Table 1 Overview of study results using the domains HRV-index test, LT and/or VT reference test, graded exercise test protocol, participants and results. Note that only values and methods included in the synthesis are reported, as some studies evaluated more than one HRV threshold concept of which the one with the best agreement was then chosen for the synthesis. In the results section the available information on test results, Bland and Altman indices and correlation analysis are presented, respectively. Additionally, effects sizes are denoted as Cohen's d, further confidence intervals (CI), means and standard deviations of values at ventilatory thresholds (VTlow/VThigh), and/or lactate thresholds (LTlow/LThigh) and heart rate variability thresholds (HRVTlow/HRVThigh) are reported

Anosov et al. 2000 [34]

HRVTlow: Hilbert-transform (time–frequency method) using first HF rise after low plateau

VTlow: GET using VCO2/VO2 (V-slope) method

Cycling ergometer, increase 20 W every min

22 (13f, 9m) untrained participants

2–14% difference at VTlow and HRVTlow

Blain et al. 2005 [39]

HRVTlow:

STFT (time–frequency method) using first fRSA increase

HRVThigh: STFT (time–frequency domain) using second fRSA increase

VTlow:

VT1 using VE and VE/VO2 method;

VThigh:

VT2 using VE and VE//VCO2 method

Cycling ergometer, increase 37.5 W every 2 min

14 sedentary men, 12 male endurance athletes

Sedentary subgroup, HF—151 ± 19.5 W at VTlow and 150.3 ± 18.7 W at HRVTlow (d = − 0.037; CI − 0.916 to 0.843), 200.3 ± 29.4 W at VThigh and 198.3 ± 28.8 W at HRVThigh (d = − 0.069; CI − 0.948 to 0.811)

Endurance athletes: 247 ± 33.6 at VTlow and 247.3 ± 32.8 W at HRVTlow (d = 0.009; CI − 0.941 to 0.959), 310.9 ± 26.7 W at VThigh and 316 ± 28.8 W at HRVThigh (d = 0.184; CI − 0.768 to 1.135)

Cassirame et al. 2015 [96]

HRVTlow:

STFT (time–frequency method) using first HF*pHF increase

HRVThigh: STFT (time–frequency domain) using second HF*pHF increase

VTlow:

VT1 using VE and VE/VO2 method;

VThigh:

VT2 using VE and VE/VCO2 method

ski-mountaineering on an alpine slope, increase 0.5 km/h every min

9 (4f, 5m) competitive ski-mountaineers

HRVTlow could only be determined for one participant, bias and standard deviation of VThigh versus HRVThigh − 0.02 bpm and 1.87 bpm, r² from linear regression between HR at VThigh and HRVThigh was 0.91

Cottin et al. 2007 [40]

HRVTlow:

SPWVD (time–frequency method) using first HF*pHF increase

HRVThigh:

SPWVD (time–frequency domain) using second HF*pHF increase

VTlow:

VT1 using first nonlinear increase VE/VO2 while VE/VCO2 remains constant

VThigh

Using nonlinear increase in VE/VCO2 and second increase in VE/VO2

Track running, increase 0.5 km/ every min

12 professional male soccer players

Speed at VTlow was 9.83 ± 1.12 km/h and 10.08 ± 1.29 km/h (d = 0.207; CI − 0.745 to 1.159) while speed at VThigh was 12.55 ± 1.31 km/h and 12.58 ± 1.33 km/h (d = 0.023; CI − 0.927 to 0.972), bias between VTlow and HRVTlow was − 0.25 km/h and − 0.05 km/h between VThigh and HRVThigh, correlation coefficients for HR at VTlow and HRVTlow and VThigh and HRVThigh were r = 0.97 and r = 0.98

Cottin et al. 2006 [36]

HRVTlow:

STFT (time–frequency method) using first HF and HF*pHF increase

HRVThigh:

STFT (time–frequency domain) using second HF*pHF increase

VTlow:

VT1 using first nonlinear increase VE/VO2 while VE/VCO2 remains constant

VThigh:

Using nonlinear increase in VE/VCO2 and second increase in VE/VO2

Cycling ergometer, 15–20 W every min

11 competitive male cyclists and triathletes

Power at VTlow was 219 ± 45 W and 220 ± 48 W at HRVTlow (d = 0.021; CI − 0.971 to 1.013) and 293 ± 45 W at VThigh and 294 ± 48 W at HRVThigh (d = 0.021; CI − 0.971 to 1.013), bias between VTlow and HRVTlow was 0.45 W and 0.91 W between VThigh and HRVThigh, correlation coefficients for HR at VTlow and HRVTlow and VThigh and HRVThigh were r = 0.97 and r = 0.98

Cunha et al. 2014 [87]

HRVTlow:

HRVT using SD1 less than 3 ms (time domain)

VTlow:

GET using modified V-slope first increase in VO2/VCO2

Cycling ergometer, running and walking tests on a treadmill

16 regularly active male students

HR in cycling, walking and running was 133 ± 7 bpm, 141 ± 11 bpm and148 ± 14 bpm at VTlow and 135 ± 9 bpm, 144 ± 16 bpm and 152 ± 15 bpm at HRVTlow (d = 0.248; CI − 0.578 to 1.074; d = 0.219; CI − 0.606 to 1.043; d = 0.276; CI − 0.551 to 1.102), bias between HR at VTlow and HRVTlow ranged from 2 to 4 bpm with standard deviation ranging from 3 to 8 for cycling, walking and running, correlation coefficients for HR at VTlow and HRVTlow for cycling walking and running were r = 0.93, r = 0.92 and r = 0.93

Di Michele et al. 2012 [90]

HRVThigh:

AT-HRV using discrete step before first increase in HF power / HF power-RSA or increase associated with highest speed if 2 increses occurred (time–frequency domain)

LThigh:

AT using lactate turnpoint determined as discrete step immediately before the observation of a sudden and sustained increase

7 × 200 m swimming test with individualized step protocol

14 (8f, 6m) high-level swimmers

HR at LThigh was 182 ± 8.1 bpm and 181.1 ± 8.2 bpm at HRVThigh (d = − 0.11; CI − 0.99 to 0.77), bias between HR at VThigh and HRVThigh was − 0.9 bpm with a standard deviation of 2.95 bpm

Dourado et al. 2010 [110]

HRVTlow:

HRVT using the first point in which the difference in SD1 is less than 1 ms in 2 consecutive stages (time domain)

VTlow:

VT (Wassermann) using VE/VO2 and VE/VCO2

Shuttle walk test, increase 0.6 km/h every minute

10 (3f, 7m) sedentary participants

Speed at VTlow was 5.04 ± 1 km/h and 5.1 ± 1.04 km/h at HRVTlow (d = 0.059; CI − 0.821 to 0.938), bias in oxygen uptake between VTlow and HRVTlow was − 0.05 l/min with a standard deviation of 0.13 l/min

Dourado & Guerra 2013 [111]

HRVTlow:

HRVT using the first point in which the difference in SD1 is less than 1 ms in 2 consecutive stages (time domain)

VTlow:

VT using first nonlinear increase in VE/VO2 while VE/VCO2 remains constant

Shuttle walk test, increase 0.6 km/h every minute

31 (17f, 14m) healthy participants

Bias between VTlow and HRVTlow was 0.04 km/h with a standard deviation of 0.67 km/h, ICC for speed at VTlow and HRVTlow was r = 0.92

Garcia-Manso et al. 2013 [97]

HRVTlow:

Aerobic (HRV) threshold using the first increase of frequency peaks of the high frequency-very high frequency band

HRVThigh:

Anaerobic (HRV) threshold using the second increase of the frequency peaks curve (time–frequency domain)

VTlow:

VT1 by visual detection using VE/VCO2 (ventilatory equivalent

CO2), VE/VO2 (ventilatory equivalent O2), PETCO2 (end-tidal partial pressure

CO2), and PETO2 (end-tidal partial pressure O2)

VThigh:

VT2 based on Gaskill, Ruby [73]

Cycling ergometry, increase 25 W every minute

8 high-performance male cyclists

Bias between VTlow and HRVTlow was 1.88 W with a standard deviation of 5.26 W, bias between VThigh and HRVThigh was − 4.38 W with a standard deviation of 12.04 W

Garcia-Tabar et al. 2013 [91]

HRVTlow:

SD1Tlow using SD1 value 1 ms above minimal SD1 (time domain)

LTlow:

AeT using basic lactate + 0.2 mmol/l

Cycling ergometry, increase 58 W every 3 min

12 world-class male cyclists

HR at LTlow was 132.7 ± 9.6 bpm and 129.6 ± 10.4 bpm at HRVTlow (d = − 0.31; CI − 1.265 to 0.646), bias expressed as relative power between LTlow and HRVTlow was 0.15 W/kg with a standard deviation of 0.29 W/kg, correlation coefficient for relative power at LTlow and HRVTlow was r = 0.88

Granell & De Vito 2018 [112]

HRVTlow:

THFp using the stabilization point of frequency peak of high frequency defined as first point where no significant further change appears

(time–frequency domain)

VTlow:

VT using V-slope

Cycling ergometry, increase 20 W every min

10 regularly active men

Bias between VTlow and HRVTlow was 5 bpm with a standard deviation of 7.65 bpm

Hamdan et al. 2016 [54]

HRVThigh:

Cardiac vagal threshold using minimum of polynomial third order fit of H_C time course (nonlinear domain)

LThigh:

Aerobic threshold according to Dickhuth, Röcker [94]

Cycling ergometry, increase 50 W every 3 min

19 healthy male students

Power at LThigh was 184 ± 28 W and 192 ± 34 W at HRVThigh (d = 0.257; CI − 0.501 to 1.015), bias between LThigh and HRVThigh was 7.95 W with a standard deviation of 18.08 W, correlation coefficient for power at LThigh and HRVThigh was r = 0.86

Karapetian et al. 2008 [35]

HRVTlow:

HRVT (SD) using SD1 and point of no further decline (time domain)

LTlow:

LT using visually determination at first increase from low-intensity exercise

VTlow:

VT using modified V-slope [73]

Cycling ergometry, increase 25 W every 3 min

24 (15f, 9m) regularly active participants

Bias between VTlow and HRVTlow expressed in oxygen consumption was 0.065 l/min and 0.05 l/min between LTlow and HRVTlow with standard deviations of 0.47 l/min and 0.46 l/min, respectively, correlation coefficient for VO2 at HRVTlow and LTlow was r = 0.82 and for VO2 at HRVTlow and VTlow r = 0.89

Mankowski et al. 2016 [42]

HRVThigh:

HRVT2 using visual inspection of nRMSSD minimum (time domain)

VThigh:

VT2 using the second nonlinear increase in VE/VCO2

Cycling ergometry, increase 12–18 W every min

11 (3f, 8m) regularly active participants

Power at VThigh was 251.3 ± 44.7 W and 243.6 ± 44.2 W at HRVThigh (d = − 0.173; CI − 1.167 to 0.821), bias between VThigh and HRVThigh was 4.9 W with a standard deviation of 20.35 W

Mateo-March et al. 2022 [56]

HRVTlow:

HRVT1 using DFAa1 = 0.75

HRVThigh:

HRVT2 using DFAa1 = 0.5

(nonlinear domain)

LTlow:

LT1 using blood lactate increase above baseline

LThigh:

LT2 using increase of over 2 mmol/l above baseline

Cycling ergometry, increase 25 W every min

38 male elite cyclists

HR at LTlow 153 ± 14 bpm and 150 ± 17 bpm at HRVTlow (d = 0.02; CI − 0.30 to 0.34), CI for bias between LTlow and HRVTlow were − 1.62 bpm to 6.99 bpm with CI for limits of agreement ranging from 20.94 to 35.80 bpm and − 30.43 to − 15.58 bpm for upper and lower limits of agreement, respectively, HR at LThigh was 176.84 ± 11.35 bpm and 173.18 ± 12.32 bpm at HRVThigh (d = 0.41, CI 0.07–0.74), CI for bias between LThigh and HRVThigh were 0.69 bpm to 6.77 bpm with CI for limits of agreement ranging from 16.34 to 26.81 bpm and − 19.35 to − 8.88 bpm for upper and lower limits of agreement, respectively

Mendia-Iztueta et al. 2016 [86]

HRVTlow:

HRVT1 visual determination using point of no further decrease of SD1

(time domain), HRVThigh:

HRVT2 visual determination using high frequency power and high frequency power considering RSA point of last increase (time–frequency domain)

VTlow:

VT1 from VE/VO2 and VCO2/VO2

VThigh:

VT2 using VE/VO2 and VE/VCO2

Five different roller-skiing techniques on a treadmill, increase every 3 min (mix of velocity and incline depending on technique)

