Extraesophageal Associations of Gastroesophageal Reflux Disease in Children Without Neurologic Defects

Extraesophageal Associations of Gastroesophageal Reflux Disease in Children Without Neurologic Defects El-Serag HB, Gilger M, Kuebeler M, Rabeneck L. Gastroenterology 2001;121:1294–9. Summary: Most published studies that link gastroesophageal reflux disease (GERD) to respiratory diseases have been limited by small size, selection bias, lack of controls, narrow focus on a single respiratory problem, and the confounding effects of including patients with neurologic disease or congenital esophageal anomalies. The authors of this study (Gastroenterol 2001;212:1294–9) attempted to avoid these deficiencies. Using Texas Children's Hospital's administrative database, which codes all inpatient and outpatient encounters with the International Classification of Disease and Current Physician's Terminology systems, the authors undertook a large “case-control” study of the association between GERD and respiratory disease. The independent variable was GERD. The study included 1,980 children between ages 2 and 18 years with the diagnosis of GERD (identified by five codes: “esophageal reflux,” “esophagitis unspecified,” “reflux esophagitis,” “other esophagitis,” and “stricture and stenosis of esophagus”) seen from October 1996 to October 2000. For patients with multiple entries in the database, only the first was counted. Children with cerebral palsy, mental retardation, and any congenital tracheoesophageal anomalies, including stenosis, were excluded. The inclusion and exclusion criteria for controls were the same as for the subjects, except that children with GERD, were not included. Four controls were selected for every case (7,920 children), matching the controls to cases by the year of entry to control for possible changes in coding practices during the study. The dependent variables included upper and lower respiratory disease (sinusitis, laryngitis, otitis media, asthma, pneumonia, bronchiectasis). Potential risk factors for GERD such as cystic fibrosis, scleroderma, and morbid obesity, and diagnostic and therapeutic procedures, such as upper endoscopy, esophageal dilation, laryngoscopy, bronchoscopy, and sinus surgery, were identified. The authors also examined demographics, comorbid illness, and frequency of follow-up visits to further analyze potential confounding differences between cases and controls. Using univariate analysis, the authors tested differences between cases and controls as to the dependent variables. They followed this analysis by multivariable logistic regression analyses to control for differences between cases and controls in demographics and potential risk factors. By univariate analysis, the GERD cases had significantly increased frequency of each of the respiratory diagnoses except otitis media (which was significantly inversely related to GERD). The patients with GERD had, as expected, an increased frequency of risk factors for GERD and had experienced more diagnostic and therapeutic procedures. However, comorbid illnesses and follow-up visits were similar in cases and controls. Oddly, each of the demographic variables examined differed significantly between cases and controls, the children with GERD being older, and more frequently female and white. The multivariable logistic regression analysis was undertaken to control for these demographic differences. Further analysis also excluded patients with cystic fibrosis to avoid confounding the analysis with patients who had a unique reason for increased frequency of upper and lower respiratory disorders. The significant association between GERD and (non-otitis) upper and lower respiratory disorders persisted even after these adjustments. Comment: Susan R. Orenstein M.D. Professor and Chief of Pediatric Gastroenterology University of Pittsburgh School of Medicine Children's Hospital of Pittsburgh Pittsburgh, Pennsylvania The clinical literature that support an association between respiratory disorders and GERD progress through four types: case reports, epidemiologic studies, therapeutic studies, and (uncommonly) experimental studies. Several decades ago, the first case reports of children with both GERD and respiratory disorders were published (Pediatr Radiol 1973;1:156–60;Chest 1983;84:301–2). Epidemiologic studies, often small-scale and beset by selection bias because of using referred patients and lacking controls for various confounding variables, followed. Some of these examined the prevalence of respiratory disease in children with GERD (Arch Dis Child 1960;35:481–3;Ann Otol Rhinol Laryngol 1980;89:450–3), and others studied the prevalence of GERD in children with respiratory disease (J Pediatr Gastroenterol Nutr 1991;13:16–22;Ann Otol Rhinol Laryngol 1997;169(suppl):1–16;Laryngoscope 2000;110:1560–2). Therapeutic studies, in which treatment of GERD is examined for its impact on coexisting respiratory disorders, are potentially helpful (Ann Otol Rhinol Laryngol 2000;109:18–23;Arch Otolaryngol Head Neck Surg 2001;127:511–4). However, in practice the “fatal flaws” of retrospective design, lack of