A New, Noninvasive Accelerometric Method for Diagnosing Acid andNon-Acid Gastroesophageal Reflux in Premature Infants: Proof of Concept

Abstract

Non-acid reflux is common in premature neonates. Current methods of diagnosing gastroesophageal reflux (GER) such as pH probes, multichannel impedance monitoring, X-rays, or endoscopy are either invasive or unable to diagnose non-acid reflux. Passage of a naso-esophageal tube is uncomfortable. Imaging studies are of short duration and may miss reflux entirely. Herein, we present proof of concept of a noninvasive accelerometric device that detects acid and non-acid reflux in premature infants. An accelerometer was taped over the subxiphoid process in patients suspected of having GER who were already scheduled for pH probe or multichannel impedance monitoring. The largest cohort was preterm infants, but term infants and toddlers were also studied. Low-frequency subaudible signals were obtained on a digital recorder (sampling rate 200 Hz) signals. Fast Fourier transforms graphically displayed the frequency and amplitude of signals. Data were then resampled at a rate of 60 Hz to create a spectrogram with a focused range of 0–30 Hz representing reflux-associated events. Proof of concept was attained through successful comparison with results from concurrent pH probes, multichannel impedance recordings, and ultrasound studies. We have thus validated accelerometry as a noninvasive method for assessing both acid and non-acid GER. The noninvasiveness of this diagnostic modality allows for repeated testing to assess the efficacy of anti-reflux medications, even when patients remain on antacids. This technology allows for more rational management of patients with GER and represents a major advance in the diagnosis and treatment of GER.

1 Introduction

Gastroesophageal reflux (GER) refers to the involuntary retrograde passage of gastric contents into the esophagus and above, commonly caused by the relaxation of the lower esophageal sphincter, with factors such as decreased gastric and intestinal motility playing a role as well, especially in preterm infants [1]. GER remains a common problem throughout life, but especially in preterm babies, young infants, and both children and adults with dysphagia or neurological disease of any cause [2–4].

Current diagnostic modalities in GER are suboptimal, being characterized by invasiveness and lack of reproducibility. In preterm babies, ≈75% of GER is nonacid in nature [5–7], yet the most commonly used diagnostic method is the pH probe, which, by definition, does not diagnose nonacid reflux. pH probes, multichannel impedance monitors, endoscopy, and various imaging techniques used in diagnosing GER are invasive, involving placement of an esophageal probe and/or use of X-rays, making repetitive use problematic. The measurement of esophageal pH is not considered a reliable method of diagnosing GER in premature infants since the stomach pH in premature infants is rarely <4 because of a higher baseline pH and frequent milk feedings that neutralize acidity [8,9]. Multichannel impedance monitoring, also invasive, suffers from lack of reproducibility [10,11] and lack of standards for neonates. Therefore, optimal diagnosis (and treatment) of GER has not been well defined, and most pharmacological treatments, besides having questionable efficacy and a variety of side effects [12], have not been studied longitudinally.

We therefore saw the need for an accurate, noninvasive methodology to diagnose both acid and non-acid reflux in preterm infants. Building on our previous experience [13–15], we have developed a new method, described below, to diagnose GER with accelerometry [16,17].

To establish proof of concept, we compared our methodology to current diagnostic methods, including pH probes and multichannel impedance monitoring, as well as real-time ultrasound imaging.

2 Methods and Procedures

A schematic of the methodology is shown in Fig. 1(a). Our system consisted of an accelerometer (Honeywell Sensotec MAQ36; Columbus, OH) taped to the skin over the subxiphoid process (Fig. 1(a)). Acoustic signals were captured by a digital recorder (DASH 2EZ+, Astro-Med, Inc., West Warwick, RI) using a 200 Hz sampling rate (to yield a 100 Hz window so that we could focus on the subaudible range (0–30 Hz) where the low-frequency signals occurred). A band stop filter at 60 Hz, and other filters as necessary, was run in order to eliminate interference from other Neonatal ICU (NICU) electronic equipment. More recently, we have used a custom-designed digital recorder in conjunction with a digital three-dimensional accelerometer (Freescale MMA8451Q), Mouser Electronics, Mansfield, TX (Fig. 1(b)).

Fig. 1.

(a) Schematic of the setup of the accelerometric Device. (b) accelerometer photo of the accelerometer used in these studies.

