From August 1985 until December 2005, we randomly recruited a family-based population sample from a geographically defined area in northern Belgium as described elsewhere . EPOGH recruited participants from 1999 until 2001. The EPOGH investigators were trained at the Studies Coordinating Centre, and applied the same protocol, questionnaires and follow-up procedures, as used in FLEMENGHO. The FLEMENGHO and EPOGH studies received ethical approval. The initial response rate at enrolment was 61.3% [10, 11]. All subjects provided informed consent in writing.
The FLEMENGHO and EPOGH participants remained in follow-up. Five centers opted to perform echocardiographic phenotyping and to assess LV function by using the new TDI indexes along with classical pulsed wave Doppler velocities of blood flow. Our current study population includes 1287 subjects, who were examined from June 2005 until September 2009. We excluded 29 subjects from analysis, because of atrial fibrillation (n = 12), a pacemaker (n = 2), or because LVDD could not be reliably determined (n = 15). Thus, the current analysis included 1258 participants: 782 FLEMENGHO participants (Noorderkempen, Belgium) and 476 EPOGH subjects from Gdańsk (n = 108) and Kraków (n = 124), Poland, Mirano, Italy (n = 106), and Novosibirsk, Russia (n = 138).
In each center one experienced physician did the ultrasound examination, using a Vivid 7 Pro (GE Vingmed, Horten, Norway), interfaced with a 2.5- to 3.5-MHz phased-array probe, according to a standardized protocol described in detail in previous publications . With the subjects in partial left decubitus and breathing normally, the observer obtained images, together with a simultaneous ECG signal, along the parasternal long and short axes and from the apical 4-, 2-chamber and long-axis views. All recordings included at least 5 cardiac cycles and were digitally stored for off-line analysis.
One experienced observer (TK) analyzed the digitally stored images, using the EchoPac software, version 4.0.4 (GE Vingmed), averaging three cardiac cycles. The LV internal diameter and interventricular septal and posterior wall thickness were measured at end-diastole from the 2-dimensionally guided M-mode tracings according to the recommendations . When optimal orientation of M-mode ultrasound beam could not be obtained, the reader performed linear measurements on correctly oriented two-dimensional images. End-diastolic LV dimensions were used to calculate LV mass by an anatomically validated formula. LV hypertrophy was a left ventricular mass index (LVMI) of 125 g/m2 in men and 110 g/m2 in women or more. We calculated LV ejection fraction (EF) from LV end-systolic and end-diastolic volumes measured from the apical 4- and 2-chambers views, using the standard biplane Simpson's method. We measured left atrial (LA) dimensions in 3 orthogonal planes: the parasternal long, lateral, and supero-inferior axes . LA volume (LAVI) was calculated using the prolate-elipsoid method  and was indexed to body surface area.
From the transmitral flow signal, we measured peak early (E) and late (A) diastolic velocities, the E/A ratio and A flow duration. The duration of PV reversal flow during atrial systole (AR) was measured from the PV flow signal. From the pulsed wave TDI recordings, we measured the early (e') and late (a') peak diastolic velocities of the mitral annulus displacement, and the e'/a' ratio at the 4 acquisition sites (septal, lateral, inferior, and posterior). We calculated the E/e' ratio by dividing transmitral E peak by e' averaged from the 4 acquisition sites. As reported previously,  the inter-observer intra-session reproducibility across the four sampling sites ranged from 4.48% to 5.34% for e' velocities and from 3.96% to 4.52% for a' velocities.
We administered a standardized questionnaire to collect detailed information on each subject's medical history, smoking and drinking habits, and intake of medications. The conventional blood pressure was the average of five consecutive auscultatory readings obtained with the subject in the seated position. Hypertension was a blood pressure of at least 140 mmHg systolic or 90 mmHg diastolic or the use of antihypertensive drugs. Body mass index (BMI) was weight in kilograms divided by the square of height in meters. Obesity was a BMI of 30 kg/m2 or higher. Central obesity was a waist circumference of at least 102 cm in men or 88 cm in women. Diabetes was a fasting blood glucose of 7.0 mmol/L or higher or the use of antidiabetic agents.
NT-proBNP in the FLEMENGHO study was measured in plasma by a competitive enzymatic immunoassay for research use (Biomedica Gruppe, Vienna, Austria) . NT-proBNP in the EPOGH study was determined in serum by an electrochemiluminescence immunoassay Elecsys 2010 (Roche Diagnostics, Indianapolis, USA) .
To generate a reference sample, we excluded participants, if one or more of the following conditions were present: hypertension (FLEMENGHO/EPOGH, n = 323/211), diabetes (n = 29/25), obesity (n = 143/134), central obesity (n = 216/144), LV hypertrophy (n = 84/36), renal failure (n = 4/4) or cardiac diseases (valvular abnormalities, n = 31/8; myocardial infarction and/or coronary revascularization, n = 24/6). The number of subjects in the reference group was 338 in FLEMENGHO, 185 in EPOGH, and 523 in total.
For database management and statistical analysis, we used SAS software, version 9.1.3 (SAS Institute, Cary, NC). We compared means and proportions by means of a large sample z-test and the χ
2-test, respectively. We performed single and stepwise linear regression to identify correlates of the Doppler indexes as measured on a continuous scale. We set the P-values for variables to enter and to stay in the regression models at 0.05. To obtain 95% confidence intervals of the percentile values of the E/A distributions, we computed the bootstrap distribution  of the thresholds by randomly resampling the study population 1000 times with replacement, using the PROC SURVEYSELECT procedure, as implemented in the SAS software. Because NT-proBNP was measured by different methods in FLEMENGHO and EPOGH, we rescaled the values by computing population-specific z-scores, which reflect the deviation of each individual measurement from the population mean.