NICU indicates neonatal intensive care unit; PHIS, Pediatric Health Information System.
Shown are each hospital’s volume of gastrostomies in infants with neurological impairment and the percentage who underwent concomitant fundoplication.
Effectiveness of Fundoplication at the Time of Gastrostomy in Infants With Neurological Impairment
Barnhart DC, Hall M, Mahant S, et al. Effectiveness of fundoplication at the time of gastrostomy in infants with neurological impairment. JAMA Pediatr. Published online August 5, 2013.
Barnhart DC, Hall M, Mahant S, Goldin AB, Berry JG, Faix RG, Dean JM, Srivastava R. Effectiveness of Fundoplication at the Time of Gastrostomy in Infants With Neurological Impairment. JAMA Pediatr. 2013;167(10):911-918. doi:10.1001/jamapediatrics.2013.334
Copyright 2013 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Gastrostomy tube (GT) placement is the most common gastrointestinal operation performed on neonates. Concomitant fundoplication is used variably to prevent complications of gastroesophageal reflux, but its effectiveness is unproven.
To compare the effect of fundoplication at the time of GT placement vs GT placement alone on subsequent reflux-related hospitalizations in infants with neurological impairment.
Design, Setting, and Participants
Retrospective, observational cohort study, defined by birth between January 1, 2005, and December 31, 2010, at 42 children’s hospitals in the United States, with a 1-year follow-up period among 4163 infants with neurological impairment who underwent GT placement with or without fundoplication during their neonatal intensive care unit stay.
Fundoplication and GT placement vs GT placement alone.
Main Outcomes and Measures
One-year postprocedural reflux-related hospitalization rates, defined as hospitalization for asthma, mechanical ventilation, gastroesophageal reflux disease, and aspiration or other types of pneumonia. Propensity to undergo concomitant fundoplication was modeled using demographics, prior procedures (tracheostomy and mechanical ventilation), and prior diagnoses (eg, pneumonia, gastroesophageal reflux disease, and other comorbidities).
Overall, 4163 of 42 796 infants (9.7%) with neurological impairment admitted to the neonatal intensive care unit underwent GT placement alone or with fundoplication. Infants who concomitantly underwent fundoplication had more reflux-related hospitalizations during the first year than those who underwent GT placement alone (mean, 1.02; 95% CI, 0.93-1.10 vs mean, 0.92; 95% CI, 0.91-1.00). Of 1404 infants who underwent fundoplication, 1027 (73.1%) were matched based on propensity scores. The mean difference of the matched cohort for any reflux-related hospitalizations was −0.05 (95% CI, −0.20 to 0.15) per year.
Conclusions and Relevance
Infants with neurological impairment who underwent fundoplication at the time of GT placement did not have a reduced rate of reflux-related hospitalizations during the first year compared with those who underwent GT placement alone, despite propensity score matching. This may be due to a lack of effectiveness of fundoplication in preventing these complications or due to differences in the patient groups that were inadequately accounted for in the matching.
Infants and children with neurological impairment account for a large percentage of pediatric inpatient resource use throughout the United States.1Quiz Ref IDThey often experience dysfunctional swallowing and gastroesophageal reflux disease (GERD). Both of these conditions place them at high risk for aspiration pneumonia,2- 4 which is the most common cause of death in children with severe neurological impairment.5- 8 Long-term feeding access, typically a gastrostomy tube (GT), is often required to provide optimal nutrition for infants with neurological impairment who have dysfunctional swallowing. At the time of GT placement, symptoms of GERD are often treated medically with antireflux medications or surgically with concomitant fundoplication.
Rates of GT placement and fundoplication operations in children with neurological impairment have increased by 25% during the past decade.1 Historically, fundoplication has been reported to be the third most common procedure performed by pediatric surgeons.9 Approximately 40% of pediatric fundoplications in the United States are performed in children with neurological impairment.10 Our clinical experience from various institutions has led us to observe institutional variability in the use of concomitant fundoplication. This stems from uncertainty in evaluating GERD in infants with neurological impairment and challenges in defining medically refractory GERD, as well as a lack of objective outcome data.
Quiz Ref IDThere are several goals of the concomitant fundoplication. The primary objective is to protect the infant with neurological impairment and GERD from aspiration of refluxed gastric contents, which causes acute and chronic pulmonary sequelae. Secondary objectives include reducing symptoms, optimizing nutrition, and improving the child and caregiver health-related quality of life, as well as decreasing future GERD-related hospitalizations.
