July 26, 2021 at 7:08 am #55966Jon44Participant
Just thought this was an interesting discussion topic. This is from the guy who developed one of the HRV apps and now offers a system (phone gets data from chest strap, outputs to watch app) to use this particular metric of HRV which he describes as being highly correlated with aerobic threshold.
Interesting because it makes some intuitive sense that a relative measure (real-time HRV and its change from some baseline) could be a better indicator than heart-rate.
So I am a PhD scientist (mathematical epidemiology) so I am used to reading papers, but I am not an exercise physiologist. Here are my thoughts:
The average VT1 GAS was 39.8 ml/kg/min (±8.9) compared to 40.1 ml/kg/min (±8.6) obtained by HRVT. The average HR at VT1 GAS was 152 bpm (±21) compared to 154 bpm (±20) obtained by HRVT. … Intraclass correlation between VT1 GAS and HRVT was 0.99 for VO2 and 0.96 for HR. The comparison of VT1 GAS and HRVT showed no differences (VO2: p = 0.347, d = 0.030; HR: p = 0.191, d = 0.091).
This is an incredibly strong correlation and unlikely to occur due to chance.
Seventeen male volunteers aged 19–52, without previous medical history, current medications or physical issues were tested.
The authors state pretty clearly my issue with this, and TBH, I’m surprised they were given IRB approval to do a study on only men, in medicine, this study would not get approved…
Although this study was done with a wide range of subject age and fitness characteristics, no female participants were tested. If the DFA a1 index behavior is to be considered as a zone 1 delimiter for the general population, further investigation using female subjects is mandatory.
Also, 17 participants is a very small sample size.
Participants did not consume caffeine, alcohol or any stimulant for the 24 h before testing.
I dont know enough about exercise physiology to know if this would be a problem, but I would assume that caffeine would impact the heart…
Another area of concern is the transfer of the DFA a1 0.75 breakpoint obtained during incremental testing to that of one found during constant load exercise, including moderate length intervals (5 min). No data is available comparing DFA a1 behavior during an incremental ramp to constant load exercise (Gronwald and Hoos, 2020), making automatic transfer of zone boundaries unclear.
Most people don’t only do constant load exercise, so I am concerned about this.
They list this as a limitation:
Another interesting subject to explore is the impact of athlete overtraining on DFA a1 behavior and VT1 prediction accuracy during exercise. Baumert et al. (2006) did show changes in DFA related scaling behavior after intense training, which may provide both a potential source of HRVT bias and an opportunity to screen for overtraining states.
But I wonder if thats not a limitation at all but a benefit.
Another area for investigation is whether DFA a1 cut off values are equivalent between chest belt and research grade ECG recordings. Although in this study, the RR intervals were recorded with a research grade ECG device, it may be possible to reproduce similar results with chest belt ECG recordings.
the question of DFA a1 value precision obtained by diverse monitoring devices possessing different sample rates and prepossessing strategies. Device sample rates have been shown to variably alter DFA a1 values at rest (Voss et al., 1996; Tapanainen et al., 1999; Singh et al., 2015) but may have more significant effects during exercise.
This seems to be a concern noted in the medium post:
The main challenge of this method seems to be getting high-quality data. Even the best chest straps out there (Polar H7 or H10), generate quite a few artifacts when running, often causing the following computations to be impacted.
and what concerns me is:
Currently, the only app that can compute DFA alpha 1 in real-time is the Heart Rate Variability Logger for iPhone and Android. … For these reasons, the HRV Logger allows you to employ an aggressive artifact removal strategy (“workout” mode under Settings).
My summary is: I think this is an interesting technology that if you have $10 that you don’t know what to do with, this seems like a cool way to spend it and do some N=1 experiments on. If multiple TFUA athletes use this app, I’d love to see a thread where people keep track of their drift test + lab test + HRV app test so we could start to gather some data.
Thanks for your summary analysis.
FYI, one of the links had this podcast discussion of the topic that sounds worth listening to:
I’ve used this app since February and find it interesting/useful/worthwhile. It took me a couple of runs to figure out how to configure the app/strap but once I did I’ve been using it as a rough proxy for my AeT.
I have found that:
* The DFAa1 .75 value correlated well (if a couple beats low) with the results of the 2 self administered lactate test I’ve done since February. Both lactate tests were done when I was fresh.
* DFAa1 Changed by a couple points almost every day – mostly reflected in how I was feeling.
* There’s an art to figuring how to use the data. Artifacts are an issue but you can download the data after the fact and see how many artifacts there are per two minute increment. (The app only gives you DFAa1 values in 2 minutes increments). I tended to disregard the interval if there were more than one or two artifacts.
* It was only good for consistent running. Because of the two minute reading interval, changes in pace under two minutes aren’t properly reflected in the readout.
