Heart and Circulatory Physiology

Chronic intermittent hypoxia in humans during 28 nights results in blood pressure elevation and increased muscle sympathetic nerve activity

G. S. Gilmartin, M. Lynch, R. Tamisier, J. W. Weiss


Chronic intermittent hypoxia (CIH) is thought to be responsible for the cardiovascular disease associated with obstructive sleep apnea (OSA). Increased sympathetic activation, altered vascular function, and inflammation are all putative mechanisms. We recently reported (Tamisier R, Gilmartin GS, Launois SH, Pepin JL, Nespoulet H, Thomas RJ, Levy P, Weiss JW. J Appl Physiol 107: 17–24, 2009) a new model of CIH in healthy humans that is associated with both increases in blood pressure and augmented peripheral chemosensitivity. We tested the hypothesis that exposure to CIH would also result in augmented muscle sympathetic nerve activity (MSNA) and altered vascular reactivity contributing to blood pressure elevation. We therefore exposed healthy subjects between the ages of 20 and 34 yr (n = 7) to 9 h of nocturnal intermittent hypoxia for 28 consecutive nights. Cardiovascular and hemodynamic variables were recorded at three time points; MSNA was collected before and after exposure. Diastolic blood pressure (71 ± 1.3 vs. 74 ± 1.7 mmHg, P < 0.01), MSNA [9.94 ± 2.0 to 14.63 ± 1.5 bursts/min (P < 0.05); 16.89 ± 3.2 to 26.97 ± 3.3 bursts/100 heartbeats (hb) (P = 0.01)], and forearm vascular resistance (FVR) (35.3 ± 5.8 vs. 55.3 ± 6.5 mmHg·ml−1·min·100 g tissue, P = 0.01) all increased significantly after 4 wk of exposure. Forearm blood flow response following ischemia of 15 min (reactive hyperemia) fell below baseline values after 4 wk, following an initial increase after 2 wk of exposure. From these results we conclude that the increased blood pressure following prolonged exposure to CIH in healthy humans is associated with sympathetic activation and augmented FVR.

  • sleep apnea
  • hypertension
  • pathophysiology

obstructive sleep apnea (OSA) is a prevalent syndrome characterized by recurrent nocturnal episodes of partial or complete pharyngeal collapse. Defined by an apnea-hypopnea index (AHI) of >5 events/h and significant daytime somnolence, OSA affects up to 5% of the middle-aged population (33). OSA is an independent risk factor for hypertension, arrhythmias, and coronary heart disease (28). Increased sympathetic activation, endothelial dysfunction, and alterations in systemic inflammatory mediators are the three primary mechanisms proposed in the pathophysiology of cardiovascular disease associated with OSA. Evidence from research protocols in OSA patients and both human and animal models of intermittent hypoxia (IH) has provided compelling evidence that IH is the major stimulus of these changes (3).

Understanding of the consequences of IH has been derived primarily from animal and cell culture models. Fletcher et al. (1113) exposed small animals to IH and explored the physiological consequences. Using a dog model, Brooks et al. (2, 3) combined respiratory effort, asphyxia, and arousal from sleep, the three stimuli associated with OSA, and determined that IH, and not sleep fragmentation, was the main stimulus involved in diurnal blood pressure elevation. The animals in both models exhibited elevated daytime arterial blood pressure, and in the model of Fletcher et al. increased sympathetic activity was seen. This increase in sympathetic activity is suppressed in carotid sinus-denervated animals, suggesting a critical contribution of the peripheral chemoreceptor in modulating the sympathoexcitation seen after hypoxic exposure (24).

Hypoxic exposure in human subjects, lasting hours (14, 21) or days (7), results in a sustained increase in ventilation, which is termed ventilatory acclimatization. In contrast to the acute increase in ventilation during acute hypoxia, ventilatory acclimatization is characterized by a sustained increase in ventilation that outlasts the exposure (7). Acclimatization is mediated through peripheral chemoreceptors and is associated with an increase in peripheral chemoreflex gain (1). The mechanism for acclimatization is thought to be altered neuromodulation of the carotid chemoreceptor, with endothelin (6), angiotensin (23), and other substances (1) believed to be contributors. Although study of IH in normal volunteers does not completely reproduce the clinical disorder of OSA, important insights can be gained from understanding the changes in hemodynamics and sympathetic activation that accompany the exposure.

