Variabilityininterpretationofthe electrocardiogram in young athletes: an unrecognized obstacle for electrocardiogrambased screening protocols

Transcription

1 Europace (2015) 17, doi: /europace/euu385 CLINICAL RESEARCH Electrocardiology and risk stratification Variabilityininterpretationofthe electrocardiogram in young athletes: an unrecognized obstacle for electrocardiogrambased screening protocols Benjamin Berte 1, Mattias Duytschaever 1,2, Juliana Elices 1, Vikas Kataria 1, Liesbeth Timmers 2,Frédéric Van Heuverswyn 2, Roland Stroobandt 2, Jan De Neve 3, Karel Watteyne 3, Elke Vandensteen 3, Yves Vandekerckhove 1, and Rene Tavernier 1 * 1 Department of Cardiology, Sint-Jan Hospital Bruges, Ruddershove 10, 8000 Bruges, Belgium; 2 Heart Center, Ghent University Hospital, Ghent, Belgium; and 3 Department of Rehabilitation and Sports Medicine, Sint-Jan Hospital Bruges, Bruges, Belgium Received 12 July 2014; accepted after revision 8 December 2014; online publish-ahead-of-print 6 February 2015 Aims To assess in young athletes (i) the variability in the percentage of abnormal electrocardiograms (ECGs) using different criteria and (ii) the variability in ECG interpretation among cardiologists and sport physicians.. Methods Electrocardiograms of 138 athletes were categorized by seven cardiologists according to the original European Society of and Results Cardiology (ESC) criteria by Corrado (C), subsequently modified by Uberoi (U), Marek (M), and the Seattle criteria (S); seven sports physicians only used S criteria. The percentage of abnormal ECGs for each physician was calculated and the percentage of complete agreement was assessed. For cardiologists, the median percentage of abnormal ECGs was 14% [interquartile range (IQR) %] for C, 11% (IQR %) for U [not significant (NS) compared with C], 11% (IQR 10 13%) for M (NS compared with C), and 7% (IQR 5 8%) for S (P, compared with C); complete agreement in interpretation was 64.5% for C, 76% for U (P, 0.05 compared with C), 74% for M (NS compared with C), and 84% for S (P, compared with C). Sport physicians classified a median of 7% (IQR 7 11%) of ECGs as abnormal by S (P ¼ NS compared with cardiologists using S); complete agreement was 72% (P, 0.05 compared with cardiologists using S).. Conclusion Seattle criteria reduced the number of abnormal ECGs in athletes and increased agreement in classification. However, variability in ECG interpretation by cardiologists and sport physicians remains high and is a limitation for ECG-based screening programs Keywords ECG Screening Athletes Introduction The European Society of Cardiology (ESC) guidelines recommend performing pre-participation screening in athletes with a questionnaire, a physical examination, and an ECG. 1 The usefulness of screening programmes in athletes to prevent sudden cardiac death (SCD) is under debate for several reasons. Although sudden death in athletes is a major issue in the public forum, the estimated incidence of sudden death in athletes is very low (1 3/ /person-years). 2,3 In this particular setting, a useful screening programme to reduce the number of SCDs should be highly sensitive. On the other hand, to avoid that athletes are refrained from sports participation without a justified cause and to limit the financial burden imposed by additional investigations, specificity should be high as well. In this respect, the * Corresponding author. Tel: ; fax: , address: rene.tavernier@azsintjan.be Dr Juliana Elices received an EHRA fellowship grant for Clinical Electrophysiology with emphasis on catheter ablation. Dr Vikas Kataria received an International EHRA fellowship grant for Clinical Electrophysiology with emphasis on catheter ablation. Published on behalf of the European Society of Cardiology. All rights reserved. & The Author For permissions please journals.permissions@oup.com.

