To date, Doppler echocardiography is the most widespread and well-recognized technique for the noninvasive evaluation of systolic pulmonary artery pressure (sPAP). However, recent studies have reported reservations about the relevance of Doppler echocardiography or the tool’s reliability in the diagnosis and follow-up of patients with pulmonary hypertension (PH). Thus, the aim of this dedicated retrospective study was to address the questions of Doppler echocardiography’s relevance and accuracy for PH diagnosis in the routine activity of a conventional echocardiography department.
Institutional databases were used to extract and analyze the records of 310 patients who underwent both hemodynamic and echocardiographic investigations within a single hospitalization period.
Despite an underestimation of absolute Doppler sPAP values compared with measurements on right heart catheterization, data analysis revealed a strong correlation ( r = 0.80, P < .00001, n = 310). Targeting a mean pulmonary pressure on right heart catheterization of 25 mm Hg for the definition of PH, receiver operating characteristic curve analysis demonstrated a strong association between sPAP and PH diagnosis (area under the curve, 0.82; n = 155). The cutoff obtained for sPAP was 38 mm Hg, and when applied on a second-test subgroup population ( n = 155), sensitivity, specificity, and accuracy were 88%, 83%, and 86%, respectively. When patients with examination intervals of <2 days were selected ( n = 115), sensitivity and specificity reached 89% and 89%, respectively. No combination of parameters produced an improvement on the initial results.
In the real-world practice of a conventional echocardiography department, Doppler echocardiography is associated with high accuracy, sensitivity, and specificity for PH evaluation, thus confirming its major position as a primary noninvasive tool.
Doppler echocardiography is the most widespread and well-recognized technique for the noninvasive evaluation of systolic pulmonary artery pressure (sPAP). Since its validation by Berger et al. and Currie et al. in 1985, multiple experiences have been published confirming the Doppler method’s reliability, independently of cardiac disease. On the basis of publications reporting correlations with invasive measurement data, Doppler echocardiography is still recommended as the primary tool for early screening and assessment of patients with clinical suspicion of pulmonary hypertension (PH). The Doppler approach’s limitations have previously been highlighted, with some restrictions pertaining to patients with pulmonary disease. Other heart diseases, such as chronic heart failure, have been targeted as good models for Doppler evaluations of pulmonary pressures (PPs).
More recently, focus has been given to patients with PH. Pursuant to this, and after a large prospective study, Rich et al. found Doppler echocardiography an unreliable technique for diagnosis and follow-up of patients with PH. Such reservations have been similarly addressed in a meta-analysis, introducing doubt with regard to the relevance of Doppler echocardiography, despite its wide and routine use for noninvasive cardiac investigations.
Therefore, we conducted a retrospective study examining Doppler echocardiography’s performance in real-world practice in a large population from our echocardiography department. We aimed both to assess the reliability of Doppler echocardiography compared with the invasive and referent pulmonary blood pressure evaluation method and to test its accuracy in PH diagnosis, all in the context of a conventional echocardiography department’s routine practice.
Data was collected from the cardiac catheterization department, which performs approximately 4,500 right heart catheterization (RHC) procedures every year, and from the echocardiography department, which conducts >15,000 transthoracic Doppler echocardiographic studies every year. In the latter department, including seven rooms, operators with different profiles perform five to 15 examinations per half day.
The operators’ team consisted of two nurses (working as sonographers) equivalent to level 2 in the American Heart Association (AHA) and American College of Cardiology (ACC) guidelines and level 1 of our university echocardiographic regional diploma (UERD) program, three cardiology fellows (level 2 in the AHA and ACC guidelines, level 2 of the UERD program), and six senior cardiologists (level 3 in the AHA and ACC guidelines, level 2 of the UERD program). The UERD program is a 2-year teaching course consisting of an initial 155 hours of theory and 120 transthoracic echocardiographic studies performed and interpreted in the course of the first year (level 1) and an additional 155 hours of theory, 120 transthoracic echocardiographic studies, and 60 transesophageal echocardiographic studies performed and interpreted in the course of the second year (level 2). All examinations performed by nurses or fellows were considered secondary and underwent systematic validation by one of the senior cardiologists.
