Testing the Association Between Gastroesophageal Reflux and Apnea in Infants

Background: There is controversy regarding the association between gastroesophageal reflux and apnea. The controversy is further confused by the variety of ways reflux is detected, the varying methods of data analysis and the heterogeneity of subjects studied.

Objectives: To determine if apnea is associated with reflux and to determine whether the association differs with acid or non acid reflux.

Methods: We prospectively evaluated infants with apparent life-threatening events or apnea by simultaneous pneumography, esophageal pH and multichannel intraluminal impedance monitoring. Apneic events and reflux frequency, duration, acidity and height were recorded. Apneic events were considered to be temporally linked to gastroesophageal reflux when they occurred during a gastroesophageal reflux episode or within 5 minutes after clearance of the refluxate. Data described with frequency tables, scatterplots and time series graphs were analyzed by regression analysis and χ² testing.

Results: Twenty-five infants (10 males) aged 1 to 19 months were studied for up to 24 hours. Of 527 total apneic episodes, only 80 (15.2%) were temporally linked with gastroesophageal reflux: 37 (7.0%) with acid reflux and 43 (8.2%) with non-acid reflux. Scatterplots revealed no significant correlation between apnea and frequency or duration of reflux episodes. Individual χ² analyses within subjects found limited associations between reflux and apnea. Regression analyses showed a significant association between apnea and reflux in 4 of 25 subjects.

Conclusions: We found little evidence for an association between apnea and total reflux, acid reflux or non-acid reflux. There was no difference between acid gastroesophageal reflux and non-acid gastroesophageal reflux in the frequency association with apnea. Either a χ² statistic for each subject or R² value computed from a lagged regression model for each subject can be used as an index of association in patient evaluation.

INTRODUCTION

A temporal association between gastroesophageal reflux (GER) and apnea in infants has frequently been suggested. However, the results and interpretations of clinical studies have been conflicting (^1-16). Major methodological differences preclude meaningful comparisons between studies. These differences include the criteria for subject selection, the methods used to detect GER, the length of observations and the criteria for defining temporal association between reflux and apnea.

Most previous studies have used esophageal pH monitoring to identify GER. This procedure has significant limitations because not all reflux episodes are detected by pH probe (^17,18). Multichannel intraluminal impedance (MII) monitoring is being investigated for its usefulness in detecting reflux independent of pH (^19,20). MII tracks the movement of boluses through the esophagus by measuring changes in electrical conductivity. Conductivity of an empty esophagus is almost zero whereas conductivity of formula, saliva or gastric contents is high. Electrical impedance, the inverse of conductivity, decreases when high conductivity boluses pass through the esophagus and increases during passage of gas boluses. Combining MII and pH measurements permits detection of both acid (AGER) and nonacid (NAGER) GER (^21,22).

Three approaches have been used to explore an association between GER and apnea. The most common has been a temporal linking strategy involving simultaneous measurement of GER and apnea, followed by analysis examining the linkage within an arbitrary time between a reflux episode and apnea (¹⁰). Another approach is the symptom index, which tests whether apnea occurs more frequently during reflux episodes than during control periods. Wenzl et al. (¹⁰) found that the mean time spent apneic during GER was greater than the mean time spent apneic during reflux-free time. Using a binomial statistical model, Menon et al. (⁹) demonstrated that apnea occurred more frequently during GER than during reflux-free periods. The clinical value of such calculations is not clear. A third, less frequently used approach is construction of a dose-response curve. This analysis uses coordinate patterns of two scatterplots: the number of apneic events plotted as a function of the number of GER episodes and the number of apneic events plotted as a function of the duration of GER. Using this approach, Barrington et al. (³) did not find a statistically significant correlation between GER and apnea. The aims of our investigation were to determine if apnea is truly associated with reflux and, if so, to determine whether the association is different with acid or nonacid reflux, and whether an apnea index calculated for each patient is a clinically relevant way of analyzing data.

