Respiratory Physiology & Neurobiology

Respiratory Physiology & Neurobiology 162 (2008) 48 54 Contents lists available at ScienceDirect Respiratory Physiology & Neurobiology journal homepage: www.elsevier.com/locate/resphysiol Bench study of auto-cpap devices using a collapsible upper airway model with upstream resistance Minoru Hirose a, Junichi Honda b, Eiji Sato a, Toshihiro Shinbo a, Kenichi Kokubo a, Toshio Ichiwata c, Hirosuke Kobayashi a,d, a Department of Clinical Engineering, School of Allied Health Sciences, Kitasato University, Kanagawa 228-8555, Japan b Clinical Engineering, Fujinomiya City General Hospital, Shizuoka 418-0076, Japan c Department of Respiratory Medicine, Dokkyo Medical University Koshigaya Hospital, Saitama 343-5888, Japan d Division of Respiratory Medicine, Kitasato University Hospital, Sagamihara, Kanagawa 228-8555, Japan article info abstract Article history: Accepted 28 March 2008 Keywords: Sleep apnea syndrome Mechanical ventilation Auto-CPAP Bench study Upper airway model The aim of this study was to investigate the response of auto-cpap devices to respiratory events (apnea, hypopnea, flow-limitation and snoring) on the same condition using a physiological upper airway model. The hypothesis of this study is that collapsibility of the flow-limiting collapsible segment of the airway is influenced by the upstream airway resistance. Five auto-cpap devices, AutoSet T, AutoSet Spirit TM, Goodnight 420E, PV10i and REMstar Auto were evaluated. Apnea: all the devices increased the auto- CPAP level, while AutoSet T and AutoSet Spirit TM did not respond to apnea for 30 s. Hypopnea: all the devices except the AutoSet T and Goodnight 420E increased pressure. Flow-limitation: all the devices except the PV10i and REMstar Auto increased pressure. Snoring: the snoring sounds disappeared when REMstar Auto and PV10i were used, and the Goodnight 420E lowered the level of snoring. In conclusion, the response of auto-cpap devices to respiratory events differed. Collapsible upper airway model with upstream resistance is useful for the first-step assessment of auto-cpap devices. 2008 Elsevier B.V. All rights reserved. 1. Introduction Sleep-disordered breathing (SDB) is accompanied by several undesirable states, such as headache at awakening and sleepiness in daytime. Obstructive sleep apnea/hypopnea syndrome (OSAHS) accounts for 90% of patients with SDB (Bradley and Phillipson, 1992). Currently available medical treatments for OSAHS include (1) life style improvement (exercise, weight loss, cessation of smoking, temperance, etc.), (2) application of nasal continuous positive airway pressure (ncpap) during sleep, (3) the use of dental splinting, such as mandible advance devices, and (4) upper airway surgery. Among these treatments, ncpap is the first choice therapy in evidence-based medicine (Hoffstein et al., 1992; Polo et al., 1994; Hudgel, 1996). When applying ncpap to OSAHS patients, it is important to determine optimal treatment pressure for individual patients (Hsu and Lo, 2003; Akashiba et al., 1999). The suitable pressure is determined by titration using the CPAP device, i.e., by Corresponding author at: Department of Clinical Engineering, School of Allied Health Sciences, Kitasato University, Kitasato 1-15-1, Sagamihara, Kanagawa 228-8555, Japan. Tel.: +81 42 778 9711; fax: +81 42 778 9711. E-mail addresses: hirose@kitasato-u.ac.jp (M. Hirose), ei-san@kitasato-u.ac.jp (E. Sato), tshinbo@kitasato-u.ac.jp (T. Shinbo), kokubo@kitasato-u.ac.jp (K. Kokubo), ichiwata@dokkyomed.ac.jp (T. Ichiwata), hiro@kitasato-u.ac.jp (H. Kobayashi). adjusting the pressure manually or adjusting the pressure automatically (auto-cpap) (Meurice et al., 1996; Sharma et al., 1996; Konermann et al., 1998). The auto-cpap devices deliver variable pressure levels in response to the detection of respiratory events (apnea, hypopnea, flow-limitation and snoring) based on their own algorithm. Auto-CPAP is now used not only for titration to determine the fixed level of CPAP, but also for the treatment of OSHAS patients at home. Although many auto-cpap devices are available now, the algorithm used by each device considerably