CT pulmonary angiography combined with echocardiography in suspected systemic sclerosis-associated pulmonary arterial hypertension

Abstract

Objective. Echocardiography is widely used in the investigation of patients with suspected SSc-associated pulmonary arterial hypertension (SSc-PAH). We investigated whether CT pulmonary angiography (CTPA) provides additive diagnostic value.

Methods. Data for 89 consecutive patients with suspected SSc-PAH undergoing echocardiography, CTPA and right heart catheterization were reviewed. Pulmonary artery diameter (dPA) and ascending aorta diameter (dAA), right and left ventricular diameter (dRV and dLV) and grade of tricuspid regurgitation (TR_CT) measured at CTPA and tricuspid gradient (TG_ECHO) at echocardiography were retrieved. A predictive equation for mean pulmonary arterial pressure (mPAP) was derived using multivariate linear regression. Multivariate Cox regression analysis was then used to assess the prognostic strength of CTPA parameters and TG_ECHO.

Results. Absolute measures of dPA and dRV correlated weakly with mPAP. However, dPA : dAA and dRV : dLV showed stronger correlations with mPAP (dPA : dAA r = 0.42, P < 0.001; dRV : dLV r = 0.51, P < 0.001). dRV : dLV correlated more strongly with pulmonary vascular resistance than did dPA : dAA (r = 0.63 vs 0.39, P both <0.001). dPA : dAA and TG_ECHO were independent predictors of mPAP. A derived CT/echo composite index had a higher predictive accuracy (area under the curve = 0.95) than dPA : dAA or TG_ECHO although negative predictive value (NPV) was only 77%. Combining the CT/echo composite index with presence or absence of TR_CT increased NPV to 100% although this observation requires further validation. dRV : dLV was the strongest prognostic factor.

Conclusion. In suspected SSc-PAH, cardiac chamber and great vessel measurements at CTPA correlate with pulmonary haemodynamics and predict survival. In combination with echocardiography CTPA increases diagnostic accuracy and may identify other potential causes of breathlessness.

Pulmonary arterial hypertension, Systemic sclerosis, Computed tomography pulmonary angiography, Echocardiography, Right ventricular diameter, Pulmonary artery diameter

Introduction

Pulmonary arterial hypertension (PAH), a condition in which increased pulmonary vascular resistance (PVR) leads to an elevation in mean pulmonary arterial pressure (mPAP), has been shown to complicate 7.85–12% of patients with SSc-PAH [1, 2]. Median survival in historical series was <1 year and although survival has improved, SSc-PAH is still associated with a poor prognosis [3, 4]. Diagnosis relies on invasive measurement of pulmonary haemodynamics at right heart catheterization (RHC). Although echocardiography is widely used as an initial screening and diagnostic tool, it is not accurate enough to confidently exclude the presence of PAH as a single test [5].

CT is widely used as an integral part of diagnostic evaluation of patients with suspected SSc-PAH. Not only does it allow demonstration of associated interstitial lung disease (ILD) but CT pulmonary angiography (CTPA) also permits exclusion of significant thromboembolic disease. Several other features reflecting pulmonary hypertension (PH) can also be measured at CTPA. The main pulmonary artery diameter (dPA) has been found to correlate inconsistently with mPAP [6–11] although this measurement may be less reliable in the presence of ILD [11]. A composite index of the ratio of dPA and diameter of the adjacent aorta (dAA) together with tricuspid gradient measured at echocardiography (TG_ECHO) has recently been demonstrated to have greater predictive value of mPAP in a heterogeneous group of patients with suspected PH [12]. The extent of tricuspid regurgitation seen at CTPA (TR_CT) has also been shown to correlate with mPAP, while right ventricular (RV) size measured by MRI in PAH and by non-gated CT in acute pulmonary embolism has been shown to be of prognostic significance [13, 14]. The importance of RV size as assessed by non-gated CT in the assessment of suspected PAH is, however, unknown. The aim of this study was therefore to assess the relationship of parameters measurable at CTPA with pulmonary haemodynamics in patients with suspected SSc-PAH. We also investigated whether these parameters could be used to increase the diagnostic strength of the currently used screening tool, namely TG_ECHO.

