Published ahead of print on December 12, 2002, doi:10.1164/rccm.200210-1130OC
© 2003 American Thoracic Society
Echocardiographic Assessment of Pulmonary Hypertension in Patients with Advanced Lung DiseaseDivision of Pulmonary, Allergy and Critical Care, Columbia University College of Physicians and Surgeons, New York, New York; and Divisions of Pulmonary, Allergy and Critical Care Medicine, Cardiothoracic Surgery, and Cardiovascular Medicine, University of Pennsylvania Medical Center, Philadelphia, Pennsylvania Correspondence and requests for reprints should be addressed to Selim M. Arcasoy, M.D., Columbia University College of Physicians and Surgeons, Lung Transplantation Program, PH-14E, Room 104, 622 West 168th Street, New York, NY 10032. E-mail: sa2059{at}columbia.edu
Doppler echocardiography is commonly used to estimate systolic pulmonary artery pressure and to diagnose pulmonary hypertension, but data relating to its utility in patients with advanced lung disease are limited. In a cohort study of 374 lung transplant candidates, the performance characteristics of echocardiography compared with right heart catheterization in the determination of systolic pulmonary artery pressure and diagnosis of pulmonary hypertension were investigated. The prevalence of pulmonary hypertension was 25% in the study population. Estimation of systolic pulmonary artery pressure by echocardiography was possible in 166 patients (44%). The correlation between systolic pulmonary artery pressure estimated by echocardiography and measured by cardiac catheterization was good (r = 0.69, p < 0.0001). However, 52% of pressure estimations were found to be inaccurate (more than 10 mm Hg difference compared with measured pressure), and 48% of patients were misclassified as having pulmonary hypertension by echocardiography. Sensitivity, specificity, and positive and negative predictive values of systolic pulmonary artery pressure estimation for diagnosis of pulmonary hypertension were 85%, 55%, 52%, and 87%, respectively. In conclusion, despite a statistically significant correlation with directly measured values, estimation of systolic pulmonary artery pressure by echocardiography is frequently inaccurate in patients with advanced lung disease and leads to considerable overdiagnosis of pulmonary hypertension.
Key Words: pulmonary hypertension • echocardiography • cardiac catheterization • lung disease Pulmonary hypertension (PH) is a common feature signaling the advanced stages of diseases of the pulmonary vasculature, parenchyma, and airways. Detection of the presence of PH provides valuable prognostic information and dictates the need for such therapies as vasodilators and supplemental oxygen (1–3). Moreover, the magnitude of PH influences decisions on the appropriateness of surgical options for treatment of advanced lung disease. In this regard, the presence of significant PH often prompts consideration of lung transplantation in the appropriate setting (4). Conversely, PH is considered a contraindication to lung volume reduction surgery for advanced emphysema (5). Although pulmonary artery pressures and other cardiac hemodynamic parameters can be accurately assessed by right heart catheterization (RHC), a simple, reliable, and noninvasive method to estimate pulmonary artery pressures and diagnose PH in patients with advanced lung disease is preferable. Doppler echocardiography (DE) has gained popularity in the last 2 decades for noninvasive estimation of systolic pulmonary artery pressure (sPAP) from the peak velocity of a tricuspid regurgitant jet. Studies in patients with cardiac disease have revealed a significant statistical correlation between sPAP estimated by DE and that measured by RHC (6–9). However, in patients with chronic pulmonary disease, DE has been reported to perform variably in the assessment of sPAP (10–12). Despite a close correlation between DE-estimated and directly measured sPAP, several studies have revealed that such estimations were possible in only a minority of patients with chronic lung disease (11–14). More importantly, by focusing exclusively on the correlation between estimated and measured pressures, previous studies have largely ignored the important issue of the true accuracy of DE estimation of sPAP. That is, if DE were to consistently overestimate or underestimate sPAP by a constant value, correlation would be high, but the accuracy of the estimation would be dictated by the magnitude of discrepancy. It is the accuracy of estimation that defines the clinical utility of DE in such circumstances as evaluation of patients for lung volume reduction surgery or lung transplantation. Patients with advanced lung disease referred for lung transplantation at our center are required to undergo both DE and RHC as part of their evaluation. Using this large cohort of patients, the aim of this study was to examine the performance characteristics of DE in estimating sPAP and detecting the presence of PH.
Consecutive patients with advanced lung disease who underwent an evaluation for lung transplantation between January 1991 and December 2000 were screened. The study population included all patients who had DE and RHC within 72 hours of each other during a formal evaluation process. DE was typically performed 24–48 hours before RHC according to our transplant evaluation protocol. The studies were performed independently by operators who had no knowledge of the results of the other study. The clinical status and treatment of these patients were unaltered during the evaluation period.
DE Technique RAP was empirically estimated as 15 mm Hg before 1997. Since 1997, RAP was estimated to be 5, 10, or 15 mm Hg based on the variation in the size of inferior vena cava with inspiration as follows (16): complete collapse, RAP = 5 mm Hg; partial collapse, RAP = 10 mm Hg; and no collapse, RAP = 15 mm Hg.
RHC
Definitions and Statistical Analysis We assessed the capacity of DE to estimate sPAP and diagnose PH in three ways. First, to examine the overall relationship of DE-estimated sPAP to RHC measurements, we compared the results as continuous variables using Pearson and intraclass correlation methods. Second, we calculated the "accuracy" of DE-estimated sPAP, defined as the percentage of studies that fell within 10 mm Hg of the pressure measured directly by RHC. Third, using RHC as the gold standard, we calculated the sensitivity, specificity, and positive and negative predictive values of two methods of diagnosing PH by DE: (1) estimation of sPAP by trans-tricuspid gradient method and (2) the presence of qualitative right ventricular structural or functional abnormalities. By the former method, a DE-estimated sPAP equal to or exceeding 45 mm Hg was considered to be positive for the diagnosis of PH. By the latter method, the presence of any one of the following findings was considered as a positive study: right ventricular hypertrophy, dilation, or systolic dysfunction. To determine whether estimation of RAP by DE impacted significantly on the accuracy of sPAP estimation, DE-derived sPAP estimates were recalculated by using the actual RAP measured by RHC in the formula, and the correlation coefficients and accuracy were recalculated. Results are expressed as means ± SD or as percentages. Comparisons of percentages among disease subpopulations were made by a chi-square test. In all analyses, a p value of 0.05 or less was considered significant. All statistical comparisons were performed using STATA version 7.0 (STATA Corp., College Station, TX). This research protocol was granted exempt status by the Institutional Review Board of the Office of Regulatory Affairs at the University of Pennsylvania.
During the study period, 592 consecutive patients underwent a formal evaluation for lung transplantation. Three hundred and seventy-four patients had DE and RHC within 3 days of each other. This group forms the study population. The clinical characteristics of study patients are shown in Table 1 . Two hundred fifty-three patients (68%) had obstructive lung disease; 106 patients (28%) had interstitial lung disease, and 15 patients (4%) had pulmonary vascular disease. By RHC, 94 patients (25%) had PH, whereas 280 patients (75%) did not. The prevalence of PH was 18% among the obstructive lung disease population, 59% among those with interstitial lung disease, and 100% among those with pulmonary vascular disease.
Estimation of sPAP by DE was possible in 166 (44%) of the 374 patients (Figure 1) . Patients with RHC-confirmed PH were more likely to have an estimated sPAP by DE compared with patients without PH (64% versus 38%, p < 0.0001). Estimation of sPAP was achieved less frequently in patients with obstructive lung diseases compared with those with interstitial lung diseases or pulmonary vascular disease (estimation possible in 38%, 54%, and 67%, respectively; p = 0.008). Similarly, the likelihood of estimating sPAP was significantly lower in patients with marked air trapping, defined as a residual volume exceeding 150% of predicted, compared with those without air trapping of this magnitude (40% versus 56%, p = 0.007).
Correlation and Accuracy In the subset of 166 patients in whom sPAP could be estimated by DE, there was a strong correlation between sPAP estimated by DE and that measured by RHC (r = 0.69, p < 0.0001 by Pearson correlation) (Figure 2) . Likewise, the intraclass correlation was strong (r = 0.60, p < 0.0001).
