Introduction

Dysphagia is a common symptom of a wide variety of underlying diseases, including neurologic and respiratory conditions [1]. Secondary complications of dysphagia may include aspiration pneumonia, respiratory failure, malnutrition, dehydration, and death [2, 3]. These complications frequently lead to increased hospital length of stay and higher medical costs [3]. For these reasons, early detection and management of dysphagia is important to minimize or avoid the expensive and sometimes life-threatening consequences of dysphagia.

Pulse oximetry has been advocated by some as a means to detect aspiration [4, 5]. Pulse oximetry is a commonly used noninvasive means of measuring peripheral capillary oxygen saturation (SpO2), i.e., the percentage of oxygenated hemoglobin in individuals at risk for hypoxemia [6, 7]. This is accomplished by measuring the ratio of light absorption of oxygenated (oxyhemoglobin) and deoxygenated (deoxyhemoglobin) blood. The underlying rationale posited for use of pulse oximetry to detect aspiration is that aspiration may lead to bronchoconstriction and/or airway obstruction leading to ventilation–perfusion mismatch [8,9,10], which in turn results in a drop in oxygen saturation [9]. Potential use of pulse oximetry to detect or screen for aspiration risk is attractive to clinicians, as it is readily available, quick, and noninvasive, and may be easily incorporated into a clinical swallowing examination. In fact, use of pulse oximetry to detect aspiration has been advocated and adopted into clinical practice by many clinicians. For instance, pulse oximetry has been incorporated into clinical swallowing assessment protocols, such as the Volume–Viscosity Swallow Test [11, 12]. However, research and expert opinions regarding the validity of pulse oximetry to indicate aspiration has yielded mixed findings. For instance, Zaidi et al. reported oxygen desaturation during swallowing to be a potential marker of aspiration in individuals following acute stroke [4]. However, others have challenged the validity of using pulse oximetry to detect aspiration [13]. Owing to the mixed conclusions of research examining use of pulse oximetry to detect aspiration, it remains a topic of debate and clinical confusion. In this regard, a systematic review of the body of this literature will shed more light on the underlying factors contributing to these discrepant findings and offer clinical direction on the validity of using pulse oximetry to detect prandial aspiration.

The purpose of this study is to systematically review evidence on the use of pulse oximetry in individuals with dysphagia to detect a decrease in SPO2 indicating aspiration during swallowing, toward the goal of further informing clinical practice in dysphagia assessment.

Methods

This systematic review was completed in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) standards [14, 15], an evidence-based set of guidelines for reporting systematic reviews. Peer-reviewed research studies directly investigating use of pulse oximetry to detect aspiration with simultaneous confirmation of aspiration via a gold standard instrumental study, i.e., modified barium swallow study (MBSS) or fiberoptic endoscopic evaluation of swallowing (FEES), in adults were included. Articles that did not meet the inclusion criteria or were published in a language other than English were excluded (Fig. 1). Conference abstracts were excluded owing to insufficient information to facilitate review. Articles that studied pulse oximetry only in combination with other aspects of the clinical swallowing exam were excluded due to the difficulty in disentangling the specific contribution of pulse oximetry to detection of aspiration. Articles examining pediatric populations were excluded due to the potential impact of differences in swallowing and respiratory anatomy on findings. Because of the inherent challenges in conducting large-scale studies involving dysphagia rehabilitation research [16], we did not exclude research based on study design or populations/settings studied. Comprehensive literature search of PubMed, CINAHL Plus, PsychINFO, AMED, and Embase databases without date limits was conducted on 8/25/16 with an updated search covering the years 2016 forward on 4/8/17. Literature search strategies included use of medical subject headings (MeSH) for the PubMed database, along with text related to dysphagia, aspiration, and pulse oximetry used with all targeted databases. See Table 1 for keywords and search strategy. To ensure literature saturation, additional hand searches of references from included articles, relevant review articles, and other sources known to the authors were completed.

Fig. 1
figure 1

Preferred Reporting for Items for Systematic Reviews and Meta-Analysis (PRISMA) flow chart

Full size image
Table 1 Search terms
Full size table

The strength of the evidence for all of the articles fully reviewed was rated via the Australian level of evidence hierarchy for diagnostic accuracy (Table 2) [17], owing to ease and reliability for characterizing quality of evidence and risk for bias in diagnostic studies. Additional information was extracted from each article to further determine risk for bias at both outcome and study levels. Data extracted from each article included demographic information, parameters of measurement for pulse oximetry in the context of swallowing, swallowing assessment methods, and study findings.

