INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The obstructive sleep apnea hypopnea syndrome (OSAHS) occurs in 2 to 4% of the adult population (1). Increased morbidity and mortality are associated with this disease in persons not treated (2, 3). Unfortunately, there are no universally effective drug therapies for this common disorder. Although nasal continuous positive airway pressure (nCPAP) is an extremely effective therapy (4, 5), compliance with nCPAP use is an issue (6). Upper airway surgery and oral appliances are treatment modalities designed to increase the caliber of the upper airway, but these treatments are effective in only subsets of patients with OSAHS (7, 8). Clearly, there is a significant need for a safe, well-tolerated, effective pharmacotherapeutic for persons with OSAHS.

In the majority of humans with OSAHS, respiration is normal during waking. During sleep, however, there are intermittent reductions in upper airway dilator muscle activity, and these reductions in muscle activity are associated with apneas and hypopneas (9). The neural mechanisms responsible for this sleep-state-dependent reduction in upper airway dilator activity are unknown. Serotonin (5-HT) is one neuromodulator that can powerfully excite upper airway motoneurons (10, 11), and 5-HT activity is very sleep-state-dependent, such that 5-HT activity is reduced in non-rapid-eye-movement sleep (NREMS) and even more profoundly reduced in rapid-eye-movement sleep (REMS) (12). We therefore hypothesize that sleep-state-dependent reductions in upper airway muscle activity occur, at least in part, because of a sleep-state-dependent withdrawal of excitatory serotoninergic input to upper airway dilator motoneurons. We have shown the importance of serotonin in the maintenance of patent upper airways and normal respiration (13) in a natural animal model of sleep-disordered breathing (SDB), the English bulldog (14). This particular breed of dogs has a narrowed upper airway with an elongated soft palate (14). These dogs are strikingly hypersomnolent, and in sleep they demonstrate intermittent reductions in upper airway dilator muscle activity, coincident with oxyhemoglobin desaturations and arousals in REMS and primarily respiratory-effort-related arousals in NREMS (15). Thus, the dogs are a natural animal model of the human obstructive sleep apnea hypopnea syndrome, best representing REMS-predominant OSAHS and upper airway resistance syndrome. To demonstrate the importance of serotoninergic drive in the maintenance of normal respiration in bulldogs with sleep-disordered breathing, we systemically administered serotonin antagonists and demonstrated that these drugs in waking could mimic sleep-disordered respiration with reduced upper airway dilator activity, upper airway collapse, and oxyhemoglobin desaturations (13). Thus, 5-HT appears important in the maintenance of upper airway muscle activity in an animal model of SDB. The precise mechanisms, however, through which 5-HT excites upper airway motoneurons remain unknown.

Although a number of human trials have been conducted evaluating the effectiveness of several serotoninergic drugs in treating the sleep apnea syndrome (16), the effectiveness of such drugs has neither been established nor excluded. That is, the study designs for serotoninergic drug trials in humans with OSAHS have typically been double-blinded, randomized trials comparing the effects of placebo and a single drug dose with one study for each condition. Most of the studies described improvements in sleep-disordered breathing in some, but not all, of the patients (16). Because there can be significant day-to-day variability in the severity of sleep apnea (21, 22), the study design of a single trial in each condition lends itself to both Type I and Type II statistical error. A second distinct possibility for Type II error in these studies is the study of one dose. That is, dose-dependency and the possibility that persons with more severe sleep apnea require greater serotoninergic tone have never been discerned. Thus, multidose studies with multiple trials for each dose are necessary to determine if serotoninergic drugs may impact positively upon the sleep apnea syndrome.

Such studies may be easier to perform in the animal model of obstructive sleep apnea, the English bulldog. The advantages of studying the animal model will also include greater control of potentially confounding factors (e.g., variability in the sleep schedule, alcohol, other medications, weight change, and sleeping position), in addition to the greater feasibility of executing a study design of multiple trials at multiple doses for each subject. We reasoned that if such a study showed that serotoninergics are indeed effective in treating sleep-disordered breathing in our animal model, then these results would add weight to the view that specific serotoninergic drugs may be effective in treating obstructive sleep apnea in humans. This, in turn, would enable the development of specific serotoninergics to treat the human sleep apnea syndrome safely and effectively.

