Sudden cardiac death (SCD) is the leading cause of mortality in athletes engaged in competitive sport (9). One of the most common cardiac genetic diseases that can lead to SCD is long QT Syndrome (LQTS). LQTS is a multifactorial disease that manifests itself as a profound QT interval prolongation on the ECG, syncope, and possibly death. It is estimated that 1 in every 2000 individuals carry some form of genetic mutation for LQTS (20,21). With the recent genetic discovery of a novel calmodulin mutation (13), 15 genes have now been linked to the congenital form of LQTS.
In an effort to identify LQTS and prevent SCD, athletes are recommended to undergo a preparticipation health screening before engaging in physical activity (15). Although not mandated in the United States, an ECG can be an important element of this exam. Athlete-specific ECG interpretation criteria have been developed (Seattle Criteria) (5) in an effort to distinguish between adaptive physiological changes and those that may be indicative of an underlying cardiac disease. The Seattle Criteria’s recommendation for identifying young athletes at risk of SCD because of LQTS is to correct the QT interval measure by using the Bazett’s correction formula (
) (6). A QTc interval greater than 470 ms for men or 480 ms for women is considered a sign of potential LQTS requiring further investigation (6).
The Bazett correction formula was established in 1920 from an analysis of less than 50 subjects (2) and has subsequently been shown to undercorrect the QT interval at low HR ranges (14,19), which are HR typically experienced by young athletes. The Bazett correction formula was implemented to enable the application of the same numerical criteria for all HR. However, because of its failure to do so, it adds systematic measurement bias in athletes, which may result in a higher false-negative rate for QT interval prolongation. This reduces the number of athletes who are identified with QT interval prolongation that may benefit from follow-up investigation.
To avoid this, it has been recommended that maneuvers to raise HR (23) be used in athletes, such as standing or exercise. This approach is complicated by the hysteresis between HR and QT interval adaptations that result in a delayed response to changes in HR (8). In an effort to improve screening for LQTS in young athletes, we evaluate preparticipation screening ECG to determine a more appropriate method for QT interval estimation.
We analyzed N = 2077 athlete ECG data that were collected as part of preparticipation health screenings. All screenings were completed within the United States between June 2010 and March 2015, and all participants signed a consent form approved by the institutional review board. Three distinct cohorts of athletes were included in the study: high school (HS), collegiate (Col), and professional athletes (Pro). The majority of athletes participated in either rowing/crew, American football, or basketball (52%).
As part of their preparticipation health screening, all athletes received a standard cardiac assessment that included the completion of the AHA preparticipation screening questionnaire, a cardiac-focused physical exam, and a recording of an ECG. Athletes with abnormalities on any portion of these assessments were consulted by an expert cardiologist specializing in sports medicine. As resting basal ECG data are required, we excluded ECG from athletes whose HR values were greater than 100 bpm (n = 26). After examination, no athletes were determined to have a preexisting cardiovascular disease precluding their participation in sports. On follow-up, to our knowledge, there have been no reports of SCD or sudden death in these athletes.
ECG recording and analysis
All ECG data were recorded using the CardeaScreen electrocardiograph system (Seattle, WA) and were conducted by personnel trained in 12-lead ECG acquisition. High-resolution 16-s ECG data were recorded at 1 kHz and band-pass filtered between 0.05 and 150 Hz. The CardeaScreen system automatically performs waveform analysis and interval measurements, and these were overread by an expert cardiologist specializing in sports medicine.
To ensure the validity of the automatically derived QT intervals, all QT intervals were plotted by HR, and if they were greater than approximately 3 SD from the population mean QT interval, they were determined to be outliers and underwent manual verification. The fiducial points of all outliers were then inspected for accuracy and manually corrected if the QRS onset or T wave offset were misidentified. Manual correction was performed using the same global QT interval used for the automatic measures of the CardeaScreen system. Overall, less than 2% of ECG required manual correction, most commonly for the misidentification of the T wave end. This can occur when the amplitude of the T wave is of insufficient voltage, making it difficult to identify the end of the QT interval. After manual correction, the QT interval was recalculated and included in the analysis data set.
To determine a model of best fit, a regression analysis was performed and exponential, linear, and power function models developed for the uncorrected QT interval data. To test the appropriateness of each correction formula, QTc intervals were calculated using the four common correction formulas: Bazett, Fridericia, Hodges, and Framingham. These were plotted by HR, and mean QTc/HR slopes were determined. In a perfectly corrected model, the QTc/HR slope would equal zero. However, in the absence of a perfect QT interval correction formula, for this study, the QT interval correction formula resulting in the lowest mean slope was deemed the most appropriate for correcting the QT interval in athletes.
