Evidence underlying breathing retraining in people with stable chronic obstructive pulmonary disease.

Despite many studies on the topic, the role of breathing retraining techniques such as pursed-lip breathing (PLB) (1) and diaphragmatic breathing (DB) (2) in the rehabilitation of people with chronic obstructive pulmonary disease (COPD) remains unclear. This is because many of the studies on the use of breathing retraining are poorly conducted, lack control groups, use small heterogeneous samples, and frequently use poorly described methods. Furthermore, researchers often examine a combination of retraining techniques, which makes it impossible to attribute outcomes to any single technique. The results from some of the small and better designed studies are clear; however, combining these results to obtain a single unified statement regarding the value of a particular breathing exercise is unrealistic because of disease heterogeneity in the subjects and differences in study designs.

One solution to this problem would be to conduct well-designed, multicenter, randomized controlled trials using state-of-the-art technology. These trials, however, are unlikely to occur because they are expensive and difficult to conduct. Another approach is to examine the literature to determine the mechanisms responsible for the therapeutic effects of each of these techniques. With this understanding, we believe that therapists can select interventions based on a patient's clinical findings and thus, in theory, improve the effectiveness of the intervention provided. The purposes of this review, therefore, are to examine the separate effects of PLB and DB and to identify the mechanisms responsible for improvements in people with COPD.

Methods

We searched PubMed (1966 to January 2003) and the Cumulative Index to Nursing and Allied Health Literature (CINAHL, 1982 to January 2003) for articles published in English. The following terms were used in the search: "breathing retraining," "breathing exercises," "pursed-lip(s) breathing," "diaphragmatic breathing," "chronic obstructive pulmonary disease" and its acronym "COPD," "chronic obstructive lung disease" and its acronym "COLD," "chronic bronchitis," and "emphysema." To obtain additional articles, we also reviewed the reference lists from relevant articles.

We restricted our review to studies that examined a single retraining technique so that the effect of the technique could be clearly assessed. Because our focus was on patients with COPD, reports that included subjects diagnosed with asthma, cystic fibrosis, primary bronchiectasis, or pulmonary fibrosis were not included in the review. We excluded investigations when the methods were not clearly defined. Studies also were excluded if inappropriate statistical analysis made it impossible to evaluate the effect of the intervention. We included studies where randomized controlled methods were not used, because very few of the studies would meet this strict requirement. We did not attempt to combine study results as in a meta-analysis. We retrieved 397 publications and excluded 375 from this review. Table 1 lists the various reasons these articles were excluded.

In this review, the severity of COPD is indicated by providing adjectives used by the authors and by the spirometric data they reported. Currently, COPD severity is commonly classified according to the standards published in the Global Initiative for Chronic Obstructive Lung Disease (GOLD). (3) In these standards, disease severity is based on airflow limitation as measured by spirometry, and spirometric values are expressed as "percent predicted" using appropriate normal values for the person's sex, age, and height. Table 2 lists the GOLD disease severity classifications.

An earlier classification system developed by Pennock et al (4) also identifies normal limits for lung volumes. They based their classification system on the Gaussian distribution of pulmonary function test results in subjects without pulmonary disease and used the coefficient of variation to determine normal ranges. According to this classification system, COPD is present when: residual volume is > 140% predicted, functional residual capacity is >130% predicted, total lung capacity is >120% predicted, or maximum voluntary ventilation is <75% predicted. The authors did not attempt to distinguish disease severity.

Pursed-Lip Breathing

Faling described PLB as "usually the easiest breathing technique to learn and it is often (but not uniformly) employed instinctively by those who benefit from its use. Patients inhale through the nose over several seconds with the mouth closed and then exhale slowly over 4 to 6 seconds through pursed lips held in a whistling or kissing position. This is done with or without contraction of the abdominal muscles." (1) (p600) Faling recommended that PLB should be used during or following any activity that makes the patient tachypneic or dyspneic. He contended that relief of dyspnea is almost immediate after starting to use the technique.

Clinical Outcomes of PLB

Several investigators have examined the effect of PLB on ventilatory parameters and arterial blood gases in people with COPD. They uniformly reported that the technique decreases respiratory rate, (5-10) minute ventilation (VE), (8) and partial pressure of carbon dioxide in arterial blood (PaC[O.sub.2]) (7-10) and increases tidal volume (VT). (7-10) Pursed-lip breathing also has been documented to increase partial pressure of oxygen in arterial blood (Pa[O.sub.2]) (7,8) and the percentage of hemoglobin sites that are bound to oxygen in arterial blood (Sa[O.sub.2]). (7,8,11) Changes in oxygen consumption (V[O.sub.2]) are less consistent. (6-8) Pursed-lip breathing has been reported to decrease dyspnea (8,11) and, therefore, may improve exercise tolerance (12,13) and reduce limitations in activities of daily living. (14)

