1 INTRODUCTION

Sickle cell disease (SCD) is an inherited blood disorder that results from an amino acid substitution in the β chain of hemoglobin, producing sickle hemoglobin (HbS).¹ So, HbS polymerizes a short time after it releases oxygen to tissues, causing the normally flexible red blood cells (RBC) to become rigid and take on a sickle-shape. Red blood cells need to bend in order to traverse the small diameter capillaries in the microvasculature. If the flow rate is slower than the time to polymer formation, the RBC are not able to exit from the narrow microvasculature into a larger vessel before this flexible-to-rigid transformation occurs, and they become lodged in the microvasculature, thereby obstructing blood flow. Thus, any factor that decreases microvascular flow could promote vaso-occlusion.² The clinical manifestation of extensive obstruction of microvascular flow is a clinically perceptible painful vaso-occlusive crisis (VOC), the hallmark of SCD.

While much of the literature in SCD has focused on nitric oxide (NO)-mediated vascular dysfunction and cellular adhesion in the post-capillary venules as key factors that may decrease microvascular flow and promote vaso-occlusion, the role played by the sympathetic and parasympathetic nervous systems in causing vasoconstriction in the pre-capillary arteriole, and decreasing microvascular flow has only recently been considered.³ Our prior work has demonstrated that anticipatory pain,⁴ experimental mental stress,⁵ and cold exposure⁶ all cause significant global vasoconstriction in SCD and non-SCD controls and these responses vary significantly among all individuals.³ Sighs (deep breaths) trigger peripheral vasoconstriction frequently in SCD but rarely in non-SCD indivuals,⁷ and cold face stimulation reduces the sensitivity of the baroreflex control of peripheral vascular resistance more in non-transfused than transfused individuals with SCD or in controls.⁸ Taken together, these data indicate that individuals with SCD have hyperactive autonomic nervous system responses.

Monitoring vasoconstriction during sleep offers an excellent, relatively well controlled opportunity to study the sympathetic nervous system activity, which causes vasoconstriction.⁹ Transient arousals from sleep represent the strongest form of sympathetic stimulus that occurs naturally and frequently in all humans.¹⁰ These fluctuations in sympathetic output lead to corresponding fluctuations in cardiovascular activity and peripheral blood flow.⁹ In the general population, the cardiovascular responses to these arousals include brief, but potent increases in respiratory effort, heart rate and blood pressure.¹¹ Blasi et al.¹² showed that the arousal-induced reductions in parasympathetic nervous system activity, which opposes the sympathetic nervous system, are transient and strongly coupled with respiration. However, the sympathetic effects last significantly longer and may accumulate if repetitive arousals occur in close succession. Periodic limb movements, which happen commonly in children with SCD,¹³ also trigger sympathetic nervous system (SNS) responses, leading to transient increases in heart rate and blood pressure during sleep.¹⁴ Given the strong biological premise that deoxygenation of HbS is the basis of the pathogenesis of SCD, our group and others have tested the hypothesis that nocturnal hypoxia is associated with an increase in the incidence rate of severe vaso-occlusive episodes.¹⁵ We showed that experimental hypoxia caused parasympathetic withdrawal only in SCD individuals, but did not decrease perfusion in SCD or control individuals.⁷ Furthermore, we showed in the rigorous multi-center prospective NIH-funded Sleep and Asthma Cohort (SAC) study in children with SCD, that there was no relationship between nocturnal decreased oxygen saturation or other standard sleep-related parameters, and increased incidence rate of severe vaso-occlusive pain eposides.¹⁶

In this report, we analyzed the photoplethysmography signal recorded during a standard sleep study, along with the clinical laboratory data and rate of severe sickle pain crisis obtained as part of the SAC study.¹⁶ This was to test the hypothesis that the magnitude of the peripheral vasoconstriction that occurred naturally during sleep would predict frequency of severe vaso-occlusive pain episodes. In doing so, we tested in humans the corollary of the Eaton-Hofrichter hypothesis² that decreased microvascular perfusion relative to the delay time for HbS polymerization, in this case due to greater levels of peripheral vasoconstriction, would result in an increased likelihood of vaso-occlusion and thus higher incidence of VOC. To test the direction of causality, we also determined whether the rate of severe vaso-occlusive pain episodes prior to the sleep study influences the nocturnal vasoconstriction parameters.

