1
JEPonline
Heart Rate Variability Thresholds Predict Lactate Thresholds in Professional World-Class Road Cyclists
Ibai Garcia-Tabar, Luis Sánchez-Medina, José F. Aramendi, Maite Ruesta, Javier Ibañez, Esteban M. Gorostiaga
Studies, Research and Sports Medicine Centre, Government of Navarre, Pamplona, Spain
ABSTRACT
Garcia-Tabar I, Sánchez-Medina L, Aramendi JF, Ruesta M, Ibañez J, Gorostiaga EM.HeartRateVariability Thresholds PredictLactate Thresholds inProfessional World-Class Road Cyclists. JEPonline2013;16(5):38-50.This study aimed to predict widely used aerobic threshold (AeT) and anaerobic lactate thresholds (AnTs) and other cycling performance variables from mathematically determined heart rate variability thresholds (HRVTs). Twelve male professional world-class road cyclists performed a continuous maximal graded cycling test. Blood lactate concentration ([La-]), heart rate (HR), and RR intervals were monitored. Four different LTs (one AeT and three AnTs) were determined. HRVTs were determined from the standard deviation of the instantaneous beat-to-beat RR intervals (SD1). The AeT and one of the HRVT were not statistically different. Significant relationships (P0.05) were found between the lactate thresholds and the HRVTs (r = 0.65-0.88). HRVTs strongly correlated with percentages of peak aerobic power (r = 0.94-0.97; P0.001) and percentages of peak HR (r = 0.87-0.95; P0.001) at which these thresholds occurred. The results indicate that lactate thresholds and percentages of peak aerobic power and peak HR at the HRVTs can be accurately predicted from SD1 valuesduring a submaximal or maximal, non-invasive, low-cost, incremental exercise test in world-class road cyclists. The AeTmight be coincidental with the vagal withdrawal of the heart.
Key Words: Heart Rate Monitor, Cardiac Vagal Activity, Training, Elite Athletes, Exercise Testing
INTRODUCTION
Maximal oxygen uptake (VO2 max) and lactate thresholds (LTs) are commonly used predictors of endurance sport performance. Sophisticated and expensive gas analyzers and participants’ maximal exertion are required to assess VO2 max. Interestingly, while VO2 max appears to be a rather insensitive variable to detect sport performance improvements in elite endurance athletes (8), LTs have become even more popular than respiratory gases in the field of training prescription (22). Either the aerobic lactate threshold (AeT) or various anaerobic lactate thresholds (AnT) are considered sensitive variables to assess endurance capacity as well as to individually prescribe endurance training (30). Nevertheless, LT determination is an invasive and uncomfortable testing method since several blood extractions are required.
Heart rate variability (HRV), a measure of the oscillations in successive RR intervals (26), has become of great interest due to its promising potential to evaluate the activity of the sympathetic nervous system (SNS) and the parasympathetic nervous system (PNS). During rest or low-intensity dynamic exercise, heart rate (HR) is largely controlled by the PNS through the vagus nerve. During incremental exercise there is an increase in HR concomitant with an exponential decrease in HRV (26), and the autonomic nervous system shifts towards a predominance of the SNS caused by an almost complete withdrawal of the vagal modulation of the heart and activation of the SNS (26). The vagal withdrawal (1,5,7,26), defined as the HRV threshold (HRVT), and the AeT (9,13) have been observed to occur at a similar relative exercise intensity (50-60% VO2 max) in untrained and recreationally trained individuals.
Recently, Karapetian et al. (12) found a very large correlation (r = 0.82) between VO2at the HRVT and VO2 at the AeT in a group of male and female volunteers. The average aerobic capacities of the participants were low (VO2 max = 27 and 35 mL·kg-1·min-1) and not homogeneous (coefficient of variation, CV, of VO2 max: ~25%). Furthermore, low-resolution (2) visual inspection techniques, without measuring the interval between the data points, were used to determine the HRVT. It is also worth noting the thresholds are not always visually determinable (5,7),thus their visual determination could differ considerably between observers (32).
The validity of the HRVT to estimate the AeT and AnT in homogeneous groups of athletes with very high aerobic capacities has not been examined yet. Estimation of the AeT and AnT through HRVT may have remarkable practical implications. HRV threshold is thought to be measurable during a safe, non-invasive, low-cost, submaximal test (20) that enables the observer to describe changes in endurance capacity and to prescribe exercise interventions without interfering in regular training programs and tapering periods before competitions. Accordingly, the purpose of the current investigation was to predict the AeT and AnT by mathematically determined HRVTs in a homogeneous group of professional world-class road cyclists. The associations between the HRVTs and other cycling performance variables were also addressed.