Changes in Ventilation and Lung Volume

In 1963, Motley (7) published data that suggested that the decreased respiratory rate associated with PLB accounted for the technique's beneficial effects. He compared spontaneous breathing with slow, deep breathing in 35 people with severe COPD (residual volume >200% predicted, vital capacity=72% predicted, maximum breathing capacity=39% predicted) and contrasted the results with those from a control group of 20 people. Motley (7) reported that when the subjects slowed their respiratory rate (average rate fell from 15 to 9), VT increased (average VT increased from 494 mL to 814 mL) and VE was unchanged, Sa[O.sub.2] was increased (from 89.5% to 92.1%), and PaC[O.sub.2] was decreased (from 40 mm Hg to 37 mm Hg).

Three years after Motley's report (7) appeared, Thoman and colleagues (10) reported on comparisons of spontaneous breathing, PLB, and slow, deep breathing. This was done in an attempt to clarify whether the effects of PLB were due to slowing the respiratory rate. They studied a group of 21 people with COPD who were comparable to those in Morley's study (residual volume=270% predicted, vital capacity=71% predicted, maximum voluntary ventilation=32% predicted). The investigators reported that PLB slowed breathing frequency (from 19 to 12 breaths per minute) and that both slow, deep breathing and PLB resulted in a similar increase in VT (from a baseline of 0.61 L to 0.82 L and 0.84 L, for slow, deep breathing and PLB, respectively). These changes resulted in a decrease in PaC[O.sub.2] (from a baseline of 55 mm Hg to 52 mm Hg and 51 mm Hg, for slow, deep breathing and PLB, respectively) that was not associated with a change in slow space volume (volume of the poorly ventilated airspaces) or functional residual capacity. The authors concluded that the effects of PLB could be attributed to slowing the respiratory rate. They also proposed that the lack of change in functional residual capacity in this group of individuals with hyperinflation (excessive air trapping resulting in an end expiratory lung volume [EELV] above normal, predicted functional residual capacity) suggested that neither technique was effective in improving the position of the diaphragm. Taken together, these findings suggest that the positive effects of PLB most likely are due to slowing the respiratory rate and that pursing the lips may only be a means to achieving the slower rate.

Tiep et al (15) and Campbell and Friend (16) also investigated the effects of PLB in people with COPD and reported results similar to those in the studies discussed. Unfortunately, the study by Tiep et al (15) must be viewed with caution. The authors used a custom-designed pneumovest to measure breathing pattern variables, but they did not include any information on the validity or reliability of data obtained using this equipment. The other significant methodological concern is related to subject selection bias. Subjects were included in the experimental group only if, after training, they could increase their Sa[O.sub.2] using PLB. The control group did not receive similar training, so the 2 groups were not comparable. Campbell and Friend (16) are often quoted in reviews of PLB; however, they investigated a combination of exercises that included relaxation exercises, slowing the respiratory rate to prolong expiration, and a form of DB. The effects of PLB alone, therefore, cannot be determined.

Dyspnea Relief

Mueller et al (8) examined the mechanisms underlying the dyspnea relief commonly associated with PLB. Twelve subjects with COPD were divided into 2 groups based on whether PLB relieved their symptoms. Mueller et al stated that all subjects had been adequately trained and were capable of performing PLB correctly. The 7 responders (forced expiratory volume in 1 second/ forced vital capacity [FE[V.sub.1]/FVC] =48% predicted, maximum voluntary ventilation=42% predicted) were classified as people who stated definitively, on more than one occasion, that PLB relieved their feelings of breathlessness. The 5 nonresponders (FE[V.sub.1]/FVC=41% predicted, maximum voluntary ventilation=36% predicted) reported no improvement in their symptoms. Ventilatory, blood gas, and gas exchange variables were measured during spontaneous breathing and PLB. Arterial blood gas measures improved similarly (PaC[O.sub.2] decreased by 4 mm Hg, Pa[O.sub.2] increased by 4 mm Hg, and Sa[O.sub.2] increased by 2%-3%) in both groups during PLB. Oxygen uptake was not different between the groups during spontaneous breathing or PLB. In contrast, although PLB reduced respiratory rate in both groups (from 16 to 9 breaths per minute in responders and from 16 to 12 breaths per minute in nonresponders), only the responders had increased VT (from 0.75 L to 1.19 L). These results suggest that the relief of dyspnea associated with PLB probably is due to changes in the mechanical function of the respiratory system and not to alterations in gas exchange or the metabolic work of breathing.

