, Journal of Machine Learning Research 14(Aug): 2349−2353.