Table 2 Australian levels of evidence.
Full size table

Results

Study Selection

Following removal of duplicate records, the search yielded 294 citations. Citations were screened in two stages. In stage I, titles and abstracts were independently screened by two co-investigators for relevance to the purpose of the search. Two hundred ten articles without a specific focus on both pulse oximetry and swallowing were excluded. In stage II, abstracts were reviewed and articles skimmed. Sixty-five articles that did not directly examine the use of pulse oximetry to detect aspiration were excluded at this stage, including review articles and abstracts. Blinded inter-rater agreement for inclusion/exclusion of citations at stage I and II screening was > 80%. Discrepancies were resolved via a third reviewer along with discussion and consensus. The remaining 19 articles underwent full blinded review by two co-investigators to determine relevance to the specific purpose, with exclusion of nine additional articles due to lack of simultaneous examination of pulse oximetry during swallowing with gold standard detection of aspiration, i.e., via MBSS or FEES. Ultimately, ten studies were qualitatively synthesized. With use of a detailed form, the ten articles meeting criteria were reviewed in full for details contained in Tables 3, 4, 5, and 6 by two co-investigators with blinded data extraction and quality of evidence rating. Disagreements were resolved through consensus discussion.

Table 3 Study demographic information
Full size table
Table 4 Pulse oximetry parameters and risk for bias
Full size table
Table 5 Swallow study parameters
Full size table
Table 6 Study findings—prediction of aspiration from pulse oximetry
Full size table

Study Characteristics

All ten citations reviewed examined SpO2 simultaneously with a gold standard method of detecting aspiration, via either MBSS [5, 18,19,20,21,22] or FEES [23–26]. All of these articles were rated at Level III-2 on the Australian level of evidence hierarchy for diagnostic accuracy (Table 2) [17].

Demographic information for the studies included is outlined in Table 3. Most of the studies focused primarily on the stroke population; others included individuals with multiple underlying diagnoses. None of the studies included a statistical power analysis. Over half of the studies did not include a comparison group; some of these studies compared instances of aspiration versus no aspiration within or between subjects. Exclusion criteria were variable or, in some studies, unspecified.

Pulse Oximetry Parameters

Pulse oximetry measurement parameters and additional factors indicating risk for bias are outlined in Table 3. Criteria for “desaturation” in conjunction with swallowing were variable between studies, either ranging from a 2 to 4% drop in oxygen saturation or unspecified with group comparisons. Because the margin of error in most pulse oximetry devices is approximately 2% [27,28,29], use of the 2% drop in SpO2 criteria for judging desaturation may lead to bias. Study efforts to prevent measurement artifact were variable and unspecified in half of the studies reviewed. Nearly half of the included studies did not report efforts to blind observations. Additional potentially confounding variables were noted across all studies and included mixed populations, populations with underlying respiratory impairments, lack of control for viscosity or texture, and lack of a comparison group (Tables 3 and 4).

Instrumental Swallow Study Parameters

Swallow study measurement parameters are outlined in Table 5. The criteria for categorizing a participant as an aspirator were not clearly specified in most of the reviewed studies. Others indicated that participants would be categorized as “aspirators” if they were observed to aspirate at least once on any food or liquid trial [22, 23]. Most of the studies reported observations with a variety of textures and amounts, but only one compared findings based on the texture or viscosity aspirated [24].

Study Findings

Study findings were mixed and highly variable (Table 6). Reported measures of sensitivity and specificity for use of pulse oximetry to detect aspiration ranged from 10 to 87% and 39 to 100%, respectively. Reported measures of positive predictive value (PPV) and negative predictive value (NPV) ranged from 35.5 to 100% and 52 to 88%, respectively. Three of the four studies examining group comparisons reported a lack of change or relationship between aspiration and oxygen desaturation. The majority of studies (7/10) failed to demonstrate an association between aspiration and O2 desaturation. Two of the ten studies that examined O2 desaturation in relation to observed events of aspiration concluded that there was no significant relationship between events of aspiration and subsequent O2 desaturation.