2 1436 B. Berte et al. What s new The present study in young athletes shows that (1) The variability in the percentage of ECG abnormalities using different criteria is high. (2) The variability in ECG interpretation using the same criteria among cardiologists and sport physicians is considerable. (3) The Seattle criteria result in the lowest prevalence of ECG abnormalities, the lowest variability in ECG classification, and a moderate overall agreement in ECG interpretation among all physicians. These novel findings are important when discussing the usefulness of ECG-based screening programmes to prevent sudden death in young athletes. initial 2005 ESC guidelines related to ECG interpretation have been updated in 2010 by Corrado et al. and modified in 2011 by Uberoi et al. and Marek et al. 4 6 More recently, Drezner et al. 7 suggested to increase the minimum screening age from 12 to 14 years and introduced stricter less ambiguous cut-off values for all individual criteria, especially for those related to T-wave inversion, ST depression, and pathological Q waves ( Seattle criteria ). Although not investigated prospectively, it is expected that these modifications increase specificity (without compromising sensitivity) and render ECG interpretation of athletes by cardiologists and sport physicians straightforward and less variable. No prior study systematically analysed the variability in ECG classifications using different criteria. The aim of our study was to assess prospectively in young athletes (i) the variability in percentage of abnormal ECGs using thedifferentproposed criteria and to assess(ii) the variability in ECG interpretation among cardiologists and sport physicians. Methods Study population and screening protocol From April till November 2012, all the youth players of a first division Belgian soccer team {n ¼ 138, all males, median age 14 years [12 18, interquartile range (IQR) 12 16], Caucasian n ¼ 100, Black African n ¼ 27, Maghrebian n ¼ 11} were screened using the American Heart Association (AHA)-questionnaire (QST), a physical examination (PE) performed by a cardiologist (BB), and an ECG. The ECGs were categorized as normal or abnormal by consensus by a panel of three electrophysiologists (MD, RT, and BB) using the ESC 2010 criteria. Further disease-specific examinations (echocardiography, 24 h ECG recording, and exercise testing) were performed if any abnormality was found. All participants gave informed consent. The study was funded by a private sponsorship. Electrocardiogram interpretation To analyse the impact of different ECG criteria on the percentage of abnormalities and to analyse the agreement in ECG interpretation among cardiologists, we instructed during a half-a-day seminar the different ECG screening criteria [Corrado et al. (ESC), Uberoi et al., Marek et al., and Drezner et al. (Seattle)] and provided them in tabular format to seven cardiologists [three cardiologists working in an academic centre actively involved in an ECG-based screening programme, one general cardiologist with 20 years of experience, one cardiologist working in a private hospital experienced in paediatric ECG interpretation, and two European Heart Rhythm Association (EHRA)-certified cardiologists] (see Appendix) and asked them to categorize each ECG as normal or abnormal accordingly. Furthermore, seven sport physicians involved in ECG-based screening protocols categorized each ECG according to the Seattle criteria using the same table. Statistical analysis For statistical analysis, GraphPad Prism version 5.0c was used. Continuous variables are expressed as mean and standard deviation or median and IQR according to their distribution. Categorical variables are expressed as absolute numbers and percentages. Categorical variables were compared using Fisher s exact test, and paired test was used when appropriate. Analysis of variance with Dunnetts correction using the Corrado data as control data was used for multiple comparisons. Kappa statistics were used to assess the agreement in ECG interpretation between different physicians. Results Screening results Patient demographics are given in Table 1. Out of 138 young