• „ Original article

Can we improve the accuracy of electrocardiographic algorithms for accessory pathway location in children?

Paola Ferrari1, Giovanni Malanchini1, Marco Racheli1, Gabriele Ferrari1, Cristina Leidi1, Michele Senni2, Paolo Della Bella3, Maurizio Malacrida4, Simone Gulletta3*, Paolo De Filippo1*

1Cardiac Electrophysiology and Pacing Unit, Papa Giovanni XXIII Hospital, Bergamo, Italy

2Cardiology Division, Cardiovascular Department, Papa Giovanni XXIII Hospital, Bergamo, Italy

3Arrhythmia Unit and Electrophysiology Laboratories, San Raffaele Scientific Institute, Milan, Italy

4Boston Scientific, Milan, Italy

*Both authors equally contributed to the study

Editorial

by Baranchuk et al.

To show differences between predicted and actual locations of accessory pathways, we considered 13 anatomical regions around the mitral and tricuspid annuli as the possible exact locations, as identified by EPS (Figure 1).

We then analyzed the possible locations predicted by each algorithm, which could include one or more of the 13 regions identified. To highlight these differences, we constructed a table (Table 1) indicating the relationship between the predicted locations and the actual ablation sites; the prediction was deemed correct if the pathway was ablated at one site in the same row of Table 1.

Descriptive statistics are reported as means (SD) for normally distributed continuous variables, or medians with 25th to 75th percentiles in the case of skewed distribution. The normality of distribution was tested through the non-parametric Kolmogorov-Smirnov test. Raw exact accuracy of correct prediction of each algorithm was assessed as the percentage of cases correctly assigned to one of the possible predicted locations by each algorithm. Corrected accuracy was calculated by division of raw accuracy the number of possible locations (13.8 and 13 respectively in the Arruda, Boersma, and Chiang algorithms). Bayesian probability (P[A|B]) considering the prediction of the algorithm attenuated by the “by chance” probability of the AP to be in a locations group was also calculated (Table 3A, B).

Without any correction, the algorithm by Boersma et al. proved the ablest to predict pathway locations that were found to be concordant with the site of successful AP ablation (Table 2).

Table 3A reports the accuracy of the 3 algorithms in predicting the location of left-sided APs vs. sites at risk for AV block. Pre-test probability (P) was calculated on the basis of the prevalence of accessory pathways in each of 3 groups of locations in our population. The algorithm by Chiang et al. displayed the greatest predictive power: when this algorithm predicted an AP location in one of these regions, there was a 72.17% chance that the pathway was located in that region. All 3 algorithms displayed a very high predictive power in locating the AP on the left side: 82.35%, 83.61%, and 83.93% probability for Arruda, Boersma, and Chiang, respectively.

We also calculated the probability of an AP being on the right or the left side (Table 3B), as predicted by each algorithm. The highest probability was achieved by the Arruda algorithm (91.4% when predicted on the right and 95.9% when predicted on the left). Similar values were obtained by applying the Boersma and Chiang algorithms.

Agreement among the algorithms

Figure 2 reports the inter-rater agreement among the 3 reference algorithms according to the identification of left-sided AP (Figure 2A) and the risk of AV block (Figure 2B).

With regard to left-sided AP identification, the strength of agreement was good between the Boersma and Chiang algorithms (κ = 0.611; 95% confidence interval [CI], 0.468–0.753), whereas it was moderate between Arruda and Chiang algorithms as well as between Arruda and Boersma algorithms (κ = 0.566, 95% CI, 0.419–0.713, and κ = 0.582; 95% CI, 0.438–0.727, respectively). Regarding sites at risk of AV block, the agreement was fairly good between Arruda and Chiang algorithms as well as between Boersma and Chiang algorithms (κ = 0.358; 95% CI, 0.195–0.520 and κ = 0.307; 95% CI, 0.192–0.422, respectively), whereas it was moderate between Arruda and Boersma algorithms (κ = 0.45; 95% CI, 0.304–0.597).

Diagnostic accuracy according to aggregated prediction score

We applied a step-down approach based on the concordance of the 3 algorithms in identifying both left-sided sites and sites at risk of AV block (Figure 3A, B).