Discussion

This systematic review has examined the use of pulse oximetry to detect aspiration in studies with simultaneous confirmation of aspiration via a gold standard instrumental study, such as MBSS or FEES. Although the research findings are mixed, the majority of the studies examined in this systematic review do not support the use of pulse oximetry for purposes of detecting aspiration (Table 6). In addition, the strength of the evidence is weakened in many of these studies by study design flaws and potential for bias, such as lack of a comparison group, blinding and inter-rater reliability, as well as inadequate sample size (Tables 4 and 5). Therefore, the evidence for use of pulse oximetry as a means to detect prandial aspiration is inadequate and not convincing.

This discussion will reflect on the reviewed research within the context of methodological factors, which may have impacted the study findings, as well as other aspects of medical knowledge critical to clinical decision making [30] in this context, including consideration of physiologic factors that could affect O2 saturation or desaturation in the context of swallowing.

Variability in Comparison Methods for Prandial Aspiration

Clinically, it is well known that individuals who aspirate do not necessarily aspirate on every swallow [31]. Further, risk for aspiration in individuals may vary depending upon the texture or liquid viscosity [32] and the size of the bolus [33, 34]. Theoretically, these same variables might impact the degree of desaturation. For instance, aspiration of a larger or solid bolus might be more likely to trigger airway occlusion or bronchospasm. This may have been a factor in the study published by Collins and Bakheit as their finding of a positive correlation between prandial desaturation and aspiration was observed in the context of much larger bolus volumes [5]. For these reasons, group and case comparisons (Table 6) that seek to classify each individual as an aspirator or non-aspirator may cloud the issue of whether or not pulse oximetry can detect aspiration for specific swallow events. Only two of the ten reviewed studies (Wang et al. [22]; Marian et al. [26]) examined a drop in SpO2 in association with specific swallows [22, 26] (Table 6). Both of these studies concluded no relationship between O2 desaturation and instances of aspiration. In addition, two of the ten studies (Chong et al. [21]; Smith et al. [23]) also categorized individuals with observed penetration (no aspiration) in the group of individuals with observed aspiration [21, 23] (Table 5). However, theoretically, it does not make sense that SpO2 would decrease due to the presence of food or liquid in the laryngeal vestibule, as bronchospasm or interference with respiration would typically occur at the level of the lower respiratory system or lungs.

Definition of “Desaturation” in the Context of Swallowing

One issue of variability lies in the definition of “desaturation” in the context of using pulse oximetry to identify instances of aspiration. Of the studies examining sensitivity/specificity of pulse oximetry to detect aspiration in this review, the majority specified a ≥ 2% drop in O2 saturation during eating as “desaturation” (Table 4). The stated rationale for this criterion is unclear and may have continued over time through precedent. The problem with this criterion is that the standard error of measurement for pulse oximeters is generally about 2% [6]. More specifically, for SaO2 values of 90% or higher, the precision of SpO2 measures to estimate the actual percent saturation of oxygen bound to hemoglobin in arterial blood (SaO2), i.e., the standard deviation of differences between the 2 measures, is < 3%, and the bias of SpO2 measures, i.e., the mean difference, is < 2% [27,28,29]. Other researchers have commented on this issue as well. For instance, Colodny stated that “controversy exists over the use of the 2% desaturation criterion as indicative of aspiration” [13]. In addition, both Leder and Colodny critiqued the 2% criterion as being arbitrary and at the margin of error for the pulse oximetry equipment [24, 25].

In addition to considering the standard error of measurement in pulse oximetry devices, it is important to note that the bias and precision of SpO2 measures tend to worsen when SaO2 is lower than 90% [27, 28]. This is not surprising considering the characteristics of the oxyhemoglobin dissociation curve (Fig. 2), a sigmoid-shaped curve that demonstrates the complex relationship between the amount of oxygen available and dissolved in plasma, i.e., partial pressure of O2 dissolved in arterial blood (PaO2), and the amount (percent) saturation of oxygen bound to hemoglobin in arterial blood (SaO2). Owing to the sigmoidal shape of the oxyhemoglobin disassociation curve [35], pulse oximetry may not detect hypoxemia as easily in individuals with higher PaO2 levels [27]. Conversely, individuals with lower SpO2 values (e.g., in the lower 90s) at baseline may show larger drops with smaller changes. It is therefore plausible that an individual’s underlying respiratory function could impact the accuracy of pulse oximetry in detecting aspiration.