* In my case, DFAa1 .75 heart rate value on flat ground is different than the .75 value uphill. Sometimes shockingly so. This alone makes me wonder what exactly is being measured based on my (limited) understanding of the concept.
But overall, despite the quirks, I’ve been using the app as a rough check in for a daily AeT. I think it saved me from a training mistake or two (when DFAa1 HR value tanked for a couple of days after a round of heavy training) and it gave me a good visual on exactly how much I had to warm up.
Hi all! I decided to give this a try. I downloaded the app yesterday and used it on my run this morning with my Polar H9 cheststrap. Should mention that my strap has been giving me some grief lately – seems to lose connection mid run and my HR display will drop until it’s in the 60s. Going to submit a warranty claim today, most likely.
Anyway, chest strap did work pretty well for a while today and I figured I likely had some decent data. I opened up the excel file afterward and did find that 0.75 corresponded with 146 bpm (hr drift had put my aet in the low 150s). However, looking at the artifacts per reading, I don’t have any readings with fewer than 6 artifacts. Average of all my readings is 41 artifacts – seems very high! Especially considering what Al said about throwing out any with more than 2 artifacts. Could potentially be related to the issue with my strap?
I’m going to research about how to prevent artifacts and see about getting my cheststrap monitor looked at. Curious what others find!
I don’t know if this app offers anything unique but it did help me for two main reasons.
First, it showed me just how much my AeT varied day to day. Even in an average training block my DFAa1 estimated AeT varied by as much as 5 points (usually under) for reasons that weren’t always obvious. So I started capping my daily runs at HR 140 instead of (lactate tested) 145. In the heavier block I was getting readings that were consistently 7-8 points under. So I went with 135 until the numbers started to improve.
Even with a lactate test I was surprised how much I had to slow down. (Which might be a theme for this website). I don’t think I got faster faster (if that makes sense) but I had better luck adding volume sooner.
Second, because of the way I’m wired, I found it enormously helpful to check in on the numbers pretty much every day. But not everybody is going to feel that way.
That’s probably more than you needed but I hope it helps.
Sorry that happened to you but it sounds like the strap. But if it isn’t, and you’ve configured the app to the recommended settings, I recommend:
* checking to see if your strap how your strap is communicating with the app. Sometimes you can find an option to swap between a bluetooth and an ANT connection. You want the ANT connection.
* making sure that your running motion isn’t moving the strap too much. When you get deep in the weeds on the technology there’s some discussion on how some running motions lead to lower readings (by moving the strap). The solutions were to tighten the strap or to wear a rash guard/tight shirt to keep it from moving.
Al, thanks again for the feedback–so by saying you had better luck adding volume sooner, it sounds like you’re saying the metric is directly helping you avoid an overtrained state. (Or at least serving as a check.) I’m in–I just need to mount a campaign to get the app developer to create an app for the Samsung watch 🙂
And FYI for you and Brittany—in the podcast I linked to above with a physician doing a lot of research in this area, they state definitively that you want Bluetooth, not ANT. (Higher sampling rate and better error correction is the way I understood it.)
Hi Jon and Al!
Agreed, I think it’s the strap. It’s linked with a Bluetooth connection, is strapped on tightly and sits under my sports bra so definitely held tightly in place. I went down an internet rabbit hole yesterday of folks with similar problems with their straps. Since I only purchased in May and it’s still under warranty, I’m going to send back to Polar to see what they can do.
I also reached out to the HRV logger team on their contact form and asked if they had any recommendations for reducing artifacts – figure that might be useful for others as well:
“We definitely see more artifacts with runners compared to cyclists and often this is because of the motion and vibration of running. There are a few steps you can try to reduce the artifacts. The first is to make sure that your Polar H9 strap is tight enough so there is no movement when you run.
Next we recommend trying to reduce the amount of interference from any other connections during your run. To do this you can put your phone in airplane mode and only turn on the Bluetooth connection. You can also make sure that other apps are closed in the background and that your phone battery is fully charged.
A final note is that depending on the intensity of your runs you may see more artifacts. At very low intensity the workout mode will identify many artifacts that are just a result of natural variation rather than relating to signal quality. If your DFA1a is much higher than 0.75 and you’re seeing many artifacts, this may be the case.”
Even with my wacky first readings, similar to what Al mentioned, I imagine this will be helpful in seeing concretely how Aet varies day to day and particularly that it is sometimes lower than I acknowledge. We’ve all read the articles and intellectually know that it varies, but when I saw that 0.77 was corresponding with 146, it made me realize that without that, I would have probably spent quite a bit of my run above 146 (but below 150 or 153) which wouldn’t have been particularly beneficial especially since I’m working in hopes of fixing some ADS.
I can’t imagine this is going to work but –
I once had a polar strap behave the same way because the battery started to move around in the casing. I fixed it by putting a small square of heavy paper on the bottom of the battery before closing the lid. Might be worth trying.
- You must be logged in to reply to this topic.