Only one other study has attempted to determine the cause of dyspnea relief attributed to PLB. Ingram and Schilder (11) used expiratory resistive loading (ERL) to simulate PLB in subjects with COPD. Subjects wore noseclips and breathed through a mouthpiece, which provided a known resistance that approximated the resistance produced during PLB. Changes in mechanical properties of the lung were measured in response to short episodes of ERL. Fifteen subjects with moderate COPD were divided into responders (n=8, FE[V.sub.1] = 45% predicted, maximum voluntary ventilation = 29% predicted) and nonresponders (n=7, FE[V.sub.1] = 45% predicted, maximum voluntary ventilation = 43% predicted) based on symptom relief while using PLB. Four people without known pulmonary disease served as control subjects. Pulmonary function testing indicated that 5 of the 8 subjects in the responder group demonstrated tracheobronchial collapse during expiration. None of the other subjects exhibited this phenomenon. Expiratory resistive loading resulted in an increase in expiratory time (data not reported in the original article), which was associated, in both groups, with decreases in mean expiratory flow (from 0.49 L/s to 0.45 L/s and from 0.54 L/s to 0.38 L/s, in responders and nonresponders, respectively) and peak expiratory flow (from 1 L/s to 0.6 L/s and from 0.9 L/s to 0.5 L/s, in responders and nonresponders, respectively).

Ingram and Schilder (11) identified subject characteristics that may have accounted for the symptom relief associated with ERL. They calculated an index of large airway collapsibility in each subject, which was used to identify the degree of airway narrowing that occurred when breathing without ERL. Responders had a higher index of collapsibility than nonresponders and control subjects. The authors also calculated the effectiveness of the external resistor in decreasing pulmonary resistance and reported that the decrease was greater in the responders than the nonresponders (51% versus 24%). When the effectiveness was plotted against the collapsibility index, it was found that ERL was most effective in decreasing lung resistance in responders who demonstrated airway collapse during expiration during unobstructed breathing. The authors hypothesized that, according to the Bernoulli effect, the decrease in expiratory flow with ERL would decrease the pressure drop across the airway wall, thus reducing airway collapse. Simply decreasing the respiratory rate also could decrease expiratory flow and the resistive pressure drop across the airway, supporting the assertions by Morley (7) and Thoman et al (10) that the effects of PLB are primarily due to slowing the respiratory rate.

Ingram and Schilder (11) also noted that ERL increased EELV in all groups, suggesting that the dyspnea relief (which occurred in responders only) was not due to improved mechanical efficiency of the diaphragm. End expiratory lung volume increased least in the responder group (125 mL versus 414 mL in the nonresponders), and the authors argued that this finding indicated that increases in airway diameter were not responsible for the drop in lung resistance and dyspnea relief found in the responders observed with ERL.

The findings of a study by O'Donnell and Webb (17) may explain the relationship of increases in EELV to symptom relief during PLB. Using regression analysis, the authors reported that dynamic hyperinflation (hyperinflation that occurs during breathing) accounted for a large proportion (38%) of dyspnea in a group of people with COPD during cycle ergometry. Theoretically, hyperinflation decreases lung compliance, which would increase the work of breathing and could reduce VT. These changes, in combination with impaired respiratory muscle function at increased EELV, could limit the body's ability to respond to the increased ventilatory demand during cycling. Thus, the increase in EELV, which occurred during ERL in the study by Ingram and Schilder, (11) may explain why the nonresponder group did not naturally use PLB.

Work of Breathing

Two other important points arise from the work of Ingrain and Schilder. (11) First, ERL increased the overall mechanical work during a breath because resistances in series (ERL+airway resistance) are additive. Expiratory resistive loading decreased lung resistance but not the total system resistance; airway resistance, therefore, was still greater than in the unobstructed state. The authors, therefore, felt that it was unlikely that dyspnea relief during PLB was related to a decrease in the work of breathing. Second, the changes in dyspnea occurred over a short period of time (6 breaths) and coincided with mechanical changes in respiratory system function. Given this time frame, the authors suggested that changes in gas exchange variables could not account for the almost immediate dyspnea relief noted by people who use PLB. (11)

O'Donnell et al (18) also investigated the effects of ERL on ventilatory mechanics in a group of people with COPD. Their findings are similar to those of other studies cited in this review. Expiratory resistive loading was associated with an increase in EELV (100 mL) and decreases in VE (from 13.1 L/min to 12.6 L/min), peak expiratory flow (from 0.56 L/s to 0.45 L/s), and mean expiratory flow (from 0.34 L/s to 0.32 L/s). The authors attributed these changes to a reduction in airway compression in the subjects who had expiratory flow limitation when breathing at rest. O'Donnell and colleagues proposed that people with COPD have "a very fine control of expiratory flow whereby intrathoracic pressure is continuously adjusted to a level that is just enough to attain maximal flow." (18) (p105) Furthermore, they proposed that this active control develops with the disease process, suggesting that imposing retraining techniques is not uniformly helpful in this population.