Patients and methods

CT scans for 89 consecutive patients with suspected SSc-PAH who underwent both CTPA and RHC between July 2001 and February 2006 in our institution were analysed. RHC was performed via the internal jugular vein using a Swan–Ganz catheter with cardiac output measurement by the thermodilution method. SSc had been diagnosed by the referring rheumatologist in the majority of cases. If a new diagnosis of SSc was made in our unit, this was confirmed by a rheumatologist according to standard criteria [15, 16]. PH was defined as mPAP at RHC (mPAP_RHC) ≥25 mmHg with PAH being diagnosed when the pulmonary capillary wedge pressure (PCWP) was ≤15 mmHg. Correlations of both mPAP_RHC and PVR with CT parameters were assessed in 81 patients who had undergone CTPA within 3 months of RHC (Group A) and with TG_ECHO in 70 patients who had undergone echocardiography within 3 months of RHC (Group B). Further analysis was then performed on 67 patients who had both CTPA and echocardiography within 3 months of RHC (Group C). Further analysis was also performed according to the presence of associated significant ILD [defined as moderate or severe fibrosis (equivalent to >20% lung involvement) or forced vital capacity (FVC) <70% predicted in the absence of significant airflow obstruction, in a similar manner to that described elsewhere] [17]. Approval for retrospective analysis of imaging techniques had been granted by the North Sheffield ethics committee.

CT acquisition and analysis

CTPA during the period of image acquisition was performed using single slice scanners (CTi Hispeed and Hi speed Fxi, GE, CT, USA). A 100 ml bolus of i.v. contrast (Optiray; Mallinckrodt Imaging, Ireland during 2001 and Ultravist 300; Bayer Schering, Germany during 2002–06) was given via an antecubital vein at a rate of 3–4 ml/s. Standard acquisition parameters of 250 mA, 120 kVp, 1.4 pitch, 3 mm slice thickness and 0.5 s rotation time were used, scanning was in a caudo-cranial direction. Bolus tracking was used to trigger contrast infusion. dPA and dAA were assessed by a single observer (M.R.), blinded to haemodynamic and echocardiographic findings, by identifying the widest dPA at the level of the bifurcation of the main pulmonary artery and then measuring the adjacent ascending aorta. TR_CT was assessed using a simplification of a previously described technique using four degrees of regurgitation: none [no reflux into inferior vena cava (IVC)], mild (reflux into IVC but not hepatic veins), moderate (reflux into proximal hepatic veins) and severe (reflux into distal hepatic veins) [18]. The maximum diameters of the RV (dRV) and left ventricle (dLV) were measured in the transverse plane as described previously [14].

Echocardiography

Echocardiography was performed using Powervision 6000 and 8000 machines (Toshiba, Japan). Multiple windows were used to obtain the optimal Doppler estimation of TR velocity (V). The TG_ECHO was then calculated using the modified Bernoulli equation (TG_ECHO = 4V²).

Statistical analysis

Patient data are expressed using mean (s.d.) or median (range). Characteristics of patients with and without PH were compared using the independent t-test, Mann–Whitney and χ² tests as appropriate. Correlation coefficients were calculated using Pearson's and Spearman's tests. Multivariate forward stepwise linear regression analysis was performed to assess the relationship of mPAP with CT parameters and TG_ECHO. Diagnostic strength was assessed using receiver operated characteristic (ROC) curve analysis. Prognostic importance was assessed using multiple forward stepwise Cox regression analysis and Kaplan–Meier analysis. The CTPA or echocardiographic parameter that independently predicted survival was then compared with other previously identified prognostic indices [4]. P < 0.05 was considered statistically significant. Analysis was performed using SPSS (v13) statistical package.

