Arthritis & Rheumatism, Volume 62,
November 2010 Abstract Supplement
Abstracts of the American College of
Rheumatology/Association of Rheumatology Health Professionals
Annual Scientific Meeting
Atlanta, Georgia November 6-11, 2010.
A Formula To Predict Mean Pulmonary Artery Pressures in Patients with Connective Tissue Disease Based on Echocardiography, NTproBNP and O2 Saturation.
Schreiber3, Benjamin E., Valerio3, Christopher J., Handler3, Clive, Keir1, Greg, Wells1, Athol U., Denton2, Christopher P., Coghlan3, J. Gerry
Pulmonary Hypertension (PH) is an important complication of connective tissue disease (CTD). Diagnosis is confirmed by right heart catheter (RHC) but no single non-invasive test accurately predicts PH. To improve selection of patients for RHC, we analysed the relationship between non-invasive tests and mean pulmonary artery pressure (mPAP) at RHC and derived a new formula to predict PAP.
Retrospective analysis of patients with CTD undergoing a first RHC. In our database we have 968 patients (811 SSc, 53 SLE, 15 RA, 42 UCTD, 8 MCTD, 12 DM, 7 PM, 6 Sjogrens, 3 antiphospholipid syndrome, 3 vasculitis). We included for analysis NTproBNP and echocardiography if performed within 3 months of RHC and pulmonary function tests if done within 6 months. Pulse oximetry was measured at the time of the RHC.
Regression analysis of these variables individually against mPAP at RHC gave the following results: DLCO (n=469) gave R2=8.7, AUC=0.68; NTproBNP (capped at 300 pmol/litre to reduce the skewing effect of large values) with n=380, gave R2=31.5, AUC=0.74 and echo derived tricuspid valve gradient (n=165) gave R2=48.9, AUC=0.81. Capping NTproBNP gave similar results to a log transformation.
Multivariable linear regression showed significant correlation with mPAP for Echo derived TV gradient (p<0.0005), capped NTproBNP and Oxygen (p=0.004). Addition of predicted DLCO, KCO, FVC, weight or height did not improve the fit. To reduce heteroskedasticity, oxygen was used categorically (1 if SpO2 >94%, 2 if 9094%, 3 if <90%). Based on all 123 patients for whom we had NTproBNP, echo and SpO2 data, the derived formula is:
Predicted mPAP = 8.37
+3.83 × Oxygen category
+ 0.328 × Echo derived Tricuspid Valve gradient
+ 0.032 × NTproBNP (capped)
This formula had R2 of 60.3%.
The area under the curve was 0.84 (95% CI 0.770.91). Using a threshold predicted mPAP of 25 it has a sensitivity of 87.3%, specificity of 55.8%, positive LR of 2.0 and negative LR of 0.2. Using a threshold predicted mPAP of 30, it has a sensitivity of 66.2%, specificity of 90.4%, positive LR of 6.9, negative LR of 0.4.
Bland-Altman analysis showed a mean agreement of -0.6 (95% CI -16.4, 16.3) for difference between predicted and actual mPAP, although as seen in the figure, the 95% CI are tighter at lower predicted values.
This compares favourably with echocardiography done within 1 hour of the RHC (Fisher MR et al, Am J Respir Crit Care Med. 2009).
This formula may help identify CTD patients requiring RHC. Patients with a formula predicted mPAP under 25 are unlikely to have PH, and those patients with predicted mPAP above 30 are very likely to have PH.
To cite this abstract, please use the following information:
Schreiber, Benjamin E., Valerio, Christopher J., Handler, Clive, Keir, Greg, Wells, Athol U., Denton, Christopher P., et al; A Formula To Predict Mean Pulmonary Artery Pressures in Patients with Connective Tissue Disease Based on Echocardiography, NTproBNP and O2 Saturation. [abstract]. Arthritis Rheum 2010;62 Suppl 10 :727