Applicability of Dmax Method on Heart Rate Variability to Estimate the Lactate Thresholds in Male Runners

Introduction The purpose of this study was to evaluate the application of the Dmax method on heart rate variability (HRV) to estimate the lactate thresholds (LT), during a maximal incremental running test (MIRT). Methods Nineteen male runners performed two MIRTs, with the initial speed at 8 km·h−1 and increments of 1 km·h−1 every 3 minutes, until exhaustion. Measures of HRV and blood lactate concentrations were obtained, and lactate (LT1 and LT2) and HRV (HRVTDMAX1 and HRVTDMAX2) thresholds were identified. ANOVA with Scheffe's post hoc test, effect sizes (d), the bias ± 95% limits of agreement (LoA), standard error of the estimate (SEE), Pearson's (r), and intraclass correlation coefficient (ICC) were calculated to assess validity. Results No significant differences were observed between HRVTDMAX1 and LT1 when expressed for speed (12.1 ± 1.4 km·h−1 and 11.2 ± 2.1 km·h−1; p=0.55; d = 0.45; r = 0.46; bias ± LoA = 0.8 ± 3.7 km·h−1; SEE = 1.2 km·h−1 (95% CI, 0.9–1.9)). Significant differences were observed between HRVTDMAX2 and LT2 when expressed for speed (12.0 ± 1.2 km·h−1 and 14.1 ± 2.5 km·h−1; p=0.00; d = 1.21; r = 0.48; bias ± LoA = −1.0 ± 1.8 km·h−1; SEE = 1.1 km·h−1 (95% CI, 0.8–1.6)), respectively. Reproducibility values were found for the LT1 (ICC = 0.90; bias ± LoA = −0.7 ± 2.0 km·h−1), LT2 (ICC = 0.97; bias ± LoA = −0.1 ± 1.1 km·h−1), HRVTDMAX1 (ICC = 0.48; bias ± LoA = −0.2 ± 3.4 km·h−1), and HRVTDMAX2 (ICC = 0.30; bias ± LoA = 0.3 ± 3.5 km·h−1). Conclusions The Dmax method applied over a HRV dataset allowed the identification of LT1 that is close to aerobic threshold, during a MIRT.


