Lactate Threshold and “Lactic Acid” in Endurance and Hybrid Performance
Lactate threshold (LT) is best understood as a systems-level “tipping behavior” in metabolism: as intensity rises, lactate appearance in blood and muscle increasingly outpaces clearance/oxidation, and the body transitions from sustainable “steady-state” work to work that is progressively less sustainable. The practical value of LT for coaches and athletes is that it anchors training intensity in a way that predicts performance and guides programming, especially in events requiring prolonged work at “hard” intensity interspersed with surges-such as HYROX-where athletes must repeatedly produce and clear lactate while maintaining running economy and muscular endurance. [1]
A key scientific correction is that, at physiological pH, “lactic acid” is almost entirely dissociated; the predominant species is lactate (and H⁺ separately). Lactate production via lactate dehydrogenase is not the cause of exercise acidosis; in biochemical terms it consumes protons and helps retard acidosis, while acidosis is more closely tied to reactions such as ATP hydrolysis when ATP turnover outstrips mitochondrial control of proton balance. [2]
For non-lab LT estimation, there is no single best field method. The “right” method depends on which threshold you need (LT1 vs LT2 / maximal steady state), the sport modality (running vs cycling vs mixed), and the athlete’s tolerance for error. Evidence supports several practical approaches-with meaningful limitations-including: 30-min time trial methods (strong validity for running LT pace/HR in trained runners), critical power/critical speed modeling (conceptually aligned with the boundary between sustainable and non-sustainable severe work, but method-sensitive), HRV-derived thresholds (promising agreement on average yet heterogeneous across studies), talk test (useful for threshold 1 boundary), RPE anchors (affordable and surprisingly informative), and smartwatch algorithms (convenient but often pace-biased and device-specific). [3]
Programming implications: (1) build a strong low-intensity base (below LT1) to improve mitochondrial capacity and lactate clearance; (2) dose LT2 / “threshold” work in a way that is sustainable and specific to event demands; and (3) integrate high-intensity work strategically (for VO₂max and speed reserve) while managing cumulative fatigue-especially in concurrent strength + endurance plans. Sex and age differences matter most in absolute outputs and recovery capacity, not in whether thresholds exist or can be trained; relative MLSS intensity can be similar between men and women, and masters athletes can maintain a high fraction of performance when training quality is preserved. [4]
Biochemical distinction and terminology that coaches should use
At the core of the “lactic acid vs lactate” confusion is acid-base chemistry. Lactic acid (HLac) has a pKa ≈ 3.86, so at physiological pH (~7.4) it is overwhelmingly deprotonated; the predominant form in blood and muscle is lactate (Lac⁻), with protons (H⁺) handled by separate buffering and transport systems. In practical coaching language: athletes are not “filled with lactic acid,” they experience rising lactate (a metabolite) and rising acidity (a shift in H⁺ balance) during high-intensity work. [2]
A second critical correction is mechanistic: lactate production does not biochemically “dump acid.” In the widely cited biochemical critique of “exercise-induced lactic acidosis,” lactate production is described as retarding, not causing, acidosis, because lactate formation consumes protons and supports glycolytic flux by regenerating NAD⁺ (which helps sustain ATP regeneration when intensity rises). Lactate is therefore a useful marker of the metabolic conditions that coincide with acidosis, but it is not the direct source of the protons that lower pH. [5]
For coaches writing about LT, the most defensible terminology is to separate:
- Lactate: a metabolite continuously produced and used (fuel, precursor, signaling). [6]
- Metabolic acidosis: a shift in acid-base status associated with high ATP turnover and limited “steady” proton handling; lactate tracks the intensity domain but is not the root cause. [7]
- Threshold concepts: multiple operational “thresholds” exist (LT, LT1/LT2, ventilatory thresholds, OBLA, MLSS, CP), and they are not interchangeable unless the method and definition are explicit. [8]
Finally, emphasize that lactate is not a “waste product.” Modern lactate shuttle theory frames lactate as a major oxidative substrate, a gluconeogenic precursor, and a signaling molecule; during sustained exercise, a large fraction of lactate disposal is via oxidation (with a smaller fraction via gluconeogenesis). [9]
Physiology of lactate production, clearance, and accumulation
Lactate dynamics are best taught as a balance between appearance (production + release) and disposal (oxidation + uptake + conversion). As intensity increases, glycolytic flux rises and lactate formation increases; this occurs even when oxygen is present-so “lactate = no oxygen” is inaccurate. Lactate rises when appearance outpaces disposal, not simply because intensity is “anaerobic.” [9]
A key applied point: blood lactate is not muscle lactate, and it is not a direct measure of “fatigue chemicals.” Blood lactate reflects whole-body production/uptake, transport, sampling site, and protocol design. Nutrition status, prior training load, heat stress, stage duration, and test modality can all change the lactate curve for the same athlete at the same “fitness.” [10]
