INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

To noninvasively measure transfer impedance (Z_tr) of the human respiratory system (1) pressure oscillations are imposed around the thorax while a subject sits in an enclosed chamber. The pressure around the chest wall, P_cw, and the flow at the mouth, _ao, are measured and Z_tr is defined as:

(1)

where P_cw() and _ao() represent the Fourier transform of the pressure and flow waveforms, respectively and  is the angular frequency in radians/s. There are two levels at which respiratory impedance has been analyzed in the past. First, distinct features of the spectra can be used to empirically indicate diseased conditions consistent with abnormalities in lung function assessed by spirometry (5). Second (1), Z_tr can be analyzed by fitting the DuBois' six element model (8) to the Z_tr data. In this model there is an airways resistance (Raw) and inductance (I_aw) and a tissue compartment having a tissue resistance (R_t), inductance (I_t), and capacitance (C_t) are separated by a gas compliance (C_g) compartment representing alveolar gas compressibility. The implication is that this noninvasive technique allows for separate estimates of tissue and airways properties. In actuality, the reliability of this approach is highly dependent on the frequency range over which Z_tr is measured (9). Indeed, recent studies have shown the separate estimates of airway and tissue properties in the DuBois model are reliable only if data is acquired to at least 50-64 Hz (1, 10). Moreover, application of this model assumes an a priori knowledge of gas compliance (C_g) based on measurements of functional residual capacity (FRC) (2).

Lutchen and Jackson have reviewed (9) how Z_tr may provide several advantages to input impedance (Z_in) in adult humans. Most notably, from 4-64 Hz, Z_in data does not provide information specific to the airways and tissues while Z_tr data does. Moreover, Z_tr data should be significantly less sensitive to influences of upper airway shunting (11). There are, however, very few studies reporting Z_tr measurements in subjects with lung disease. Ying and colleagues made Z_tr measurements on nine patients with COPD (3) and Marchal and coworkers (4) in asthmatic adults and children. However the frequency range of these studies was low (< 36 Hz) calling into question the reliability of the separation between airways and tissue properties.

The goal of this study was to measure Z_tr in a patient population with respiratory disease and investigate the utility of Z_tr to provide sensitive and specific information on lung mechanical properties and function. We compared features of Z_tr spectra as well as parameters from the six-element model with standard spirometric indices. These comparisons were also made after respiratory therapy using bronchodilators.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Subjects

We studied 11 normal individuals and 26 patients with various respiratory diseases (see Table 1). The normal subjects ranged from 17-53 yr of age, and reported no history of any lung disease. One of the normals (Subject 11) was an asymptomatic smoker but showed normal spirometry. Several patients exhibited multiple conditions including some combination of asthma, COPD, chronic cough, dyspnea, lung cancer and sarcoidosis. Over the entire patient group, a total of 12 had previously diagnosed COPD and 12 were smokers. This "smoker" group consisted of one patient with COPD, six patients with dyspnea (one of which had asthma and one chronic cough), one patient with lung cancer, three other patients with chronic cough, and one normal patient.

Experimental Set-up

A leg and neck out partial body plethysmograph (PBP) was used to seal the patient from thigh to neck following the method of Lutchen and coworkers (1). Two loudspeakers (Pyle, 15'' Professional High Fidelity Woofer, P/N 1560) are mounted on each side of the PBP. The speakers were driven with a digital signal stored on a computer after being passed through a digital to analog (D/A) converter (Metrabyte) sampled at 512 Hz, amplified (Crown D150A) and low pass filtered at 160 Hz.

Pressure was measured in the PBP using a Microswitch transducer (0-5 cm H₂O) and flow was measured at the mouth using a Celesco transducer (0-2 cm H₂O) mounted across a pneumotachometer (Fleisch No. 2). Pressure and flow were band passed filtered (Ithaco) from 2 to 160 Hz to remove the breathing signal and its higher harmonics. Both signals were monitored on a dual channel oscilloscope. Pressure and flow were then sent to the A/D board, sampled at 512 Hz and stored in the computer.