Oxygen Consumption

Changes in V[O.sub.2] and, by inference, changes in the work of breathing that occur with PLB appear to result from alterations in muscle function associated with the technique. Jones et al (6) reported a decrease in V[O.sub.2] (from 175 mL [O.sub.2]/min to 165 mL [O.sub.2]/min) during PLB, whereas other researchers (7,8) have reported that V[O.sub.2] is essentially unchanged by the technique. Only one study has investigated the pattern of respiratory muscle use during PLB in people with COPD. (5) Breslin (5) studied 13 people with COPD (FE[V.sub.1] = 37% [+ or -] 12% predicted [[bar.X] [+ or -] SD] during spontaneous breathing and PLB. She assessed rib cage and accessory muscle activity and abdominal and diaphragm muscle activity by measuring esophageal and gastric pressures, respectively. Breslin reported that, during PLB, decreased diaphragm activity during inspiration was accompanied by increased use of rib cage muscles. Both abdominal and rib cage accessory muscle activity increased during expiration. Respiratory rate decreased (from 22 to 15 breaths per minute) as did the duty cycle (inspiratory time divided by total cycle time; from 0.48 to 0.35). Unfortunately, interpretation of these results is limited because the author did not measure V[O.sub.2] or VT during PLB. In addition, Breslin's estimate of the resting diaphragm tension-time index (0.17) indicates that, as a group, the subjects were above the diaphragm fatigue threshold (0.15) described by Bellemare and Grassino. (19) A breathing pattern above this threshold purportedly leads to imminent respiratory failure. (19) These data are surprising because all of the subjects lived in the community and were medically stable.

Thompson et al (20) examined inspiratory muscle recruitment during ERL in 14 people with severe COPD (FE[V.sub.1]/FVC=29% predicted) and 10 control subjects without pulmonary disease. Respiratory muscle function was assessed noninvasively and included both diaphragm and accessory muscle activity. Thompson and colleagues (20) discovered that ERL decreased total inspiratory muscle activity by 12% in the COPD group. They were careful to note, however, that changes in expiratory muscle function, as noted by Breslin, (5) could account for the lack of change in Sa[O.sub.2] and end tidal C[O.sub.2] they observed and the increase in total V[O.sub.2]. Interestingly, although Motley (7) stated that V[O.sub.2] did not change for the experimental group as a whole, data plots demonstrated that PLB increased V[O.sub.2] in at least 12 of the 35 subjects. It is difficult to assess this discrepancy because he did not present the means and standard deviations for his data. Changes in EELV and respiratory system resistance (11) could account for increases in respiratory muscle activity and V[O.sub.2].

Some of the reports we cite here have used ERL as a surrogate for PLB. The advantage of using ERL is that it allows investigators to apply a known airway resistance. During PLB, resistance could vary greatly among subjects, especially if they are unfamiliar with the technique. (21) Some investigators, (20,21) however, have suggested that ERL does not adequately mimic PLB, and, therefore, they recommend that it should not be used to examine the breathing retraining technique.

Recently, Spahija and Grassino (21) investigated these differences in a group of subjects without known pulmonary disease at rest and during cycle ergometry. They concluded that the 2 techniques resulted in different breathing patterns and EELV. Expiratory resistive loading and PLB both decreased expiratory flow (from a baseline value of 0.35 L/s to 0.29 L/s and 0.33 L/s for ERL and PLB, respectively), but ERL produced a much smaller prolongation in expiratory time (37% versus 71%) and failed to produce the changes in respiratory rate and VT that were associated with PLB. Pursed-lip breathing caused the respiratory rate to fall from 15 to 10 and VT to rise from 0.80 L to 1.3 L, whereas ERL produced a respiratory rate of 13 and a VT of 0.9 L. In contrast, the changes in breathing pattern in response to ERL in people without pulmonary symptoms that were reported by Hill et al (22) and Gothe and Cherniack (23) were qualitatively similar to the changes stimulated by PLB in the work of Spahija and Grassino. (21) Spahija and Grassino found that ERL increased EELV, whereas it remained unchanged with PLB. Interestingly, Spahija and Grassino did not consider the 3% to 4% increase in the 6-L vital capacity to be functionally meaningful. Ingram and Schilder (11) and Thompson et al (20) demonstrated increases in EELV in people without pulmonary disease in their studies (600 mL and 130 mL, respectively). No one has compared PLB and ERL in people with COPD. These findings suggest that clinicians should use caution when interpreting studies that use ERL to examine the effects of PLB and that more investigations are needed to examine the comparability of these 2 techniques in people with COPD.