Two large multicenter studies11,12 examined reflux-related hospitalizations using administrative data and demonstrated reductions following fundoplication in children with neurological impairment compared with the prefundoplication period. We undertook this study to test the hypothesis that infants with neurological impairment who underwent fundoplication at the time of GT placement would have fewer reflux-related hospitalizations than those who underwent GT placement alone.
Quiz Ref IDData for this retrospective, observational cohort study were obtained from the Pediatric Health Information System administrative database developed by the Children’s Hospital Association (Overland Park, Kansas). The database contains inpatient demographic, diagnostic, and procedural data from 42 not-for-profit, freestanding children’s hospitals in the United States. The data warehouse function for the Pediatric Health Information System database is managed by Thomson Reuters (Evanston, Illinois). Data are subjected to several reliability and validity checks and are processed into data quality reports. Patients are tracked for repeat hospitalizations at the same children’s hospital using a unique identifier. This study was approved by the institutional review boards of Children’s Hospital, Boston, Massachusetts, and the University of Utah Health Sciences Center and Primary Children’s Medical Center, Salt Lake City.
The cohort was created using the following 4 inclusion criteria: (1) birth date during the study enrollment period, Month X, 2005, to Month X, 2010; (2) admission to a neonatal intensive care unit within the first 90 days of birth; (3) an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9 CM) diagnostic code for neurological impairment; and (4) an ICD-9 CM code for placement of a GT (procedure codes 43.0, 43.19, 43.11, 43.1, and 97.02 or diagnosis codes V44.1, V55.1, and 536.4x) during the initial neonatal intensive care unit admission. The cohort was subdivided into 2 groups for analysis based on whether a fundoplication (codes 44.66 and 44.67) was performed at the time of initial GT placement. The cohort was followed up for 1 year after GT placement to capture subsequent reflux-related hospitalizations. Because of difficulties in defining and anticipated inconsistencies in diagnosing pathologic gastroesophageal reflux in neonates, the presence or absence of GERD was not used as an inclusion criterion.
Neurological impairment was defined as static or progressive, central or peripheral neurological diagnoses associated with chronic functional or intellectual impairment. The ICD-9 CM codes that have been used to operationalize this definition (eg, cerebral palsy, hydrocephalus, leukodystrophy, and epilepsy) have been previously published.1 To assure that there were no children with neurological impairment in whom this was not initially recognized but was subsequently diagnosed, the occurrence of neurological impairment codes during the first 3 years was compared with that at the time of initial discharge. The cohort of infants with neurological impairment diagnosed at the index admission was used as the study group because few infants developed previously unrecognized impairment and this group more accurately represents the clinical scenario in which decisions about fundoplication are made.
Demographic characteristics included sex, birth weight, and disposition (home, died, or other). Race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, Asian, or other) was also recorded.
Complex chronic conditions were defined as renal, metabolic, respiratory, gastrointestinal, malignant neoplasm, congenital or genetic condition, hematological or immunological, and cardiovascular diagnoses expected to last longer than 12 months that involve several different organ systems or one organ system severe enough to require specialty pediatric care and hospitalization.13,14 We did not use the neuromuscular category because the study cohort was assembled with neurological impairment as a criterion for inclusion. Areas of special interest for GERD complications are anomalies in the upper airway and requirement for tracheostomy. The procedural and diagnostic codes used to identify these conditions have been previously published.12
The use of medications to treat GERD during the index hospitalization was summarized for each patient. These were identified from pharmacy billing data. Histamine2 blockers included cimetidine hydrochloride, famotidine hydrochloride, nizatidine, and ranitidine. Proton pump inhibitors included esomepraozle, lansoprazole, omeprazole, and pantoprazole sodium. Motility agents included metoclopramide hydrochloride and oral erythromycin ethylsuccinate. The use of cisapride was evaluated but was rarely documented. Sucralfate was the only buffering or protective medication used.
The extent of diagnostic testing before GT placement was assessed for each patient with billing data. Contrast radiographic tests included cinegraphic esophagography, esophageal fluoroscopy, and upper gastrointestinal contrast studies with and without small-bowel follow-through studies. Endoscopy included esophagoscopy and esophagogastroduodenoscopy with or without biopsies. The use of prolonged pH probe monitoring studies was summarized, but impedance monitoring was not commonly used during the time frame of the study and did not have a separate billing code. The use of manometry and nuclear medicine was assessed. Both of these modalities were rarely used (<1% of patients) and were excluded from further analyses. The extent of diagnostic evaluation for each patient was summarized as fluoroscopy, pH monitoring, and endoscopy.
