Author + information
- Received October 29, 2008
- Revision received January 27, 2009
- Accepted February 3, 2009
- Published online June 23, 2009.
- Esteban González-Torrecilla, MD, PhD⁎ (, )
- Jesús Almendral, MD, PhD,
- Angel Arenal, MD,
- Felipe Atienza, MD, PhD,
- Leonardo F. Atea, MD,
- Silvia del Castillo, MD and
- Francisco Fernández-Avilés, MD, PhD
- ↵⁎Reprint requests and correspondence:
Dr. Esteban González-Torrecilla, Electrophysiology Unit, Cardiology Department, Hospital General Universitario Gregorio Marañón, Doctor Esquerdo 46, 28007 Madrid, Spain
Objectives The aim of this study was to assess the independent predictive contribution to the electrocardiogram (ECG) of bedside clinical variables to distinguish the major forms of paroxysmal supraventricular tachycardias.
Background Atrioventricular nodal re-entrant tachycardias (AVNRTs) and orthodromic reciprocating tachycardias (ORTs), through concealed accessory pathways, are major mechanisms of paroxysmal atrioventricular re-entrant tachycardias.
Methods We prospectively included 370 consecutive patients undergoing an electrophysiologic study for paroxysmal, regular, narrow-QRS complex tachycardias without pre-excitation in sinus rhythm. A diagnostic interpretation of ECG recordings was performed by 2 observers blinded to invasive diagnosis used as gold standard. The independent diagnostic contribution of basic clinical variables from a 7-item questionnaire was analyzed alone and in combination with the ECG interpretation by stepwise logistic regression.
Results AVNRTs and ORTs were demonstrated in 262 and 108 patients, respectively. Age at symptom onset (odds ratio [OR]: 1.27), presence of palpitations in the neck (OR: 3.54), and female sex (OR: 2.96) (all p = 0.0001) were the clinical variables with significant diagnostic power for AVNRT diagnosis. These variables were selected by the logistic model as predictors of the tachycardia diagnosis when the ECG interpretation was included in the analysis (C statistic = 0.81 vs. 0.75 with clinical variables alone; p = 0.003). Neck palpitation was the only predictor of AVNRT when positive ECG findings were lacking.
Conclusions Age at the onset of symptoms, sensation of rapid regular pounding in the neck during tachycardia, and female sex are the only significant clinical variables in the differential diagnosis of paroxysmal atrioventricular reciprocating tachycardias in patients without pre-excitation in sinus rhythm. Their consideration adds significant diagnostic information to the ECG.
Atrioventricular nodal re-entrant tachycardias (AVNRTs) and orthodromic reciprocating tachycardias (ORTs) using a concealed accessory pathway are the major mechanisms of paroxysmal atrioventricular (AV) re-entrant tachycardias when electrocardiographic (ECG) features of ventricular pre-excitation in sinus rhythm are lacking. Although several studies in the published data have analyzed the univariate associations of clinical variables in both arrhythmic mechanisms (1–10), its noninvasive differential diagnosis has been typically based on subjective, inaccurate ECG interpretation. However, the independent predictive contribution to the ECG of simple bedside clinical variables in that differential diagnosis is not clear. The predictive accuracy of several combinations of these immediate clinical variables would have implications for diagnostic and therapeutic management of these patients. Therefore, the aim of this study was to assess the independent predictive contribution of these clinical variables and their combinations with the ECG in differential diagnosis in a large series of patients with these arrhythmias who underwent an electrophysiologic study.
We included 370 consecutive patients >10 years of age (231 female subjects, 139 male subjects; age 46 ± 18 years, range 11 to 89 years) who underwent an electrophysiologic study for paroxysmal, regular, narrow-QRS complex tachycardia without pre-excitation in sinus rhythm. The prospective inclusion covered from February 2003 to June 2008. The documented arrhythmias occurred either spontaneously (n = 313, 85%) or were induced during the electrophysiologic study. The patients were not receiving antiarrhythmic drugs at the time of the clinical tachycardia documentation. Patients with atrial tachycardias without prior noninvasive diagnosis (9 patients), showing >1 tachycardia mechanism (n = 5), or unable to fulfill a structured clinical questionnaire (n = 3) were excluded. No significant organic heart disease was observed in all but 9 cases.