The purpose of this study, then, was to ascertain the overall therapeutic potential of serotoninergic drugs in treating sleep-disordered breathing in the English bulldog. There are numerous categories of serotoninergic drugs, e.g., selective reuptake inhibitors, 5-HT production enhancers, 5-HT releasing agents, drugs that slow the breakdown of 5-HT, and 5-HT agonists, which may be broad spectrum or target one or more of the 14 5-HT receptor subtypes. A combination therapy, using drugs active through several of these mechanisms, was studied in an effort to maximize the serotonin effect, given that the precise mechanisms and receptor subtypes through which 5-HT excites upper airway motoneurons are not known. At the same time, consideration was made to use drugs that would not significantly disrupt sleep or prevent specific sleep states from occurring. We elected to target increased 5-HT production and release by using L-tryptophan, a 5-HT precursor that increases 5-HT production relative to the amount of precursor administered and does so in a nonfiring rate-dependent fashion (23). Further, L-tryptophan has been shown to reduce, albeit slightly, obstructive apneas in patients with obstructive sleep apnea (16). We also chose to target the most likely 5-HT receptor subtypes involved in the excitation of upper airway dilator motoneurons, 5-HT_{2A, 2C, 6, 7} (11, 24). For the latter targets, we selected trazodone. Trazodone is a complicated drug, and although the parent drug is likely to have activity as a 5-HT_2C antagonist (28), it has as a metabolite meta-(chlorophenyl)piperazine (m-CPP), a powerful 5-HT_{2A, 2C, 6, 7} agonist (27). Importantly, the overall effect of systemically administered trazodone on motoneurons is excitatory. That is, the effect of systemically administered trazodone on spinal motoneuronal activity is an augmentation of activity, presumably through its metabolite m-CPP (27). Administration of m-CPP itself is not well-tolerated and is associated with anxiety, insomnia, and nausea (29), whereas trazodone (30) and L-tryptophan (31) have been shown to improve sleep.

This study, therefore, evaluates the effectiveness of trazodone with L-tryptophan in alleviating sleep-disordered breathing in our animal model of the sleep apnea syndrome. The study design implemented was a series of 16 randomized sleep studies per dog with four trials under each of four conditions, placebo and three doses of the drugs. We report a substantial improvement in NREMS-disordered breathing events and moderate improvements in REMS-disordered breathing events. Individual responses for each dog were significant, as well as for the group. Further, this therapy is associated with consolidation of sleep, enhanced slow-wave sleep, and less suppression of upper airway dilator muscle activity in NREMS and REMS.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Animals

The English bulldogs were housed individually in a University Laboratory Animal Research colony at the University of Pennsylvania. The animals were kept on a 12:12 light-dark cycle with food and water ad libitum. The surgical procedure and experimental protocol were approved by the Institutional Animal Care and Use Committee at the University of Pennsylvania. All dogs had been trained to sleep in the recording cage.

Animal Preparation

To obtain optimal long-term electroencephalographic and electromyographic recording signals, a telemetry transmitter and electrode system (32) (TL 10M3 Bioimplant; Data Sciences, St. Paul, MN) was implanted surgically into each bulldog. Dogs were premedicated with 0.005 mg/kg atropine sulfate intramuscularly and 0.2 mg/kg dexamethasone intravenously to reduce upper airway secretions and edema. General anesthesia was induced intravenously with 25 to 35 mg/kg thiopental. The dogs were endotracheally intubated, and general anesthesia was maintained with inhaled isoflurane (1 to 2.5%). A 3-cm dorsal midline thoracic incision was made, through which a subcutaneous pocket was created 7 cm in diameter to hold the transmitter. A second 3-cm dorsal incision was made on the neck, and a 5-cm midline scalp incision was made to expose the dorsum of the skull. The final incision was a midline 3-cm incision on the ventral surface of the neck. Under sterile conditions, the telemetry transmitter was placed into the subcutaneous pocket, and all electrodes were fed subcutaneously to the dorsal neck incision. Here two electrodes were secured into the dorsal nuchal musculature to serve as nuchal electromyography (nEMG) electrodes. Three additional electrodes were fed subcutaneously from the neck to the skull. Prior to implantation of these electrodes, the skull surface was cleaned of its periosteum and dried. The three electrodes were screwed into the skull over the frontal and parietal cerebrum. Two of these served as electroencephalography (EEG) electrodes, and one served as a ground electrode. These were secured in place with dental acrylic. Two additional electrodes were tunneled subcutaneously to the ventral neck incision and implanted with sutures into the left sternohyoid (SH) muscle. All incisions were sutured closed. Animals were then allowed to recover from anesthesia. After 12 h of close observation, they were returned to their home cages. Animals were allowed 3 wk recovery prior to experimentation.

Electrophysiologic Recording Procedures

Sleep studies were performed with the dogs unrestrained within a 2-cubic-meter electrically shielded Plexiglas recording chamber. EEG, nuchal EMG, and SH EMG signals were transmitted to a receiver (RLA 2000; Data Sciences) within the recording chamber. From the receiver, EEG and EMG signals were passed to a receiver multiplexor (RMX 10; Data Sciences), and then amplified and band-passed filtered (BMI-830; CWE, Inc., Ardmore, PA) using settings of 1 to 100 Hz for the EEG and 50 to 1,000 Hz for the EMG signals. The SH EMG signals were then passed through a moving averager (MA-821; CWE, Inc.) using a time constant of 100 ms. All of these signals were sent to a polygraph recorder (TA11; Gould Instruments, Cleveland, OH). The other measured parameters were hemoglobin oxygen saturation, abdominal and rib cage movement, snoring, and nasal airflow. Oximetry was measured continuously at the dog's pinnae (Biox IIA; Ohmeda, Boulder, CO). All changes in oximetry readings were recorded manually on the chart recorder paper. Rib cage and abdominal movements were measured with inductive plethysmography bands (Respitrace; Ambulatory Monitoring, Ardsley, NY) with signals sent to the polygraph. Snoring was measured with a snore sensor attached to the ventral neck of the dog (Snore Mike 1200; EPM Systems, Sylmar, CA). At rest, bulldogs are exclusively nasal breathers. Thus, only nasal airflow was measured in this study. Airflow was measured indirectly and qualitatively with a thermistor (Easy Flow; EPM Systems, Midlothian, VA) suspended just over the nares user a wire halter we developed specifically for this purpose. Signals were sent directly to the chart recorder.