All echocardiographic measurements were performed according to American Society of Echocardiography and European Association of Echocardiography recommendations, in line with UERD standard operating procedure. Accordingly, tricuspid regurgitant maximum velocities were measured on the basis of adequate acquisitions, and estimated sPAP levels were determined using the modified Bernoulli equation when tricuspid regurgitant jets were analyzable, in conjunction with echocardiographic estimation of right atrial pressure (RAP) ( Figure 1 ). No contrast methods were used. In case of atrial fibrillation, five to 10 beats were used for averaging velocities. Echocardiographic RAP estimation was performed on the basis of inferior vena cava (IVC) size and collapsibility. RAP was estimated to be 3 mm Hg when the IVC diameter was <21 mm with >50% collapsibility, 8 mm Hg when the IVC diameter was <21 mm with <50% collapsibility, and 15 mm Hg when the IVC diameter was >21 mm with <50% collapsibility.
Senior physicians specializing in invasive techniques performed the catheterizations according to standard procedures. Fluid-filled catheters were used for patients not sedated. Systolic, diastolic, and mean PPs were averaged and calculated over five beats. Systolic RHC PP was used for direct comparison with sPAP using echocardiography and mean PP for the diagnosis definition of PH using catheterization.
Our institution’s databases were used for this retrospective analysis. To this end, measurement data from the RHC and echocardiography departments were automatically transferred to the Structured Query Language database, which was copied into the central institutional database. Only new records from the site were automatically added to this central database every night. Patient records, including diagnoses and treatments, laboratory results, and hemodynamic and echocardiographic investigations, were extracted from this central database using dedicated software for database queries (SAP Business Object Enterprise XI version 12.3.6, version 601; SAP AG, Walldorf, Germany).
The inclusion criterion for patient enrollment was having undergone both hemodynamic and echocardiographic investigations (with estimation of PP by Doppler on tricuspid regurgitant flow) between June 2011 and March 2012, irrespective of the causal disease, during a single hospitalization period. The exclusion criterion was a lack of PP estimation on the basis of tricuspid regurgitant flow using Doppler echocardiography.
The local ethics committee approved the study. All patients hospitalized at our institution were informed that their personal medical data might be used for research purposes.
Continuous variables are expressed as mean ± SD and categorical variables as percentages and numbers of patients. Chi-square tests or Fisher’s exact tests were performed for qualitative variable comparison. Two-tailed paired t tests or nonparametric Mann-Whitney U tests were used for comparisons of quantitative data from Doppler echocardiographic and RHC examinations.
Systolic PP comparisons between RHC and Doppler echocardiography were performed on the entire selected population, using paired two-tailed t tests or nonparametric Mann-Whitney U tests, linear regression with coefficient correlation calculation, and Bland-Altman analysis.
Potential confounding parameters that could influence PP estimation, such as time between echocardiography and catheterization or operator’s skill, were sought by multivariate linear regression analysis.
Receiver operating characteristic curve analyses were performed on a randomly selected subgroup of 155 patients for different echographic parameters, including the maximum velocity of tricuspid regurgitation, PP, acceleration time on pulmonary flow, and right chamber and IVC dimensions, to determine cutoffs for PH diagnosis (defined as a pulmonary mean pressure >25 mm Hg on catheterization ). Parameter cutoffs were then applied to the second subgroup of 155 patients to determine the tests’ sensitivity, specificity, and accuracy. Finally, an algorithm combining selected parameters was tested to optimize sensitivity and specificity levels for PH diagnosis.
