For those interested in the relationship between heart rate and lactate:
It's been over a year now since I started taking lactate samples. I have accumulated over a 100 samples from runs since, and a few more from stationary rides when I got injured recently.
I'll keep this about running though. This might become long, depending on how much I feel on rambling, so I'll add a TL,DR at the footnote of the post.
Objective
Establish whether the relationship between heart rate [HR] and lactate values is strong enough (good correlation) to use it as a primary gauge for lactate-orientated training.
Data
Samples were taking by a Lactate Plus meter. Heart rate was recorded via Garmin HRM-Pro Plus chest strap.
I categorized my 110 samples into three color codes.
1- Green, which means I'm confident about the sample reading being correct.
2- Yellow, which means I'm uncertain and it might be contaminated.
3- Red, which means I'm certain the reading is wrong.
A total of 18 samples fall into the yellow category, while 7 samples are red. That leaves 85 samples in the green category.
I also state whether the samples was taking indoor on the treadmill, or outside.
The list sheet contains rows for each samples, in which the Rep distance/Rep duration/Avg Pace/Avg HR/ Max HR/Temperature/Relative Humidity/Feels Like index/Incline are also logged.
Methods
Comparing the two parameters is done by plotting average heart rate on the X-axis versus lactate values on the Y-axis which is the basis for a linear regression model. The R squared coefficient is then calculated.
This coefficient of determination runs from 0, which means there is absolutely no correlation whatsoever between the two parameters, to 1, which means the variance in the model can be completely explained by one of the parameters.
Six sets of samples were taken into account when plotting the regression model and the subsequent correlation comparisons:
Set #1: All green samples only.
Set #2: All green + yellow samples.
Set #3: Indoor green samples only.
Set #4: Outdoor green samples only.
Set #5: Indoor green + yellow samples.
Set #6: Outdoor green + yellow samples.
Red samples have been omitted from any analysis that affects the next section.
Results
A total of 37 samples were taken outside, with the other 73 samples being indoors on the treadmill. One thing to note is that 6 out of 7 red samples fall into the outside category, illustrating how sweat contamination/poor sampling technique is more prevalent outside versus inside.
Set #1 coefficient is = 0.5212.
Set #2 coefficient is = 0.421.
Set #3 coefficient is = 0.4875.
Set #4 coefficient is = 0.2876.
Set #5 coefficient is = 0.3246.
Set #6 coefficient is = 0.2866.
Discussion
I have to preface by saying this is a N = 1 experiment, but I tried to keep things repeatable and thus give myself the best chance of meaningful data and thus meaningful models. I also must mention that such an analysis has limitations. I won't go into much detail about all that. Other limitations include the sampling size, which is on the lower side for smaller subsets such as "Outdoor green samples only".
The first set has a correlation of 0.5212, which suggests that around 52% of the variability observed in the target variable is explained by the linear regression model.
In comparison, the R square coefficient drops down to 0.421 when including the yellow samples into the mix. In this case, approximately 42% of the variance is explained by one of the model parameters, making it even less correlated.
The correlation goes slightly up a few percent when we look at set #3 where the about 48% of the variance is explained.
However, the correlations go down significantly when we start looking at sets 4, 5 and 6, where the range is 28-32%.
Lactate levels at any HR seemed to be considerably lower on the treadmill vs outside. Secondly, even when comparing identical workouts on the treadmill in different times of the year, I could have a differential of 1.6 mmol/L at an identical heart rate, even when controlling for pace and weather.
Case study: I compared a 6 x 1K workout done on the same treadmill.
Workout #1 took place in October 2023 when I was in peak shape tapering for my HM in the middle of the month. I tested lactate after the last rep and it was 1.8 mmol/L at 3:47 min/km pace with an average heart rate of 164 bpm and a maximum heart rate of 171 bpm.
Workout #2 took place in March 2024, after an extended break from running (4 weeks off in February, only biking on a smart trainer and strength training). Again, I tested lactate after the last rep and this time I got 3.4 mmol/L at 3:48 min/km pace, despite having an average heart rate of 167 bpm and a maximum heart rate of 174 bpm.
These two workouts were in almost identical conditions (24.3 vs 23.4 Celsius, 44.5% vs 47% RH, respectively). The only difference is that the treadmill was set to an incline of 0.5% during workout #2.
The difference between the two lactate readings is quiet significant. If I had chosen to train by a pre-set HR target on the first workout, I would have probably ended up running too hard/fast.
I have other examples on how much the lactate/HR can drift apart under constant HR, but for simplicity and repeatability, I only showcased the 6 x 1K treadmill workout. I made the same observation for outdoor workouts.
Conclusion
In this (on-going) experiment of comparing lactate with heart rate, my experience suggests that there exists a HR/lactate decoupling effect, where the two can be significantly uncorrelated.
As it stands though, training by a set heart rate inferred from a certain protocol might not be suitable in certain conditions (think hot, humid summers). Values obtained from the treadmill do not always transfer well into the outdoor realm as well. One must take into account all these variables and control for them for a meaningful target.
If the desired outcome in a threshold workout is a specific lactate value, there is no going around the fact that a lactate meter is the way to go. Lactate values can be skewed to multiple reasons, including nutrition.
More testing over a longer period will give a better insight about the behavior of the different subset models.
TL,DR
I took over a 100 lactate samples over the past year in different settings and compared them to HR. The relationship between HR and lactate is not strong enough to establish a set target HR for training IF the training is lactated-based (i.e., the goal is to hit a specific lactate value).
Jiggy out!