A pooled community-based population sample of 17,139 North American children, adolescents, and adults aged from birth to 75 years was used to evaluate a variety of one-, two-, and three-parameter formulas for correction of the QT interval for heart rate throughout a wide range of heart rates in sinus rhythm. QT measurements were made by a computer program from simultaneously sampled standard 12-lead or orthogonal XYZ leads, and all QT measurements were visually verified using a high-resolution display terminal. A random subsample of 1,920 was drawn 3 times by allocating 20 subjects to each heart rate subinterval of 1 beat/min, and the performance of a set of 13 QT prediction formulas was compared by ranking them according to the Akaike Information Criterion. The traditional Bazett's square root formula failed at low heart rates and the cube root formula of Fridericia at high rates. None of the linear two-parameter functions of R-R interval or heart rate performed with adequate accuracy. The best one-parameter formula for predicted QT(QTp) was obtained by regressing the inverse of QT on heart rate, with the expression QTp(ms) = 656/(1 + 0.01 H R), and by adding the term 0.4 x Age - 25 for age-related trend correction in males 15-50 years old. Percentile population distributions for the QT index (QTI = (QT/QTp) x 100) produced a convenient and stable 98 percentile normal QT range spanning 10% above and below the population median QTI = 100 in all major subgroups by age, sex, and race.(ABSTRACT TRUNCATED AT 250 WORDS)