INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Obstructive sleep apnea (OSA) is recognized with increasing frequency as a disease of overweight, middle-aged males. It causes symptoms of excessive daytime sleepiness, memory impairment, and hypertension and contributes to other cardiovascular complications and overall increased mortality (1, 2). The recurrent apneas cause repetitive episodic falls in blood oxygen, which may occur as often as 600 times per night. Each episode may be associated with transient elevation of systemic and pulmonary artery pressure, possibly resulting in sustained systemic and pulmonary hypertension in some patients. Reports of left ventricular dysfunction in OSA are rare. An early report by Lugaresi and coworkers showed fluoroscopic enlargement of the cardiac shadow and congestion of the pulmonary vessels (3). Chaudhary and colleagues have reported three cases of transient radiographic infiltrates and deranged blood gases in patients with typical OSA (4). The authors were able to document normal daytime left ventricular function in two patients by right heart catheterization or radionuclide studies. In one patient, recurrent pulmonary edema stopped after treatment of the obstructive apnea by tracheostomy.

Pulmonary edema due to upper airway obstruction has been reported in children with laryngospasm or epiglotitis (7, 8). Some reports of pulmonary edema in adults labeled as having "laryngospasm" may actually have occurred in patients with a body habitus and symptoms compatible with OSA (9). Pulmonary edema has been described in adults with tumor, hanging, strangulation, near drowning, and extubation postanesthesia. There are a variety of pathophysiologic processes taking place during sleep with OSA, and several factors could contribute to the development of increased lung edema.

To explore the possibility that left ventricular dysfunction and/or lung edema occurs during recurrent OSA, we examined a canine preparation under recurrent obstructive apneas for 8 h. Our hypotheses were that (1) recurrent severe obstructive apneas over a period of 8 h (analogous to one night of human sleep) can lead to lung edema as evidenced by light and electron microscopy (EM) examination and lung wet/dry weight ratios; (2) lung edema would then lead to worsening parameters of gas exchange to include worsened alveolar-arterial oxygen gradient (AaDO₂) and shunt fraction (S/T). Furthermore, desaturation during the later part of the 8-h period should be worse than the beginning as gas exchange deteriorates (10). Multiple cardiopulmonary parameters were measured or calculated including gas exchange, cardiac output (), pulmonary artery pressure (Ppa), pulmonary capillary wedge pressure (Pcw), systemic artery pressure (Psa), and wet/dry lung weight.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Thirteen mongrel dogs weighing 20-29 kg were first anesthesized with ketamine-HCl (5 mg/kg, intramuscular) and an intravenous dose of thiopental (12 mg/kg). They were then orally intubated with a cuffed endotracheal tube and given a loading dose (3.6 mg/kg) of -chloralose. Subcutaneous injections of 1-3 ml of 1% lidocaine were used at vascular cutdown sites to minimize pain. A transurethral bladder catheter was passed into each animal at the time of vascular access in order to allow continuous monitoring of urinary output.

A fiberoptic catheter (4F; Baxter-Edwards, Irvine, CA) was introduced by cutdown into the left femoral artery to continuously measure arterial oxyhemoglobin saturation (Sa_O2), Psa, and heart rate (HR). A continuous cardiac output thermal dilution, flotation catheter (8F, CCO-SvO2-746H-8S; Baxter-Edwards) was introduced by cutdown into the femoral vein to measure continuous Ppa, , and mixed venous oxyhemoglobin saturation (Sv_O2). Arterial catheter Sa_O2 and central venous catheter Sv_O2 readings were calibrated and verified hourly by bench oximeter (OSM2; Radiometer, Copenhagen, Denmark) readings to blood drawn anaerobically from the respective catheters. All measurements were recorded on a 12-channel recorder (model 78; Grass, Quincy, MA). Transducers were referenced to midthorax with dogs in the left decubitus position.

Chloralose was titrated continuously by hourly injection of 10 ml of a 1% solution. -Chloralose was used to maintain anesthesia with minimal respiratory, cardiovascular, or neurologic depression, to achieve a stable, long-lasting light anesthesia (11). The depth of anesthesia was continually assessed by testing pedal withdrawal (toe web pinch) and palpebral reflexes (lightly touching edge of eyelid). At no time during the procedure did the animal withdraw the limb, vocalize, or exhibit muscle twitches or spasms. The palpebral reflex remained absent or slow. The animals were not considered to be "asleep" by electroencephalographic criteria, nor was the electroencephalograph monitored. Core body temperature was monitored and maintained at 37-39° C with a warming blanket. Negative intrathoracic pressure (NIP) was used as a measure of respiratory drive throughout the experiment. NIP was computed by subtracting maximum inspiratory diastolic Ppa from end-expiratory diastolic Ppa (mm Hg) during both resting ventilation (controls) and maximum obstructed effort.