To analyze the QT interval and to evaluate the effect of the Bazett’s correction formula, as recommended within the Seattle Criteria, the 95th and the 99th percentile measures of the uncorrected QT interval were determined and grouped by HR. The 99th percentile measures were compared with the Bazett’s corrected QT interval groups using classification analysis, determining the subjects that would and would not have received secondary investigation. Summary statistics were used to compare these groups. All analysis was performed using Microsoft Excel (Seattle, WA), NCSS (Kaysville, UT), and IBM SPSS Statistics for Windows (Version 20.0, Armonk, NY).
Two thousand and seventy-seven athletes were evaluated (n = 597 HS, n = 1207 Col, and n = 273 Pro), and only those with an HR less than 100 bpm were analyzed. The mean age was 19 ± 3.5 yr (14–35 yr), and the majority of the athletes were male and Caucasian (Table 1).
Overall mean (
) QTc interval was 403 ± 33 ms and expectedly varied between men and women (400 ± 52 vs 410 ± 33 ms, p < 0.0001). The QT interval was plotted against HR for all subjects by sex (Fig. 1). The overall linear correlation between the QT interval and the HR was (R = −0.62), with women having a slightly stronger linear correlation than men (R = −0.82 vs −0.78). The QT interval also varied by cohort, with high school athletes having a lower mean QT interval than college or professional athletes (HS = 388 ± 30, Col = 410 ± 33, Pro = 407 ± 27, p < 0.0001).
Model of best fit
To identify the model that best represented the relationship between the QT interval and the HR in young athletes, linear, exponential, and power function models were developed (Fig. 2). A nonlinear power function with an exponent of −0.349 fit the data the best (R2 = 0.64), slightly outperforming both exponential (R2 = 0.63) and linear (R2 = 0.62) functions.
QT interval correction
To determine the most appropriate method for QTc interval estimation in young athletes, standard linear and nonlinear correction formulae were applied. These were plotted, and the residual correlations to HR were determined (Fig. 3). Of the four correction formulas tested, the Fridericia correction formula resulted in the lowest mean QTc/HR slope (m = −0.10), slightly lower than that derived with the Framingham formula (m = −0.11). The Seattle Criteria’s recommended Bazett’s correction formula (m = 0.96) and the Hodges correction formula had significant residual dependence to HR (m = −0.41) after application.
QT interval thresholds
Table 2 presents the 95th and the 99th percentiles for uncorrected QT for HR bands. These uncorrected QT values represent the gold standard for identifying those at greatest risk of manifesting the LQTS. The performance of the Seattle Criteria’s recommended Bazett’s correction formula in identifying athletes requiring secondary investigation was compared with the method of using the 99th percentile uncorrected QT intervals, grouped by HR (Table 3). In this classification analysis, seven subjects had Bazett’s corrected QT interval prolongation warranting further investigation for LQTS (five men and two women), whereas only five athletes (four men and one woman) warranted further investigation for LQTS when the Fridericia correction formula was applied. In using the Bazett or the Fridericia correction formula, we find that up to 75% of athletes with an uncorrected QT interval greater than the 99th percentile values within this athlete population (n = 23) are not identified as needing secondary investigation for LQTS. Importantly, all athletes identified by either the Bazett’s or Fridericia correction formulae are represented within the group with an uncorrected QT interval greater than the 99th percentile, demonstrating comparative screening sensitivity.
This is the first study to attempt to determine the optimal method for QT interval estimation in athletes. We find that the Bazett’s correction formula, recommended by the latest athlete ECG interpretation guidelines (Seattle Criteria) (5), has the least favorable performance of four typically used correction formulae, with significant residual dependence to HR (m = 0.96). Of the other correction formulas that have been used extensively in clinical practice and research, although still imperfect, the Fridericia correction formula resulted in the lowest residual dependence to HR (m = −0.10). One explanation for this is the comparative equivalence between the Fridericia correction formula (QTc = QT/RR1/3) and the exponential of the model of best fit (y = 1702.1x−0.349) (1/3 or 0.33 vs 0.35). As the coefficients verge, it is obvious that the similar Fridericia formula would outperform the other formula.
Two likely reasons why Bazett’s correction formula continues to be used in clinical practice and research, despite its well-documented measurement biases, are that it is relatively simple to calculate and it has historical precedence. We propose a risk stratification method that forgoes the need to apply a QT interval correction formula. This approach is supported by the wide scatter of the QTc/HR data plots for any of the correction formula, compared with the uncorrected HR/QT data plots (Figs. 1 and 3). Because of the measurement biases inherent in the application of the Bazett’s formula, we find that up to 75% of subjects who had an uncorrected QT interval greater than the 99th percentile for the population were not identified as having QT interval prolongation after the application of the Bazett’s formula and, therefore, did not receive investigation for LQTS. As these uncorrected QT intervals were grouped by HR, these truly represent the longest 1% of all athlete’s QT intervals. Whether these cases represent true positives for LQTS is unknown and is difficult to say with confidence without confirmed cases of LQTS in our data; however, it does reveal missed opportunities to investigate athletes for LQTS who possess true QT interval prolongation.
