Determination of Lactate Thresholds in Maximal Running Test by Heart Rate Variability Data Set

1. Background

The relationship between parasympathetic (PNS) and sympathetic nervous system (SNS) activation can be assessed with heart rate variability (HRV) analysis (1, 2). Increasing the exercise intensity may thus reflect a reduction in HRV during the initial stages of a maximal incremental exercise test, thereafter a significant PNS activation withdrawal occurs between 50% - 60% of the maximal oxygen uptake (VO_2MAX) (3). That intensity has been related with the first physiological threshold (i.e. first lactate or ventilatory threshold - LT₁ and VT₁, respectively) (4). In contrast, there is a gradual increment in SNS activation from intensities corresponding to LT1 and VT₁, with remarkable increase at exercise intensities related to the second physiological threshold (i.e. second lactate and ventilatory threshold - LT₂ and VT₂, respectively) (3, 5). As a result, studies have suggested that HRV analysis may be a potential and useful tool to estimate the exercise intensity of the first and second physiological thresholds (4, 6-8).

Using a cycling exercise mode, Cottin et al. (7) found that nonlinear increases in HRV data, assessed by high frequency (HF) and peak frequency spectrum (ƒHF) of the beat-to-beat RR intervals, identified VT₁ and VT₂. They also showed that instantaneous HF multiplied by ƒHF (HF.ƒHF) provided a reliable and accurate index to identify these thresholds. In a subsequent study, these authors expanded these findings to running exercise mode. Using an exhaustive incremental running test performed on a 200 m track, Cottin et al. (8) showed that VT₁ and VT₂ could be assessed through a simple HRV monitor. Recently, these results were further confirmed in obese individuals (9), corroborating the usefulness of the HRV method to identify the first and second physiological threshold in different individuals, during different exercise modes.

However, despite the support to use HRV data set when determining VT₁ and VT₂ (7-9), some aspects should be addressed. No study has currently verified the agreement between breakpoints identified through HRV and LT₁ and LT₂, as no lactate measures were included in those previous studies (7, 8). In this regard, it should be highlighted that lactate thresholds (10-13) is considered as a gold standard method when predicting running aerobic performance. The agreement between ventilatory and lactate thresholds is still an ongoing debate, as these indices are used to identify identical exercise intensities. Since work by Gaesser and Poole (14) nearly thirty years ago, studies have shown different results when comparing VT₁ and VT₂ to LT₁ and LT₂, respectively (15-17).

On the other hand, Karapetian et al. (4) have suggested the identification of a HRV threshold by using time domain indexes (standard deviation - SD and mean successive difference – MSD), which may match lactate and ventilatory thresholds during maximal incremental cycling test. Furthermore, it is important to point out that those studies have reported a nonlinear pattern in HRV data during maximal incremental exercise tests, thus making possible the identification of a HRV threshold when visually inspecting abrupt breakpoints in HRV responses during maximal incremental exercises. Consequently, it has been suggested that the HRV threshold, regardless of the method used for its determination, may be used to estimate the first and second physiological thresholds with different methodologies of HRV data analysis such as frequency domain (7, 8), time domain (4) or Poincare plot (18). Unfortunately, however, an important aspect that remains to be confirmed is the possibility of estimating LT₁ and LT₂ by using HRV data set in incremental running test.

The use of HRV data set for estimation of these traditional thresholds may be of interest, as this method is less expensive and time consuming, and requires uncomplicated and accessible HR monitors. Traditional VT₁ and VT₂ determination requires sophisticated and expensive metabolic gaseous exchange analyzers, while the LT₁ and LT₂ require invasive procedures with multiple blood sample collections, making these methods impractical for clinical assessment and daily endurance training sessions. Therefore, the study of alternative HRV methods for estimation of the first and second physiological threshold has practical relevance.

Reliability of the HRV thresholds method has been provided during maximal incremental cycling tests (18), but no study has been designed yet, to verify the HRV thresholds’ validity and reproducibility when determining LT₁ and LT₂ in maximal incremental running test. Thus, the purpose of this study was to determine if HRV thresholds could identify the lactate thresholds’ intensity during a maximal, incremental running test. We expected that HRV thresholds may elicit reliability and reproducibility when identifying lactate thresholds in treadmill running test.

2. Objectives

Therefore, the first aim of this study was to verify if the HRV method would determine valid HRV breakpoints during running test and if these corresponds to the first and second physiological threshold such as LT₁ and LT₂. The second aim of this study was to verify the reproducibility of the HRV method in running test.

3. Methods

3.1. Participants

Nineteen recreational long-distance male runners (30.4 ± 4.1 years; body mass of 74.3 ± 8.5 kg; height of 176 ± 6.4 cm and body fat of 13.8 ± 4.6 %) volunteered to participate in this study. All participants were healthy, non-smokers, with no cardiovascular or neuromuscular diseases. They were informed about risks and benefits before participating in the study, thereafter they provided their written informed consent. The study protocol was approved by the institutional ethics committee of the University of São Paulo, and conducted according to international standards (19).

3.2. Study Design

The experimental design consisted of two maximal incremental running tests interspersed by a washout period of 3 - 7 days. The tests were performed at the same time of the day and in standard laboratory conditions (humidity of ≈ 50% and temperature of ≈ 22°C). Participants were asked to avoid intense exercises, alcohol and caffeine beverages 24 hours before each test, and to consume a light meal 3 hours before the tests. Before starting the experimental procedures in each test session day, we asked participants about the accomplishment of these recommendations so that only participants who had followed them were evaluated.

3.4. Maximal Incremental Running Test Protocol

Participants wore a cardio belt for beat-to-beat heart rate (HR) measures (s810 Polar®, Kempele, Finland). Thereafter, the participants rested for 20 min (10 min supine + 10 min sitting) for baseline measures of HR and lactate concentrations. Lactate concentrations [La] were determined from a 25 µL blood sample, drawn from the tip of the forefinger (YSI 1500, Yellow Springs, OH, USA). Then the participants moved to the treadmill (CEFISE TK35, Nova Odessa, Brazil) and warmed up for 3 min at 5 km.h^-1 and 1% gradient. The test started at 5 km.h^-1, with 1 km.h^-1 increases every 3 min, until exhaustion (10). The HR data set were recorded continuously throughout the tests, while 25 µL of blood samples were collected during the last 30 s of every stage, while the participant was running.

3.5. Determination of Physiological Thresholds

LT₁ and LT₂: The speed corresponding to a 2 mmol.L^-1 blood lactate concentration was used to determine LT₁ (20), while the speed corresponding to 3.5 mmol.L^-1 identified the LT₂ (10).

D_max: The lactate threshold was further determined according to the D_max method (LT_Dmax) suggested by Cheng et al. (11), thereby providing individualized lactate threshold values. Briefly, LT_Dmax reflects the longest perpendicular distance between [La] values predicted by a 3º order polynomial function over actual [La] values and values derived from a linear regression calculated with the outermost lactate values (i.e. the first and last values of that curve).

HRV thresholds: Beat-to-beat RR interval data were analyzed for aberrant values. A RR interval value was interpreted as a premature beat if it deviated from the previous interval by more than 30% (1, 4). The Poincare plot analysis was performed on RR interval time series to calculate the standard deviation of the instantaneous (SD1) and continuous long-term RR intervals (SD2) (3). Poincare plot is a nonlinear measure of temporal correlations within time series of inter-beat intervals in which an RR interval is plotted against its first predecessor. The SD1 has been shown to correlate strongly with vagal tone (PNS) and SD2 has been shown to correlate with the PNS and SNS. Thereafter, raw RR intervals recorded during the last 60 s of each stage of exercise were normalized by the mean RR interval and multiplied by 1000 (SD1_n and SD2_n) (1). Thus, in accordance with behavior of these variables in previous studies (3, 18), SD1_n and SD2_n were visually used by three evaluators to determine two HRV thresholds. The first (HRVT₁) and second (HRVT₂) HRV thresholds were identified at the first breakpoint in SD1_n and SD2_n curves, respectively. Figure 1 illustrates typical SD1_n and SD2_n curves and the HRV thresholds.

The First and Second HRV Thresholds (HRVT₁ and HRVT₂) Visually Determined by SD1_n and SD2_n, Respectively

Figure 1.

The First and Second HRV Thresholds (HRVT₁ and HRVT₂) Visually Determined by SD1_n and SD2_n, Respectively

All physiological thresholds, LT₁, LT₂, LT_Dmax, HRVT₁ and HRVT₂ were expressed in absolute speed (km.h^-1), HR (bpm), milliseconds beat-to-beat RR interval (ms) and blood lactate concentration (mmol.L^-1). Additionally, thresholds were expressed relatively (%) to maximal values.

3.6. Statistics

Values were expressed as mean and standard deviation (± SD). After ensuring Gaussian data distribution (normality and homoscedasticity), a repeated-measure design ANOVA with Bonferroni’s post-hoc test were used to verify possible differences among different thresholds identified in maximal incremental running test. Typical error of measurement (21) was used to verify the test-retest HRV breakpoints’ reproducibility, as well as to compare them with lactate thresholds. The Bland and Altman plot further provided a visual analysis of the data set bias (22), in addition to the reliability analysis with intra-class coefficient correlation (ICC). All analyzes were performed with SPSS (19.0), and with a spreadsheet designed to specifically calculate the reliability (available on www.sportsci.org). Results were interpreted as significant when P was < 0.05.

4. Results

Baseline HRV values (423.1 ± 28 ms vs. 994.6 ± 139.3 ms; P < 0.05), but not baseline lactate values (1.34 ± 0.5 mmol.L^-1 vs. 1.18 ± 0.3 mmol.L^-1. P > 0.05), were significantly different between test and retest. With regard to values recorded at the exhaustion in maximal incremental running tests, no significant difference was observed between test and retest (P > 0.05) in speed (16.5 ± 1.8 km.h^-1 vs. 16.6 ± 1.7 km.h^-1), HR (192 ± 6 bpm vs. 191 ± 6 bpm) and blood lactate concentration (9.2 ± 1.9 mmol.L^-1 vs. 9.2 ± 1.9 mmol.L^-1).

4.1. Identification of Physiological Thresholds

Overall results were that neither the first nor the second. Physiological threshold was different when determined by lactate and HRV thresholds expressed in absolute as well as in relative values. For example, while LT₁ matched HRV₁, LT₂ matched HRV₂. Interestingly, LTDmax matched the first (LT₁ and HRV₁), but not the second physiological threshold (LT₂ and HRV₂.). Table 1 shows in full results of the first and second physiological thresholds identified during the maximal incremental running test.

Pearson correlation coefficient analysis in thresholds expressed as speed revealed significant correlation (P < 0.01) between HRVT₁ and LT₁ (r = 0.71), HRVT₁ and LT_Dmax (r = 0.68), LT1 and LT_Dmax (r = 0.75), HRVT₂ and LT_Dmax (r = 0.65). Additionally, significant correlation was observed between HRVT₂ and LT₂ (r = 0.86).

Figure 2 shows the magnitude of differences and bias between HRV and lactate thresholds. A good agreement was observed between HRVT₁ and LT₁ (-0.26 ± 2.07 km.h^-1), as well as between HRVT₂ and LT₂ (0.00 ± 1.29 km.h^-1), given the small bias reported.

Bland and Altman Plot for the Agreement Between HRVT and LT₁ (A) and LT₂ (B) Methods.

Figure 2.

Bland and Altman Plot for the Agreement Between HRVT and LT₁ (A) and LT₂ (B) Methods.

4.2. Reproducibility of the HRV Thresholds

Figure 3 presents the Bland-Altman plot with bias analysis for the HRVT₁ and HRVT₂ reliability. There was a reasonable random distribution in test-retest residuals for HRVT₁ and HRVT₂, thus suggesting a homoscedastic distribution as a function of the speed. Results further suggested a consistent reproducibility for thresholds determined through the HRV data set, given the typical error of measurement reported at ≈ 6% (≈ 0.45 km.h^-1) and the ICC showed high correlations (P < 0.001) (Table 2).

Bland and Altman Plot for the Reliability of HRVT₁ (A) and HRVT₂ (B)

Figure 3.

Bland and Altman Plot for the Reliability of HRVT₁ (A) and HRVT₂ (B)

5. Discussion

The main findings of the present study were that the HRV data set provided fair estimation of LT₁ and LT₂ through HRVT₁ and HRVT₂, respectively, during maximal running test. Furthermore, HRV thresholds elicited good reproducibility with low bias during incremental running test.

5.1. HRV Data Set and the Identification of Lactate Thresholds

The results of this study showed that HRVT1, identified by analysis on time domain of the HRV data matched LT₁ and LT_Dmax, either expressed in absolute or relative values of speed and HR. The significant vagal withdrawal during incremental exercises has been well documented in the literature using time domain HRV analysis (4, 18, 23). In fact, HRVT₁ showed relative values of speed around 70.6%, close to values (74.6%) reported by Candido et al. (18) when using a cycle ergometer and similar methods for HRV analysis. In the present study, although the correlation between either HRVT₁ and LT1 or HRVT₁ and LT_Dmax (expressed as km.h^-1) has elicited significant coefficient values (r = 0.71 and r = 0.68, respectively), this was lower than the correlation reported by a previous study by Karapetian et al. (r = 0.82) (4) which used similar visual analysis to identify the vagal withdrawal, that is the point at which the HRV showed no further decrease. However, it is important to note that in contrast to the present study which used a treadmill running incremental test, Karapetian et al. (4) used a cycle ergometer, thus limiting deeper comparisons as the values were expressed as power output (watts) instead of speed (km.h^-1).

Furthermore, others have shown that HRV is a useful method to identify the second physiological breakpoint (6-8). Accordingly, we showed that HRVT₂ was similar to the second physiological threshold identified through the traditional lactate method (LT₂). The variable SD2 was used because it represents the parasympathetic and sympathetic nervous system activity, and this variable shows a non-linear pattern in intensities close to heavy and severe domains (3). At intensities above the first physiological threshold as the LT₁ the sympathetic system starts to intensify its activation (24), so that a significant activation occurs in intensities close to the second physiological threshold, the LT₂. Analyzing the results of Tulppo et al. (3), it was observed a breakpoint in SD2 very nearby intensities ≈ 80% - 90% VO_2MAX during maximal progressive cycling test, having the participants being submitted or not to a parasympathetic system activity blockage before the test. In the present study, HRVT₂ showed relative values of speed and HR around 86.1% and 90.5%, respectively. These values are close to intensities related to the second physiological threshold (5, 24).

Despite some criticism, the identification of two physiological thresholds using different variables is possible; due to a cascade of integrated neurophysiological events that occurs in organs and tissues responsible for locomotion (16). In maximal incremental exercises, the first metabolic transition zone matches a significant PNS withdrawal. Concurrently, there are increases in CO₂ production derived from the bicarbonate buffering in active muscles at the same time (4, 25). The identification of a plateau in SD1, an index referred to the vagal component of the autonomic balance, has been coincident with the intensity around the first metabolic transition zone (P > 0.05) (3).