differ each other. Since quality assessments of these auto-cpap devices are usually performed during direct clinical application to individual patient, it is difficult to compare the auto-cpap devices under the same upper airway and breathing condition. The aim of this study was to investigate the response of auto- CPAP devices to several respiratory events using a physiological upper airway model. Recently, bench studies have been proposed (Farré et al., 2002; Abdenbi et al., 2004; Lofaso et al., 2006; Rigau et al., 2006). Among them Abdenbi et al. (2004) developed a physiological collapsible upper airway model, but they did not attach an upstream airway resistance to the collapsible rubber tube, i.e., the resistance between auto-cpap device and the collapsible tube. We first tried to examine auto-cpap devices without the upstream airway resistance, but we could not observe the tube collapse and flow-limitation. If there is no upstream resistance, the 1569-9048/$ see front matter 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.resp.2008.03.014

M. Hirose et al. / Respiratory Physiology & Neurobiology 162 (2008) 48 54 49 pressure in the collapsible tube should be atmospheric pressure, and the tube will be collapsed only when surrounding pressure is larger than the atmospheric pressure. However, this situation is considered to be limited in some extremely severe OSHAS patients, since surrounding pressure should not be so higher than the atmospheric pressure. The hypothesis of this study is that collapsibility of the flowlimiting collapsible segment of the airway is influenced by the upstream airway resistance, and based upon this hypothesis we have added the upstream airway resistance to the collapsible tube. We used pressure-controlled ventilation on the lung-side to simulate breathing rather than volume-controlled ventilation as Abdenbi et al. (2004) used, since volume-controlled ventilation creates stable ventilation and flow, irrespectively of the airway resistance, or there might be air-leak somewhere between the collapsible tube and the lung, which is not physiological. We attached the upper airway resistance between the auto- CPAP devices tested and the collapsible rubber tube, since the pressure in the collapsible airway is influenced by its upstream resistance on inspiration and the pressure detected by auto- CPAP device is also influenced by the resistance between the device and the collapsible site, and we verified the responses of several currently available auto-cpap devices to the simulated respiratory events. We found that the response of auto-cpap devices to the respiratory events considerably differed each other. 2. Methods 2.1. Collapsible upper airway model with upstream resistance The cross-area of an upper airway is determined by balance between the pressure in the airway and the pressure in the surrounding tissue (Remmers et al., 1978; Brouillette and Thach, 1979), and a net negative pressure results in the narrowing of the upper airway. The cross-sectional area of the upper airway decreases during inspiration and increases during expiration. Therefore, the upper airway model was designed as an assembly equipped with a collapsible elastic-tube (with a diameter of 18 mm and a length of 48 mm) housed in a Plexiglas box to simulate flow-limitation and snoring by adjusting the pressure surrounding the elastic-tube in the box (Fig. 1). This collapsible upper airway model is similar to the set-up used by Abdenbi et al. (2004), but we further added upstream (mouth-side) resistance of 50 cmh 2 O/(L s) to their model. Fig. 1. Upper airway model. An upper airway model equipped with an elastic-tube was housed in a Plexiglas box. The surround pressure was adjusted by applying a positive pressure to the interior of the Plexiglas box. 2.2. Verification of flow-limitation and snoring in the upper airway simulation Flow-limitation in the upper airway was verified using a test circuit as follows (Fig. 2). A vacuum pump attached to a watersealed equipment was connected downstream of the upper airway model to adjust the downstream negative pressure by altering the height of water in the water-sealed