Introduction
e autonomic cardiac drive can be investigated by the heart rate variability (HRV), which is characterized as a variation quantified in milliseconds between RR intervals [1]. A predominance of parasympathetic nervous system (PNS) activity is observed at rest and low effort intensities. In approximately 50-60% of the maximum oxygen uptake (VO 2MAX ), a significant vagal withdrawal occurs [2]. e aerobic threshold (AeT) has been related to that intensity [3][4][5], i.e., the exercise intensity which lactate concentrations [La] initiate to increase beyond resting values and are frequently called "lactate threshold" (LT) [6,7]. On the other hand, in intensities above the AeT, there is a gradual and constant increase in activation in the sympathetic nervous system (SNS), and a marked increase in the physiological responses related to the anaerobic threshold (AnT) can be observed [3,4]. at intensity is corresponding to maximal lactate steady state (MLSS), i.e., the highest constant exercise intensity output that can be maintained over time without continual [La] accumulation [3,6].
e Poincaré plot is a nonlinear HRV analysis method that uses time domain markers [21] and is an important research area, since it allows its use in nonstationary data, a characteristic inherent to HRV, especially during the increase of effort intensity [22]. e Poincaréplot analysis provides the calculation of the standard deviation of instantaneous (SD1) and continuous long-term RR intervals (SD2) [1]. e SD1 marker has been shown to correlate strongly with vagal tone (PNS), and previous studies have pointed to an abrupt point of change in their behavior in intensities related to AeT [2,5,9,11]. SD2 marker has been shown to correlate with the PNS and SNS, and this variable shows a nonlinear pattern in intensities close to heavy and severe domains [2,11]. e applicability and efficacy of the Poincaré plot in the estimation of the AeT and AnT have been confirmed in previous studies in running [11] and cycling [9,15,23].
However, in addition to the heart rate (HR) which presents theoretical support for a nonlinear pattern, especially in intensity close to AnT [24], some aspects need to be better elucidated when using HRV markers for the estimation of AeT and AnT. Firstly, it would be the validity of the method since the majority of studies used visual analysis for HRVT identification [9][10][11][12]16], which is influenced by the subjective aspect and experience of the evaluator. In order to remedy this limitation, Cheng et al. [25] proposed the Dmax method to identify the lactate and ventilatory thresholds. Previous studies demonstrated greater reliability of the Dmax method than visual analysis or the use of fixed [La] [26].
e Dmax method presents an important advantage which a breakpoint can always be detected [25], although a maximal test is needed. Only one study of our knowledge used the Dmax method to identify HRVTs [23]. eir results surprisingly on the contrary, as reported by the authors, pointed to the visual analysis as better indicators of reliability in the SD1 and RMSSD (square root of the mean squared differences of successive RR intervals) markers than the Dmax method.
is way, it is doubted if the Dmax method is better or not than the visual analysis for the identification of AeT or AnT, when using HRV dataset. Nevertheless, the results of the aforementioned study [23] were not compared with traditional methods to estimate AeT and AnT, as [La] or gaseous exchanges; therefore, greater inferences are limited. Secondly, no study to date has analyzed the possibility to identify the HRVTs by the Dmax method in the different situations and conditions compared to lactate and ventilatory thresholds. Finally, the reproducibility of the method must be verified in relation to different situations and conditions, since it has only been tested on the cycle [15,23]. e identification of HRVTs and consequently the estimation of MLSS can be a framework very important to control and monitor training workloads, as well as to assess the improvement in performance during an endurance training program [21]. e applicability of HRV thresholds in a single-day test perhaps can be very attractive for research studies of sports science, trainers, and practitioners users. erefore, the aim of the study is to investigate the application of the Dmax method originally proposed by Cheng et al. [25], on HRV dataset to estimate the AeT and AnT, during a maximal incremental running test (MIRT) in male runners. Firstly, the hypothesis is that the HRVT identified by SD1 marker (HRVT DMAX1 ) could be used to estimate the AeT, since this marker has been shown to correlate with the PNS [2,9,11,15]. Secondly, the hypothesis is that the HRVT identified by SD2 marker (HRVT DMAX2 ) could be used to estimate AnT, since this marker has been shown to correlate with a significant PNS and mainly with the SNS [11]. e reproducibility of the HRVT DMAX1 and HRVT DMAX2 , as well as AeT and AnT, will be verified.

Participants.
Nineteen male recreational long-distance runners (30.4 ± 4.0 years; body mass of 74.3 ± 8.4 kg; height of 176 ± 6.3 cm; body fat of 13.8 ± 4.5%) volunteered to participate in this study. All participants were healthy, without cardiovascular or orthopedic problems, nonsmokers, and not taking any medication. e study protocol complied with the Declaration of Helsinki for human experimentation [27] and was approved by the Institutional Ethics Committee of the University of São Paulo.

Study Design.
All the participants performed two MIRTs interspersed by a washout period of 3-7 days. e 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 instructed to avoid intense exercises, alcohol, and caffeine beverages 24 hours before each test and to consume a light meal 3 hours before the tests.

Maximal Incremental Running Test Protocol.
Before MIRT, participants used a cardio belt for beat-to-beat heart rate (HR) measures (S810 Polar ® , Kempele, Finland), during rest for 20 min (10 min supine + 10 min sitting) for baseline measures of HR, HRV, and [La]. e [La] was obtained from a 25 μL blood, drawn from the tip of the forefinger, and blood samples were then stored in Eppendorf tubes containing 50 μL of 1% NaF in a − 30°C environment, according to the recommendations of the manufacturer (YSI 1500 Sport, Yellow Springs, OH, USA). Later, the samples were analyzed using enzyme electrode technology (YSI 1500 Sport, Yellow Springs, OH, USA). en, the participants were directed on the treadmill (CEFISE TK35, Nova Odessa, Brazil) and warmed up for 3 min at 5 km·h − 1 and 1% gradient. e test started at 8 km·h − 1 , with 1 km·h − 1 increases every 3 min, until exhaustion, being a protocol adapted by Heck et al. [28]. e HR dataset was recorded continuously throughout the tests. Blood samples of 25 μL were collected during the last 30 s of every stage, while the participant was running.