Eight dogs were studied during eight consecutive 50-min periods of repetitive obstructive apneas (approximately 50 apneas per hour). Five similarly intubated and anesthesized dogs with the same vascular catheters but without induced apneas were simultaneously monitored on adjacent operating tables as nonapnea controls. The apneas were induced by clamping the endotracheal tube at end expiration for 45 s followed by release, with 30 s of spontaneous breathing, and the cycle repeated. Complete cessation of air flow was verified by a flow turbine attached to the end of the endotracheal tube. Apneas were stopped for 7-10 min of each hour to recalibrate transducers and oximeters, record steady state saturations and hemodynamic parameters, and draw nonapnea arterial and venous blood gases. At that time, control dogs were given several large tidal volumes or "sighs," to avoid atelectasis. Using previously described formulas (12), with a respiratory quotient of 0.85, and a P₅₀ for blood of 29.0 mm Hg (13), S/T and AaDO₂ were calculated from steady state blood.

Animals were exposed to water ad libitum until 2 h before surgery. After intubation and surgery, an initial fluid bolus of 500 ml of normal saline was infused into each animal to replace any preexisting volume deficit or operative blood loss. Throughout the experiment the infused volume of saline and anesthesia solution was limited to a rate equaling hourly urinary output.

At the end of the 8-h experiment, a 200-mg bolus of Beuthanasia-D was administered intravenously and ventilation was stopped. The thorax was opened by midline sternotomy and the heart and lungs were removed en bloc. After removing the heart and great vessels to the point where they entered the hilum, blood was allowed to drain by gravity. The lungs were patted dry and weighed. The time between death and weighing of the lungs was between 5 and 10 min. After weighing, the right lung (nondependent) was oven dried for 3 to 5 d to constant weight. Four to five cubic (2-4 mm) sections of the left lung (dependent) were placed in 10% buffered formalin for light microscopy processing while a similar number of samples were placed in 2.5% glutaraldehyde for EM processing. Light microscopy samples were examined blindly by two veterinary pathologists for alveolar fluid and interstitial widening. EM specimens were mounted on scored, numbered grids. Grid numbers were selected randomly in the first animal, and then the same number grids were photographed in each successive animal. Five micrographs per specimen were made at an approximate magnification of ×30,000. The micrographs were number coded and scored independently and in a blinded fashion by two lung electron microscopists.

Animals were considered to have lung edema if they had any of the following: (1) alveolar edema by light microscopy, (2) interstitial edema by electron microscopy, or (3) a wet/dry lung weight ratio of  6.0. Alveolar fluid as an isolated finding by electron microscopy was not considered to be lung edema.

Data are reported as means ± standard deviation. Variables for each group measured at the end of each hour were compared with baseline values by one-way analysis of variance (ANOVA) for repeated measures with post hoc Bonferroni and Student t tests when applicable. The null hypothesis was rejected at p < 0.05. To examine group interaction over time, individual parameters were compared using ANOVA for repeated measures with simple contrast. The hypotheses were analyzed as "path models" (14) with H1 positing a direct effect of recurrent apnea on lung edema and H2 predicting an indirect effect of recurrent apnea on gas exchange via its effect on lung edema. The path model consists of a set of recursive linear regression equations. Specifically, edema = ₁ × apnea and gas exchange = ₂ × apnea. Throughout this article, steady state refers to values taken during spontaneous breathing either at baseline or during the 5- to 7-min period at the end of each hour without apnea, when blood gas and hemodynamic variables were allowed to seek stable levels. Intraapneic refers to values of blood gases or hemodynamic measures taken during the nadirs of the last three apneas of a 50-min apnea cycle.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Lung Edema

The path regression equations show a significant direct positive effect of recurrent apnea on lung edema (₁ = 0.625; p = 0.022; and R² = 0.391) and a significant direct effect of edema on the worsening of gas exchange (₂ = 0.648; p = 0.017; and R² = 0.391). A path model that tested for a direct effect of recurrent apnea on gas exchange did not show a significant direct effect when controlling for the effects of edema. The path analysis indicates that recurrent apnea affects gas exchange through its impact on lung edema.