equipment; the upstream pressure was adjusted by the surrounding pressure of the collapsible tube. The flow and pressure before and after the upper airway model were measured using two pressure-flow analyzers (Calibration-Analyzer RT-200, Timeter Instruments Co., St. Louis, MO, USA), and the surrounding pressure of the elastic-tube was varied from 0 to 20 cmh 2 O for flow-limitation and 10 cmh 2 Ofor snoring. Flow-limitation was then created by changing the level of the surrounding pressure with fixed upstream-resistance of 50 cmh 2 O/(L s). This level of upstream-resistance was determined, since snoring sounds were generated by adjusting the surrounding pressure of the elastic-tube at that upstream-resistance level. The pressure vibration was measured directly by the pressure-flow analyzer, and the snoring sounds were recorded by a digital recorder (Voice-Trek DS-10, Olympus Co. Ltd., Tokyo, Japan). The pressure frequency of snoring was checked using both the pressure record and the frequency spectrum of the sounds analyzed using FFT analysis software (FFT Wave, E.N. Software, Tokyo, Japan). Fig. 2. Test circuit for upper airway model. A vacuum pump was connected downstream of the upper airway model, and the flux in the pressure difference before and after the upper airway model was measured to confirm flow-limitation. The negative pressure was adjusted according to the height of water in the water-seal equipment.

50 M. Hirose et al. / Respiratory Physiology & Neurobiology 162 (2008) 48 54 Fig. 3. Test circuit for evaluating algorithms of auto-cpap devices. Each auto-cpap device was evaluated using the test circuit. Spontaneous breathing was simulated as a negative pressure swing downstream of the simulated upper airway model. The negative pressure swing was generated in one chamber of the test lung, to which another chamber was firmly fixed and was ventilated with a ventilator (Puritan Bennett 7200ae, Puritan Bennett Co., Carlsbad, CA, USA) in the pressure-controlled ventilation mode with a presumed tidal volume of 0.5 L and a respiratory rate of 10 breaths/min. In this manner, the positive pressure swing generated by the ventilator was converted to a negative pressure swing downstream (lung-side) of the upper airway model (collapsible tube connected with mouth-side airway resistance). 2.3. Evaluation of the response of auto-cpap devices to respiratory events Each algorithm of five auto-cpap devices was evaluated: AutoSet T (Resmed Ltd., Sydney, Australia), AutoSet Spirit TM (a newer version of AutoSet T), Goodnight 420E (Tyco Healthcare, USA), PV10i (BREAS, Sweden), and REMstar Auto (Respironics, CA, USA). The auto-cpap devices were evaluated using a test assembly (Fig. 3) in which spontaneous breathing was simulated as a negative pressure swing downstream of the simulated upper airway model, to which the two-chamber Michigan test lung (TTL, Michigan Instruments Inc., Grand Rapids, MI, USA) was connected as was used by Lofaso et al. (2006). The negative pressure swing was generated in one chamber of the test lung, to which another chamber was firmly fixed and was ventilated with a ventilator (Puritan Bennett 7200ae, Puritan Bennett Co., Carlsbad, CA, USA) in the pressure-controlled ventilation mode with a presumed tidal volume of 0.5 L and a respiratory rate of 10 breaths/min. In this manner, the positive pressure swing generated by the ventilator was converted to a negative pressure swing downstream (lung-side) of the upper airway model (collapsible tube connected with mouth-side airway resistance). The Auto-CPAP devices were connected to the mouth-side airway resistance of the upper airway model, and the flow and pressure upstream and downstream of the upper airway model were measured. The time course of the pressure level generated by each auto-cpap device was measured under four kinds of simulated respiratory events, as follows: Fig. 4. Flow pattern of sleep events: (a) control changing to apnea; (b) control changing to hypopnea; (c) control changing to flow-limitation; (d) control changing to snoring. Each flow pattern was confirmed before the pressure response of auto-cpap devices was measured.

