Determinants of Rib Motion in Flail Chest

Determinants of Rib Motion in Flail Chest MATTEO CAPPELLO, ALEXANDRE LEGRAND, and ANDRÉ DE TROYER Laboratory of Cardiorespiratory Physiology, Brussels School of Medicine, and Departments of Chest Medicine and Surgery, Erasme University Hospital, Brussels, Belgium We have previously developed a canine model of isolated flail chest to assess the effects of this condition on the mechanics of breathing, and these studies have led to the conclusion that the respiratory displacement of the fractured ribs is primarily determined by the fall in pleural pressure ( Ppl) and the action of the parasternal intercostal muscles. The present studies were designed to test the validity of this conclusion. A flail was induced in six supine anesthetized animals by fracturing both dorsally and ventrally the second to fifth ribs on the right side of the chest, after which the phrenic nerve roots were bilaterally sectioned in the neck. Sectioning the phrenic nerves caused a 34% decrease in Ppl, associated with a 39% increase in parasternal intercostal inspiratory EMG activity (p 0.05), and resulted in a marked reduction in the inspiratory inward displacement of the ribs. In three animals, the inward rib displacement was even reversed into a small outward displacement. When the airway was then occluded at end-expiration to increase Ppl during the subsequent inspiration, all animals again showed a clear-cut inward rib displacement. These observations therefore confirm that in dogs with flail chest, the inspiratory displacement of the fractured ribs is set by the balance between the force related to pleural pressure and that generated by the parasternal intercostals. These observations also point to the critical importance of the pattern of inspiratory muscle activation in determining the magnitude of rib cage paradox in such patients. Cappello M, Legrand A, De Troyer A. Determinants of rib motion in flail chest. AM J RESPIR CRIT CARE MED 1999;159:886 891. Flail chest is a common occurrence in blunt chest trauma and has serious adverse effects on the mechanics of breathing (1). However, blunt chest trauma usually involves additional events, such as pneumothorax and pulmonary contusion, which also adversely affect the respiratory system. In addition, blunt chest trauma produces severe pain (1). The confounding influence of these events is so prominent that measurements of thoracoabdominal motion in patients with flail chest have yielded variable results (2). To assess the isolated effects of flail chest on the mechanics of breathing, we therefore developed a canine model (3). As anticipated, the segment of the rib cage thus disconnected from the rest of the chest wall was observed to move inward during inspiration and outward during expiration. However, the fractured ribs continued to move cranially during inspiration, thus indicating that pleural pressure was not the only determining factor. Subsequent electromyographic studies revealed that flail elicits an increased inspiratory activity in the external intercostal muscles connecting the fractured ribs, probably through an increased activation of the muscle spindles (4). Since the inspiratory activity recorded from the diaphragm and the parasternal intercostals remained unaltered, it was therefore speculated that the increased external intercostal activity was an important determinant of the respiratory displacement of the ribs (4). Selective muscle denervation, (Received in original form July 16, 1998 and in revised form October 26, 1998) Correspondence and requests for reprints should be addressed to Matteo Cappello, M.D., Chest Service, Erasme University Hospital, Route de Lennik, 808, 1070 Brussels, Belgium. Am J Respir Crit Care Med Vol 159. pp 886 891, 1999 Internet address: www.atsjournals.org however, did not support this speculation (5). Severing the external intercostal muscles proved, in fact, to have little influence on rib displacement. In contrast, denervating the parasternal intercostals accentuated the inspiratory inward displacement of the ribs and reversed their persistent cranial displacement into a caudal displacement, and this led to the conclusion that in dogs with flail chest, regardless of the increased external intercostal activity, the respiratory displacement of the ribs is primarily determined by the balance between the force related to the fall in pleural pressure and that generated by the parasternal intercostals (5). The present study was designed to test the validity of this conclusion. We have thus created a flail chest in a group of anesthetized dogs and recorded the respiratory displacement of the fractured and nonfractured ribs. We then induced a complete