In the setting of ventilatory acclimatization that accompanies exposure to IH in normal volunteers sympathoexcitation has been demonstrated (17, 18). This, again, has been demonstrated in response to exposures of various durations in humans and has been seen both during exposure and upon return to normoxia (20, 25). An important, and to date incompletely answered, question in human subjects is whether the changes seen after exposure to cyclic IH (as seen in OSA) are similar to those seen after sustained IH [as seen in chronic obstructive pulmonary disease (COPD) and obesity hypoventilation syndrome]. This, we would argue, is an important question, as the oxygen saturation fluctuations in the clinical disorder of OSA are cyclic drops in saturation rather than a sustained drop in oxygen saturation during periods of sleep. To date, this has been evaluated in animal models, with only limited study in humans.

To test this hypothesis, we have recently established (31) a new model for extended CIH exposure in healthy humans. In this model, normal human subjects are exposed to 20–30 cycles/h of IH for 8–9 h/day for up to 4 wk. Exposure to chronic IH (CIH) in this model produces increases in morning blood pressure in a setting of enhanced peripheral chemosensitivity, arguing for ventilatory acclimatization to hypoxia (31). In the present study, we hypothesized that exposure to prolonged duration (28 nights) of nocturnal cyclic IH in humans would result in sustained elevations in muscle sympathetic nerve activity (MSNA) and altered vascular reactivity that would contribute to this elevation in daytime blood pressure following exposure.


This study was approved by the Beth Israel Deaconess Medical Center Committee on Clinical Investigations; all subjects provided written informed consent before participation.


Seven healthy, nonsmoking, normotensive subjects who were free of vasoactive medications were studied. The six men and one woman between the ages of 20 and 34 yr (26 ± 5 yr) had body mass indexes between 19.2 and 25.5 kg/m2 (22 ± 2.6 kg/m2). The single female subject tested negative for pregnancy [urinary human β-chorionic gonadotropin (β-hCG)] before participation. In this subject, exposure was begun the first week after menses to minimize the possible confounding effects of hormonal changes on vascular function. Only the findings for changes in blood pressure after 2 and 4 wk of exposure have been previously reported (31).

General Procedures

Cardiovascular and hemodynamic variables were recorded at baseline and after 14 and 28 nights of CIH. MSNA was measured at baseline and after 28 nights of CIH, with recordings performed under normoxic conditions during the morning following either the baseline night or the 28th night of exposure between 3 and 4 h after removal from the tent enclosure. A schema of the testing protocol is provided in Fig. 1. Subjects were studied in the supine position at a room temperature of 24°C. CIH exposure and data collection were conducted at the same time of day across all subjects. Cardiovascular variables and sympathetic nerve activity were recorded continuously, digitized at 128 Hz (model DI-720 series; DataQ Instruments, Akron, OH), and stored locally on a computer hard disk. The data were analyzed off-line with signal processing software (Windaq; DataQ Instruments).

Fig. 1.

Four-week exposure protocol. Measurements were made at 3 time points: before hypoxic exposure, after 2 wk of hypoxic exposure, and after 4 wk of hypoxic exposure. MSNA, muscle sympathetic nerve activity; MAP, mean arterial pressure; HR, heart rate.


Subjects were exposed to 9 h of IH between the hours of 10 PM and 7 AM for 28 consecutive nights with the use of a commercially available normobaric “altitude tent” (Colorado Altitude Training, Colorado Springs, CO). Ambient inspired O2 fraction (FiO2) in the tent chamber was maintained at 0.13 continuously, and supplemental oxygen was provided intermittently to provide “reoxygenation” resulting in cyclic desaturations throughout the period of exposure. A Macromatic time delay relay (TR-63128; Milwaukee, WI) connected to a Festo valve (CPE14) was used to intermittently administer supplemental oxygen to subjects through a nasal cannula. With an oxygen flowmeter (Ohio Medical Products, Madison, WI), 2–4 l/min of oxygen was administered for 10–21 s, alternating with 2–4 l/min of nitrogen for 150–240 s. The flow rate and timing for each subject were adjusted to obtain a drop in oxygen saturation of ∼10% (i.e., 82% to 92%) every 3 min during exposure. Oxygen saturation was continuously monitored by finger oximetry (Biox 3740; Ohmeda, Louisville, CO).