M. Hirose et al. / Respiratory Physiology & Neurobiology 162 (2008) 48 54 51 (a) Apnea: the upstream tube in the upper airway model was closed for over 36 s until pressure change was observed. (b) Hypopnea: the tidal volume of the ventilator was reduced from 0.5 to 0.2 L for 48 s. (c) Flow-limitation: the pressure surrounding the elastic-tube was changed from 0 to 20 cmh 2 O. (d) Snoring: a surrounding pressure of 0 20 cmh 2 O was applied and snoring sounds were generated. Each flow pattern was confirmed (Fig. 4) before the pressure response of auto-cpap devices was measured, and to assess repeatability of the response of auto-cpap device, each test was repeated three times after resetting the device (power ON/OFF). The pressure at the start of measurements was set 4.0 cmh 2 Oin each auto-cpap device. 3. Results 3.1. Verification of flow-limitation and snoring in the upper airway simulation The flow through the upper airway model increased as the pressure difference between the upstream and downstream regions of the upper airway model increased but reached a lower plateau level when the pressure surrounding the elastic-tube was increased (Fig. 5), indicating a flow-limitation. Sounds that resembled snoring were generated when a surrounding pressure of 10 cmh 2 O was applied. The center frequencies of the snoring ranged from 280 to 300 Hz, as confirmed by both the pressure record and the frequency spectrum of the sounds (data not shown). These frequencies correspond to the reported center frequencies of natural snoring (Agrawal et al., 2002). 3.2. Evaluation of the response of auto-cpap devices to respiratory events 3.2.1. Apnea The pressure patterns generated by each auto-cpap device in response to apnea were as follows (Fig. 6). Goodnight 420E: the pressure began to raise 6 s after the start of apnea and rose to 5.0 cmh 2 O at 12 s after the start of apnea. REMstar Auto: the pressure began to raise 12 s after the start of apnea and rose to 5.0 cmh 2 O at 18 s after the start of apnea. PV10i: the pressure gradually began to raise 6 s after the start of apnea and rose to 5.3 cmh 2 O at 30 s after the start of apnea. AutoSet T and AutoSet Spirit TM : Fig. 6. Pressure response during apnea. All of the auto-cpap devices responded to apnea and increased the pressure. However, each device generated the pressure differently. the pressure did not change with apnea at the first time, but it suddenly began to rise 36 s after the start of apnea at the second time and rose to 9.2 cmh 2 O with the AutoSet T and 7.7 cmh 2 O with the AutoSet Spirit TM. 3.2.2. Hypopnea The pressures generated by each auto-cpap device in response to hypopnea were as follows (Fig. 7). AutoSet T and Goodnight 420E did not respond to hypopnea. PV10i: the pressure gradually began to rise 6 s after the start of hypopnea and rose to 5.3 cmh 2 O after about 30 s. REMstar Auto: the pressure began to rise after 12 s of hypopnea and rose to 5.0 cmh 2 O at 18 s after the start of hypopnea. AutoSet Spirit TM : the pressure began to raise 18 s after the start of hypopnea and rose gradually to 4.7 cmh 2 Oat 48 s after the start of hypopnea. 3.2.3. Flow-limitation The pressures generated by each auto-cpap device in response to flow-limitation were as follows (Fig. 8). AutoSet T: the pressure began to rise to 6.2 cmh 2 O at a surrounding pressure of 15 cmh 2 O and to about 10 cmh 2 O at a surrounding pressure of 20 cmh 2 O. Goodnight 420E: the pressure began to rise to 4.5 cmh 2 O at a surrounding pressure of 15 cmh 2 O and to 6.5 cmh 2 O at a surrounding pressure of 20 cmh 2 O. PV10i: the pressure did not rise within the surrounding pressure range investigated, i.e., from 0 to 20 cmh 2 O. REMstar Auto: the pressure began to rise to 7.5 cmh 2 O only when Fig. 5. Verification of flow-limitation. Flow-limitation was verified, since the flux rate at a large driving pressure was limited to a lower level as the surrounding pressure around the elastic-tube increased. Fig. 7. Pressure response during hypopnea. The AutoSet T and Goodnight 420E devices did not respond to hypopnea, whereas the other devices responded and increased the pressure.