METHODS
Subjects
Twelve male professional world-class road cyclists volunteered to participate in this study. The subjects’ mean (± SD) age, height, body mass, and percentage of body fat were 26.9 ± 3.8 yrs, 1.78 ± 0.06 m, 69.6 ± 6.4 kg, and 6.6 ± 0.9%, respectively. The road cyclists belonged to the same International Cycling Union World Tour road cycling team. The study was conducted between December and January, just 4 wks prior to the beginning of the competitive season. At the time of testing, the subjects had cycled 800-1000 km as part of their pre-season training.
This investigation was approved by the Local Institutional Review Board, and it was carried out according to the Declaration of Helsinki. A written informed consent was obtained from all the subjects. Additionally, the cyclists completed a specific health questionnaire to ensure that they were not taking any supplement and/or medication, which was corroborated by the team’s medical staff.
Procedures
All testing sessions were performed during the morning in a climatically-controlled laboratory under similar environmental conditions (temperature, 19.3 to 21.4°C; relative air humidity, 30 to 32%). The pre-exercise meal for eachsubject was the same in an effort to standardize nutritional intake. All subjects were previously familiarized with the exercise testing equipment and procedures.
Peak aerobic power (Wpeak) was estimated by a continuous maximal graded test protocol to volitional exhaustion performed on a computer-controlled cycle-ergometer (SRM High Performance Ergometer, Ingenieurbüro Schoberer, Germany). Saddle, handlebar, and crank length were individually adjusted so that the dimensions of each cyclist’s own bicycle set up were obtained. In order to allow cyclists to wear their own cycling shoes, clip-in pedals were attached to the ergometer. Both starting intensity and workload increments were 58 W and stages were 3 min long. The subjects were instructed not to speak during the test, to remain seated on the saddle throughout the test, and to maintain a constant cycling pedalling cadence of 90 rev·min-1. Exhaustion was defined as the cyclist not being able to maintain the required cadence.
Heart rate was continuously monitored throughout the trial (RS800CX, Polar Electro Oy, Finland). Capillary blood samples from a hyperemic earlobe were obtained at rest, 10 sec before the end of each stage, at the completion of the test and at the 3rd min of recovery. After cleaning and puncturing, the single-use enzyme-coated electrode test strip was directly filled by a 5μl whole-blood sample and blood lactate concentration [La-] was amperometrically determined (Lactate Pro LT-1710, Arkray KDK Corporation, Japan). The Lactate Pro was checked beforehand to ensure correct operation in accordance with the manufacturer’s instructions.Peak aerobic power was calculated as described elsewhere (14).
Determination of Blood Lactate Thresholds
The key variable of interest at the thresholds was power output per kilogram of body mass (W·kg-1), as commonly reported in the literature. Individual data points for the exercise [La-] values were plotted as a continuous function against workload. The exercise [La-] curve was fitted with a third-order degree polynomial equation. The correlation coefficient (r) of every individual equation was r > 0.99 (P0.001). The highest [La-] used for LT assessment was the [La-] obtained at the last completed stage.
Four different LTs (one AeT and three AnTs) that have been proposed as important determinants of endurance capacity (30) were identified: (a) in order to overcome the error associated with the analyzer (31), the workload corresponding to an elevation in [La-] of 0.2 mmol·l-1 above baseline [La-] was chosen as a threshold and named AeT+0.2; (b) the workload associated with a [La-] of 1 mmol·l-1 above baseline [La-] (22) was also determined (AnT+1); (c) as well as fixed [La-] of 2.5 mmol·l-1 (AnT2.5) (31); and (d) 4 mmol·l-1 (AnT4) (24). The workloads corresponding to the aforementioned LTs were computed from the individual equations describing the exercise [La-] curves.
Determination of SD1 Thresholds
RR intervals were collected by the HR monitor and uploaded to a computer. RR intervals from the last 2 min of each stage of exercise were used for HRV analysis(12). Polar ProTrainer 5 (Polar Electro Oy, version 5.40.172, Finland) was used for the computation of the standard deviation of the instantaneous beat-to-beat RR intervals (SD1). The RR filtering algorithm of the Polar software was not used in order to avoid modifying intrinsic RR variability. Indeed, despite being able to accurately calculate HRV variables in men, it has been demonstrated that the automated filtering feature of the Polar ProTrainer 5 does not properly correct and edit RR data(27).