Subjects acclimatized to the hypoxic exposure with three increases in “altitude,” starting with one night at sea level, one night at “7,500 ft,” and one night at “10,000 ft,” before completing 28 nights at “13,000–15,000 ft.” Independent evaluation with an oxygen sensor showed the FiO2 at 13,000 ft to be 0.13 (Crowcon; Gasman). Altitude setting was adjusted across exposure to maintain stable FiO2 at 0.13. CO2 removal from the closed tent system was accomplished with soda-phosphate crystals and a fan-driven system that allowed continuous passage of tent gas across the system. In our previous study under identical tent conditions (15), verification of CO2 levels as performed by an automated CO2 monitoring system (Realterm; Colorado Altitude Training) yielded 0.4% mmHg as an average value during the exposure (range 0.1% to 0.52%).

Cardiovascular Variables

Heart rate (HR) was determined from the electrocardiogram. Dominant arm arterial pressure was measured at 5-min intervals while subjects underwent testing with an automated arm-cuff sphygmomanometer (Cardiocap/5; Datex-Ohmeda).

Forearm Blood Flow

Blood flow was measured by venous occlusion plethysmography (EC6 plethysmography; Hokanson, Bellevue, WA) using mercury-in-Silastic strain gauges. After the arm was placed in a passive position at the level of the right atrium, a strain gauge was placed at the midpoint of the forearm, between the distal and proximal venous occlusion cuffs. Before data collection a series of occlusions were performed to determine the venous occlusion pressure that resulted in the steepest slope of the arterial inflow curve. This typically yielded venous occlusion pressures of 45–50 mmHg. One minute before the start of measurements, the wrist arterial occlusion cuff was inflated to 200 mmHg. The collecting cuff positioned above the elbow was rapidly inflated above venous pressures for 8 s every 15 s. The average of four to six flow measurements was taken at each time point. Forearm blood flow (FBF) is expressed in milliliters per 100 ml of limb tissue per minute. Forearm vascular resistance (FVR) was obtained by dividing mean arterial pressure (MAP) by FBF.

Reactive Hyperemia

Forearm ischemia was induced for 15 min, after which time the cuff was deflated and FBF was measured for 2.5 min by venous occlusion plethysmography. Peak blood flow was evaluated as an average of the first three FBF measurements after ischemia. The area under the curve (AUC) was calculated both at baseline (before exposure to hypoxia) and after 14 and 28 nights of exposure. Vascular reactivity was considered the total excess blood flow above FBF before cuff occlusion, calculated as the difference between the AUC during reactive hyperemia (RH) and FBF immediately before cuff occlusion.

Muscle Sympathetic Nerve Activity

After localization by surface stimulation, standard tungsten microelectrodes were inserted into the peroneal nerve in the popliteal fossa. Signals were filtered, amplified, and full-wave rectified. The rectified signal was integrated for display on an oscilloscope and for recording (Nerve Traffic Analyzer, model 662c-3; Bioengineering Dept., University of Iowa, Iowa City, IA). Electrode position in muscle fibers was confirmed by pulse synchronous bursts of activity occurring during diastole only, reproducible activation during the second phase of the Valsalva maneuver, elicitation of afferent nerve activity by mild muscle stretching, and the absence of response to startle. Sympathetic bursts were identified with a specific algorithm described by Hamner and Taylor (19) and Matlab software (The Mathworks, Natick, MA). For purposes of quantification MSNA was reported in 5-min periods and expressed as burst frequency (bursts/min) and burst incidence [bursts/100 heartbeats (hb)].

Data Analysis

MSNA was averaged over 5-min windows before and after CIH exposure. HR and MAP were averaged over the corresponding time interval during which plethysmographic forearm flow measurements were made. Baseline values were compared with 2-wk and 4-wk exposure values by one-way analysis of variance (ANOVA) and with 4-wk exposure by Student's t-test for paired data. Differences between hyperemic flows were compared by one-way ANOVA. Except where otherwise noted, data are reported as means ± SE. P values < 0.05 were considered statistically significant.


