Modeled VO2 During Ramp Tests and Intervals

We previously looked at how measured pulmonary VO2 (pVO2) compares to WKO-modeled VO2 (VO2mod) during more-or-less constant work rate VO2max intervals. My conclusion was that modeling VO2 from power was unnecessary and would lead to less reliable and less valuable information than simply using power itself. This is mostly because VO2mod is unable to capture the VO2 Slow Component (VO2sc) and therefore under-estimates pVO2 during severe intensity workloads above CP/FTP.

This limitation reflects an outdated assumption that VO2 for a constant workload can be modeled linearly from power during a ramp-incremental test. This is related to an issue I’ve touched on before. How ramp tests using different ramp rates will inevitably produce different results for ‘VO2max power’ (PVO2max) and estimates of Aerobic (AeT: LT1, VT1, etc.) & Anaerobic (AnT: LT2, VT2, MLSS, RCP, etc.) thresholds.


(top) Two ramp-incremental tests with different ramp rates showing VO2 vs Time, producing very different peak power outputs (POpeak, or PVO2max).

(bottom) VO2 vs Power for the same ramp tests as above, showing
similar LT1 (AeT) but very different RCP (AnT)

From Keir et al, 2018 Fig 4

This inconsistency between ramp test protocol is one persistent issue. Another is then translating the resulting workloads based on the ramp test results into prescription of constant work rate intervals.

Protocol Matters! This is something I want to dig into another time.

For now, the best way to understand the differences between ramp-incremental and constant work rate intervals, and the appropriateness of a linear VO2-power model in either context, is to see some real-world examples.

VO2 During Ramp-Incremental Test

Recently I performed a series of maximal graded exercise tests to exhaustion under a few different protocol. The first was a traditional incremental-ramp test, where the workload was increased continuously at a ramp rate of 30 W/min until task failure.

Let’s look at how measured pVO2 compares to modeled VO2mod for this typical incremental-ramp protocol.

Ready for the transition from dark WKO4 charts to light WKO5 charts?


Ramp-incremental test at 30 W/min
Pulmonary pVO2 measured by VO2 Master Pro and WKO-modeled VO2mod

  • Power in yellow
  • Blue line shows Right leg Power (good leg)
  • Pink line shows Left leg Power (bad leg)
  • Heart Rate in red, highlighted above 90% HRmax
  • Modeled VO2mod in Dark Blue (in L/min) in the foreground, highlighted above 90% VO2mod max.
  • Measured pVO2 in Light Blue (in mL/min) in the background, highlighted above 90% pVO2max

Top Left Report:

  • Time >90% HRmax and >90% VO2max
  • pVO2peak is the peak 30sec pVO2 achieved during this ramp test. This value (5270 mL/min) will be referenced as pVO2max on subsequent charts, since this test is the most common ‘VO2max test’ protocol used in the literature.
  • VO2mod peak is the peak 30sec VO2mod achieved during this ramp test. This value (5.150 L/min) will also be referenced as VO2mod max on subsequent charts. Importantly, I found this experimentally derived value to be more accurate than the historical estimate of VO2mod max (below)
  • 90-day VO2mod max is the 90-day modeled VO2max, estimated from historical power data and the WKO power-duration curve. For whatever reason, possibly incomplete data informing the curve, this value is less accurate than the experimentally observed VO2mod max value (above)
  • R2 VO2 shows how closely the two VO2 lines are correlated

Very close! On first glance clearly both VO2 lines are closely related, with a high R2 coefficient. This is where modeled VO2 shines. The algorithm used to infer VO2mod from power would have been originally derived from a ramp-incremental test such as this.

Interestingly VO2mod possibly appears systematically lower than pVO2 as intensity increases through the ramp. This could already reflect a subtle influence of VO2sc above Anaerobic Threshold that cannot be accounted for by the model. VO2peak values are very close at 5.170 L/min and 5270 mL/min, respectively. The peak power output during this test was 443 W.

The second incremental-ramp test I performed was at a slightly higher ramp rate of 40 W/min until task failure. This ramp rate is far higher than would be typically performed. It was actually part of another experiment, but it provides a good comparison.


Ramp-incremental test at 30 W/min

Top Left Report:

  • Note that pVO2max (5270 ml/min) and VO2mod max (5.170 L/min) are retained from the previous ramp test.
  • Compare how closely both VO2peak values during this ramp test compare to previously.
  • R2 VO2 is virtually identical to previous.

The ramp rates 30 W/min vs 40 W/min are close enough that the characteristics should be very similar for both tests. And indeed that is what we see. 30 W/min brought me to VO2max and task failure within 12min, while 40 W/min took only 9min. VO2peak values at failure were very close to each other and to previous VO2max values at 5210 mL/min and 5.230 L/min.

The first inevitable difference with a steeper ramp rate and shorter test duration, is that I will be less affected by fatigue and peak power will be higher. For this 40 W/min ramp test my peak power output was quite a bit higher at 455W at the end of ~9min compared to 443 W at ~12min. This also explains the slightly higher VO2mod peak during this test.

We again see a possible slight under-estimation by VO2mod at higher intensities. This does seem to occur around my expected Anaerobic Threshold, suggesting it could be evidence of a VO2sc effect. However around the same intensity my Left leg also begins to drop power to preserve tissue oxygenation (see the R/L power imbalance on the charts). So the systemic change in pVO2-power may be exacerbated by this condition.

VO2 During Step Test

What about constant work rate intervals with a much slower ramp rate? To get a better look at VO2 during constant workloads I performed a modified 5-1-5 assessment. This is a novel incremental-step test designed by Juerg Feldmann and the team at Moxy to be used with muscle oxygenation (SmO2) to assess physiological strengths and limiters across the intensity spectrum.

The 5-1-5 assessment consists of typically 5 load steps. Each load step is repeated for two 5min work intervals, with 1min passive rest (no pedaling) between each work interval.

Screen Shot 2019-07-02 at 1.10.34 PM

Example of a 5-1-5 assessment with load steps, from Moxy

This test typically isn’t used to assess precise physiological thresholds or training zones, nor does it need to be performed to exhaustion and VO2max. But with some modifications I think the 5-1-5 protocol can be adapted to assess thresholds and zones with better reliability and relevance to constant work rate intervals, than an incremental-ramp test.


Modified 5-1-5 Incremental-step assessment at 50 W/step

Top Left Report:

  • Note the close match between pVO2 and VO2mod at low intensities, as previously discussed. But as intensity increases during these longer intervals, pVO2 begins to outpace VO2mod
  • pVO2peak reaches very close to the same pVO2max as the previous ramp-incremental tests, however the model thinks VO2mod peak was nearly a full litre of oxygen consumption per minute lower during this step test, because of the lower peak power achieved (355 W vs 430-440 W)

Whereas the ramp-incremental tests brought me to exhaustion in 10-15min, the full 5-1-5 step test took just over an hour, clearly contributing to a lower end-test peak power. My peak power was 350 W during the 5-1-5 assessment, compared to 445-455 W peak power during the ramp-incremental tests. We’ll get back to the implications of these workloads in a moment.

Note that during the 2x5min work steps, pVO2 began to outpace VO2mod as early as the third work step at 200 W. This is somewhere close to, but still I would expect under my Aerobic Threshold. Certainly the pVO2-power relationship changes below Anaerobic Threshold, where the VO2 Slow Component (VO2sc) would be expected to affect the linear VO2-power relationship.

I’ve seen this lower-than-expected non-linearity between pVO2 and VO2mod with other athletes as well, so I don’t think this is cause by my particular blood flow limitation. Rather I suspect it reflects a true difference in pVO2 reaching homeostasis during these 2x5min work steps at an elevated VO2 compared to what VO2mod would expect based on the ramp-incremental test.

Implications of Ramp Test VO2 vs Constant Work Rate VO2

Ramp-incremental tests are designed to elicit a linear relationship between workload (power) and physiological response (VO2, blood lactate, SmO2, etc.) in order to bring the athlete to task failure quickly enough such that VO2sc and other sources of fatigue have a negligible effect on that linear relationship. The idea is to be able to use that linear relationship to prescribe training intervals which are expected to elicit the same physiological response during constant work rate, as during the ramp test.

However this linear relationship clearly doesn’t hold for even modest 5-10 minute constant work rate intervals, as demonstrated in the 5-1-5 assessment. Nevermind for longer intervals around Anaerobic Threhsold where VO2sc would further exaggerate this difference.

The incremental-ramp test design basically elicits a constant Phase II rise in VO2 kinetics. The body’s physiological response is constantly trying to catch up to the increasing workload, without having time to reach homeostasis. This design allows measurement of maximum VO2 well enough, but much of the information on changing internal states and metabolic efficiency (eg. physiological thresholds) along the way is lost.

I’ve argued about the importance of understanding metabolic efficiency across the intensity spectrum before, and the importance of internal measurements like Heart Rate, VO2, and SmO2. Knowing how our body produces power, not just how much power, will let us optimize our training to target our particular physiological limiters, our goals, and to what our bodies can tolerate on any given day.

Prescribing Constant Work Rate VO2max Intervals

To conclude, let’s tie it all back to VO2max power and interval prescription.

For roughly the same VO2peak during all three tests (~5200 mL/min), my peak power varied from 350 W during the long duration 5-1-5 assessment, to 455 W during the fastest 40 W/min ramp test. That’s a massively different output for the “same” internal work!

Traditionally this peak power output might be prescribed as my target for VO2max intervals. But clearly the sustainable duration at these workloads will be drastically different. 455 W sounds horribly unsustainable, with or without a bad leg…

Using the estimated GE method of deriving aerobic contribution to power at 5270 mL/min pVO2max, at my estimated gross efficiency this equates to a range of 345-400 W (yeah, there are some large error bars here). The 350 W achieved at the end of the 5-1-5 test is right in this range. So this might be a more appropriate target for maximizing volume >90% VO2max?

Which makes sense, since I was able to hold 350 W for a full 5min interval during that test. I failed during the second 5min 350 W step likely due to accumulated fatigue and only having a brief 1min recovery interval. I would expect to be able to perform around 4x5min intervals at 350 W with sufficient recovery intervals. ie. a classic VO2max interval workout. Whereas I know I wouldn’t be able to complete the same amount of time at 450+ W.

This sounds like something testable! I would be very interesting in seeing how much time >90% VO2max I can accumulate at 450 W and at 350 W, using constant work rate intervals. I only have so many high intensity workouts I can tolerate with my bad leg, but I should be able to do this experiment over the next few weeks. After I finish my current experiment comparing VO2-guided and SmO2-guided 30/15s microbursts.

6 thoughts on “Modeled VO2 During Ramp Tests and Intervals

  1. I’ve used a 2.5′ step protocol that I found to produce pVO2 max numbers that are useful for designing intervals. I like that I go into the last 2 steps a bit fatigued so that I get a more realistic number. Just curious if you loaded the files in WKO4 and got a pVO2 max value. If so, did the numbers from the 1′ protocols vary much?


    1. Hi Todd,

      The PVO2max (power at VO2max) numbers that WKO reports would either (1) just reverse the expression that models VO2 from power. So basically it would just end up reporting peak power from the test, as I reported in the article (430 W, 440 W, and 360 W peak 60sec power for each of the three test protocols, respectively.

      or (2) the WKO model for PVO2max uses the 90-day power-duration curve rather than the numbers from any one particular test or activity. Which as mentioned in the article isn’t really relevant for me, unfortunately, because my power-duration curve isn’t fully filled out to reflect my actual fitness. EDIT: Actually now that I look at it, WKO models my 90-day PVO2mod max at 395 W for a 90-day VO2mod max of 4.850 L/min. So while VO2max is low, PVO2max is ball-park within the estimate range from the 3 test protocol. Apparently my Gross Efficiency is lower than that predicted by the ACSM formula 😅. So it is a relevant number to use for prescribing intervals!

      I’ll turn the question right back at you: When you have done a 2.5min step-test, how do you determine peak power or PVO2max? Peak 1min power? Peak 2.5min power? How does it match up to WKO-modeled PVO2max?


  2. Thx for the thorough reply. I’ve seen that WKO4 takes the avg of the last two steps (5′). I’ve found 95-7% of this number is pretty good for max aerobic intervals of 3-3.5 min’s. I’ve also used 90% of this number for “Seiler’s” 8 min intervals. A test with shorter steps might be a bit easier to fit in a training wk, even if the numbers are a bit higher.
    Unrelated to power, this is my first season in ‘cross using a HR strap and it’s pretty crazy; 45’ @ 95-98% of max HR. But WKO4’s relative VO2 chart only has me at a handful of minutes at 95% or >. 😦


    1. Just saw your post on the WKO FB discussion as well. Let’s discuss! Why do you think the CX race showed such a different HR-power relationship? Or HR to modeled VO2.. which again is just looking at power to estimate VO2?

      Take a look at the expression for VO2 in that WKO4 chart. You’ll see ‘ewma(bikepower,25)’ somewhere. That means it looks at the exponentially weighted moving average (ewma basically means ‘smoothed’) power over 25sec. Then dig into your CX power data. How many continuous high power efforts are there longer than 25sec?

      Much of the energy expenditure during CX in the form of short spiky anaerobic efforts, or “rests” where you might not be pushing on your power meter, but your body is still working hard, eg. running, bike handling, cornering, maintaining balance, etc. It will all elevate HR and clearly the physiological strain is real. The VO2 chart probably just isn’t the most relevant thing to look at for a CX race.


      1. Ha. Thx. There were A LOt of short high efforts. My Pmax and FRC are low, so I approach these races knowing that I have to try to start as well as I can and work my way through the field as the laps tick by. There were only a couple of efforts/lap that were about 25.” In these sections I fully opened it up, knowing that these were the only sections where I could make up time lost in the early laps. So if you think the strain is real, do you think there is a fitness bump from doing cx races? I’ve heard some people say yes and others disagree. My prep this year was a bit diff thatn in years past; I took a more polarized approach and increased training hrs (13-15 tot hrs/wk), with most of the riding at ~70-75% FTP and the other time @ 95-100% of pVO2. I recover better and I feel like I can go a bit deeper when I have to go hard. Thx for all your comments. They’re really appreciated.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.