Individual data points for the exercise SD1 values were plotted as a continuous function against workload. In an attempt to improve previously used visual methods, three new mathematical methods were used to objectively determine the SD1 thresholds (SD1Ts) (Figure 1). First, the workload corresponding to a SD1 value of 1 ms above the lowest SD1 value (SD1Tlow). Determination of the SD1Tlow requires performing a maximal test to find out the lowest SD1 value. In an endeavour to determine a SD1T from a submaximal incremental test, two other thresholds were defined. Second, the workload corresponding to a SD1 value of 3 ms above the SD1 value of the first exercise intensity at which there was a decrease in SD1 0.5 ms between two consecutive stages (SD1T0.5). Third, the workload corresponding to a SD1 value of 0.5 ms above the SD1 value of the first exercise intensity at which there was a decrease in SD1 2.5 ms between two consecutive stages (SD1T2.5). In order to improve applicability, linear interpolation was used to identify the workloads at the SD1Ts.
All LTs and SD1Ts data processing was performed off-line using specific routines developed in a commercial software package (MATLAB R2008a, The MathWorks Inc., USA). HR values at the LTs and SD1Ts were computed from the individual HR-workload linear regression equations (r > 0.98; P0.001).
Figure 1.Identification of the SD1 Thresholds based on Different Methods. ∆ 2.5, a decrease in SD1 smaller than 2.5 ms; ∆ 0.5, a decrease in SD1 smaller than 0.5 ms; Min., lowest SD1 value. The arrows above the curve denote the SD1 thresholds (SD1Tlow, SD1T2.5 and SD1T0.5)
Statistical Analyses
Statistical analysis was carried out using the statistical package SPSS 17.0 (SPSS Inc., USA). A prospective calculation of sample size for a minimum value of a very large correlation (r = 0.80) and based on prior data (12) was performed. Assuming a power of 80% and a type I error of 0.05, the estimated sample size required to accomplish this study was 7 cyclists (10). Prior to performing the statistical analysis, assumptions of normality, homoscedasticity, and when appropriate sphericity, were checked. Differences in mean [La-] and SD1 at different workloads were analyzed using one factor analysis of variance (ANOVA) with repeated measures. When significance was found, the Student’s paired t-test with Bonferroni correction for multiple comparisons was used to identify the significant difference. Differences in power outputs between thresholds were evaluated by the Student’s paired t-tests. Linear regression analyses were performed to determine the relationship between the variables of interest. The strength of each linear relationship was assessed by Pearson’s product-moment correlation coefficient (r). With the aim to compare our results with recently published results (12), Bland and Altman plot analysis (6) was used to asses agreement between the AeT+0.2 and SD1Tlow. Correlations’ magnitudes were interpreted according to Hopkins et al. (11). Statistical significance was set at P0.05. Data in the text, tables and figures are reported as mean values and standard deviation (± SD).
RESULTS
Maximal Incremental Exercise Test
Mean Wpeak, Wpeak relative to body mass, peak HR (HRpeak), and peak blood lactate concentration ([La-peak]) attained in the maximal incremental exercise test were 423 ± 18 W, 6.1 ± 0.4 W·kg-1, 192 ± 11 beats·min-1 and 9.4 ± 2.0 mmol·l-1, respectively.
The average workload vs. [La-] and HR curves, and the average workload vs. SD1 curve are presented in Figure 2. Heart rate increased linearly and significantly (P0.001) over the duration of the test (Figure 2a). During the first four stages [La-] did not noticeably change. From the fourth stage on (232 W) [La-] significantly increased (P0.01) in each subsequent exercise stage. As shown in Figure 2b, SD1 decreased exponentially over the course of the test. SD1 values of the first three stages differed significantly from SD1 values of the rest of the stages (P0.05). Thereafter (232 W) SD1 did not significantly change.
Table 1. Descriptive Features of Blood Lactate and SD1 Thresholds (n = 12).