Study Design and Protocol

Each of the five English bulldogs studied was randomized to the order of 16 separate sleep studies (four placebo studies and four studies at each of the three drug dose conditions). The order for each of the 16 studies was determined using a random numbers sheet for numbers 1 to 4 for each of the four experimental conditions. The four experimental conditions varied by drug dose administered. These conditions were as follows. Condition 1: placebo. Condition 2: trazodone 3.3 mg/ kg and L-tryptophan 44.3 mg/kg (Dose 1). Condition 3: trazodone 6.7 mg/kg and L-tryptophan 89 mg/kg (Dose 2). Condition 4: 13.3 mg/kg trazodone and 174.3 mg/kg L-tryptophan (Dose 3). A validated washout period of 3 d occurred between each of the 16 studies. Dogs, as well as polysomnography recorders and scorers, were blinded to the conditions. Drugs and placebo were administered orally 30 min prior to the beginning of a 6-h sleep study conducted from 10:00 A.M. to 4:00 P.M. The dogs were allowed to settle in the cage for 15 min prior to the start of the study while electrical signals were checked.

Sleep State Scoring

A method of scoring sleep in the dogs was adapted from the methods most commonly used to identify human sleep states (33). The differences in scoring events in the dogs were that (1) unlike humans, the dogs with frontoparietal electrodes have neither prominent alpha activity nor K complexes, and (2) scoring epochs were 1 min long, and the sleep state for each epoch was defined as the predominant stage for the epoch. We defined sleep states in the dogs as follows. Waking: low voltage, high frequency EEG activity and an absence of spindles or delta frequency activity. Spindle NREMS: low voltage, mixed frequency EEG activity with spindles occurring at least once per minute. Slow-wave NREMS: waves of delta frequency for > 50% of the epoch. REMS: eyes closed, phasic twitches of facial musculature and low voltage, high frequency EEG. REMS is more difficult to determine in dogs than in humans, in that postural muscle atonia is not a pervasive phenomenon. For that reason, we required visual monitoring of the dog and documentation on the polygraphic record of the presence of phasic twitches to define REMS. The onset of REMS was defined as the beginning of either neck atonia or phasic facial twitches and an EEG pattern of high frequency. Records were scored independently by two scorers. The percentage agreement between scorers for all epochs in one entire record for each dog was > 98%.

Sleep Architecture

In addition to identifying the sleep state for each epoch, we calculated sleep efficiency, sleep latency, percentage of each sleep state, and the arousal index. Sleep efficiency was defined as the percentage of total test time in which the subject was in either NREMS or REMS. Sleep latency was defined as the time in minutes from lights out to the first three consecutive epochs of sleep. The percentage of each behavioral state, including waking, was calculated by dividing the time spent in each particular sleep/wake state by the overall test time. The NREMS arousal index was calculated as the number of arousals per hour of NREMS. Arousals were defined as > 10-s duration epochs with an abrupt shift to a faster frequency in NREMS or a > 10-s long increase in nuchal EMG activity in REMS.

Scoring Sleep-disordered Breathing Events

Sleep-disordered breathing events were classified as respiratory effort-related arousals (RERAs), or as apneas or hypopneas. The RERAs were identified as arousals (33) associated with either increased snoring immediately preceding an arousal or with paradoxical movements of the rib cage and abdomen preceding the arousal but without desaturations of at least 2%. An example of a RERA in the bulldog is provided in Figure 1, showing a brief period of paradoxical respiratory movement and end-inspiration, a prolonged reduction in airflow followed by an arousal, but no oxyhemoglobin desaturation. RERAs occur in humans with sleep-disordered breathing, and although the relative clinical significance of these events compared with hypopneas and apneas remains to be more fully defined, it is probable that such events contribute to daytime hypersomnolence (34). In addition, when using a thermistor to measure airflow, albeit approximately, these RERAs, at least in part, may represent hypopneas (35). Hypopneas in our study were defined as reductions in airflow by 50 to 90% associated with oxyhemoglobin desaturations of at least 2% with an arousal or of at least 4% if there was no arousal. Apneas were defined as cessations in airflow or reductions by > 90% and desaturations 2 to 3% with an arousal or 4% without an arousal. Apneas, hypopneas and RERAs were grouped together as respiratory disturbances, and the respiratory disturbance index (RDI) was calculated as the average number of apneas, hypopneas and RERAs per hour of sleep. The total time (in seconds) during SDB events spent with oxyhemoglobin saturations < 90% was calculated for each study and expressed as the RDI desaturation time.