Results

Patient characteristics

Eighty-nine patients were evaluated and baseline demographics are shown in Table 1. Median duration of follow-up to death or census was 3.64 (range 0.02–8.3) years. Median time between RHC and CT was 2 (range –524 to 1329) days with only 8 (9%) patients having an interval of >3 months. Sixty-one (69%) patients were diagnosed with PH (55 PAH and 6 PH in association with left heart disease). Nineteen (21%) patients had associated significant ILD. Eighty (90%) patients had been diagnosed with lcSSc and 9 (10%) with diffuse SSc. Autoantibodies were available for 77 patients; 72 (94%) were ANA positive, 51 (66%) had ACAs and 9 (12%) were positive for Scl-70 antibodies. Median duration of SSc before RHC was 6 (range 0–3) years.

Table 1

Baseline demographics

Variable	Whole group (n = 89)	No PH (n = 28)	PH (n = 61)	P-value
Age, years	66 (10)	65 (10)	67 (10)	0.58
Gender: female, %	85	82	87	0.56
WHO II/III/IV, %	26 : 61 : 14	60 : 40 : 10	10 : 70 : 20	<0.001
Limited : diffuse, %	90 : 10	89 : 11	90 : 10	0.90
FVC, % predicted	97 (22)	100 (20)	95 (23)	0.34
TLCO, % predicted	45 (16)	57 (15)	40 (13)	<0.001
mRAP, mmHg	6 (5)	4 (2)	7 (5)	<0.001
mPAPRHC, mmHg	33 (14)	19 (4)	40 (12)	<0.001
PCWP, mmHg	9 (5)	8 (3)	9 (5)	0.67
CI, l/min/m2	2.9 (0.8)	3.2 (0.5)	2.8 (0.9)	0.02
PVR, median (range), dyn s cm−5	336 (83–1709)	153 (83–274)	560 (126–1709)	<0.001
mVO2, %	68 (8)	73 (4)	66 (8)	<0.001

Data are expressed as mean (s.d.) unless otherwise noted. WHO: world health organization functional class; TL_CO: gas transfer; mRAP: mean right atrial pressure; CI: cardiac index.

Correlation with pulmonary haemodynamics

Relatively weak correlations existed between both dPA and dRV and mPAP_RHC; these correlations strengthened when corrections for dAA and dLV, respectively, were made (dPA : dAA r = 0.42, P < 0.001; dRV : dLV r = 0.51, P < 0.001; Table 2). dRV : dLV correlated more strongly with PVR than did dPA : dAA (r = 0.63 vs r = 0.39, P both <0.001). TR_CT grade also correlated moderately with both mPAP_RHC and PVR (Table 2). The diameter of the left atrium, measured at CTPA, correlated moderately strongly with PCWP (r = 0.456, P < 0.001).

Table 2

CT and haemodynamic comparisons

	Group A (n = 81)		Group A, no ILD (n = 63)		Group B (n = 70)
Variable	r	P-value	r	P-value	r	P-value
dPA
mPAPRHC	0.345	0.002	0.567	<0.001
PVR	0.255	0.022	0.441	<0.001
dPA : dAA
mPAPRHC	0.420	<0.001	0.615	<0.001
PVR	0.392	<0.001	0.589	<0.001
dRV
mPAPRHC	0.332	0.002	0.298	0.018
PVR	0.334	0.002	0.309	0.015
dRV : dLV
mPAPRHC	0.510	<0.001	0.521	<0.001
PVR	0.630	<0.001	0.688	<0.001
TRCT
mPAPRHC	0.445	<0.001	0.439	0.001
PVR	0.491	<0.001	0.487	<0.001
TGECHO
mPAPRHC					0.640	<0.001
PVR					0.756	<0.001

Group A: RHC within 3 months of CTPA; Group B: RHC within 3 months of echocardiography.