52 M. Hirose et al. / Respiratory Physiology & Neurobiology 162 (2008) 48 54 Fig. 8. Pressure response during flow-limitation. All the devices except PV10i responded to flow-limitation and increased the pressure. Fig. 9. Pressure response during snoring. Snoring sounds disappeared when the REMstar Auto and PV10i devices were used, and the Goodnight 420E lowered the snoring sound level. The AutoSet Spirit TM and AutoSet T devices did not change the snoring sound level. the surrounding pressure was 15 cmh 2 O and did not rise at a surrounding pressure of 20 cmh 2 O. 3.2.4. Snoring The pressures generated by each auto-cpap device in response to snoring were as follows (Fig. 9). At a surrounding pressure of 10 cmh 2 O, snoring sounds were generated at a CPAP pressure of 4cmH 2 O, the sound level decreased when the CPAP pressure was raised to 4.5 cmh 2 O, and the snoring sounds disappeared at a pressure of 5 cmh 2 O. The AutoSet T and AutoSet Spirit TM devices did not respond to the simulated snoring. Goodnight 420E: the CPAP pressure gradually increased after 24 s and rose to 4.5 cmh 2 O after 48 s, when the snoring sounds disappeared. PV10i: the CPAP pressure rose gradually 90 s after the start of snoring, and the snoring sounds disappeared about 3 min after the start of snoring. REMstar Auto: the CPAP pressure suddenly rose to 5 cmh 2 O about 1 min after the start of snoring and the snoring sounds disappeared. 4. Discussion 4.1. Examination of auto-cpap devices under the same condition The number of patients who use auto-cpap device is increasing worldwide. New auto-cpap devices are being developed, and many kinds of devices are being introduced into the medical market, including devices aimed at medical care at home. When a patient uses an auto-cpap device, the optimal type of auto-cpap must be determined for each patient, but the comparison of several kinds of auto-cpap devices under the same condition in a clinical test is not feasible. To evaluate the response of different auto-cpap devices to respiratory events, each response must be evaluated under the same bench-test condition. In order to perform the bench testing of auto-cpap device, the upper airway simulation model which can imitate respiratory events (apnea, hypopnea, flow-limitation and snoring) are required. Farré et al. (2002) evaluated the response of auto-cpap devices using an open-loop computer-controlled breathing waveform. This breathing waveform was created by a servo-controlled pump which was able to produce flow pattern with any respiratory events stored in the computer. Rigau et al. (2006) further developed this method using a closed-loop feedback control. Lofaso et al. (2006) also evaluated auto-cpap devices using the similar method. These studies, however, did not use physiological upper airway model, whereas only Abdenbi et al. (2004) evaluated auto-cpap devices using a physiological collapsible upper airway model. However, Abdenbi et al. (2004) did not take upstream resistance (the airway resistance between the auto-cpap device and the collapsible tube) into consideration. The pressure in the collapsible airway is influenced not only by the surrounding pressure but also by its upstream resistance on inspiration, and the pressure detected by auto-cpap device is also influenced by the resistance between the device and the collapsible site. Therefore, we considered that the resistance between the auto-cpap device and the collapsible site should be taken into account. This model has a potential advantage over previous bench methods in investigating the interaction between the applied CPAP and re-opening of upper airway in a more physiological setting. In this study we confined air-leak to the conventional leak of auto-cpap devices, and did not evaluated the effect of mask-leak on the response of auto-cpap devices, since the mask-leak is variable among patients and depends on individual patient and mask fitting. We did not place cardiogenic oscillation between the collapsible tube and the lung component. It should be noted that the influence of the mask-leak as well as the cardiogenic oscillation to the response of auto-cpap devices was out of the scope of this study. In this study we focused on relatively acute response (within 1 min) of auto-cpap devices, since long-term (20 45 min) fixed upper airway condition is unlikely in clinical settings and long term response is considered to be cumulative steps of acute responses. 