Summary of the Evidence Regarding Pursed-Lip Breathing

The experimental work presented here suggests that PLB does relieve dyspnea in selected subjects; however, we are not yet able to identify those people beforehand. The evidence suggests that there is little benefit in continuing to teach the technique to patients who do not learn it rather quickly. In those people who do experience dyspnea relief while using PLB, the technique seems to optimize the mechanical function of the lungs, limiting increases in EELV and, therefore, the deleterious effects of hyperinflation. The evidence suggests that the symptom relief during PLB occurs despite an increase in the metabolic work of breathing. Patients need to be taught to use "just enough" positive airway pressure. Allowing patients to use excessive resistance may increase the work of breathing to the point where the cost-benefit ratio is no longer favorable. If the positive effects associated with PLB can be maintained during exercise, activity tolerance should improve. To date, however, no studies have systematically investigated this aspect of therapy.

Diaphragmatic Breathing

Gosselink et al described DB as "facilitating outward motion of the abdominal wall while reducing upper rib cage motion during inspiration." (24)(p1140) Accordingly, individual skill in performing DB is assessed by observation or measurement of abdominal excursion during the respiratory cycle. Patients can be taught to perform the maneuver while maintaining abdominal muscle relaxation, (25) a technique that ensures that inspiration begins from functional residual capacity. Alternatively, abdominal muscle contraction during expiration may be encouraged. (12) Theoretically, expiratory muscle activation lengthens the diaphragm and increases its force-generating capacity. (26) The use of additional muscles during active expiration, however, may increase the energy cost of breathing.

Improvement in Ventilation

Researchers have examined how DB affects overall ventilation, the regional distribution of ventilation, and gas exchange. Changes in overall ventilation have been measured using inhaled radioactive xenon in 6 subjects with COPD (residual volume: 200%-580% predicted). (27) Xenon washout times indicated that overall ventilation was unaffected in 3 subjects, yet increased by 20% in another 3 subjects. In the subjects who showed an increase in overall ventilation, DB was accomplished by large increases in VT (56%-106% of eupneic VT) and no change in VE, which indicates a substantial slowing of respiratory frequency. Using data from the article to calculate respiratory rate, we determined that the average resting breathing rate was 22 breaths per minute and the average diaphragmatic breathing rate was 12 breaths per minute. Brach et al (27) concluded that any individual changes in ventilation were probably due to the shift to slower, deeper breathing regardless of a DB pattern.

Campbell and Friend (16) postulated that the increased abdominal motion during DB may shift ventilation toward the bases of the lungs. In the study cited above, Brach et al (27) found that DB did not alter regional ventilation for the group as a whole. Two people, however, increased ventilation to the base of one lung by more than 20%. Unfortunately, the authors were not able to explain this finding. Although the study by Sackner et al (28) is often quoted, it is not included here for 2 reasons: several subjects had asthma, and respiratory inductive plethysmographic measurements demonstrated that half of the subjects had minimal abdominal displacement during DB.

Improvements in ventilation also might be reflected as decreases in carbon dioxide levels or improvements in oxygenation. Becklake et al (29) measured oxygenation using oximetry at rest and during exercise in 6 subjects with chronic emphysema (mean residual volume = 156% predicted). Each subject received at least 3 instructional sessions from a physical therapist regarding the importance of abdominal movement, expiration, and general relaxation during DB. Citing previous literature, (30) Becklake et al (29) considered a >4% change in Sa[O.sub.2] to be clinically meaningful, given the day-to-day fluctuations in the disease state. Oxygen saturation improved by >4% in only 1 subject, whereas it dropped by >4% in 2 subjects. For the group, DB had no clinically meaningful effect on Sa[O.sub.2] measurements taken at rest (a 0.7% increase) or during exercise (a 3% decrease). Although Miller (31) and Vitacca et al (32) have reported much larger changes in Pa[O.sub.2] and PaC[O.sub.2], these studies combined DB with other breathing modifications, which make it impossible to determine the effect of DB alone.

Normalization of Breathing Pattern

Sackner et al (33) reported the effects of DB on VT and respiratory frequency in 9 patients with COPD (FE[V.sub.1] = 55% [+ or -] 19% predicted [[bar.X] [+ or -] SD], FE[V.sub.1]/FVC= 52% [+ or -] 13% predicted [[bar.X] [+ or -] SD]). Respiratory inductance plethysmography was used to measure chest wall movements, and these changes were calibrated using spirometric measurements to indicate actual volume changes. Subjects were studied in the supine position with and without visual feedback about abdominal excursion. The presence of visual feedback did not affect the results Diaphragmatic breathing was associated with an average 22% increase in VT and 25% increase in inspirator time. Respiratory frequency and VE did not change because of intersubject variability. A report with similar methods by Willeput et al (34) is not included here because 3 of the 11 subjects had tuberculosis and because they had neither emphysema nor chronic bronchitis.