2 METHODS

2.2 Data analysis

The information used for the detection and quantification of peripheral vasoconstriction episodes during sleep was derived from two channels of the polysomnogram recordings: (a) the sequences of R-peaks extracted from the electrocardiogram (ECG), from which we determined the R-to-R interval (RRI) on a beat-to-beat basis over the entire sleep duration; and (b) the finger photoplethysmogram (PPG) waveform acquired from the non-oxygen sensitive component of the optical signal from the pulse oximetry sensor (Masimo SET [v2], Irvine, CA). From each PPG pulse, we derived the peak-to-trough amplitude (PPGa), which reflects the pulsatile change in blood volume caused by arteriolar blood flow in the fingertip.^{17, 18} The PPGa is largely modulated by the autonomic nervous system on a beat-to-beat basis.¹⁹ Thus, the relative changes in PPGa were used to quantify the degree of peripheral vasoconstriction/vasodilation at any given time.

The PPG recording in each polysomnogram was visually examined to exclude segments with obvious artifacts due to motion, interruptions to the sleep study, signal clipping or other abrupt alterations in gain of the channel that made the signal quality inadequate for analysis. Using the algorithm that we developed (supplemental materials), the magnitude of each vasoconstriction event detected during the overnight recording of the PPGa signal was calculated.

Figure 1A displays an example of such a vasoconstriction episode, immediately following an arousal from sleep (shown by the red vertical line at time zero). As is the case with most arousals, there is an increase in heart rate indicated by a decrease in R-to-R interval (Figure 1A, top tracing). For each identified vasoconstriction episode, the algorithm generated the following quantitative descriptors: (a) vasoconstriction duration (T_vasoc), for example, the time from the onset to the end of vasoconstriction; (b) vasoconstriction “area” (A_vasoc), computed by integrating the difference between the PPGa and the pre-vasoconstriction baseline over T_vasoc; and (c) vasoconstriction magnitude (M_vasoc). Note, M_vasoc is calculated by dividing A_vasoc by T_vasoc for that particular vasoconstriction episode and represents the average reduction in PPGa over T_vasoc. For each participant, the algorithm also computed the frequency of vasoconstrictions per hour (f_vasoc). Thus, the median values of these extracted descriptors over the total duration of sleep for each participant represent the basal vasoreactivity and characterize the nocturnal vasoconstriction phenotype of each participant

Autonomic balance based on heart rate variability was determined by computing the following standard indices for each individual based on the electrocardiogram obtained during the polysomnogram: (a) the root mean square differences of successive R-R intervals (RMSSD), which assesses parasympathetic (vagal) modulation of heart rate; and (b) the ratio of low-frequency power to high-frequency power of heart rate variability (LHR_RRI), which represents the balance between sympathetic and vagal control of heart rate.²⁰ We also used other standard sleep study indices available from the centrally adjudicated polysomnograms in the SAC clinical datasets.¹⁶ These included the full night obstructive apnea-hypopnea index (OAHI = number of apneas or hypopneas per hour of sleep), arousal index (AI = number of arousals per hour of sleep), and the number of periodic limb movements per hour of sleep (PLMI = periodic limb movement index) as potential predictors of post-PSG pain rate.