paralysis of the diaphragm to elicit simultaneously an increased inspiratory activity in the parasternal intercostals and a reduction in the pleural pressure swings (6 10). After the diaphragm was paralyzed, we also occluded the airway at end-expiration for a single breath to increase the pleural pressure swings. If the respiratory displacement of the fractured ribs was indeed determined by the balance between the force related to pleural pressure and that generated by the parasternal intercostals, then the paradoxical inward rib displacement should be reduced after diaphragmatic paralysis and augmented again during occluded breaths. METHODS Six adult mongrel dogs weighing 21 to 31 kg and anesthetized with pentobarbital sodium (initial dose 25 mg/kg, intravenously) were studied. The animals were placed in the supine posture and intubated with a cuffed endotracheal tube, and a venous catheter was inserted in

Cappello, Legrand, and De Troyer: Determinants of Rib Motion in Flail Chest 887 the forelimb to give maintenance doses of anesthetic. The rib cage and intercostal muscles were then exposed on the right side of the chest from the first to ninth ribs by deflection of the skin and underlying muscle layers, and the second to fifth ribs were prepared to be sectioned, as previously described (3). On each of these four ribs, 2-cmlong incisions were thus made dorsally, 1 cm ventral to the rib angle, and ventrally, 1 to 2 cm lateral to the chondrocostal junction, through the periosteum, and the periosteum of each exposure was then slit and peeled on the external and internal aspects of the rib, so that the ribs could be sectioned easily later. When this procedure was completed, the C5, C6, and C7 phrenic nerve roots were isolated bilaterally in the neck, and loose ligatures were placed around them. Measurements All measurements were obtained while the animal was breathing spontaneously. Airflow was measured at the endotracheal tube with a heated Fleisch pneumotachograph connected to a differential pressure transducer (Validyne Corp., Northridge, CA), tidal volume was obtained by electronic integration of the flow signal, and the changes in pleural pressure (Ppl) were measured with a balloon catheter system placed in the midesophagus and filled with 0.5 ml of air. The respiratory displacements of the fourth and sixth ribs along the laterolateral axis of the rib cage were measured by using displacement transducers, as previously described (7). Thus, a hook was screwed into each rib in the midaxillary line and connected to a linear displacement transducer (Schaevitz Engineering, Pennsauken, NJ) through a long inextensible thread. These threads were oriented perpendicular to the sagittal plane of the animal s body and laid over pulleys placed at the side of the animal. The inward and outward displacements of the ribs during breathing were therefore leading to up and down motions of the core of the transducers. In each animal, we also recorded the electromyograms of the parasternal intercostal and external intercostal muscles in the third interspace with pairs of silver hook electrodes spaced 3 4 mm apart. Each pair of electrodes was placed in parallel fibers. The parasternal intercostal electrodes were inserted in the sternal, most active portion of the muscle (11), and the external intercostal electrodes were placed midway between the dorsal and ventral rib exposures. The two electromyographic (EMG) signals thus obtained were processed with amplifiers (model 830/1; CWE Inc., Ardmore, PA), band-pass filtered below 100 and above 2,000 Hz, and rectified prior to their passage through leaky integrators with a time constant of 0.2 s. Protocol The animal was allowed to recover for 30 min after instrumentation, after which baseline measurements of airflow, tidal volume, Ppl, rib motion, and intercostal EMG activity were made. Ribs 2 to 5 were then sectioned both dorsally and ventrally, and 1-cm-long segments of bone were removed. When this segment of the rib cage was completely disconnected from the rest of the bony thorax, measurements were repeated. The C5 C7 phrenic nerve roots on both sides of the neck were subsequently infiltrated with 2% lignocaine and sectioned, and after a 15-min recovery period, a third set of measurements was obtained. In each condition, three periods of resting breathing were recorded over 30 min. Lastly, a Hans-Rudolph valve was attached to the pneumotachograph and the tracheal tube, and the inspiratory line of the valve was blocked during the expiratory pause for a single inspiratory effort. Three occluded breaths separated by 10 to 15 unimpeded breaths were obtained in each animal. Although the animals were breathing spontaneously throughout