athletes, 24(17%) had abnormal findings: abnormal QST n ¼ 5(3.6%) (familial history n ¼ 4, palpitations n ¼ 1), abnormal PE n ¼ 8(5.8%) (all systolic murmur 2/6), and abnormal ECG n ¼ 16(11.6%) [T-wave inversion n ¼ 10, abnormal Q wave n ¼ 5, right ventricular hypertrophy n ¼ 1, ST depression n ¼ 1, left anterior hemiblock (LAHB) n ¼ 1 and manifest pre-excitation n ¼ 1]. The prevalence of any ECG abnormality [19/138(14%)] was not different between athletes 14 years [10/79(13%)] and younger [9/59(15%)]. Echocardiography excluded underlying structural heart disease in all 24 patients. The athlete with pre-excitation was asymptomatic and pre-excitation was only intermittently present. Based upon these results, nobody was excluded from sports participation. Percentage of abnormal electrocardiograms using the different electrocardiogram criteria The results are given in Table 2. For all cardiologists and for all criteria, the percentage of abnormal ECGs ranged from 3 to 25%, with a median of 11% (IQR 8 13%). Overall, the percentage of abnormal ECGs decreased from a median of 14% (IQR %) using the Corrado criteria (C) to a median of 7% (IQR 5 8%) using the Table 1 Baseline characteristics N or Mean + SD... Age (years) Gender (M/F) 138/0 Caucasian/Magrebian/Black 100/27/11 Heart rate (bpm) PR (ms) QRS (ms) Axis (8) QTc (ms)

3 Variability in interpretation of the ECG in young athletes 1437 Seattle criteria (S) (P, 0.005). This reduction was seen for each cardiologist. In contrast to the Seattle criteria, the criteria as suggested by Uberoi (U) and Marek (M) did not result in a significant reduction in the percentage of abnormal ECGs. For all sport physicians, the percentage of abnormal ECGs using the Seattle criteria ranged from 3 to 19% with a median of 7% (IQR 7 11%) [P ¼ not significant (NS) compared with cardiologists using Seattle criteria]. Abnormal electrocardiogram findings using Corrado and Seattle criteria In Table 3 we show for each ECG category, the number of abnormal ECGs according to Corrado and Seattle criteria. The reduction in abnormal ECG findings using Seattle criteria was mainly related to less T-wave inversion [median 9 (IQR 8 10) vs. median 5 (IQR 4 6), P, 0.005]. Interestingly, none of the ECGs of Black athletes was considered as abnormal by the ESC or the Seattle criteria based upon T-wave inversion. Less abnormal T-wave inversion using Seattle criteria was exclusively attributed to the classification of T-wave inversion in V1 and V2 as a normal pattern. The other reasons for the reduction in abnormal ECG findings were less short QTc [median 4 (IQR 0 9) vs. median 0 (IQR 0 0), P, 0.05], and less QRS prolongation [median 4 (IQR0 5)vs.median0(IQR0 0),P, 0.05]. Trends in reduction were similar for each individual cardiologist. Agreement in electrocardiogram interpretation among and between cardiologists and sport physicians The results are summarized in Figure 1 and Table 4. The overall agreement in ECG interpretation (i.e. normal or abnormal by each Table 2 Percentage of abnormal ECGs according to the different criteria (among cardiologists and sport physicians) % of abnormal ECGs % of abnormal ECGs % of abnormal ECGs % of abnormal ECGs % of abnormal ECGs by Corrado by Uberoi by Marek by Seattle by Seattle... CA SP 1 7 CA SP 2 7 CA SP 3 11 CA SP 4 7 CA SP 5 7 CA SP 6 19 CA SP 7 3 Median * Median 7 IQR IQR 7 11 Range Range 3 19 CA, cardiologist; SP, sport physician; IQR, interquartile range. *P, compared with Corrado. Table 3 Number of ECG abnormalities for each cardiologist Cardiologist... P-value Median (IQR)... C S C S C S C S C S C S C S C S... TWI (n) (8 10) 5 (4 6),0.005 Q waves (n) (2 3) 2(2 3) NS Short QTc (n) (0 9) 0(0 0),0.05 ST depression (n) (1 3) 1(0 1) NS prolonged QRS (n) (0 5) 0(0 0),0.05 WPW (n) (1 1) 1(1 1) NS AVB I (n) (0 0) 0(0 0) NS AVB II (n) (0 1) 0(0 1) NS RVH (n) (0 1) 0(0 0) NS LAHB (n) (0 1) 0(0 1) NS PVC (n) (0 1) 0(0 1) NS TWI, T-wave inversion; AVB, atrioventricular block; RVH, right ventricular hypertrophy; LAHB, left anterior hemiblock; PVC, premature ventricular contraction; C, Corrado criteria; S, Seattle criteria; IQR, interquartile range; WPW, Wolff Parkinson White pattern.