PATIENTS AND MATERIALS

Study Population

Twenty-five patients (10 male) aged between 1 and 19 months (median age, 4 months) were studied. Eleven were born prematurely (gestational age <38 weeks). Median corrected age at the time of study was 4.5 months and postconceptional age at the time of study (TOS) was >38 weeks in 24 of 25 patients. One patient was 36 postconception weeks (Table 1). Subjects had been referred to our facility after experiencing an apparent life-threatening event. GER had been suspected clinically as the etiology of the apparent life-threatening event. We defined apparent life-threatening events as episodes including some combination of apnea (central or obstructive), color change (cyanosis, pallor, or plethora), change in muscle tone (usually hypotonia), choking or gagging.

Table 1

Study Protocol

Subjects were monitored by combined MII/pH and pneumography for 24 hours. Acid suppressing medications were discontinued 5-7 days before study. Prokinetics were discontinued 3 days before study. Patients were fed their regular diet 4 to 6 times during the 24-hour period. The impedance catheter (PRZ-61B00004, Sandhill Scientific, Denver, Colorado) was passed transnasally following a 3-hour fast. We used fluoroscopy to locate the lower esophageal sphincter and its distance from the nares. Then 87% of the total distance was calculated and the catheter was positioned with the pH electrode 13% of esophageal length above the lower esophageal sphincter. Positioning was confirmed by radiography.

The combined impedance and pH (MII-pH) catheter has an outer diameter of 2.13 mm (6.4-French). There are 7 impedance sensors in the form of 4-mm cylindrical ring electrodes placed 1.5 cm apart starting at the distal tip of the catheter. The segment between each pair of sensors (impedance sensor spacing) corresponds to 1 impedance channel. Thus, the catheter with 7 impedance sensors has 6 channels. The pH electrode is positioned between the two distal impedance sensors.

Apnea was detected by pneumography records which included nasal air flow, oxygen saturation, heart rate, electrocardiogram and chest movements. The combined MII/pH measurements were synchronized with pneumography data using the Time Synchronization Device™ (Sandhill, Englwood, CO).

DEFINITIONS

Apnea

In this study, an apnea event was defined as a decrease in air flow to 10% of baseline lasting 10 seconds or more. Apneic events with simultaneous cessation of oronasal flow and chest movement were defined as central apnea. Events involving cessation of oronasal flow but not chest movement were defined as obstructive apnea. Mixed apnea was defined as a pause in breathing where cessation of oronasal flow was later followed by cessation of chest movement. Any pause in the breathing, whether central, obstructive or mixed, was considered an apneic event if it lasted for ≥10 seconds. An apneic event was considered temporally linked with a GER episode if the apneic event occurred either during the episode or within the ensuing 5 minutes.

Reflux Episodes

Impedance was measured in ohms (Ω). A reflux episode (GER) was defined as a fall in impedance of ≥50% of basal which progressed retrograde over two or more of the most distal channels. An episode was considered acid (AGER) when the pH of the esophagus decreased and remained <4 for at least 5 seconds during the episode. An episode was considered nonacid (NAGER) when the pH increased, remained unchanged or decreased by less than 1 pH unit while remaining <4 (¹⁹).

Extent of Reflux Migration

The height reached by the refluxate was considered limited to the distal esophagus if it was confined to the two most distal impedance channels (≤3 cm above the pH electrode). The reflux was considered intermediate in extent if the refluxate migrated to either or both of the middle two but not the proximal channels. The refluxate was considered proximal if it reached either or both of the most proximal channels up to 9 cm above the pH electrode.

METHOD OF ANALYSIS

Analysis of the combined pneumogram-impedance studies was performed in 3 phases. In phase 1, an experienced investigator blinded to the clinical condition scored the pneumogram. In phase 2, the impedance tracing was visually inspected and analyzed without knowledge of the pneumographic results. Evaluations were limited to the time during which both impedance and pneumogram recordings occurred simultaneously. In phase 3, phase 1 and phase 2 analyses were combined to look for event correlations.

STATISTICAL ANALYSES

Statistical methods with SAS version 9.1 (SAS Inc. Cary, NC), including χ², Mantel Haenszel, Breslow-Day, and regression analyses, were used. Statistical significance was established by a P value <.05.

