We explored the influence of global longitudinal strain (GLS) measured with two-dimensional speckle-tracking echocardiography on left ventricular mass regression (LVMR) in patients with pure aortic stenosis (AS) and normal left ventricular function undergoing aortic valve replacement (AVR). The study population included 83 patients with severe AS (aortic valve area <1 cm(2)) treated with AVR. Bioprostheses were implanted in 58 patients (69.8 %), and the 25 remaining patients (30.2 %) received mechanical prostheses. Peak systolic longitudinal strain was measured in four-chamber (PLS4ch), two-chamber (PLS2ch), and three-chamber (PLS3ch) views, and global longitudinal strain was obtained by averaging the peak systolic values of the 18 segments. Median follow-up was 66.6 months (interquartile range 49.7-86.3 months). At follow-up, values of PLS4ch, PLS2ch, PLS3ch, and GLS were significantly lower (less negative) in patients who did not show left ventricular (LV) mass regression (all P < 0.001). Baseline global strain was the strongest predictor of lack of LVMR (odds ratio 3.5 (95 % confidence interval 3.0-4.9), P < 0.001), and GLS value ≥-9.9 % predicted lack of LVMR with 95 % sensitivity and 87 % specificity (P < 0.001). Other multivariable predictors were the preoperative LV mass value (cutoff value ≥147 g/m(2), P < 0.001), baseline effective orifice area index (cutoff ≤0.35 cm(2)/m(2), P = 0.01), and baseline mean gradient (cutoff ≥58 mmHg, P = 0.01). Finally, we failed to find interactions between GLS and other significant parameters (all P < 0.05). Global longitudinal strain accurately predicts LV mass regression in patients with pure AS undergoing AVR. Our findings must be confirmed by further larger studies.