Another study by Sackner et al (35) showed that DB was associated with distorted chest wall motion. Seven people with COPD (FE[V.sub.1] = 52% [+ or -] 20% predicted [[bar.X] [+ or -] SD]) were studied in the supine position, and they were asked to breathe in 3 ways: normally or with increased abdominal or rib cage movement. Chest wall motion was recorded using respiratory inductance plethysmography, and signals were analyzed for paradoxical motions (abdomen or rib cage moving in opposite directions) and asynchronous motions (difference in rate of change of movement in the rib cage or abdominal compartment). Diaphragmatic breathing caused increased paradoxical and asynchronous movements of the rib cage. The amount of asynchrony was not correlated with disease severity.

Decrease in the Work of Breathing and Dyspnea

Two groups of researchers (6,24) have measured V[O.sub.2] to assess the effect of DB on the work of breathing. Jones et al (6) showed that DB decreased resting V[O.sub.2] by 5% in 30 people with moderately severe COPD (FE[V.sub.1] = 39% [+ or -] 13% predicted [[bar.X] [+ or -] SD]) who were placed in the supine position. They attributed most of this change to the 14% fall in respiratory rate, rather than to the increased movement of the abdominal wall. Subjects were included in the study if they were observed performing the technique adequately. However, in the study by Sackner et al (33) cited previously, subjects were included who were observed to be able to perform DB, but half of the subjects had minimal increases in abdominal movement when measured by respiratory inductance plethysmography.

Gosselink et al (24) report increased V[O.sub.2] and work of breathing in 7 people with severe COPD (FE[V.sub.1] = 34% [+ or -] 7% predicted [[bar.X] [+ or -] SD]) who performed DB with no spontaneous changes in respiratory frequency. Although the subjects' position was not stated, the seated posture was probably used. We assumed that this posture was used because chest wall motion was assessed by measuring the transit time of sound waves through flexible hollow tubes, and a recumbent position would have made this measurement difficult. After 3 weeks of training, only those people who were able to at least double their abdominal movement were included in the study. Measurements were taken while the patients performed either spontaneous breathing or DB, while at rest or while breathing against a threshold load. Threshold loading requires a person to generate enough inspiratory pressure to open a valve before inspiratory air flow can occur. This method of loading the inspiratory muscles may mimic the additional respiratory muscle load generated by dynamic hyperinflation that occurs in some people with COPD during exercise. The work of breathing (mechanical efficiency) was calculated according to the method of Collett et al, (36) which relates changes in mechanical work to changes in V[O.sub.2] during loaded breathing. Gosselink et al (24) found that, during resting breathing, DB increased V[O.sub.2] (by an average of 17 mL/min) but had no effect on VT, breathing frequency, duty cycle (inspiratory time divided by total respiratory cycle time), or VE. When compared with natural breathing under loaded conditions, DB during loaded breathing was associated with a lower mechanical efficiency, increased paradoxical rib cage motions, and no change in V[O.sub.2] or dyspnea. We retrieved no other studies that investigated the effect of DB alone on dyspnea in people with COPD.

Summary of the Evidence Regarding Diaphragmatic Breathing

Evidence indicates that DB does not change regional ventilation in people with COPD. This technique may increase total ventilation but, if so, the evidence suggests this may be due to the slower, deeper breathing pattern that may occur during DB rather than an exaggeration of abdominal motion. Diaphragmatic breathing also may increase asynchronous and paradoxical rib cage motion, which may account for the increase in the work of breathing noted by some authors. There is little information on the effect of DB in altering dyspnea, although one study (24) showed that there was no effect under conditions of threshold loading. Consequently, the evidence suggests that there is no benefit to training people with COPD to perform this skilled maneuver beyond the benefit that might be achieved by simply slowing the breathing rate or using PLB.

Limitations to the Literature Reviewed

There are several limitations to the literature we cited in this review. The researchers used small sample sizes, which decreases the power to detect treatment effects. Most of the researchers examined the effect of PLB or DB interventions on impairments. Only one study (24) investigated the functional limitations of dyspnea. There are no reports of investigations of the effect of either PLB or DB on disability, morbidity, or mortality. In addition, only short-term effects have been measured. Last, there appear to be no data available that define clinically meaningful changes in respiratory rate, VT, or carbon dioxide level. In one research report, (30) a 4% change in oxygen saturation was stated to be clinically meaningful. The alinear relationship between Sa[O.sub.2] and Pa[O.sub.2], (37) however, indicates that the clinical meaningfulness of a given change in Sa[O.sub.2] will depend, in part, on the initial Sa[O.sub.2] value.