View larger version (20K): [in this window] [in a new window]   Figure 1.   A polygraphic tracing showing a respiratory effort-related arousal from NREMS. Note that halfway through this epoch, there is a brief arousal detected in the electroencephalogram (EEG) followed by increased snoring, reduced activity of the sternohyoid muscle, seen in both the raw signal (SH EMG) and the rectified moving average (SH MA). The snoring intensity increases until an arousal is noted in the EEG signal. Oxyhemoglobin saturation was stable at 96%, despite this prolonged drop in airflow.

Measuring Upper Airway Dilator Muscle Activity

The sternohyoid muscle was chosen as the representative upper airway dilator because this muscle, compared with other upper airway muscles in the bulldog, demonstrates more pronounced respiratory-related phasic activity (15) and has been shown to have sleep-state- dependent reductions in activity (36). We looked exclusively at respiratory drive, measured as the amplitude of respiratory-related phasic activity (the amplitude difference between the peak amplitude of the moving average minus the tonic end-expiratory amplitude in arbitrary units) (36). Comparisons were made in respiratory drive across state changes between placebo and the three drug conditions. To accomplish this, respiratory drive for the SH was measured for six breaths prior to state change and the six breaths just after state change, excluding transitions in which either a sigh or SDB event occurred within 10 bursts prior to state change. The state changes for which the measurements were obtained were from waking to spindle NREMS and from slow-wave sleep (SWS) to REMS. The state change from spindle to SWS was not examined because there is no increase in the frequency of SDB events after this state change. For each animal, these changes in respiratory drive were recorded for all such state changes associated with respiratory-related phasic activity of the SH muscle throughout each study to obtain an average percent change for each animal in the group.

Data Analysis

Data analysis was designed to answer the following questions. (1) Does the administration of trazodone and L-tryptophan affect sleep efficiency, sleep latency, sleep state distribution, or sleep fragmentation (arousal index) in our animal model of the sleep apnea syndrome? (2) Does trazodone with L-tryptophan reduce the frequency (RDI) or severity (oxyhemoglobin saturation nadir, RDI, and arousal index) at any of the three doses compared with baseline? (3) Does this therapy impact upon state-related reductions in the phasic respiratory drive for upper airway dilator activity? (4) Does the combination therapy reduce the total study time for which SDB resulted in oxyhemoglobin saturations < 90%? (5) How is sleep architecture (the percentage of time spent in each of the sleep stages) affected? (6) Are any of these effects dose-dependent? Given our study design of placebo and three drug doses, the first five questions were addressed using two-way repeated-measures analysis of variance (ANOVA) with Tukey's pair-wise comparisons if overall significances were found, with the null hypotheses being rejected at a p < 0.05 level. In all of these statistical analyses, the sample size was n = 5, with df = 4. Dose-dependency was determined using nonparametric linear correlations (Spearman's rho correlation) between condition (placebo = 0, Dose 1 = 1, Dose 2 = 2, and Dose 3 = 3) and each of the above variables.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Five English bulldogs successfully completed all 16 randomized studies to determine the effects of three levels of trazodone and L-tryptophan dosing on sleep-wake behaviors and sleep-disordered breathing. Neither adverse effects nor changes in waking behaviors were observed during the studies.

Drug Effects on Behavioral State and Sleep-Wake Architecture

The combination of trazodone and L-tryptophan significantly increased sleep efficiency in the five dogs (group ANOVA, p = 0.0001). Group mean sleep efficiency ± standard error (SE), as defined by the percentage of the total 6-h test time spent asleep, was 45.6 ± 6.8% during placebo trials. For the three doses of trazodone and L-tryptophan, sleep efficiency, compared with placebo, was: Dose 1, 49.4 ± 6.5% (NS); Dose 2, 56.4 ± 4.3% (p = 0.09); Dose 3, 71.6 ± 3.3% (p = 0.005, n = 5). Sleep efficiency during Dose 3 was significantly greater than that for Dose 2 (p = 0.03). There was a dose-dependency effect of this drug on sleep efficiency, as evidenced by the linear correlation between drug dose and sleep efficiency (r² = 0.84, p = 0.04). Sleep efficiency responses to drug are shown in Figure 2A in individual dogs.

View larger version (14K): [in this window] [in a new window]   Figure 2.   Individual data are presented for the five dogs comparing the effects of trazodone and L-tryptophan with placebo on sleep efficiency and on the arousal index. Linear dose-responses in sleep efficiency and in the arousal index are evident for each subject. Mean data for the group are presented in RESULTS. For the group, there were significant dose-dependent increases in sleep efficiency and dose- dependent reductions in sleep fragmentation.

In contrast, sleep latency was not affected by trazodone and L-tryptophan. Group mean sleep latencies and standard errors for placebo and the three drug doses were: placebo, 17 ± 7 min; Dose 1, 21 ± 8 min; Dose 2, 7 ± 2 min; Dose 3, 12 ± 5 min. The overall differences by repeated measures ANOVA were not significant.