Severe pain episodes, as defined in the SAC study,¹⁶ were used to calculate pain rates. Post-PSG pain rate for any given participant was calculated as the number of SCD-associated pain episodes per year severe enough to require hospital admission and treatment with opioids that occurred between the date of the PSG and the end of follow-up (T_post-psg, years), divided by the duration of follow-up (T_post-psg). Pre-PSG pain rate for the same participant was calculated as the number of severe SCD-associated pain episodes per year, as defined above, occurring between birth or T_post-psg years before PSG, whichever is smaller, and the date of the PSG (T_pre-psg, years), divided by T_pre-psg. More details on the severe pain episode definition and quality control in the SAC study¹⁶ are provided in the supplemental methods.

3 RESULTS

4 DISCUSSION

Over four decades ago, Eaton and Hofrichter postulated that the relation between retention of RBC in the microvasculature as a consequence of decreased regional blood flow and time to flexible-to-rigid transformation of the HbS-containing RBC, is the fundamental mechanism of vaso-occlusive crisis.² Thus, two basic factors determine whether vaso-occlusion occurs: (a) the delay time between deoxygenation and polymerization of HbS; and (b) how fast the RBC are moving through the microvasculature.²² If the microvascular blood flow decreases significantly, whether from adhesion in the postcapillary venules or constriction of pre-capillary arterioles, such that the time to transit the capillary network and to reach a vessel larger than the RBC diameter is longer than the time to HbS polymerization, the RBC become rigid in small vessels and lodge, thereby occluding flow. The vast majority, if not all, of the proposed mechanisms for SCD pathology to this day involve delay time to HbS polymerization or microvascular transit time as the final common path to VOC.

In the course of a number of studies in humans using peripheral perfusion as a biomarker to understand the triggers of VOC, we have shown that cold,⁸ mental stress,⁵ respiration,⁷ and pain itself,^{4, 23} all factors mentioned by patients to precede VOC, cause significant peripheral vasoconstriction, and that each study participant, both SCD and controls, seems to have a characteristic baseline vasoconstriction pattern that was observed over several experimental visits.⁶ We suspected that individuals with high inherent vasoconstriction would have more frequent VOC².

The present data based on the amplitude of the PPG signal from a commonly used clinical sensor provides a noninvasive means of quantifying the magnitude of sympathetically mediated constriction of the arterioles upstream of the capillary, thereby enabling the tracking of periods when microvascular blood flow is significantly reduced. The median value of the vasoconstriction magnitude parameter, M_vasoc, measured over several hours during sleep clearly reflects the magnitude and duration of vasoconstriction, as seen when comparing Figure 1B (low M_vasoc) to Figure 1C (high M_vasoc). Importantly, based on a standard definition of severe SCD VOC events that was adopted by the SAC, vasoconstriction magnitude quantified by median M_vasoc strongly predicts subsequent severe VOC rate. Obviously, individuals with SCD experience many more VOC events that are less severe and do not result in hospitalization. This information was not available in the SAC dataset. Nevertheless, it is clear that sympathetic nervous system-mediated vasoconstriction predicts future frequency of severe VOC events. Furthermore, the fact that M_vasoc is strongly related to incidence of VOC over a median of 5 years after the polysomnogram is consistent with the vasoconstriction magnitude being a characteristic of the individual that persists over time.

An implicit assumption in our model is that high propensity to peripheral vasoconstriction leads to high frequency of future VOC episodes. Could past VOC-related pain occurring before the polysomnogram increase peripheral vasoconstriction instead? When we analyzed the relationship between median M_vasoc and the rate of severe pain events measured in the period preceding the polysomnogram, we found no association between pre-PSG pain rate and median M_vasoc (Table 3), providing strong support for the direction of causality, and suggesting that high vasoconstriction magnitude predisposes to frequent VOC, and not the other way around. Taken together, these data suggest that autonomic-mediated vasoreactivity plays a major role in promoting VOC in SCD and provides the first experimental evidence in humans validating the Eaton-Hofrichter model² on a macro scale.