4 1438 B. Berte et al. ECG interpretation by cardiologists (n = 7) ECG interpretation using Seattle 100% NL by all ABNL by all Disagreement 100% NL by all ABNL by all Disagreement % of ECGs 75% 50% 25% 0% Corrado Uberoi Marek Seattle Figure 1 Electrocardiogram interpretation by cardiologists. Table 4 Kappa values for different ECG classification criteria among cardiologists and sport physicians Kappa-value P-value Lower Upper CI CI... Corrado CA 0.37, Uberoi CA 0.50, Marek CA 0.49, Seattle CA 0.45, Seattle SP 0.28, Seattle CA + SP 0.34, CA, cardiologists; Sport, sport physicians; CI, confidence interval. of the seven cardiologists) was 64.5% for Corrado, 76% for Uberoi (P, 0.05 compared with C), 74% for Marek (NS compared with C), and 84% for Seattle criteria (P, compared with C). This was exclusively related to higher agreement among all cardiologists on normality (62 83%, P, for S) without any increase in agreement on abnormality (2 and 1.5% for S, P ¼ NS). On the other hand, the agreement between each of the seven sport physicians using Seattle criteria was only 72% (96/138, P, 0.05 compared with cardiologists using Seattle criteria) (Figure 2). Overall, the agreement between cardiologists and sport physicians(n ¼ 14) using theseattle criteria was 65% (91/138). The kappa values ranged from 0.28 to 0.49 (Table 4). Discussion Main findings The present study in young athletes shows that (i) variability in the percentage of ECG abnormalities using different criteria is high, (ii) variability in ECG interpretation using the same criteria among cardiologists and sport physicians is considerable, and (iii) Seattle criteria result in the lowest prevalence of ECG abnormalities, the lowest variability in ECG classification, and a moderate overall agreement in ECG interpretation among all physicians. These novel findings are important when discussing the usefulness of ECG-based screening programmes to prevent sudden death in young athletes. % of ECGs 75% 50% 25% 0% Cardiologists (n =7) Sport physicians (n =7) Cardiologists and sport physicians (n =14) Figure 2 Electrocardiogram interpretation by cardiologists and sport physicians using Seattle criteria. The European Society of Cardiology criteria by Corrado: prevalence of abnormal electrocardiograms and inter-observer variability In 2000, Pelliccia et al. 8 studied 1005 electrocardiograms of highly trained athletes and classified 402 ECGs as abnormal (40%). These included 233/402 ECGs (80%) with an isolated increase in QRS voltage and 59/402 ECGs (20%) with an early repolarization pattern. In the Recommendations for Interpretation of 12-lead electrocardiogram in the athlete (as proposed by Corrado and the ESC in 2010), these ECG abnormalities are now classified as normal (common and training related). This resulted in a decrease from 40 to 11% abnormal ECGs. 4 This latter percentage is in line with the prevalence of abnormal ECGs using the Corrado criteria in our population (11% after consensus between three electrophysiologists and a median of 14%, range 9 25% between seven cardiologists). The trend for a higher prevalence in our population could be related to differences in population characteristics (age, ethnicity, and gender) and the unique sports activity. No prior study addressed the inter-observer variability using the Corrado criteria. We observed a low overall agreement in ECG interpretation of 64.5%. This finding was paralleled by a wide IQR in % of abnormal ECGs ( %). These observations indicate a considerable inter-observer variability that can be related to the inherent difficulty in ECG interpretation, especially in athletes but probably also due to the subjectivity of the criteria proposed by the ESC. The modified criteria: influence on the prevalence of electrocardiogram abnormalities and inter-observer variability To reduce the number of abnormal ECGs and in an effort to identify not all ECG abnormalities but only those with findings suggestive of disorders associated with sudden cardiac death, modified criteria have been put forward. Uberoi and Marek introduced modification

