INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The imbalance of A/ relationship is a major mechanism of abnormal arterial blood gases in chronic obstructive pulmonary disease (COPD), as demonstrated by the multiple inert gas technique (MIGET). Even clinically stable mild or moderate COPD patients exhibit A/ mismatching (1). This accounts almost completely for the observed arterial hypoxemia, as the intrapulmonary (right-to-left) shunt is negligible under stable clinical conditions in these patients. The pattern and the severity of A/ mismatch differ among patients with COPD and change with time, more or less according to the evolution of the disease and the clinical state of the patient (2). The degree of A/ inequality in patients with COPD does not correlate very well with the severity of airflow obstruction, nor with the outcome of the disease (1, 3, 4). It is, however, accepted that the deterioration of the A/ distribution with time reflects progressive structural changes of the lung affecting both the airways and the blood vessels (2).

In the study of Marthan and coworkers, emphysema severity correlated positively with the alveolar-arterial pressure difference for oxygen (PA  a_O2) and negatively with Pa_O2 (3). Furthermore, emphysema severity was related not only to the dispersion of alveolar ventilation but also to that of blood flow. The abnormalities in the dispersion of pulmonary perfusion both in low and high pattern ventilation perfusion distribution patients (1) was clearly not a negligible determinant of abnormal A/ ratios. These abnormalities may be due to at least two different factors. First, they may result from a remodeling of the pulmonary capillary network in emphysematous zones (3). Second, they may be related to changes in vasoreactivity. It has been shown that the thickening of the intimal layer of pulmonary muscular arteries interferes with their vascular reactivity to oxygen (5). The more severe the lesions of the vessel walls, the less the reversal of hypoxic vasoconstriction by oxygen. It has therefore been suggested that the lesions of the pulmonary artery wall play a key role in the pathogenesis of pulmonary hypertension (6) during the late phase of COPD. These lesions affect the level of vascular reactivity that contributes to the maintenance of adequate A/ matching at an earlier stage of the disease, well before pulmonary hypertension develops (7, 8).

Pulmonary arterial hypertension develops late in the course of COPD and is associated with a poor prognosis, hence the traditional focus on pulmonary pressure measurements (9, 10). Because in patients without severe hypoxemia, pulmonary arterial pressure and pulmonary vascular resistance are usually normal or only slightly elevated at rest (11), the hypothesis that pulmonary circulation plays a determinant role in A/ modulation early in COPD has often been overlooked. However the relationship between early pulmonary vascular abnormalities and the changes in A/ pattern in patients with "mild" COPD has been clearly supported by morphologic and immunochemical studies performed on pulmonary muscular arteries specimens obtained in resected lungs (4, 5, 12). The lack of a reasonably noninvasive and simple method to monitor lung blood flow distribution has hampered further physiological studies on possible early pulmonary circulation remodeling, vasoreactivity changes, and pulmonary blood flow redistribution. The distribution function of pulmonary transit times of an intravascular indicator reflects both perfused vascular volume and contact time between blood and vascular surface area (13, 14). Various methods to determine the transit times distribution have been proposed (15). However, the need for catheterization and other methodological difficulties (e.g., corrections for intravascular indicator injection and sampling transport functions) have hampered the clinical relevance of this approach. We recently demonstrated that the pulmonary transit time distribution through the lung can be assessed almost noninvasively in patients, by first pass angiocardioscintigraphy of an intravascular indicator (albumin ^99mTc) (15). The purpose of the present study was to assess whether significant changes of pulmonary transit time distribution can be evidenced early in the course of COPD and whether these changes are related to hypoxemia or hypoxia.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