Steady State Blood Gases and Saturation

The study and control dogs did not differ in mean body weight (24.5 ± 3.1, 24.4 ± 2.2 kg, respectively). There was no significant change in Sa_O2 in the control group over the 8-h period. Despite a slight downward trend in Sa_O2 in the apnea animals (measured during steady state periods), this trend was significantly different from baseline only at Hour 6, which included data from Dog 8, which expired with severe gas exchange abnormalities (Figure 1). Both groups showed a progressive fall in Sv_O2 from baseline, significant at Hours 3, 4, and 6 in study dogs and at Hours 2 and 5 in controls (Figure 1). There was a downward trend in Pa_O2 and Pv_O2 (mixed venous O₂ pressure) in both the control and apnea groups over the 8-h period (Figure 2), but the curves did not differ from each other. Again, at Hour 6, Pa_O2 was significantly lower in the study group (p < 0.04), accounted for in part by the rapid deterioration in gas exchange in the animal that expired in the seventh hour.

View larger version (14K): [in this window] [in a new window]   Figure 1.   Mean arterial (SaO2) and mixed venous (SvO2) oxygen saturation from whole blood for apnea dogs (solid circles) and control dogs (open squares). An asterisk (*) indicates a significant difference from baseline (p < 0.02-0.04 when n = 8). SvO2 values for Hours 3, 4, and 6 and SaO2 values for Hour 6 in apnea animals lose significance if the value for Dog 8, which died of acute cardiac failure, is removed from the calculations. Repeated-measures ANOVA with simple contrast does not detect a difference in trends between groups. In this and all subsequent figures, 0 represents baseline, and 1 through 8 are measured at the end of each hour of apnea (under steady state conditions). Symbols are the same in all figures, and n = 8 for Hours 1-6, n = 7 for Hours 7 and 8 because one animal expired early in Hour 7.

View larger version (14K): [in this window] [in a new window]   Figure 2.   Mean arterial and mixed venous oxygen tension for apnea and control dogs. The asterisk (*) indicates a significant difference from the baseline value (p < 0.04 when n = 8). The difference loses significance if the PaO2 for Dog 8 is removed from the calculations. Repeated-measures ANOVA with simple contrast does not detect a difference in trends between groups.

The downward trend in oxygenation in both groups during steady state periods is accounted for by deterioration in parameters of gas exchange, including widening of AaDO₂ (Figure 3) and increase in S/T (Figure 4). AaDO₂ and S/T in apnea animals were significantly different from baseline in Hours 4-8 and 5-8, respectively. If the values for the animal that expired before the end of the experiment are removed from analysis, values at Hours 6, 7, and 8 remain significantly different from baseline for both AaDO₂ and S/T (Figures 3 and 4, respectively). The upward trends in AaDO₂ and S/T (n = 8, study; n = 5, control) do not differ significantly over the 8-h period (ANOVA with simple contrast). Nadir Sa_O2 and Sv_O2 became lower during each successive period of apneas, correlating with the gradual deterioration in gas exchange in the study animals (Figure 5).

View larger version (13K): [in this window] [in a new window]   Figure 3.   Mean alveolar-arterial oxygen gradient for apnea and control. An asterisk (*) indicates a significant difference from the baseline value (p < 0.05-0.01 when n = 8). Values marked with a plus symbol (+) lose significance if AaDO2 for Dog 8 is removed from the calculations. Repeated-measures ANOVA with simple contrast does not detect a difference between groups in the upward trend.

View larger version (13K): [in this window] [in a new window]   Figure 4.   Mean venous admixture for apnea and control dogs. An asterisk (*) indicates a significant difference from the baseline value (p < 0.04). The value marked with a plus symbol (+) loses significance if the S/ T value for Dog 8 is removed from the calculations. Repeated-measures ANOVA with simple contrast does not detect a difference between groups in the upward trend.

View larger version (26K): [in this window] [in a new window]   Figure 5.   Continuous arterial and mixed venous oxygen saturation (fiberoptic catheter) for two dogs (Dog 16, apnea; Dog 17, control) run simultaneously over an 8-h experimental period. The breaks between apneas were for recalibration and measurement of steady state blood gases and hemodynamics. Top channel: SaO2 by fiberoptic catheter in apnea dog. Middle channel: SvO2 and continuous cardiac output in apnea dog. Bottom channel: SvO2 and continuous in control dog. Note the greater fall in SvO2 with each successive apnea period compared with the control dog.

Intraapnea Blood Gases and Saturation

A gradual worsening of gas exchange during each successive hour of apnea is best seen in the individual analog tracings of steady state and intraapnea Sa_O2 and Sv_O2 in an apnea dog with lung edema compared with its simultaneously run control (Figure 5). Figure 5 shows eight successive hours of apnea, each resulting in more severe arterial and venous desaturation than in the preceding hour, while the control animal showed a modest deterioration in Sv_O2. Figure 6 summarizes the progressive decline in nadir and recovery saturations for the five apnea animals that met criteria for lung edema. After the first hour, steady state Sv_O2 as well as nadir Sv_O2 are significantly lower (except in Hour 8) than Hour 1 values, which were considered baseline. Steady state Sa_O2 values, although trending downward, were not different from Hour 1 values, but nadir Sa_O2 at Hours 5, 6, and 7 were significantly lower than Hour 1 values. All Sa_O2 and Sv_O2 values for these data were read from the fiberoptic catheters. The nadir apneic values were computed from the Sa_O2 and Sv_O2 of the last three apneas of each 50-min apnea cycle.

View larger version (23K): [in this window] [in a new window]   Figure 6.   Mean SaO2 and SvO2 values read from fiberoptic catheters inserted in five apnea animals judged to have lung edema; values were determined at baseline and every hour during the experiment during rest (open and closed circles) and during nadirs of the last three apneas of the hour (open and closed triangles). Note that mean nadir values for SaO2 and SvO2 become progressively lower over time. An asterisk (*) indicates a significant difference from the baseline value (p < 0.05).

Steady State Pulmonary and Systemic Hemodynamics

There was no significant difference in at baseline between apnea and control animals. ANOVA for repeated measures over time with simple contrast revealed that the for both groups showed a significant downward trend but that the trend did not differ between groups (Figure 7). The for Hours 3-8 were significantly lower than baseline in the apnea dogs. There was no change in ,we from beginning to the end of the study for either group (Figure 8). Despite a slight upward trend in both groups, steady state Ppa did not increase significantly in either group when measured during the hourly 10-min steady state periods. measured as the mean of the last three apnea peak Ppa values of each 50-min period did show a progressive increase over the 8-h period. There was no difference in steady state Psa trends between the control or apnea animals (Figure 9). End apnea increased 14 mm Hg from preapnea in apnea animals.

View larger version (13K): [in this window] [in a new window]   Figure 7.   Mean cardiac output as determined by fiberoptic catheter at baseline and every hour during the experiment (under nonapneic conditions) for apnea dogs (solid circles) and control dogs (open squares). An asterisk (*) indicates a significant difference from baseline (at p < 0.04). The slopes of the change were not significantly different between groups by ANOVA for repeated measures with simple contrast.

View larger version (17K): [in this window] [in a new window]   Figure 8.   Top: Mean pulmonary artery pressure calculated as two-thirds of pulse pressure from Swan-Ganz catheter readings at baseline and every hour during the experiment at rest (closed circles) and during the nadir SaO2 of the last apnea of the hour (closed triangles), and in the nonapneic control dogs (open squares). There was no significant change in resting Ppa in either group but all nadir readings were higher than baseline. The Ppa tended to become more elevated with progressive apneas of the same duration. Bottom: Pulmonary capillary wedge pressure (Pcw) for apnea and control dogs throughout the 8-h experiment measured each hour under steady state, nonapnea conditions.

View larger version (16K): [in this window] [in a new window]   Figure 9.   Mean systemic blood pressure (Psa) calculated as two-thirds of pulse pressure from arterial catheter readings at baseline and every hour during the experiment during rest (closed circles) and during nadir SaO2 of last apnea of the hour (closed triangles), and in nonapneic control dogs (open squares). There was no significant change in resting Psa in either group, but there was a tendency toward decreased Psa relative to baseline in apnea animals.

In the study animals, the interapneic (resting) heart rate (Table 2) showed a fall from the first hour (p < 0.01) and remained lower during Hours 2-8 whereas control dogs showed no such change. Nadir-apneic heart rate showed relative bradycardia in Hours 2-8, which was lower than the initial slowing in Hour 1. Interapneic (resting) NIP tended to be greater in Hours 2-8 compared with Hour 1 for both study and control dogs (see Table 2 for specific values). Nadir-apneic NIP became significantly greater (representing increased effort) with each 50-min period after Hour 1 (mean of last three apneas before apnea termination), indicating that neither progressive hypoxia, extended muscular effort, nor -chloralose anesthesia depressed respiratory drive or output.

Mortality

One apnea dog expired early in the seventh hour of the study (Dog 8). Hemodynamic data indicate that this was from acute left ventricular failure as Ppa increased from 16/9 to 50/36 mm Hg and Pcw increased from a baseline of 4 mm Hg to a high of 16 mm Hg just before death (Table 3, Figure 10). , which ranged from 4.0 to 4.3 L/min throughout the first 6 h of study, increased acutely to 6.1 L/min at the end of the sixth hour of apnea, then fell precipitously to less than 1.5 L/min just before death. Postmortem examination of this animal did not reveal pre-existing pathology of the lung, heart, or great vessels. There was acute biventricular dilatation, and proteinaceous fluid covering the cut surfaces of the lung, compatible with pulmonary edema. It also showed evidence of alveolar edema by light microscopy and interstitial edema by electron microscopy. It appears that the dog expired of severe biventricular dysfunction accompanying hypoxemia.

View larger version (26K): [in this window] [in a new window]   Figure 10.   Continuous tracing of multiple physiological parameters during several repetitive apneas at the beginning, middle, and just before death in Dog 8. Top channel: Pulmonary artery pressure. Second channel: Systemic artery pressure. Third channel: Tidal volume. Fourth channel: Esophageal pressure (Pes). Fifth channel: SaO2. Bottom channel: SvO2. Note that there is a progressive deterioration of SaO2 and SvO2, with acceleration of this drop just before death at 16:35. Pulmonary artery pressure becomes markedly elevated in the final hour before death, although systemic pressure and cardiac output remain near baseline. The terminal events include a rapid drop in Ppa and Psa with a fall in cardiac output just before death (far right). The breaths at the far right were manually given.

Lung Fluid

The mean lung wet/dry weight for the apnea group (n = 8) was 5.40 ± 0.93, whereas that for the control group was 5.00 ± 0.67 (NS). Generalized alveolar filling by eosinophilic fluid, revealed by light microscopy, along with interstitial edema, determined by electron microscopy, were observed in three dogs (Dogs 8, 12, and 20) (Table 1, Figure 11). The wet/dry weight ratio in these three animals was 5.10, 5.19, and 7.30, respectively. Two other study dogs (Dogs 14 and 16) but no controls had interstitial edema agreed on by both electron microscopists. Their wet/dry ratios were 5.0 and 6.5, respectively. The Pcw, Ppa, and values in the five animals with light microscopy or EM evidence of pulmonary edema were not discriminatory from other study or control animals except for Dog 8, described above. These five animals (Dogs 8, 12, 14, 16, and 20) had other EM findings compatible with alveolar-interstitial damage (interstitial fibrin, epithelial and endothelial cell breakage) but such changes were also seen in control animals and thus were not specific to apnea dogs.

View larger version (146K): [in this window] [in a new window]   Figure 11.   Histologic section of lung from apnea animal 20, showing accumulation of alveolar edema fluid. Stain, hematoxylin- eosin; original magnification, ×40.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The aims of this experiment were to determine the following: (1) Can lung edema be produced in an acute animal preparation of obstructive apnea? (2) Does cardiac dysfunction occur during a long period of repetitive apneas? (3) If lung edema is produced in this setting, is it "cardiac" or "noncardiac"? (4) Does gas exchange deteriorate in animals that show lung edema? The important findings related to repetitive apneas are that (1) five of eight animals displayed evidence by light microscopy and/or EM, and wet/dry lung weight ratios, consistent with lung edema, (2) overall hemodynamic parameters (Ppc,we, Ppa, and Psa) remained relatively stable except for a slight fall in (but not different between groups), (3) one animal expired with what appeared to be acute cardiac decompensation with pulmonary edema, (4) there was significant deterioration of gas exchange in those animals with histologic and/or EM evidence of lung edema compared with those without edema and controls, and (5) there was a substantial fall in nadir level of apnea desaturation related to the deterioration of gas exchange.

Five of the eight study animals showed physiologic and tissue changes consistent with lung edema whereas one of these expired with acute cardiac failure. The lung wet/dry ratios tended to be higher in the five animals with positive histologic/ EM changes (5.8 ± 1.0) compared with controls (mean, 5.0 ± 0.7) or the remaining study animals (mean, 4.8 ± 0.5). Likewise, the gas exchange deterioration tended to be worse in animals with histologic or EM changes consistent with lung edema formation, although Dog 18, without histologic findings of lung edema, did show a S/T increase from baseline of 13.4% by the end of the study. We believe that part of this inconsistency is due to tissue sampling error and part is due to the insensitivity of lung wet/dry weight ratios. For example, tissue for light microscopy and EM sections was selected randomly (without regard to gross appearance) from the first dog and then from equivalent areas from subsequent dogs, in order not to bias histologic and EM findings. Early lung edema may be patchy and could have been missed in some dogs (e.g., Dog 18). This is especially true of the EM sections that were selected out of the same numbered grids for each animal. Also, there was not always agreement on EM findings between the two readers. Thus, while light microscopy and EM are specific when changes compatible with increased lung water are found, there may be some false negatives because of sampling error.

Furthermore, histologic examination often does not agree with lung weights in experimental lung edema. The wet/dry weight ratio is a relatively insensitive technique for detecting mild interstitial edema (15). Within the interstitial space, fluid accumulation is seen primarily around arteries owing to lower interstitial pressures. "Observation of perivascular edema on histological sections in the face of normal wet/dry ratios result from the relative sensitivities of the two techniques" (16). The animals with only slightly elevated lung wet/dry weight ratios may have had sufficient interstitial edema around vessels to cause abnormal gas exchange but not enough to grossly elevate the wet weight of the lung.

The animal that expired in the seventh hour of repetitive apneas showed a rapid terminal deterioration of gas exchange as evidenced by a drop in Pa_O2 and marked increase in venous admixture (Table 3) beginning in the fifth and sixth hours (Figure 10). Pcw began increasing in the fifth hour but blood pressure and were maintained until the last few minutes and then deteriorated rapidly below the catheter range minutes before death. Interstitial edema with ventilation perfusion mismatch would best account for the gas exchange deterioration in this animal. Gross lobar atelectasis was not seen in this or any animal but we cannot rule out some areas of microatelectasis. The immediate cause of death was severe hypoxemia with acute cardiac failure (ventricular dilatation) and pulmonary edema.

In this experiment, during the nonapneic periods, we assumed a normal P₅₀ (29 mm Hg in dogs) and a respiratory quotient (R) of 0.85, knowing that P₅₀ could possibly shift rightward owing to acidosis and hypercarbia, and that R could possibly increase during the course of the experiment owing to increased sympathetic activity. While we had no way of computing the actual values for either of these parameters, we did test the theoretical effect of such shifts on AaDO₂ and S/T. Holding all parameters constant, we tested the effect of an upward shift in R to 0.95 and 1.0, and then a rightward shift in P₅₀ from 29 to 35 and 40 mm Hg in the situations of mild (6-9%) and severe (50-70%) shunt. Increased R raised AaDO₂ by about 3-5 mm Hg in both mild and severe levels of abnormal gas exchange, but percentagewise had a greater effect on AaDO₂ and S/T in the setting of mild gas exchange abnormality. Increasing P₅₀, on the other hand, had no effect on AaDO₂ and caused only 1.5% higher S/T in mildly impaired animals. By assuming a normal R and P₅₀, we in effect underestimated the impairment of gas exchange in mildly impaired animals. Thus, we may have missed a small deterioration of gas exchange in mildly effected control or study animals. In making statistical comparisons, it is preferable to err on the conservative side, yet we still found significant differences from baseline in the final 3 h of study (Figures 3 and 4) even after removing from the analysis the data on the dog that expired.