4.2. Features of the auto-cpap devices Five kinds of auto-cpap devices and their response were evaluated using our collapsible upper airway with upstream resistance, modeling spontaneous breathing and the respiratory events (apnea, hypopnea, flow-limitation, and snoring), and the features of these devices were as follows. 4.2.1. AutoSet T and AutoSet Spirit TM As designed by the manufacturer, the pressure generated by both AutoSet T and AutoSet Spirit TM did not increase during apnea episode and rose only after the apnea improved, but the pressure levels after apnea at the first time increased higher than those of other auto-cpap devices. It is very likely that these auto-cpap devices were designed not to produce an extremely high pressure during apnea which would be needed to open the closed upperairway, but were designed to maintain sleep quality, by preventing repeated apnea. AutoSet T and AutoSet Spirit TM maintained a sufficient flow, even if the surrounding pressure of the elastic-tube was increased from 0 to 20 cmh 2 O. Although they did not respond to single episodes of hypopnea, their generated pressure increased during episodes of hypopnea with flow-limitation. Although the characteristic responses of AutoSet T and AutoSet Spirit TM to respiratory events were almost the same, the reaction to hypopnea was added to AutoSet Spirit TM. When the AutoSetT and the AutoSet Spirit TM were compared, the AutoSet Spirit TM was con-

M. Hirose et al. / Respiratory Physiology & Neurobiology 162 (2008) 48 54 53 sidered to be more competent at detecting flow-limitation than the AutoSet T, since the AutoSet Spirit TM was able to react potently to the increased surrounding pressure. Therefore, AutoSet T and AutoSet Spirit TM devices are considered to be suitable for patients with frequent strong airway obstruction or narrowing, such as patients with thick layers of tissue surrounding their upper airway or those who have a large tongue, and who may sleep poorly with a large pressure increase during apnea as is generated with other auto-cpap devices. 4.2.2. Goodnight 420E Although the Goodnight 420E responded to apnea, it did not respond to hypopnea. Although the device responded to flow-limitation, flow was not maintained when the surrounding pressure of the elastic-tube was more than 15 cmh 2 O. Although the level of snoring sound was lessened, the snoring sound did not disappear completely, since the pressure supplied by the Goodnight 420E was mild. Consequently, the Goodnight 420E is considered to be suitable for patients with mild upper airway obstructions or narrowing. 4.2.3. PV10i Although the PV10i responded to apnea, hypopnea, and snoring, it did not respond to flow-limitation. The response of PV10i for respiratory events differed from those of other models. Other models raised the pressure linearly, but the PV10i raised the pressure gradually until it reached a suitable pressure. The responses for adjusting the pressure level in other models were determined by the number and the duration of respiratory events. The algorithm of the PV10i, in which a method similar to speech recognition technology is used according to the manufacturer s data book, differs from the algorithm of other devices, and PV10i raised the pressure gradually compared with other models, monitoring the pressure level every 2.5 s. Therefore, the PV10i is suitable for patients who are sensitive to sudden increase in applied pressure and for those who do not have strong upper airway obstruction. 