Heart Rate Variability resholds.
e Dmax method was used to analyze the behavior of SD1 and SD2 markers, to identify the HRVT DMAX1 and HRVT DMAX2 , respectively ( Figure 1). e Dmax method was determined according to Cheng et al. [25], thereby providing individualized lactate threshold values according to the following equation: where Y represents the predicted values of SD1 or SD2 at a given workload (km·h − 1 ); a 3 , a 2 , a 1 , and a 0 are the intercepts; and x is the speed. Briefly, the Dmax method reflects the longest perpendicular distance between SD1 and SD2 values predicted by a third-order polynomial function over actual (SD1 and SD2) values and values derived from a linear regression calculated with the first and last values of each curve, respectively. e SD1 marker was used because it has been shown to correlate strongly with vagal tone (PNS), and previous studies have pointed to an abrupt point of change in their behavior in intensities related to AeT [2,5,9,11,15]. SD2 marker has been shown to correlate with the PNS and SNS, and this variable shows a nonlinear pattern in intensities close to heavy and severe domains, being these related to AnT [2,11]. ereafter, raw RR intervals were recorded during the last 60 s of each stage of the exercise, and then the Dmax method was applied on the measured values.

Lactate resholds.
e first lactate threshold (LT 1 ) (i.e., AeT) was determined as the lowest value of the ratio [La]/speed [29]. After, the second lactate threshold (LT 2 ) (i.e. AnT) was determined as the running speed at 1.5 mmol·L − 1 above LT 1 (Figure 1) [29]. e LT 1 and LT 2 derived from the Dickhuth et al.'s [29] methods were used as criterium measures, because LT 1 has a high correlation with MLSS [7], and LT 2 (+1.5 mmol·L − 1 ) was used because it showed a high concordance with MLSS in runners during the MIRT with stages of 3 min [30].

Statistical
Procedures. Values were expressed as mean and standard deviation (±SD). After ensuring Gaussian data distribution (normality and homoscedasticity), a spreadsheet was used for the analysis of concurrent validity [31] and statistical standards were followed [32]. Cohen's [33] (d) effect sizes and ANOVA with Scheffe's post hoc test were used to compare the magnitude of the differences between the thresholds LT 1 and LT 2 with HRVT DMAX1 and HRVT DMAX2 , respectively. Additionally, the standard error of the estimate (SEE), the bias ± 95% of limits of agreement [LoA] of the Bland and Altman analysis [34], and the Pearson product-moment correlation were used to evaluate the association between the different methods for identifying thresholds. For measures, reliability determination, the intraclass correlation coefficient (ICC), and the typical error of measurement (TEM) were performed using a Hopkins spreadsheet [31]. e d values were interpreted using the following scale: <0.20 (trivial), 0.2-0.6 (small), 0.6-1.2 (moderate), 1.2-2.0 (large), 2.0-4.0 (very large), and >4.0 (extremely large) [33]. Additionally, the ICC and the Pearson correlation coefficients (r) were interpreted as follows: <0.10 (trivial), 0.30 (small), 0.50 (moderate), 0.70 (large), 0.90 (very large), 0.99 (nearly perfect), and 1 (perfect) [31]. e data analysis was performed using the SPSS (19.0). e significance adopted was set at p < 0.05.  Table 1 shows the results of all the methods. Table 2 shows in detail the results of the Pearson correlation coefficient between the methods expressed as absolute and relative values for speed, lactate, RR, and HR. Figure 2 shows the magnitude of differences between HRVT DMAX1 and LT 1 , and HRVT DMAX2 and LT 2 . e Bland-Altman and regression analysis showed between HRVT DMAX1 and LT 1 the bias ± LoA � 0.84 ± 3.7 km·h − 1 and SEE � 1.2 km·h − 1 (95% CI, 0.9-1.9), and between HRVT DMAX2 and LT 2 the bias ± LoA � − 1.07 ± 1.8 km·h − 1 and SEE � 1.1 km·h − 1 (95% CI, 0.8-1.6).  Table 3 shows all results of ICC and TEM for all the methods (Figure 3).