All subjects were exposed to 9 h of nocturnal IH for 4 wk. Mean overnight digital pulse oxygen saturations were 87.9 ± 0.67%. Figure 2 shows a representative oxygen saturation profile during 1 h of nocturnal CIH.

Fig. 2.

Representative plot of oxygen saturation from 1 subject over 1 h of an 8-h nocturnal exposure.

Hemodynamic Variables

HR and MAP.

As previously reported, diastolic blood pressure (DBP) was unchanged after 2 wk but increased significantly after 4 wk of exposure (71 ± 1.3 vs. 74 ± 1.7 mmHg, P < 0.01) (31). There was no change in MAP or systolic blood pressure after either 2 or 4 wk of exposure. HR was unchanged from baseline values after 4 wk of exposure. These results are illustrated in Fig. 3, which shows the set point of blood pressure and HR.

Fig. 3.

Systolic blood pressure and HR before (○) and after (●) 4 wk of exposure.

FVR and FBF.

There was significant decrease in FBF after both 2 and 4 wk of exposure. The decrease seen after 2 wk (3.05 ± 0.56 vs. 1.8 ± 0.24 ml−1·min·100 g tissue−1, P < 0.05) was sustained after 4 wk (3.05 ± 0.56 vs. 1.96 ± 0.4 ml−1·min·100 g tissue−1, P < 0.01). Consistent with the increase in DBP and the decrease in FBF, FVR increased significantly from baseline after 2 wk of exposure (35.3 ± 5.8 vs. 56.9 ± 8.3 mmHg·ml−1·min·100 g tissue, P < 0.01). FVR remained elevated after 4 wk of exposure (35.3 ± 5.8 vs. 55.3 ± 6.5 mmHg·ml−1·min·100 g tissue, P < 0.01). These results are summarized in Fig. 4.

Fig. 4.

Forearm blood flow (A) and forearm vascular resistance (B) at baseline and after 2 and 4 wk of exposure. **P < 0.01.


As a group, subjects demonstrated an increase in mean MSNA after 4 wk of exposure to intermittent cyclic hypoxia [9.94 ± 2.0 to 14.63 ± 1.5 bursts/min (P < 0.05); 16.89 ± 3.2 to 26.97 ± 3.3 bursts/100 hb (P = 0.01)]. Five subjects demonstrated a clear increase in MSNA expressed as either burst frequency (bursts/min) or burst incidence (bursts/100 hb). One subject demonstrated a small decrease in burst frequency (20.33 to 19.69 bursts/min) while demonstrating a modest increase in burst incidence (31.8 to 37.5 bursts/100 hb). One subject did not demonstrate a significant change in MSNA quantified by either burst frequency or burst incidence following exposure. A representative tracing from one subject is shown in Fig. 5, and individual subjects' responses are shown in Fig. 6. As shown in Fig. 7, there was a significant increase in mean MSNA between baseline and postexposure measurements. Figure 7 also illustrates that the documented change in sympathoexcitation correlates with a consistent increase in DBP. However, there was not a significant correlation with the magnitude of change in MSNA and the magnitude of change in FVR in individual subjects [r2 = 0.07 and r2 = 0.01 for ΔMSNA (bursts/min) and ΔMSNA (bursts/100 hb), respectively].

Fig. 5.

Representative recording of 90 s of leg MSNA before exposure (baseline) and after 4 wk of intermittent nocturnal hypoxia.

Fig. 6.

Plot of individual data points for subjects' MSNA [expressed in bursts/min (A) and in bursts/100 heartbeats (hb) (B)] before exposure and after 4 wk of intermittent nocturnal hypoxia.

Fig. 7.

Evolution of the set point for diastolic blood pressure and MSNA (bursts/min) before (○) and after (●) 4 wk of exposure. *P < 0.05.

Reactive Hyperemia

Figure 8 shows the temporal response to RH following 15 min of ischemia. There was an overall increase in hyperemic FBF after 2 wk of exposure. However, there was no significant change in the peak hyperemic FBF after 2 wk of exposure. After 4 wk of exposure, hyperemic FBF was attenuated to below baseline values. In addition, there was a significant decrease in the peak hyperemic FBF from baseline to 4 wk. There was also a significant decrease in the peak hyperemic FBF from 2 wk to 4 wk of exposure (Table 1).

Fig. 8.

Reactive hyperemia (15-min ischemia) before exposure (baseline) and after 2 wk and 4 wk of exposure.

View this table:
Table 1.

Peak flow in response to reactive hyperemia at baseline (before exposure), after 2 wk of hypoxia, and after 4 wk of hypoxia


The novel findings of this study are that prolonged exposure to nocturnal IH in healthy humans results in 1) increased muscle sympathetic activity, 2) increased FVR, and 3) initial increased blood flow response to forearm ischemia at 14 days of exposure that falls below baseline levels after 28 days of exposure. These results provide significant and novel insight into the hemodynamic changes following prolonged exposure to CIH in healthy humans.

Although prolonged exposure to CIH in rodents and dogs has been demonstrated to contribute to blood pressure elevations mediated, in part, through sympathoexcitation (3, 10), there are only limited data in human subjects. In humans, previous studies using brief hypoxic exposure have demonstrated increases in sympathetic tone that outlast exposure (26, 30). The changes seen with more prolonged exposure of 8-h to 2-wk duration of continuous IH have been associated with various changes in sympathetic activation in the setting of ventilatory acclimatization (15, 32). In our recent report (31), cyclic IH, as defined within this study as well to represent exposure to oxygen saturation fluctuations at a rate of 20 desaturations/h across 9 h of nocturnal exposure, was demonstrated to produce ventilatory acclimatization after 2 wk of exposure and may be contributing to sympathetic activation in our subjects.

Augmented sympathetic activity is one mechanism driving blood pressure elevation occurring after CIH exposure in the rodent (9). We report here that sympathetic activity measured by MSNA increases after exposure. It is important to note that baseline activity as well as sympathetic activity following exposure, as absolute values, are relatively low in this group. This is attributable in large part to the approach to analysis, as defined by Hamner and Taylor (19), with strict criteria for identification and quantification of sympathetic activity resulting in a systematic underestimation of activity compared with alternative methods. As this approach was used in all data analyzed here we have good confidence in the findings of increased activity following exposure; however, comparison of absolute values of activity in studies in which different approaches to analysis of MSNA have been used must be undertaken with caution. Although the mechanism of this sympathoactivation needs to be further investigated, it is likely that this sympathoactivation contributes to blood pressure elevation after CIH exposure in our human subjects. Increased sympathetic activity could be related to altered central sympathetic processing or to an alteration in sympathetic regulatory reflex systems (i.e., peripheral chemoreflex, arterial baroreflex, cardiopulmonary reflex, and mechano- and metaboreflex). Both the arterial baroreflex and the peripheral chemoreflex have been implicated as causes of sympathetic activation in OSA patients (5, 27). Using a similar hypoxic exposure, we previously reported (31) an increase in peripheral chemoreception that may contribute to sympathetic activation.

Our study shows that FVR increases with exposure to CIH at 2 wk and remains elevated at 4 wk. It is likely that this increase is driven in part by augmented MSNA. In evaluation of individual subjects' responses we did not demonstrate a significant correlation between the magnitude of change in MSNA and the magnitude of change in FVR following exposure to hypoxia, which would suggest that additional mechanisms are contributing to regulation of local vascular resistance, and these alternative mechanisms are reasonable targets for future investigation. In addition, in the present study we did not measure systemic vascular resistance but only forearm vascular resistance, which is dominated by muscle blood flow regulation. However, we hypothesize that this increase in vascular resistance would contribute to the increase in blood pressure. Finally, as is suggested in Fig. 7, there may well be resetting of the baroreflex that contributes to the sustained elevation of blood pressure upon return to normoxia, although this mechanism was not tested dynamically in the present protocol.

RH was performed as a stimulus to achieve maximal vasodilation. Hyperemic FBF following 15 min of ischemia provides information from both the peak flow and the recovery. The peak flow response corresponds to vessel muscle tone without either the myogenic response or endothelial vasodilation. The return to baseline flow is the result of the balance between the myogenic response (vasoconstriction) and shear stress and vascular factors (vasorelaxation). This vasodilation involves several factors including nitric oxide (NO), prostaglandin, and adenosine (29).

The increase in RH peak flow seen after 2 wk of exposure may reflect adaptations to maintain peripheral oxygen delivery, although this finding is in contrast to the findings of a decrease in baseline FBF and an increase in baseline FVR at this time point during exposure. This would suggest that mechanisms regulating baseline FBF have been altered in ways that are separate and distinct from those that regulate response to RH testing as performed here. It is reasonable to hypothesize that sympathetic activation, which may be the dominant influence in decreases in FBF at baseline, is overwhelmed by enhanced NO-mediated vasodilation in response to RH at the 2 wk time point during exposure (29). The decrease in hyperemic FBF to below baseline conditions after 4 wk of exposure may signify an increase in baseline vasoconstrictor tone as well as a decrease in baseline vasodilator response during RH, suggesting a decrease in NO-mediated vasodilation at completion of exposure. The combined effect of an increase in resting MSNA and shift toward vasoconstriction in the vasorelaxation/vasodilation balance is likely to explain the increase in DBP after prolonged hypoxic exposure, but, in the setting of only 2-wk duration of exposure, enhanced vasodilatory forces as evidenced by enhanced RH blood flow buffer the effects of sympathoexcitation and result in no significant change in blood pressure. This shift in the balance between vasorelaxation and vasodilation has been previously described in sleep apnea patients, in whom such a shift has been attributed to increased sympathetic tone (4, 27), an altered response to pharmacological α-adrenergic agonist and β-blocker (16), or a lack of vasodilation (8, 22). Further testing would be required to confirm these hypothesized mechanisms in normal volunteers undergoing hypoxic exposure of this nature.

Several limitations must be acknowledged in this study. Our exposure does provide cyclic desaturations with a desaturation frequency of 20–30 cycles/h. This is an event frequency similar to that seen in moderate-severity OSA. As has been reported, the exposure does result in evolution of central respiratory events during “reoxygenation.” The impact of this phenomenon is difficult to assess because there is no significant change in sleep quality as a result of the events, and we would suggest that the exposure does primarily represent IH rather than an impact of sleep fragmentation despite the evolution of central respiratory events during exposure (34). It does not, however, reproduce the respiratory events seen in OSA as hypercarbia and compromised airway patency are not present. It is important to note that CO2 levels in the tent chamber were not measured during this exposure. In our previous use of the tent chamber with identical conditions (15), CO2 levels were maintained at 0.4%mmHg during exposure with a range of 0.1–0.52% and a trend to increase across the duration of the exposure night. Extension of these findings to explain the blood pressure elevations seen in patients with OSA must therefore be undertaken with caution. In addition, blood pressure elevations seen in the subjects reported here are modest and do not replicate the changes reported in rodent models of CIH. The magnitude of blood pressure change in our subjects may be a reflection of the modest desaturations experienced by our subjects (mean saturations of 87%), species differences, or the duration of the exposure.

In addition, there was no control group as part of this exposure. The subjects did serve as their own control with paired comparison to baseline (preexposure) conditions, and this does aid in study design to allow minimization of the impact of interindividual differences. A true control group of subjects exposed to 28 nights of normoxia in conditions otherwise identical to the hypoxic subjects is clearly desirable and is important to be included in future study designs. In addition, full evaluation of the changes was only conducted after 28 nights of exposure, and an understanding of the time course of these changes would require future study of more varied durations of exposure. Finally, ventilatory response testing was not performed in this group of subjects directly, and we do depend on the findings in healthy humans undergoing identical exposure for the suggestion that ventilatory acclimatization was present (31). Future study will have to incorporate direct testing of the magnitude of the change in ventilatory acclimatization as a result of the exposure of this duration to more accurately quantify the potential contribution from chemoreflex and baroreflex changes in driving the changes seen.


Exposure of healthy human subjects to cyclic IH is associated with sympathetic activation and altered FVR. Changes in FBF following 15 min of forearm ischemia are dynamic across the exposure despite sustained sympathetic activation. Altered vascular resistance seen after exposure may not be fully accounted for by the sympathoexcitation seen. These findings may contribute to an understanding of the elevations in systemic blood pressure in patients with disease states characterized by cyclic IH, such as OSA.


No conflicts of interest, financial or otherwise, are declared by the author(s).


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
View Abstract