Predicting success of intensive dialysis in the treatment of uremic pericarditis

Abstract

To identify predictors of the success or failure of daily intensive dialysis in uremic pericarditis, a retrospective examination was made of initial clinical, laboratory, and echocardiographic data in 97 patients using univariate and multivariate statistical analysis. In this group, 67 patients showed response to intensive dialysis, and 30 patients did not (22 required surgery and eight died). By univariate analysis, nine factors correlated with intensive dialysis failure (p < 0.10): admission temperature over 102 °F, rales, admission blood pressure under 100 mm Hg, jugular venous distension, peritoneal dialysis treatment only because of severe hemodynamic instability, white blood cell count over 15,000/mm 3, white blood cell count left shift, large effusion by echocardiography, and both anterior and posterior effusion by echocardiography. Echocardiographic left ventricular size and function were not useful predictors of success or failure; there was no difference in response to hemodialysis in patients with pericarditis before dialysis (69 percent) versus patients with pericarditis during a maintenance program (67 percent). By discriminant analysis, a seven-variable function was constructed that divided the patients into three groups: (1) those likely to show response to intensive dialysis (48 patients, predictive value of 98 percent), (2) those with an intermediate (38 percent) chance of showing response to intensive dialysis (30 patients), and (3) those unlikely to show response to intensive dialysis (14 patients, predictive value of 100 percent). When the function was applied prospectively to 12 patients (eight with success and four with failure), all were classified correctly. Thus, discriminant analysis of patients with uremic pericarditis allows improved selection of patients with uremic pericarditis likely to have response to daily intensive dialysis and early consideration of alternative forms of treatment in patients unlikely to show response to intensive dialysis. However, the model should be validated in the particular institution where it is to be used before its application.