The study was conducted according to institutional ethical regulations. Patients were stratified on the basis of pulmonary function tests (PFT) and cardiac performance (Tables 1, 2, and 3). Group 1 (control, n = 11) had normal PFTs, cardiac index (CI > 3.1 L min¹ m²) and ventricular function (left ventricular ejection fraction, LVEF > 0.50, right ventricular ejection fraction, RVEF > 0.45). Group 2 (COPD, n = 11) included stage 1 COPD patients according to ATS staging (16). Group 3 (low CI, n = 11) included patients with normal PFTs and chest radiographs, no clinical signs of pulmonary disease, but a low cardiac index (< 3.1 L min¹ m²). Group 4 (restrictive, n = 7) included hypoxemic patients (Pa_O2 < 80 mm Hg) with a total lung capacity (TLC) below 80% predicted and no signs of obstructive disease on the flow-volume curve. Group 5 (acute hypoxia, n = 8) was designed to obtain an index of changes in distribution occurring during moderate acute hypoxia. We incorporated patients with normal cardiopulmonary function after corrective surgery for interatrial defect during randomized sequences of normoxia and acute hypoxia (15 min, FI_O2 = 0.12).

Tracer

Albumin ^99mTc was injected through an external jugular vein via an indwelling intravenous catheter (15). A total activity of 10 MBq kg¹ was injected in three aliquots. Aliquot 1(1 MBq kg¹) served to position the camera in the best septal position, aliquot 2 (4 MBq kg¹) for the first pass left anterior oblique (LAO) recording, and aliquot 3 for a right anterior oblique (RAO) recording.

Imaging Procedure

The first pass of indicator was recorded with a SOPHA gamma camera using 0.5 s frames. RAO data acquisition was performed in list mode sequence (17). We also recorded the superior vena cava time activity curve to check that bolus fragmentation did not occur.

Data Analysis

Right and left ventricular time activity curves were derived from regions of interest (ROI). We deconvoluted the left ventricular curve, the lung output function, by the right ventricular curve, the lung input function, to obtain the distribution of transit times of the tracer through the lung (18) as described by Capderou and coworkers (15). Cardiac output (CO) was computed from left ventricular time activity curves (19). Left (LVEF) and right (RVEF) ventricular ejection fractions were computed from gated ventriculography (17).

We based our analysis on the comparison of the moments of the distributions (20). We computed the mean transit time (mtt) of the pulmonary distribution (h[t]), mtt = t · h(t)/h(t), and other moments (Mn) from Mn = h(t) · (t  mtt)ⁿ/h(t). The square root of the second moment (M2), the standard deviation (disp), was used to assess the dispersion of transit times around the mean. We computed from these moments (1) the skewness (sk) defined as the ratio M3/ M2^3/2, an index of the curve asymmetry and (2) the kurtosis (kr) defined as the ratio M4/M2², an index of the curve "flatness" (21, 22). The relative dispersion (rel disp) was defined as the ratio disp/mtt. We also compared distributions once normalized by their mean transit time (23) to correct for variations solely related to fluctuations in cardiac output. Statistical analysis was conducted using ANOVA (21). p < 0.05 was considered significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

An example of ventricular time activity curves and the resulting h(t) is given in Figure 1. The bolus criteria defined in METHODS were met in all experiments. Cardiac output (CO) data, cardiopulmonary blood volume data, and ejection fractions are listed in Table 3. Distribution descriptors are given in Table 4. Superimposed point-by-point averaged distributions for groups 1 to 4 before and after normalization are given, respectively, in Figures 2 and 3, following the mode of representation suggested by Presson and coworkers (24). Because of individual variations in the time width of the distributions, error bars representing SEM are only meaningful for the normalized curves.

View larger version (20K): [in this window] [in a new window]   Figure 1.   (A) Right and left time activity curves and (B) the resulting lung transport function, h(t), for a patient with COPD.

View larger version (18K): [in this window] [in a new window]   Figure 2.   Distributions of pulmonary transit times for control (group 1), COPD (group 2), low cardiac index (group 3), and restrictive (group 4) patients. Each curve represents the point-by-point mean values for the group (see text for explanations). Each distribution is significantly different from all others. The difference is due to two factors: a change in distribution linked to the difference in cardiac output between groups and a redistribution linked to each pathology.