In addition to the items to be covered in the history of any patient with palpitations (10), a structured 7-item questionnaire was given to every patient at admission by 1 of the investigators (Table 1).The 12-lead ECGs were recorded at a paper speed of 25 mm/s and gain setting of 10 mm/mV, and filter settings were 0.05 to 150 Hz for laboratory recordings. All ECGs were reviewed independently by 2 experienced electrophysiologists (E.G.-T. and S.C.) who were blinded to patient's clinical information and tachycardia mechanism. Spontaneously occurring tachycardia recordings were preferred for ECG analysis when available. In case of discordant judgment, the final ECG diagnosis was reached by consensus. Great care was taken in using ECG diagnostic criteria consistent with those applied in previous studies (11,12). Specifically, the presence of the following criteria were evaluated in each patient to reach a presumptive ECG diagnosis of the tachycardia mechanism: 1) pseudo r′ deflection (V1) and/or pseudo S-wave (inferior leads), favoring AVNRT; 2) identifiable P-wave after the QRS complex, suggesting ORT; and 3) QRS alternans, suggesting ORT. Comparing the sinus rhythm ECGs helped identify the first 2 ECG criteria (Fig. 1).An ECG AVNRT diagnosis was performed when any of these positive ECG findings were lacking (12). A random subset of 50 ECGs were selected to determine the interobserver and intraobserver concordance for categorical variables. In every patient, the mechanism of arrhythmia was defined during the electrophysiological study with widely established criteria (13) and confirmed by the efficacy of radiofrequency ablation (n = 298) or cryoablation (n = 67).
Continuous and categorical variables are expressed as mean value ± SD and as percentages, respectively. Comparisons between normally distributed continuous variables were made with the Student ttest. Differences between categorical variables were analyzed with the chi-square test or MacNemar test when appropriate. The independent diagnostic contribution of basic significant clinical variables was analyzed alone and in conjunction with the ECG interpretation by stepwise multiple logistic regression. First order interactions among selected covariates were tested. Pre-specified default probability for stepwise values (p-in [0.05] and p-out [0.10]) were used in these analyses, with likelihood ratio test for variable selection. Optimal cutoff value of selected continuous variables to predict tachycardia mechanism was determined by receiver-operator characteristic (ROC) curve analysis. To evaluate the ability of the logistic models to classify the arrhythmia mechanism, we plotted ROC curves for the respective predictive probabilities (14,15). The C statistic, a measure of the area under the ROC curve, was calculated and compared for logistic regression models with and without the ECG diagnosis. Intraobserver and interobserver agreement in analyzing the subjective ECG diagnosis were calculated through the kappa statistic. All univariate and multivariable analyses were performed with the use of SPSS software version 13.0 (SPSS, Chicago, Illinois). The ROC curves and C statistics were generated and compared (paired analysis) with the use of the STATA package version 9.1 (Statacorp LP, College Station, Texas). A 2-tailed p value of <0.05 was considered statistically significant.
Descriptive results and univariate analyses
The AVNRT and ORT were demonstrated in 262 and 108 patients, respectively. Atypical AVNRT (ventriculo-atrial interval >100 ms) was observed in 23 patients. In contrast, accessory pathway location was in left free wall in 68% of ORT cases, and 31 of these patients (29%) had a concealed septal AV bypass. Multiple concealed accessory pathways were observed in 4 of these patients. The patients' general and demographic characteristics are listed in Table 2.Female sex, presence of associated structural heart disease, and identifiable precipitating factors were more prevalent in AVNRT patients. In addition, arrhythmias in patients with concealed accessory AV pathways tend to occur at a younger age than in AVNRT patients. Figure 2shows the age at the onset of symptoms ROC curve. The cutoff point that better optimizes the sensitivity and specificity for an AVNRT diagnosis is for values ≥30 years.
A correct ECG interpretation (68% for total study population) was obtained more frequently in the ORT group. The Cohen's kappa values for the interobserver and intraobserver concordance in the ECG diagnosis were 0.7 and 0.83, respectively.