Forcing functions containing frequency components from 4-128 Hz in 4 Hz increments or 1-128 Hz in 1 Hz increments were used. The energy at the low frequency end (< 32 Hz) was enhanced to improve the signal-to-noise. The phases were such that the crest factor was minimized so that the influence of nonlinear distortion was reduced. The crest factor is equal to the peak-to-peak pressure oscillation divided by the rms pressure, P_p-p/P_rms (12). Also, the use of small broadband oscillations has been shown to aid in producing a linearized response (13). Pressure and flow measurements were digitally compensated using the method of Renzi and associates (14). Impedance was calculated as described by Michaelson and coworkers (15).

Protocol

The following pulmonary function tests were conducted in each patient: spirometry to determine FVC, FEV₁, FEV₁/FVC, PEF, and FEF_25-75 and helium dilution to determine lung volume (lung volume for 3 subjects was measured using body plethysmography). Spirometry for 9 of the 11 normal subjects was not performed. Patients were then put into the plethysmograph and Z_tr was measured. Impedance measurements were made by applying 24 pseudorandom noise (PRN) bursts to the thoracic cage. Unreasonably high impedance values for individual bursts, presumably as a result of glottis closure, were discarded. All noncorrupted data was stored for subsequent analysis. In five healthy subjects we also measured plethysmographic airway resistance, Raw, for subsequent comparison to model estimates.

Twenty-three of 30 patients were given a bronchodilator (ventolin or albuterol). For those receiving bronchodilator, spirometry was repeated. In seven of these 23, lung volume measurements were also repeated. All patients receiving bronchodilator treatment had Z_tr measurements repeated.

Parameter Estimates

Parameters for the six-element model were estimated using a gradient optimization technique. This technique minimizes the squares of the difference between measured and model impedance data values. The sum of the squares of the difference is given by a performance index (P.I.)

(2)

where d and m represent data and model respectively and i is the frequency index. The C_g value had to be preassigned as based on the FRC estimate from the gas-dilution method (see Table 1). The same value of Cg was used before and after bronchodilator in the patients except in the seven subjects whose FRC measurements were made in both conditions. However, as shown below we did not find a statistically significant change in FRC after bronchodilator in these seven subjects.

Data Comparisons

Pooled comparisons of the spectral data were done at two levels. First, the normal subjects were compared with all diseased subjects pooled together. Then normals were compared with the COPD subjects and with the smoker subjects as separate groups (there were insufficient numbers of distinct asthmatics and sarcoidosis subjects). One of the smokers was included in both groups. Also, while anecdotal at best, we compared the data from a single asthmatic pre and post bronchodilator with the normal population. In these comparisons we identified key spectral features and examined statistical differences among these groups for these features. Finally, we performed correlations between various spirometric measures and spectral features as well as model parameters after fitting the six-element model to these data.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Z_tr Data: Healthy versus Disease

Figure 1 shows the average and standard deviation bounds on Z_tr spectra from the 11 normal subjects. The real part of the healthy subjects Z_tr data is approximately 5 cm H₂O/L/s at the lowest frequency of 4 Hz (Re₄), remained relatively constant out to about 16-20 Hz, and then decreased thereafter. The real part crosses the zero axis near 35 Hz (mean R_o = 35 ± 3). The imaginary part (reactance) is negative at low frequencies, increases with frequency reaching a first resonance (i.e., crossing zero) near 5 Hz (RES₁ = 5 ± 1.5). Reactance reaches a maximum of approximately 6.5 cm H₂O/L/s (Im_max = 6.5 ± 2) near 38 Hz (f Im_max = 38), and decreases with increasing frequency, again becoming negative at approximately 65 Hz (RES₂ = 65 ± 5). We point out that several of these features occurred above 32 Hz, that is at frequencies above those measured by other investigators (2) in many previous studies. Also, we found that beyond 72 Hz the data became extremely unreliable and variable and thus was not included in our analyses.

View larger version (25K): [in this window] [in a new window]   Figure 1.   Mean (inverted triangles) and ± standard deviation (solid line) bounds on normal real and imaginary parts of Ztr. Also shown are the key spectral features described in detail in the text and used to compare Ztr in healthy and sick subjects.

