Guadagnoli Livia, Geeraerts Annelies, Geysen Hannelore, Pauwels Ans, Vanuytsel Tim, Tack Jan, Van Oudenhove Lukas
Translational Research Center for Gastrointestinal Disorders, Katholieke Universiteit Leuven, Leuven, Belgium; Laboratory for Brain-Gut Axis Studies, Translational Research Center for Gastrointestinal Disorders, Katholieke Universiteit Leuven, Leuven, Belgium; Leuven Brain Institute, Katholieke Universiteit Leuven, Leuven, Belgium.
Translational Research Center for Gastrointestinal Disorders, Katholieke Universiteit Leuven, Leuven, Belgium.
Gastroenterology. 2023 Oct;165(4):848-860. doi: 10.1053/j.gastro.2023.06.019. Epub 2023 Jun 30.
BACKGROUND & AIMS: Physiological and psychological factors have been found to influence esophageal symptom reporting. We aimed to evaluate which of these factors are associated with 3 reflux symptom severity outcomes (ie, Total Reflux, Heartburn, and Sleep Disturbance) through a traditional statistical and a complementary machine-learning approach.
Consecutive adult patients with refractory heartburn/regurgitation symptoms underwent standard 24-hour pH-impedance monitoring and completed questionnaires assessing past and current gastrointestinal and psychological health. In the traditional statistical approach, hierarchical general linear models assessed relationships of psychological and physiological variables (eg, total number of reflux episodes) with reflux severity scores. Mediation analyses further assessed pathways between relevant variables. In the machine-learning approach, all psychological and physiological variables were entered into 11 different models and cross-validated model performance was compared among the different models to select the best model.
Three hundred ninety-three participants (mean [SD] age, 48.5 [14.1] years; 60% were female) were included. General psychological functioning emerged as an important variable in the traditional statistical approach, as it was significantly associated with all 3 outcomes and mediated the relationship between childhood trauma and both Total Reflux and Heartburn Severity. In the machine-learning analyses, general psychological variables (eg, depressive symptoms) were most important for Total Reflux and Sleep Disturbance outcomes, and symptom-specific variables, like visceral anxiety, were more influential for Heartburn Severity. Physiological variables were not significant contributors to reflux symptom severity outcomes in our sample across reflux classifications and statistical methodology.
Psychological processes, both general and symptom-specific, should be considered as another important factor within the multifactorial processes that impact reflux symptom severity reporting across the reflux spectrum.
已发现生理和心理因素会影响食管症状报告。我们旨在通过传统统计学方法和辅助机器学习方法评估这些因素中哪些与3种反流症状严重程度结果(即总反流、烧心和睡眠障碍)相关。
连续纳入有难治性烧心/反流症状的成年患者,进行标准的24小时pH阻抗监测,并完成评估过去和当前胃肠道及心理健康状况的问卷。在传统统计学方法中,分层一般线性模型评估心理和生理变量(如反流发作总数)与反流严重程度评分之间的关系。中介分析进一步评估相关变量之间的途径。在机器学习方法中,将所有心理和生理变量输入11种不同模型,并比较不同模型之间的交叉验证模型性能,以选择最佳模型。
纳入了393名参与者(平均[标准差]年龄为48.5[14.1]岁;60%为女性)。在传统统计学方法中,一般心理功能成为一个重要变量,因为它与所有3种结果均显著相关,并介导了童年创伤与总反流和烧心严重程度之间的关系。在机器学习分析中,一般心理变量(如抑郁症状)对总反流和睡眠障碍结果最为重要,而特定症状变量,如内脏焦虑,对烧心严重程度的影响更大。在我们的样本中,无论反流分类和统计方法如何,生理变量对反流症状严重程度结果的影响均不显著。
在影响整个反流谱反流症状严重程度报告的多因素过程中,应将一般和特定症状的心理过程视为另一个重要因素。