1.

Introduction

Most interventions during the early hours after ischemic stroke onset aim to maximize and stabilize cerebral blood perfusion to diminish the neurological damage and improve the long-term outcome.¹ Furthermore, these patients could be at risk of possible detrimental cerebral blood flow (CBF) fluctuations in the case, for instance, of undiagnosed breathing disorders (BDs). Disordered breathing includes a range of respiratory disorders,² often during sleep, characterized by repetitive reduction (hypopneas) or cessation (apneas) of airflow.

Different types of BDs with different physiological origins have been reported in association with stroke in 10% to 70% of the patients.^3–12 However, it is not yet clear if the presence of a BD is associated with the long-term outcome. Although some works support the idea that BDs are associated with poor recovery from stroke,^10,13–16 one study has reported that BDs in stroke patients are associated with early neurological worsening but not with long-term outcome at 6 months.⁴ The reason for the divergence of the results of these studies could be not only the physiological origin of BDs (upper airway obstruction, heart failure, disease, or injury involving the brain)¹⁷ but also the possible, variable amount of detrimental impact of these events on the cerebral hemodynamics. Thus, the diagnosis, the quantification of these respiratory events, and the characterization of their effect on the cerebral hemodynamics may be relevant for the management of the ischemic stroke patients.

The dynamics of local, microvascular CBF are particularly interesting as, hypothetically, they could reveal that large, repeated changes in the cerebral perfusion could be detrimental when cerebral autoregulation is damaged in the acute stages of stroke.¹⁸ This is relevant during these acute stages as ischemia in the brain is a potentially reversible process dependent on the restoration of the microvascular CBF in the penumbra within a time window of cellular viability that varies on duration and severity of CBF cessation.¹⁹ However, if cerebral autoregulation is damaged, CBF can abruptly increase or decrease without following the needs of the brain and without restoring the needed CBF in the penumbra. If so, BD treatment could be used to avoid these unwanted changes. However, there are no established methods to monitor microvascular CBF at the bedside.

In non-stroke subjects, cerebral hemodynamic fluctuations due to apneas or hypopneas have been extensively studied and characterized previously by several groups by measuring the cerebral blood flow velocity (CBFV) in the middle cerebral artery by transcranial Doppler ultrasound (TCD).^20–24 However, macrovascular CBFV is only an indirect measurement of what happens in the distal microcirculation.

Fluctuations of cerebral hemodynamics have previously been observed in acute stroke patients. A study of nine patients with acute stroke²⁵ reported arterial desaturations from unknown origin that correlated with cerebral hemodynamics measured with near-infrared diffuse optical spectroscopy (NIRS-DOS), and showed an asymmetry of the changes in the two cerebral hemispheres. Pizza et al.²⁶ observed in seven patients with acute/subacute middle cerebral artery stroke that obstructive and central apneas were linked to cerebral oxygenation changes by NIRS-DOS. Asymmetrical patterns of cerebral hemodynamic changes were again found with significantly larger changes on the unaffected hemisphere compared to the affected one. However, neither NIRS-DOS nor TCD provide a direct measure of microvascular CBF.

In this pilot study, we have utilized diffuse correlation spectroscopy (DCS), which uses the speckle statistics of near-infrared diffuse light to noninvasively monitor local, microvascular CBF.^27–29 DCS does not need any exogenous tracers and has been validated against different modalities.^29,30 In previous studies,³¹ we have observed unexpected fluctuations of microvascular CBF, heart rate (HR), arterial oxygen saturation (${\mathrm{SpO}}_2$), and end-tidal carbon dioxide in three (18%) of the seventeen acute ischemic stroke (AIS) patients during the first week from stroke onset, which has motivated the design of this study to investigate the origin of these fluctuations. Here, we have aimed to test the hypothesis that these cerebral and systemic fluctuations were associated with apneic and hypopneic events after stroke.

2.