Figure 2 summarizes the data for COPD (Figure 2A) and smoker (2B) patients as a group relative to the bounds on healthy Z_tr data. The COPD patients as a group showed elevated real part outside the standard deviation bounds for healthy subjects for f < 56 Hz. The difference is greater at the lower frequencies. As a consequence, the COPD Re₄ and R_o are substantially and significantly elevated relative to control. The differences in the imaginary parts between COPD and healthy subjects are far less evident and not statistically significant. If there were differences, the tendency was for a lower imaginary part between approximately 8 and 32 Hz causing an elevated RES₂ and f Im_max. The smokers showed less difference between both the real and imaginary parts of Z_tr compared with the healthy Z_tr (Figure 2B).

View larger version (27K): [in this window] [in a new window]   View larger version (23K): [in this window] [in a new window]   Figure 2.   Mean (open symbols) and ± SD (solid line) Ztr bounds for COPD (A), and smoker (B) subjects. For each case we also show the ± standard deviation bounds from the normals of Figure 1 (dashed lines).

Z_tr Data: Pre versus Post Bronchodilator

In the normal subjects there were no differences between the pre and post bronchodilator Z_tr or spirometry. In contrast, the impedance spectra were found to change noticeably in many patients. In fact, of the 11 patients showing improved spirometry after bronchodilator therapy (i.e., > 10% improvement in one or more spirometric index), all showed significantly more normal Z_tr spectra as well. Figure 3 shows examples of pre- and post-bronchodilator Z_tr data relative to normal Z_tr data for two subjects who showed improved spirometry: one asthmatic (Subject 17) and one COPD patient (Subject 21), respectively. Specifically, before bronchodilator, Subject 17 showed elevated real and imaginary parts. After therapy, both the real and imaginary parts of the Z_tr spectra were shifted within the bounds determined for normals at virtually all frequencies. Consequently, Re₄, R_o, Im_max and f Im_max decreased, however, RES₁ and RES₂ did not change appreciably. Likewise, before bronchodilator, the Z_tr spectra of Subject 21 (COPD) showed increased real part at all frequencies, while the imaginary part resembled that of a normal individual. After medication, his real part also decreased and was within the healthy range at all frequencies. As a result, he also experienced decreased Re₄, R_o, and f Im_max compared with his pre-bronchodilator spectra. The imaginary part showed only a slight reduction remaining within the bounds of a healthy subject. Again, there was very little change in RES₁ and RES₂.

View larger version (26K): [in this window] [in a new window]   View larger version (24K): [in this window] [in a new window]   Figure 3.   Example of asthmatic (A) and COPD (B) subject Ztr data before (open squares) and after (inverted triangles) administering a bronchodilator. Also shown is the ± standard deviation bounds from the normals of Figure 1.

In summary, patients showed significant differences in their Z_tr spectra from that of the healthy group. While the largest difference occurred for the COPD subjects, the Z_tr of smokers were also abnormal in several respects. Moreover, all patients that had a spirometric improvement with bronchodilator also had improved Z_tr spectra. In the following sections we explicitly compare the distinct features from Figure 1 and the six- element model parameters as estimated from these data.

Z_tr Spectral Features and Model Parameters

Figures 4 and 5 compared the key Z_tr spectral features (Figure 4) and the six-element model parameters (Figure 5) of abnormal versus normal and pre versus post-bronchodilator data. We have compared these values for all patients averaged together and each of the main patient groups (COPD and smoker) separately relative to normals. Focusing first on abnormal (pre) versus normal values we see that with respect to the Z_tr spectra Re₄, R_o, f Im_max, RES₁, and RES₂ were all statistically significantly elevated in both patient groups. With respect to the model analysis, Table 2 shows that the Raw from model analysis compared well with the Raw from standard plethysmography. During disease we found that Raw from the six-element model applied to Z_tr was always elevated (Figure 5), with the strongest statistical difference occurring in the COPD group (p < 0.0005). Other parameter differences were less clear, particularly when broken down into specific disease groups. In fact, only one other parameter showed differences at the p < 0.01 level or less. This was C_t for the COPD group.