Temporal Linking

Apnea was linked to GER if the apneic event was preceded by GER within 5 minutes. GER events were identified as acid or nonacid and by height (no reflux, proximal, intermediate, or distal). Each 5-minute interval during the study could theoretically contain a GER event only, an apneic event only, both an apnea and a GER or neither.

Structured Time Series

In the time series analysis, the study period was divided into 5-, 10-, and 20-minute intervals and the number of times the four possible events discussed above occurred was tabulated. Separate analyses were conducted for each interval. If one or more apnea events and one or more GER events occurred within a 5-minute interval it was considered linked. In 24 hours there are 288 5-minute intervals, 144 10-minute intervals, and 72 20-minute intervals. χ² statistics were calculated for each subject as well as the overall Mantel-Haenszel statistic, which was used to test the association between apnea and GER stratified by subject. The Breslow-Day test was used to evaluate the homogeneity of the odds ratios (the consistency of the relationship over patients). The test was repeated for each of the three interval lengths (5, 10 and 20 minutes) considering acid and nonacid reflux separately.

Regression Method

A regression time series model accounting for lagged GER duration was used to test whether the association between the duration of apnea and GER might depend on events leading up in time to the apneic event. A quadratic equation was fitted to lagged GER in the 12 5-minute time periods (1 hour) before the initiation of apnea in each 5-minute period. Three simple models regressed the duration of apnea on the duration of total GER, acid GER and nonacid GER. The duration of apnea was also regressed on a single multiple regression model that included the lagged variables for acid, nonacid, and height index. The associations between the durations of apnea and GER, including lagged effects of GER duration and GER height before the apneic event, were tested with the PDLREG procedure from SAS. A polynomial model with linear and quadratic components was used to evaluate the lagged effect of GER and height during the 12 intervals (1 hour) before each apneic event. Durations of apnea and GER were Log₁₀ transformed. Because apneic events between time segments may not be independent, Durbin-Watson statistics were used to identify autocorrelation. The duration of apneic events occurred one time period before the time interval of interest was used to test and control for the autocorrelation within individual time series. R² values are reported to describe the proportion of variance accounted for by GER components. The sensitivity of the time interval approach was tested by shifting the starting time interval 2.5 minutes and reanalyzing all the models.

This study protocol was reviewed and approved by the Institutional Review Board (IRB) of Children's Hospital in Columbus, Ohio. The parents or guardians of subjects all gave written informed consent.

RESULTS

Frequency of GER

Apnea and GER were detected in all patients; 1211 episodes of GER (613 AGER and 598 NAGER) occurred with a mean frequency of 13 ± 10 (SD) per patient (range, 0 to 39) for AGER and 12 ± 7 per patient (range, 0 to 28, P = 0.88) for NAGER. Five infants had only NAGER and one had no NAGER. Figure 1 shows the frequency of AGER and NAGER per individual patient. There was no significant difference in esophageal distribution of reflux between AGER and NAGER. Refluxate reached the proximal esophagus in 84%, intermediate esophagus in 14% and distal esophagus in 3% of episodes.

Fig. 1

Apnea

Of the 527 apneic events recorded (median, 14 per patient), 224 (42.5%) were obstructive, 137 (26.0%) were central and 166 (31.5%) were mixed. Frequency of the three types was similar (P = 0.21). Mean duration of apneic events was 11.8 ± 1.3 seconds (range, 10.0 to 15.5 seconds).

Relationship Between Total Apnea and Total GER

Figures 2, 3 and 4 illustrate the relationship between the total number of GER events and the total number of GER events per subject in panels (A) and between the total duration of GER events and duration of apnea per subject in panels (B). None of the R² in any of the figures is statistically significant indicating no dose response relationship between GER and apnea.

