Empirically treated Pneumocystis carinii pneumonia in London, 1983-1989.

For 227 episodes of Pneumocystis carinii pneumonia (PCP) treated at St Mary's between 1983 and 1989, factors predictive of fatal outcome were age, haemoglobin levels, peripheral lymphocyte count and alveolar-arterial oxygen gradient. Case fatality for the 47 empirically-treated episodes was significantly higher compared with the 180 cytologically proven episodes (55% vs 18%, chi 2 = 25.7, P less than 0.0001). Case fatality for episodes which could not be bronchoscoped was significantly higher compared with bronchoscopy negative cases (66% vs 25%, chi 2 = 4.5, P less than 0.05). Predictive factors for fatal outcome differed significantly for cases which could not be bronchoscoped and cytologically proven cases: haemoglobin level (10.7 g/dl vs 12.0 g/dl, P less than 0.001), lymphocyte count (0.64 x 10(9)/l vs 0.87 x 10(9)/l, P = 0.05) and oxygen gradient (77.7 mmHg vs 58.9 mmHg, P less than 0.02). Such differences were not observed between bronchoscopy negative and cytologically proven cases. Case fatality decreased significantly over time (b = -0.39, SE = 0.14, P less than 0.05). Total and non-fatal first time episodes displayed an inverse relationship between oxygen gradient and time (r = -0.22, P less than 0.006 and r = -0.24, P less than 0.01, respectively). Mean oxygen gradient of fatal episodes for sequential years increased significantly from 73 mmHg in 1983 to 102 mmHg in 1989 (r = 0.92, P less than 0.01). This suggests that medical intervention as well as presentation with less severe disease both contributed to improved case fatality over time.