We previously studied edema and left ventricular pressure-volume relations in a porcine heart model in which edema occurred even with hyperosmolar crystalloid cardioplegia. This susceptibility to edema was attributed to venous occlusion and an initial 20-minute period of ischemia. Results did not demonstrate reversal of edema by hyperosmolar perfusates. Accordingly, in the present study, heart weight, myocardial water content, and left ventricular pressure-volume curves were measured before and after perfusion-induced edema in eight isolated, arrested, hypothermic porcine hearts. Cardioplegic solution was infused 2.1 +/- 0.8 minutes after the onset of ischemia, and the atrioventricular ring was not clamped during the administration of cardioplegic solution. Cardioplegic solution (1 L) was infused at intervals of 33 +/- 6 minutes at 4 degrees C. Solution osmolarity was 380 (Stanford solution) or 294 mOsm/L (Plegisol solution). The perfusion sequence was 380-1, 380-2, 294-1, 380-3. Pressure-volume relations were assessed with the use of left ventricular volume at a pressure of 10 mm Hg and the ventricular chamber stiffness constant, beta, derived from P = alpha e beta V. Perfusions 380-1 and 380-2 did not affect the pressure-volume curve. Perfusion 294-1 increased heart weight and water content (p < 0.05) and decreased left ventricular volume at 10 mm Hg compared with perfusions 380-1, 380-2, and 380-3. In addition, beta increased (0.023 +/- 0.005 versus 0.029 +/- 0.006, p < 0.05) after perfusion 294-1, compared with 380-1. Correlation coefficients for linear regressions between left ventricular volume at 10 mm Hg and heart weight and water content were r = 0.84 and r = 0.70, respectively. We conclude that under conditions similar to those used clinically, the left ventricle of the pig does not develop edema with Stanford solution (380 mOsm/L). Edema does follow Plegisol solution (294 mOsm/L) cardioplegia. Edema and reduced compliance are incompletely reversed by hypertonic cardioplegia. The porcine left ventricle can usefully replicate the clinical model.