Combined use of non-invasive techniques to predict pulmonary arterial pressure in chronic respiratory disease.

The value of non-invasive procedures for predicting pulmonary arterial pressure was investigated in 370 patients with chronic obstructive lung diseases and in 73 with fibrosing alveolitis in a combined study at nine centres in six European countries. Measurements included forced expiratory volume in one second, arterial blood gas tensions, standard electrocardiogram, radiographic dimensions of pulmonary artery, right ventricle dimensions by M mode echocardiography, and myocardial scintigraphy with thallium-201; and certain clinical signs were also used. No single variable was correlated closely enough to allow accurate prediction of pulmonary arterial pressure. Four methods were used to incorporate several variables into mathematical functions for predicting pulmonary arterial pressure. In patients with chronic obstructive lung disease multiple stepwise regression explained 49% of the variance in pulmonary arterial pressure but was not useful for prediction. Discriminant analysis allowed patients to be allocated to bands of pulmonary arterial pressure, as did two non-parametric procedures, in which decision trees were established using either the Kolmogoroff-Smirnoff statistic or Fisher's exact test. Patients with a pulmonary arterial pressure of 30 mm Hg or more were identified with a sensitivity of 83% and a specificity of 91%. The non-parametric tests gave better results than discriminant function. A further 54 patients were studied to validate the functions. Of these, 90% with a pulmonary arterial pressure above 20 mm Hg were correctly identified, and 80% of those with a pulmonary arterial pressure above 29 mm Hg. Similar results were obtained in subjects with fibrosing alveolitis. These mathematical functions allow the use of combinations of non-invasive procedures to select from populations at risk of pulmonary hypertension those in whom direct measurement is required. The mathematical functions are capable of further development by incorporation of variables from newer non-invasive procedures.