We have previously shown that experimentally induced hypoxia causes loss of parasympathetic modulation of heart rate in SCD individuals, but not in controls. Importantly, transient hypoxia does not decrease microvascular flow in SCD individuals.⁷ Consistent with this finding, Willen et al.¹⁶ showed no significant association between low mean nocturnal SpO₂ and incidence of VOC in the SAC participants studied here. On the other hand, we^{5, 7, 8, 24, 25} and others^26-29 have shown significant dysautonomia in SCD. Therefore, we examined measures of cardiac ANS balance (LHR, RMSSD) but found no association with VOC rate (modesl 8-9 Table S2). In contrast, parasympathetic modulation of heart rate (RMSSD) was negatively correlated with OAHI (Table S4), while the sympathetic stimuli, OAHI, AI, and PLMI were all associated with increased median M_vasoc (models 3-5; Table 3). These findings show that sympathetic stimuli during sleep cause vasoconstriction as measured by M_vasoc and are consistent with existing knowledge about the detrimental effects of sleep-disordered breathing on autonomic and cardiovascular function.^{9, 30} These data raise the possibility that the net sympathetic nervous system activity quantified by M_vasoc derived from PPG signals may be the common path that links sleep-disordered breathing and cardiovascular disease.^{31, 32}

The PPG signal is recorded ubiquitously in sleep studies for monitoring SpO₂ and pulse rate, but changes in the magnitude of the PPG are ignored. Essentially, every wearable device that reports cardiac activity uses the PPG signal to measure heart rate but does not utilize the changes in PPG amplitude. Adaptation of these devices to measure M_vasoc augurs well for future large-scale SCD population studies, in particular to determine if transient increases in M_vasoc predict VOC in the immediate future. In fact, the recognition that the vasoreactivity detected by the PPG signal represents an easily accessible and readily quantifiable biomarker of sympathetic nervous system activity and tissue perfusion likely has implications for the study of human disease far beyond SCD.

It is important to emphasize that we do not argue that the distinct episodes of vasoconstriction we observed during sleep would predict immediate transition into VOC. Instead, the current data add neural-mediated vasoconstriction to the list of events than can increase transit time and the likelihood of more frequent severe VOC. We believe that the magnitude of the decrease in PPGa depicted in Figure 1 represents the collective effects of all inputs causing vasoconstriction, including modulation of the sensitivity of the vessel to constriction by nitric oxide or endothelin-1, the resistance to outflow by cellular adhesion in the postcapillary venules and the sympathetic nervous system triggered vasoconstriction, all of which contribute to the increase in microvascular transit time that promotes vaso-occlusion. Furthermore, the data showing that the vasoconstriction status at one point in time predicts the frequency of VOC over the next year or more suggest that the propensity to vasoconstrict measured by median M_vasoc is an inherent, or at least lasting, characteristic of each individual. And, that those who have more intense, more frequent, and more prolonged vasoconstriction events are more likely to have more frequent severe VOC.

The present data reveal for the first time that peripheral vasoconstriction strength, as a surrogate for microvascular transit time, predicts increased likelihood of severe sickle pain crisis frequency. The fact that individuals have intrinsic vasoreactivity phenotypes that can be identified by analysis of the PPG suggests that vasoreactivity is a biophysical biomarker for sympathetic balance that may have a role in predicting clinical outcomes in SCD as well as other sympathetic nervous system-related vascular disorders. Autonomic balance and vasoreactivity likely contribute to the known variability in the incidence rate of severe VOC pain and represent a plausible target for future therapeutic intervention.

CONFLICT OF INTEREST

The Aauthors declare no conflict of interest.

AUTHOR CONTRIBUTIONS

TC conceived the project and approach; PC, MK developed the key parameters and analysis approach; PC, MK analyzed the data; PC, MK, TC interpreted the data; YJ, PC, MK, CR, MD developed the data extraction approach; MD was the lead PI on the SAC study, YJ extracted and assembled the data; PC, MK, TC wrote the paper; PC, MK,TC, CR, MD edited the paper.