Regarding symptom frequency during tachycardia episodes (Table 3),only the sensation of rapid regular pounding in the neck during tachycardia was observed more frequently in AVNRT patients. Forty percent of those patients refer palpitations in the neck. Specifically, the percentage of AVNRT patients with palpitations in the neck was significantly greater in the typical forms of this tachyarrhythmia (51% vs. 25%, p < 0.001).
The presence of syncope during any of the episodes was uncommon in both groups of patients. Syncope/pre-syncope during tachycardia episodes were more commonly found in older patients (age 56 ± 17 years vs. 45 ± 16 years; p < 0.0005).
Age at the onset of symptoms (adjusted odds ratio [OR]: 1.27), the presence of palpitations in the neck (OR: 3.54), and female sex (OR: 2.96) were the only clinical variables with significant diagnostic power in the identification of AVNRT (vs. ORT) in the multivariate analysis. These 3 variables were selected by the logistic model as significant predictors of the tachycardia mechanism when the ECG subjective diagnosis was included in the analysis (Table 4).No significant interactions were found between selected covariates. Adjusted diagnostic probabilities >70% and >80% were found in 66% and 54%, respectively, of our patients with different combinations of these clinical variables. When ECG interpretation was included in the analysis, these diagnostic probabilities were found in 71% and 52% of the patients, respectively. We constructed ROC curves for the model incorporating selected clinical variables with and without inclusion of the ECG interpretation, with the AVNRT diagnosis as the outcome (Fig. 3).The C statistic for that diagnosis was greater for the model including both clinical variables and ECG interpretation (0.81 vs. 0.755, p = 0.003).
In the clinical ECG predictive logistic model, 21 of 46 false positive cases for the diagnosis of AVNRT (45%) were in patients with incorrect ECG diagnosis, and 71.5% of them were female subjects. Similarly, 22 of 29 false negative predictions for AVNRT diagnosis (76%) had an incorrect ECG presumptive diagnosis, and only 3 of those patients referred palpitations in the neck (10%). Nine of those false negative cases (31%) were in patients with atypical forms of AVNRT.
In 116 patients (31%) in whom pre-specified ECG criteria were lacking (78%, AVNRT patients) (Fig. 1C), the presence of neck palpitations appeared as the only significant predictive variable for an AVNRT diagnosis (OR: 7.5). This clinical covariate led to 77.5% correct classifications as compared with only 60% correct classifications by ECG interpretation in this subset of patients (p < 0.001). Sensitivity, specificity, and positive and negative predictive values of that clinical finding for identifying AVNRT in this subgroup of patients were 50%, 88.5%, 94%, and 34%, respectively. However, age at onset of symptoms (OR: 1.34), female sex (OR: 2.9), and the presence of neck palpitations (OR: 2.9) added significant predictive accuracy to the ECG subjective diagnosis in the remaining 254 patients showing positive pre-specified ECG criteria (Table 5).Similarly, sensitivity, specificity, and positive and negative predictive values of women age >30 years for AVNRT diagnosis in this subset of patients were 41%, 90%, 90%, and 42%, respectively.
Our study shows the independent additional diagnostic value of simple bedside clinical variables and their combinations for the differential diagnosis of paroxysmal AV re-entrant tachycardias in patients without pre-excitation during sinus rhythm. Age at the onset of symptoms, the presence of palpitations in the neck, and female sex were, in order of predictive impact, the only clinical variables with significant diagnostic power in the identification of AVNRT (vs. ORT) in the multivariate analysis. The inclusion of these simple bedside clinical variables into classical ECG interpretation added significant predictive information to the differential diagnosis of the major mechanisms of paroxysmal, regular supraventricular tachycardias. Despite current over-reliance on laboratory tests, our findings further corroborate the usefulness of simple clinical history findings in the assessment of patients with these common tachyarrhythmias.