View larger version (20K): [in this window] [in a new window]   Figure 3.   Distributions of pulmonary transit times after normalization by mean transit time for the four groups of Figure 1. Changes in distribution induced by independent variations of cardiac output are eliminated by the normalization (see text for explanation). The normalized low CI distribution does not differ from control. Normalized COPD and restrictive distributions differ significantly from both low CI and control and also from each other. Error bars represent mean ± SEM.

Groups 1 to 4 had comparable morphometric data (Table 1), with the exception of group 2 (COPD), which had a lower bodyweight (p < 0.05) than all others. Pa_O2, Pa_CO2, and [H⁺] in the COPD and restrictive groups were not statistically different from each other but differed from the low cardiac index group (Tables 2 and 3). LVEF and RVEF differed from controls only in the restrictive patients. However they remained within the normal range.

Mean pulmonary transit time was significantly longer in the low CI group. This was not linked to any change in cardiopulmonary blood volume (Table 3). The ratio of the mean pulmonary transit time to the R-R interval, which expresses the pulmonary blood volume in number of stroke volumes (SV), was significantly higher in all patients than in control subjects. This was related to significantly lower values of SV. In the low CI group the low SV value was concordant with a normal heart rate. Standard deviation of the distribution, which is a dispersion descriptor (22, 23), was significantly increased in the three groups of patients compared with control subjects. This indicates heterogeneity of pulmonary blood flow, but is not sufficient to isolate the influence of changes in cardiac output from that of other factors. Indeed, there were important individual variations in the restrictive and COPD groups. Dispersion was similar in restrictive and patients with COPD. In low CI patients, it tended to be lower. The distribution shape was also significantly flatter in patients with COPD than in control and low CI patients (see kurtosis data in Table 4). Because independent variations in CO could have interfered with other anatomical or physiological factors it was essential to normalize dispersion for cardiac output. This was done using the relative dispersion defined as the ratio of dispersion to mean transit time (22, 23). According to this index, the low CI group was not different any more from the control group, suggesting that CO was a key element to explain the observed difference. Furthermore, the relative dispersions of the COPD and of the restrictive groups were significantly different from one another and from that of the control group. Therefore, factors other than a change in CO were responsible for the differences observed in the shape of the distributions. Again, the most significant changes were observed in the COPD group.

The reconstructed mean distributions as a function of time for each group are displayed in Figure 2. Each curve was obtained by averaging individual values at each sampling time. Although this representation tends to smooth individual variations, it gives a reasonable image of shape differences between groups (24). However, it cannot be used for statistical analysis because of the dispersion of the distributions and of intersubjects CO variability. We therefore normalized each curve by its own mean transit time (15, 22). This procedure presents not only the advantage of ignoring cardiac output variations when they are not linked to changes intrinsic to the pulmonary circulation as stated before, but also the advantage of standardizing time intervals so that the curves can be averaged in the time domain (Figure 3). The distributions of the low CI group were almost superimposed to those of the control group after normalization. The normalized distributions in the restrictive group were significantly different from that of control subjects, with an earlier rise and a lower peak. The normalized distributions of the COPD group were significantly different from those of all other groups, with a fast start, a slowly rising slope, an early and very flattened peak, and a widened descending part. During normoxia the normalized distributions of the young subjects of group 5 did not differ from the normalized distributions of the older patients of control group 1 (Figure 4). The distributions and normalized distributions of group 5 subjects were significantly different during acute hypoxia when compared with normoxia (Figure 5 and Table 5). Interestingly, the normalized distributions observed during acute hypoxia were similar to the distributions observed in the restrictive group, but were significantly different from the COPD group (Figure 6). Although actual measurements could not be performed because arterial sampling was not permitted in these subjects, the Pa_O2 level induced by an FI_O2 of 0.12 in the group 5 patients could reasonably be predicted to be around the Pa_O2 measured in the COPD and restrictive groups.

View larger version (15K): [in this window] [in a new window]   Figure 4.   Normalized distributions of pulmonary transit times in the three normoxic and normoxemic groups: control (group 1), low cardiac index (group 3), baseline of acute hypoxia (group 5). The distributions are not significantly different. Error bars represent mean ± SEM.