5 Variability in interpretation of the ECG in young athletes 1439 of the Corrado criteria primarily focused on more stringent definitions of abnormalities related to Q waves, right ventricular hypertrophy, T-wave inversion, ST depression, and QT interval. 5,6 Marek found in a retrospective cohort study that out of US high school students aged between 14 and 19 years only 817 or 2.5% had abnormal ECGs requiring further evaluation. In our study there was a trend to a lower % of abnormal ECGs using the criteria proposed by Uberoi and Marek but still reached 11 and 11%, respectively, with an IQR of % and 10 13%. These relatively wide IQR are reflected in an overall agreement in ECG interpretation of only 76 and 74%, respectively. It is expected that the Seattle criteria, with more stringent criteria for QT interval abnormalities, abnormal T-wave inversion, and QRS prolongation would further decrease the percentage of abnormal ECGs to around 4%. 7 Overall, our prospective data confirm that these criteria indeed decrease the prevalence of abnormal ECG findings (especially due to a decrease in the prevalence of abnormal T-waves, short QT intervals, and prolonged QRS duration) in young athletes but to a lesser degree than expected (7%). This is in parallel with a recent study by Berge et al. evaluating 587 soccer players. 9 The prevalence of abnormal ECGs was 29.3% according to the ESC s recommendations and 11.2% using the Seattle criteria. Sheikh et al. found that Seattle criteria reduced abnormal ECGs to 18.4% in Black athletes and 7.1% in White athletes (vs and 16.2%, respectively, using ESC criteria). 10 Our study also shows that these criteria not only decrease the prevalence in abnormal ECGs but also result in a higher overall agreement in ECG interpretation of 84% among cardiologists. Agreement in electrocardiogram interpretation among physicians No systematic analysis of variability using C, U, M, and S was performed before. We showed that overall agreement in ECG classification of athletes went up from 64.5% using the Corrado criteria to 84% using the Seattle criteria, but that the overall agreement using the Seattle criteria went down again from 84% to 72 65% depending on the expert level of physician and the total number of physicians. Kappa statistics, however, indicate that overall agreement in ECG classification can maximally be considered as moderate irrespective of the criteria used. Variability in ECG interpretation has been studied in other populations. Drezner asked four groups of physicians (cardiologists, sport physicians, primary care attendings, and primary care residents) to analyse 40 previously selected ECGs (28 normal and 12 truly abnormal ECGs, including long QT syndrome n ¼ 2, Brugada type 1 n ¼ 1, Wolff Parkinson White syndrome n ¼ 2, arrhythmogenic right ventricular cardiomyopathy n ¼ 1, hypertrophic cardiomyopathy n ¼ 5, and left ventricular non-compaction n ¼ 1) using an online ECG-interpretation tool. 11 Cardiologists classified 96% of the ECGs correctly and sport physicians 91% of the ECGs (P ¼ NS). These data suggest a high agreement in ECG interpretation among physicians. These data are in contrast to our data and reports by Hill et al. 12,13 In a first study by Hill et al., 53 members of the Western Society of Paediatric Cardiology were asked to classify a series of 18 ECGs that represented conditions causing paediatric sudden cardiac death (1 with long QT syndrome; 4 with hypertrophic cardiomyopathy; 2 with Wolff Parkinson White syndrome; 1 with pulmonary arterial hypertension; 2 with myocarditis) or normal hearts (n ¼ 8 with some demonstrating common findings for athletic hearts). 12 The average percentage of correct ECG interpretations per respondent was 69% but the range of 34 98% indicated a high variability like in our study. A more recent study from the same group showed that paediatric electrophysiologists did not interpret screening ECGs more accurately than paediatric cardiologists with average number of correct ECG interpretations of 73 and 69%, respectively. 13 Most likely the observed differences are largely explained by differences in study design and methodology. Our population consisted of an unselected prospective group of athletes where a high prevalence of normal ECGs is anticipated. In contrast, Drezner used a group of ECGs with an artificial high prevalence of clearly abnormal ECGs not representative for findings in a screening population. In this respect, it is interesting to note that in the study by Drezner, the normal ECGs were most often incorrectly categorized suggesting that in athletes, the difficulty lies in the identification of the normal ECG and not in the recognition of the clearly abnormal ECG. Consequences of variability in electrocardiogram interpretation Our study suggests that depending upon the screening physician and the used criteria, up to 20% of screened athletes could be referred for variable additional tests. This questions the value of screening programmes. From a society point of view, these data indicate that the workload and financial burden of a screening program can vary markedly. From a patients point of view, it could be argued thatthis is not important as long as all abnormalities are detected. Our data fail to answer this question since we only performed additional examination to exclude underlying heart disease in 24/138 athletes. The data from Hill, however, raise concern since abnormal ECGs were often classified as normal. 