We observed a significant discrepancy between RAP estimated by DE and measured by RHC (15 ± 1 versus 8 ± 5 mm Hg, p < 0.001). The magnitude of this discrepancy was not affected by the method of estimating RAP (universal assumption of an RAP of 15 mm Hg versus estimation of RAP by the pattern of inferior vena caval collapse). To determine whether the estimation of RAP by DE impacted significantly on the correlation between estimated and measured sPAP, the actual RAP measurements obtained by RHC were used to recalculate DE-estimated sPAP values. The correlation coefficient improved only slightly when RAP measured by RHC was substituted for that estimated by DE and DE sPAP was recalculated (r = 0.74, p < 0.0001). Despite a statistically significant correlation between sPAP estimated by DE and that measured by RHC, DE-estimated sPAP was accurate (within 10 mm Hg of RHC measurement) in only 48% (95% CI, 40 to 56%) of the studies. Moreover, 47 of 166 (28%) studies revealed a difference between DE-estimated and RHC-measured sPAP of more than 20 mm Hg, and in 15 studies (9%), the difference was more than 30 mm Hg. The accuracy of DE was higher in patients with obstructive lung disease compared with those with interstitial lung disease and pulmonary vascular disease (56% versus 37% versus 20%, p = 0.018). The accuracy of DE was not different in patients with or without PH as determined by RHC (50% versus 47%, p = 0.89). However, in patients without PH, DE was more likely to overestimate sPAP, whereas in those with PH, DE was as likely to overestimate as underestimate sPAP (Figure 3) . When DE-estimated sPAP was recalculated using RAP measured by RHC rather than estimated RAP, the overall accuracy of DE improved only modestly (55% accuracy compared with 48% when estimated RAP was used).
Conversely, when estimated sPAP was 45 mm Hg or more by DE, it was accurate in only 31% of the studies compared with 72% of studies in which estimated sPAP was less than 45 mm Hg (p < 0.0001). Importantly, 48 of 99 patients (48%) with sPAP of 45 mm Hg or more by DE did not have PH by RHC (Figure 1). In all but four of these false-positive cases, the difference in sPAP obtained by the two methods was more than 10 mm Hg.
Diagnosis of PH by DE
Next, the performance characteristics of DE-identified right ventricular abnormalities as a surrogate diagnostic marker of PH were examined (Table 2). The right ventricle was adequately visualized in 372 of the 374 patients. When applied to this group, the sensitivity, specificity, and positive and negative predictive values of right ventricular abnormalities in diagnosing PH were 82% (95% CI, 73 to 89%), 57% (95% CI, 51 to 62%), 39% (95% CI, 32 to 46%), and 90% (95% CI, 85 to 94%), respectively. The utility of using right ventricular findings in patients in whom sPAP could not be estimated was then specifically scrutinized (Figure 1). In 208 patients in whom sPAP could not be estimated, right ventricular findings were present in 88 (42%), but only 29 of these 88 (33%) patients had PH by RHC. In 120 patients who had no right ventricular findings, 115 (96%) did not have PH by RHC. Thus, the finding of right ventricular abnormalities led to considerable overdiagnosis of PH, but the absence of these abnormalities reliably excluded PH. Finally, we calculated the performance characteristics of DE-estimated sPAP and right ventricular abnormalities in establishing a diagnosis of PH for each of the two main disease subgroups: obstructive lung disease and interstitial lung disease (Table 2). Similar to the group as a whole, the specificity of these DE parameters was generally lower than their sensitivity, a discrepancy that was particularly apparent in relationship to DE-estimated sPAP among the interstitial lung disease group. The positive predictive value of DE studies was low in both disease subgroups. In contrast, the negative predictive value of DE studies exceeded 90% in the obstructive lung disease group, whereas corresponding values were lower and more variable in the interstitial lung disease population. The universal prevalence of PH among patients with pulmonary vascular disease precluded calculation of all but sensitivity in this group, which was 100% for DE-estimated sPAP and 93% for the presence of right ventricular abnormalities.