Trazodone with L-tryptophan significantly reduced sleep fragmentation, measured as the number of arousals per hour of sleep (p = 0.009, n = 5). The arousal index at baseline (placebo) for the group was 13.6 ± 3.0 arousals/h. The group mean values ± SE for each of the three doses, and Tukey's pairwise-comparisons to placebo were as follows. Dose 1 was 9.4 arousals/h (NS); the arousal index at Dose 2 was 6.2 ± 0.5 (p = 0.08), and for Dose 3 it was 5.0 ± 0.5 (p = 0.03, n = 5). A strong dose-dependency in response in arousal index was evidenced by the r² = 0.85 (r = 0.92), p = 0.001. The responses in the individual animals are illustrated in Figure 2B.

The effects of trazodone and L-tryptophan on sleep architecture may be best described as a shift from time spent in waking to increased time in slow-wave sleep, with minor effects on the amount of spindle (light) NREMS and REMS. The effects on wakefulness mirror the changes described for sleep efficiency with percentages of time spent in waking, significantly different overall (p = 0.003, n = 5). Percentage of study time spent in wakefulness for placebo and Doses 1, 2, and 3, respectively, were: 55.6 ± 6.7%, 49.2 ± 6.0%, 40.4 ± 5.9%, and 33.3 ± 5.7%. Compared with placebo, significant reductions in wakefulness were seen at Dose 2 (p = 0.01) and at Dose 3 (p < 0.004, n = 5). The difference between Doses 2 and 3 was not significant (p = 0.08). There were no significant effects on spindle NREMS compared with placebo, 22.0 ± 2.6% versus 19.9 ± 3.3% for Dose 1 (NS), 21.8 ± 1.8% for Dose 2 (NS), and 16.5 ± 1.8% for Dose 3 (NS). SWS, however, increased dramatically, with an overall p = 0.0004, n = 5. The placebo percentage of SWS was 17.8 ± 4.2%. With Dose 1 there was an increase to 26.2 ± 4.4% (p = 0.002) compared with placebo (n = 5). At Dose 2 there was an increase in SWS to 35.6 ± 4.8% (p = 0.0004), and at Dose 3, there was an increase to 47.0 ± 3.9% (p = 0.0004, n = 5). REMS, as mentioned above, was not affected. The percentages of REMS for placebo and the three doses were as follows: 4.8 ± 1.0%, 4.8 ± 0.4%, 4.0 ± 0.8%, and 3.7 ± 0.5% (NS). The percentage distributions of behavioral states under the four conditions are shown in Figure 3.

View larger version (48K): [in this window] [in a new window]   Figure 3.   The effects of trazodone and L-tryptophan on sleep architecture for the group (n = 5). Data are presented as mean for the group ± standard error. The upper left panel (A) shows the group effect on the percentage of wakefulness during the 6-h study. Significant reductions in time spent awake were seen, as well as a dose-dependency in response. See RESULTS for details. The upper right panel (B) shows a lack of effect of combination therapy on spindle (light) NREMS. The lower left panel (C ) shows significant increases in slow-wave NREMS in a dose-dependent fashion. The lower right panel (D) shows minimal effects on the percentage of REMS during study.

Further Characterization of SDB Events in the Bulldog in NREMS

With placebo, the predominant respiratory events were RERAs, comprising > 75% of all respiratory NREMS events in all dogs. An example of such an event is provided in Figure 1. Apneas with desaturations of > 4% comprised < 9% of the events. The group mean NREMS hypopnea index was 1.0 ± 0.3 hypopneas/ h, and the NREMS apnea index was 0.2 ± 0.1 apneas/h.

Effects of Trazodone and L-tryptophan on Sleep-disordered Breathing

Trazodone with L-tryptophan reduced the frequency of sleep-disordered breathing in both NREMS and REMS in a dose-dependent fashion. These reductions in sleep-disordered breathing were seen in all dogs individually, as well as in the group. The individual RDI responses in NREMS and REMS are illustrated in Figure 4. Although apneas and hypopneas are rare, the frequency of both types of respiratory events was reduced with Dose 3: 0.2 ± 0.1 hypopneas/h (p < 0.03, n = 5) and 0.0 ± 0.01 apneas/h (p = 0.05, n = 5). All dogs showed, individually, reductions in the NREMS RDI with trazodone and L-tryptophan at Doses 2 and 3, and the group effect of drug on NREMS SDB was significant (p = 0.02). Specifically, during NREMS (spindle and slow-wave sleep), the group RDI during placebo was 7.0 ± 1.7 events/h with placebo. The group mean RDI at Dose 1 was 5.2 ± 3.1 (p = 0.46). The RDIs for the group at Doses 2 and 3 and Tukey's comparisons to placebo were as follows: at Dose 2 the RDI was 2.2 ± 1.5 (p = 0.09); at Dose 3 the RDI was 0.8 ± 0.3 (p = 0.03). There was a negative correlation between dose and the NREMS RDI (r = 0.57, p < 0.04), suggesting a dose-dependency in response.

View larger version (17K): [in this window] [in a new window]   Figure 4.   Individual data for each dog are presented for combination drug effect (average for four trials ± SE) on the respiratory disturbance index, as defined by the number of respiratory effort- related arousals, hypopneas, and apneas in NREMS (A) and REMS (B). Individual dose-dependent responses in the REMS apnea and hypopnea indices are evident in all subjects. In comparison, lower doses of drug alleviated NREMS events compared with REMS events in all but the most severely affected dog.

