See also -
DFA a1 - running ramp and sample rate observations with the Movesense ECG
A popular theme of endurance training is how to measure the various physiologic thresholds. Not only is this helpful in monitoring current performance level, but perhaps as useful is it's role in delineating training zone markers. Zone 3 (of a 3 zone model) is heralded by various concept ideas such as MLSS, RCP, VT2 and LT2. These have been discussed before. More importantly (in my opinion) is the transition from zone 1 to 2. If inappropriate time is spent in zone 2 many negative consequences could occur. As we age and recovery strategies become important, improper time in zone 2 becomes even more problematic.
The non linear HRV index, DFA a1 may have great promise as a means of measuring exercise intensity and training markers. Whether this will translate to a practical, non invasive clue for zone 1 to 2 transition remains to be seen. In order to answer this question, further studies are needed. As shown previously, some particular hurdles need to be jumped over to obtain accurate data, including avoidance of artifacts in the RR series. Perhaps of equal importance is the method of ascertaining the change of the index as intensity rises. A time honored approach involves doing an incremental exercise ramp (to failure). As an example, the "gold standard" incremental ramp for VO2 max, also includes measurement of VT1 or LT1, both accepted markers of zone 1 to 2 transition. LT1 can also be measured by longer steady state intervals of 4 to 8 minutes in length and they are usually equivalent (depending on ramp slope).
With a non linear HRV index that's measured over a span of 2 minutes (hopefully representing a state of equilibrium), we can't be assured that a calculation is proper or accurate since the ramp is not steady state. In the past I have obtained DFA a1 values from 5 minute constant power cycling intervals, with values taken over the last 2 minutes. However, most (?all) currently published research on the DFA a1 relation to intensity is based on the incremental ramp. For example here is a figure from studies of Hautala (first) and Gronwald (second):
Gronwald et al:
The question arises, would a ramp derived value for DFA a1 correspond to one taken from the end of a constant 5 minute segment? Based on the article last year from the Murias group, a slow ramp would largely negate the "mean response time" lag of VO2 parameters. The following post aims to provide some preliminary data comparing slow ramps vs 5 minute constant power intervals in regards to DFA a1 behavior and value range.
Methods - all studies were done after identical light meals, about 8am, using the Hexoskin for RR recording, on an Elite Suito smart trainer, Assomia Duo pedals and with Zwift controlling both ramp progression as well as constant power.
As an aside, the Hexoskin provides an ideal platform for obtaining artifact free single lead ECG data for HRV. Here is a screen recording from my android phone during one of the sessions:
The waveform is downloadable and important for both arrhythmia detection and artifact confirmation.
Kubios software (latest version 3.3.1) was used to determine both average heart rates and DFA a1 for the 2 minute measurement windows. There were no artifacts except in one tracing where a single APC (atrial premature contraction) was noted.
5 minute constant power intervals:
After a 20 minute warmup, 5 minute intervals were done at about VT1-15W (152 or 157 watts), VT1+5W (172 or 177 watts) and VT1+25W (192 or 197 watts), then a 5 minute active recovery at VT1-20W (152W) for 5 minutes - then repeat x 2. The constant power intervals were done on two different days with slightly different wattage on each day (5 watts higher on the second day).
A total of 3 segments for each of the 3 intensities was plotted by averaging the DFA a1 at the last 2 minutes of each active segment.
For example here is the series of data points from the 4/22 session:
The average DFA a1 for the 3 intervals done at 177 watts (in yellow) was about .71 (red arrow).
DFA a1 vs Watts
Blue arrows represent the point where DFA a1 reaches .7 and .5 values. For more detail on what these numbers mean please read this or this.
- Both ramps have very good linear plots with R>.95
- As expected, DFA a1 declines with intensity from correlated, complex values near 1, pass through an intermediate stage until reaching uncorrelated, white noise levels of .5.
- DFA a1 at .7 (between correlated and uncorrelated) was 173, 176 watts - essentially the same.
- DFA a1 at .5 (white noise) occurred at 191 and 190 watts.
- Both ramps showed quite similar results despite using slightly different cycling power intervals.
How does average heart rate vs DFA a1 look?
- The curve R values were .89 and .99.
- DFA a1 reached .7 at about 132 or 128 bpm.
- DFA a1 reached .5 (white noise) at 135 or 132 bpm.
- Not as tight as the power figures, but useful if we are trying to avoid the .5 value - just stay below 130 bpm.
To put this information in some sort of personal historic perspective, here is the data presented in my case report several months ago (also at 5 minute constant power):
The overall figures are in line with the above results.
Now that we have an idea what 5 minute constant power interval data looks like, the next step is to compare this to a continuous ramp.
The first ramp protocol was 10 watts per minute which is a commonly used protocol (otherwise known as 30 watts per 3 minute), starting at 100 watts (after a 20 min warm up).
Zwift ramps are a bit different though, they are truly continuous - there is no abrupt jump between stages. So if you were to do 10 watts per minute incremental rise, the trainer would steadily increase power every second as an equal fraction to maintain that pattern.
The following is a 100 to 310 watt ramp to almost failure, done over 21 minutes, therefore 10 watts per minute:
This is a standard data extract from Kubios with measurement windows every 120 seconds:
Although one can draw a line through the points to get reasonable regression formulas, I decided to try something different. The following is also an extract from Kubios, but with recalculation of 2 minute data every 20 seconds:
DFA a1 vs Avg HR:
- Three different regression curves were placed. I decided to use the one in gray - a linear fit but just through the mid portion. The R was .95 and the arrows in blue again point to heart rates at DFA a1 .7 and .5 values.