Methods

This study was conducted at the Stroke Unit of Hospital de la Santa Creu i Sant Pau, in Barcelona, Spain, from August 2016 to March 2017. The study protocol was approved by the local ethical committee (EC/15/130). Either the patients or their relatives gave their informed written consent.

The inclusion criteria were being older than 18 years, confirmed ischemic stroke in the anterior circulation, measurement time $\le 7$ days from the stroke onset and written informed consent. The exclusion criteria were resting HR $<40$ or $>110$ beats per minute (bmp), ${\mathrm{SpO}}_2$ $<92\%$, use of oxygen therapy, symptomatic lacunar stroke, previous diagnosis of BDs, previous transient ischemic attack, minor stroke, or intracranial hemorrhage.

The study included two different parts (see Fig. 1, for a flow chart of the procedure). First, patients who fit the clinical inclusion criteria were screened with a pulse oximeter for a minimum of 4 h between 8 am and 2 pm. Afterward, if the pulse oximeter showed periodic decreases of $\ge 3\%$ in ${\mathrm{SpO}}_2$, the patients were measured in the afternoon with DCS and also with a respiratory polygraphy device simultaneously for 1 h in the most comfortable head-of-bed position for the patients.

Fig. 1

Methodology of the two parts of the study. The subjects included in each part are depicted in gray. ${\mathrm{SpO}}_2$, arterial oxygen saturation.

2.3.

Statistical Analysis

To calculate the percentage of CBF change ($\mathrm{\Delta }\mathrm{rCBF}$) due to each apneic or hypopneic event, we have estimated a basal BFI (${\mathrm{BFI}}_{\mathrm{bl}}$) as the average of the cerebral BFI from 30 s before the apnea/hypopnea start up to 30 s after the event end. This was considered to correct for slight changes in the probe position during the measurement. Afterward, the CBF changes were calculated as $\mathrm{\Delta }r\mathrm{CBF}=(\frac{\mathrm{BFI}}{{\mathrm{BFI}}_{\mathrm{bl}}}-1)\times 100$. Similarly, $\mathrm{\Delta }\mathrm{HR}$ was defined as $\mathrm{\Delta }\mathrm{HR}=\mathrm{HR}-{\mathrm{HR}}_{\mathrm{bl}}$ and $\mathrm{\Delta }{\mathrm{SpO}}_2$ was defined as $\mathrm{\Delta }{\mathrm{SpO}}_2={\mathrm{SpO}}_2-{\mathrm{SpO}}_{2\mathrm{bl}}$. In all cases, the ${\text{subscript}}_{bl}$ indicates an average over the same period.

Afterward, the events were parametrized for the characterization of the induced $\mathrm{\Delta }\mathrm{rCBF}$, $\mathrm{\Delta }\mathrm{HR}$, or $\mathrm{\Delta }{\mathrm{SpO}}_2$ for each event. The goal was to study if individual apneic or hypopneic events were associated with significant increases or decreases in these variables as it was hypothesized. This step consisted in considering each event as a function dependent on time [$\mathrm{\Delta }\mathrm{rCBF}(\mathrm{t})$, $\mathrm{\Delta }\mathrm{HR}(\mathrm{t})$, and $\mathrm{\Delta }{\mathrm{SpO}}_2(\mathrm{t})$], and then, the extrema of these functions along a specific time interval were calculated. The reference time (time zero) was considered as the end of each event. The time windows to find these extrema were considered from $-5$ to 15 s for the first extremum on $\mathrm{\Delta }\mathrm{rCBF}$ (see Fig. 2, as an example), from 0 to 15 s for the $\mathrm{\Delta }\mathrm{HR}$, and from 5 to 35 s for the $\mathrm{\Delta }{\mathrm{SpO}}_2$. These time windows were chosen by visual observation of all the apneas and hypopneas plotted together from $-30$ to 60 s and according to the literature.^20,41 These analysis methods have previously been verified in nocturnal measurements of chronic obstructive sleep apnea patients.⁴²

Fig. 2

Cerebral blood flow characterization of individual sleep events, where the apnea end is considered as a pivot point. The light gray region indicates the sleep event. The dark gray region indicates the time window used for finding the first extremum value. The first and second extrema are labeled in blue.