All 3 algorithms identified the site as left-sided in 38 (31.7%) cases; 2 algorithms identified it as left-sided in 13 (10.8%) cases; 1 algorithm identified it as left-sided in 24 (20.0%) cases, and all algorithms identified the site as non-left-sided in 45 (37.5%) cases. Concordance (all algorithms reported the same result) was found in 69.1% (n = 83) of cases. On applying a first-step diagnostic evaluation when concordance was achieved, we were able to correctly identify left-sided or non-left-sided sites in 96.4% (n = 80) of cases (correctly identified as left-sided: 37/38 — 97.4%; correctly identified as non-left-sided: 43/45 — 95.6%). Regarding the evaluation of sites at risk of AV block, in 15% of cases (n = 18), all the 3 algorithms identified the predicted site as being at risk of AV block; in 22.5% of cases (n = 27), 2 algorithms identified it as being at risk of AV block; in 24.2% (n = 29), 1 algorithm identified it as at risk of AV block, and in 46 (38.3%) all algorithms identified the site as not being at risk of AV block. Concordance was observed in 53.3% (n = 64) of cases. On applying a first-step diagnostic evaluation when concordance was achieved, in 92.2% (n = 59) of cases we were able to correctly identify sites at risk or not at risk of AV block (correctly identified as being at risk of AV block: 15/18 — 93.3%; correctly identified as not at risk of AV block: 44/46 — 95.7%).

DISCUSSION

The 3 algorithms tested (those introduced by Arruda et al. [4], Boersma et al. [5], and Chiang et al. [6]) showed several differences and were constructed on relatively large populations of patients with ventricular pre-excitation. The sensitivity, specificity, and accuracy reported in the original studies ranged from 90% to 99%. However, the performance of these algorithms is reported to be reduced in children [14, 15]. In the present study, the overall accuracy of each of the selected algorithms was found to be higher than previously reported in children by Wren et al. [14]. Indeed, the mean accuracy of these algorithms was reported to be 39% in previous studies [14], with the highest value being achieved by the Boersma algorithm (reported accuracy of 48%). Our results are partly in line with these findings, but the mean accuracy of the 3 algorithms in our population was 62.23%. The greatest accuracy (71.7%), without any correction, was found using the Boersma algorithm, which was originally developed for use in children.

A practical implication of the present analysis is the division into 3 groups of locations around the mitral and tricuspid annuli: a left-sided group, a group of locations at risk of damaging the conduction system (AV block risk), and a right free-wall group (Figure 1). On considering these 3 subgroups, the algorithms by Arruda and Boersma seem to distinguish APs better than the one by Chiang (83.3% and 84.1% vs. 67.5%). These results could be regarded as discordant with previous studies [13, 14], which reported that the algorithms developed in adults displayed low accuracy in children. Nevertheless, we only included patients with a structurally normal heart, who have some similar characteristics to adults with manifest ventricular pre-excitation.

The most important implication of our analysis is that integrating the 3 algorithms could dramatically improve their predictive power. Indeed, while inter-algorithm agreement in the study population ranged from good to fair when all 3 algorithms were concordant in predicting the need (or not) for a left-sided (trans-septal or retrograde aortic) access, their predictive power rose to 97.4% and 95.6%. Similarly, the risk of AV block, which is considerable when the location of the pathway is right anterior, anteroseptal, or mid-septal, can be predicted or ruled out with 92.2% probability when the results of the 3 algorithms are integrated. This important finding could help to plan accurately the procedure of ablation of the accessory pathway in terms of selection of vascular access, catheter choice, and the type of energy delivery. Moreover, specific tools, such as echocardiographic monitoring and anesthesiologist support, can be made promptly available during the procedure. Moreover, it is noteworthy that the improvement in prediction accuracy was not achieved by adding a difficult or novel tool, but by using well-known algorithms that are already used in everyday clinical practice. The proposed approach can be applied to a vast majority of cases of pre-excitation where concordance between algorithms is present, in our analysis 69% of cases.

In a recently published study, the authors claimed that the inaccuracy of published algorithms for predicting accessory pathways in children was due to differences in ECG parameters [14]. They proposed a new algorithm, which was validated in a prospective cohort, to overcome the limitations of the earlier algorithms in identifying septal locations. However, while grouping together every septal pathway (from anteroseptal to postero-septal) may reflect ECG vectors, it has a limited value when considering the practical need for a left-sided approach and the risk of AV block in small-sized pediatric patients.

The simple and powerful added value of combining more than one algorithm to predict pathway location could also help to guide therapeutic indications and discussion with parents during electrophysiological consultation. Since the possible implications (malignant arrhythmias) and adverse events related to an accessory pathway ablation could be devastating, a good precision in planning and estimating the procedural risk is more than important. Moreover, this has become more relevant in the recent era, when a lower threshold for prophylactic ablation is considered by some electrophysiologists [19].