The selected population consisted of 310 patients, 162 of whom were men, with a mean age of 64.8 ± 15.9 and a mean body mass index of 26.3 ± 5.7 kg/m 2 ( Table 1 ).
|Age (y)||64.8 ± 15.9|
|Weight (kg)||73.0 ± 18.4|
|Height (cm)||166.1 ± 10.8|
|BMI (kg/m 2 )||26.3 ± 5.7|
|HR (beats/min)||74.6 ± 15. 7|
|SBP (mm Hg)||118.7 ± 36.0|
|DBP (mm Hg)||66.0 ± 20.6|
|Systolic PP on RHC (mm Hg)||53.2 ± 25.7|
|Mean PP on RHC (mm Hg)||32.7 ± 15.3|
|sPAP on echocardiography (mm Hg)||49.6 ± 21.7|
|Delay (d)||2.0 ± 2.9|
|Grade 1||12.9% (40)|
|Grade 2||3.9% (12)|
|Grade 3||83.2% (258)|
|Heart failure||18% (55)|
|Ischemic cardiomyopathy||16% (50)|
|Respiratory failure||15% (46)|
|PH grade I||12% (37)|
|PH grade II||12% (37)|
|Dilated cardiomyopathy||11% (35)|
In total, 136 patients underwent echocardiographic investigation specifically for cardiac evaluation (27% for valvulopathy, 18% for heart failure, 16% for ischemic cardiomyopathy, and 11% for dilated cardiomyopathy), and 80 did so for pulmonary evaluation (15% for respiratory failure, 12% for primary PH, and 12% for secondary PH). Two-dimensional ultrasound window quality (from the parasternal or apical views) was good in 61% of examinations, medium in 30%, and poor in 8% (defined by the failure to visualize major heart structures). Three different operators’ skills were identified: 13% of examinations were performed by nurses, 4% by fellows, and 83% by senior physicians. The mean time between the two examinations was 2.0 ± 2.9 days.
Systolic and diastolic blood pressures from echocardiography and catheterization measurements were not significantly different (122 ± 20 vs 122 ± 19 mm Hg, P = .14, and 67 ± 12 vs 68 ± 13 mm Hg, P = .42, respectively). Other parameters are listed in Table 1 .
Systolic PP Correlation
Systolic PP from RHC was significantly higher than that from Doppler echocardiography (53.2 ± 25.7 vs 49.7 ± 21.9 mm Hg, t -test P < .01). However, a significant, strong, and positive correlation was found between the two measurements ( r = 0.80, P < .00001; Figure 2 A). The mean difference was 3.6 ± 15.3 mm Hg, with 95% limits of agreement ranging from −26 to 33.4 mm Hg ( Figure 2 B). Graphically, the highest differences were found for high pathologic systolic PP values, with an underestimation of echocardiographic sPAP values compared with the referent RHC measurements. When focusing analysis on PP values <60 mm Hg (159 values), the mean error measurement was reduced to −2.09 mm Hg, with 95% limits of agreement ranging from −22.5 to 18.3 mm Hg.
The estimated mean RAP on echocardiography was 8.25 ± 4.23 mm Hg, compared with 7.39 ± 5.27 mm Hg on RHC ( t -test P < .01). A significant but modest correlation was found ( r = 0.43, P < .001).
Analysis of Confounding Parameters
To evaluate the potential impact of parameters such as the interval between the two examinations or operators’ training, a multivariate linear regression analysis was performed, using as the dependent variable the absolute difference of systolic PPs obtained by the two methods. Independent variables were body mass index, heart rate, systolic blood pressure, time interval between the two examinations, operator’s skill, and sonographic window quality. As shown in Table 2 , only the time interval was shown to be positively associated with the absolute difference of systolic PPs, without reaching statistical significance.