View larger version (18K): [in this window] [in a new window]   Figure 5.   Distributions of pulmonary transit times (A) and normalized distributions of pulmonary transit times (B) during normoxia (FIO2 = 0.21) and during acute hypoxia (FIO2 = 0.12) in the young nonpulmonary patients group. Distributions and normalized distributions become significantly different from baseline as a result of acute hypoxia. Error bars in B represent mean ± SEM.

View larger version (18K): [in this window] [in a new window]   Figure 6.   Normalized distributions of pulmonary transit times in the three hypoxemic groups: COPD (group 2), restrictive (group 4), and acute hypoxia (group 5). The restrictive and the acute hypoxia distributions do not differ from each other. The COPD distribution differs significantly from the restrictive and the acute hypoxia distributions. Error bars represent mean ± SEM.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Assessment of the pulmonary vascular impact of COPD relies mostly on pulmonary artery and pulmonary artery occlusion pressure data. Pulmonary arterial hypertension generally occurs late in the evolution of the disease. Because pressure changes can result only from major changes in either vessels structure or reactivity or both, they reflect the late stage of a chronic process about which little is known. Exercise-related pulmonary hypertension during exercise in patients with COPD with a normal baseline pulmonary arterial pressure has already led to the suspicion that COPD-related vascular abnormalities might occur much earlier during the course of the disease than previously thought. Because histological material is difficult to obtain in early or mild COPD, few data on early changes in the pulmonary vasculature are available. Enlargement of the intimal layer in pulmonary muscular arteries is associated with a higher degree of A/ inequality in mild COPD (6, 8). The possible role of an inflammatory process in the pathogenesis of pulmonary vascular abnormalities in the early stage of COPD has also been recently documented, but the relationship between these abnormalities and possible early functional changes in the pulmonary circulation remains speculative. (8). Indeed hemodynamic investigation techniques are neither appropriate nor sensitive enough, at these stages and their use in patients with a moderate disease is ethically questionable.

The calculation of the pulmonary distribution of transit times of a ^99mTc intravascular tracer by adequate deconvolution techniques from standard angiocardioscintigraphy, beside being not very invasive, is a powerful tool for the study of the pulmonary circulation. It is both sensitive and reproducible (15). Furthermore normalization for cardiac output allows distinguishing changes in pulmonary transit time distribution directly linked to cardiac output variations from changes related to structural or functional modifications of the pulmonary vascular tree. In this study, we used this approach in so-called moderate (ERS staging) or stage 1 (ATS staging) patients with COPD (16) to establish if early functional changes in the distribution of pulmonary transit times were detectable. The study demonstrates that the distribution of pulmonary transit times in patients with stage 1 COPD differs from the distributions observed in all other groups, including normals, but more interestingly from patients both restrictive and hypoxemic with comparable Pa_O2 and Pa_CO2 levels, and acute hypoxic-hypoxemic subjects. The low cardiac index observed in patients with COPD could offer a simple explanation for this difference. Because the normalized distribution of pulmonary blood flow of patients with normal lung function and the same cardiac index was identical to that of control subjects, this explanation appears unlikely. It was therefore important to assess, at least indirectly, whether the observed changes in distribution were coherent with ventilation-perfusion mismatch. For this purpose we computed the average ventilation-perfusion ratio (A/) for each group, from blood gases and cardiac output data, using the P O₂-PCO₂ Rahn diagram, assuming a normal respiratory exchange ratio of R = 0.8 (25). Calculations were not performed for the young subjects of group 5 because blood gases were unavailable in this group as stated above. We found a value of 1.12 in the COPD group, 1.08 in the low cardiac index group, and 0.81 in the restrictive group. These values are in accordance with A/ ratios directly measured using MIGET in comparable patients (26). If we make the assumption that the alveolar ventilation A was either normal or slightly lower than normal in our patients with COPD, then the increased in the computed A/ ratio is consistent with our observation of a lower than normal pulmonary blood flow in the COPD group. The alternate explanation of a right-to-left shunt was ruled out by the angioscintigraphic data. Consequently, we suggest that significant alterations in the structure or the reactivity of the pulmonary vascular bed may occur very early in the course of the disease, on par with the early modifications of airway structure and function. It is interesting to note that lung vascular modifications were more important in our COPD group than in our restrictive group for a given level of hypoxemia. Changes in cardiopulmonary blood volume paralleled differences in distributions between groups. Cardiopulmonary blood volume was significantly lower in patients with COPD compared with the restrictive and normal groups but the ratio of cardiopulmonary blood volume to stroke volume was higher in patients with COPD compared with normals. This was due in part to a low stroke volume. The fact that we observed a rise in mean transit times in patients with COPD although cardiopulmonary blood volume was lower compared with control subjects supports the hypothesis that changes in normalized distribution shapes were more representative of changes in blood flow distribution than of changes in blood volume. However, cardiopulmonary blood volume data must be interpreted with extreme caution because it includes possible variations in cardiac as well as in lung blood volumes.

