A method and apparatus for optimizing cardiac resynchronization therapy are provided. An iterative optimization procedure is performed to test the systolic hemodynamic effects of varying A-V-V timing schemes. The hemodynamic effect is assessed based on a surrogate of stroke volume. The stroke volume surrogate is derived from a sensor signal proportional to the blood pressure in the aorta or a major artery. The A-V-V timing scheme corresponding to the highest magnitude stroke volume, as indicated by the stroke volume surrogate, is identified and automatically programmed to maintain optimal A-V-V settings acutely and chronically.