Effect of self-correlation on the relation between QT interval and cardiac cycle length

The electrocardiographic QT interval, with a wide variety of practical and theoretic applications in medicine, pharmacology, and cardiology, is critically dependent on heart rate as expressed by cycle length (RR). To account for this and permit intraindividual and interindividual comparisons of changing QT intervals, a wide variety of formulas have been offered as single rate-correcting expressions over the range of heart rates. Yet as rate rises, the QT interval occupies more and more of the cycle length, until at the highest heart rates almost the entire cycle length is QT interval, and the QT-RR correlations become largely self-correlation. To elucidate this we investigated 150 consecutive patients divided into high, mid, and low rate terciles. For the entire group, and in each tercile, we determined the level of correlation between the QT interval and cycle length (RR interval). To remove the self-correlation, we repeated the calculations, correlating QT interval with RR interval minus QT interval. For the group, correlation (r) of QT with RR was 0.65. For the high, mid, and low rate terciles, correlation fell sharply, respectively: 0.58 to 0.36 to 0.25. When the self-correlation was removed by subtracting QT from cycle length (RR-QT), all correlations were even further sharply reduced for each tercile, respectively: 0.35, 0.16, and 0.03 (the latter being virtually no correlation). We conclude that at low heart rates, when it is not being correlated with itself, the correlation between QT and RR is low; at high heart rates the higher correlation is spurious because it is mainly self-correlation.