Thursday, September 16, 2021

Atrial fibrillation - warning signs from chest belt recordings

I've been putting off writing this post for some time for several reasons.  One, the subject matter is complex and usually requires a knowledge of both physiology and medicine.  The other problem is that I don't know if I could do the subject "justice" in a single post.  However, after some additional thought, I decided to go ahead since an education about cardiac arrhythmia and single lead ECG interpretation could potentially help prevent some life changing events.  In conjunction with this post, please read these as well:

Movesense HR+ and Medical ECG review  

Polar H10 ECG tracing - a short how to guide   

ECG artifact strips from Fatmaxxer - a guide

ECG arrhythmia and artifact visualization tips  

So here goes:

The problem:

 (Taken from this review)
  • Even though there are innumerable health benefits from long standing endurance exercise there are also some consequences.  
  • The one we will address in this post centers around the risk of atrial fibrillation.

What is the magnitude of the problem? 

Here is a good review of the etiology, risk factors and theory:

In this review several studies are discussed that describe why this may occur:

One of the first studies on AF in athletes was conducted by
Karjalainen et al., who reported a prevalence of AF of 5.3% in vet-
eran orienteers (aged 47 ± 7 years). Grimsmo et al. reported a
higher prevalence of AF (12.8%) in older (59–88 years) cross-country
. By contrast, Pelliccia et al. reported a very low prevalence
of paroxysmal AF of 0.3% in young elite endurance athletes (mean age
24 ± 6 years). Thus, the diagnostic efforts for identifying SEE-
associated AF should mostly focus on middle-aged long-term endur-
ance male athletes (~45–65 years). This target group comprises
middle-aged men who have been engaged in regular structured SEE
training for years (such as veteran marathon runners who still have
some performance expectations in Masters categories), and especially
those with previous experience at the ‘elite’or competitive level. At a
younger age, the prevalence of AF is usually very low and beyond the
age of 65 AF is likely to be caused, at least partially, by age-related struc-
tural heart disease independent from SEE itself.

In a recent review, Carpenter et al. highlighted that ‘vagal AF’
remains an under-recognized entity, with no universal definition,
etiology, or diagnostic criteria. The observation of atrioventricu-
lar block, asystolic periods, sinus bradycardia and an increase in
heart rate variability (HRV) have been used as criteria to define
vagal AF. Vagal stimuli such as eating, sleeping, relaxation
following stress or exercise and alcohol consumption have been im-
plicated. Carpenter and colleagues underlined that vagal-driven AF
might actually explain many cases of paroxysmal AF among endur-
ance athletes

What exactly is atrial fibrillation?  To answer that we need to first consider the normal electrical flow through the heart leading to orderly contraction of all chambers.

The sino-atrial node (pacemaker area) usually controls impulse initiation in the atria (small chamber on top of the heart).  


The atria contracts first, filling the main chamber (ventricle), which in turn contracts due to the impulses traveling through the atrioventricular node (then His bundle, right and left bundles as above), pumping out a bolus of blood to the systemic circulation (from left ventricle) or to the lungs (from right ventricle).  If the atrial cells (or nearby cells) become inappropriately excitable and initiate contractile activity chaotically, that is atrial fibrillation.  Since the activity is without any coordination, no real pumping action occurs and the atria simply quiver.  If an isolated (non ventricular) area creates a single abnormal impulse, an "atrial premature contraction" APC occurs.  If this arises in the ventricle, a ventricular premature contraction VPC is the result.

Why is atrial fibrillation (AF) dangerous?  By itself, AF can lead to very rapid ventricular rates with heart rates quite high, enough to cause a heart attack in a vulnerable individual with coronary occlusion.  The AV node may not transmit each atrial impulse, so various degrees of block may occur.  In addition, the ventricular bundles may have varying degrees of "refractory times" - periods immediately after an impulse where they can't be excited again until they recover.  This could lead to the atrial impulse being conducted in an aberrent fashion, with abnormal looking electrical activity seen on ECG.

Back to AF.

What's the other danger of AF?  The other consequence may be even worse - the occurrence of stroke.  Why - AF leads to stasis of blood in the atria (since there is no real organized contraction) and the chance of a blood clot forming is high.  If the AF spontaneously corrects and goes back to "sinus rhythm", organized contractile function returns and the clot may be pushed out of the heart to the systemic circulation including the brain.  When the clot lodges in a cerebral vessel, blockade of blood flow will cause a stroke with potential permanent damage to that area.