None of the controls had alveolar edema by light microscopy but all had mild deterioration in gas exchange, suggesting that at least part of the gas exchange abnormalities in this preparation (both study and controls) may be accounted for by microatelectasis. Alveolar leak would not be expected in the spontaneously breathing, nonapneic animals, making microatelectasis the only reasonable possibility as a cause of worsening gas exchange. Microatelectasis will not be obvious on histologic examination. The fact that gas exchange deterioration was worse over the 8-h period and the lungs slightly heavier in the apnea group suggests that increased lung water may have contributed to the gas exchange deterioration in the experimental group. Despite a slight downward trend in , which was not different between groups, the animals that displayed either alveolar or interstitial edema by microscopy maintained and Pcw throughout the 8 h of apnea, mitigating against left ventricular failure as the cause of increased lung water, except in the animal that died (Table 3). Thus, it appears that subclinical, noncardiac lung edema may be more common than suspected after severe repetitive obstructive apneas. The death of Dog 8 leaves open the possibility that acute left ventricular failure may occasionally occur.

Several mechanisms are proposed to account for pulmonary edema in the setting of severe hypoxia. An important effect of hypoxia is to increase pulmonary microvascular hydrostatic pressure (Pmv) through constriction of pulmonary arterioles and postcapillary venules, altering Starling forces favoring fluid filtration into the lung (17). In addition, arterial hypoxemia can cause a massive CNS sympathetic discharge resulting in peripheral vasoconstriction and systemic hypertension, shunting blood centrally and overloading the vascular system (18). Moss and coworkers have caused pulmonary edema in animals by selective cerebral perfusion with hypoxemic blood (19). The resulting systemic hypertension could also increase left ventricular afterload, causing pulmonary hydrostatic pressure to increase. Hypoxemia during apnea could directly suppress left ventricular contractility, compounding unfavorable Starling forces. Arterial hypoxemia has been associated with increased brain endorphins resulting in detrimental effects on respiration and circulation. In examining the hemodynamic data from the animals adversely affected by the repetitive apneas, mean steady state systemic blood pressure () was not higher in apnea versus the control group (Figure 9). The intraapneic Psa was markedly higher than steady state Psa for each apnea animal and theoretically could have contributed to unfavorable Starling forces during apnea with transient increased left ventricular afterload. However, the extremes of intraapneic Psa increase were no higher in the lung edema apnea dogs than in the nonedema apnea dogs. Extreme, transient acute pulmonary or systemic hypertension does not appear to be a mechanism contributing to lung edema in the setting of this study but cannot be proven by our study design.

During obstructive apneas, tremendous negative intrathoracic pressure (NIP) (40 mm Hg or lower) is generated against a closed airway, resulting in a variety of mechanical effects. This extreme NIP is transmitted to the perivascular space, decreasing perivascular hydrostatic pressure and favoring fluid filtration (20). Because the lymphatic duct is an intrathoracic structure that empties into an extrathoracic vein, it is exposed to the same stresses of extreme NIP, which may retard the egress of lymph during inspiration. In addition, wide fluctuations in NIP could theoretically cause rapid alteration of pulmonary blood flow velocity, creating disruptive sheer forces, damaging capillaries, and creating protein leaks (21). The accentuated NIP can also increase venous return (increased preload) while decreasing pulmonary vascular bed emptying through leftward septal shift (increasing left ventricular diastolic filling pressure) (22). Accentuated negative pressure around the heart increases transmural pressure and left ventricular afterload. The net effect of these mechanical loads may be to alter Starling forces with increased fluid filtration with decreased lymph removal, increasing lung water.

The many factors interacting in obstructive asphyxia may be better understood by examining forces acting to cause fluid accumulation Qf = Kf[(Pmv  Ppmv)  @ (ii mv  ii pmv)] where Qf is the fluid filtration rate, Kf is the filtration coefficient, Pmv is the microvascular hydrostatic pressure, Ppmv is the interstitial pressure surrounding the microvasculature, @ is the reflection coefficient, and ii mv and ii pmv are the osmotic pressures in the plasma and interstitial fluid. Because osmotic pressure and Kf should be constant before and during apnea, Qf should depend mainly on Pmv and Ppmv.

