In 881 middle-aged men from one Italian cohort of the Seven Countries Study, QT and RR intervals were measured in lead 2 from resting ECGs (25 mm/sec) and fitted separately with 10 mathematically different QT prediction formulas. The relative accuracy of fit to data was assessed from the minimum mean-squared residual and the minimum Akaike Information Criterion values. Using the Minnesota code, 588 men had normal (group 1) and 293 had abnormal (group 2) ECGs. A better fit to QT-RR data by all formulas was observed in group 1, compared with group 2. Among one-parameter equations in both groups, the cubic root Fridericia's formula is better suited to fit the data than the Bazett's square root or other formulas. The former compares favorably with multiparameter equations or with the inverse relation and gives the best fit in group 2. Thus the cubic root equation might be more accurate than the square root or several complex formulas for correcting measured QT intervals for cardiac cycle length in middle-aged men.