4.2.4. REMstar Auto The REMstar Auto responded to all the respiratory events: apnea, hypopnea, flow-limitation, and snoring. According to the manufacturer s data sheet, the REMstar Auto has been designed to respond to apnea twice per minutes when 80% reduction in ventilation occurs for 10 s. Therefore, the REMstar Auto is suitable for patients having all kinds of respiratory events, and it may also be used for the titration of CPAP pressure. However, REMstar Auto reacted only when the surrounding pressure was 15 cmh 2 Oinflowlimitation, but did not react when the surrounding pressure was less than or more than 15 cmh 2 O in flow-limitation. The reaction of REMstar Auto to flow-limitation may be unstable for patients with variable level of upper airway collapse. 4.3. Comparison with other bench studies The same devices evaluated in other bench studies were compared. AutoSet T was evaluated by Farré et al. (2002), where the pressure increase in auto-cpap device were observed in repetitive respiratory events and in prolonged flow-limitation for 5 min. Although we evaluated them for a shorter time period, AutoSet T similarly raised its pressure rapidly to apnea, but did not respond to mild hyponea. However, AutoSet T responded to hyponea with snoring in their study, whereas it did not respond to snoring in our study, possibly due to difference in upstream resistance, thereby resulting in reduction in snoring pressure swing. PV10i, Goodnight 420E, AutoSet Spirit TM and REMstar Auto were evaluated by Rigau et al. (2006), where the pressure increase in auto-cpap device were observed for 20 min. PV10i and Goodnight 420E responded to each respiratory event as were observed in our study: PV10i did not respond to flow-limitation and Goodnight 420E did not respond to hypopnea. PV10i, Goodnight 420E and REMstar Auto were also evaluated by Lofaso et al. (2006), but in their study REMstar Auto consistently failed to detect flow-limitation. The reason of these responses, which differ from our observation, is not clear, but the fixed constant flow pattern generated in their study might cause the difference in the responses. PV10i and REMstar Auto exhibited unexpected responses with all their test protocols, i.e., auto-triggering occurred after 5 min of constant pressure in PV10i and slow increase and decrease in pressure recurred every 540 s in REMstar Auto. We did not observe pressure for a long time period of respiratory events, and further studies are required for the response of auto-cpap devices to prolonged flowlimitation. AutoSet T, PV10i and REMstar Auto were evaluated by Abdenbi et al. (2004). The pressure responded similarly to respiratory events. It rose in AutoSet T earlier, whereas pressure raised little by little in PV10i in response to apnea. AutoSet T did not respond to snoring as was observed in our study, but in contrast to their study, it responded to flow-limitation without snoring in our study. The addition of upstream resistance between auto-cpap device and collapsible airway may affect the response of auto-cpap devices to respiratory events, possibly because of the lessen pressure (and flow) swing due to the resistance. Previous studies showed the cumulative short-term response on the long-term trace, and showed the capping of applied pressure at a higher level. Auto-CPAP devices respond to sleep events and their response changes the airway condition. For example, flow-limitation will disappear after the short-term response of the auto-cpap device, and long-term stable flow-limitation is not realistic in the response of the auto-cpap device. We considered that the long-term response is the result of cumulative shortterm response, and we focused on short-term response in this study. Long-term man-machine feedback fluctuation, if it occurs, is an interesting theme to be investigated, and warrant further studies. 4.4. Limitation of this upper airway model The various proprietary responses designed by the manufacturers make it possible to develop highly efficient auto-cpap devices. However, bench studies including our study have disclosed that the auto-cpap devices differed in their responses to each respiratory event. Bench tests are useful to better understand principles of operation of each device, and will probably help medical stuffs to expect the performance in clinical application. However, it should be noted that bench tests could not allow medical stuffs to draw any conclusion about the applicability of the device to each patient before the clinical use. The collapsible airway model