placebo control, and lack of masked investigators nearly always have biased such studies. The current study is epidemiologic. The authors have taken pains to avoid some of the problems of previous epidemiologic studies, specifically the lack of controls and selection bias. The large size of the study groups also has produced precise estimates with narrow confidence intervals. The incorporation of controls for confounding variables is a notable strength of the study. The authors prospectively matched cases and controls by year of diagnosis to control for disease-coding practices. Finding that demographics actually did differ between their cases and controls, the authors took pains to control for this in the subsequent multivariable analysis. To examine for other confounders unrelated to GERD, they examined cases and controls for presumably unassociated comorbid conditions and found no differences. Because the study is retrospective and code based, it should be unaffected by investigator or recall bias. Being code related, the disease definitions should be quite precise. Precision is increased by the exclusion of forms of GERD that may differ from the general childhood form they focus on, such as those associated with neurologic disease, congenital esophageal anomalies, and (excluded during their analysis) cystic fibrosis. The authors acknowledge several potential limitations of their study. Care that patients and controls received elsewhere was not assessed. Because the study was hospital based, it likely dealt with more, and more severe, respiratory disease and GERD than is prevalent in the general population. One of the challenges in studies like this is the bias among the clinicians assigning the International Classification of Disease codes. Many of us have accepted the notion that GERD and some respiratory disorders are related. As a consequence, we may explore respiratory symptoms in more detail in a patient with GERD than we would if we did not suspect GERD. Likewise, otolaryngologists, pulmonologists, and other clinicians who deal with upper and lower respiratory diseases are now identifying GERD as a problem in many of their patients, with or without objective testing. Because the diagnosis of GERD underlies the assignment of children in this study to patient or control groups, accuracy and lack of bias in this diagnosis is crucial to our ability to generalize from this study's data. Another challenge related to the diagnosis of GERD is the lack of precision in the codes themselves. Although the study maintained rigorous precision and avoided bias by defining variables on the basis of codes, the original assignment of codes might have been quite imprecise in practice. This problem affects the dependent variables, such as the respiratory disease, but even more importantly affects the diagnosis of GERD itself. A further challenge emanates from what is also one of the study's greatest strengths—its huge size. The power provided by such a large study enables the detection of statistically significant differences that may be clinically insignificant. For example, the older age of the GERD patients (mean, 9.16 years) compared with controls (mean, 8.64 years) was statistically significant (P < 0.003). However, it is difficult to imagine that this difference is clinically significant. Several additional pieces of information would have been interesting. The decreased incidence of otitis was contrary to the authors' hypotheses. Would inclusion of children with mild or occasional cases of otitis have confounded the otitis data? It would have been interesting to select and evaluate the subgroup of children who had severe enough middle ear disease to require tympanostomy tubes, for example. One might also question whether the International Classification of Disease code 530.19 for “other esophagitis” should have been included in the GERD diagnosis group at all. Could this code have been used to identify unrelated conditions such as allergic or eosinophilic esophagitis? Exploration of this code group in more detail would have been helpful. This large, powerful, epidemiologic study has approached the issues around the association between GERD and respiratory disease better than any similar study to date. It has raised an important question regarding a negative association of GERD with otitis media. As the authors note, “further studies are needed to examine whether a cause–effect relationship exists between GERD and these potential extraesophageal disorders.” It is time to turn to the difficult but important work of clarifying these cause–effect relationships. If the authors' institution diagnoses approximately 500 children a year with GERD, of whom at least 13% have respiratory disorders, it should not take long to do a prospective, placebo-controlled, double-blind therapeutic trial, using the powerful modern pharmacotherapies for GERD. Although such a study would delay GERD therapy for children assigned to placebo, it would provide the hard data needed to justify antireflux therapy in children with respiratory diseases.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.