Conclusion

Careful examination of the literature on DB and PLB reveals that the use of PLB appears to be an effective way to decrease dyspnea and improve gas exchange in people with moderate to severe, but stable, COPD. These positive effects appear to be related to the technique's ability to decrease airway narrowing during expiration, an effect attributed to decreasing the resistive pressure drop across the airway wall. Thus, PLB could only be expected to be beneficial to those people with narrowing of larger airways during expiration, which would exclude people with mild disease. Only a few studies demonstrated positive effects during DB. These effects appeared to be associated with slowing the respiratory rate and not improving ventilation or V[O.sub.2]. Pursed-lip breathing is often adopted naturally, and DB requires skill and extensive training. Our interpretation of the evidence is that PLB can be a valuable rehabilitation tool in selected cases and that there is no rationale for teaching DB to this patient population. Traditionally, physical therapists classify DB and PLB as breathing retraining techniques. To date, no studies were found that investigated patients' ability to use these techniques during functional activities, which may require use of the techniques over prolonged periods of time. This should be a focus of future research.

Table 1.
Reasons for Excluding Articles From Review

Reason                                                 N

Article was article review                              51
Article was descriptive article                         68
Researchers used multiple interventions                 22
Researchers used the wrong technique                    77
  (eg, airway clearance)
Researchers studied the wrong patient population        78
  (eg, asthma)
The technique used in the study was not clearly          3
  defined
The article addressed new instrumentation                6
The article addressed beliefs about treatment            2
The article was an editorial or commentary               9
The article addressed a cost-benefit analysis of the     3
  intervention
The article assessed mechanical ventilation             27
The article assessed medication                          4
The article assessed mechanics                          10
The article assessed neural regulation of breathing      6
Miscellaneous reasons                                    9
Total                                                  375

Table 2.
Disease Severity Classifications for Chronic Obstructive Pulmonary
Disease (COPD) According to Global Initiative for Chronic
Obstructive Lung Disease (GOLD) Standards (3)

Severity       Measurement (a)

Mild COPD      FE[V.sub.1]/FVC < 70% predicted
               FE[V.sub.1] [greater than or equal to] 80% predicted

Moderate COPD  FE[V.sub.1]/FVC < 70% predicted
               30% [less than or equal to] FE[V.sub.1] < 80% predicted

Severe COPD    FE[V.sub.1]/FVC < 70% predicted
               FE[V.sub.1] < 30% predicted or FE[V.sub.1] < 50%
                 predicted plus respiratory failure or clinical signs
                 of right heart failure

(a) FE[V.sub.1] = forced expiratoty volume in 1 second, FVC = forced
vital capacity. Spirometer values are expressed as "percent predicted"
using appropriate normal values for the person's sex, age, and height.

References

(1) Faling LJ. Pulmonary rehabilitation: physical modalities. Clin Chest Med. 1986;7:599-618.

(2) Cahalin LP, Braga M, Matsuo Y, Hernandez ED. Efficacy of diaphragmatic breathing in persons with chronic obstructive pulmonary disease: a review of the literature. J Cardiopulm Rehabil. 2002;22:7-21.

(3) Global Initiative for Chronic Obstructive Lung Disease. Global Strategy for the Diagnosis, Management, and Prevention of Chronic Obstructive Pulmonary Disease: NHLBI/WHO Workshop Report. Rockville, Md: National Institutes of Health, National Heart, Lung, and Blond Institute; April 2001. NIH Publication No. 2701.

(4) Pennock BE, Gottrell JJ, Rogers RM. Pulmonary function testing: what is "normal"? Arch Intern Med. 1983;143:2123-2127.

(5) Breslin EH. The pattern of respiratory muscle recruitment during pursed-lip breathing. Chest. 1992:101:75-78.

(6) Jones AY, Dean E, Chow CCS. Comparison of the oxygen cost of breathing exercises and spontaneous breathing in patients with stable chronic obstructive pulmonary disease. Phys Ther; 2003;83:424-431,

(7) Motley HL. The effects of slow deep breathing on the blood gas exchange in emphysema. Am Rev Respir Dis. 1963;88:484-492.

(8) Mueller RE, Petty TL, Filley GF. Ventilation and arterial blood gas changes induced by pursed lips breathing. J Appl Physiol. 1970;28: 784-789.

(9) Petty TL, Guthrie A. The effects of augmented breathing maneuvers on ventilation in severe chronic airway obstruction. Respir Care. 1971; 16:104-111.

(10) Thoman RL, Stoker GL, Ross JC. The efficacy of pursed-lips breathing in patients with chronic obstructive pulmonary disease. Am Rev Respir Dis. 1966;93:100-106.

(11) Ingram RM, Schilder DP. Effect of pursed lips expiration on the pulmonary pressure-flow relationship in obstructive lung disease. Am Rev Respir Dis. 1967;96:381-388.

(12) Humberstone N, Tecklin JS. Respiratory treatment. In: Irwin S, Tecklin JS, eds. Cardiopulmonary Physical Therapy. 3rd ed. St Louis, Mo: Mosby; 1995:357-360.