So to summarize:

  • AF is dangerous and can lead to stroke.
  • It is not uncommon in athletes who have done long standing endurance exercise for many years - just the kind of training that seems so desirable.
  • It is associated with higher vagal activity including higher HRV.  This seems almost paradoxical since higher resting HRV is usually felt to be a sign of better cardiac health and fitness.  While this may be so, one needs to be alert to the fact that AF is an associated finding.
  • It may be associated with some fibrosis and also structural change in chamber dynamics.

This diagram (taken from here) illustrates the U shaped risk of AF with increasing endurance exercise:

What can we do about reducing the risk of AF (besides cutting back on exercise)?

As noted above, certain life style changes such as caffeine and alcohol reduction may help, although I'm unaware of any prospective studies examining this.  I don't think reducing training volume will go over big with most readers....

What else can we do - Observation of early warning signs of AF.  

What are the early risk factors we can see?

Although these studies were not done in athletes, it seems sensible to presume that excessive atrial premature activity would also be a risk factor in athletic AF.

How do we pick up this activity with a chest belt monitor (Polar H10)?

It's actually not difficult.  It can be done in a number of ways, but the best is by using Fatmaxxer to record the RR data.  Other ways include looking for excess artifacts in Kubios/Runalyze etc, then going to the trouble of getting a Holter monitor (24 hr recording) through your physician.  If you have Kubios premium and a Movesense ECG, you have a perfect arrhythmia detection combo.  Coaches and researchers may want to consider that investment.

Fatmaxxer will record a 10 sec strip of good quality ECG data (sample rate 130 Hz) from your Polar H10.  Please see this for details.  It will not record "normal" RR intervals which greatly simplifies the process.  We will need to differentiate noise from atrial or ventricular activity which is as follows.

What does a normal strip look like?

  • Regular, narrow complexes, separated by the same spacing.  Although there is heart rate variability in here, it's too small to see visually.


The electrical baseline is jumping and that is producing distortions - they are not of any concern.

How about multiple atrial premature complexes?

  • I circled the normal ones in green, the early, premature beats in red.  They are all narrow and the QRS is identical looking.  There is a reset of the SA node by the APC (non compensatory pause) noted in orange since the gap is > 2x the RR distance:

Ventricular beats are going to be wider, not have a P wave and have a "compensatory" pause:


And here is a classic VPC captured by Fatmaxxer during one of my sessions:

  • The wide complex is circled in red, the lack of a P wave is circled in orange and the larger than normal T wave is circled in green.

How many are too many?

That's a tough one to answer.  Sometimes it's obvious.  Here is a section from one of the participants from out Frontiers DFA a1 validation study (whose data was excluded):

I circled the normal in black, APCs in red - they run in a paired pattern - normal, then two abnormal then repeat.  It almost resembles AF, but there are P waves visible:

  • I circled the P waves and the arrow shows where the P is "hidden" under the T wave.
  • This would be something to take seriously and investigate.

I would recommend some of the recommendations in the above studies - high order ectopy (doublets, triplets) and high frequency APCs (>30 per hour).



  • Atrial fib is common in endurance athletes.  The high risk candidate is middle age with a history of many years of endurance training (sound familiar).
  • Additional risk factors include frequent APCs, either at rest and/or during exercise.  High resting HRV and high vagal tone also may be involved.
  • A fib can lead to stroke which can be irreversible.
  • With simple consumer HRM equipment and free software, one can identify APCs, their frequency and consult a specialist before a problem develops.

Sunday, September 12, 2021

Rapid HRVT graphing and interpretation

One of the most important questions we can answer with DFA a1 observation is where is our aerobic threshold?  To do so, either constant power intervals or an incremental ramp (not to failure) is needed.  Over the months, I've been asked by athletes and coaches to look over data and weigh in on an interpretation.  What I'd like to do in this post is give a brief approach on how to do this yourself and then present the data in an attractive Excel graph.  To keep this both simple and quick, I recommend using Runalyze as your source of data.  Why?  They use a similar DFA a1 computation method as Kubios, can output a1, HR and power in fine increments (every 5s recalculation) and according to my testing, provide accurate results.  Although we can plot DFA a1 over time (and work out the time vs power or HR), the following approach is easier and will provide comparable results.  I also discussed some general graphing and analysis tips in a set of older posts, here we will just review the HRVT.

