The general purpose of this thesis was to investigate new cardiac resynchronization therapy (CRT) optimization techniques and to assess their reliability when compared to invasive measurements of left ventricular contractility (LV dP/dtmax).We first assessed whether cardiac output (CO) measured by trans-thoracic impedance cardiography could correctly identify the optimal interventricular (VV) pacing interval while using invasive measurements of LV dP/dtmax as reference. We did not find any significant statistical correlation between the two optimizing methods when their corresponding optimal VV intervals were compared. We also tested the hypothesis that measurements of right ventricular contractility (RV dP/dtmax) could be used to guide VV delay optimization in CRT. The comparison of optimal VV intervals obtained from the left and right ventricular dP/dtmax did not show a statistically significant correlation; however, a positive correlation was found when broader VV intervals were evaluated and we concluded that this finding deserves further investigation. An interesting alternative for CRT optimization is the use of device integrated algorithms or sensors capable to adapt the CRT settings to the current needs of the individual patient. In this respect we investigated the use of cardiogenic impedance (CI) measurements obtained through the CRT-D device as a method for CRT optimization with invasive measurements of LV dP/dtmax as a reference. Our results showed that CI could be measured through the device after implantation and that a patient-specific impedance-based prediction model was capable to accurately predict the optimal AV and VV delays. To follow up on these positive results we re-evaluated the patient-specific impedance-based prediction models 24 hours post implantation and investigated the possibility of calibrating them using parameters derived from non-invasive measurements of arterial pressure obtained by finger pelthysmography at implantation.The results showed that the patient-specific impedance-based prediction models did not perform as well on the follow-up data as they did on the data from implantation day and that they correlated poorly with plethysmographic parameters. Our studies suggest that novel methods for CRT optimization should be thoroughly evaluated and compared to established measures of left ventricular function prior to introduction into clinical practice.