Fig. 2
figure 2

Image source Wikimedia Commons contributors, 'File:Oxyhaemoglobin dissociation curve.png', Wikimedia Commons, the free media repository, 19 February 2017, 01:04 UTC, (https://commons.wikimedia.org/w/index.php?title=File:Oxyhaemoglobin_dissociation_curve.png&oldid=234162700) [accessed 26 November 2017]

Oxyhemoglobin dissociation curve. The sigmoidal shape demonstrates the varying affinity of hemoglobin for O2 [35]. Note that once PaO2 is below 60 mmHg, the curve becomes steep and smaller changes in PaO2 will reduce SaO2. The affinity of hemoglobin for O2 will vary with a number of factors. For instance, the curve will shift left (increasing oxygen affinity for hemoglobin) with increased pH, reduced PaCO2, reduced 2,3-diphosphoglycerate (DPG), and reduced temperature. Conversely, the curve will shift right (reducing oxygen affinity for hemoglobin) with decreased pH, increased PaCO2, increased DPG, and increased temperature. PaO 2 partial pressure of O2 dissolved in arterial blood; SaO 2 saturation of oxygen bound to hemoglobin in arterial blood

Full size image

Control for Artifact in Pulse Oximetry Measures

In addition to considering the limits of pulse oximetry devices and the impact of the oxyhemoglobin dissociation curve, other factors may lead to erroneous pulse oximetry measures. Measurement artifact may occur for a variety of reasons, including motion artifact, nail polish, skin pigmentation, low perfusion state (e.g., low cardiac output, vasoconstriction, or hypothermia), use of intravenous dyes, and anemia [27, 36]. Elevated levels of dyshemoglobins, such as carboxyhemoglobin (associated with exposure to carbon monoxide) and methemoglobin, may also lead to inaccurate SpO2 readings [27]. Use of a waveform display, available on many pulse oximeters, can help distinguish a normal signal from a signal with various forms of artifact [27]. Many of the studies included in this review documented efforts to minimize artifact, but at least four did not specify these efforts (Table 4). The time it takes for a pulse oximeter to detect a change in SpO2 is also a consideration with attempts to use pulse oximetry to detect aspiration. Jubran reported that the length of time to detect a decrease in SpO2 was > 1 min [27]. All articles included in this review specified the duration of time for observation after each swallow (Table 4).

Possible Impact of Breathing Swallowing Coordination

Alteration of breathing–swallowing coordination might potentially explain desaturation in the context of swallowing for some individuals. The upper airway serves multiple purposes in humans, including breathing, swallowing, and speech. Swallowing must therefore be coordinated with respiration. The key indices of breathing–swallowing coordination include the respiratory phase pattern (observation of inspiration or expiration before and after each swallow) and the swallow apnea duration [37]. The swallow apnea duration refers to the duration of time of a centrally controlled cessation of breathing that occurs during each swallow. Swallow apnea duration may change in the context of underlying disease. For instance, longer swallow apnea durations have been observed in the elderly [37] and in individuals with neurologic disease [38]. In addition, in individuals with underlying respiratory or pulmonary disease, such as those with an elevated respiratory rate, the time needed for the swallow apnea (approximately 1 s) could compete with the individual’s ventilatory needs. For instance, oxygen desaturation associated with eating and swallowing—without aspiration—was observed in some of the reviewed studies. Theoretically, the time needed for the swallow apnea in individuals with underlying respiratory impairments and/or prolonged swallow apnea duration owing to underlying neurologic disease might lead to the fluctuation of O2 saturation. Further research would be needed to confirm this possibility.

Conclusion

Current evidence does not support the use of pulse oximetry to detect aspiration. This may be partially related to methodological factors in existing studies, including variability in comparison methods for measuring prandial aspiration, defining ‘desaturation’ in the context of swallowing within the standard error of measurement, and lack of control for artifact in pulse oximetry measures. It is possible that factors other than aspiration might explain desaturation in the context of swallowing in some individuals, such as alterations to swallow apnea duration. Although utility for detection of aspiration is not confirmed, pulse oximetry may aid generalized judgments of patient homeostasis during swallowing assessment.