Dysphagia was diagnosed if any of the following four criteria was met:

Food intake level scale (FILS) (Kunieda et al., 2013) less than Level 10. This scale categorizes the severity of dysphagia primarily based on methods of feeding. Subjects who receive nutrition via non-oral pathways are categorized as Levels 1–3, subjects who receive nutrition by both oral dysphagia diet and alternative nutrition are categorized as Levels 4–6, subjects who receive nutrition by various degrees of dysphagia diet are categorized as Levels 7–9, and normative subjects are categorized as Level 10.

FILS was recorded for all patients, and all subjects scored <10 (7.3 ± 2.8). Since the severity of dysphagia in these patients was mild, each subject additionally underwent at least one of three swallowing screening tests to confirm the functional impairment (RSST: 2.2 ± 1.3, n = 28; MWST: 4.4 ± 0.5, n = 14; WST: 1.9 ± 0.3, n = 10).

For those who were recruited from the Daito City Cohort and at the health promotion festival, the Japanese version of the 10-item eating assessment tool (EAT-10) was performed to facilitate clear differentiation of healthy subjects within the elderly population. EAT-10 score of 3 or higher was considered as possible dysphagia (Belafsky et al., 2008). According to this criterion, 269 subjects belonged to the healthy group, and 24 subjects belonged to the possible dysphagic group. The mean ages of the healthy, possible dysphagic group and of the dysphagic group were not statistically different (healthy: 75.0 ± 6.1 years old, range 56–92 years old, 21 males and 248 females, possible dysphagia: 75.9 ± 6.6 years old, range 66–96 years old, 3 males and 21 females, dysphagia: 73.3 ± 9.9 years old, range 55–89 years old, 15 males and 15 females, p = 0.324).

To evaluate breathing–swallowing coordination, we developed a unique swallow monitoring device with minimal instrumentation, which enabled even ambulant monitoring of the swallowing function (Yagi et al., 2016). The device consists of a nasal cannula-type flow sensor connected to a differential pressure transducer, a film-type piezoelectric sensor attached on the surface of the thyroid cartilage, and a logging system that stores signals in a microSD card. The piezoelectric sensor has a wide dynamic range (0–4 kHz) so that both the laryngeal motion and sound are captured. Swallowing periods are extracted semi-automatically with an algorithm using the respiratory flow, the swallowing sound, and the laryngeal motion.

We analyzed the following parameters associated with breathing–swallowing coordination, as shown in Figure 1, using custom-made programs, written in MATLAB R2014b (Mathworks, Natick, MA, USA). A total of 2,648 swallows were analyzed.

We used soft jelly and water for test foods. The properties of the soft jelly, i.e., hardness, adhesiveness, and cohesiveness are strictly controlled to meet the criteria of Level 0 dysphagia diet according to the Japanese Society of Dysphagia Rehabilitation (JSDR) specification described in IDDSI (International Dysphagia Diet Standardization Initiative) report (Cichero et al., 2016). Subjects were in an upright position sitting on a chair, and swallowed voluntarily about 3 g of Level 0 test food from a teaspoon and 3 ml of water from a 5-ml syringe, two to five times for each. Subjects were instructed to swallow the Level 0 jelly without chewing.

The correlations between parameters were evaluated using Pearson's product-moment correlation analysis. Comparisons of the parameters between different breathing–swallowing coordination types were performed using single-measurement, simple-factorial analysis of variance (ANOVA), followed by post hoc analyses with Tukey-Kramer test. Differences in the frequency of breathing–swallowing coordination patterns between three different timings of swallow in the respiratory cycle, and differences in timings of swallow in the respiratory cycle between healthy, possible dysphagic, and dysphagic groups were compared using the chi-squared test followed by Haberman's residual analysis. All data are presented as mean ± standard deviation. P-values were two-sided, and P < 0.05 was considered as statistically significant.

For analyses in this subsection, we utilized all data from the healthy, possible dysphagic, and dysphagic groups. We found a strong correlation between swallowing latency and pause duration (r = 0.867, p < 0.0001; Table 4), and weak correlations between old phase and swallowing latency (r = 0.316, p < 0.0001), between old phase and pause duration (r = 0.216, p < 0.0001), and between old phase and I-SW frequency (r = −0.259, p < 0.0001). The swallowing latency was variable, which resulted in scattered phase resetting characteristics shown as a co-phase plot (Figure 2). To gain insight into the variability of swallowing latency and pause duration, we quantified the variability of swallowing latency and pause duration (deglutition apnea) for those who had more than eight swallow records. Neither swallowing latency nor variability [coefficient of variation; (SD/mean) ^* 100] of swallowing latency was correlated with age (Figure 3). Further, neither pause duration nor variability of pause duration was correlated with age.

Swallows with E-SW-I pattern had a significantly larger old phase, smaller co-phase, and shorter time to the following inspiration as compared to those with E-SW-E pattern, whereas swallows with I-SW-E pattern had a significantly smaller old phase, larger co-phase, and longer time to the following inspiration (Table 5). To further elucidate the relationship between old phase and breathing–swallowing discoordination, we divided the old phase into three groups, early (old phase <0.4), intermediate (old phase between 0.4 and 1.0), and late (old phase >1.0). Swallows with an early old phase had a greater chance of I-SW pattern as compared to those with intermediate and late old phases, whereas swallows with a late old phase had a greater chance of SW-I pattern as compared to those with early and intermediate old phases (Table 6). This relationship between old phase and breathing–swallowing discoordination was common in both healthy subjects and patients with dysphagia, suggesting that the characteristic does not depend on a specific etiology of dysphagia.

Parameters associated with breathing–swallowing coordination in healthy, dysphagic, and possible dysphagic groups are listed in Table 7. When we put together all swallows from Level 0 and water swallows, the old phase, pause duration, and swallowing latency were significantly larger in dysphagic group, however in possible dysphagic group, only pause duration was significantly larger as compared to healthy group. Further, the frequencies of E-SW-I and I-SW-I patterns were significantly higher in dysphagic group, but not in possible dysphagic group, as compared to healthy group. Timings of swallow in patients with dysphagia were shifted to either early or late timings (Figure 4, Table 8). The difference in the timings of swallow and the I-SW frequency between healthy and dysphagic groups was more marked in water swallows.

We evaluated correlations between parameters associated with breathing–swallowing coordination and those associated with severity of dysphagia to determine whether breathing–swallowing discoordination is deteriorated in proportion to the severity of dysphagia (Table 9). A significant negative correlation was found between MWST and SW-I frequency. MWST of subjects in the present study was either 4 or 5 points, thus the result implies that patients who could swallow twice additionally within 30 s after 3 ml water swallow (5 points) had fewer SW-I patterned swallows. Significant negative correlations were also found between WST and I-SW or SW-I frequency. WST of subjects in the present study was either 1 or 2 points, thus the result implies that patients who could swallow water at once in 5 s had, ironically, a higher chance of I-SW or SW-I swallow patterns as compared to those who needed to swallow more than twice. FILS, RSST, and EAT-10 were not correlated with I-SW frequency or SW-I frequency, indicating that these dysphagia screening tools could not detect breathing–swallowing discoordination.

We performed univariate and multivariate logistic regression analyses in an attempt to discriminate healthy subjects and patients with dysphagia using parameters associated with breathing–swallowing coordination. For the logistic analysis, we introduced a new variable, the frequency of “old phase out of range,” which was defined as the old phase either within the range of early old phase (<0.4) or late old phase (>1.0). We calculated the area under the curve (AUC) of the receiver operating characteristic (ROC) curve to evaluate the model performance for different combinations of parameters. The model parameters were selected from those having a small p-value in univariate regression analysis (Table 10). We found that the combination of SW-I frequency, the frequency of old phase out of range, and pause duration showed the best performance (AUC = 0.730) in multivariate logistic regression analyses.

When an intrinsic oscillator is perturbed by external stimuli, the timing at which a specific phase initiates is shifted (either advanced or delayed). The amount of phase shift depends on the timing and strength of the stimulus; the relationship between timing and amount of shift is termed phase resetting characteristic, which characterizes the oscillator (Glass and Mackey, 1988; Oku and Dick, 1992). Swallows are one of natural perturbations of the respiratory cycle. Paydarfar et al. (1995) investigated the relationship between timing of swallows and phase shift of the respiratory timing, and found that swallows strongly reset the respiratory rhythm (type-0 resetting, Glass and Mackey, 1988), and swallows at the expiratory-to-inspiratory phase transition have the smallest co-phase (i.e., the shortest time to the next inspiration), and swallows at early expiration have the longest co-phase (i.e., the longest time to the next inspiration). Although they did not discuss the frequency of SW-I pattern, the shortest time to the next inspiration for swallows at the expiratory-to-inspiratory phase transition implies an increase in chance of SW-I pattern, which is consistent with the results of the present study.

The major difference between phase resetting by involuntary swallowing and that by voluntary swallowing in the present study is the variability of swallowing latency in voluntary swallowing. It has been reported that the onset of respiratory pause for swallowing is approximately coincident with the bolus propulsion of liquid into the pharynx (Martin-Harris et al., 2005), however, the pause duration is highly variable when eating solid foods (Palmer and Hiiemae, 2003; Matsuo et al., 2008). Even with water swallows, we observed a considerable variability in swallowing latency. The pause in breathing often began substantially before the swallow, and the old phase sometimes exceeded the whole respiratory cycle length. We think that this is because voluntary swallowing is controlled by the cortex, while the occurrence of involuntary swallow in the respiratory cycle is regulated by the interaction between the central pattern generators of breathing and swallowing within the brainstem. We assume that the variability of cortically-controlled swallowing latency causes the scattering of the co-phase plot and breathing–swallowing discoordination.

By using transgenic mice in which Channelrhdopsin-2 (ChR2) or Archaerhodopsin (Arch) is specifically expressed in glycinergic neurons, it has been recently shown that activation of glycinergic neurons in the pre-Bötzinger complex (preBötC) leads to interruption of respiratory activity, whereas silencing of glycinergic neurons in the preBötC induces inspiration (Sherman et al., 2015). Simulation using a neural mass model consisting of five types of respiratory neurons in the Bötzinger Complex (BötC), the pre-BötC, and the rostral ventrolateral respiratory group (rVRG) could reproduce the experimental results (Oku and Hulsmann, 2017). On the other hand, the nucleus tractus solitarii (NTS) is the primary relay nucleus which receives afferent signals from pharyngeal mechanoreceptors and chemoreceptors primarily via the superior laryngeal nerve (SLN) (Jean, 2001; Miller, 2008). What we still do not know is the relay pathway from the solitary nucleus to respiratory neurons in the BötC and preBötC, and the types of glycinergic respiratory neurons that are activated or inhibited by the afferent signals. Neurons orthodromically activated by electrical stimulation of the SLN, which exhibited burst firing at the onset of swallowing, project to NTS, the nucleus ambiguus, the hypoglossal nucleus, the medullary reticular formation, and the dorsal motor nucleus of the vagus (Ezure et al., 1993; Sugiyama et al., 2011). These neurons may be involved in the phase resetting associated with swallowing.

We found a strong correlation between the duration of respiratory pause for swallowing and the swallowing latency. This relationship holds true regardless of the type of food, and regardless of whether the subjects are in healthy, dysphagic, or possible dysphagic groups. We do not know the neural mechanisms of this tight coupling, or mechanisms by which the duration of respiratory pause is regulated. One of the reasons for the strong correlation might be the fact that, as the respiratory pause lengthens, the swallowing latency occupies the majority of the respiratory pause.

As we discussed in the previous subsection, a swallow perturbs the respiratory cycle, and shifts the timing of the following inspiration depending on the timing in the respiratory cycle that the swallow occurred, however, this phase resetting theory does not consider factors associated with swallowing latency and pause duration. Paydarfar et al. (1995) reported that the duration of respiratory pause does not depend on the timing of swallow within the respiratory cycle. However, we found that both swallowing latency and pause duration are weakly correlated with the old phase, the timing of swallow in the respiratory cycle. Our analysis indicates that the swallowing latency and pause duration lengthen as the old phase becomes greater, however, interestingly, swallows with I-SW-E pattern, i.e., with early old phase, also had a longer swallowing latency and longer pause duration. This issue is further discussed below.