Previous studies have shown the univariate significant associations of age at onset of first arrhythmic event (1,2,8,9), sex prevalence (1–4,9), and the presence of regular palpitations in the neck during tachycardia episodes (4,16,17) with the major mechanisms of paroxysmal, regular supraventricular tachycardias. Their biological plausibility has been discussed in detail elsewhere (9,16–19). However, this is the first study that prospectively assessed the adjusted predictive power of the different combinations of these selected bedside clinical variables either in isolation or in combination with an ECG presumptive diagnosis. In fact, the presence of positive findings in ≥2 of those clinical variables strongly favors (>80% of predicted probability) an AVNRT diagnosis. Therefore these clinical findings might be useful when ECG information is lacking or limited, such as in ambulatory Holter monitoring. That predicted probability increases to >90% when ≥2 positive clinical findings are present in conjunction with a presumptive ECG diagnosis of AVNRT.
However, the increment in the diagnostic power of the ECG with clinical information, although statistically significant, seems to be at most moderate. In fact, the percentage of patients with predicted probabilities >80% for a correct diagnosis were quite similar, regardless of the use of ECG information in conjunction with basic clinical information. In previous experiences, most false-negative cases derived from ECG interpretation did correspond to patients with atypical or uncommon forms of AVNRT (12), a subgroup of patients with a lower prevalence of a significant predictive covariate such as the presence of rapid regular pounding in the neck during tachycardia episodes. The latter might partially offset the predictive accuracy provided by clinical covariates in the total study group. However, an area under the ROC curve of >0.8 demonstrates an excellent discrimination power of our clinical ECG predictive logistic model (15), in which almost two-thirds of patients with incorrect classification also had an erroneous subjective ECG diagnosis. In addition, a bedside clinical variable such as the presence of neck palpitations emerged as the sole predictor of AVNRT when positive pre-specified ECG criteria were not identified. That most of these are AVNRT patients (78%) is likely to be an underlying cause for the absence of significant predictive impact of other clinical variables in that subgroup of patients.
Interrogation inconsistencies inherent in any anamnestic data and history-taking are a major limitation of the study (20). Moreover, the reproducibility in the presumptive ECG diagnosis of the tachycardia is another limitation of our study and an inherent drawback to the subjective interpretation of surface ECG. In addition, our data analyses are restricted to the described clinical symptoms. A further limitation is that our study is based on a binary logistic regression analysis excluding patients with atrial tachycardia who are a minority in previous consecutive series of cases with paroxysmal, regular supraventricular tachycardias. In fact, patients with atrial tachycardias without prior noninvasive diagnosis would constitute only approximately 2% of our study population. Another limitation of the study is that there might have been a spectrum and referral bias, because all patients had to be referred for an electrophysiology procedure to be included in the study. Finally, these predictive regression models need to be validated prospectively from an external cohort.
Age at onset of first arrhythmic event, sensation of palpitations in the neck during tachycardias, and sex are, in order of predictive impact, the only significant clinical variables in the differential diagnosis of paroxysmal AV reciprocating tachycardias in patients without pre-excitation in sinus rhythm. In addition, the presence of neck palpitations emerges as the sole significant AVNRT predictor when positive ECG criteria are not identified. Precise probabilities for a correct diagnosis associated with combinations of those simple clinical criteria are presented for the first time. Their consideration adds moderate but significant predictive accuracy to the ECG in that differential diagnosis.
- Abbreviations and Acronyms
- atrioventricular nodal re-entrant tachycardia
- odds ratio
- orthodromic reciprocating tachycardia
- receiver-operator characteristic
- Received October 29, 2008.
- Revision received January 27, 2009.
- Accepted February 3, 2009.
- American College of Cardiology Foundation
- Maury P.,
- Zimmermann M.,
- Metzger J.
- Bottoni N.,
- Tomasi C.,
- Donateo P.,
- et al.
- Braunwald E.
- Kalbfleisch S.J.,
- El-Atassi R.,
- Calkins H.,
- Lanberg J.J.,
- Morady F.
- González-Torrecilla E.,
- Almendral J.,
- Arenal A.,
- Del Castillo S.,
- Fernández-Avilés F.
- Josephson M.E.
- Sing T.,
- Sander O.,
- Beerenwinkel N.,
- Lengauer T.
- Hosmer D.W.,
- Lemeshow S.
- Cohen M.I.,
- Wieand T.S.,
- Rhodes L.A.,
- Vetter V.L.