The spectrogram was then de-jittered by soft thresholding with a 100 μV threshold. The remaining signals are processed using a fast Fourier transform with a segment size of 1024. After the de-jitter, if any spectral density above 60 Hz exceeds 2 μV, the entire segment (from 0–100 Hz) is excised in order to eliminate potential artifact. The data are then resampled at 60 samples per second to create a spectrogram with a segment size of 512 and a focused range of 0–30 Hz, and the mean amplitude in μV is calculated for the entire recording. An average value of >1 μV in the processed signal is considered indicative of reflux.

As a first step for proof of concept, accelerometric readings were recorded simultaneously with pH probes in 155 preterm infants [16,17] suspected of having GER. Full-term [18], and older (1–15 month) infants (n = 57), as well as ten adults have also been studied, but will be presented at a later date. At the outset, it was key to demonstrate that accelerometric data correlated with acidic pH probe (pH < 4) and positive ultrasound and multichannel impedance recording data. An average value of ≥1 μV for the entire recording period (4–8 h) is considered indicative of reflux based on the fact that the cutoff of a mean μV of 1 “fit” the overall dataset best when comparing results from all babies who had concurrent pH and accelerometric studies.

Inclusion criteria included all premature infants with gestational age <37 weeks. While most babies had pH probe studies or impedance studies already planned for suspect GER, in some cases, accelerometric studies were performed even when pH probes or impedance studies were not done (scheduling issues, obtaining normal/control infants, term infants, etc.).

No babies with upper gastrointestinal tract anomalies or gastrointestinal obstruction were included.

Informed consent for accelerometer use was obtained prior to each study. Other studies, such as pH probe (pHNS-P, 5-French, single channel probe, ComforTEC™; ZepHyr 2000-A monitor, Sandhill Scientific, now Diversatek Healthcare, Milwaukee, WI), multichannel impedance monitor (Sleuth ZPN-BS-46, 6.4 French probe, ComforTEC; Z07-2000B impedance/pH monitor, Sandhill Scientific, now Diversatek Healthcare), or X-rays were performed as part of the infant's standard of care, at the discretion of the attending physician. This protocol has been approved by the IRB of Michigan State University and Sparrow Hospital.

3 Results

Given the concerns regarding the current state of the art with respect to the diagnosis of GER, especially in infants, (invasiveness, lack of reproducibility, use of acidity (in pH probes) as the primary outcome), we hypothesized that accelerometry could prove to be useful in capturing GER-related events noninvasively.

Figure 2 is a result from a concurrent pH probe recording and an accelerometric signal. As can be seen (Fig. 2(b)), when the pH begins to dip (from about 7 to 2 at approximately 10 am, before recovering and increasing over the next 15 min) the accelerometric signals (Fig. 2(a)), with the portion of interest occurring in the subaudible range (<10 Hz), increases at the same time. It can be readily seen that the accelerometric recordings matched the pH changes consistently, with high accelerometric readings in the period between 9:58 and 10:13 am.

Fig. 2.

Concurrent pH probe and accelerometric recordings. This representative pH recording (b) shows a change from ≈7 to ≈2 at 10:00, slowly returning to a reading of >4 by ≈10:10. At the same time, the accelerometric readings (a) in the range of <30 Hz show a concurrent increase in μV, continuing through 10:13.

We then compared the concurrent accelerometric and pH recordings in 100 preterm infants suspected of having GER. Using the pH probe cutoff for “abnormal” recordings (Boix-Ochoa score > 16.6 [19] and our preliminary cutoff for accelerometric recordings (μV > 1.0), we found that when the pH recordings were positive (i.e., pH scores were indicative of significant GER), the accelerometric recordings were also indicative of GER (n = 20). There were only two points where our method was negative even though the pH probe was positive (91% sensitivity). Thus, in a clinical setting, the accelerometric technique appeared to be a good alternative to other, more invasive methodologies.