Indeed, when discussing the indications for EPS and ablation with the parents/caregivers of a young child, it is important to provide them with a clear description of the procedure, an estimate of its duration, and its possible complications. In addition, the availability of a more detailed plan of ablation could be very useful, even to expert operators, when dealing with children of small size and low weight.

This study has some limitations. Its retrospective nature makes any conclusion driven by this analysis hypothesis-generating. Therefore, prospective data are needed in the future to strengthen the feasibility and accuracy of the proposed integrated approach. Patients with multiple accessory pathways were excluded given the reported inaccuracy of multiple non-invasive methods in locating more than one accessory pathway. Secondly, the gold standard used for AP location was the site of ablation, which, in children, offers less accurate spatial discrimination than other methods, such as autopsy or direct visualization during open-heart surgery. Patients were treated with radiofrequency or cryo-ablation according to operator preference, and this may result in a different risk of damage to cardiac structures; however, the study is limited to algorithms accuracy, and no complications were reported. Moreover, different numbers of possible locations and differences in the nomenclature of regions around the AV-valvular-annuli make the comparison between algorithms liable to some assumptions and inherent errors. However, Table 1 details how we overcame these peculiarities. Moreover, for this reason, grouping accessory pathways into 3 different regions seems to be an attractive approach. Another limitation can be the choice of the 3 algorithms we used in the study, these are surely not all published in this field, but we believe that a more inclusive approach could be misleading, and, therefore, we selected the 3 as a reference and proved higher accuracy than observed in previous studies. The last limitation of the proposed method is the fact that it is applicable only if the concordance of the 3 referred algorithms is achieved. In our study, the algorithms concordance was found in 69% of cases — a vast majority of the study population.

CONCLUSIONS

In our multicenter experience involving a pediatric population, an aggregated prediction score based on 3 reference algorithms yielded a high level of precision in predicting an accessory pathway location. This approach could be used to safely plan invasive procedures and to target ablation sites and may be a valuable alternative to a single-algorithm prediction strategy.

Article information

Conflict of interest: MM is an employee of Boston Scientific, Inc. Other authors declare no conflict of interest.

Open access: This article is available in open access under Creative Common Attribution-Non-Commercial-No Derivatives 4.0 International (CC BY-NC-ND 4.0) license, allowing to download articles and share them with others as long as they credit the authors and the publisher, but without permission to change them in any way or use them commercially. For commercial use, please contact the journal office at kardiologiapolska@ptkardio.pl.