View larger version (58K): [in this window] [in a new window]   Figure 4.   Summary comparison of Ztr spectral features (defined in Figure 1) in diseased versus normal subjects and within diseased subjects pre versus post bronchodilator. Comparisons were performed using a t test and were done with all diseased subjects lumped together ("D") as well as with subjects grouped into specific disease categories of COPD ("C") or smokers ("S"). Note the p values preceded by @ are for the diseased group feature relative to the normal group and p values preceded by * are for the diseased group feature post bronchodilator relative to the same feature in that group pre-bronchodilator.

View larger version (56K): [in this window] [in a new window]   Figure 5.   Summary comparison of six-element model parameters derived from model fitting of Ztr in diseased versus normal subjects and within diseased subjects pre versus post bronchodilator. Parameters include airway resistance and inertance (Raw and Iaw), tissue resistance, inertance and compliance (Rt, It, Ct) and lung gas compressibility (Cg). Comparisons were performed using a t test and were done with all diseased subjects lumped together ("D") as well as with subjects grouped into specific disease categories of COPD ("C") or Smokers ("S"). Group feature relative to the normal group and p values preceded by * are for the diseased group feature postbronchodilator relative to the same feature in that group prebronchodilator.

After bronchodilator, R_o, Re₄, and Raw significantly decreased among diseased patients grouped together. When considered separately, all diseased groups experienced a significant decrease in R_o, whereas a significant decrease in Raw and Re₄ was only seen in COPD patients.

Correlations with Spirometry Z_tr Spectral Features

All Z_tr spectral features and six-element model parameters were correlated with five spirometric indices: FVC, FEV₁, FEV₁/FVC, FEF_25-75, and PEF. Likewise, changes in model parameters and spectral features were correlated with changes in these spirometric indices (Tables 3 and 4). With respect to absolute values, the spectral feature most correlated with abnormal spirometry was R_o. The R_o was well correlated with % predicted FVC (r = 0.71) and % predicted FEV₁ (Figure 6A, B). Also, the regression between R_o and FEV₁ among all patients yielded an r value of 0.47, but if patient NS (Patient 12, smoker), which is an outlier, is removed, the correlation improved significantly (r = 0.62, Figure 6B). Also, Re₄ was correlated with % predicted FVC (r = 0.58) (Figure 6C). With respect to six-element model parameter estimates, airway resistance (Raw) was slightly correlated with % predicted FVC (r = 0.49) (Figure 6D). Tissue resistance (R_t) correlated strongly with % predicted FEF_25-75 (r = 0.86), FEV₁/ FVC (r = 0.82), FEV₁ (r = 0.67), and PEF (r = 0.62). However, we found the remaining parameters, I_aw, I_t, and C_t did not correlate with any spirometric indices.

View larger version (28K): [in this window] [in a new window]   Figure 6.   Regression data for some MEFV indices versus Ztr spectral features or six-element model parameters that displayed higher correlations (see Table 3). All data is for pre-bronchodilator conditions and only features or parameters that also exhibit high correlations for pre-versus-post analysis (Table 4) are shown.

With respect to pre- versus post-bronchodilator data, parameter changes were correlated against changes in spirometric indices for those patients given medication (Table 4). Figure 7 shows changes in those features that were well correlated with changed spirometry. Change in R_o was well correlated with change in % predicted FEV₁ (r = 0.74) (Figure 7A). Worth noting is the relationship between R_o and FEV₁. The correlations are strong when relating both absolute R_o and FEV₁ (pre bronchodilator data only) as well as relating changes in these values due to bronchodilator. Change in Re₄ correlated well with both % predicted PEF (r = 0.58) and % predicted FEV₁ (r = 0.71) (Figure 7C, D).

View larger version (32K): [in this window] [in a new window]   Figure 7.   Regression data for changes in some MEFV indices versus changes in key Ztr spectral features or six-element model parameters that displayed higher correlations (see Table 4).