On the basis of this equation, we can examine the factors most likely to increase lung water in obstructive asphyxia. Hansen and coworkers have examined the effect of hypoxia, and hypoxia with hypercapnia, on fluid filtration in newborn lambs (25). Although Qf increased by 74% owing to increased Pmv with hypoxia alone and an additional 33% with hypoxia and hypercapnia, neither lymph protein concentration nor extravascular lung water increased over baseline because lymph removal equaled formation. It should be noted that these animals were mechanically ventilated so that Ppmv was more positive than usual. A single 2-min period of obstructive asphyxia did not change the preceding parameters. In a separate experiment, Hansen and colleagues found no increase in filtration pressure, Qf, or lymph protein concentration in spontaneously breathing newborn lambs when inspiratory resistance was applied to accentuate NIP (26). The authors concluded that accentuation of NIP through added inspiratory resistance lowered Pmv and Ppmv equally so that Qf remained unaffected. Because the animals breathed 30-40% O₂, pulmonary vasoconstriction was avoided, limiting a possible increase in Pmv. Several other authors have shown that more negative NIP alone does not increase lung water. Bo and colleagues have shown increased Qf when Pmv was held constant and alveolar pressure was decreased (27). During airway obstruction, a drop in pleural pressure (accentuated NIP) resulted in even further fluid filtration (28). Polianski and coworkers were unable to increase extravascular lung water in dogs with 20 cm H₂O negative intrathoracic pressure for up to 6 h (29). On the other hand, Haddy and coworkers (30) and Stalcup and Mellins (31) both showed increased extravascular lung water with accentuated NIP. While these results conflict, most do not include severe hypoxia as a study condition, which is regularly seen in OSA. Hypoxia could theoretically aggravate edema formation by increasing Qf as demonstrated by Hansen and colleagues (25). Both hypoxia-induced pulmonary vasoconstriction (increased Pmv) and accentuated NIP (increased Ppmv) may be necessary to increase extravascular lung water. Because the thoracic duct is an intrathoracic structure that empties into an extrathoracic vessel, perhaps the same pressure forces that act to cause lung blood pooling also act to keep lymph within the thoracic duct and favor increased interstitial fluid accumulation.

One report has added new clinical significance to the association of left ventricular dysfunction and OSA. Malone and coworkers examined left ventricular ejection fraction (LVEF) in eight subjects with "idiopathic" cardiomyopathy (LVEF < 55%; negative for myocarditis or coronary artery disease) and clinical suspicion of OSA (32). After elimination of OSA by effective therapy with nasal continuous positive airway pressure (NCPAP) for 4 wk, subjects increased their mean LVEF from 37 ± 5 to 49 ± 5%. In four subjects, cessation of NCPAP for 1 wk caused a return of LVEF to pretreatment values. The same laboratory has also documented improvement in heart rate, LVEF (21.2 to 28.9%), and New York Heart Association symptom classification in 12 patients with congestive heart failure and Cheyne-Stokes respiration after 12 wk of NCPAP therapy (33). Appropriate non-NCPAP-treated controls showed no change in these parameters. Hedner and colleagues have shown that left ventricular wall thickness by echocardiography is increased in normotensive OSA patients, implying that recurrent hypoxia or extreme negative intrathoracic pressure may chronically affect left ventricular function (34).

In conclusion, we have demonstrated that recurrent obstructive apneas over an 8-h period can cause lung edema and deterioration of gas exchange in more than half of animals exposed to severe repetitive apneas. Severe apnea can, on occasion, cause acute cardiac decompensation. However, in other than the single animal that expired, no animal exhibited overt clinical signs of acute pulmonary edema (frothy endotracheal tube secretions, elevated Pcw, hypotension, or shock). Changes over the experimental period were limited to those detected by histologic or EM examination and by parameters of gas exchange deterioration. With clinically evident acute pulmonary edema being rare in OSA patients (based on the current literature), this study would indicate that the lungs of apneic humans and animals appear to be relatively resistant to gross alveolar flooding, although interstitial edema might be more common. If analyzed by more sensitive means (measures of lung water flux), subclinical pulmonary edema might be common. We found no consistent EM picture of endothelial or capillary epithelial damage in apnea dogs that were not seen in controls. The lack of correlation between lung wet/dry ratios and light microscopic findings appear to be due to insensitivity of the former and sampling error in the latter.

There are two clinical settings to which the findings of this study may relate. One is that if lung water increases throughout the night, gas exchange may worsen, which could contribute to steeper dSa_O2/dt and lower nadir saturation for a given apnea duration, late in the night (10). Second, one could speculate that recurrent left ventricular strain over the course of 8 h of apnea might cumulatively lead to or contribute to chronic left ventricular failure similar to the patients described by Malone and coworkers (32).