Discussion
e main findings of the present study were that the application of the Dmax method on HRV dataset (SD1 and    I  II  III  IV  I  II  III  IV  I  II  III  IV  I  II  III  IV  HRVT    e results of the SD1 marker, which was used to identify the HRVT DMAX1 , showed values of approximately ≈73.4% of the peak speed value, being these intensities related to the AeT [3,4]. However, the values were slightly above the values reported in previous studies such as Garcia-Tabar et al. [9] analyzing a homogeneous group of professional male worldclass road cyclists (≈36-52% W PEAK ), Candido et al. [23] analyzing healthy individuals (≈50-60% W PEAK ), Sales et al. [20] analyzing individuals with type 2 diabetes (≈64-66% VO 2PEAK ), and Tulppo et al. [2] investigating complete or not parasympathetic blockade.
e HRVT DMAX1 has elicited significant correlation when compared to LT 1 (r � 0.46). e results of the present study in relation to HRVT DMAX1 are slightly below those found by Garcia-Tabar et al. [9], which used the same marker of PNS (SD1) to estimate the LT (r � 0.66-0.88), although different protocols and methodologies were applied. However, it is important to note that only a study of our knowledge by Nascimento et al. [11] used the same ergometer when analyzing HRV indices by the Poincaré plot and [La], in case the treadmill, and all other studies used a cycle ergometer. Consequently, greater comparisons are limits due mainly to the specificity and differences in the movement gesture as well as in the recruitment of motor units [35,36].
It is important to note that the SD1 marker, which presents the prevalence of PNS activity, correlates with indices representing high-frequency bands (HF), such as in the Fourier Transform, when analyzing the behavior of HRV by frequency domain [1]. In addition, studies suggest that the respiratory pattern has a strong effect on the HF-HRV bands, both at rest and at exercise [37][38][39]. During exercise in heavy domain occurs an increase in respiratory frequency with a constant final volume, triggering a mechanical effect on the sinus node, concomitant with an increase in HF. is can be demonstrated in previous studies which identified changes in HF behavior below and above of ventilatory threshold [37]. e SD2 marker used in the present study to identify the HRVT DMAX2 showed significant differences in relation to LT 2 when expressed for absolute and relative values of speed and HR, but not when expressed to lactate and RR, respectively. On the other hand, it is important to note that HRVT DMAX2 has elicited moderate coefficient values (r � 0.48) and significant coefficient values (r � 0.71) when compared to LT 2 , with the values expressed for speed and HR, respectively. is variation in the correlations may be explained in part by the size and heterogeneous characteristic of the sample (body fat coefficient of variation (CV) is 33%). e values found in relation to the HRVT DMAX2 were approximately ≈72.9% of the peak speed value, being these intensities related to the AeT [3,4]. Previous studies showed a breakpoint in SD2 to intensities of approximately ≈80%-90% VO 2MAX during maximal progressive cycling test [2] and intensities of approximately ≈86.1% of the peak speed value during MIRT [11]. However, as previously mentioned in relation to the possibility of identifying a single breakpoint by the Dmax method, these results suggest that the SD2 marker shows a nonlinearity behavior during a MIRT. erefore, in addition to a breakpoint in intensity close to the AnT, as already demonstrated in previous studies [2,11], a significant change point in SD2 also occurs at near intensities related to the AeT.
In the present study, the approximation of HRVT DMAX2 with the AeT, possibly, is due to the use of the Dmax method. is method allows the identification of only one breakpoint, although it is a nonsubjective method when compared to the visual method, for example. However, there are questions concerning the determination of the breakpoint by the Dmax method, in relation to the amount of values used in the model construction, being that the initial and final values can influence and compromise greater inferences when comparing the identification of the AeT or AnT [40].