REFERENCES

1. 1. Brugada J, Katritsis DG, Arbelo E, et al. 2019 ESC Guidelines for the management of patients with supraventricular tachycardia. The Task Force for the management of patients with supraventricular tachycardia of the European Society of Cardiology (ESC). Eur Heart J. 2020; 41(5): 655–720, doi: 10.1093/eurheartj/ehz467, indexed in Pubmed: 31504425.
2. 2. Brugada J, Blom N, Sarquella-Brugada G, et al. European Heart Rhythm Association, Association for European Paediatric and Congenital Cardiology. Pharmacological and non-pharmacological therapy for arrhythmias in the pediatric population: EHRA and AEPC-Arrhythmia Working Group joint consensus statement. Europace. 2013; 15(9): 1337–1382, doi: 10.1093/europace/eut082, indexed in Pubmed: 23851511.
3. 3. Saul JP, Kanter R, Abrams D, et al. PACES/HRS expert consensus statement on the use of catheter ablation in children and patients with congenital heart disease. Heart Rhythm. 2016; 13(6): e251–e289, doi: 10.1016/j.hrthm.2016.02.009.
4. 4. Arruda MS, McClelland JH, Wang X, et al. Development and validation of an ECG algorithm for identifying accessory pathway ablation site in Wolff-Parkinson-White syndrome. J Cardiovasc Electrophysiol. 1998; 9(1): 2–12, doi: 10.1111/j.1540-8167.1998.tb00861.x, indexed in Pubmed: 9475572.
5. 5. Boersma L, García-Moran E, Mont L, et al. Accessory pathway localization by QRS polarity in children with Wolff-Parkinson-White syndrome. J Cardiovasc Electrophysiol. 2002; 13(12): 1222–1226, doi: 10.1046/j.1540-8167.2002.01222.x, indexed in Pubmed: 12521337.
6. 6. Chiang CE, Chen SA, Teo WS, et al. An accurate stepwise electrocardiographic algorithm for localization of accessory pathways in patients with Wolff-Parkinson-White syndrome from a comprehensive analysis of delta waves and R/S ratio during sinus rhythm. Am J Cardiol. 1995; 76(1): 40–46, doi: 10.1016/s0002-9149(99)80798-x, indexed in Pubmed: 7793401.
7. 7. d’Avila A, Brugada J, Skeberis V, et al. A fast and reliable algorithm to localize accessory pathways based on the polarity of the QRS complex on the surface ECG during sinus rhythm. Pacing Clin Electrophysiol. 1995; 18(9 Pt 1): 1615–1627, doi: 10.1111/j.1540-8159.1995.tb06983.x, indexed in Pubmed: 7491305.
8. 8. Fitzpatrick AP, Gonzales RP, Lesh MD, et al. New algorithm for the localization of accessory atrioventricular connections using a baseline electrocardiogram. J Am Coll Cardiol. 1994; 23(1): 107–116, doi: 10.1016/0735-1097(94)90508-8, indexed in Pubmed: 8277067.
9. 9. Iturralde P, Araya-Gomez V, Colin L, et al. A new ECG algorithm for the localization of accessory pathways using only the polarity of the QRS complex. J Electrocardiol. 1996; 29(4): 289–299, doi: 10.1016/s0022-0736(96)80093-8, indexed in Pubmed: 8913903.
10. 10. Xie B, Heald SC, Bashir Y, et al. Localization of accessory pathways from the 12-lead electrocardiogram using a new algorithm. Am J Cardiol. 1994; 74(2): 161–165, doi: 10.1016/0002-9149(94)90090-6, indexed in Pubmed: 8023781.
11. 11. Pambrun T, El Bouazzaoui R, Combes N, et al. Maximal Pre-Excitation Based Algorithm for Localization of Manifest Accessory Pathways in Adults. JACC Clin Electrophysiol. 2018; 4(8): 1052–1061, doi: 10.1016/j.jacep.2018.03.018, indexed in Pubmed: 30139487.
12. 12. Moskal P, Bednarski A, Kiełbasa G, et al. Increased preexcitation on electrocardiography improves accuracy of algorithms for accessory pathway localization in Wolff-Parkinson-White syndrome. Kardiol Pol. 2020; 78(6): 567–573, doi: 10.33963/KP.15378, indexed in Pubmed: 32438794.
13. 13. Maden O, Balci KG, Selcuk MT, et al. Comparison of the accuracy of three algorithms in predicting accessory pathways among adult Wolff-Parkinson-White syndrome patients. J Interv Card Electrophysiol. 2015; 44(3): 213–219, doi: 10.1007/s10840-015-0057-6, indexed in Pubmed: 26384677.
14. 14. Wren C, Vogel M, Lord S, et al. Accuracy of algorithms to predict accessory pathway location in children with Wolff-Parkinson-White syndrome. Heart. 2012; 98(3): 202–206, doi: 10.1136/heartjnl-2011-300269, indexed in Pubmed: 21917661.
15. 15. Baek SM, Song MiK, Uhm JS, et al. New algorithm for accessory pathway localization focused on screening septal pathways in pediatric patients with Wolff-Parkinson-White syndrome. Heart Rhythm. 2020; 17(12): 2172–2179, doi: 10.1016/j.hrthm.2020.07.016, indexed in Pubmed: 32681992.
16. 16. Teixeira CM, Pereira TA, Lebreiro AM, et al. Accuracy of the electrocardiogram in localizing the accessory pathway in patients with Wolff-Parkinson-White pattern. Arq Bras Cardiol. 2016; 107(4): 331–338, doi: 10.5935/abc.20160132, indexed in Pubmed: 27627222.
17. 17. Bar-Cohen Y, Khairy P, Morwood J, et al. Inaccuracy of Wolff-Parkinson-white accessory pathway localization algorithms in children and patients with congenital heart defects. J Cardiovasc Electrophysiol. 2006; 17(7): 712–716, doi: 10.1111/j.1540-8167.2006.00467.x, indexed in Pubmed: 16836664.
18. 18. Cohen J. A coefficient of agreement for nominal scales. Educa- tional and Psychological Measurement. 1960; 20(1): 37–46, doi: 10.1177/001316446002000104.
19. 19. Fleiss JL, Levin B, Paik MC. Statistical methods for rates and proportions, 3rd ed. John Wiley & Sons, Hoboken 2003.
20. 20. Moskal P, Jastrzębski M, Pitak M, et al. Malignant ventricular arrhythmias and other complications of untreated accessory pathways: an analysis of prevalence and risk factors in over 600 ablation cases. Kardiol Pol. 2020; 78(3): 203–208, doi: 10.33963/KP.15161, indexed in Pubmed: 31994490.