Change in Raw correlated well with change in % predicted FEV₁, r = 0.60) (Figure 7B) and slightly with change in % predicted PEF (r = 0.49). No other model parameters showed changes in response to bronchodilator that were correlated with changes in any spirometric index (Table 4). Indeed, while R_t correlated well with % predicted FEF_25-75, FEV₁/FVC, FEV₁, and PEF, changes in R_t were completely uncorrelated with changes in these spirometric indices (Table 4). In other words, changes in lung physiology that occur due to bronchodilation capable of improving lung function are uncorrelated with changes that may occur in the tissue resistance as estimated from Z_tr approach. It is worth noting that R_t from Z_tr is based on data above 4 Hz while the viscoelastic nature of lung tissue would cause lung tissue resistance to have its greatest influence at much lower frequencies (16, 17).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

The technique of standard spirometry to assess lung mechanical function requires considerable patient cooperation. In comparison, Z_tr requires virtually no patient cooperation and provides a noninvasive approach for acquiring data reflective of lung mechanics. Moreover, one form of data analysis (application of the six-element model) provides direct estimates of specific lumped mechanical properties associated with the airways and tissues, but only if data out to 64 Hz are acquired. Nevertheless, there have been virtually no studies (over this frequency range) that have evaluated Z_tr as a clinical test of lung function.

The current study presents a comprehensive evaluation of whether Z_tr can be used to assess the degree and, perhaps, type of impaired lung function. This assessment was performed at three levels. First, we evaluated the sensitivity of specific Z_tr spectral features and mechanical properties (as estimated from the six-element model) to diseased lung conditions. Second, we correlated specific Z_tr features and mechanical properties to spirometric indices. Third, we compared the sensitivities of the Z_tr features and mechanical properties with the sensitivities of spirometry to administration of a bronchodilator. With respect to the six-element model analysis we note that the model presumes a monoalveolar behavior. However, for all diseased subjects the Z_tr data were always fit well by this model and a more complex model was not necessary. This suggests that Z_tr data from 2-72 Hz is not sensitive to lung inhomogeneities. Nevertheless, some spectral features and mechanical properties were found to be both highly sensitive to impaired lung function and to improved lung function resulting from bronchodilator treatment. Moreover, many of these features and properties require data over an expanded frequency range (at least 2-64 Hz), a range over which clinical data have not been previously published.

One concern is the potential for flow limitation to occur in COPD patients during the impedance measurements. One might suspect the potential for a fixed maximum flow during expiration regardless of forcing pressure for such a condition. We did not examine the maximum flow-volume loops compared with the tidal loops to discern if flow limitation occurred. However, we contend that wave propagation velocity (i.e., flow rate at which flow limitation occurs) is very frequency dependent (18). Thus even though flow limitation may occur for very low frequency normal breathing, the high frequency forced oscillations may not be flow limited. Moreover, the final Z_tr from a single individual represents the average of several spectra occurring throughout the breathing cycle. Since Z_tr measurements were made during inspiration and expiration, if flow limitation was occurring and distorting our data, one would expect highly variable spectra. However, the standard deviation bounds for the COPD patients as a group were not much larger than those for the normal subjects (Figure 2A). This further suggests a minimal distortion associated with any potential expiratory flow limitation.

Correlations and Z_tr Spectral Features

Several Z_tr spectral features were sensitive to the existence of and degree of lung disease. In particular Re₄ and R_o were substantially and significantly elevated in COPD patients and in smokers. Moreover, both features decreased substantially (Figures 3A, B and 4) after bronchodilator treatment. In several subjects the entire real part of Z_tr was elevated pre-bronchodilator, becoming essentially normal post-bronchodilator (e.g., Figure 3). This suggests that Z_tr features are sensitive to more peripheral airways that are responsive to bronchodilators.