- DFA a1 at .7 occurred at a heart rate of 132 bpm
- DFA a1 at .5 occurred at a heart rate of 137 bpm
How are we going to compare DFA a1 to power?
Step one - plot power over time:
Step two - Then use the regression equation to calculate what each watt figure is at each time interval on the DFA a1 curve:
Giving us the following:
We have many more points and several different choices for plotting regression.
- There are many ways to model the regression lines. In light blue is a second order polynomial with an R of .91. However, as in the heart rate curve above, I prefer to just select the mid portion of data (in orange) and do a simple linear regression through those points (R was .95).
- DFA a1 at .7 occurred at 173 watts.
- DFA a1 at .5 (white noise) occurred at 190 watts.
Issue of Reproducibility?
Advantages of not continuing past moderate power:
In academic testing situations as well as for athletic guidance, this process may need to be repeated on a regular basis. Doing a long ramp to failure is not going to fit into a polarized training model nor is something most athletes have time for. However, based on curve shape there are alternatives.
As noted above, at intensities past 200 watts, the curve is relatively flat. Therefore, to minimize physical stress from repeatedly taking the test and avoiding unwanted time in zone 2, perhaps a restricted range ramp can work as well. In addition, using an even lower ramp incremental rise may be helpful (5 watt per min instead of 10 watt per min). The 2 minute window of measurement will still be used but recalculated every 20 seconds as a running data point.
In view of these thoughts, I decided to try an even slower rise through the mid portion - 5 watts per minute which translates to - 130 watts to 230 watts over 20 minutes.
There were two sessions, separated by a few days. Conditions were identical as the constant power tests above. Blue arrows refer to the point on the regression line of reaching a DFA a1 of .7 and .5:
Heart rate vs DFA a1:
DFA a1 vs Watts:
- Linear regression was plotted as in the longer ramps with R values in the .83 to .89 range.
- DFA a1 of .7 occurred at 179 watts or 131 bpm.
- DFA a1 of .5 occurred at 201 watts or 138 bpm.
The same protocol was repeated several days later:
- The outliers (circled) represent the effect of just one APC (atrial premature contraction). Unfortunately, Kubios does not handle this well. Here is a screenshot directly from Kubios, with heart rate and RR (red arrow) on the top, DFA a1 recalculated every 20 sec with 2 minute windows on the bottom. The area in yellow is the result of just one APC (red arrow on top). Correction method was "automatic".
- However, with medium threshold correction, values drop to near .2 around the APC.
- I tried to adjust the regression accordingly which yielded:
- DFA a1 at .7 of 180 watts or 134 bpm.
- DFA a1 at .5 of 193 watts or 142 bpm.
- "Constant 1" and "Constant 2" were composed of three x 5 minute intervals, repeated three times. Each "Ramp" value was derived from one incremental ramp test.
- The DFA a1 values at .7 are all very near my officially measured VT1 (172w).
- At just 20 watts above the VT1, DFA a1 reaches .5, uncorrelated white noise.
- It seems that a variety of incremental ramps will show the same ballpark "power vs DFA a1" profile with respect to values at .7 and .5 as the constant power intervals . A 10 watt per minute to near failure as well as 5 watt per minute to only moderate intensity seem equivilant. However, this does not mean it is reasonable to use a fast ramp such as 30 watts per minute to evaluate the index.
- Based on this very preliminary look at DFA a1 behavior comparing incremental ramps with constant power intervals in just one subject, I am encouraged that multiple valid paths exist to observe the index change with exercise intensity.
- Potential use case senario - If subject "X" was to be tested, they probably should perform over the full range to near failure to find the intensity associated with DFA a1 curve flattening, but thereafter only restricted range testing would be required to minimize stress.
- Hopefully this topic can be evaluated more thoroughly in the future with a careful look at multiple subjects doing both ramps and constant power sessions.
- Even 1 APC can throw off the DFA a1 values in Kubios. With auto correction, values of DFA a1 were higher than usual. Threshold correction was much worse with DFA a1 below .2. Be careful with trend interpretation if any premature beat is present.
- Based on my previous experience, Kubios artifact correction will have significant effects on DFA a1 values. Artifact free recordings are essential until such time as it's proven otherwise.
- Finally, on a personal level, DFA a1 values near .7 seem to correspond to my VT1 as measured with gas exchange.
- At an intensity only 20 watts higher than VT1 (about 190 watts), the DFA a1 pattern becomes uncorrelated (white noise values of .5).
- DFA alpha1, HRV complexity and polarized training
- HRV artifact avoidance vs correction, getting it right the first time
- VT1 correlation to HRV indexes - revisited
- DFA a1 and Zone 1 limits - the effect of Kubios artifact correction
- HRV artifact effects on DFA a1 using alternate software
- A just published article on DFA a1 and Zone 1 demarcation
- Analysis of Hexoskin binary RR interval and respiratory .wav data
- DFA a1 decline with intensity, effect of elevated skin temperature
- Movesense HR+ and Medical ECG review
- Movesense Medical Module and dynamic heart rate variability
- Movesense Medical ECG V2.0 Firmware brief review
- Movesense Medical ECG - improving the waveform and HRV accuracy
- DFA a1 and the aerobic threshold, video conference presentation
- DFA a1 calculation - Kubios vs Python mini validation
- Real time Aerobic thresholds and polarized training with HRV Logger
- DFA a1 and exercise intensity FAQ
Post a Comment