Averages and standard deviations have been performed considering all apneic and hypopneic events for each patient and also for the population. The Wilcoxon signed-rank test was used to assess the difference between ipsilesional and contralesional $\mathrm{\Delta }\mathrm{rCBF}$ responses. The same test was used to assess if the apneic/hypopneic response of different variables was significantly different from zero. All analyses were performed with MATLAB (Mathworks, Natick, Massachusetts). A $p$-value $<0.05$ was considered to be statistically significant.

3.

Results

During the 8-month period, we have screened 28 ischemic stroke patients with pulse oximetry, of whom 5 ($n=5$, 18%) presented fluctuations ($\ge 3\%$) in ${\mathrm{SpO}}_2$ and were then included in the second part of the study. Table 1 shows the clinical characteristics of these five patients. Four were male, had an age of $81\pm 6$ years ($\text{mean}\pm \text{standard deviation}$), an admission NIHSS of 8 (2, 16) [median (first quartile, third quartile)], and a time from stroke onset to measurement of 4 (3, 5) days.

Table 1

Demographics and prestroke clinical characteristics (top) and poststroke clinical characteristics (bottom) of the five ischemic stroke patients.

Table 2 shows the results of the DCS and the respiratory polygraphy measurements. In all five patients, the respiratory polygraphy confirmed the presence of altered breathing patterns; two presented periodic breathing [cases 1 and 3; clusters of breaths separated by intervals of apnea (no breathing) or hypopnea], two presented a predominance of obstructive and hypopneic events (cases 2 and 4; the airflow ceased or partially ceased due to upper airway obstruction), and one presented central apneas (case 5; the airflow ceased or partially ceased due to diminished or absent effort to breath).

Table 2

Polygraphic characteristics (top) and maximum CBF, HR and SpO2 changes due to each apneic/hypopneic event (bottom) of each patient.

Figure 3 shows a representative trace (from case 4) of 10 min of continuous oronasal flow, abdominal and thoracic movements, HR, ${\mathrm{SpO}}_2$, and CBF changes. This specific interval shows two types of events with good signal-to-noise ratio in the optical data and a “close-up” revealing details. Both apneic (four obstructive apneas, green shading) and hypopneic (three, yellow shading) events are visible and the simultaneous monitoring shows time correlated variation of microvascular CBF, HR, and respiratory variables. From the polygraphy recording, oronasal flow appears to be reduced during the hypopneas and ceases during the obstructive apneas, whereas abdominal and thoracic movements are reduced during these events. HR does not follow a general pattern for this trace. ${\mathrm{SpO}}_2$ shows a drop with a delay relative to the apneic/hypopneic event. In this time period with frequent events, as commonly observed, the ${\mathrm{SpO}}_2$ drop of the previous event is overlapped with the next event. It can also be observed that the CBF reaches its maximum at the end of an event. Finally, for this specific interval, periodic breathing is diagnosed but the rest of the acquisition is dominated by obstructive apneic events, as shown in Table 2.

Fig. 3

A 10-min trace of systemic fluctuations of physiological variables alongside the fluctuations of CBF. Red and black CBF traces correspond to the ipsilesional and contralesional hemispheres, respectively. Hypopneas (yellow) and obstructive apneas (green) are shown from the measurements of an 89-year-old male with a left middle cerebral artery ischemic stroke (case 4). The patient presented a left middle cerebral artery stroke with insular, temporal, and basal ganglia ischemic areas. The NIHSS score was 16 when measured. The patient died after 24 h of the measurement.

On the contrary, Fig. 4 shows 10 min of similar traces where neither fluctuations nor apnea/hypopnea events are found. This patient (case 3) is the only one with no apneic/hypopneic events during those 10 min.

