Page 36 - Assessing right ventricular function and the pulmonary circulation in pulmonary hypertension Onno Anthonius Spruijt

2
Earlier studies mainly focused on PA/AAAX to predict PH and showed that a PA/AAAX>1 has a sensitivity and specificity ranging from 58 to 87% and 73 to 95%, respectively *14, 15, 18-20+. This is in line with our results (PA/AAAX>1: sensitivity 75% and specificity 92%).
Multivariable binary logistic regression analyses and the significantly higher AUCs of model 2 and 3 compared to model 1, showed that there is a statistically significant improvement of the prediction model when ventricular and pulmonary artery measurements are combined. DCA confirmed the clinical relevance of this approach. Arguing that, missing the diagnosis is worse than performing unnecessary diagnostic tests, we assigned a higher weight to false negatives than to false positives and focused on a range of low threshold probabilities. We showed that, even at this range of low threshold probabilities, in comparison to model 1, models 2 and 3 allowed a decrease in number of false positives without an increase in the number of false negatives. As such, adding ventricular measurements to pulmonary artery measurements statistically improves the prediction model with clinical relevance.
We are aware of only one other study investigating ventricular measurements on CTPA to predict PH. Chan et al measured the RV/LV ratio in the axial view and found that a RV/LV>1.28 predicted PH with a sensitivity of 85.7% and 86.1% *16+. There are no studies that used a combination of ventricular and pulmonary measurements to improve the predictive value of CTPA.
Manual reconstructed 4CH-views for determining ventricular diameters on standard CTPA have not been previously used in radiological studies of PH. In studies of patients of acute PE, some investigators indicated that the RV/LV determined in the 4 chamber view provided superior prediction of subsequent adverse events than the same ratio measured in the axial view, although other studies didn’t find any differences *18, 19, 23+.
In this study, ROC analyses showed no significant difference between model 2 and 3 (p=0.266) and also the net benefits determined with DCA were almost equal in both models. Therefore, determination of the RV/LV ratio in the axial view seems preferable as it does not require a manual reconstruction of the image.
We analyzed a range of cut-off values for the RV/LV ratio and did not use ROC analysis, as this method may not necessarily yield a clinically relevant cut-off value. To avoid missed diagnosis, the most suitable cut-off value for defining precapillary PH in this study was RV/LV ≥ 1.20 (model 2: sensitivity 94%, specificity 80%, PPV 91%, NPV 87%; model 3: sensitivity 96%, specificity 80%, PPV 91%, NPV 91%).