The severity of pulmonary hypertension was evaluated by right cardiac catheterization in 89 patients with stable chronic obstructive pulmonary disease, both at rest and during maximum treadmill exercise. Thirty-one patients were found to have pulmonary hypertension at rest, defined as a mean pulmonary arterial pressure of 20 mm Hg or more. Although the remaining 58 patients had normal mean pulmonary arterial pressure at rest, three developed pulmonary hypertension during exercise (mean pulmonary arterial pressure greater than or equal to 35 mm Hg). Multiple anthropometric, spirometric, radiographic, and gas-exchange variables were analyzed and correlated with the hemodynamic data to define their value in predicting mean pulmonary arterial pressure. While arterial oxygen pressure (PaO2) at maximum exercise was the variable most highly correlated with resting mean pulmonary arterial pressure (r = -0.67), stepwise multiple linear regression analysis indicated that measurement of the diameter of the right descending pulmonary artery and arterial carbon dioxide tension (PaCO2) also contributed to the prediction of mean pulmonary arterial pressure. Spirometric indices of airflow obstruction, hyperinflation, and the diffusing capacity of the lung for carbon monoxide correlated poorly with the severity of pulmonary hypertension and consequently were not useful predictors of mean pulmonary arterial pressure. The threshold criteria of a PaO2 less than 60 mm Hg or a PaCO2 more than 40 mm Hg were reasonably accurate for a diagnosis of pulmonary hypertension. These arterial blood gas criteria were superior to the spirometric and radiographic variables examined in predicting pulmonary hypertension prior to the development of clinically overt cor pulmonale.