Perhaps the use of different mathematical models [40] on HRV dataset, even with the possibility of submaximal tests [9], could allow greater accuracy in the estimation of AeT and AnT in different situations and populations involved. Moreover, both HRVT DMAX1 and HRVT DMAX2 are methods relatively simple to analyze and do not require a fixed number of RR intervals nor long recording periods [41], which facilitates the evaluation during incremental exercise. e usefulness of the HRVT DMAX1 and HRVT DMAX2 to estimate the AeT and determine aerobic capacity to prescribe exercise training intensities in sports performance and rehabilitation programs, from SD1 and SD2 values, is of great interest.
e HRVTs may be objectively and noninvasively determined and are of lower cost than lactate or ventilatory threshold assessments, where blood lactate or gas analysis equipment is required as well as specialized professionals.
e LT 1 and LT 2 methods demonstrated a high level of reliability (ICC � 0.90 and 0.97; p < 0.001; TEM � 0.75 and 0.40 km·h − 1 ; TEM � 6.4 and 2.8%, respectively). ese results presented lower values than previous studies using similar markers for determination of LT 1 and LT 2 (TEM � 2.8 and 3.6%, respectively) [42]. Bland and Altman presented low values for difference between the test and retest as a function of their mean (bias ± LoA � − 0.73 ± 2.0 km·h − 1 ; bias ± LoA �− 0.08 ± 1.1 km·h − 1 , respectively). Regarding the HRVT DMAX1 , significant values were found in relation to RR and HR (ICC � 0.80 and 0.81; p < 0.001, respectively), but on the other hand, moderate values were found when expressed in relation to speed (ICC � 0.48), being below those found in a previous study (ICC � 0.73) using the same marker for identification of AeT [15], but through visual identification instead of the Dmax method. Regarding the TEM, results were slightly above the values found in the aforementioned study (TEM � 8.5%). Bland and Altman presented low values for difference between the test and retest as a function of their mean (bias ± LoA �− 0.18 ± 3.4 km·h − 1 ). Already to the HRVT DMAX2 , significant values were found when expressed in relation to lactate (ICC � 0.60; p < 0.05) and RR and HR (ICC � 0.80 and 0.82; p < 0.001, respectively), with low values only when expressed in relation to speed (ICC � 0.30; p < 0.22), corroborating previous studies that used HRV markers by the Dmax method [23].
Due to its low cost, noninvasive nature, and high applicability, HRVTs is an important framework for researchers, trainers, and race practitioners. In the present study, its simplified and nonsubjective identification by the Dmax method suggests the possibility of planning a training program with a safe intensity in a metabolic transition zone, being slightly above the values found for AeT. e present study certainly has some limitations that must be considered in the analysis of the results and their applicability. Aspects such as heterogeneous characteristic of the sample may try to explain in part the variation of correlation values between the different methods, since previous studies report that the greater the heterogeneity of a group, the greater the magnitudes of correlation [32]. Another fact may be just the recruitment of male runners and the need to perform a MIRT. In addition, MLSS was not used as a gold standard, but LT 2 had a good approximation of MLSS in runners [30]. erefore, it is suggested to carry out studies with different characteristics of samples, such as gender, age, and levels of training.

Conclusion
To the best of our knowledge, this is the first report on the application of the Dmax method on a HRV dataset to identify the lactate thresholds, during a MIRT. e results of this study showed that the Dmax method applied over a set of HRV during a MIRT in male recreational long-distance runners allowed the identification of HRVTs approaching the AeT. e HRVTs are of low cost, noninvasive nature, and high applicability. A limiting factor for the interpretation of the data is the recruitment of only males and trained, which do not allow the generalization of results to different populations.
us, further studies are needed to confirm the reproducibility of HRVT as well as its use in different protocols, genders, age groups, and levels of training.

Disclosure
e level of evidence is Level II Study of Diagnostic Test.

HRV:
Heart rate variability VO 2MAX : Maximum oxygen uptake AeT: Aerobic threshold PNS: Parasympathetic nervous system SNS: Sympathetic nervous system AnT: Anaerobic threshold HRVT: Heart rate variability threshold SD1: e standard deviation of instantaneous RR intervals SD2: e continuous long-term RR intervals HR: Heart rate MIRT: Maximal incremental running test LA: Lactate concentrations.

Data Availability
e data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest
e authors declare no conflicts of interest regarding the publication of this manuscript.