In this study, we investigated the value of DE as a noninvasive means of estimating sPAP and establishing or excluding a diagnosis of significant PH in a large cohort of patients with advanced lung disease of various etiologies. We demonstrated a reasonably strong and statistically significant correlation between DE-estimated sPAP and RHC-measured sPAP. Despite this significant correlation, several important limitations of DE were identified. First, sPAP estimation by DE was possible in less than one-half of all patients. Second, there was a discordance of greater than 10 mm Hg between estimated and measured sPAP in 52% of patients, and in 28%, the discordance was greater than 20 mm Hg. Third, DE-estimated sPAP was associated with a poor positive predictive value for the diagnosis of PH, with confirmation of PH by RHC in only one-half of those patients in whom DE suggested its presence. Reliance on findings of right ventricular abnormalities did not enhance the poor positive predictive value of DE. The technique employed in this study to estimate sPAP by DE involves calculation of the trans-tricuspid gradient from measurement of the peak velocity of the tricuspid regurgitation jet using the modified Bernoulli equation (6, 7, 9, 15). Right ventricular pressure and, indirectly, sPAP are then determined by adding an estimated RAP to the calculated trans-tricuspid gradient. The success and accuracy of this technique are dependent on the ability to identify and precisely measure the velocity of the tricuspid regurgitant jet and, therefore, are influenced by factors altering the position of the heart in relationship to the DE probe, such as obesity and hyperinflated lungs. Intraobserver and interobserver variability and limitations in the estimation of RAP also impact the accuracy of this technique. The use of this technique has been most extensively and convincingly validated in patients with cardiac disease. In three of the largest studies, encompassing a total of over 250 patients, sPAP could be estimated in 59–87% of patients, and the correlation with measured sPAP was excellent, with a correlation coefficient of 0.93–0.97 (6, 7, 9). In contrast to these studies, we have focused on a population with advanced pulmonary disease, comprised chiefly of patients with severe chronic obstructive pulmonary disease and interstitial lung diseases. Our study comprises the largest cohort of patients with pulmonary disease studied to date and both corroborates and extends observations made in previous smaller series. We have demonstrated that both the likelihood of estimating a sPAP and the correlation with measured sPAP are lower in this population compared with those with cardiac disease. Regarding the first issue, we were able to estimate sPAP in only 44% of patients. It is likely that factors related to pulmonary disease contributed significantly to this. Specifically, changes associated with chronic obstructive pulmonary disease, including marked increase in intrathoracic gas, expansion of the thoracic cage, and alterations in the position of the heart, adversely influence the ability to detect and measure the velocity of the tricuspid regurgitant jet. Our demonstration that the likelihood of obtaining a DE-estimated sPAP was significantly lower in the subset of patients with obstructive lung disease and those with a residual volume exceeding 150% of predicted support this claim. The 38% success rate in estimating sPAP in patients with obstructive lung disease falls within a range documented in several previous studies. In a study of 100 patients with chronic obstructive pulmonary disease, Tramarin and colleagues obtained estimates by DE in only 30% (12). Bach and associates documented a success rate of 26% among 207 patients with advanced chronic obstructive pulmonary disease undergoing evaluation for lung volume reduction surgery (11). In contrast, Laaban and colleagues obtained estimates of sPAP in 66% of 41 patients with severe chronic obstructive pulmonary disease (10). Whether changes in chest wall configuration and cardiac orientation also limit DE estimation of sPAP in patients with interstitial lung diseases remains unclear. Although we documented a statistically significant correlation between DE-estimated sPAP and measured values by RHC, the correlation coefficient of 0.69 was well below that reported in studies involving cardiac patient populations. This suggests that there is greater variability between estimated and measured pressures in patients with lung disease, again likely due to factors specific to this patient population that limit accurate visualization and measurement of the velocity of the tricuspid regurgitant jet. Three previous studies involving patients with pulmonary disease have found similar correlation coefficients, in the range of 0.65–0.73 (10–12). One additional study documented a correlation coefficient of 0.92, but sPAP was determined by DE in only 24% of patients, suggesting that the patients in whom pressures were estimated may have been meticulously selected (13). In contrast to these previous studies that focused largely on definition of a correlation coefficient as a means of characterizing the variability between estimated and measured pressures, we have attempted to provide a more clinically relevant parameter by determining the frequency with which estimated sPAP fell within 10 mm Hg of the actual value. We found that DE provided an "accurate" (i.e., less than 10 mm Hg discrepancy) estimate of sPAP in only 48% of cases. Accuracy improved only modestly to 55% when RAP measured by RHC was substituted for estimated RAP in the calculation of sPAP, suggesting