The QT interval represents the ventricular action potential from depolarization to repolarization [1]. Long QT Syndrome (LQTS) is both an acquired and inherited arrhythmia disorder characterized by a prolonged QT interval. The disorder can clinically manifest with syncope, palpitations, and in severe cases sudden cardiac arrest due to Torsade de Pointes [2]. Over 17 genes contributing to ion-channel function have been implicated in congenital LQTS [2–4]. Accuracy of QT measurement, therefore, is imperative for the early detection of LQTS and initiation of appropriate therapy to prevent morbidity and mortality. As QT interval is strongly affected by heart rate (HR), several formulae have been proposed to correct for this relationship. The ideal formula would eliminate any relationship between HR (or RR interval on ECG) and the corrected QT interval (QTc).

Bazett and Fridericia describe an exponential relationship between QT vs. RR or HR [5,6]. Bazett developed a formula estimating the duration of systole with varying pulse rates from a study of 39 subjects, which was later adapted to relate QT prolongation to acute rheumatic carditis in children [5,7,8]. Fridericia proposed a relationship between the duration of ventricular repolarization (QT) and pulse period (RR) through a study of 50 subjects with a cutoff of 430 milliseconds (ms) as the definition of prolonged [9]. On the other hand, Framingham, Hodges, and Rautaharju suggested a linear relationship between QT vs. RR or HR [10]. The Framingham formula integrated data from 5,018 subjects with a median age of 44 years comprising the Framingham Heart Study to create a linear regression model for correcting QT according to RR cycle length [10]. In response to the discrepancies identified with the non-linear Bazett and Fridericia formulae, Hodges developed a linear QTc formula from a study of 607 subjects [11]. Rautaharju, a newer formula, derived from 57,595 patients ranging from five to 89 years of age with a cutoff of 477 ms; Rautaharju additionally introduced a sex coefficient to account for observed differences in QT duration by sex [12].

The parameters of each formula are detailed in Table 1. Across all formulae, QTc is equivalent to QT at a HR of 60 bpm (frequency of 1 Hz) [13].

Both intra-subject [14,15] and inter-subject [16,17] differences in QTc/HR patterns have been demonstrated. Although the robustness of different QTc formulae against intra-subject and inter-subject variations in HR has been well-explored in adults, it has been less examined in the pediatric population that has intrinsically greater extremes of HR and is more susceptible to errors in QT measurement. Our study aims to determine which formula provides the best rate correction within the pediatric population and to assess if the computer-algorithm utilized by our ECG machine is a reliable method of QT correction compared to manual calculation.

This is a descriptive retrospective study. We included patients in the pediatric cardiology practice of a community hospital who had an ECG performed in the year 2017. We reviewed the age, sex, past medical history, and relevant medications of each participant. We excluded all participants with a past medical history (including cardiovascular disease) as well as those on regular medications. Only the first ECG was retained for each participant. The demographic and ECG parameters associated with each participant may be available upon request. We examined the ECGs for rhythm, HR, QT, RR, and QRS intervals. ECGs with missing leads or excessive noise were removed, along with those with a QRS duration above 120 ms and those not in sinus rhythm. Participant demographics were evaluated using mean and standard deviation (SD) for continuous variables, and frequency and percent for categorical variables. We calculated the QTc interval using the Bazett, Fridericia, Framingham, Hodges, and Rautaharju formulae as well as the computer algorithm. The Marquette^TM 12SL^TM ECG Analysis Program was utilized for the computer calculation of QTc, which reports the Bazett-corrected measurement of QTc [18].

The overall cohort is described through descriptive statistics. Wilcoxon rank sum test compared clinical characteristics and QTc values between male and female participants. Spearman’s rank correlation tests evaluated for potential correlations between two variables. Two-sample paired t-test or Wilcoxon signed-rank test, as appropriate, assessed for potential differences between every combination of two QTc methods, with p-values adjusted according to the Benjamini-Hochberg method. Separate univariate linear regression models determined potential associations between QTc measurements and HR or RR. Bland-Altman Plots evaluated for agreement between two QTc formulae.

All analyses were also performed by excluding the top and bottom 2% of the data as sorted by HR ranges according to the method adopted by Luo et al. [19]. This group adopted the top and bottom 2% of data instead of the traditional mean ± 2 SD to demarcate the upper and lower normal limits of QTc distributions for each formula on the basis that the characteristics of QT distributions are not clear, possibly following a skewed distribution [20].

All p-values are two-sided with statistical significance evaluated at the 0.05 alpha level. All analyses were performed in R Version 4.1.2 (The R Foundation for Statistical Computing).

This study was reviewed and approved by the Institutional Review Board at the New York Presbyterian Brooklyn Methodist Hospital (IRB number 1792968).