Given the large drug effect on SWS time, we analyzed the RDI for light (spindle) NREMS separately by comparing placebo to Dose 3 of the drugs. With placebo, the RDI during light NREMS was 12.46 ± 3.18 events/h. At Dose 3, the light NREMS RDI dropped to 2.21 ± 1.02 (p = 0.03, n = 5).

Four of the five dogs had higher RDIs in REMS than in NREMS with placebo. Overall, there was a significant effect of trazodone and L-tryptophan on the group mean REMS RDI (p = 0.007). The placebo REMS RDI for the group was 31.4 ± 6.12 events/h. Dose 1 significantly reduced the REMS RDI to 18.4 ± 5.78 (p = 0.04). For Dose 2, however, there was not a significant reduction in RDI, 22.3 ± 7 (p = 0.27). Dose 3 reduced the REMS RDI to 11.5 ± 4.28 events/h (p = 0.002). A dose-dependency is shown with Spearman's rho = 0.53 (p = 0.02). The individual responses in the five dogs for NREMS and REMS respiratory disturbance indices are shown in Figure 4.

The overall effects of trazodone and L-tryptophan on oxygenation were a reduction in the total recording time spent with saturations < 90% (p = 0.0001). The group mean ± SE total time spent with Sa_O2 < 90% during placebo was 151 ± 67 s. At Doses 1 and 2, the mean time was not significantly different, 144 ± 89 s and 91 ± 37 s, respectively. There was, however, a significant reduction in the time spent with Sa_O2 < 90% at Dose 3, 83 ± 63 (p = 0.04). For the group, the mean oxyhemoglobin saturation nadirs during respiratory events were unchanged with drug treatment at any dose: placebo = 84 ± 2%, Dose 1 = 84 ± 3%, Dose 2 = 85 ± 3%, Dose 3 = 86 ± 3%. Results of the effects of trazodone/L-tryptophan on oxyhemoglobin saturations are summarized in Table 1. At all levels of trazodone and L-tryptophan, the REM sleep-onset central apneas persisted, whereas the frequency and severity of events later in REMS were reduced with this combination therapy.

The Effects of Trazodone and L-Tryptophan on Upper Airway Dilator Activity

Each of the five dogs had some respiratory-related phasic EMG activity on the sternohyoid muscle. In all of these dogs, however, respiratory-related EMG activity of the sternohyoid was visually apparent on the polygraphic records for just 10 to 50% of the recording period. As is true with EMG activity in general, the SH EMG activity was positional, such that with an arousal and a head turn, the SH phasic activity would change significantly. Therefore, we elected to determine if there were changes in the peak phasic component of the SH EMG moving average amplitude (a measure of respiratory drive) across two different state changes: quiet waking into spindle NREMS and slow-wave sleep into REMS. Because we did not measure voltage and calibrate signals, tonic activity could not be compared. SDB events are most likely to occur, as in humans, during light NREMS and REMS. Thus, the state changes across which we measured the amplitude change in SH phasic respiratory activity were from quiet wakefulness to light NREMS and from slow-wave sleep NREMS to REMS. We compared phasic amplitude of SH moving average amplitude by averaging all bursts occurring 1 min just prior to state change, with all bursts occurring 1 min after state change. For each condition (drug dose), data were averaged for each dog from all samples of the specific state changes that had respiratory activity for each of the four studies. The individual data are presented in Figure 5A for the state change from wakefulness to light NREMS and in Figure 5B for SWS to REMS. As a group, there were no significant changes from wakefulness to light NREMS (p = 0.21) in SH phasic respiratory activity at any of the three doses compared with placebo. The percentage change in SH peak amplitude from wake to light NREMS with placebo was 3.2 ± 7.1%; at Dose 1, the change was 22.2 ± 13.2%. At Dose 2, the change was 20.8 ± 13.9%, and at Dose 3, the change was 21.1 ± 6.3%. In contrast, trazodone and L-tryptophan did not reduce the REMS-associated suppression of upper airway dilator activity (p = 0.014). The values, expressed as percentage change across sleep-state transition for placebo, Doses 1, 2, and 3 were as follows: 38.8 ± 9.4%, 38.0 ± 5.1%, 44.0 ± 9.1%, and 16.1 ± 6.8%. The significance was seen only at the highest dose (p < 0.02). Whenever phasic SH activity could be visualized during a SDB event, suppression of SH peak activity was noted to coincide with the event. This was true for NREMS as well as for REMS. There were, however, further differences in muscle suppression in NREMS when compared with REMS. First, the percentage suppression for a given animal was always greater in REMS than in NREMS. Second, REMS almost always began with an apnea, whereas it was rare to have an NREMS event occur at state transition.