Athletes within the top 1% or possibly even 0.5% for QT interval duration warrant further investigation. As shown in Table 2, the current QTc interval thresholds of 470 ms for men and 480 ms for women inadequately capture those athletes within the top 1% for QT interval prolongation. In screening athletes for potential LQTS, we believe a slightly higher false-positive rate would be acceptable. Having to conduct a minor number of follow-up investigations is preferred, versus the risk of missing one case of LQTS and SCD due to a false-negative finding using more stringent, and as we have shown, inaccurate methodology.
The secondary assessment for a patient with QT interval prolongation involves a more focused detailed personal and family history related to LQTS, in an effort to identify symptoms suggestive of a need for further genetic investigation. If the 99th or the 99.5th percentile is used as an upper-bound threshold for risk, this equates to 1% or less of athletes undergoing further secondary investigation, and although we do not demonstrate data within this study, only a subset of these athletes will ever have symptomatology requiring more intensive follow-up and genetic testing. With current ECG abnormality rates for preparticipation screening around 6% (18), we believe this recommendation represents a very conservative approach to LQTS screening.
Forgoing the need to correct the QT interval has been suggested previously (7); however, this is the first study to recommend using uncorrected QT interval measurements for the risk stratification of LQTS in athletes. This is an important consideration. Research has previously demonstrated significant interindividual variability in QT and HR dynamics (14). Therefore, population-based correction formulas invariably add considerable QT interval estimation error, as no one person is represented by a population-level correction formula. Stemming from this consideration, individual correction formulas have been suggested, especially related to the assessment of drug-induced QT interval prolongation (14). However, although preferred over population-level corrections formulas, an individual correction is impractical to perform during a preparticipation screening with a 12-s ECG. The use of uncorrected QT interval measurements, or a “QT nomogram” for risk stratification, has been tested in acquired LQTS and is shown to outperform traditional correction factors (3,11,25). However, these studies often use an uncorrected upper 95th percentile risk threshold, which is lower than that proposed within our study (99th percentile).
Risk thresholds are important aspects of any screening program. Importantly, there have not been any studies that determine QT interval risk thresholds specifically in athletes. Much of the recommendations within the Seattle interpretation criteria for athletes are based on data derived from nonathlete populations, often from large ambulatory populations (16,22). The application of these traditional QT interval thresholds (470 ms men and 480 ms women) may not be applicable to athletes for LQTS risk stratification. As athlete’s hearts are typically different from nonathletes (4,10), especially with lower resting HR, the use of the Bazett’s correction formula contributes significant measurement bias, specifically when applied to the young athlete population.
The combination of a correction formula that introduces measurement bias and the use of thresholds that are more appropriate for adult nonathlete populations remain problematic. This is clear from the nearly 10-ms difference between the upper thresholds shown within our young athlete cohort and a reference data set of more than 70,000 individuals published by Mason et al. (16). Therefore, in addition to recommending the use of the uncorrected QT interval in HR groups, we identify a need to establish new reference ranges for QT interval prolongation in young athletes. We have used a threshold of the 99th percentile population-level QT interval. However, with larger data sets and identification of LQTS cases, this should undergo further examination.
Not correcting the QT interval measure will also significantly reduce measurement error associated with QT interval monitoring. The miscalculation of the QTc interval has been shown to be a major determinate for diagnostic miscues in subjects referred for LQTS evaluation (22). Studies consistently demonstrate that physicians are unable to reliably measure and correct the QT interval during testing conditions, which we believe are more favorable than in the clinical or prescreening setting (1,12,24). Although few studies have assessed the abilities of other providers who may participate in preparticipation health screenings (nurse practitioners, physician assistants, and athletic trainers) (17), we assume that the performance of these groups and others would be no better than that of the physicians studied.
Within this data set, no athletes were determined to have LQTS. Therefore, we are unable to determine the sensitivity and specificity for our proposed new screening method. Furthermore, without athletes with known LQTS, we cannot determine the true extent of misclassification when using the Seattle Criteria–recommended Bazett’s correction formula. Future studies with robust samples that include subjects with known LQTS are needed.
This study analyzed the ECG of young athletes (N = 2077) and determined that the current recommendations for QT interval measurement within the Seattle ECG Interpretation Criteria for Athletes are inadequate. Up to 75% of athletes possessing an uncorrected QT interval greater than 99% of the population adjusted for HR are not identified for LQTS investigation. We propose a new method of risk stratification that forgoes QT interval correction. Further study is needed to establish QT interval distribution thresholds in athletes and to develop appropriate risk stratifications for QT interval monitoring in this population for the prevention of SCD.