With respect to spirometry, the spectral feature with the strongest correlation was R_o versus % predicted FVC. Two other spectral features, f Im_max and Re₄, also correlated with FVC although not as reliably. Only two spectral features (R_o and Re₄) showed strong correlations with spirometric indices both in absolute terms and with respect to correlating changes in these features with changes in spirometric indices (Figure 4 and Tables 3 and 4). It is important to note that the spectral feature most sensitive to disease, R_o, often occurs above 32 Hz. In summary, patients with abnormal R_o and Re₄ are highly likely to have abnormal spirometry. Likewise, patients who display reduced R_o and Re₄ after bronchodilator are highly likely to have improved lung function as assessed by spirometry. These results are encouraging with respect to the potential application of Z_tr as an alternative (but much easier to perform) lung function test.

Z_tr versus Z_in. Studies similar to ours have been performed using input impedance (Z_in) but with far less success. Van Noord and coworkers (6) compared the real part at low frequency denoted R_{rs 6} (their first acceptable impedance was at 6 Hz rather than 4 Hz), with spirometry. Similar to our results, they found the change in R_{rs 6} to be correlated with change in FEV₁, r = 0.55 (and with FVC, r = 0.32), but these correlations were weaker than we found with the Z_tr based R_o. Of particular interest is that we did not find that RES₁ from Z_tr was sensitive to lung disease and it did not correlate with spirometric indices. Alternatively, for Z_in, Lutchen and associates (20) found a strong correlation between the first resonance and FEV₁ (r = 0.69) when forcing is imposed at the mouth (i.e., Z_in) (20). Also, Chalker and coworkers (5) found strong correlations between the first Z_in based resonance and both FVC and FEV₁/FVC (r = 0.81, and r = 0.65, respectively) using Z_in. Finally, previous studies (e.g., 21) have not shown any Z_in spectral features above 2 Hz that are significantly altered in smokers while we showed substantial sensitivity of R_o and Re₄ in smokers relative to normal subjects.

Use of Z_tr Data to Identify Impaired or Improved Lung Mechanics

The spirometric parameters often used in clinical diagnosis are FVC, FEV₁, FEV₁/FVC, and FEF_25-75. A knowledge of each in addition to knowing lung volumes can aid in the diagnosis of disease. FVC, FEV₁, and FVC/FEV₁ values below 80% predicted often are indicative of abnormal lung function. Ten of 12 subjects with R_o > 45 Hz had a % predicted FVC < 80%. All seven subjects with R_o < 40.3 Hz had a % predicted FVC > 80%. Eleven subjects had an R_o value between 40.3 and 45.0 Hz. Of these, five had a % predicted FVC < 80% while six were above. The lowest FVC of this group was 55% of predicted and the highest 89%. In summary, our data suggest that R_o < ~ 40 Hz indicates a normal FVC, and R_o > ~ 45 Hz an abnormal FVC. Those values of R_o between 40-45 Hz imply borderline FVC values.

Bronchodilator treatment is considered successful if spirometric indices improve by approximately 10% of the original predicted value. We found high correlations between changes due to the bronchodilator in R_o and changes in several spirometric indices. Specifically, a decrease in R_o will imply increased FEV₁ and, to a lesser extent, FVC. Consistent with our analysis below, these changes are a consequence of reduced Raw and increased lung volume (i.e., C_g). Also, Re₄ was significantly reduced in subjects with improved FEV₁ and PEF after treatment. Again, a large decrease in this feature will also suggest improved lung function.

Alternatively, we do not expect that Z_tr data shows high specificity to the explicit pathophysiology of lung disease. Most notable, COPD is not a specific disease, but a syndrome to reflect several patients, often with variable pathophysiology (e.g., emphysema, chronic bronchitis, and asthma). Nevertheless, with our COPD subjects, the Z_tr data fell within rather tight bounds (Figure 2a) such that this group was completely distinguishable from the healthy Z_tr group. Also, the variability in the COPD spectra was similar to that of the spectra in healthy subjects. Moreover, the correlations we performed suggest that regardless of the underlying cause, a subject with low FVC (%) and low FEV₁ (%) is more likely to have high R_o, Re₄, and Raw (Figure 6). Also, subjects that respond to bronchodilator with increased FEV₁, FVC, and PEF are likely to show decreased R_o, Re₄, and Raw. If their lung condition is not improved by the dilator, these other spectral features or physiological properties will also not improve.