|Independent Variable||β ± SD||t||P|
|Constant||13.25 ± 19.96||0.664||.51|
|BMI||0.003 ± 0.066||0.045||.96|
|HR||−0.047 ± 0.060||−0.774||.44|
|SBP||0.0006 ± 0.026||0.023||.98|
|Delay||0.574 ± 0.313||1.831||.07|
|Skill||0.974 ± 1.365||0.714||.48|
|Ultrasound window||2.588 ± 1.571||1.648||.10|
Echocardiographic PP for PH Diagnosis
Targeting a mean RHC PP limit value of 25 mm Hg for PH definition, receiver operating characteristic curve analysis revealed a strong association between echocardiographic sPAP and PH diagnosis (area under the curve, 0.82). The cutoff obtained for echocardiographic sPAP with the highest sensitivity and specificity was 38 mm Hg ( Figure 3 ). When applied to the second subgroup population, sensitivity, specificity, and accuracy were 88%, 83%, and 86%, respectively. Positive predictive value was 91% and negative predictive value 78.2%.
When patients with examination intervals of <2 days were selected ( n = 115), there were 69 true-positives, 32 true-negatives, nine false-negatives, and four false-positives. Sensitivity was 89%, specificity 89%, positive predictive value 94%, and negative predictive value 78%. Accuracy (patients classified correctly) was 89%. Characteristics and a comparison of misclassified patients are presented in Table 3 .
|Variable||Correctly classified ( n = 134)||Misclassified ( n = 21)||P||False-positive ( n = 9)||False-negative ( n = 12)||P|
|Age (y)||65.8 ± 13.7||63.7 ± 18.8||NS||66.7 ± 22.3||61.4 ± 16.3||.24|
|Men||50.0% (67)||57.1% (12)||NS||33.3% (3)||75.0% (9)||.09|
|Weight (kg)||74.2 ± 19.3||66.9 ± 17.26||.09||58.6 ± 16.0||73.1 ± 16.0||.06|
|Height (cm)||165.8 ± 9.2||165.3 ± 10.1||NS||160.2 ± 8.8||169.2 ± 9.5||.05|
|BMI (kg/m 2 )||26.9 ± 5.8||24.2 ± 4.7||.03||22.7 ± 5.6||25.3 ± 3.7||.20|
|HR (beats/min)||73.6 ± 13.5||73.9 ± 14.6||NS||76.5 ± 10.5||72.0 ± 17.3||.20|
|SBP (mm Hg)||125.8 ± 27.2||125.8 ± 22.8||NS||130.6 ± 30.6||122.7 ± 17.3||.44|
|DBP (mm Hg)||69.2 ± 14.7||71.3 ± 17.9||NS||69.1 ± 15.8||72.6 ± 19.7||NS|
|Delay (d)||1.9 ± 3.0||2.5 ± 3.7||NS||2.1 ± 1.6||2.7 ± 4.8||.2|
|Grade 1||13% (18)||10% (2)||NS||22% (2)||0% (0)||.17|
|Grade 2||4% (6)||0% (0)||NS||0% (0)||0% (0)||—|
|Grade 3||82% (110)||90% (19)||NS||78% (7)||100% (12)||.17|
|Good||57% (77)||67% (14)||NS||56% (5)||75% (9)||.40|
|Medium||31% (42)||30% (6)||NS||30% (4)||17% (2)||.33|
|Poor||11% (15)||5% (1)||NS||0% (0)||8% (1)||NS|
|Valvulopathy||24% (32)||43% (9)||.07||33% (3)||50% (6)||NS|
|Heart failure||16% (22)||19% (4)||NS||11% (1)||25% (3)||NS|
|Ischemic cardiomyopathy||12% (16)||24% (5)||.14||22% (2)||25% (3)||NS|
|Respiratory failure||19% (25)||10% (2)||.30||22% (2)||0% (0)||.17|
|PH grade I||20% (27)||10% (2)||.25||10% (1)||8% (1)||NS|
|PH grade II||14% (19)||5% (1)||.23||0% (0)||8% (1)||NS|
|Dilated cardiomyopathy||7% (9)||10% (2)||NS||0% (0)||17% (2)||.49|