12 Furthermore, one should not ignore the additional stress imposed on an athlete by additional screening. On top of this, additional investigations will not prevent all sports-related sudden deaths nor prevent the unnecessary exclusion of athletes from sport participation. Finally, it should also be noted that, up until now, no prospective data are available, which confirm that the application of modified Corrado criteria or Seattle criteria not only reduce the number of ECGs considered as abnormal but also will still detect all abnormalities and reduce the number of sudden deaths in athletes. Limitations of the study Only ECGs of male young athletes were studied. Before puberty ECG interpretation can be more cumbersome and one can wonder if a 13-year-old boy involved in competitive soccer can be considered as a highly trained athlete. The prevalence of ECG abnormalities in an older cohort performing other sports (e.g. endurance sport) may have been different and could influence the reproducibility of ECG interpretation. We did not assess the presence of structural heart disease in each participant and therefore no gold standard is available. However, the aim of our study was to assess variability in ECG interpretation. Finally, we tested the variability in ECG interpretation

6 1440 B. Berte et al. among seven cardiologists and seven sport physicians. As the number of physicians interpreting the ECG will further increase, it is logic that the % of overall agreement will further decrease. Conclusions The high variability in ECG interpretation using the Corrado, Uberoi, Marek, or Seattle criteria questions their usefulness in ECG-based screening protocols. Application of the Seattle criteria offers the advantage of lowest albeit still considerable variability in ECG interpretation and one should acknowledge that the prospective value of these criteria has not been addressed. Conflict of interest: none declared. Funding This work was supported by The Club Academy of Football Club Brugge. References 1. Corrado D, Pelliccia A, Bjornstad HH, Vanhees L, Biffi A, Borjesson M et al. Cardiovascular pre-participation screening of young competitive athletes for prevention of sudden death: proposal for a common European protocol. Eur Heart J 2005;26: Maron BJ, Haas TS, Ahluwalia A, Rutten-Ramos SC. Incidence of cardiovascular sudden deaths in Minnesota high school athletes. Heart Rhythm 2013;10: Holst AG, Winkel BG, Theilade J, Kristensen IB, Thomsen JL, Ottesen GL et al. Incidence andetiologyofsports-related sudden cardiacdeathindenmark implications for preparticipation screening. Heart Rhythm 2010;7: CorradoD, PellicciaA, Heidbuchel H, Sharma S, Link M, BassoCet al. Recommendations for interpretation of 12-lead electrocardiogram in the athlete. Eur Heart J 2010; 31: Uberoi A, Stein R, Perez MV, Freeman J, Wheeler M, Dewey F et al. Interpretation of the electrocardiogram of young athletes. Circulation 2011;124: Marek J, Bufalino V, Davis J, Marek K, Gami A, Stephan W et al. Feasibility and findings of large-scale electrocardiographic screening in young adults: data from 32,561 subjects. Heart Rhythm 2011;8: Drezner JA, Ackerman MJ, Anderson J, Ashley E, Asplund CA, Baggish AL et al. Electrocardiographic interpretation in athletes: the Seattle Criteria. Br J Sports Med 2013;47: Pelliccia A, Maron BJ, Culasso F, Di Paolo FM, Spataro A, Biffi A et al. Clinical significance of abnormal electrocardiographic patterns in trained athletes. Circulation 2000;102: Berge HM, Gjesdal K, Andersen TE, Solberg EE, Steine K. Prevalence of abnormal ECGs in male soccer players decreases with the Seattle criteria, but is still high. Scand J Med Sci Sports 2014 Jun 24 [Epub ahead of print]. 10. Sheikh N, Papadakis M, Ghani S, Zaidi A, Gati S, Adami PE et al. Comparison of electrocardiographic criteria for the detection of cardiac abnormalities in elite Black and White athletes. Circulation 2014;129: Drezner JA, Asif IM, Owens DS, Prutkin JM, Salerno JC, Fean R et al. Accuracy of ECG interpretation in competitive athletes: the impact of using standardised ECG criteria. Br J Sports Med 2012;46: Hill AC, Miyake CY, Grady S, Dubin AM. Accuracy of interpretation of preparticipation screening electrocardiograms. J Pediatr 2011;159: Harbison AL, Hill AC, Motonaga KS, MiyakeCY, Dubin AM. Do pediatric electrophysiologists readpre-participation screeningelectrocardiograms moreaccurately than general pediatric cardiologists? J Pediatr 2013;163: