Muscle Oxygen Training: Combining NIRS and DFA a1 for critical intensity estimation

Preface - before we discuss the details about combining NIRS related muscle O2 desaturation breakpoints and HRV "surrogate" threshold estimation techniques, I would like to provide some background on how this project came to be. Well over a year and a half ago, I began thinking about ways to improve localization of the second exercise intensity threshold (AKA - RCP, MLSS, MMSS, VT2, LT2, CP, FTP etc.). As I explained in the prior post, would combining "surrogate" thresholds of two disparate methodologies yield a better personalized agreement with a gold standard test. Many of these surrogate concepts are well rooted in published data, but each single method generally has a large per individual error, despite group averages being very close to a gold standard. My initial method of choice was to average the HRVT2 and NIRS/muscle O2 desaturation breakpoints (HHb BP) resulting in a "Combo" hybrid. NIRS has a long and sometimes controversial history of claims plus industry "hype" which I've covered before in this blog (many times). In fact, the reason I started this blog was the misleading statements coming out of the Moxy forums and elsewhere. But, having said that, there is no question that there is one scenario where NIRS related HHb BP may be helpful - identification of the "critical intensity" (RCP/MLSS or Maximal Metabolic Steady State, MMSS) which we'll abbreviate as "CI". Unfortunately, as with the HRVT2, personalized identification of CI is quite different from getting the group average to agree. Since the underlying basis for the HRVT2 and HHb breakpoints are very different, perhaps looking at combining each would give us a better idea of a given individual's true CI. Testing this hypothesis should not be too difficult - we would need incremental ramps with NIRS probes/gas exchange and (good) HRV recordings, then analyze the data and go from there. Ideally, a research group who excelled at NIRS breakpoint and gas exchange methodology would be a perfect fit for this endeavor. For me, that choice was obvious. One of the premier experts in that field was a researcher that I have corresponded with and followed for many years, Juan Murias. We began discussing the potential project (with his PhD student Pablo Fleitas-Paniagua) and went on to do the study. Their group did the full testing "suite" (the only data I received were the RR intervals and Fatmaxxer ECG snips). My job was to find the HRVT2 HR and elapsed time on the ramp (for VO2 determination). Pablo and the Calgary team did the rest. As we will see, the results were quite interesting. Many months later, I evaluated the ECG respiratory rate threshold combined with DFA a1 HRVT (previous post), but NIRS/HRVT2 was always the primary project. It just took longer to write, get reviewed and published.

I would like to express my sincerest thanks to Juan, the Calgary group and Pablo for working on this project. Pablo did a fantastic job on his first major journal publication - he has a very promising future. Additionally, I wanted to thank Juan for agreeing to add this study to his already busy set of projects. Over the years of reading the sports literature and writing this blog, his work has always stood out as being of the highest quality. This project was not only fun but an incredible learning experience for me. Okay, enough of that - let's look at the study:

First off - what is "Critical Intensity"

As many of you already know, we can split up endurance exercise intensity into 3 zones;

Zone 1 - done below the first lactate or ventilatory threshold
Zone 2 - done between the first and second lactate/ventilatory threshold
Zone 3 - done above the second lactate/ventilatory thresholds which represent the critical intensity (CI) (From the intro - Above this boundary, within the severe intensity domain, intracellular homeostasis becomes unsustainable)

However, lactate threshold definitions are definitely "fuzzy" and if using lactate as a guide for the second threshold, the maximum lactate steady state (MLSS) is really the only way to go. That's difficult to do as well as exhausting. Another way to derive the CI is to do a formal gas exchange test - which is not that simple to arrange. Other methods used to investigate the CI include the FTP and critical power (CP). Both rely on achieving maximum volitional intensities of varying length. I discussed the FTP test in a previous post and the reasons why I am not a huge fan. Many of these same reasons apply to the CP as well. Depending on your pacing, motivation and whether the legs are "fresh" or not, error is certainly possible. These also may require several sessions to exhaustion to get results.

For these reasons, we were especially interested in whether the combo of muscle O2 desaturation breakpoints and HRVT2 could be a viable way to find the CI.

Why is it important to establish the CI?

Intensities as low as 10 watts above the CI can have profound effects on neuromuscular fatigue and lactate. Therefore, knowledge of this physiologic departure point can help avoid inappropriate intensities during training (or racing). Spending too many minutes above the CI could end up costing an individual a portion of their peak potential later on during the session or race. On the other hand, many polarized training programs encourage and include zone 3 training (above the CI) rather than zone 2 (below the CI).

A step by step review of the paper:

INTRODUCTION
Accurate identification of the critical intensity of exercise is crucial for a precise prescription of endurance training as understanding the metabolic disruptions elicited by a given stimulus is key for predicting the acute responses to individual training sessions, as well as longer-term training outcomes. Below this boundary, within the heavy exercise intensity domain, constant load activities can be sustained for relatively long periods of time (i.e., >50 min) while maintaining intracellular homeostasis, as blood lactate concentration ([Lac]b) and oxygen consumption (V̇O2) can be stabilized. Above this boundary, within the severe intensity domain, intracellular homeostasis becomes unsustainable and [Lac]b and/or V̇O2 can no longer reach steady state responses, which leads to a faster task failure due to greater metabolic instability.
Although a subject of continuous debate, two different methods are commonly used in research studies to get the best possible estimation of the metabolic and work rates associated with the critical intensity of exercise: the maximal lactate steady state and the critical power. Despite their sound theoretical and even practical support, these methods require that participants complete several visits to the laboratory, some of which need performing either time to task failure trials or sustaining high intensity bouts for long periods of time. This might be inconvenient in some populations, and simply unfeasible in others.
A different approach to determine the boundary separating the heavy from the severe intensity domains is by identifying the respiratory compensation point (RCP) during ramp incremental (RI) exercise, which has been shown consistently to occur at the metabolic rate (i.e., the V̇O2) associated to the critical intensity of exercise. This method, combined with constant load square wave transitions within the moderate and the heavy intensity domains of exercise before and after the RI test, respectively, allows for determination of the work rate associated to the critical intensity of exercise. However, this approach relies on the use of gas exchange evaluations, mostly performed in laboratory settings.
Within the past two decades, the use of the deoxygenated hemoglobin signal ([HHb]) from near infrared spectroscopy (NIRS), as well as measures of heart rate variability (HRV) have been proposed to estimate the critical intensity in a single laboratory visit using less expensive equipment. For example, the break point in the vastus lateralis [HHb] signal ([HHb]BP) towards the end of a RI test has been shown to correspond to the metabolic rate at the RCP. This measure has been reported to be repeatable, and to track changes in fitness levels and environmental conditions.
Similarly, the use of detrended fluctuation analysis of HRV has been indicated to provide a close approximation of the critical intensity of exercise. This approach involves the measurement of the cardiac beat sequence to ascertain fractal correlation patterns that reflect the degree of self-similarity of the sequence over different time scales. Multiple studies have indicated that the degree of fractal self-similarity declines with increasing exercise intensity, with the decline in the detrended fluctuation analysis and its short-term scaling exponent alpha1 (DFA a1) below 0.5 being associated with the RCP or the second lactate threshold in recreational runners and professional cyclists, respectively.
As opposed to more complex and/or invasive modalities such as gas exchange or lactate testing, current technologies allow evaluations of [HHb] and HRV at a relatively low cost and without the need to access a laboratory. Thus, there is great potential for using these technologies in the field for establishing the boundary separating the heavy from the severe exercise intensity domains. However, despite studies showing strong agreement of group averages of [HHb]BP to maximal lactate steady state (MLSS), CP, and the RCP or DFA a1 second threshold (HRVT2) to the RCP, differences might be larger than desirable for exercise prescription when considered individually. Therefore, making efforts to minimize biases/errors when estimating the heavy to severe domain exercise intensity boundary is important. An interesting strategy that has not yet been considered is to blend both NIRS and HRV threshold data for critical intensity determination, as they each are derived from different physiological subsystems, likely underlying the same physiological phenomenon. Hence, it is possible that an erroneous estimation from one subsystem could be counterbalanced by determination of the other. Therefore, combining threshold estimations from these two different approaches might further improve the reliability and precision of the outcome.
Thus, the aims of the present study were to determine: (i) the difference between the HR and V̇O2 responses associated to the [HHb]BP, the HRVT2, and the RCP; (ii) if by averaging the HR and V̇O2 responses associated to the [HHb]BP and the HRVT2 (H&HAv) during a RI test, improved the accuracy of identifying the HR and V̇O2 associated to the RCP. It was hypothesized that, even though the differences in the HR and V̇O2 responses associate to the [HHb]BP and the HRVT2 would not be significantly different to the RCP, with H&HAv responses, the variability in relationship to the RCP would decrease in comparation with the variability expressed by each method evaluated independently.

Highlights:

Definition of CI
Current methods of determination (with pitfalls)
Surrogate options including muscle O2 desat breakpoints and HRV (DFA a1 based)
Proposal to combine 2 surrogates for a combined average that is hopefully in better agreement to gold standard on a per individual basis

Methodology:

Data collection
RI test. This test was performed on an electromagnetically-braked cycle ergometer (Velotron, RacerMate, Seattle, USA). The RI test included 4 min of baseline cycling at 20 W, 6 min of moderate intensity (60 – 80 W), and 4 min at 20 W followed by 15 W·min-1 incremental rate. The RI test stopped at task failure, which was defined as the inability of participants to maintain a cycling cadence of at least 60 rpm for more than 5 consecutive s, or at volitional exhaustion despite strong verbal encouragement. During the baseline and moderate intensity cycling part, participants maintained a cadence of 60-70 rpm, whereas during the RI test participants self-selected the cadence. Participants were blinded to the work rate and elapsed time but received visual feedback on their cadence.
Gas exchange, and ventilatory variables. All gas exchange and ventilatory responses were measured breath-by-breath using a metabolic cart (Quark, Cosmed, Rome, Italy). The system was calibrated before all tests according to the manufacture’s recommendation and consisted of a low dead space turbine as well as oxygen (O2) and carbon dioxide (CO2) gas analyzers; a syringe of known volume (3 L) and a gas-mixture of known concentration (16% O2; 5% CO2; balance N2), respectively, were utilized for calibration.
Near Infrared Spectroscopy. NIRS-derived [HHb] was measured continuously over the vastus lateralis muscle belly of the dominant leg (Oxiplex TS; ISS, Champaign, USA) at a sampling rate of 10 Hz throughout the entire protocol. The NIRS probe was placed on the belly of the vastus lateralis muscle midway between the inguinal crease and the proximal border of the patella. The probe was secured in place by double-sided tape and an elastic strap to prevent any movement, and it was covered by an optically dense, black vinyl sheet and an elastic bandage to minimize both the intrusion of external light and movement.
RR Measurements: Each participant’s RR time series was recorded by a Polar H10 strap (Polar Electro, Kempele, Finland) with a sampling rate of 1000 Hz. The Polar strap electrodes were covered with conductive gel and firmly fitted to the sub pectoral area with the module centered over the sternum. Before testing, the Polar H10 ECG waveform was visually evaluated with the Android app ECG Logger (https://ecglogger.en.aptoide.com/app). To optimize DFA a1 measurements, the strap was shifted slightly to the left if the R peak amplitude was lower than the S wave. H10 data was transmitted via Bluetooth to an Android smartphone running an open-source recording application (FatMaxxer, https://github.com/IanPeake/FatMaxxer) and offloaded for further analysis.

Data analyses
V̇O2 data: during the RI tests the V̇O2 data were cleaned by removing data points laying ±3 SD from the local mean and linearly interpolated to 1 s intervals (Origin, Origin Lab, Northampton, MA). The V̇O2 values were computed from a 20 s rolling average during the RI test. The RCP was assessed by three independent experienced evaluators, based on previous recommendations. Briefly, the RCP corresponded to the second disproportional increase (second breakpoint) in the V̇E/V̇O2 relationship, where the end-tidal pressure of CO2 began to fall after a period of isocapnic buffering. The relationship between V̇E/V̇CO2 against V̇O2 was also considered for confirmation of the RCP. The average value from the three evaluators was used. If the evaluators had a disagreement of more than 100 mL∙min-1 in the result, a second round of evaluation was performed together until a consensus was reached. The V̇O2 mean response time (MRT) was calculated as previously described by averaging the last minute of the steady-state V̇O2 responses from the 60-80 W constant work rate cycling that preceded the RI test, and superimposing on the V̇O2 vs power output relationship of the ramp-incremental exercise. The difference in power output between the steady-state V̇O2 and the ramp-derived V̇O2 at the point of intersection with the linear fit was converted to time to retrieve the MRT. MRT was used to correct V̇O2 associated [HHb]BP.
[HHb] signal: The [HHb] raw data during the RI tests were averaged into second-by-second responses using Origin software (Origin Lab, Northampton, MA). As previously described, the [HHb]-time relationship was modeled with the following piecewise model:
ƒ= if (x < BP, g(x), h(x))
g(x) = i1 + (s1 · x)
i2 = i1 + (s1 · BP)
h(x) = i2 + [s2 · (x − BP)]
fit ƒ to y,
where ƒ is the piecewise function, x is time and y is [HHb], BP is the time coordinate corresponding to the interception of the two regression lines (i.e., the [HHb]BP), i1 and i2 are the intercepts of the first and second linear function, respectively and s1 and s2 are the slopes. Model parameter estimates for each individual were determined by linear least-square regression analysis. The piecewise fit was started at the onset of the systematic increase in the [HHb] signal until the last data point corresponding to the end of the test. Associated HR was determined by the time at which [HHb]BP occurred. Five participants with low signal and/or excessive noise were excluded from analysis.
RR data for each participant was imported into Kubios HRV Software (Version 3.5, Biosignal Analysis and Medical Imaging Group, Department of Physics, University of Kuopio, Kuopio, Finland). Kubios preprocessing settings were set to the default values including the RR detrending method which was kept at “Smoothness priors” (Lambda = 500). DFA a1 window width was set to 4 ≤ n ≤ 16 beats. Visual inspection of the entire test recording was done to determine sample quality, noise, arrhythmia, and missing beat artifact. The RR series of the included participants was then corrected by the Kubios “automatic method” and relevant HRV results exported as text files for further analysis. Acceptable percent artifact occurring during threshold interpretation segments was set to below 5%. Two participants with excessive atrial, ventricular ectopy and/or artifact above 5% were excluded from analysis.
In order to determine the V̇O2 or HR corresponding to a DFA a1 of 0.5, the DFA a1 was calculated over time using 2 min HRV measurement windows with a recalculation every 5 s throughout the test. Two-minute time windowing was chosen based on the minimal recommended RR intervals required for accurate computation. Plotting of DFA a1 vs. time or HR was performed. The relationship generally showed a reverse sigmoidal curve, with a stable area above 1.0 at low work rates, a rapid, near linear drop reaching below 0.5 at higher intensity, then flattening without major change. A linear regression line was drawn through the appropriate section with the HRVT2 defined as the RI time or HR where DFA a1 equaled 0.5. The V̇O2 of DFA a1 reaching 0.5 was then derived from the V̇O2 vs. time relation based on gas exchange analysis. The HR where DFA a1 reached 0.5 was calculated based on the regression equation from the linear section of the HR vs DFA a1 plot. Figure 1 (below) displays the different approaches to estimate thresholds for RCP, [HHb]BP and, HRVT2.

Matching everything up - how the NIRS breakpoint, RCP and HRVT2 behave during the ramp:

Although the above is a very good "fit" for the NIRS BP and HRVT2 (DFA a1 = .5), not every case was this well-matched.

Highlights

The Vastus lateralis was chosen for HHb BP - the breakpoint presents as a plateau of the O2 desaturation
Fatmaxxer was used to collect the RR intervals and evaluate any arrhythmia (which was found in two individuals).
The Polar H10 belt was adjusted to maximize the ECG R peak by rotation.

Results:

Highlights:

Both HRVT2 and HHb BP averages were statistically equivalent to the RCP (gold standard CI).
The combined group had better individual agreement.
If one test failed due to technical reasons, the other method was usually (always in this case) able to provide a decent result.

Discussion

"The present study compared the V̇O2 and the HR responses associated to the RCP, the [HHb]BP, the HRVT2, and the H&HAv. The main findings were that: i) in agreement with previous literature, independently comparing the [HHb]BP and the HRVT2 to the RCP resulted in no significant differences and negligible biases for V̇O2 and HR values associated to any of the techniques employed for threshold determination; ii) by averaging the V̇O2 and the HR responses associated to the [HHb]BP and the HRVT2 for comparison with the responses observed at the RCP, the mean bias and the LOA were reduced (albeit not significantly) compared with evaluating each threshold marker (i.e., [HHb]BP and the HRVT2) to the RCP independently. These findings not only highlight the value of using [HHb]BP and the HRVT2 as valid tools to establish the V̇O2 and the HR responses associated with the critical intensity of exercise, but also provide novel information on the value of combining these two signals to potentially improve the accuracy of threshold determination.
In this study, the RCP was used as the reference for the metabolic rate at which the critical intensity occurs. Some studies have questioned the use of the RCP as a demarcation of the critical intensity of exercise based on the idea that the PO associated to the RCP following an incremental test to task failure is typically greater than the PO at the critical intensity during constant load exercise. However, those studies did not consider the dissociation between V̇O2 and PO that exists during incremental compared to constant load exercise due to the existence of the V̇O2 MRT and the inability of the V̇O2 slow component to be fully expressed during incremental exercise testing. In fact, recent studies have demonstrated a coincidence in the PO associated to the RCP and the critical intensity of exercise when proper adjustments are made. Regardless, that the metabolic rate at the RCP coincides with that observed at the critical intensity of exercise has been extensively demonstrated, which justifies the use of this marker to evaluate the V̇O2 and the HR associated to the critical intensity of exercise.
Considering the relevance of establishing the critical intensity for adequate exercise prescription that allows to maximize positive adaptations to exercise training, identification of this boundary using simpler and accessible tools becomes important. Within the past two decades, the use of the [HHb]BP and the HRVT2 have been proposed as valid options to determine the threshold separating the heavy from the severe domains of exercise, with both approaches providing promising outcomes, but also generating debate in relation to their precision to accurately estimate what they intend to evaluate.
In connection to this, data from the present study have shown strong agreement between the V̇O2 and the HR at which the [HHb]BP and the RCP occurred. This provides further support to the validity of using the [HHb]BP as a tool to approximate the critical intensity of exercise. Although some reports disagree with this idea, the [HHb]BP has been consistently shown to occur at a metabolic rate that coincides with that of the RCP and the critical intensity of exercise and/or the MLSS response. It has been speculated that the [HHb]BP reflects the change in intracellular metabolic conditions that also result in the RCP, which likely trigger local vasodilatory responses that improve O2 availability within the active tissues. This event results in less reliance on O2 extraction to support the increased metabolic rate, and thus a plateau or a decrease in the slope of the [HHb] signal. Importantly, the [HHb]BP has been indicated to coincide with the metabolic rate at the critical intensity in different groups43, following endurance exercise training, and under different experimental conditions that challenged the oxygen transport system.
Data from the present study also have shown strong agreement between the V̇O2 and the HR at which the HRVT2 and the RCP took place. A DFA a1 of 0.5 represents uncorrelated, random behavior of the interbeat pattern with loss of fractal properties. This has been proposed as a marker for a potential loss of homeostasis and described to be sustainable for only short time spans44. The underlying basis of these correlation patterns is largely due to changes in the autonomic nervous system induced sinoatrial pacemaker function. During exercise there is both a withdrawal of parasympathetic and an increase of sympathetic activity resulting in a change of HRV. Alterations in DFA a1 provide a view of autonomic balance from rest to severe intensity domains different from other established systemic internal load markers of intensity that depend on gas exchange (i.e., V̇O2, V̇CO2), ventilatory (i.e., V̇E) , and biochemical (lactate) variables, as well as from measures of external load (i.e., speed/power). The HRVT2 has been evaluated in different studies in males during running and cycling using step incremental tests, as well as in females during ramp incremental test.
In addition to providing further support to the use of the [HHb]BP and the HRVT2 as valid tools to establish the critical intensity of exercise, this study demonstrated that, by combining these two measurements, the mean bias and the LOA were reduced in relation to the individual evaluation of each threshold, compared to the RCP. Given the increasingly lower costs associated to portable NIRS systems and HR monitoring straps that allow for precise evaluation of HRV, combining these two evaluation techniques seems like a viable option to either improve the confidence of the estimation, or to ensure that at least one outcome is obtained. For example, it is known that each modality has its own technical limitations/challenges. For example, NIRS may be negatively affected by subcutaneous fat tissue and probe movement. On the other hand, HRV is greatly affected by the quality of the RR/ECG recording, which is highly dependent on normal, non aberrant heart beats and good ratio between signal and noise. Thus, attempting to obtain outcomes from both devices would be advisable as, in addition to improving the precision of the outcomes when combining measures, when an individual test fails, a single test result can still represent a valuable marker that is only marginally less precise than the dual average. This is an important consideration because, as it can be gathered from Table 2, even though missing data occurred for both the [HHb]BP and the HRVT2, at least one of these demarcation points was obtained in each participant in this study.
The literature is replete with surrogate markers for the demarcation of critical intensity including the RCP, CP, MLSS, Functional threshold power (FTP), incremental lactate thresholds, HRV and NIRS breakpoints. It is noteworthy that some comparative studies are often highly variable as to precision and concordance between methods. In this context, reliance on markers other than “gold standards” such as MLSS and CP may be problematic in any given individual, with no guarantee of complete accuracy. However, athletes, coaches and patients undergoing rehabilitation are still in need of guidance as to training targets and intensity thresholds. It is with this aim that we present an initial attempt at improving threshold accuracy through combining disparate physiologic paradigms. Thus, it is not the intention to present the use of the [HHb]BP and the HRVT2 as replacements to other models, but rather as simpler, more available and lower cost tools to help practitioners and athletes improve exercise prescription.
Methodological consideration: Although the HR associated to the [HHb]BP and the HRVT2 should be a good representation of the HR associated to the critical intensity of exercise, the cardiac drift that occurs during prolonged constant work rate exercise should be taken into account as it is known that constant exercise at the same HR will result in a decrease in the absolute exercise intensity. Additionally, this study did not consider power output data as an outcome for the comparison at which each threshold occurred. Although in this study the power output values were not statistically different when comparing RCP, [HHb]BP and HRVT2, these data are not presented as the it is well-stablished that the power outputs at which these thresholds occur do not reflect the power output associated to the critical intensity of exercise due to the dissociation between V̇O2 and power output during ramp incremental compared to constant work rate exercise.
In conclusion, this study demonstrated that both the [HHb]BP and the HRVT2 independently provided V̇O2 and HR responses that strongly agreed with those observed at the RCP. More importantly, combining the [HHb]BP and the HRVT2 resulted in estimations of the V̇O2 and HR associated to the critical intensity of exercise with smaller biases and LOA. Thus, the information presented in this study support the use of the [HHb]BP and the HRVT2 as valuable tools for approximating the V̇O2 and HR responses at the critical intensity within practical settings."

Why having two threshold methods available is better than one?

Let's examine the participant list from the perspectives of "accuracy failure" and "technical failure". A failure in accuracy would represent having a high discrepancy between HR at RCP and the surrogate marker in question (for the present example, let's say 10 bpm differential for HRVT2 as in the first row below). A failure in "technique" would be a lack of any result from HRV (due to artifact/noise) or NIRS (from probe issues) such as the blank rows in the listing.

The list above is the raw data table from the article showing RCP, each surrogate method and the Combo for each participant. I added a few columns to the right: 1) the difference from the RCP (gold standard) to the HRVT2, 2) the difference from the RCP (gold standard) to the HHb BP and 3) the difference from the RCP (gold standard) to the Combo (HRVT2/HHb BP average). Orange shaded values represent 10 or more bpm difference than true CI, (which I considered the limit of what I would accept as decent agreement with gold standard).

As an example, the first participant had a -10 for the HRVT2 but a +25 bpm diff of the HHb BP to the RCP of 171 bpm (146 vs 171 bpm). However, the Combo difference was only 7 bpm, a major improvement. This illustrates the issue quite well, we have great group agreement, but this particular unfortunate person would have had totally inappropriate critical intensity markers with single surrogate methods.

"Accuracy failure" - Looking at "outliers" (discrepancy of 10 bpm or higher), there were 4 cases of HRVT2, 5 of HHB BP compared to only 3 for the combo group. More importantly, the 3 people in the combo group represented a much smaller fraction of the total since 2 and 5 participants were excluded from the HRVT2/HHb BP groups respectively. In other words, 4/19 in the HRVT2, 5/16 in the HHb BP and only 3/21 in the combo group had 10 or greater bpm difference from the RCP. Here is a tabular look at "accuracy failure" rate (as defined as 10 or above bpm differential from RCP):

Comparison to the recent Frontiers study:

An important conclusion of a recent Frontiers study using NIRS to estimate the RCP/CI was that although group averages did work out, there was too much individual variation to rely on the single method, because of the poor regression and Bland Altman scatter. Below is a comparison of their data and ours:

The correlation coefficient of the combo was higher than individual methods and was actually quite good. Although our study had a bit higher R2 than the Frontiers had, using HRVT2 definitely added value.
In fact, even standing by itself, HRVT2 had a tighter fit than NIRS (but sample size was too low for direct comparison).

"Technical failure"

Clearly, using "dual methods" gets us fewer "technical failures" since if one is excluded because of poor NIRS signal or high artifact/arrhythmia, the other may suffice. As discussed in the text, 2 participants were unable to have valid HRVT2, and 4 participants unable to have valid NIRS BPs. If they were using single method testing, they would be out of luck for results. As can be seen, even the single methods were generally close to the RCP.

Therefore, we have two solid reasons to take advantage of the HRVT2/NIRS BP Combo

better individual agreement
backup, in cases of single entity technical failure.

Why use Fatmaxxer to record RRs?

There are several options available to record RRs from the Polar H10 (Garmin devices, smartphone apps and Kubios HRV mobile). As discussed before, knowing whether artifacts are due to noise or potentially dangerous cardiac arrhythmias is important. In this study, Fatmaxxer was used, not for it's DFA a1 monitoring ability, but for it's usage of the Polar API to record RRs and ECG snips if artifact is seen. In one of the participants, there were many VPCs seen that triggered an evaluation (luckily, nothing major found):

Thanks to Ian Peake for creating such a useful app.

Although Kubios mobile could have been used, the app was not available at that time. In addition, the licensing rules (per user) would make it difficult for me to use it personally at the same time as the Murias team running it in their lab.

Final remarks:

Accurate identification of the CI is important for proper training intensity distribution and perhaps even race strategy.
Both HRVT2 and HHB BP are viable surrogate methods by themselves, but combining them into a single HR (or VO2) gets one closer to gold standard RCP on an individual basis.
As mentioned in the discussion (and elsewhere), the cycling power seen during ramp thresholds will probably not correspond to the power seen during constant load intervals at the MLSS/MMSS. If you are interested in obtaining the power at the CI, ramps with a shallow slope (5-10 w/min rise) are most appropriate.
I see no reason to limit combining threshold data methods to just the current two. Adding EDR, and even perhaps CP/FTP for those able to do those tests may get even closer to "gold standard". However, doing CP/FTP trials are very taxing, lengthy and depend on participant "effort" which IMO can be problematic (even though they are popular methods). For now, it seems that combining HRVT2 and NIRS BP techniques during an incremental ramp is the best compromise to accuracy and effort.
Cost - Unfortunately, though the cost of HRVT is near zero (if you already have a Polar H10), an NIRS probe is much higher (the Moxy is about 900$). I wish the old BSX unit was still in production as it was small, accurate and less expensive than the competition. I'm still hopeful someone will come along with a reasonably priced NIRS module for consumer usage.
This is the first study where the position of the HRM chest belt was adjusted individually to record an optimum ECG signal (strong positive R peak).
Lastly, even though we are closer to achieving agreement to the gold standard, there are still going to be cases where the difference between surrogate combo and "true" is large enough to make an impact on training targets. As always, common sense should be employed when reviewing results.

Once again, I would like to thank Juan Murias, the co-authors and Pablo Fleitas-Paniagua for making this study happen.