Conflicting data exist about the effects of cardiac resynchronization therapy (CRT) on diastolic function (DF). Aim of the study was to assess if and how CRT affects DF in systolic heart failure patients. We also investigated potential relations between CRT-induced left ventricular changes and the composite clinical endpoint of progressive heart failure and cardiac death over 3 years follow-up. 119 CRT patients underwent clinical evaluation and echocardiography before CRT and 4 months later. DF was quantified by transmitral velocities [E/A waves, deceleration time (DT), E/DT], early diastolic mitral annulus velocity (E'), E/E' ratio and 2-D speckle tracking strain rate during isovolumetric relaxation (IVR, SRivr). End-diastolic pressure-volume relationship (EDPVR) was also assessed noninvasively using a single-beat method. Overall stiffness was quantified by ventricular stiffness (Klv) normalized to end-diastolic volume (EDV). New York Heart Association class improved at 4 months (from 2.7 ± 0.7 to 1.9 ± 0.6, p < 0.001) as did ventricular filling (E/DT from 0.48 ± 0.29 to 0.39 ± 0.31 cm/s(2), p = 0.01). In contrast, relaxation (E', SRivr) and filling pressures (E/E', E/SRivr) did not change. Slope of EDPVR did not change with CRT. Such finding, together with an unmodified Klv/EDV and a 7 ± 18 % reduction in EDV (p = 0.001), suggested reverse remodelling towards a smaller equilibrium volume. Finally, end-systolic LV volume decreased from 147 ± 59 to 125 ± 52 ml and ejection fraction increased from 0.26 ± 0.07 to 0.32 ± 0.09 (both p < 0.001). Using a Cox regression model we found that only changes (Δ) in diastolic, but not systolic indexes, correlated with the composite clinical endpoint, with increments in ΔEDV20 and ΔE/DT, single or combined, greatly increasing risk of heart failure and/or cardiac death (p = 0.003). Ventricular reverse remodelling, together with improvement in ventricular filling, rather than improvements of systolic function, predict clinical prognosis long-term post-CRT.