W·kg-1 / %Wpeak / %HRpeakMean ± SD / Range / Mean ± SD / Range / Mean ± SD / Range
AeT+0.2 / 3.34 ± 0.60 / 2.2 - 4.4 / 54.7 ± 9.0 / 39 - 66 / 69.1 ± 5.0 / 60 - 76
AnT+1 / 4.27 ± 0.47 / 3.7 – 5.4 / 69.8 ± 5.1 / 62 - 79 / 79.2 ± 3.6 / 74 - 88
AnT2.5 / 4.53 ± 0.48 / 3.9 - 5.7 / 74.2 ± 5.7 / 66 - 84 / 82.1 ± 3.8 / 77 - 90
AnT4 / 5.17 ± 0.53 / 4.4 - 5.2 / 84.5 ± 4.9 / 78 - 94 / 88.9 ± 4.3 / 82 - 99
SD1Tlow / 3.19 ± 0.61 / 2.6 - 4.1 / 52.1 ± 8.8 / 37 - 63 / 67.5 ± 5.4 / 56 - 77
SD1T0.5 / 2.19 ± 0.50 / 1.4 - 2.9 / 35.7 ± 7.4 / 23 - 46 / 56.5 ± 4.2 / 49 - 62
SD1T2.5 / 2.84 ± 0.60 / 1.8 - 3.5 / 46.2 ± 8.2 / 33 - 54 / 63.5 ± 5.6 / 54 - 70
W·kg-1 watts relative to body mass, %Wpeak percentage of peak aerobic power, %HRpeak percentage of peak heart rate
Blood Lactate and SD1 Thresholds
Average workloads (W·kg-1), percentages of Wpeak (%Wpeak), and HRpeak (%HRpeak) at each of the LTs and SD1Ts assessed are presented in Table I. Paired t-tests indicated that values at the AeT+0.2 and SD1Tlow were not statistically different. Values corresponding to the SD1T0.5 and SD1T2.5 were, respectively, 34% and 15% lower than those corresponding to the AeT+0.2 (P0.001). Average workloads needed to calculate the SD1T0.5 and SD1T2.5 were 66% (range 50-86%) and 54% (range 38-70%) of Wpeak respectively, and 79% (range 64-90%) and 70% (range 60-82%) of HRpeak, respectively. As expected, values at the AnT+1, AnT2.5 and AnT4 were higher than those at the AeT+0.2 (P0.01).
Figure 2.Mean (± SD) Blood Lactate, Heart Rate (a) and SD1 (b) Values at the End of Each Stage throughout the Incremental Exercise Test. Maximal workload is indicated with dashed lines (- -). *Significantly different from all other stages (P0.05)
r / 95% CI / SEE / Regression equationAeT+0.2
SD1Tlow / r = 0.88*** / 0.56-1.23 / 0.29 W·kg-1 / y = 0.8685x + 0.5727
SD1T0.5 / r = 0.82*** / 0.34-0.99 / 0.29 W·kg-1 / y = 1.0195x + 1.0664
SD1T2.5 / r = 0.66* / 0.13-1.20 / 0.47 W·kg-1 / y = 0.6569x + 1.4787
AnT+1
SD1Tlow / r = 0.81*** / 0.53-1.59 / 0.37 W·kg-1 / y = 1.049x – 1.2895
SD1T0.5 / r = 0.68* / 0.17-1.27 / 0.38 W·kg-1 / y = 0.7026x – 0.8117
SD1T2.5 / r = 0.71** / 0.27-1.55 / 0.44 W·kg-1 / y = 0.91x – 1.0483
AnT2.5
SD1Tlow / r = 0.84*** / 0.58-1.53 / 0.34 W·kg-1 / y = 1.0536x – 1.5891
SD1T0.5 / r = 0.67* / 0.14-1.22 / 0.39 W·kg-1 / y = 0.6831x – 0.9101
SD1T2.5 / r = 0.71** / 0.26-1.49 / 0.45 W·kg-1 / y = 0.875x – 1.1319
AnT 4
SD1Tlow / r = 0.72** / 0.27-1.38 / 0.44 W·kg-1 / y = 0.8226x – 1.0616
SD1T0.5 / r = 0.54 / 0.07-1.05 / 0.44 W·kg-1 / y = 0.4884x – 0.3357
SD1T2.5 / r = 0.65* / 0.14-1.34 / 0.48 W·kg-1 / y = 0.7383x – 0.9785
Table 2. Pearson’s Product Moment (r) Between Lactate and SD1 Thresholds; 95% Confidence Intervals for r (CI); Standard Error of the Estimate (SEE), and Regression Equations (n = 12).
*P0.05; **P0.01; ***P0.001
Regression and Agreement Analyses
Results of the regression analyses between LTs and predictor variables are illustrated in Table 2. Pearson’s product-moment (r) indicated that the AeT+0.2 correlated best with the HRVTs (r = 0.66-0.88; P0.05). Regression analysis between the AeT+0.2 and predictor variable SD1Tlow is presented in Figure 3a. The Bland and Altman plot revealed agreement between these two thresholds (Figure 3b).