10 (5f, 5m) national level cross-country skiers

In five different activities (diagonal striding, two skating rhythms, double poling and Nordic walking) bias in HR between VTlow and HRVTlow ranged from − 9 to 9 bpm with standard deviations of 6.63 to 12.75 bpm and bias between VThigh and HRVThigh ranged from − 18 to − 1 bpm with standard deviations of 3.06 to 18.62 bpm

Mourot et al. 2014 [89]

HRVThigh:

VT2 from HRV method 4 using point before increase in high frequency, high frequency considering RSA and considering locomotion

(time–frequency domain)

VThigh:

VT2 using second increase in VE with concomitant increase in VE/VO2 and VE/VCO2

Roller-skiing on a treadmill, increase 0.3 km/h and 1% gradient every min

16 trained male ski-mountaineers

Power at VThigh was 186.4 ± 6.7 bpm and 185.1 ± 6.8 bpm at HRVThigh (d = − 0.193; CI − 1.017 to 0.632), bias between VThigh and HRVThigh was 0.1 bpm with a standard deviation of 3.06 bpm, correlation coefficient for HR at VThigh and HRVThigh was r = 0.825

Nascimento et al. 2017 [41]

HRVTlow:

HRVT1 using Dmax based on SD1

HRVThigh:

HRVT2 using Dmax based on SD2

(time domain)

LTlow:

LT1 using running speed at 2 mmol/l blood lactate value

LThigh:

LT2 using running speed at 3.5 mmol/l blood lactate value

Treadmill running, increase 1 km/h every 3 min

19 trained male long-distance runners

HR at LTlow was 155 ± 17 bpm and 155 ± 15 bpm at HRVTlow (d = 0.00; CI − 0.755 to 0.755) and 173 ± 12 bpm at LThigh and 173 ± 9 at HRVThigh (d = 0.00; CI − 0.755 to 0.755), bias between LTlow and HRVTlow expressed as running speed was − 0.26 km/h with a standard deviation of 2.07 km/h and bias between LThigh and HRVThigh was 0 km/h with a standard deviation of 1.29 km/h

Nascimento et al. 2019 [92]

HRVTlow:

HRVT1 using Dmax based on SD1

HRVThigh:

HRVT2 using Dmax based on SD2

(time domain)

LTLOW:

LT1 using lowest value of lactate to speed ratio

LTHIGH:

LT2 using LT1 plus 1.5 mmol/l blood lactate

Treadmill running, increase 1 km/h every 3 min

19 trained male long-distance runners

HR at LTlow was 151 ± 14 bpm and 155 ± 15 bpm at HRVTlow (d = 0.276; CI − 0.483 to 1.034) and 173 ± 9 bpm at LThigh and 159 ± 15 at HRVThigh (d = − 1.132; CI − 1.945 to − 0.319), bias between LTlow and HRVTlow expressed as running speed was 0.84 km/h with a standard deviation of 1.45 km/h and bias between LThigh and HRVThigh was − 1.07 km/h with a standard deviation of 0.91 km/h

Park et al. 2014 [113]

HRVTlow:

HF times fHFTW determination using intersection of trend lines based on data points below and above VT

(time–frequency domain)

VTlow:

VT using VE against power in W [99]

Cycling ergometry, increase 20 W every min

15 (8f, 7m) trained cyclists

Power at VTlow was 182.8 ± 38 W and 185.1 ± 50.3 W at HRVTlow (d = 0.052; CI − 0.798 to 0.901), bias between VTlow and HRVTlow expressed as power was − 2.4 W with a standard deviation of 23.1 W, correlation coefficient for power at VTlow and HRVTlow war r = 0.90

Queiroz et al. 2016 [114]

HRVTlow:

HRVTSD1 using SD1 of less than 3 ms

(time domain)

VTlow:

VT using increase of VE/VO2 without increase of VE/VCO2

Cycling ergometry, increase 15 W every min

31 untrained healthy men

HR of the healthy sub-group at VTlow was 140.6 ± 14.53 bpm and 137.8 ± 6.32 bpm at HRVTlow (d = − 0.25; CI − 0.843 to 0.343), bias between oxygen consumption at VTlow and HRVTlow was 0.26 l/min with a standard deviation of 0.26 l/min

Ramos-Campo et al. 2018 [85]

HRVTlow:

HRTV1 using point where the difference between SD1 values of 2 consecutive stages was less than 1 ms and no longer changed (time domain),

HRVThigh:

HRVT using final abrupt increase in peak high frequency (time–frequency domain)

VTlow:

VT1 using first increase ov VE/VO2 over workload

VThigh:

VT2 using disproportionate increase in VCO2 over workload and respiratory exchange ratio greater 1

Treadmill running, increase 1 km/h every min

24 professional male basketball players

HR at VTlow was 142.5 ± 9.4 bpm and 140.1 ± 10.5 bpm at HRVTlow (d = − 0.241; CI − 0.915 to 0.433) and 173.5 ± 10.9 bpm at VThigh and 175.1 ± 11.5 bpm at HRVThigh (d = 0.143; CI − 0.53 to 0.815), bias between HR at VTlow and HRVTlow was 1.52 bpm with a standard deviation of 1.2 bpm and bias between HR at VThigh and HRVThigh was 1.16 bpm with a standard deviation of 0.87 bpm, correlation coefficients for HR at VTlow and HRVTlow and VThigh and HRVThigh were r = 0.57 and r = 0.90, respectively

Rogers et al. 2021 [24]

HRVTlow:

HRVT using DFAa1 = 0.75

(nonlinear domain)

VTlow:

VT1 using V-slope, VE/VO2 and excess CO2 [73] and PetO2

Treadmill running, increase 1.3 km/h and 2% gradient every 3 min

17 regularly active men

HR at VTlow was 152 ± 21 bpm and 154 ± 20 bpm at HRVTlow (d = 0.098; CI − 0.701 to 0.896), bias in HR between VTlow and HRVTlow was − 1.9 bpm with a standard deviation of 5.3 bpm, correlation coefficient for HR at VTlow and HRVTlow was r = 0.75

Rogers et al. 2021 [51]

HRVThigh:

HRVT2 using DFAa1 = 0.5

(nonlinear domain)

VThigh:

VT2 associated HR was determined using Oxynet http://oxynetresearch.promfacility.eu

Treadmill running, increase 1.3 km/h and 2% gradient every 3 min

17 regularly active men

HR at VThigh was 174 ± 12 bpm and 171 ± 16 bpm at HRVThigh (d = − 0.212; CI − 1.012 to 0.588), bias in HR between VThigh and HRVThigh was − 4 bpm with a standard deviation of 10.2 bpm, correlation coefficient for HR at VThigh and HRVThigh was r = 0.78

Shiriashi et al. 2018 [93]

HRVTlow:

HRVT using coefficient of component

variance (CCV) L/H elevated to a value of > 0.1, following a

decrease to a value of < 5 ms2 in high frequency continuously for 60 s (frequency domain)

VTlow:

VT using V-slope, VE/VO2 and excess CO2 [73]

LTlow:

LT using visual inspection based on first increase of blood lactate above baseline

Cycling ergometry, increase 20 W every min

30 (9f, 21m) healthy participants

Relative oxygen consumption at VTlowwas 23.3 ± 5.3 ml/kg/min and 23.7 ± 3.7 ml/kg/min at LTlow (d = 0.088; CI − 0.513 to 0.688) and 23.3 ± 5.7 ml/kg/min at HRVTlow (d = − 0.182; CI − 0.784 to 0.42) (d = 0.291; CI − 0.312 to 0.895), correlation coefficients between VTlow and HRVTlow and LTlow and HRVTlow were r = 0.921 and r = 0.853, respectively

Stergiopoulos et al. 2021 [98]

HRVThigh:

HRVT2 determined as point of first abrupt increase of high frequency product after having reached the minimum (time–frequency domain)

VThigh:

VT2 using the concomitant breakaway of VE/VO2 and VE/VCO2

Treadmill running, increase 0.5/h every min and 1 min protocol of Leger et al. 1988 [101]

15 male soccer players

Running speed at VThigh from the treadmill test was 12 ± 1 km/h and 12 ± 0.9 km/h at HRVThigh (d = 0.0; CI − 0.849 to 0.849), speed at VThigh from the shuttle run test was 10.5 ± 0.4 and 10.6 ± 0.4 at HRVThigh(d = 0.25; CI − 0.603 to 1.103), bias of the pooled data from both tests expressed as running speed was 0.02 km/h with a standard deviation of 0.3 km/h