the studies, they were maintained under light surgical anesthesia. Supplementary doses of pentobarbital sodium (2 3 mg/kg) were given at intervals to ensure that there was no flexor withdrawal of the forelimbs, no pupillary light reflex, and no spontaneous movements of the fore- or hindlimbs, including during section of the ribs and the phrenic nerve roots. In addition, rectal temperature was maintained constant at 37 1 C with infrared lamps. At the end of the experiment, the animal was given an overdose of anesthetic (30 40 mg/kg intravenously). Data Analysis The inspiratory displacements of the fourth and sixth ribs were assessed relative to the rib relaxation position, as determined by hyperventilation-induced apnea, and were measured at the peak pleural pressure. Consequently, the passive outward rib motion due to the relaxation of the triangularis sterni and internal intercostal muscles at the end of expiration (12) was discarded from the data analysis. The analysis also ignored the transient, abrupt outward rib motion that is occasionally seen after the cessation of inspiration and is related to the elastic recoil properties of the rib cage (13). The rib displacements thus measured were, therefore, exclusively related to the fall in Ppl and the action of the inspiratory muscles. These displacements were averaged over 10 consecutive breaths from each run in each condition of the study and over the three occluded breaths obtained after phrenic nerve section. In addition, the inspiratory displacements of the fourth (fractured) rib during unimpeded breathing were also plotted against the displacements of the sixth (control) rib. By convention, positive displacements indicate outward displacements, and negative displacements indicate inward displacements. Phasic inspiratory EMG activity in the parasternal intercostal and external intercostal muscles in the same breaths was quantified by measuring the peak height of the integrated EMG signal in arbitrary units. To allow comparison between the different animals, however, activity in each muscle was subsequently expressed as a percentage of the activity recorded in the control condition. Data of peak inspiratory rib displacement and peak EMG activity were finally averaged for the animal group, and they are presented as means SE. Statistical assessments of the effects of flail, phrenic nerve section, and airway occlusion were made by using Wilcoxon s signed-rank tests; the criterion for statistical significance was taken as p 0.05. RESULTS The effects of flail (ribs 2 to 5) on the respiratory displacements of the ribs are illustrated by the records of a representative animal in Figure 1, and Figure 2 shows plots of the inspiratory displacements of rib 4 against the inspiratory displacements of rib 6 for each individual animal. During control, all ribs moved outward during inspiration, such that the peak inspiratory displacements of ribs 4 and 6 in the animal group averaged 2.49 0.75 and 2.61 0.44 mm, respectively. In contrast, after flail, whereas rib 6 continued to move outward, the fractured rib 4 consistently moved inward. The peak inspiratory displacements of the two ribs thus amounted to 3.24 0.56 and 6.52 1.80 mm (p 0.05), respectively, and this alteration in rib displacement was associated with a consistent increase in external intercostal inspiratory EMG activity to 477 176% of the control value (p 0.05). The inspiratory EMG activity recorded from the parasternal intercostals, however, remained unaltered at 101 9% of the control value (NS). Tidal volume (control: 405 55 ml; after flail: 418 68 ml; NS) and Ppl (control: 7.2 0.8 cm H 2 O; after flail: 7.8 0.8 cm H 2 O; NS) remained also unchanged. These alterations were similar in all respects to those described in our previous communications (3 5). Sectioning the phrenic nerve roots in the neck resulted in a decrease in tidal volume to 253 42 ml (p 0.05), a decrease in Ppl to 5.2 0.7 cm H 2 O (p 0.05), and an increase in parasternal intercostal inspiratory EMG activity to 139 17% of control (p 0.05). In addition, although external intercostal inspiratory activity was unchanged at 485 176% of the control value (NS), the respiratory displacements of the ribs were markedly altered (Figures 1 and 2). Specifically, the outward displacement of rib 6 at peak inspiration increased in all animals from 3.24 0.56 to 9.45 1.30 mm (p 0.05), and the inspiratory inward displacement of rib 4 was substantially reduced. In three of the six animals (dogs no. 2, 3, and 5 in Figure 2), in fact, this inward displacement was reversed into an outward displacement. For the animal group, therefore, the displacement of rib 4 at peak inspiration was only 0.36 1.69 mm (p 0.05).