View larger version (22K): [in this window] [in a new window]   Figure 5.   The effect of trazodone and L-tryptophan on individual dogs' sleep-state-dependent reductions in upper airway muscle activity. Values are averaged for the four trials for each dose for each animal. The left panel (A) has data for the state transition of waking to light NREMS, and the right panel (B) has the state transition from SWS-NREMS to REMS. Notice that this effect has less of a dose-dependency when compared with the effect on SDB indices.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have shown, in our natural animal model of OSAHS, that the combination of trazodone and L-tryptophan significantly reduces sleep-disordered breathing in both NREMS and REMS. Additionally, this combination therapy improves sleep efficiency, increases slow-wave sleep time, reduces sleep fragmentation, and reduces the magnitude of sleep-related suppression of upper airway dilator muscle activity.

We believe that these findings are clinically relevant to the human OSAHS. The need for a well-tolerated effective drug therapy for human OSAHS is readily evident, and there are extensive data supporting the English bulldog as an animal model for the human OSAHS. Although the sleep-disordered breathing in the dogs best represents REMS-related OSAHS and upper airways resistance syndrome, rather than obstructive sleep apnea, we believe that all of these diseases are part of a spectrum of diseases with very similar pathophysiology. The pathogenesis of OSAHS in the bulldogs (14) is very much like that observed in humans. Like humans, bulldogs with OSAHS have intrinsically narrowed upper airways (14). During wakefulness and normal respiration, both humans (38) and bulldogs (15) with sleep apnea have increased activity of their upper airway dilator muscles. It is only in sleep when these muscles are intermittently suppressed that persons (9) and bulldogs (36) with sleep apnea have obstructive sleep-disordered breathing. In this study, we report that like humans, these events may be apneas, hypopneas, and respiratory effort-related arousals and may occur in NREMS as well as in REMS. In the present study, the severity of NREMS SDB may have been underestimated by the use of thermistors instead of pressure transducers.

There are also data to suggest that the sleep-state-dependent neural mechanisms underlying sleep-related suppression in upper airway dilator control may be somewhat similar among mammalian species. In all mammalian species studied to date, serotonin augments motoneuronal activity in the mature animal (for review, see Reference 39). In particular, serotonin powerfully excites hypoglossal (10, 11) and trigeminal (25) motoneurons, and there are data to support the hypothesis that serotonin plays an important role in the maintenance of hypoglossal tone in rats (11) and in cats (24) and in the maintenance of upper airway dilator activity in the English bulldog (13). In humans with obstructive sleep apnea, however, trials of serotonniergic drugs show small to no group effect on sleep-disordered breathing (16). We argue that these minimal effects in humans may relate more so to study design and drug selection than to serotonin being less effective in humans.

In human trials of serotoninergics for OSAHS, the most promising findings were seen with L-tryptophan, with which all persons (n = 12) had an improvement in the apnea/hypopnea index (16). Six of the patients were studied after chronic therapy (> 2.5 mo of L-tryptophan). In the humans studied, L-tryptophan had its impact primarily on NREMS SDB rather than on REMS SDB. A greater improvement in NREMS events was also seen in our bulldogs, defined as a greater percentage reduction in the frequency of events in NREMS compared with REMS and a normalization of respiration in NREMS (15). This state-dependent difference in effect may, at least in part, be due to more profound reductions in serotoninergic activity in REMS than in NREMS (12).

A second group of drugs that augments endogenous 5-HT activity are the selective serotonin reuptake inhibitors (SSRIs). One of these SSRIs, fluoxetine, has been evaluated as a treatment for OSAHS in a study of 12 patients covering the whole range of severity of OSAHS (17). The drug caused a statistically significant reduction in the apnea-hypopnea index (AHI), although one patient worsened and one did not change at all, and overall, the changes were small (from 57/h to 34/h). As with L-tryptophan, there was no effect on the AHI in REMS. A second preliminary study of paroxetine, another 5-HT reuptake inhibitor, showed very similar findings, a mild effect on NREMS AHI (from 37/h to 30/h) and no effect on REMS AHI (19).

Currently, it is not known whether 5-HT reuptake inhibitors can prevent sleep-state-dependent reductions in 5-HT levels in brainstem motor nuclei, or if a precursor such as L-tryptophan can maintain high 5-HT activity across states. More importantly, it is not known whether serotoninergics that work exclusively by increasing endogenous 5-HT can increase 5-HT at upper airway dilator motoneurons sufficiently to prevent sleep-disordered breathing in REMS, even at higher doses. It may be that neither type of serotonniergic (SSRIs and precursor drugs) can boost endogenous 5-HT levels adequately to prevent REMS-related suppression in upper airway muscle activity. This may relate, not only to the profound suppression of 5-HT activity, but also to withdrawal of other excitatory influences and direct REMS-related inhibitory influences at brainstem motoneurons. It has been shown, however, that serotonin microinjected into the hypoglossal nucleus can prevent suppression of hypoglossal activity in the carbachol model of REMS atonia (40). Thus, to alleviate REMS RDI, it may be necessary to use 5-HT agonists, not only to maintain a wakefulness level of 5-HT activity at upper airway motoneurons in sleep, but to further augment 5-HT activity to offset the nonserotoninergic mechanisms of upper airway dilator suppression. One possible explanation for the improvement in the REMS SDBI with trazodone and L-tryptophan in the bulldogs is that the metabolite of trazodone, m-CPP, a broad spectrum 5-HT agonist, helped reduce REMS-related suppressions in upper airway dilator activity.