Fig. 4

A 10-min trace with no typical fluctuations of cerebral blood and no apneas or hypopneas. Red and black CBF traces correspond to the ipsilesional and contralesional hemispheres, respectively. Measurements of an 82-year-old male with a right anterior cerebral artery ischemic stroke (case 3) are provided. The patient had a blockage of the right anterior cerebral artery with a frontal ischemic area. The NIHSS score was 8 when measured.

For completeness, the full data set for the five patients can be found in Sec. 5 Appendix.

From the five patients, 239 ($n=239$) apneic/hypopneic events were analyzed. In all patients, the CBF changes were biphasic. In other words, significant bilateral CBF increases of $27.1\%\pm 17.7\%$ ($p<0.001$) for the ipsilesional and $29.0\%\pm 17.4\%$ ($p<0.001$) for the contralesional hemispheres were followed by decreases of $-19.3\%\pm 9.1\%$ ($p<0.001$) for the ipsilesional and $-21.0\%\pm 8.9\%$ ($p<0.001$) for the contralesional hemispheres were found due to each apneic/hypopneic event. There was no interhemispheric difference between the maximum increases ($p=0.189$) or the maximum decreases ($p=0.053$). However, this result suggested a trend toward a potential difference between hemispheres, as it has been previously found by NIRS-DOS,^25,26 as mentioned in Sec. 1. Also, as expected, HR increased ($p<0.001$) by $6.5\pm 6.8$ bpm and ${\mathrm{SpO}}_2$ decreased ($p<0.001$) by $-4.9\%\pm 16.8\%$.

The two patients (cases 4 and 5) with the highest AHI, one with predominance of an obstructive apnea pattern and the other with predominance of a central apnea pattern, had the two highest NIHSS both on admission and after the measurement and the lowest ASPECTS. Unfortunately, both had a mRS of 6, i.e., they had died during the 3 months after stroke onset. Of these two patients, the one with the highest AHI (case 5) also presented the lowest left ventricular ejection fraction. However, neither of these two patients presented the highest physiological or cerebral hemodynamic changes in response to the respiratory events. One of the characteristics of the remaining three patients (cases 1, 2, and 3) with lower stroke severity was that they presented mainly hypopneic events, in comparison to the obstructive and central apneas found in the two most severe patients (cases 4 and 5). We note that, given the small population, these were just anecdotal statements and should not be considered generalizable.

4.

Discussion

In this study, we have followed up on our previous observation of unexpected bilateral fluctuations in microvascular CBF in patients with anterior, cerebral ischemic stroke that were observed in different protocols with DCS measurements.³¹ We have hypothesized that these fluctuations were due to BDs, in particular, due to apneic and hypopneic events. Therefore, a protocol based on prescreening for these events with a pulse oximeter was introduced to enroll patients without a previous diagnosis of BDs for follow-up study combining respiratory polygraphy and DCS. A careful evaluation was carried out to identify the origin of these fluctuations and to characterize their details by combining this recording of different respiratory variables, ${\mathrm{SpO}}_2$, and HR. Apneic and hypopneic events were determined to be associated with the effect of cerebral hemodynamics and this effect was characterized.

As observed by the full data set in the Sec. 5 Appendix, the precise nature of the cause of the microvascular CBF changes is not absolutely clear from this study, i.e., some CBF fluctuations lead the apnea/hypopnea, whereas others follow it. This may be partially due to errors in the precise synchronization of the measurements but it may also have further implications indicating secondary events that may have caused respiratory effects, such as cortical spreading depression^43–45 in the brain stem.^46,47 However, our study did not gather definitive evidence (e.g., recordings at multiple locations to demonstrate any spreading activity and/or recordings of the electrical activity) to accept or refute this possibility. These results have encouraged us to look further and deeper into this issue in future studies by combining multimodal, precisely synchronized diffuse optical recordings with polygraphy, capnography, and electroencephalography, and a larger head coverage with multiple source and detector pairs.