Why not use Runalyze HRVT data?  The Runalyze graph uses all data points and even if you just recorded the ramp, we don't want to plot the beginning, stable a1 values (the top of the reverse S shaped curve) or the bottom unchanging nadir values (bottom of the S).

Step 1 - do a ramp in either Zwift or equivalent, with a start power well below your easy pace.  The rate of rise can be from 5 to 30w/min but at the higher rates, power at the HRVT will not be accurate.  If you are interested in cycling power, do a 5 to 10w/min ramp rise.

What about constant power intervals?  Yes, these can be very helpful, but their best use case is with real time monitoring (Fatmaxxer).  For example, cycling at a fixed power for 5-6 minutes may allow one to get an idea at what power level DFA a1 drops to below .75.  We really don't need Excel plotting to do that.

Step 2 - have Runalyze process the file.  The default time for the re computation window is every 60s, we need to change that to every 5s:

Set the window overlap to 115 sec (yeah, it's not intuitive).  Keep the window length and artifact correction as is.  Click Submit.

Then you will see this:

It contains all our needed data.

Step 3 - Click export CSV (in green).  Save the file and open it with Excel

I highlighted the important columns including time (duration end), DFA a1 (alpha), pArtifacts (%artifacts), heart rate and power.

Before you do anything else, save the file as an xlsx:

If this is not done, none of our graphics will be saved.

Step 4 - locate and inspect the data:

We don't want to plot the entire session or even the entire ramp.  What we are looking for is the near linear section from an a1 of 1.0 to about .5 (or less).  Too many repeats of similar a1 values at the start or end will throw off the linear plot.

For example, here is where I'm going to start the plot (line in yellow):

You can see that the alpha column will start with values of about 1 and then drift down.

Ending with these (last plot line in yellow):

As you scan through the data between the start and end plot, make sure the % artifact is reasonable - below 3% (will show as .03 in the column) would be best.  In this case they were zero to 1%.

Step 5 - Selecting the data to graph:

We are going to take a few shortcuts and perform a couple of tricks here to make life simpler.  Since we first want to graph HR vs a1, lets isolate just those two data columns.  Click on the start cell of HR to select it, scroll down to the end HR cell and click with the shift key held down to select that entire batch.  Choose copy.  


Add a new page (bottom of window) and paste those values into A2, to leave room for a heading:

I pasted into the second row of the new page (HR plot, in green) and in the top of the column, then put a label of HR (green also)

Repeat for alpha 1 column:

Now we have this:

Step 6 - HR vs DFA a1 graph and HRVT estimation.  

Click and drag the cursor over columns A and B (not A1 to B1, use the actual top of the columns)

With both columns selected, go to Menu - Insert - Charts and pick "scatter" on the top left of the choice list:

Excel will then automatically place a graph of HR vs a1 on the page.

Step 7 - improving the looks of the plot and getting the HRVT HR

I moved and enlarged the graph area by dragging and resize.

Let's add some axis titles and trendlines:

Select the graph, hit the plus sign then check both axis title and trendline boxes.  I re-titled the graph, axis labels already

Right click the trendline (dotted) and add the R squared and equation if desired.  You can also go to the paint can icon and change the line style/color.

Format the Y axis (a1) - We want an appropriate range and increments of 0.25 with a line at a1=.75.

In green highlights, the min (.25), max (1.25) Bounds and major Units (.25) are listed.  This will make it easier to see where the HRVT falls without resorting to equations.

Format the HR axis:

The min, max Bounds and major Units are set (green)

  • The HRVT HR is about 137 bpm.

Step 8 - HRVT power: 

In this case we will essentially follow the same steps but substitute Watts for HR - Two columns, Watts and DFA a1, plot both and adjust the axis.

I also went into "text options" for each graph component to darken the fonts

  • HRVT power was 204 watts.

To export any graph as a picture, just right click the graph and choose "save as picture"

There you have it!

Yes, if you are an expert at Excel you certainly don't need this, but for those who who aren't and would like a nicer looking, more accurate HRVT plot, this may be helpful.

One more tip - once you have a graph customized the way you want, you can "copy" the "format" to another unfinished graph (in another day's session on a different xlsx).

Step 1 - Control C (copy) the formatted graph

Step 2 - Click once on the unformatted graph to select it.


Step 3 - Go to the "Home" tab, upper left corner is "Paste" - select "Paste Special"

Step 4 - Choose "Format only"

Press "OK"

Note - You may need to right click on the X axis to reset the "Bounds" if the HR or power has changed.

Heart rate variability during dynamic exercise