Monday, April 26, 2021

DFA a1 and the HRVT2 - VT2/LT2

Until recently, if asked  - can the DFA a1 index be used to find the "anaerobic" threshold (VT2, LT2), I would have said ... no way.  My reasoning would be that the dynamic range is becoming compressed at this data range and artifact related bias would obscure the detailed results needed.  However, when we decided to take a more formal look at data from high quality ECG, near zero artifact tracings, it does appear that a relationship does exist.  Moreover, it seems to coincide with the DFA a1 value of .5, which has mathematical significance as an indicator of a random correlation pattern.

Today our article was published and I would like to share some of my own tracings looking for the HRVT2.

The following was done with a Polar H10, BSX Insight muscle O2 sensor, smart trainer and Zwift to control power.

Based on previous gas exchange and lactate testing, my second threshold was about 270 watts.  I decided to cycle 15w below and above this power to look at a1 and the muscle O2 curve of the Rectus femoris.

  • I crammed everything in one graph.
  • Muscle O2 (orange) is quite stable during the interval below LT2, but rapidly declines at 15w above it.  This is typical for rectus femoris behavior below and above the LT2 and confirms my previous testing.
  • DFA a1 average during each of the 6 minutes was .65 for the below, .51 for above the threshold power.  Artifact was below 2%.

Ramp data

The next was a typical ramp, 10w/min until near failure.

  • HRVT2 came in about at a HR of 155 and power of 275w.  Artifact was 3 to 6% toward the end, so the a1 figures might be biased somewhat.
  • Although this does agree with my lactate figures, a redo with fewer artifacts would be highly recommended. 
  • The presence of artifact is a major limitation to this technique.


  • Either a traditional ramp or 5-6 minute sequential intervals could be done to look for the HRVT2, the point at which DFA a1 crosses the .5 value. 
  • However, it could be a "heavy lift" - artifacts, device bias and software bias (HRV logger vs Kubios) could be complicating factors to getting accurate values
  • In addition, even our study ECG data had relatively wide "limits of agreement" and moderate-strong correlation coefficients (as opposed to very strong).
  • Future research is needed to better understand, confirm and apply this information to athletic training and intensity distribution.


  1. Hi,
    How large do you think the individual variation in the DFA a1 value is in relation to the both the aerobic and anaerobic thresholds?
    I just did a quick analysis of some of my own data and the DFA a1 values was a bit higher then I would have expected in relation to the lactate values I have measured one the different loads.

    Do you think the DFA a1 can be use to measure (indicate when to retest) short to moderate term performance changes?


    1. The value spread for the second threshold is definitely wider than the first (see both article Bland Altman plots).
      Depending on how you calculated your a1 data (HRM device, kubios vs hrv logger etc), yes, there certainly can be variation. And remember, even the ECG derived thresholds were not all "perfect".
      Your last point is key, we believe that an improvement in a1 threshold translates to an improvement in corresponding lactate threshold.

    2. Yes, but the variation is not too high.
      It is definitely useable to get an idea in what rage the thresholds should be, especially for the aerobic threshold.
      But of course as you say they difference in HRM-bands and software will be important to take in to account.

      I am really interested using the measurement to evaluate performance improvement, so I going to test it out for that purpose.

      Thx for the reply!