We found that the timing of swallows is delayed in patients. Considering that a swallow with a delayed timing tends to be SW-I pattern, we suggest that the increase in SW-I pattern in patients is the consequence of the delayed timing of swallows in the respiratory cycle.

We found that the duration of respiratory pause for swallowing is prolonged in patients with dysphagia. This is consistent with the study of Wang et al. (2015). We also found that the swallowing latency is prolonged in patients, and the pause duration was closely correlated with swallowing latency both in healthy subjects and in patients with dysphagia. It is known that the onset of swallowing reflex relative to a bolus invasion into the pyriform recess is delayed in patients with dysphagia (Miyaji et al., 2012). In addition, the time for food to be propelled into the pharynx may also be delayed in patients with dysphagia. We therefore suggest that the prolongation of the pause duration is a consequence of the delayed onset of swallowing reflex (relative to the respiratory phase) in patients with dysphagia.

Interestingly, we found that the swallowing latency as well as the pause duration are lengthened in swallows with I-SW-E pattern as compared to E-SW-E swallows. Therefore, I-SW-E swallows may be an adaptive behavior to compensate for the delayed onset of swallowing reflex. In contrast, I-SW-I pattern seen in patients with dysphagia would be unsafe swallows, since the time to the next inspiration is much shorter than that in swallows of E-SW-E pattern.

Currently, EAT-10, RSST, and WST are widely used for screening dysphagia. The FILS is generally used to describe the severity of dysphagia in Japan (Kunieda et al., 2013). It is noteworthy in that none of these tests are correlated with either I-SW frequency or SW-I frequency. This might be because patients we recruited had relatively mild dysphagia. Further studies are needed to elucidate whether breathing–swallowing discoordination correlates with the severity of dysphagia. Nevertheless, it should be noted that for patients with mild dysphagia, breathing–swallowing discoordination cannot be predicted by screening tools such as EAT-10, RSST, or WST.

We found that the combination of SW-I frequency, the frequency of old phase out of range, and pause duration was the best logit model to differentiate between healthy and dysphagic subjects. However, AUC of this model was only 0.730, suggesting that there is considerable overlap between healthy and dysphagic groups. Indeed there were a number of healthy subjects who showed a high frequency of I-SW/SW-I patterns. We do not know whether these subjects will manifest symptoms of dysphagia as they become older. Discoordination between breathing and swallowing may be a personal trait, which is modified by age and disease (e.g., stroke, Parkinson's disease, and COPD), and could be corrected by rehabilitation (Martin-Harris et al., 2015).

Although we focused on the discoordination between breathing and swallowing, multiple factors affect the vulnerability of aspiration during swallowing. These include anatomical abnormalities and physiological impairments associated with the triggering of the pharyngeal stage, pharyngeal motor response, and the esophageal stage of swallowing (Lundy et al., 1999). Troche et al. (2010) showed that increasing expiratory muscle strength attenuates aspiration without changing the timing of swallow, presumably due to improved hyolaryngeal complex movement. Therefore, comprehensive measures should be taken to ameliorate the vulnerability for aspiration.

Coughing is another, and probably the most important airway defensive mechanism. The coordination of breathing, swallowing, and coughing has been recently studied by a group of the University of Florida (Pitts et al., 2012, 2013; Bolser et al., 2013). Although we did not record coughs, these studies are a natural extension exploring the mechanisms of airway protective reflexes, and should be further extended to human studies.

We tested only 30 patients with dysphagia. In addition, these patients had diverse etiologies, and were poorly characterized, which might have obscured the unique manifestations of swallowing-breathing coordination in patients with specific diseases or etiologies. Therefore, a future direction of study would be to test a sufficient number of patients with a uniform etiological background.

The authors declare that this study received funding from CareIdo Co., Ltd. and J Craft Co., Ltd. The funders were not involved in the study design or collection, analysis, or interpretation of the data. This study was conducted under industry-academia collaboration contracts among Hyogo College of Medicine, Kyoto University, Kobe University, Hiroshima Prefectural University, Foodcare Co., Ltd., CareIdo Co., Ltd., J Craft Co., Ltd., Murata Manufacturing Co., Ltd., and EuSense Medical Co., Ltd.