Multivariate regression analysis using three variables (TG_ECHO, dPA : dAA, dRV : dLV) in the 67 patients in whom both CTPA and echo were performed within 3 months of RHC (Group C) identified TG_ECHO and dPA : dAA as being independently related to mPAP_RHC. The regression equation produced was:

This disease-specific CT/echo composite index correlated moderately strongly with mPAP_RHC (r = 0.72, r² = 0.52; P < 0.001, Fig. 1) with stronger correlation than its constituent elements (TG_ECHOvs mPAP_RHCr = 0.65, r² = 0.43; P < 0.001 and dPA : dAA vs mPAP_RHCr = 0.57, r² = 0.33; P < 0.001). The area under the curve (AUC) for the diagnosis of PH by the CT/echo composite index was 0.95, which compared well with the AUC for dPA (0.86), dPA : dAA (0.91), dRV (0.69), dRV : dLV (0.72) and TG_ECHO (0.89). ROCs identified a CT/echo composite index of 28 as the optimal threshold for diagnosing mPAP_RHC [sensitivity 89%, specificity 89%, positive predictive value (PPV) 95% and negative predictive value (NPV) 77%, Table 3]. To evaluate whether the NPV could be increased, the additional information provided by the presence (mild, moderate or severe) or absence of TR was then assessed. A threshold of CT/echo composite index ≥28 or at least mild TR produced a sensitivity of 100%, specificity of 63%, PPV of 87% and NPV of 100%.

Fig. 1

Correlation of CT/echo composite index and mPAP in Group C (n = 67).

Table 3

Sensitivity, specificity, PPV and NPV for diagnosis of PH in Group C ( n = 67)

Variable	Sensitivity	Specificity	PPV	NPV
dPA : dAA <1 vs ≥1	80	89	95	61
dRV : dLV < 1 vs ≥1	80	35	73	39
TRCT present vs absent	70	71	86	46
TGECHO <30 vs ≥30	96	40	79	80
TGECHO <35 vs ≥35	89	50	81	67
TGECHO <40 vs ≥40	79	80	90	62
CT/echo composite index <28 vs ≥28	89	89	95	77
CT/echo composite index <28 vs ≥28 or TRCT present	100	63	87	100
CT composite index <30 vs ≥30	89	79	91	75
CT composite index <30 vs ≥30 or TRCT present	98	58	85	92

Group C: RHC within 3 months of CTPA and echocardiography.

As accurate estimation of TG_ECHO is not possible in a proportion of patients with suspected SSc-PAH, the diagnostic utility of CTPA parameters alone was assessed by performing multivariate linear regression using dPA : dAA and RV : LV. The regression equation for this CT composite index was:

ROC analysis identified a CT index of 30 (with an AUC of 0.91) as the optimal threshold for diagnosis of PH (sensitivity 89%, specificity 79%, PPV 91% and NPV 75%, Table 3). A threshold of CT composite index ≥30 or at least mild TR resulted in a sensitivity of 98%, specificity of 58%, PPV of 85% and NPV of 92%.

Survival analysis

Univariate Cox regression analysis identified CT parameters (dPA : dAA, dRV : dLV, CT/echo composite index, TR_CT and TG_ECHO) as predictors of survival from date of CT or echocardiography (Table 4). Multivariate analysis (assuming an identical echo and CT date for the purpose of this calculation alone) identified dRV : dLV as the strongest prognostic factor. Three-year survival in patients with an dRV : dLV ≥ 1.17 (median value) was 40% compared with 75% in patients with a dRV : dLV < 1.17 (P < 0.001; Fig. 2). In the UKCTD associated PAH Registry (to date the largest registry of patients with SSc-PAH), four variables were found to independently predict poorer survival (higher age, male gender, higher WHO functional class, lower mixed venous oxygen saturation—mVO₂) [4]. When the prognostic strength of age, gender and WHO functional class together with dRV : dLV (substituted for mVO₂) was assessed in Group C of the present study both WHO functional class and dRV : dLV independently predicted outcome. When mVO₂ was added to the model, however, dRV : dLV lost its predictive significance. A similar effect was observed if other pulmonary haemodynamic measurements such as PVR were inserted into the model, rather than mVO_2, suggesting that although dRV : dLV has prognostic importance this is not independent of the pulmonary haemodynamic abnormalities it reflects.