This study was unfunded. Victor Froelicher, MD, and David Hadley, PhD, are owners of CardeaScreen, which manufactures the device used within this study. Results of the present study do not constitute endorsement by the American College of Sports Medicine.
1. Al-Khatib SM, Allen LaPointe NM, Kramer JM, et al. A survey of health care practitioners’ knowledge of the QT interval. J Gen Intern Med
2. Bazett HC. An analysis of the time-relations of electrocardiograms. Heart
3. Chan A, Isbister GK, Kirkpatrick CMJ, Dufful SB. Drug-induced QT prolongation and torsades de pointes: evaluation of a QT nomogram. QJM
4. Corrado D, Biffi A, Basso C, Pelliccia A, Thiene G. 12-Lead ECG
in the athlete: physiological versus pathological abnormalities. Br J Sports Med
5. Drezner JA, Ackerman MJ, Anderson J, et al. Electrocardiographic interpretation in athletes: the ‘Seattle Criteria’. Br J Sports Med
6. Drezner JA, Ackerman MJ, Cannon BC, et al. Abnormal electrocardiographic findings in athletes: recognising changes suggestive of primary electrical disease. Br J Sports Med
7. Fossa AA, Wisialowski T, Magnano A, et al. Dynamic beat-to-beat modeling of the QT–RR interval relationship: analysis of QT prolongation during alterations of autonomic state versus human ether a-go-go-related gene inhibition. J Pharmacol Exp Ther
8. Hadley DM, Froelicher V, Wang PJ. A novel method for patient-specific QTc—modeling QT–RR hysteresis. Ann Noninvasive Electrocardiol
9. Harmon KG, Asif IM, Klossner D, Drezner JA. Incidence of sudden cardiac death in National Collegiate Athletic Association athletes. Circulation
10. Holly RG, Shaffrath JD, Amsterdam EA. Electrocardiographic alterations associated with the hearts of athletes. Sports Med
11. Isbister GK, Page CB. Drug induced QT prolongation: the measurement and assessment of the QT interval in clinical practice. Br J Clin Pharmacol
12. LaPointe NMA, Al-Khatib SM, Kramer JM, Califf RM. Knowledge deficits related to the QT interval could affect patient safety. Ann Noninvasive Electrocardiol
13. Makita N, Yagihara N, Crotti L, et al. Novel calmodulin mutations associated with congenital arrhythmia susceptibility. Circ Cardiovasc Genet
14. Malik M, Färbom P, Batchvarov V, Hnatkova K, Camm AJ. Relation between QT and RR intervals is highly individual among healthy subjects: implications for heart rate correction of the QT interval. Heart
15. Maron BJ, Thompson PD, Ackerman MJ, et al. Recommendations and considerations related to preparticipation screening for cardiovascular abnormalities in competitive athletes: 2007 update: a scientific statement from the American Heart Association Council on Nutrition, Physical Activity, and Metabolism: endorsed by the American College of Cardiology Foundation. Circulation
16. Mason JW, Ramseth DJ, Chanter DO, Moon TE, Goodman DB, Mendzelevski B. Electrocardiographic reference ranges derived from 79,743 ambulatory subjects. J Electrocardiol
17. Pickham D, Shinn JA, Chan GK, Funk M, Drew BJ. Quasi-experimental study to improve nurses’ QT-interval monitoring: results of QTIP study. Am J Crit Care
18. Pickham D, Zarafshar S, Sani D, Kumar N, Froelicher V. Comparison of three ECG
criteria for athlete pre-participation screening. J Electrocardiol
19. Puddu PE, Jouve R, Mariotti S, et al. Evaluation of 10 QT prediction formulas in 881 middle-aged men from the seven countries study: emphasis on the cubic root Fridericia’s equation. J Electrocardiol
20. Schwartz PJ, Crotti L, Insolia R. Long-QT syndrome: from genetics to management. Circ Arrhythm Electrophysiol
21. Schwartz PJ, Stramba-Badiale M, Crotti L, et al. Prevalence of the congenital long-QT syndrome. Circulation
22. Taggart NW, Haglund CM, Tester DJ, Ackerman MJ. Diagnostic miscues in congenital long-QT syndrome. Circulation
23. Viskin S, Postema PG, Bhuiyan ZA, et al. The response of the QT interval to the brief tachycardia provoked by standing: a bedside test for diagnosing long QT syndrome. J Am Coll Cardiol
24. Viskin S, Rosovski U, Sands AJ, et al. Inaccurate electrocardiographic interpretation of long QT: the majority of physicians cannot recognize a long QT when they see one. Heart Rhythm
25. Waring WS, Graham A, Gray J, Wilson AD, Howell C, Bateman DN. Evaluation of a QT nomogram for risk assessment after antidepressant overdose. Br J Clin Pharmacol