From the first physiological threshold during incremental exercises there is an increase in muscle and blood pH levels. Together, there is a greater CO₂ production as the exercise progresses, thus intensifying the activation of higher respiratory control centers which leads to an abrupt increase in ventilation minute at the second threshold. Changes in HRV are mediated mainly by alterations in the respiratory sinus arrhythmia and the hyperventilation mechanical effects on the sinus node at this point (26). As a result, there are concomitant changes in respiratory, muscular and CNS systems. In this regard it is important to point out that, regardless of the debate involving physiological thresholds, the present study showed a consistent agreement between different methods identifying physiological thresholds, thus indicating that HRV data set is a valuable tool for physiological assessments used in exercise intensity prescription.

5.2. Reproducibility of the HRV Thresholds

Results of ICC (≈ 0.88; P < 0.001) and Bland and Altman analysis indicated good agreement between test-retest HRV thresholds, with acceptable typical error of measurement between them (≈ 6%). Together with others (7, 8), these results show that thresholds identified by HRV data set, even when analyzing HRV in time domain, is a useful and reliable method to determine the first and second lactate thresholds as identified by LT₁ (and LT_Dmax) and LT₂ during maximal running test.

The choice for HRV time domain analysis was based on previous results (27), which reported an unsuccessful spectral analysis in non-stationary HRV data when characterizing the balance between PNS and SNS during exercise. Thus, we used Poincare’s plot after normalization of SD1 and SD2 values by the mean RR, multiplied by 1000. This procedure is similar to that previously reported and allowed us to individualize analysis of the HRV (28).

Furthermore, an important aspect should be considered. Instead of using cycling as the exercise mode, we used running and participants performed a maximal incremental running test on treadmill. Most studies have investigated HRV thresholds in cycling. Therefore, these results are novel because they show that the HRV data set may further identify physiological thresholds during running on treadmill.

Some may suggest that the use of fixed blood lactate concentrations to determine LT₁ and LT₂, that is 2.0 and 3.5 mmol.L^-1, respectively, may be interpreted as a limitation of the present study. Some have criticized the use of fixed blood lactate concentrations as this may provide arbitrary and non-individualized threshold intensities (29). In the present study we first identified LT₁ and LT₂ individually, by using visual inspection of lactate curves (29). Probably due to a highly exponential lactate-speed relationship, there was a very low agreement between evaluators when determining LT₁, having this threshold been identified at intensities much lower (~ 40% Speed_MAX) than those suggested for moderate-well trained middle-distance runners performing maximal running tests, that is between 60% - 70% Speed_MAX (24, 29). Thus, we decided to use fixed blood lactate concentrations to determine LT₁ and LT₂, Therefore, we acknowledge that fixed blood lactate concentrations may be limited in some situations, however this method provided non-subjective determination of two thresholds in all participants, thus agreeing with the classical widespread threshold conceptual framework for performance and training prescription (29). In fact, both thresholds were identified within the intensity range suggested for LT₁ (60% - 70% Speed_MAX or VO_2MAX) and LT₂ (80% - 90% Speed_MAX or VO_2MAX) in moderate-well trained middle-distance runners (5, 24, 29).

On the other hand, we identified the lactate threshold according to the LT_Dmax in order to determine an individualized threshold. While this method has been considered as a valuable tool to determine lactate thresholds, two aspects should be pointed out. First, this method is considerably susceptible to minimal changes in blood lactate concentration responses over time during exercise, thereby causing “artificial changes” in lactate thresholds sometimes, mainly by fluctuations in lactate values during initial and final stages (30). Second, despite providing a non-subjective individualized threshold, only one physiological threshold can be determined by it so that we could have been unable to properly answer the research question of the present study.

5.3. Conclusions

This study showed that HRV could be a feasible, noninvasive method to determine traditional and new lactate thresholds during maximal running test, given the good agreement between thresholds identified by HRV data set, with those identified by traditional and more recent lactate techniques (LT₁ and LT₂ and LT_Dmax). In addition, the HRV thresholds showed good reproducibility between test and retest. Thus, thresholds derived from HVR data set may be a reliable and practical tool to determine traditional lactate thresholds during a maximal running test.