This technology also yields itself to a qualitative assessment of GER severity. As can be seen from Fig. 3, accelerometric readings can range from minimal, to intermediate or high, suggesting that a quantitative scoring system (using mean μV during the study, number of high μV readings, highest μV reading, length of high μV readings, and longest high reading, analogous to the pH probe algorithm) based on our recordings can be developed. Presumably, a high reading would indicate that the baby is having more than average reflux and would require closer monitoring for signs of potential aspiration and possible need for treatment. This is currently in progress, although at the present time we are only using mean μV throughout the study. Accelerometry was also compared with simultaneous abdominal ultrasound examinations (Fig. 4(a)), with correlation noted between visible lower esophageal GER and increased low-frequency signals on the accelerometer (Fig. 4(b)). Videos of positive and negative ultrasound and concurrent accelerometric readings are available in the Supplemental Materials on the ASME Digital Collection.

Fig. 3.

Graded accelerometric readings. Accelerometric readings from normal to minimal, intermediate to high. This indicates that accelerometric readings lend themselves to quantitative analysis.

Fig. 4.

Simultaneous ultrasound and accelerometer recordings. Representative abdominal ultrasound examinations (a) showing back-and-forth movement of abdominal lower esophageal contents were compared with simultaneous increased low-frequency voltage on the accelerometer (b). Videos of positive and negative ultrasound and cotemporal accelerometric readings are available in the Appendix.

Finally, our accelerometric technique was compared with simultaneous multichannel impedance monitor recordings. Figure 5(a) shows an example of positive acid reflux correlating within seconds with our device; a similar example of correlation between accelerometry and weakly acidic reflux is seen in Fig. 5(b). Figure 5(c) shows a nonacid negative study using both modalities.

Fig. 5.

Simultaneous accelerometric and impedance/pH recordings. (a) At ≈12:00 pm, there is an episode of acidic reflux on the multichannel impedance monitor, accompanied by a drop in pH to approximately 3 and a sharp increase in μG on the accelerometric monitor. (b) At ≈5:19 pm, there is a non (or weakly (briefly reaching pH = 4)) acidic GER episode, accompanied by increases in μV on the accelerometric recording. (c) At normal 30-min period where no significant changes in impedance are noted, the pH is about six and accelerometric readings are minimal.

4 Discussion

In the medical sphere, to date, accelerometers have been mainly used to assess physical activity in adults and children [20]. They have also been used in speech research in dysarthric children to assess lingual-palatal articulation [21] and in our work [13–15] on the maturation of feeding in infants. Our GER Monitor consists of a noninvasive sensor (accelerometer), which detects acceleration/movement in the lower esophagus (regardless of the acidity of the refluxate).

Besides being able to diagnose GER events, the noninvasiveness of our device allows for repeated monitoring of patients suspected of having GER disease (GERD) over time, allowing for an assessment of treatment efficacy and thus assisting the clinician to more rationally tailor therapy.

The data indicate that the accelerometric signal would often be elevated for many minutes, as also seen in the concurrent ultrasound recordings. This suggests that the accelerometer is measuring an actual back and forth movement of the refluxate. Of interest, feeding did not appear to result in significant changes in accelerometric signals, suggesting that normal peristaltic movements are not robust enough to register on our machine, while the reflux movements are.

Of note, although recordings could have just as easily been obtained by placing the accelerometer on the back of the patient, we chose not to do so, since current recommendations (for the prevention of Sudden Infant Death Syndrome) are for the babies to sleep on their backs [22].

The data were recorded over 4–8 h, as that was the limit of the current device. This usually allowed us to monitor 2–3 feeds in each neonate. The pH probe was used for 24 h. It was impossible to compare the exact concurrent times when both monitors were on the baby, since the Boix scoring system can only be used for 24 h. Nevertheless, in spite of this, the two systems matched up well, since the time of day is less important in preterm babies, as they are fed round the clock (q 3 h), and sleep most of the day (unlike adults who are upright most of the day, and eat widely spaced meals, but then are supine or prone at night and are fasting). Thus, there should be less variability between times in neonates compared to adults, where morning studies might be quite different from nighttime recordings (where position changes and fasting is the norm).

Approximately 50% of normal term infants have reflux (vomiting 2 or more feeds/day) at the age of 2 months, but only 1% still have reflux at 1 year of life [6]. Over the past 20 years, GER has become a major diagnostic and therapeutic conundrum in preterm infants in neonatal intensive care units and has been associated with episodes of vomiting, apnea, poor feeding and growth delay, and, of most concern, aspiration into the lungs with resultant respiratory compromise. However, since preterm babies cannot localize their problem, diagnosis is often difficult and requires invasive testing. Indeed, Marino et al. found that 63% of infants discharged from the Neonatal Intensive Care Unit (NICU) had evidence of GER [7] and 25% of all graduates from the NICU are discharged on anti-GER meds, often for many months [22].