Over 25 formulae have been proposed to correct for the HR-dependence of the QT interval, testifying to the controversy surrounding the endeavor towards a universal formula [23]. Existing methods have been repeatedly criticized with numerous attempts to identify more appropriate ones. Studies noting an intra-subject variability in the QT/HR relationship have further complicated the endeavor, suggesting the need for an individual QT correction method derived from subject-specific data [24–26]. Still, other studies recommend accounting for age or sex differences in QT interval [27].

Despite the controversy, Bazett continues to be most widely implemented in clinical practice, including pediatrics. The Schwartz scoring system for the clinical diagnosis of LQTS, for example, integrates the Bazett method to calculate for risk of disease [28]. As Bazett is known to over-correct QT at higher HRs [10], the use of the formula is of especial concern in the pediatric population where the normal HR of neonates ranges from 110 to 160 bpm, and children’s HR do not decrease to below 100 bpm until adolescence. In a study of 332 children with a mean age of 10.7 years, Bazett led to a high number of false positives in LTQS screening; results were exacerbated by increases in HR due to postural maneuvering [11]. Additionally, in a study of 54 healthy children with a median age of 9.9 years, Bazett (as well as Hodges) had a positive QTc and HR relationship, resulting in a prolonged QTc at peak exercise; the study recommended the use of Fridericia or Framingham calculated at one minute after maximum exercise for superior congenital LQTS assessment [29]. Our study provides additional evidence that Bazett over-corrects at higher HRs.

Still, other studies support the continued use of Bazett in pediatric cardiology [13,30–34]. In a study of 2,500 randomly selected ECGs from neonates, Bazett provided the most HR independent QT correction and accurately identified neonates affected by LQTS [34]. In a larger scale prospective study of more than 33,000 infants, Bazett successfully identified infants with a prolonged QTc and at significant risk for sudden death [30]. In a review of ECGs from 702 children with 81% less than 2 years of age, Bazett produced the most consistent QTc values across HRs with the support of 460 ms as the best threshold for prolonged QTc [33].

In our study, Fridericia was most independent of HR and RR. This finding supports the conclusion of other studies, which report Fridericia to provide more consistent calculations at faster HRs [24,29,35]. A study on children and adolescents from six to 17 years of age found the ideal HR correction formula to be QTc = QT/(RR^0.38), which most closely resembles the Fridericia formula [36]. Regulatory agencies such as the U.S. Food and Drug Administration recommend the concurrent use of Fridericia and Bazett in clinical trials on drug safety [37,38]. The robustness of Fridericia to changes in HR may attribute to the method of formula development, which studied the QT/HR relationship to identify a coefficient that would minimize the impact of RR on QTc [6,10]. The Bazett formula, alternatively, was not based on such population-based calculations [5].

Framingham and Rautaharju had strong correlations with HR and RR in our study. Our study does not recommend the use of these formulae for QT correction, although the limitations of our study, including its small sample size, must be considered. Hodges had modest performance in the Spearman’s rank correlation and linear regression analyses against HR and RR.

Computer-calculated Bazett formula performed equally with a manual calculation of Bazett in our population. Our findings suggest that if Bazett is an appropriate and validated formula for QTc calculation in the pediatric population as seen in other studies, the computer algorithm may be a comparable screening tool to detect QTc prolongation. Computer-derived algorithms and QTc calculators on smart devices may represent valuable point of care resources in settings outside of cardiology practice where health professionals may be less familiar with measuring QT intervals [39]. Moreover, given intrinsic variations in manual QT measurement across health care personnel and facilities, a standardized method of measurement through a computer device may increase both the diagnosis and monitoring of patients with prolonged QTc. Studies on the accuracy and reliability of computer-based algorithms and QTc calculators remain conflicting, however [40,41], with many displaying good precision, but low accuracy [41,42].

Limitations of our study include its retrospective design and small population size consisting of only healthy patients (N = 95). Despite a small sample size, statistically significant differences were observed between formulae. As a community hospital, our population may not be representative of other pediatric populations. Larger scale prospective studies are needed to confirm our findings. Our study assessed the performance of QTc formulae against variations in HR and RR across individuals, while variations in the parameters have also been noted within individuals.

Many formulae have been proposed to decrease the impact of HR on QT with conflicting perspectives on the ideal formula for pediatric practice. While Bazett remains most utilized, many have raised concerns regarding the formula’s tendency to over-correct at higher HRs. This finding is especially relevant to the pediatric population, which is characterized by extremes of HR, particularly within neonates. Our findings support the use of Fridericia to minimize the influence of HR on QT. Manual and computer algorithm-based calculations of Bazett exhibited moderate linear correlations with HR and RR, while Framingham and Rautaharju had strong correlations. The results of our study must be weighed against its limitations, including a small population size. If Bazett is a reliable and valid method of QT correction as seen in other pediatric study cohorts, our findings suggest that the computer algorithm may be a comparable method for use in clinical practice to standardize QT assessment.