Potentially, the mechanical compression of lung microvessels may induce changes in distribution similar to those observed in patients with COPD. The experimental mechanical compression of lung capillaries during a Valsalva maneuver can induce zone 1 conditions, slowing transit times in some parts of the lung and therefore widening the transit time distribution (27). In our spontaneously breathing, supine, patients with stage 1 COPD, the existence of zone 1 conditions, extended enough to explain a change in the distribution shape, is unlikely. However, this may become an important factor in patients suffering from a more advanced form of the disease, in particular in the presence of intrinsic PEEP.

Besides mechanical factors and anatomical remodeling, another hypothesis to explain our results involves active vasoconstriction, linked to alveolar hypoxia or to a low Pv_O2. Alveolar hypoxia appears an attractive hypothesis because it is likely to be found in patients with COPD (were we found maximal changes in distribution), and less likely in restrictive patients (in whom we found less marked changes in distribution). However, as stated in the result section, the changes observed in the acute hypoxic subjects were more like the changes observed in the restrictive patients. Indeed it is questionable to compare the pulmonary vascular reactivity of young subjects during an acute experiment to that of moderately hypoxic older patients with COPD. It is also important to remember that the lungs of the young subjects during hypoxia were exposed to global hypoxia whereas hypoxia was most likely to be inhomogeneously distributed in the lungs of the patients with COPD. In spite of these limitations, the strong possibility suggested by our data that the expected normal physiological changes in vascular reactivity induced by alveolar hypoxia did not fully explain the shape of the COPD distribution needs to be discussed further. The degree of arterial hypoxemia was similar in the restrictive and COPD groups. Hypoxemia alone, that is to say in the absence of a change of vasoreactivity caused by the disease, was therefore unlikely to have caused vasoactive redistribution in one group and not in the other. However, because cardiac output was significantly lower in the COPD group, compared with the restrictive group, the possibility that Pv_O2 was low enough in the latter to potentiate redistribution of pulmonary blood flow in a lung already hypoxic cannot be ruled out (28). Therefore the most likely hypothesis to explain the COPD distribution data remains an early change in structure or reactivity of the vascular tree of these patients. In any case, a significant change in distribution does not require extremely important changes in the tone of all the vessels, because limited modifications in the pathways of flow may result in important changes in functional distribution of A/ as shown by West (29).

We conclude that redistribution of blood flow, whether by passive or active mechanisms, or both, may occur early during the course of chronic obstructive pulmonary disease. Our study correlates well with the histological and immunochemical data obtained on pulmonary muscular arteries in patients with "mild" COPD (6). The computation of the distribution of pulmonary transit times by first pass angiocardioscintigraphy may therefore offer a powerful tool for the noninvasive follow-up of vascular impairment in the disease, including modifications of pulmonary vascular reactivity induced by various physiological or pharmacological challenges such as oxygen, prostaglandins, or corticosteroids.