Fig. 2

Survival from date of CTPA according to dRV : dLV above or below median (1.17).

Table 4

Survival analysis

CT parameter	Univariate HR	P-value	Multivariate HR	P-value
dPA : dAA	5.86 (1.06, 32.4)	0.043
dRV : dLV	3.11 (1.76, 5.44)	<0.001	2.85 (1.60, 5.07)	<0.001
CT/echo composite index	1.04 (1.02, 1.07)	0.001
TRCT : none (ref)
TRCT : any	2.38 (1.25, 4.53)	0.008
TGECHO	1.016 (1.01, 1.02)	<0.001

Group C: RHC within 3 months of CTPA and echocardiography. Ratios are presented with 95% CI. HR: hazards ratio.

Discussion

A reliable non-invasive method of identifying patients likely to have SSc-PAH, and therefore requiring RHC, is needed. In an effort to diagnose PAH earlier, screening of patients with SSc is now an established part of routine clinical practice. The most commonly used modality is estimation of TG_ECHO at transthoracic echocardiography. In the largest study of the utility of this method Mukerjee et al. [5] observed that although there was a strong PPV for the presence of PH using a threshold of 40 mmHg, the NPV using a variety of thresholds remained poor. In the present study, we likewise observed that although a TG threshold of 40 mmHg provided a reasonable PPV, no threshold could be identified that produced a high enough NPV for the adequate exclusion of PH.

Previous studies have demonstrated that although dPA and dPA : dAA correlate with mPAP_RHC, these measurements alone are not sufficient to confidently exclude PH [9, 10]. In a recent paper, Devaraj et al. [12] developed a composite index for predicting mPAP in an extremely heterogeneous group of patients, with the majority having suspected respiratory disease-associated PH. They found that dPA : dAA provided additional diagnostic information to TG_ECHO alone, but this index alone did not identify all patients with PH. As the present article involved a single underlying disease, we elected to formulate a disease-specific composite index and assessed which CT parameters measured using non-gated CTPA provided additional diagnostic information to using TG_ECHO alone in the assessment of suspected SSc-PAH. Although the CT/echo composite index had greater diagnostic strength than its constituent elements it could not completely exclude the diagnosis of PH. In the population under investigation, the incorporation of the presence or absence of TR_CT did, however, identify all cases of PH. Although this suggests that the combination of a CT/echo composite index with the presence or absence of TG_ECHO may provide an important tool for the non-invasive exclusion of SSc-PAH, it must be noted that several patients with a CT/echo composite index >28 did not in fact have any TR_CT. The possibility that the 10 patients with a CT/echo composite index <28 who had PH all had TR_CT by chance cannot be discounted.

Adequate estimation of TG_ECHO is not possible in up to 34% of patients with suspected SSc-PAH [19]. We therefore also assessed whether a CT composite index could be used to confidently exclude PH without the use of TG_ECHO. A CT composite index incorporating dPA : dAA and dRV : dLV had a reasonable sensitivity although the NPV was relatively poor. The incorporation of TR_CT again significantly improved the diagnostic utility with a sensitivity of 98% and NPV of 92%, although the specificity of this technique was poorer at 58%.

Although increased RV volumes measured using non-gated CTPA have previously been shown to be predictive of a poorer outcome following an acute pulmonary embolism [14], the information provided by this measurement in chronic PH has not been reported. The measurement of cardiac chamber size and ejection fraction using cardiac-gated CT has been well described [20]. Furthermore, the prognostic significance of RV volumes, in patients with heterogeneous forms of PH, and of MRI-derived RV and LV masses in suspected SSc-PAH has been previously demonstrated [13, 21]. These are, however, specialized imaging modalities not uniformly available in all hospitals, whereas non-gated CTPA is a widely available technique. dRV : dLV correlated moderately well with mPAP, although multivariate regression analysis identified that TG and dPA : dAA were the strongest predictors of mPAP_RHC. dRV : dLV was, however, the strongest prognostic marker. This may be explained by the observations that of the CT parameters assessed, dRV : dLV correlated most strongly with PVR and that PVR has been found elsewhere to be a better prognostic factor than mPAP_RHC in patients with PAH [13, 22].