By the polygraphy, HR and ${\mathrm{SpO}}_2$ changes were found during these events. These HR and ${\mathrm{SpO}}_2$ changes followed the apneic/hypopneic dynamics expected by the literature in patients with BDs,^{20,24,41,48,49} showing an increase during the event or at the end of it for the HR, and a decrease after the end of the event for the ${\mathrm{SpO}}_2$. No periodicity in the respiratory changes was found, as shown in the full data set in Sec. 5 Appendix, as the periodicity depended on the pathogenic mechanism, on the sleep stage, and on the position, among others.⁵⁰

In non-stroke patients with BDs, several studies have reported CBFV changes measured by TCD as showing an increase close to the end of the apnea.^20,23,24 Our findings are similar; we have observed a CBF increase of $27.1\%\pm 17.7\%$ and $29.0\%\pm 17.4\%$ for the ipsilesional and contralesional hemispheres, respectively, and a decrease of $-19.3\%\pm 9.1\%$ and $-21.0\%\pm 8.9\%$ for the ipsilesional and contralesional hemispheres, respectively, close to the end of the apneic/hypopneic events. This increase is similar to the CBF increase of $30\%\pm 17\%$ and the decrease of $-20\%\pm 12\%$ described previously by us using DCS in obstructive sleep apneic events.⁴² It is also consistent with the CBFV of 22% to 42% increase found by Alex et al.²³ and also to the $14.6\%\pm 14\%$ increase right after the apnea end by Bålfors and Franklin.²⁰ However, other authors have found larger CBFV increases.^21,24

These results show that there is a decrease in cerebral perfusion due to an apneic or hypopneic event. If these intermittent decreases lead to ischemia, they can cause hypoxic/ischemic brain injury, especially if cerebrovascular reactivity and regulation are impaired, which is often the case in ischemic stroke patients.^51,52 Consequently, this could lead to a worsened prognosis, which may call for interventions such as continuous positive airway pressure (CPAP) therapy. We note that, to date, no direct link among a degree of microvascular CBF changes, the long-term outcome, and the use of CPAP therapy has been studied.

Several works have studied the general effect of CPAP therapy, mainly in obstructive disorders or adaptive servo-ventilation (ASV) treatment for central disorders to avoid apneas and hypopneas.^53–55 In stroke patients, whereas some authors support the benefits of the CPAP treatment on BDs,^15,56–61 others claim that these benefits are not clear^62,63 or that there is no benefit at all.⁶⁴ Nevertheless, as mentioned earlier, BDs diagnosis and CPAP or ASV treatment are not the common practice in the stroke units.

Future work with a larger sample size to investigate if the CPAP or ASV could be effective in those patients who show the presence and/or large fluctuations in CBF (e.g., above a threshold) in relation to apneic/hypopneic events is warranted. In addition, it would be of interest to study if all ischemic stroke patients with apneic/hypopneic events show similar CBF fluctuations, or if there is an interindividual variability in the response to apneas or hypopneas, and, consequently, the response to treatment may be different.

In our cohort, the two patients (cases 4 and 5) with the highest AHI were those with the worse stroke severity (higher NIHSS score) at the moment of the measurement. This has been previously reported in patients with stroke showing central BDs.¹³ Similarly, the ASPECTS score of these two patients were the lowest of the group, which was compatible with the idea that large acute cerebral hemispheric lesions were associated to BDs.^7,65 One of these two severe patients (case 5; case 4 had no information available) presented heart failure [low left ventricular ejection fraction ($<40\%$)] and mainly central events, as it has been found in stroke patients in association with central BDs.⁶ The BD present in this patient could be related to it. Insular and frontal ischemic regions were present in our cohort and, interestingly, these regions have already been associated to BDs of central origin.^13,65,66 However, other works have supported the idea of the lack of an association between the presence of BDs of central origin and the location of stroke.^5,9,11 Nevertheless, more patients are needed to clearly identify clinical associations.