that errors in estimating RAP were only a minor source of the observed discrepancy. Of note, the current practice of estimating RAP by the pattern of inferior vena caval collapse resulted in overestimation of the actual RAP to the same degree as did the earlier practice of universally assuming an RAP of 15 mm Hg in all patients. We observed significant differences in the accuracy of DE in estimating sPAP among the disease subgroups. Accuracy was highest in the obstructive lung disease group compared with the groups with interstitial lung disease and pulmonary vascular disease, despite the significantly lower likelihood of obtaining an estimation of sPAP in obstructive lung disease. The reasons for this apparent paradox are not immediately clear. It is at first tempting to speculate that this simply reflects a greater likelihood of correctly estimating within 10 mm of actual pressure when pressures are low, as in the obstructive lung disease population, as opposed to when they are moderately or markedly elevated, as encountered in the interstitial lung disease and pulmonary vascular groups. However, our finding that accuracy was similar in those with and without PH refutes this argument. Alternatively, given the greater difficulties in visualizing the tricuspid regurgitant jet in patients with obstructive lung disease and hyperinflation, the interpreting cardiologists may be biased in favor of estimating pressures more selectively; restricting these estimates to only those patients in whom measurements of the regurgitant jet velocity could be made with a high level of confidence. A comparison of the diagnostic performance characteristics of DE for the two main disease subgroups, obstructive and interstitial lung disease, revealed several interesting trends. In neither disease group was the positive predictive value of DE parameters sufficiently high to permit definitive interpretation of a positive study. Similarly, the negative predictive value of DE studies was low in the interstitial lung disease group. In contrast, the negative predictive value of both estimated sPAP and right ventricular abnormalities exceeded 90% in the obstructive lung disease group. Given that the sensitivities of the DE studies were comparable in these two disease groups, the observed discrepancies in negative predictive values reflect marked differences in the prevalence of PH in these two populations. We documented a prevalence of PH of only 18% among patients in the obstructive lung disease group compared with 59% in the group with interstitial lung disease. Scharf and colleagues recently reported a similarly low prevalence of 13% among 120 patients with severe emphysema evaluated for participation in the National Emphysema Treatment Trial (17). Given the limitations of DE, what is its ultimate role in screening patients with pulmonary disease for the presence or absence of significant PH? Using a cutoff of 45 mm Hg as indicative of PH, estimated sPAP proved to be a relatively sensitive but nonspecific marker with a poor positive predictive value. Thus, an estimated sPAP in excess of 45 mm Hg must be interpreted with caution, as approximately half of the time it will represent a false-positive finding. In situations in which precise definition of the magnitude of PH is critical, such as in the assessment of suitability for lung volume reduction surgery, demonstration of an elevated sPAP by DE should prompt corroboration by RHC. The absence of an elevated sPAP by DE appears to reliably exclude significant PH and therefore to obviate the need for more invasive testing in the obstructive lung disease population but is less definitive when documented in the context of advanced interstitial lung disease. Assessment by DE of right ventricular abnormalities appears to be of greatest utility as an adjunctive means of excluding significant PH in patients with obstructive lung disease, particularly in situations in which sPAP cannot be estimated. One potential limitation of our study is that DE and RHC were not performed simultaneously but rather within a 3-day window. In a study by Currie and colleagues, 43 patients underwent RHC and both simultaneous and sequential DE, with an average of 3 days separating the two DE studies (7). Notably, the trans-tricuspid gradients were not significantly different between the two DE studies, but the correlation between DE and RHC pressures dropped slightly over time (r = 0.97 for simultaneous DE and 0.87 for delayed DE). This study suggests that separation of RHC and DE by up to 3 days has at best a minimal impact. Furthermore, all patients in our study were clinically stable and observed daily with no changes in any medications made during their evaluation period. In conclusion, estimation of sPAP by DE is frequently inaccurate (i.e., discrepant by at least 10 mm Hg) in patients with underlying advanced lung disease. In light of the poor positive predictive value of DE-estimated sPAP and of findings of right ventricular abnormalities in this population, reliance on this noninvasive technique can potentially lead to overdiagnosis of PH. When the exact magnitude of PH is of critical importance, such as in the assessment of patients for lung volume reduction surgery or lung transplantation, echocardiographic findings suggestive of PH should be confirmed by RHC.
These findings were presented in part at the American Thoracic Society Meeting, San Francisco, California, May 2001.
Supported by the Craig and Elaine Dobbin Pulmonary Research Fund. Received in original form October 3, 2002; accepted in final form December 8, 2002
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