888 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 159 1999 Figure 1. Respiratory displacements of the fourth and sixth ribs along the latero-lateral axis of the rib cage (a) in the control condition, (b) after flail, and (c) after phrenic nerve section in a representative animal. Upward deflections indicate outward displacements, and downward deflections indicate inward displacements. Traces of lung volume and pleural pressure (Ppl) are also shown. I marks the duration of inspiration. Note that with flail, the normal inspiratory outward displacement of the fourth rib was reversed into an inspiratory inward displacement. This paradoxical displacement, however, was virtually abolished after phrenic nerve section. When the airway was occluded at end-expiration for a single breath, peak parasternal intercostal activity remained unchanged, relative to unimpeded breathing, at 122 9% of control (NS), and peak external intercostal activity increased slightly to 570 127% of control (p 0.05). More importantly, there was a large increase in Ppl to 9.5 1.3 cm H 2 O (p 0.05) and a clear-cut inspiratory inward displacement of rib 4, as shown in Figure 3. This paradoxical displacement was seen in all animals, including in the three animals that had an outward displacement during unoccluded breaths, and averaged 5.49 1.58 mm for the animal group (p 0.05). Airway occlusion also resulted in a moderate reduction in the inspiratory outward displacement of rib 6 to 7.14 0.85 mm (p 0.05). DISCUSSION Paralysis of the diaphragm in anesthetized and unanesthetized animals is well known to produce a decrease in tidal volume and an increased EMG activity in the inspiratory intercostal Figure 2. Individual plots of the respiratory displacements of the fourth rib (ordinate) versus the inspiratory displacements of the sixth rib (abscissa). The thin solid line in each panel corresponds to the control condition, the dashed line corresponds to flail chest, and the heavy solid line corresponds to phrenic nerve section. Positive rib displacement indicates outward displacement; negative rib displacement indicates inward displacement.

Cappello, Legrand, and De Troyer: Determinants of Rib Motion in Flail Chest 889 Figure 3. Effects of end-expiratory airway occlusion on the lateral displacements of the fourth and sixth ribs after phrenicotomy in a representative animal with flail chest. Same animal as in Figure 1. Upward deflections indicate outward displacements and downward deflections indicate inward displacements. Note that with airway occlusion (arrows), Ppl was greater and the inspiratory inward displacement of the fourth rib was markedly increased. muscles, in particular the parasternal intercostals (6 10). Sectioning the phrenic nerves had similar effects in the animals of this study, causing a 35 40% reduction in tidal volume and Ppl associated with a 39% increase in parasternal intercostal inspiratory EMG activity. More importantly, there was a marked reduction in the inspiratory inward displacement of the fractured ribs. The magnitude of this reduction was such that the rib inward displacement was reversed into an outward displacement in three animals (Figure 2) and essentially abolished for the animal group. Yet phrenic nerve section did not elicit any increased external intercostal inspiratory activity, which might have contributed to the disappearance of the paradoxical rib motion. In contrast, when the airway was occluded at end-expiration such that Ppl during the subsequent breath was augmented, all animals again showed a clear-cut inspiratory inward displacement of the fractured ribs in spite of the increased external intercostal activity. These observations, therefore, fully support our previous conclusion (5) that in dogs with flail chest, the inspiratory displacement of the ribs is primarily determined by the balance between pleural pressure and the force generated by the parasternal intercostals. If the force developed by these muscles during breathing is assumed to be proportional to EMG activity and is given an arbitrary value of 1.0 in the control condition, then a more quantitative analysis of the relationship between lateral rib displacement, pleural pressure, and parasternal intercostal force after flail can be made, as shown in Figure 4. The values (mean SE) of rib lateral displacement per unit parasternal intercostal force (expressed as mm) measured before phrenic nerve section, after phrenic nerve section, and during occluded breaths are plotted on the ordinate, and the corresponding values of Ppl per unit parasternal intercostal force (expressed as cm H 2 O) are plotted on the abscissa. The rib lateral displacement per unit parasternal intercostal force and the Ppl per unit parasternal intercostal force were closely related to each other; the coefficient of correlation (r) thus calculated was 0.967. If lateral rib displacement is denoted Yr and parasternal intercostal force is denoted Ps, one can therefore write the following equation: Yr Ps = A+ B ( Ppl Ps), (1) Figure 4. Relationship between rib lateral displacement, pleural pressure (Ppl), and parasternal intercostal force in flail chest. Mean SE data obtained in six animals during resting breathing after flail, during unimpeded breathing after phrenic nerve section (diaphragm paralysis) and during occluded breaths after phrenic nerve section (occlusion). The rib lateral displacement per unit parasternal intercostal force is shown on the ordinate, and the change in Ppl per unit parasternal intercostal force is shown on the abscissa. Negative sign on the ordinate indicates inward rib displacement. Parasternal intercostal force was given an arbitrary value of 1.0 in the control condition (before flail).