Fig. 2

Fig. 3

Fig. 4

Temporal Linking

Only 80 of the 527 apnea events (15.2%) were temporally linked with GER, 37 (7%) with AGER and 43 (8.2%) with NAGER (Figure 5). The number of apneic events temporally linked with GER in each patient ranged from 1 to 15 (median, 4) compared with 1 to 53 (median, 13) apneic events not linked with GER. There was no clustering of apneic episodes in any of the 10 30-second intervals in the 5 minutes after an episode of reflux. At least one apneic event was temporally linked with either AGER or NAGER in 7 subjects, only AGER in 2 subjects or only NAGER in 9 subjects. In the remaining 7 subjects, apnea was never linked with either AGER or NAGER. At least one NAGER episode was linked with at least one apneic event in 64.0% (16 of 25) of the subjects, whereas at least one AGER episode was linked with at least one apneic event in 44.0% (9 of 25).

Fig. 5

The AGER and NAGER events temporally linked with apnea did not differ in height. The refluxate reached the proximal esophagus in 84%, the mid esophagus in 14% and only the distal esophagus in 2% of apnea events. The apneic events linked with GER were obstructive in 47%, central in 20% and mixed in 33% with no difference in distribution between those associated with AGER or NAGER. Figure 6 illustrates the distribution of GER linked apneic events across the 30-second increments of the 5-minute intervals.

Fig. 6

ASSOCIATION BETWEEN GER AND APNEA BASED ON EVENTS OCCURRING DURING 5-MINUTE INTERVALS

Comparison of Temporal Linking and Time Interval Approaches

The temporal linking method did not provide a means of quantifying nonlinked events. The time interval approach identified 4706 (76.2%) of 6173 total time intervals during which neither an apneic event nor a GER occurred. Figure 7 shows that of the 528 intervals with one or more apnea events, 89 intervals (1.45%) contained both apnea and GER. Comparing all episodes of GER and apnea identified per time presented in Figure 7 to the findings from temporal linking in Figure 5 reveals that the number of temporally linked apneic events was similar in the two methods (89 versus 80 linked events). When the sensitivity of the time interval method was tested by shifting the starting time 2.5 minutes forward, there were no changes of significance.

Fig. 7

There were very few statistically significant chi-square associations within individual patients (Table 2). Although the overall Mantel-Haenszel statistics were significant at 5- and 10-minute intervals for accumulated data across subjects, only 3 patients had significant AGER and apnea event associations. The P values for the 5- and 10-minute intervals calculated by Breslow-Day test (Table 3) suggest a lack of consistency in the association between AGER and apnea across patients. Apnea and NAGER were not significantly associated as tested by Mantel-Haenszel.

Table 2

Table 3

Association between the duration and height of GER and the duration of apnea including events that occurred before the apneic interval.

We tested several models to determine an association with apnea duration; the independent variables included: 1) total GER duration, 2) acid duration, 3) nonacid duration and 4) acid duration plus nonacid duration plus height index (Table 4, column 1). Included in this method is the linear lag of reflux before the onset of apnea. That is, those reflux events that occur within the time interval during an apnea event is given the most weight, those reflux events in the interval just previous have less weight and those in the second interval before the apnea event have the least weight. We considered the over-all R² as a function of several combinations of the independent variables. In Table 4 we show the results of regression analyses that consider the type of reflux duration and reflux height.

Table 4

Four of the 25 patients had significant R² in the basic model in which duration of apnea was regressed on duration of GER (Table 4). Those 4 patients were 4 females aged 1, 2, 10 and 14 months at the time of study. All 4 patients had significant apnea that required fundoplication in 3 and gastrojejunal feeding in one of them.

The means of the coefficients for each of the parameters in all of the models were tested against zero. There were no coefficients that were significantly different from zero (the quadratic term for the height index was significant at P = 0.028) indicating that there was no overall relationship between GER and apnea as measured by the pooled individual regressions (Table 4). Autocorrelation of apneic events was present in 7 of the 25 cases, so apnea at lag 1 was entered as an additional independent variable. Two of the four patients with significant R² were among those demonstrating significant autocorrelation. After controlling for autocorrelation among the apnea events, there were no significant GER related associations with apnea. Figure 8 illustrates an example of the time series analysis.

Fig. 8

DISCUSSION

We found few data to support a significant association between apnea and GER. In the few patients in whom apnea was associated with GER, there was no significant difference between acid and nonacid reflux events. The methods of analysis we used suggest that an index of association for each subject can be used for the evaluation of individual patients and might be acceptable for subsequent studies of this relationship.