Mechanisms Contributing to Z_tr Features/Sensitivity

This study was not designed to provide an explicit, statistical evaluation of whether Z_tr can permit diagnosis of specific forms of lung disease. The goal was to establish whether Z_tr spectral features and derived parameters were sensitive to disease. Our data suggest that this is the case. Moreover, our data show that Z_tr (over this frequency range, 1-72 Hz) is more sensitive than Z_in to lung disease and this leads to the following questions: what mechanisms could have contributed to the improved sensitivity of R_o and Re₄ in Z_tr data to disease? Why is RES₁ less sensitive to disease for Z_tr than for Z_in?

One method to address this question is to apply the six-element model to these data. From Figure 5, it appears that only one model parameter, Raw, is significantly elevated (with high statistical certainty) during disease as well as responsive to bronchodilator. The fact that Raw from Z_tr compared well with that from plethysmography in healthy subjects (Table 2) provides some measure of confidence in the Z_tr approach. Unfortunately, we did not measure Raw in the diseased subjects. However, it is expected that Raw from plethysmography is overestimated due to upper airway shunting and/or inhomogeneities (cf. 4). Both mechanisms would be less important during Z_tr (11 and see below). Note that absolute values of R_t were well correlated with spirometry, but changes in R_t were uncorrelated with changes in spirometry. Since the R_t is based on data above 4 Hz, it reflects predominantly the purely viscous component of chest wall resistance (16, 17) and would not be expected to be a consistent marker of disease that is influenced by therapy. We might expect C_t to also show strong sensitivity to disease and changes in disease state. The COPD subjects displayed significantly lower compliance. This is at first surprising since at least emphysematic COPD is classically thought of as producing an increase in lung tissue compliance. Perhaps these COPD patients were considerably hyperinflated so the C_t estimate is overly influenced by chest wall effects. We further point out that the C_t estimate is not likely to be statistically reliable because C_t predominantly influences data well below 2 Hz (9, 22). In fact, in COPD one would expect dynamic compliance to decrease with frequency. This could distort the C_t estimate based on data from 4 Hz and above to a lower value that would be estimated from a static or very low frequency measurement.

Our model analysis suggests that R_o and Re₄ are highly sensitive to pathological changes that cause decreased airway caliper, and are less sensitive to changes in other lumped properties of the system. To examine this hypothesis a sensitivity analysis of the Z_tr spectral features to changes in six-element model parameters was performed. Each model parameter was varied from 0.1-2.0 times the normal baseline value while the remaining parameters were held constant. Changes in Z_tr spectral features were calculated with respect to the baseline condition for each parameter set. The greatest contributors to increased R_o and Re₄ are decreased thoracic gas volume (i.e., FRC, C_g) and increased Raw (Figure 8). Decreased I_aw or R_t can also cause increased R_o, but by lesser amounts. It is interesting that, in fact, both I_aw and R_t were also significantly lower in COPD patients pre-bronchodilator compared with normals (Figure 5). Decreased C_t can increase Re₄ substantially, but has little effect on R_o. For the seven patients whose lung volumes were measured after bronchodilator, the mean FRC did increase (which according to Figure 8 would decrease R_o since C_g would increase). However, the increase was very small and not statistically significant (Figure 5). Nevertheless, it is worth noting that the range of decreases in R_o due to bronchodilator was 0-23% which, from Figure 8, could easily occur with a physiologically small combined increase in FRC and decrease in Raw.

View larger version (22K): [in this window] [in a new window]   Figure 8.   Sensitivity analysis of two key Ztr features (Ro and Re4) to changes in each of the six-element model parameters. X-axes is the fractional change in the respective parameter from the baseline values indicated. Parameters were changed independently (i.e., keeping others fixed to the baseline values).