Figure 3.Relationship between SD1Tlow and AeT+0.2. Linear regression analysis(a). Regression equation solid line (-); 95% confidence interval dashed lines (- -). Bland and Altman plot analysis (b).
Extremely large correlations between the SD1Tlow and %Wpeak and %HRpeak at which SD1Tlow occurred were observed (Figure 4a, 4b), as well as between SD1T2.5 and %Wpeak and %HRpeak at which SD1T2.5 occurred (Figure 4c, 4d). Likewise, SD1T0.5 correlated with %Wpeak at SD1T0.5 (r = 0.96; P0.001; SEE = 0.14 W·kg-1) and %HRpeak at SD1T0.5 (r = 0.87; P0.001; SEE = 0.25 W·kg-1). Correlations of similar magnitude were found between the AeT+0.2 and %Wpeak (r = 0.92; P0.001; SEE = 0.24 W·kg-1) and %HRpeak (r = 0.95; P0.001; SEE = 0.20 W·kg-1) at which AeT+0.2 occurred. Among the HRVTs, SD1T2.5 correlated very largely with Wpeak (r = 0.71; P=0.01; SEE = 0.29 W·kg-1). No other significant correlations between SD1Ts and maximal variables were found.
Figure 4.Percentages of Peak Aerobic Power and Percentages of Peak HR at SD1Tlow (a, b) and at SD1T2.5 (c, d). Regression equation solid lines (-); 95% confidence interval dashed lines (- -).
DISCUSSION
This is the first study to estimate different LTs through HRVTs using mathematical methods to determine the thresholds in a homogenous group of professional male world-class road cyclists. The major findings are as follows: (a) SD1 values (SD1Tlow) enable us to accurately and objectively predict the AeT (AeT+0.2) and AnTs; (b) extremely large relationships were found between the SD1Tlow and the %Wpeak and %HRpeak at which SD1Tlow occurred;and (c) SD1T0.5 and SD1T2.5, two other mathematically determined HRVTs measurable during a submaximal incremental exercise test, also correlated with the AeT+0.2, AnTs, %Wpeak and %HRpeak at which these HRVTs occurred. These findings indicate that the AeT+0.2, AnTs and %Wpeak and %HRpeak at which SD1Ts occur can be estimated from HRV measures obtained during a maximal or submaximal non-invasive, low-cost, incremental exercise test in a homogeneous group of professional male world-class road cyclists.
While the average relative Wpeak recorded in the present study (6.1 W·kg-1) is typical of the top-level competitive cyclists (>5.5 W·kg-1) (18), they showed slightly lower average Wpeak and OBLA or AnT4 and similar [La-peak] and %Wpeak and %HRpeak at OBLA or AnT4 compared to other professional world-class road cyclists tested more than a decade ago (16,18,19). The lower average Wpeak and OBLA or AnT4 observed in the present study may be associated with the different testing procedures used (i.e., stage duration, load increment, continuous or discontinuous protocol and type of cycle-ergometer used). The time of the season when testing was carried out is likely to be the main factor that explains these differences. This study was conducted between December and January when the subjects had only cycled 800-1000 km as part of their pre-season training. Participants in the aforementioned studies were tested in April (18,19) or 4 to 6 wks before their particular “peak” moment of the season (16), when they had already cycled more than 10,000 km during training and competition. Therefore, the subjects in this study are considered a representative sample of professional world-class road cyclists that had the typical physical characteristics of pre-season.
Visual inspection of Figure 2 indicates that mean workload just before [La-] begins to increase (~232 W, Figure 2a), usually termed “lactate threshold” (29), was very close to the point at which SD1 almost completely withdrew (Figure 2b). This was corroborated by the use of mathematical methods that showed that mean workload at the AeT+0.2 did not differ from mean workload at the SD1Tlow. In addition, a verylarge correlation (r = 0.88) and strong agreement were found between both thresholds (Figure 3). A lower magnitude of correlation (r = 0.82) between visually identified individual AeTs and HRVTs during a continuous graded cycling protocol to exhaustion has been previously reported in an heterogeneous mixed group of male and female participants with a wide range of fitness level (CV of VO2 max, ~25%) (12). The present results indicate that the AeT+0.2 could be accurately and objectively estimated from the SD1Tlow in a homogeneous group (CV of Wpeak=4.3%) of male world-class road cyclists.