Using combined impedance and pH instead of pH monitoring alone allowed detection of both nonacid and acid reflux. We found no difference between AGER and NAGER in frequency or in height of esophageal penetration. Our data show that the occurrence of both GER and apnea in a patient does not prove a cause and effect relationship between the two.

We feel our study has several strengths. Counting the apneic events occurring during each 30-second interval up to 5 minutes after the GER episode allowed us to measure the maximum number of apneic events occurring after reflux and to detect any pattern of their occurrences. Some of the protocols for testing the temporal association between GER and apnea reported in the literature involve the identification of apnea occurring within a given time interval either after or before a GER episode. Targeted intervals reported include from 20 to 300 seconds before a reflux to 20 to 300 seconds after onset of reflux (2, 3, 4, 8, and 10). As there is no generally accepted interval after episodes in which GER-linked apnea should be suspected, we elected to identify every apneic event occurring in the 5 minutes after clearance of the bolus of impedance detected reflux episodes. We then plotted these events by 30-second intervals in the 5 minutes after reflux. Even using this liberal definition of linking, we found no significant association.

Dividing each subject's study session into different time periods allowed us to evaluate both the presence and absence of an apneic event in the presence and absence of GER for an individual subject. Analyzing data by index of association is innovative and applicable for individual patients in clinical practice. Finally, the 24-hour duration of the study period is very important when studying GER in infants. In similar studies using 6-hour observations spanning 2 feedings, Dreizzen et al. (²³) showed that physiological GER occurs more frequently during the afternoon than any other time of the day. Vandenplas (²⁴) suggested that 25% to 50% of GER episodes may be missed by investigations conducted only during the nocturnal hours. Shorter duration studies do not appear to reflect the results of longer studies (^23,24). This is especially important when evaluating patients with apnea as respiratory symptoms may have diurnal variations and may relate to specific sleep periods and stages.

A limitation of this study is the wide age range of our subjects. Nevertheless, time series analysis considered each subject independent from the group allowing us to verify significance at an individual level. Many previous investigators have summarized similar data on GER and apnea across subjects. However, treating each GER episode and apneic event as independent events across subjects fails to recognize that some patients clearly have more GER episodes and/or more apneic events than others. Calculations made across subjects do not accurately define the specific characteristics of each subject in the population.

Analyzing our data by temporal linking, the same method used by Wenzl et al. (¹⁰), we found that the incidence of linkage between apnea and GER (AGER and NAGER) was lower in our study (15.2%) than theirs (30%). This may relate to the differences in patient population and the definition of apnea. We only considered a respiratory pause of ≥10 seconds an apneic event while Wenzl et al. defined apnea as a pause lasting ≥5 seconds. Our definition of apnea derived from that of Loughlin et al., who defined apnea as a pause in breathing lasting at least 2 breathing cycles (²⁵). Because our subjects had respiratory rates of 24 to 32 breaths/minute, the 10-second interval is equal to a pause of 1.9 to 2.5 breathing cycles.

Skopnik et al. (¹⁹) also studied the association of GER and apnea by MII/pH monitoring and found that the majority of refluxes were nonacid (72.6%). We found no statistical difference between the frequency of AGER and NAGER. AGER in their study, however, was defined as pH <4 for 15 seconds. We defined AGER as occurring when the pH fell below 4 during impedance defined reflux episodes for at least 5 seconds. Our definition allows detection of brief acid exposure upon bolus contact. Without defining the minimum duration of a GER episode, the same research unit (¹⁰) also found that the vast majority of GER episodes were NAGER (88.5%). The lower incidence of NAGER among the subjects in our study is likely related to the different methodology used.

In summary, we did not find an association between GER and apnea. In light of this finding, clinicians should re-evaluate the practice of prescribing antireflux therapy for apnea in the absence of other reflux symptoms. It is clear that there are some patients in whom apnea may be associated with GER regardless of its acidity. Identifying these patients requires the impedance technique. The data gathered from each patient should be divided into short intervals and analyzed either by χ² statistic or the R² value computed from a lagged regression model for each subject.