The multiplicity of methods to diagnose GERD currently in use reflects the fact that none are without problems. The most commonly used methods are the pH probe and the impedance monitor (usually in conjunction with a pH probe). The invasiveness of a nasogastric tube argues against the newer method of multichannel impedance monitoring/pH probe; impedance monitoring also suffers from a lack of interobserver agreement [10,11] and poor reproducibility even when performed on two consecutive days, especially for non-acid reflux [24]. Normal values have not been published for preterm infants or neonates in general. In addition, Funderburk et al. found that GER as measured using impedance techniques was rarely associated with symptoms commonly seen with GER (irritability, bradycardia, and desaturation) [23,25]. Automated analysis of impedance seems better than manual analysis because of the suboptimal interobserver agreement; however, Loots et al. [10] caution that manual analysis may still be required because of low specificity rate using automated analysis, and that further refinement of such analysis is needed.pH probes also detect reflux episodes not detected by intraluminal impedance, with as many as 59% of pH positive episodes not detected by impedance monitoring [24,26], raising further questions about the accuracy of such testing.

The pH probe, used alone, is also hampered by its invasiveness (use of a nasogastric tube to assay esophageal pH; use of X-rays to determine correct placement of the esophageal tube) and, at least in neonates, by the fact that it does not measure non-acid reflux, which is far more common than acid reflux in neonates [5–7], resulting in false negative studies. The reliance on acidity as the outcome measure may be further flawed by the fact that 90% of the time preterm infants have a gastric pH of >4 [9,25]. Milk feeds further buffer the gastroesophageal pH [8,9,26]. pH probes also require the discontinuation of antacids prior to use of the pH probe, making it a poor tool to measure therapeutic efficacy (i.e., response to therapy). In addition, pH probes suffer from relatively low reproducibility even when performed only one day apart [27,28].

More recently, a “wireless” system for esophageal pH monitoring has been introduced [29]. It is placed transorally 6 cm above the squamo-columnar junction and attached to the mucosal wall of the esophagus. Using an antimony electrode, it then transmits a signal via a pH telemetry capsule to an external receiver via radio-telemetry. While the lack of a catheter makes this more comfortable for long-term recordings, it requires an initial transoral placement to locate the capsule properly as well as a surgical attachment to the internal esophageal wall, making it technically invasive. Of course, it is also based on pH recording, with all of the problems noted above for pH probes.

Other commonly used ways of diagnosing GERD include a variety of X-ray studies and scans. These have the problem of being episodic (i.e., involving only a short testing period, allowing reflux episodes to be missed), a problem with ultrasonography, as well. The inherent risk of repeated X-ray exposure is also to be considered.

One of the advantages of the impedance methodology is that it allows an assessment of the height of the refluxate, by demonstrating how high the impedance changes are noted on the multiple sensors [8]. At the present time, we are only using one sensor on our device, as seen in the present report. However, we are currently developing an array of accelerometers, which will give us the ability to assess the height of the refluxate and hence to assess the risk of aspiration, as well.

5 Conclusion

Since our methodology is noninvasive and does not rely upon acidity, it allows us, for the first time, to repetitively measure GER, before and after therapy. The ability to repetitively test patients also allows for a better understanding of the efficacy of various medicines and treatment regimens for GER. Indeed the current inability to track treatment efficacy has led to the disheartening fact that many NICU “graduates” are sent home on reflux medications and may stay on those medications for many months [22], often without proven efficacy and with potential side effects [12]. We are currently in the process of developing an automated quantitative scoring system to allow for assessing severity of GER as well as developing the use of multiple accelerometers to quantify the extent to which the refluxate ascends up the esophagus, as a way of assessing risk of aspiration. This technology can potentially have a major diagnostic, therapeutic, and economic impact on the overall management of gastroesophageal reflux.

Supplementary Material

Supplementary Material

Supplementary pptx File

Funding Data

• Michigan State University (Funder ID: 10.13039/100007709).

• Targeted Support Grants for Technology Development from Michigan State University (IHG).

• Michigan Initiative for Innovation and Entrepreneurship Technology Commercialization Fund (IHG).