ABC | Volume 115, Nº1, July 2020

Original Article Benchimol-Barbosa et al. Dynamic AV-conduction to RR-interval Coupling Arq Bras Cardiol. 2020; 115(1):71-77 The present study evaluated beat-by-beat AV conduction time ( AVCT ) and RR-interval variabilities in elite runners and healthy sedentary subjects, at rest, aiming at assessing the effect of physical fitness status on spontaneous AVCT to RR-interval duration coupling. Methods Detailed information about study protocol, Ethics Committee approval, and ECG data acquisition has been described elsewhere. 6 The present study analyzed raw high resolution ECG data from the ECG data bank of the Biomedical Engineering Program, using a novel technique for extraction of atrioventricular conduction time and RR-intervals . 14 Data sampling procedure was described elsewhere. 6 The study population comprised 20 volunteers divided in two groups: the ‘Athlete’ group, comprising ten elite long- distance runners (≥ 16.0 maximal metabolic equivalents [MET] calculated as the maximal oxygen consumption achieved during stress test divided by 3.5 ml·kg -1 ·min -1 , [mean ± SD] 19.5 ± 1.3 MET; aged 25.1 ± 7.1 years), and the ‘Control’ group, comprising ten male healthy sedentary volunteers (≤ 11.5 MET; 8.7 ± 1.9 MET; aged 29.0 ± 5.4 years). It is worth mentioning that the term ‘MET’ is employed throughout the text as the maximal metabolic equivalent achieved during stress test. The athletes’ training program consisted of six to eight training sessions/week; 90 to 120 min/session; 90 to 110 km/week. The waves and fiducial point detection were carried out on ECG acquired using XYZ modified Frank leads, in the resting supine position, using low- pass filter at 15 Hz ( Butterworth , 4th order). For the analysis of the RR-interval duration, artefacts and ectopic beats were adequately excluded. 15,16 The distance between the peak of the P -wave and the peak of the R -wave in normal beats defined the PR-peak interval and was employed to analyze instantaneous AVCT adaptation over the cardiac cycle. 14 The PR-peak to RR-intervals coupling was assessed in a beat-by-beat basis throughout the whole ECG recording, using the lead showing the tallest P -wave, usually the Y lead. The RR-interval was assessed as the time between the peaks of the R-waves of two consecutive normal beats. The PR-peak interval was assessed immediately before the second beat of the respective RR-interval . For each subject, the mean (M) and standard deviation (SD) of all consecutive normal RR-intervals ( M-RR and SD-RR ) and respective PR-peak interval ( M-AVCT and SD- AVCT ) were calculated. PR-peak intervals were correlated to the respective RR-intervals and calculated regression line slopes ( RR-AVCT slope ). Statistical Analysis The variables were expressed as mean ± SD or median and interquartile range, when appropriate. Data normality was assessed using Kolmogorov-Smirnov test, and all analyzed variables had their normality assumption accepted. Variables were compared between groups using non-paired Student’s t-test. To assess the optimal cut-off values, ROC curves were calculated from the regression line slopes ( AVCT vs. RR- interval) in both groups. A multivariate linear regression model was developed to explain the MET based on significant AVCT and RR-interval parameters. Pearson’s correlation coefficient r was tested for significance (significance level was set at 5%). A concordant AV conduction was arbitrarily defined as AVCT and RR-interval increased and decreased in the same direction in consecutive cardiac cycles, and discordant otherwise. AVCT was assessed as PR-peak-interval . A validation bootstrap resampling procedure applied to the multivariate model was carried out using two different approaches. In the first approach, 1100 replications with replacement were carried out in the whole sample of both groups to assess mean and SD estimates of independent variables. In a second approach, both groups were split in a test group, comprising 67% of subjects of each group, and a validation group, with the remaining 33%. The MET estimated by the multivariate model was employed to classify Controls and Athletes in all sets of bootstrap procedures. Raw ECG signals were processed using custom-made programs written in Matlab R2007a (The MathWorks, Inc) language. Statistical analysis was carried out using MS-Excel 360 (Microsoft Corporation) and Medcalc version 11 (Medcalc Software bvba). The significance level adopted in the statistical analysis was 5%. Results The Athletes had significantly higher M-RR and SD-RR values than the Controls, whereas there were no significant differences between M-AVCT and SD-AVCT values (Table 1). Examples of subjects from the Control (a) and Athlete (b) groups are shown in Figure 1, where regression lines and respective r of AVCT vs RR-intervals scatterplots are shown. RR-AVCT slope values are positive in Controls (Figure 1-a), whereas they are negative in Athletes (Figure 1-b). Overall, RR-AVCT slope in Controls and Athletes resulted in significant between-groups’ differences (Table 1). Variables showing significant intergroup differences were entered into a multivariate linear regression model, taking MET as the dependent variable in the bootstrap procedure. SD-RR (p = 0.0099) and RR- AVCT slope (p = 0.006) were independent explanatory variables of MET, showing 90% specificity, 100% sensitivity and 95% total accuracy (Table 2). The average C-statistic in test and validation groups were, respectively, 0.97 ± 0.06 and 0.87 ± 0.13; p < 0.001 for both. The multivariate linear regression analysis and respective bootstrap procedures results are summarized in Table 2. The RR- AVCT slope values for each group, including interquartile range and 95% confidence intervals (CI) are shown in figure 2-a. Sensitivity, specificity and total accuracy were computed utilizing the optimal cutoff value shown in table 1, and exhibited as inset. To highlight the association between spontaneous decremental conduction and physical status, a regression line of RR- AVCT slope vs. MET is shown in figure 2-a. It shows a significant correlation (r = 0.70; p < 0.05) and a negative slope, demonstrating that RR- AVCT slope decreases as MET increases. An example of a short sequence of sinus beats showing spontaneous decremental conduction, registered during supine rest in a 19 y.o. athlete (MET 16.8 METs) is shown in Figure 2-b. 72