Reflux-related hospitalizations included 2 broad categories of complications of GERD, namely, gastrointestinal and pulmonary. Gastrointestinal complications included hospitalizations with a diagnosis of esophagitis or GERD. These diagnoses were included because they were considered evidence of failure of therapy for GERD. Pulmonary complications included pneumonia, aspiration pneumonia, and requirement for mechanical ventilation. These outcomes were selected because pulmonary complications represent a major cause of morbidity and mortality in children with neurological impairment, as well a common cause of admission. All pneumonias were included rather than limiting to only aspiration pneumonias in recognition of the imprecision in clinically distinguishing these entities. Requirement for mechanical ventilation was included because it is an indicator of the severity of pulmonary disease. The ICD-9 CM codes used to identify these diagnoses have been previously published.12
Demographic characteristics were summarized using frequencies and percentages for categorical variables and means (SDs) for continuous variables. Comparisons between the 2 patient groups (GT placement only vs fundoplication and GT placement) were made with χ2 test or t test as appropriate.
Using Poisson regression, unadjusted counts of admissions for reflux-related causes during a follow-up year were compared between those who underwent GT placement only and those who underwent concomitant fundoplication. The 2 groups were compared for individual diagnoses, all pulmonary diagnoses, all gastrointestinal diagnoses, and all reflux-related hospitalizations. Outcomes were truncated at a 1-year period.
To account for potential differences between infants who underwent GT placement only and those who underwent concomitant fundoplication, propensity scores were used to match patients. To ensure that propensity scores were an appropriate method, we examined the center volume and the percentage of infants with neurological impairment who received a concomitant fundoplication to assess if there was an overwhelming center bias in performing either of these procedures. The propensity to undergo concomitant fundoplication was modeled using the following variables during the index admission: sex, birth weight, race/ethnicity, age at admission, upper airway anomaly, type of neurological impairment, the extent of preoperative evaluation, the presence of complex chronic conditions, diagnosis of aspiration pneumonia or pneumonia, tracheostomy or requirement for mechanical ventilation, and diagnosis of GERD (ICD-9-CM codes 530.11 and 5301.81) or the use of GERD medications. Each patient who underwent fundoplication was matched with a patient who underwent GT placement only using nearest-neighbor matching with a caliper set at one-quarter of the SD of the logit of the propensity score.15 Pairs were forced to match on the diagnosis of GERD. Results for the propensity score–matched analyses are expressed as differences (95% CIs) in reflux-related hospitalizations for the matched pairs, and statistical significance was determined using Wilcoxon signed rank test.
The results of the cohort identification are summarized in Figure 1. The demographic characteristics of the 2 study groups are given in Table 1.
Complex chronic conditions were frequent in these infants with neurological impairment who underwent GT placement (Table 1). The mean (SD) number of complex chronic conditions per infant was 1.8 (1.2). Renal, metabolic, gastrointestinal, cardiovascular, and congenital or genetic conditions were more frequent in those who underwent GT placement only. Those who underwent concomitant fundoplication were more likely to have had aspiration pneumonia or pneumonia and to have undergone a tracheostomy and required mechanical ventilation during the initial neonatal intensive care unit hospitalization. As might be anticipated, this group was also more likely to carry the diagnosis of GERD, although 27.0% who underwent fundoplication did not have this diagnosis and 26.4% who underwent GT placement alone did.
GERD medications were frequently used in both groups (Table 2), with 73.8% of infants being treated with at least 1 of these medications. For each class of medication, there was significantly higher use prevalence among the concomitant fundoplication group.
The use of preoperative diagnostic testing is summarized in Table 2. Most infants who underwent GT placement underwent contrast radiographic examination only. In total, 9.4% of infants who underwent fundoplication had pH monitoring, despite the fact that this study is considered the sole quantitative assessment of gastroesophageal reflux. Similarly, 4.3% of infants who had a fundoplication underwent endoscopic evaluation.
Figure 2 shows each hospital’s volume of gastrostomies in infants with neurological impairment and the percentage who underwent concomitant fundoplication. Although some hospitals preferentially perform concomitant fundoplication in these infants (the left side of the figure) or GT placement alone (the right side of the figure), many centers seem to be selective in choosing between these 2 procedures.
For the entire cohort of 4163 infants, the mean number of reflux-related hospitalizations during the first year after discharge from the neonatal intensive care unit was 0.95 (95% CI, 0.91-1.00) and was greater in those who underwent fundoplication. The unadjusted mean numbers of admissions for these reflux-related hospitalizations based on the cause of admission are given in Table 3. In total, 4.7% of infants in the GT-only group underwent a subsequent fundoplication within 12 months.
Propensity-based matching was achieved for 1027 of 1404 infants (73.1%) who underwent fundoplication. This matching forced concurrence between pairs for the diagnosis of gastroesophageal reflux and gastrointestinal complex chronic conditions. The characteristics of these matched pairs are summarized in Table 4. After matching, no differences were observed between the 2 groups in any of the characteristics assessed except race/ethnicity and requirement for mechanical ventilation during the first admission. Race/ethnicity did not show an obvious pattern. The use of mechanical ventilation was 4.9% higher in the GT-only group.
After matching, no differences were found between the 2 groups in terms of medication use (overall or for specific medications). Comparison between the GT-only group vs the concomitant fundoplication group demonstrated comparable use of fluoroscopy (84.4% vs 84.2%, P = .90) and pH monitoring (10.0% vs 9.1%, P = .45). Endoscopy was performed more frequently in the GT-only group (12.7% vs 5.3%, P < .001).
The mean differences in admission rates between the matched pairs are as follows: overall (−0.05; 95% CI, −0.20 to 0.15), pneumonia (0.04; 95% CI, −0.02 to 0.09), esosphagitis (0.00; 95% CI, −0.01 to 0.00), aspiration pneumonia (−0.02; 95% CI, −0.05 to 0.02), GERD (−0.14; 95% CI, −0.26 to −0.02), and requirement for mechanical ventilation (0.07; 95% CI, 0.00 to 0.14). In interpreting these paired differences, negative values indicate a lower risk of hospitalization after concomitant fundoplication, while positive values indicate less risk after GT placement only. No difference was observed between the matched pairs in the overall reflux-related hospitalization rates. Similarly, no differences were found between these groups in the cause of the reflux-related hospitalizations.
Quiz Ref IDIn this multicenter retrospective study of infants with neurological impairment who underwent GT placement alone or concomitant fundoplication, there appeared to be a higher rate of reflux-related hospitalizations among those who underwent both procedures. However, differences were observed in the associated comorbidities between those who underwent concomitant fundoplication and those who did not. Propensity score–matched analysis was performed to control for these differences. Although this matching eliminated the increased rate of admission in the fundoplication group, it demonstrated no therapeutic benefit of fundoplication in terms of reflux-related admissions. In those who underwent fundoplication and those who did not, the use of GERD medications was common, and preoperative GERD diagnostic evaluation was limited. The overall proportion of infants with neurological impairment undergoing fundoplication at the time of GT placement was variable across centers.
Two large multicenter studies examined reflux-related hospitalizations using administrative data and lend some bearing on our findings. The first study, by Goldin et al,11 examined children cared for throughout the state of Washington and found that reflux-related hospitalization rates decreased following fundoplication in a group of approximately 500 children with developmental delay who had a reflux-related hospitalization before surgery. However, the study reported no change in aspiration pneumonia rates following fundoplication. The second study,12 which used similar methods, examined age-specific models for reflux-related hospitalizations in infants with neurological impairment and also found overall decreased admissions in infants younger than 12 months and reductions in specific reflux-related hospitalization rates (ie, GERD, aspiration pneumonia, and requirement for mechanical ventilation) but not pneumonia (nonaspiration) hospitalizations following a first fundoplication.
In contrast to these studies, we found no difference in overall reflux-related hospitalization rates or in any of the specific reflux-related causes of hospitalization in infants having neurological impairment who underwent a concomitant fundoplication compared with infants having neurological impairment who received GT placement alone. It is important to note that the present study seeks to answer a different question than the prior studies. The earlier studies examined whether fundoplication was effective at decreasing reflux-related hospitalizations in children who had preceding admissions related to reflux. In contrast, this study tested whether there was a benefit of concomitant fundoplication in infants with neurological impairment undergoing GT placement. In essence, these operations may be performed prophylactically in infants who may or may not have experienced complications of gastroesophageal reflux. In addition, methodological differences between the studies could help explain these divergent findings. Unlike the other studies, which used a prepost study design on the same cohort of patients, we used a matched comparison group. This matching was done to reduce the possibility that the infants who underwent fundoplication had a higher baseline risk of reflux-related hospitalization, which could potentially mask the benefit of fundoplication. In observational studies, propensity matching should maximize the possibility of detecting a therapeutic effect of fundoplication if one exists.