is different from practical upper airway, for example effects upon the sympathetic nervous system, and biological complicated effects with the higher respiratory control returned from apnea/hyponea events are neglected. It is very likely that the upper airway resistance and the collapsibility of the flow-limiting segment are variable in each patient during sleep, and the role of variable upstream resistance in respiratory events is an important point to be investigated. The results of bench tests could be used at the first step assessment to formulate hypothesis about the most appropriate way of use of each device in clinical application, and to create a new con-

54 M. Hirose et al. / Respiratory Physiology & Neurobiology 162 (2008) 48 54 cept in developing devices. It is possible that tailor-made algorithm will be designed for each patient with sleep apnea/hypopnea in future. 5. Conclusion The response of auto-cpap devices to respiratory events considerably differed from each other. Collapsible upper airway model, a collapsible tube with resistance between auto-cpap device and the tube, was considered to be useful for the first-step assessment of auto-cpap devices. Acknowledgment The authors thank the manufacturers for lending their auto- CPAP devices used in this study. References Abdenbi, F., Chambille, B., Escourrou, P., 2004. Bench testing of auto-adjusting positive airway pressure devices. Eur. Respir. J. 24, 649 658. Agrawal, S., Stone, P., Mcguinness, K., Morris, L., Camilleri, A.E., 2002. Sound frequency analysis and the site of snoring in natural and induced sleep. Clin. Otolaryngol. 27, 162 166. Akashiba, T., Minemura, H., Yamamoto, H., Itoh, D., Kosaka, N., Saitoh, O., Horie, T., 1999. Effects of nasal continuous positive airway pressure on pulmonary haemodynamics and tissue oxygenation in patients with obstructive sleep apnoea. Respirology 4, 83 87. Bradley, T.D., Phillipson, E.A., 1992. Central sleep apnea. Clin. Chest Med. 13, 493 505. Brouillette, R.T., Thach, B.T., 1979. A neuromuscular mechanism maintaining extrathoracic airway patency. J. Appl. Physiol. 46, 772 779. Farré, R., Montserrat, J.M., Rigau, J., Trepat, X., Pinto, P., Navajas, D., 2002. Response of automatic continuous positive airway pressure devices to different sleep breathing patterns. Am. J. Respir. Crit. Care Med. 166, 469 473. Hoffstein, V., Viner, S., Mateika, S., Conway, J., 1992. Treatment of obstructive sleep apnea with nasal continuous positive airway pressure. Patient compliance, perception of benefits, and side effects. Am. Rev. Respir. Dis. 145, 841 845. Hsu, A.A., Lo, C., 2003. Continuous positive airway pressure therapy in sleep apnoea. Respirology 8, 447 454. Hudgel, D.W., 1996. Treatment of obstructive sleep apnea. A review. Chest 109, 1346 1358. Konermann, M., Sanner, B.M., Vyleta, M., Laschewski, F., Groetz, J., Sturm, A., Zidek, W., 1998. Use of conventional and self-adjusting nasal continuous positive airway pressure for treatment of severe obstructive sleep apnea syndrome: a comparative study. Chest 113, 714 718. Lofaso, F., Desmarais, G., Leroux, K., Zalc, V., Fodil, R., Isabey, D., Louis, B., 2006. Bench evaluation of flow limitation detection by automated continuous positive airway pressure devices. Chest 130, 343 349. Meurice, J.C., Marc, I., Series, F., 1996. Efficacy of autocpap in the treatment of obstructive sleep apnea/hypopnea syndrome. Am. J. Respir. Crit. Care Med. 153, 794 798. Polo, O., Berthon-Jones, M., Sulliran, C., Douglas, N.J., 1994. Management of obstructive sleep apnoea/hypopnoea syndrome. Lancet 344, 656 660. Remmers, J.E., degroot, W.J., Sauerland, E.K., Anch, A.M., 1978. Pathogenesis of upper airway occlusion during sleep. J. Appl. Physiol. 44, 931 938. Rigau, J., Montserrat, J.M., Wohrle, H., Plattner, D., Schwaibold, M., Navajas, D., Farre, R., 2006. Bench model to simulate upper airway obstruction for analyzing automatic continuous positive airway pressure devices. Chest 130, 350 361. Sharma, S., Wali, S., Pouliot, Z., Peters, M., Neufeld, H., Kryger, M., 1996. Treatment of obstructive sleep apnea with a self-titrating continuous positive airway pressure (CPAP) system. Sleep 19, 497 501.