890 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 159 1999 and rearranging this equation yields Yr = A Ps + B Ppl. (2) The coefficients A and B thus describe the relative effects of parasternal intercostal force and pleural pressure on lateral rib displacement. Since the linear relationship yields A 4.5 mm and B 1.3 mm/cm H 2 O, it may therefore be concluded that in these animals, the outward rib displacement produced by the parasternal intercostals after flail is about three to four times greater than the inward rib displacement produced by a 1 cm H 2 O fall in pleural pressure. In other words, since Ppl after flail was 7.8 cm H 2 O, the force developed by the parasternal intercostals on the fractured ribs counterbalanced about half the force due to Ppl. This analysis also predicts that if phrenic nerve section had induced a 39% increase in parasternal intercostal EMG activity without any reduction in Ppl, the ribs would have continued to move 3.9 mm inward at peak inspiration. On the other hand, had phrenic nerve section induced a 35% reduction in Ppl without any concomitant increase in parasternal intercostal EMG activity, the persistent inspiratory inward displacement of the ribs would have been only 2.2 mm. The increased parasternal intercostal activity, combined with the reduction in Ppl, did not only abolish the inspiratory inward displacement of the fractured ribs after diaphragmatic paralysis, but it also augmented the outward displacement of the nonfractured ribs. The increase in outward displacement of these ribs (6.2 mm) was, in fact, identical to the change in displacement of the fractured ribs (6.2 mm). As a result, the net difference between the lateral displacements of the fractured versus nonfractured ribs was unaltered, and this emphasizes the importance of the perspective in qualifying the overall effect of phrenic nerve section on flail chest. In other words, the paradoxical displacement of the fractured ribs could equally be qualified as being abolished or maintained, depending on whether the perspective is from the head (an immobile structure in the present context) or from the intact ribs. As we have previously pointed out, the animals of this study did not show any increased external intercostal EMG activity after phrenic nerve section. This absence of external intercostal muscle response made interpretation of the mechanical events easier (see above), yet it is atypical. Indeed, when paralysis of the diaphragm is induced in dogs with an intact rib cage, the reduction in tidal volume elicits an increase in Pa CO2, which leads to an increased inspiratory activity in both the external intercostals and the parasternal intercostals (8, 10). In addition, recent studies have established that stimulation of phrenic nerve afferents induces a reflex inhibition of activity in the canine inspiratory intercostal muscles, in particular the external intercostals (14). Thus, sectioning the phrenic nerves causes an increased chemical respiratory drive and concomitantly removes an inhibition from the contracting diaphragm, such that external intercostal inspiratory activity increases more than parasternal intercostal activity (8, 10). The consistent increase in parasternal intercostal activity seen after phrenic nerve section in this study suggests that these mechanisms also operated in our animals. Therefore, external intercostal activity should have increased as well. The reason for which this activity remained unaltered is uncertain, but it might be related to the effect of phrenic section on the disconnected segment of the rib cage. We have previously shown that in flail chest, the inspiratory inward displacement of the fractured ribs causes the normal inspiratory short- ening of the external intercostal muscles to be reversed into an inspiratory muscle lengthening (4). Consequently, the numerous muscle spindles contained in these muscles are triggered, so that their global efferent -motor activity is increased. On the other hand, after phrenic nerve section, the inspiratory inward displacement of the fractured ribs was abolished. It is most likely, therefore, that the inspiratory lengthening of the external intercostals connecting these ribs was reduced or suppressed. As a result, the muscle spindles must have been unloaded, leading to a reduction in excitatory postsynaptic potentials in the corresponding -motoneurones. This reduction in proprioceptive inputs may have compensated for the increased efferent -motor activity to the external intercostals that diaphragmatic paralysis would have produced otherwise. Although the present observations cannot be extended to patients with flail chest without caution, they point to the critical importance of the pattern of inspiratory muscle activation in determining the magnitude of rib cage paradox in this setting. Thus, if the pain due to the rib fractures is not properly controlled, inhibition of the parasternal intercostals and other rib cage inspiratory muscles may leave pleural pressure unopposed. In this condition, a given tidal volume will be associated with more prominent rib cage paradox. The paradox might be further accentuated by pulmonary contusion, leading to a decreased pulmonary compliance and larger pleural pressure swings (15). In contrast, should the trauma involve an injury of the upper abdomen, substantial inhibition of the diaphragm might occur (16 18). Adequate relief of pain in this condition would not affect the diaphragm (18) but would cause the rib cage inspiratory muscles to generate most of the tidal volume; consequently, as in our animals after phrenic nerve section, rib cage paradox would be attenuated or suppressed. The present findings, therefore, may help in understanding the variability in thoracoabdominal motion that has been previously reported in patients with flail chest (2). References 1. Todd, T. R. J., and F. Shamji. 1985. Pathophysiology of chest wall trauma. In C. Roussos and P. T. Macklem, editors. The Thorax, part B. Marcel Dekker, New York. 979 997. 2. Tzelepis, G. E., F. D. McCool, and F. G. Hoppin, Jr. 1989. Chest wall distortion in patients with flail chest. Am. Rev. Respir. Dis. 140:31 37. 3. Cappello, M., C. Yuehua, and A. De Troyer. 1995. Rib cage distortion in a canine model of flail chest. Am. J. Respir. Crit. Care Med. 151:1481 1485. 4. Cappello, M., C. Yuehua, and A. De Troyer. 1996. Respiratory muscle response to flail chest. Am. J. Respir. Crit. Care Med. 153:1897 1901. 5. Cappello, M., and A. De Troyer. 1997. Actions of the inspiratory intercostal muscles in flail chest. Am. J. Respir. Crit. Care Med. 155:1085 1089. 6. Nochomovitz, L., M. Goldman, J. Mitra, and N. S. Cherniack. 1981. Respiratory responses in reversible diaphragm paralysis. J. Appl. Physiol. 51:1150 1156. 7. De Troyer, A., and S. Kelly. 1982. Chest wall mechanics in dogs with acute diaphragm paralysis. J. Appl. Physiol. 53:373 379. 8. De Troyer, A., and G. A. Farkas. 1989. Inspiratory function of the levator costae and external intercostal muscles in the dog. J. Appl. Physiol. 67: 2614 2621. 9. Katagiri, M., R. N. Young, R. S. Platt, T. M. Kieser, and P. A. Easton. 1994. Respiratory muscle compensation for unilateral or bilateral hemidiaphragm paralysis in awake canines. J. Appl. Physiol. 77:1972 1982. 10. Brichant, J. F., and A. De Troyer. 1997. On the intercostal muscle compensation for diaphragmatic paralysis in the dog. J. Physiol. (Lond.) 500:245 253. 11. De Troyer, A., and A. Legrand. 1995. Inhomogeneous activation of the parasternal intercostals during breathing. J. Appl. Physiol. 79:55 62. 12. De Troyer, A., and V. Ninane. 1986. Triangularis sterni: a primary muscle of breathing in the dog. J. Appl. Physiol. 60:14 21.

Cappello, Legrand, and De Troyer: Determinants of Rib Motion in Flail Chest 891 13. De Troyer, A., and G. A. Farkas. 1990. Linkage between the parasternals and external intercostals during resting breathing. J. Appl. Physiol. 69: 509 516. 14. De Troyer, A. 1998. The canine phrenic-to-intercostal reflex. J. Physiol. (Lond.) 508:919 927. 15. Oppenheimer, L., K. D. Craven, L. Forkert, and L. D. H. Wood. 1979. Pathophysiology of pulmonary contusion in dogs. J. Appl. Physiol. 47: 718 728. 16. Ford, G. T., W. A. Whitelaw, T. W. Rosenal, P. J. Cruse, and C. A. Guenter. 1983. Diaphragm function after upper abdominal surgery in humans. Am. Rev. Respir. Dis. 127:431 436. 17. Road, J. D., K. R. Burgess, W. A. Whitelaw, and G. T. Ford. 1984. Diaphragm function and respiratory response after upper abdominal surgery in dogs. J. Appl Physiol. 57:576 582. 18. Simonneau, G., A. Vivien, R. Sartene, F. Kunstlinger, K. Samii, Y. Noviant, and P. Duroux. 1983. Diaphragm dysfunction induced by upper abdominal surgery: role of postoperative pain. Am. Rev. Respir. Dis. 128:899 903.