LV Filling: LV filling was studied using transmitral blood velocities recorded by pulsed Doppler. Doppler measurements were averaged from at least three consecutive beats. Transmitral blood velocities were obtained from the apical four-chamber view, positioning the sample volume at the mitral valve leaflet tips. Doppler velocity curves were recorded at 100mm/s. Peak velocity and VTI of the initial flow (E wave), representing the early filling phase, and of the late flow (A wave), representing the atrial contraction, were measured. The peak velocities ratio (E/A) and the ratio of the A wave VTI to the total VTI (relative contribution of atrial contraction to the total LV filling) were calculated. Other following variables were measured: the deceleration pressure half time of early diastolic transmitral flow (pressure half-time) and isovolumetric relaxation time (IVRT). The IVRT was the interval from the aortic valve closure signal to the mitral valve opening signal.

RV Diameters: RV end-systolic diameter and RV end-diastolic diameter (RVEDD) were measured by M-mode echocardiography from the left parasternal long-axis views.

RV-Right Atrial Pressure Gradient: As we did not measure the right atrial (RA) pressure, we expressed our results as instantaneous systolic pressure gradient from RV to RA, without any calculation of systolic pulmonary artery pressure. Tricuspid regurgitant flow was identified in continuous Doppler mode from the apical four-chamber view. Instantaneous systolic pressure gradient from RV to RA (RV/RAg) was calculated with the modified Bernoulli equation from the peak velocity of the tricuspid regurgitant signal: RV/Rag = 4V2, where V is the maximal regurgitant velocity in meters per second. Continuous variables were expressed as mean ± SD. Statistical tests were run on statistical software (Sigma Stat; SPSS; Chicago, IL). Each subject has served as his/her own control. Two series of measurements were obtained: the first as control, and the second after scuba diving. Data distribution was studied using a Kolmogorov-Smirnov test. When data distribution reflected a normal distribution, we used a t test for paired data. In the case of cohorts of variables not having a normal distribution, comparisons were done with the Wilcoxon matched-pair signed-ranks test; p values < 0.05 were considered as significant.