Our proof-of-concept study has some limitations, including the small number of patients included in the final analysis, the different types of possible BDs with different physiological origins, and the short recording time of DCS simultaneously with respiratory polygraphy. Future studies will add more subjects and will measure these subjects during several hours continuously in order to better identify the type of BD, and moreover, to link it to the degree of the cerebral hemodynamic responses, among other previously mentioned goals.

Another limitation is that only one measurement per patient has been performed for this study, so we could not track the start and cessation of the fluctuations together with the clinical stroke evolution. Anecdotally, in our previous protocols,^31,67 the physiological fluctuations observed have prevailed in the different measurements performed in each patient during the first week after stroke. Accordingly, it is expected that BDs could be present during the first week after stroke in the case of patients with BDs of central origin, as BDs have been found to resolve after several weeks.⁶⁶ Future studies should include daily measurements to track when these fluctuations cease and to relate it to clinical parameters.

Owing to the small number of subjects considered for this analysis, we have avoided a group analysis. Two important parameters to understand CBF changes are the blood pressure and the end-tidal carbon dioxide; future studies should include their simultaneous measurement with DCS. Different studies have suggested these two parameters as different primary drivers to the CBFV changes during apneas. Przybylowski et al.⁶⁸ suggested that carbon dioxide is the primary regulator, and not the blood pressure, of the rise in CBFV (measured by TCD) during apnea and in the reduction of CBFV in the following apnea. On the contrary, Yaggi and Mohsenin⁶⁹ suggested that if autoregulation is impaired or the changes in blood pressure are sufficiently rapid that an intact autoregulation does not have time to adapt, then the changes in TCD velocity could be related to the changes in blood pressure. However, these studies only focus on the macrocirculation level; on the microcirculation level, these parameters may have a different influence on microvascular CBF. Moreover, hypoxia has been found during apneic events,^70,71 which could also contribute to the local microvascular change.

Finally, diffuse optical spectroscopy techniques not only provide information from the brain but also contain scalp and skull contributions.^29,72 In this work, instead of including a small source–detector separation (1 cm) to study the superficial effects, we have kept all detectors for large separations to focus on improving the signal-to-noise ratio, as a source–detector separation of 2.5 cm has been found to be a good compromise.^29,30,73 Future implementations could utilize more source–detector pairs as well as a pressure modulation protocol to minimize this effect.⁷⁴

In conclusion, DCS has successfully revealed bilateral fluctuations in CBF of ischemic stroke patients with undiagnosed BDs, in accordance with systemic fluctuations of physiological variables. These cerebral hemodynamic fluctuations are related to apneic and hypopneic events in the five patients tested. Further studies are needed to study the possible detrimental impact of the magnitude and presence of these cerebral hemodynamic fluctuations, especially in the early hours after stroke where the interventions are more relevant to the outcome, for the recovery of stroke patients.

5.

Appendix

The full data set for the five patients measured is shown in Figs. 5Fig. 6Fig. 7Fig. 8–9.

Fig. 5

The full trace recording of case 1 showing periodic respiration, with mainly hypopneic events. Hypopneas are depicted in yellow and central apneas in light blue. Red and black CBF traces correspond to the ipsilesional and contralesional hemispheres, respectively.

Fig. 6

The full trace recording of case 2 showing only hypopneic events (in yellow). Red and black CBF traces correspond to the ipsilesional and contralesional hemispheres, respectively.

Fig. 7

The full trace recording of case 3 showing periodic respiration, with mainly hypopneic events. Hypopneas are depicted in yellow, central apneas in light blue, obstructive apneas in green, and mixed apneas in magenta. Red and black CBF traces correspond to the ipsilesional and contralesional hemispheres, respectively.

Fig. 8

The full trace recording of case 4 showing both obstructive (in green) and hypopneic (in yellow) events. Red and black CBF traces correspond to the ipsilesional and contralesional hemispheres, respectively.

Fig. 9

The full trace recording of case 5 showing central apneic events (in light blue). Obstructive apneas are depicted in green. Red and black CBF traces correspond to the ipsilesional and contralesional hemispheres, respectively.