The present article has also been able to validate various other previously reported associations between CT parameters and pulmonary haemodynamics for the first time in suspected SSc-PAH. Groves et al. [18] reported moderate correlations between TR_CT grade and mPAP_RHC in patients with chronic thromboembolic pulmonary hypertension. In the current article, we also observed significant correlations between TR_CT and mPAP_RHC and PVR. Devaraj et al. [11] previously reported that dPA : dAA correlated more strongly with mPAP_RHC in patients without diffuse pulmonary fibrosis. We also observed that the exclusion of patients with significant ILD improved the correlation of mPAP_RHC with dPA : dAA, although the composite index still performed well in the group as a whole.

There are limitations to the current article. Many patients did not have echocardiography and CT performed contemporaneously with RHC, and even when a threshold of 3 months from investigation to RHC was used, a proportion of patients were excluded from further analysis. This threshold of 3 months has been adopted in other similar studies and would seem to be a reasonable compromise between reliability of calculated relationships between parameters and adequate patient numbers [5]. As the studied cohort comprised patients referred for further investigation of suspected SSc-PAH, they may not be completely representative of a population of SSc patients within a general rheumatology setting. CTPA images for the present study were obtained on a single-slice scanner as opposed to the 64-slice (or higher) scanners currently used. Although the difference in image resolution is unlikely to be significant for the parameters assessed in this study, it is possible that the slightly higher contrast infusion rate (5 ml/s) and cranio-caudal scanning direction used with more modern scanners may increase the concentration of contrast in the IVC leading to a new gradation for regurgitation.

What clinical lessons can be taken from our findings? First, PH cannot be confidently predicted or excluded on the basis of a quick eyeball of CT findings, specifically the absolute or relative sizes of the pulmonary artery and RV. Although a dPA : dAA ≥ 1 has good specificity and PPV for a diagnosis of PH, a dPA : dAA < 1 cannot be used to confidently exclude PH. dRV fares even less well, with dRV : dLV ≥1 or <1 having poor diagnostic discriminatory power, although survival is significantly superior in patients with dRV : dLV < 1. Second, as previously reported elsewhere, no echocardiographic parameter threshold such as TG_ECHO exists that can confidently exclude PH in patients with suspected SSc-PAH [5, 21]. Third, a disease-specific CT/echo composite index has better discriminatory power than its constituent elements. In the present cohort combination of this index with the presence or absence of TR_CT allowed PH to be confidently excluded. Although this observation requires validation in a further cohort of patients, an assessment of the breathless SSc patient using both echocardiography and CTPA provides a potentially attractive non-invasive method of excluding other important causes of dyspnoea (e.g. pulmonary emboli or significant ILD), while also discriminating those patients with a high or low chance of having PH. It is important to emphasize that due to the high pre-test probability of PH in patients with SSc, as compared with the population as a whole, SSc patients with significant dyspnoea that is unexplained by imaging and spirometry, should proceed to RHC even if echocardiographic and composite CT/echo parameters do not suggest that PH is likely.

In conclusion, the present study has demonstrated that cardiac chamber and great vessel measurements at CTPA correlate with pulmonary haemodynamic measurements obtained at RHC and have additive value to echocardiography in the assessment of patients with suspected SSc-PAH. This article also reports for the first time the prognostic importance of dRV : dLV measured at non-gated CTPA in patients with suspected pulmonary hypertension.

Acknowledgements

R.C., C.A.E. and D.G.K. are co-investigators in the National Institute for Health Research (NIHR) Sheffield Cardiovascular Biomedical Research Unit.

Disclosure statement: The authors have declared no conflicts of interest.

Author notes

*Robin Condliffe and Mark Radon contributed equally to this work.