The only 5-HT agonist tested in humans for its effectiveness against SDB has been buspirone, a 5-HT_1A partial agonist. In the adult mammal, there is no appreciable 5-HT_1A activity at hypoglossal motoneurons (10, 11, 24). 5-HT_1A agonists, however, have been shown to excite medullary respiratory neurons (41). The optimal 5-HT receptor subtypes to target for OSAHS are not known. There may be several different 5-HT receptor subtypes involved in the excitation of upper airway dilator motoneurons (24, 42, 43). A study in adult rats has shown that mRNA for many different 5-HT receptor subtypes is present in the area of the hypoglossal nucleus (43). The 5-HT receptor subtypes that were within this region and have been shown at other loci to demonstrate excitatory activities are 5-HT_{2A, 2C, 6, 7} (43). Novel 5-HT receptors subtypes are rapidly being discovered, such that it is also possible that a yet unidentified 5-HT receptor subtype plays an important excitatory role at upper airway dilator motoneurons. In pursuit of a specific 5-HT agonist for OSAHS, it will be important to discern if the receptor subtypes in humans and in animals with normal respiration parallel those in beings with narrowed upper airways and OSAHS.

From the present study, it is not possible to determine which aspects of the drug combination of trazodone and L-tryptophan were important in reducing SDB and improving sleep quality. Potential serotoninergic mechanisms for reducing SDB include direct 5-HT excitatory effects at upper airway motoneurons (10, 11) through increased production of 5-HT after administering its precursor, L-tryptophan (23), or through direct excitation by trazodone's metabolite, m-CPP (27), as well as excitation of respiratory-related premotor neurons through similar mechanisms (47, 48). 5-HT does not always excite respiratory activity. In the neonatal rat, 5-HT_1A receptors are expressed at the hypoglossal nucleus and the effect of locally applied 5-HT in the neonatal rat hypoglossal nucleus is inhibition of respiratory activity. Additionally, activation of 5-HT_{2A, 2C/or 3} receptors at the nodose ganglion markedly suppresses respiration. This latter effect of serotonin may explain results obtained with systemically administered 5-HT drugs in anesthetized rats with infranodal vagotomies. Under these conditions, systemically administered 5-HT₂ antagonists excite respiratory activity.

In addition to possible serotoninergic effects on upper airway motoneurons or on respiratory-related neurons, it is also possible that these drugs impacted upon sleep-disordered breathing indirectly by stabilizing sleep. As in humans, sleep-disordered breathing in the bulldogs is typically not seen in Stage III/IV (or slow-wave) NREMS. In the bulldogs, trazodone with L-tryptophan produced a marked dose-dependent increase in slow-wave sleep, without a change in light NREMS. Thus, the percentage of NREMS time spent in the more vulnerable light NREMS went from 56% of NREMS time to 29% of NREMS time at the highest dose, approximately a 50% reduction in the light NREMS proportion of NREMS. The percentage drop in the NREMS SDBI was, however, far greater (88%), such that the reduction in SDB cannot be explained exclusively by the changes in sleep architecture. Further, the significant decline in the light NREMS RDI with drug also supports the idea that the drugs' effects were at least in part due to increased respiratory drive.

The changes in respiratory drive to the sternohyoid muscle were rather small compared with the changes in SDB, also raising the possibility that the drugs acted through mechanisms other than to augment serotonin activity at upper airway motoneurons. The NREMS changes in SH muscle activity were insignificant; although there were not enough data to measure NREMS-related suppression only at SDB events, as was done for REMS, where a SDB almost always occurs at REMS onset. Thus, the NREMS-related changes with and without drug are likely minimized by our methods of data analysis. In our previous work with serotonin antagonists in bulldogs, rather small reductions (20 to 50%) in geniohyoid and sternohyoid activity were associated with complete collapse of the upper airway in two dogs with severe SDB. Therefore, it is possible that the improvement in upper airway activity does not need to be of great magnitude to alleviate SDB. Further, the sternohyoid is just one of the many upper airway dilator muscles, and differential responsiveness to serotonin has been shown for brainstem motoneuronal pools (49).

Before initiating human trials of this combination therapy for the OSAHS, there are unresolved concerns with the safety of L-tryptophan in humans (50). It remains unclear whether the serious adverse effects of L-tryptophan (eosinophilic-myalgia syndrome, pulmonary hypertension, and death) were related to L-tryptophan or to contaminants in processing the drug. That is, manybut not allcases of adverse effects were traced to several manufacturers. A second potential concern with this drug combination is that trazodone, especially at high doses equivalent to those used in the present study, may cause severe priapism. With these concerns, we believe that a major significance of the present study is that in this study we have shown that a serotoninergic combination therapy was universally effective and well-tolerated and did not disrupt sleep in our animal model. This work, therefore, demonstrates the potential for serotoninergics in treating OSAHS and provides further justification to elucidate the precise mechanisms through which serotonin activity relates to sleep-disordered breathing. With this elucidation, some targeted approaches to pharmacotherapy are likely to emerge.