The strengths of this study include the multicenter nature of the data source, which provided many patients and increased the generalizability of the results. Our outcome was objective because readmission for these diagnoses is clearly defined, and we studied children who were cared for in a neonatal intensive care unit since birth, reducing potential selection bias.
There are limitations to this study. Although our propensity score model adjusted for measured confounders, unmeasured confounding likely remains. Clinical symptoms, including dysfunctional swallowing and intolerance of gastric feedings, likely influence the decision to perform a concomitant fundoplication; however, these were unavailable in the database. The severity of neurological impairment, although difficult to measure, might be another unmeasured confounder. The effect of these potential confounders cannot be measured. Some reflux-related hospitalizations may have occurred at hospitals not captured in the database. However, we believe that this effect is likely limited. Because these patients were all infants with neurological impairment, care is typically rendered at a children’s hospital, and definitive admission to general hospitals would be infrequent. Another limitation is that we had no results of GERD testing. However, minimal numbers of pH probe studies or endoscopies were performed in this population. Postoperative clinical variables that might influence ongoing primary aspiration (such as continuation of oral feedings) were unable to be obtained. Mechanical ventilation may be used for some elective surgery admissions as opposed to true reflux-related admissions. We were unable to accurately identify placement of gastrojejunal feeding tubes (an alternative antireflux procedure) in the Pediatric Health Information System database. Conversion to gastrojejunal feeding tubes may have occurred differently between the 2 groups and may have confounded the results. Finally, we were unable to capture complications of fundoplication that did not require readmission such as recurrent gastroesophageal reflux, gas bloat, and dumping syndrome. These are well described in the literature and should be considered in an overall risk-benefit assessment.
We failed to prove our hypothesis that infants with neurological impairment who underwent fundoplication at the time of gastrostomy would have fewer reflux-related hospitalizations than those who underwent GT placement alone. This hypothesis was founded on the notion that if fundoplication had a strong effect on the occurrence of reflux-related hospitalizations, this effect would not be masked despite uncontrolled confounders. This was not the case. Either our propensity score model failed to account for factors that are important in determining when a concomitant fundoplication is performed (such as intolerance of gastric feedings), or there is no clear treatment effect on the risk for future reflux-related hospitalizations. Only adequately powered multicenter prospective trials that collect broader clinical data will allow this question to be definitively answered. These trials should be pragmatic and consider how the diagnosis of GERD is made, especially given how infrequently invasive GERD testing occurs in this group of children. Pragmatic trials will have a higher likelihood of influencing practice if the study design closely matches current clinical practice.
Quiz Ref IDIn conclusion, this large multicenter observational cohort study with propensity score matching of infants with neurological impairment failed to show a difference in reflux-related hospitalizations in those who underwent GT placement alone compared with concomitant fundoplication. Either no difference exists, or prospective studies that include further clinical factors will be necessary to definitively demonstrate a difference. Until such a comparative effectiveness study is performed, the present variable practice is likely to continue.
Accepted for Publication: January 17, 2013.
Corresponding Author: Douglas C. Barnhart, MD, MSPH, Department of Surgery, Primary Children’s Medical Center, University of Utah, 100 N Mario Capecchi Dr, Ste 2600, Salt Lake City, UT 84113 (Douglas.Barnhart@imail.org)
Published Online: August 5, 2013. doi:10.1001/jamapediatrics.2013.334.
Author Contributions:Study concept and design: Barnhart, Hall, Mahant, Goldin, Berry, Dean, Srivastava.
Acquisition of data: Barnhart, Hall, Goldin.
Analysis and interpretation of data: Barnhart, Hall, Mahant, Goldin, Berry, Faix.
Drafting of the manuscript: Barnhart, Hall, Goldin, Berry, Srivastava.
Critical revision of the manuscript for important intellectual content: Barnhart, Mahant, Goldin, Berry, Faix, Dean, Srivastava.
Statistical analysis: Barnhart, Hall, Goldin, Berry.
Administrative, technical, and material support: Barnhart, Goldin, Faix, Dean.
Study supervision: Barnhart, Goldin, Berry, Dean.
Conflict of Interest Disclosures: None reported.