Elsevier

Heart Rhythm

Volume 12, Issue 1, January 2015, Pages 130-136
Heart Rhythm

Modest agreement in ECG interpretation limits the application of ECG screening in young athletes

https://doi.org/10.1016/j.hrthm.2014.09.060Get rights and content

Background

Athlete ECG screening has been recommended by several international sporting bodies; however, a number of controversies remain regarding the accuracy of ECG screening. An important component that has not been assessed is the reproducibility of ECG interpretation.

Objective

The purpose of this study was to assess the variability of ECG interpretation among experienced physicians when screening a large number of athletes.

Methods

A sports cardiologist, a sports medicine physician, and an electrophysiologist analyzed 440 consecutive screening ECGs from asymptomatic athletes and were asked to classify the ECGs according to the 2010 European Society of Cardiology criteria as normal (or demonstrating training related ECG changes) or abnormal. When an abnormal ECG was identified, they were asked to outline what follow-up investigations they would recommend.

Results

The reported prevalence of abnormal ECGs ranged from 13.4% to 17.5%. Agreement on which ECGs were abnormal ranged from poor (κ = 0.297) to moderate (κ = 0.543) between observers. Suggested follow-up investigations were varied, and follow-up costs ranged from an additional A$30–A$129 per screening episode. Neither of the 2 subjects (0.45%) in the cohort with significant pathology diagnosed as a result of screening were identified correctly by all 3 physicians.

Conclusion

Even when experienced physicians interpret athletes’ ECGs according to current standards, there is significant interobserver variability that results in false-positive and false-negative results, thus reducing the effectiveness and increasing the social and economic cost of screening.

Introduction

Preparticipation screening inclusive of a 12-lead ECG is recommended by the European Society of Cardiology (ESC) for all young persons engaging in competitive sport1 in an effort to reduce the incidence of sudden cardiac death, as reported in the Veneto region of Italy after mandated screening.2 Although the strength of this evidence is still debated, many sporting bodies are implementing cardiac screening inclusive of an ECG.

The 12-lead ECG is a relatively cheap screening tool; however, the accuracy of the test is entirely dependent on individual interpretation of the pathologic findings, and its cost efficacy is dictated by a cascade of subsequent investigations to elucidate underlying pathology. There has been recent focus on refinement of criteria for ECG interpretation in athletes in an effort to reduce false-positive results of screening.3, 4, 5 However, there is a paucity of data on the extent of variability in ECG interpretation among experienced physicians when faced with a real-world sample of athletes’ ECGs in which the incidence of underlying pathology is low. Therefore, we sought to assess the variability of ECG interpretation among experienced physicians and their ability to identify true abnormal results when faced with a large number of screening ECGs.

Section snippets

Methods

The first 440 consecutively collected ECGs in our prospective study, which commenced in June 2011, were used for analysis.6 ECGs were deliberately unfiltered and nonenriched so that the prevalence of disease approximated that which could be expected in an athletic cohort. The inclusion and exclusion criteria have been described in detail elsewhere.6, 7 In brief, all subjects were elite athletes (age 16–35 years) and were not known to have any preexisting cardiac conditions.

These 440

Results

The 440 elite athletic subjects were aged 21 ± 5 years (range 16–35 years), and 36 (8.2%) were female. The cohort was predominantly (81.6%) Caucasian, with 14.5% indigenous Australian, Torres Strait Islander, or Pacific Islander/Maori, 1.6% Asian, and 2.3% African/South American.

Discussion

This study conveys important new information on the variability of detection of pathologic ECG abnormalities in a large unselected cohort of athletes undergoing ECG screening. The main finding was that interphysician variability of en masse athlete ECG interpretation is high, with at best moderate agreement between physicians experienced in athlete ECG interpretation. Suggested resultant investigations were many and varied, and the small number of athletes with cardiac pathology was missed on

Conclusion

There is significant variability in ECG interpretation, even among physicians with prior experience in athlete ECG analysis. This limitation must be considered before implementing population-based ECG screening for young asymptomatic athletes.

Although guidelines have aided in the recognition of normal physiologic ECG findings in athletes, we demonstrated that even among experienced physicians, agreement on what constitutes an abnormal ECG is poor. Unlike other screening tools with laboratory

References (20)

There are more references available in the full text version of this article.

Cited by (46)

  • Detection of Hypertrophic Cardiomyopathy Using a Convolutional Neural Network-Enabled Electrocardiogram

    2020, Journal of the American College of Cardiology
    Citation Excerpt :

    The large-scale model creation and testing reported herein represents the first application of a multilayer CNN that was truly agnostic to any specific ECG features during its development. Several sets of ECG criteria for HCM in young athletes have been proposed with an emphasis toward distinguishing athletic adaptation and normal ECG variants from HCM, including the European (16), Stanford (17), and Seattle criteria (18), yet follow-up studies and attempts for external validation of these criteria have revealed inconsistencies in their diagnostic performance (19–22). However, the use of an AI-based deep learning network has the potential to overcome these challenges as it is not confined to any “classic” ECG criteria and it is trained to detect even the most subtle ECG patterns associated with structural changes in HCM that may be undetectable by the human eye or by traditional automated algorithms.

View all citing articles on Scopus

Dr. Brosnan is the recipient of a National Health and Medical Research Council (NHMRC) PhD scholarship. Dr. LaGerche is the recipient of an NHMRC postdoctoral scholarship. Dr. Kumar is the recipient of a Neil Hamilton Fairley Overseas Research Scholarship co-funded by the National Heart Foundation of Australia and NHMRC. Dr. Kalman is the recipient of an NHMRC practitioner fellowship.

View full text