(13) Watchie J. Cardiopulmonary Physical Therapy: A Clinical Manual. Philadelphia, Pa: WB Saunders; 1995:208, 211-212.

(14) Levenson CR. Breathing exercises. In: Zadai CC, ed. Pulmonary Management in Physical Therapy. New York, NY: Churchill Livingstone; 1992:135-155.

(15) Tiep BL, Burns M, Kao D, et al. Pursed lips breathing training using ear oximetry. Chest. 1986;90:218-221.

(16) Campbell EJ, Friend J. Action of breathing exercises in pulmonary emphysema. Lancet. 1955;268:325-329.

(17) O'Donnell DE, Webb KA. Exertional breathlessness in patients with chronic airflow limitation: the role of lung hyperinflation. Am Rev Respir Dis. 1993;148:1351-1357.

(18) O'Donnell DE, Sanii R, Anthonisen NR, Younes M. Expiratory resistive loading in patients with severe chronic air-flow limitation: an evaluation of ventilatory mechanics and compensatory responses. Am Rev Respir Dis. 1987;136:102-107.

(19) Bellemare F, Grassino A. Effect of pressure and timing of contraction on human diaphragm fatigue. J Appl Physiol. 1982;53:1190-1195.

(20) Thompson WH, Carvalho P, Souza JP, Charan NB. Effect of expiratory resistive loading on the noninvasive tension-time index in COPD. J Appl Physiol. 2000;89:2007-2014.

(21) Spahija JA, Grassino A. Effects of pursed-lips breathing and expiratory resistive loading in healthy subjects. J Appl Physiol. 1996;80: 1772-1784.

(22) Hill AR, Kaiser DL, Lu JY, Rochester DF. Steady-state response of conscious man to small expiratory resistive loads. Respir Physiol. 1985; 61:369-381.

(23) Gothe B, Cherniack NS. Effects of expiratory loading on respiration in humans. J Appl Physiol. 1980;49:601-608.

(24) Gosselink RA, Wagenaar RC, Rijswijk H, et al. Diaphragmatic breathing reduces efficiency of breathing in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med. 1995;151: 1136-1142.

(25) Gaskell CV. Introduction to the treatment of medical chest conditions. In: Cash JE, ed. Chest, Heart, and Vascular Disorders for Physiotherapists. London, England: Faber and Faber; 1975:123-124.

(26) Criner GJ, Celli BR. Effect of unsupported arm exercise on ventilatory muscle recruitment in patients with severe chronic airflow obstruction. Am Rev Respir Dis. 1988;138:856-861.

(27) Brach BB, Chao RP, Sgroi VL, et al. [sup.133]Xenon washout patterns during diaphragmatic breathing: studies in normal subjects and patients with chronic obstructive pulmonary disease. Chest. 1977;71: 735-739.

(28) Sackner MA, Silva G, Banks JM, et al. Distribution of ventilation during diaphragmatic breathing in obstructive lung disease. Am Rev Respir Dis. 1974;109:331-337.

(29) Becklake MR, McGregor M, Goldman HI, Braudo JL. A study of the effects of physiotherapy in chronic hypertrophic emphysema using lung function tests. Dis Chest. 1954;26:180-191.

(30) Briscoe WB, Becklake MR, Rose IW. Intrapulmonary mixing of helium in normal and emphysematons subjects. Clin Sci. 1951;10: 37-51.

(31) Miller WF. A physiologic evaluation of the effects of diaphragmatic breathing training in patients with chronic pulmonary emphysema. Am J Med. 1954;17:471-477.

(32) Vitacca M, Clini E, Bianchi L, Ambrosino N. Acute effects of deep diaphragmatic breathing in COPD patients with chronic respiratory insufficiency. Eur Respir J. 1998;11:408-415.

(33) Sackner MA, Gonzalez HE, Jenouri G, Rodriguez M. Effects of abdominal and thoracic breathing on breathing pattern components in normal subjects and in patients with chronic obstructive pulmonary disease. Am Rev Respir Dis. 1984;130:584-587.

(34) Willeput R, Vachaudez JP, Lenders D, et al. Thoracoabdominal motion during chest physiotherapy in patients affected by chronic obstructive lung disease. Respiration. 1983;44:204-214.

(35) Sackner MA, Gonzalez H, Rodriguez M, et al. Assessment of asynchronous and paradoxic motion between rib cage and abdomen in normal subjects and in patients with chronic obstructive pulmonary disease. Am Rev Respir Dis. 1984;130:588-593.

(36) Collett PW, Perry, C, Engel LA. Pressure-time product, flow, and oxygen cost of resistive breathing in humans. J Appl Physiol. 1985;58: 1263-1272.

(37) West JB. Pulmonary Pathophysiology: The Essentials, 5th ed. Baltimore, Md: Lippincott Williams & Wilkins; 1998:50-56.