We do not fully understand why Raw and FEV₁ were not as highly correlated to FEV₁ as R_o was. Since Z_tr is not greatly influenced by airway wall compliance, the corresponding Raw estimate from the DuBois model should be a more pure measure of airway caliber. In contrast, MEFV parameters are a function of airway caliber, lung elastic recoil, and airway wall compliance. Therefore we might expect a close correlation between Raw and MEFV only in those patients whose principal lesion is airway caliper related. We further point out that Raw showed some correlation with FVC (% predicted) (Table 3 and Figure 6) and that changes in Raw were highly correlated with changes in FEV (% predicted) (Figure 7). One might expect absolute values of Raw to correlate less as they are not normalized in any way. Generally, though, abnormal Raw extracted from Z_tr data is somewhat predictive of abnormal spirometry (in FVC) and changes in Raw following bronchodilator therapy are highly predictive of an increase in FEV₁.

We point out that R_o is not exclusively influenced by Raw. The sensitivity study (Figure 8) examines the influence of changes in one parameter at a time keeping all others at their baseline values. The R_o can increase primarily due to an increase in Raw, or a decrease in I_aw, and FRC (i.e., C_g) and R_t. As mentioned, it is unlikely that R_t is important or reflective of the lung. Second, the COPD subjects did not show increased C_g at baseline. This leaves the possibility that R_o reflects a combination of the increase in Raw and a decrease in I_aw. It is not clear why I_aw would decrease during COPD. Nevertheless, we speculate that since R_o is influenced by several parameters rather than just Raw alone, it tends to amplify the changes that occur during decreased airway caliper associated with COPD and becomes more sensitive to abnormal spirometry than just Raw alone.

How can we explain the poor RES₁ correlations with spirometry using Z_tr while previous studies (5, 20) report fairly high correlations when using Z_in? The answer lies in the difference in the degree to which upper airway shunting can effect Z_in versus Z_tr (11). For both Z_tr and Z_in impedance, RES₁ is inversely proportional to the product of the net total respiratory compliance (C_rs) and total respiratory inertance (I_rs). With Z_in and Z_tr the I_rs primarily reflects the upper and central airways. In Z_in the C_rs in healthy subjects is primarily the parenchymal and chest wall C_t with little influence from the airway wall compliance. However, as peripheral airway constriction increases, the net impedance "downstream" of the I_aw increases. Specifically, the influence of airway wall shunting increases. Since airway wall compliance is substantially less than that of the parenchymal tissue, the net effect of increased peripheral resistance is to increase the RES₁. For Z_tr, the increased peripheral lung impedance will have a far less effect on C_rs and thus on RES₁. In particular, when forcing at the chest wall, the impedance "downstream" of the constriction remains unaltered, and consists primarily of a low impedance pneumotachometer and the upper airways. In effect, during lung constriction one would expect different values of RES₁ for Z_in and Z_tr, each of which reflects different components of the lung and experimental system. The RES₁ for Z_tr would be less correlated with spirometric indices because it is less influenced by airway wall shunting. This further implies that during lung constriction, six-element model analysis of Z_tr better reflects intrathoracic Raw. This would explain why Z_tr features are more sensitive than Z_in features in smokers or with changes due to bronchodilation.

	SUMMARY

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

In summary, we measured Z_tr up to 128 Hz in 11 normal subjects and 26 patients with lung disease diagnosed by standard MEFV pulmonary function test and analyzed the data with the six-element monoalveolar model as proposed by DuBois (8). Model parameters and Z_tr spectral features were correlated with spirometric data. Changes in model and spectral features were also correlated with changed spirometry.

The Re₄ was also significantly elevated in diseased subjects. Ying and coworkers (3) used a similar spectral feature, the mean value of the real part of 0-30 Hz, and found it to be significantly elevated in diseased subjects. However, the effects of bronchodilator had not been investigated using Z_tr data. Moreover, we have identified that the feature R_o is even more effective in predicting normal versus abnormal spirometry. To this date, this feature has not been evaluated since it is typically above 32 Hz (well above 32 Hz in diseased individuals) and not encompassed in previous clinical studies. This feature can also track changes due to bronchodilator and be used to evaluate therapeutic efficacy. We have also established normal ranges for several parameters and features, specifically, Raw, R_t, R_o, R_a, f Im_max, and Re₄ and demonstrated that they are affected by disease.