Training alters lactate kinetics through multiple mechanisms, including increased mitochondrial oxidative capacity, improved oxygen delivery, and changes in lactate transport capacity (monocarboxylate transporters). Exercise training has been shown to upregulate lactate/H⁺ transport capacity and MCT proteins in muscle in some contexts, though responses vary by protocol and measurement methods. [11]
Misconceptions to address explicitly in a coach/athlete review:
“Lactic acid buildup causes soreness the next day.” Delayed onset muscle soreness (DOMS) is not explained by lactate persisting in muscle; DOMS is more consistent with exercise-induced muscle damage and subsequent processes rather than “lactate staying in the legs.” [12]
“Flush the lactic acid.” Lactate is cleared relatively quickly post-exercise, and active recovery typically accelerates lactate clearance compared with passive rest, with some evidence that intensities around a fraction of LT are efficient for clearance dynamics. [13]
Conceptual map of lactate handling
This simplified shuttle map is consistent with contemporary lactate shuttle framing: continuous production and utilization, substantial oxidation-based disposal during exercise, and lactate’s role as a major gluconeogenic precursor. [14]
What “lactate threshold” means and why definitions matter
Most coaching confusion comes from treating LT as a single point. In practice, there are at least two useful “threshold boundaries,” often aligned with intensity domains:
- First threshold (LT1 / VT1 / “aerobic threshold”): marks the transition from easy/moderate work to “steady but noticeably harder” work; ventilation changes and lactate begins to rise above baseline. This boundary is especially relevant for durability, high-volume training, recovery runs/rides, and long-event pacing. [15]
- Second threshold (LT2 / VT2 / MLSS / CP-adjacent concept): approximates the upper boundary of heavy-intensity exercise and relates to the highest sustainable oxidative steady state; above it, physiology drifts and time-to-exhaustion becomes much shorter. This boundary is central for threshold intervals, tempo work, and performance in events around ~20-60+ minutes depending on the definition and athlete. [16]
Critically, operational definitions differ. In the “maximal metabolic steady state” debate, MLSS is often determined by repeated 30-min constant-load trials and a criterion such as <1 mmol/L rise in lactate within a specified time window, whereas critical power/speed is derived from the hyperbolic power-duration or speed-duration relationship. Methodological differences can yield different “threshold” numbers for the same athlete, and the agreement depends heavily on protocol rigor and chosen criteria. [17]
For a review article aimed at coaches: it is defensible to treat LT as two practical anchors (LT1 and LT2) and then state explicitly which non-lab methods best approximate each anchor in the real world. [18]
Non-lab methods to estimate LT with evidence, protocols, and tradeoffs
The table below emphasizes four realities for applied coaching:
1) different methods estimate different thresholds (LT1 vs LT2),
2) cross-device and cross-modality transfer (running ↔ cycling) is imperfect,
3) accuracy depends on protocol discipline, environment, and athlete familiarity,
4) “good enough for training” may be different from “valid vs lab standard.”
Comparison of practical non-lab LT estimation approaches
30-min running time trial (pace + HR)
Best aligns with: LT2 proxy
Protocol summary: Warm up, then run ~30 minutes at the fastest sustainable pace. Use the average pace and heart rate (often the last 20 minutes) as the estimate. Repeat under similar conditions for consistency.
Equipment: Track or treadmill, heart-rate monitor
Typical time: ~45–60 minutes including warm-up and cooldown
Accuracy / reliability: In trained runners, estimated LT velocity with low standard error (~0.21 m/s) and HR error ~8 bpm versus lab LT4.0 criteria. Requires good pacing ability.
Key validation sources: McGehee et al., 2005
VDOT / race-performance model
Best aligns with: LT2 proxy (pace-based)
Protocol summary: Use recent race performances across different distances to calculate threshold training pace using a VDOT calculator.
Equipment: Race results and calculator
Typical time: A few minutes
Accuracy / reliability: Larger standard error (~0.41 m/s) compared with 30-minute time trials. Accuracy depends heavily on accurate race results under similar conditions.
Key validation sources: McGehee et al., 2005
3,200 m / 2-mile time-trial model
Best aligns with: LT2-like proxy (pace-based)
Protocol summary: Run a hard 3,200 m (2-mile) time trial and use regression models to estimate LT pace.
Equipment: Track and timer
Typical time: ~20–30 minutes
Accuracy / reliability: Evidence is mixed. Some studies show it tends to overestimate LT velocity compared with direct measurements.
Key validation sources: Weltman 1987; McGehee 2005
Conconi heart-rate deflection test
Best aligns with: Intended as LT2 proxy, but unreliable
Protocol summary: Perform an incremental running or cycling test with increasing intensity and identify the point where the heart-rate curve “deflects.”
Equipment: Track, treadmill, or bike with HR monitor
Typical time: ~30–60 minutes
Accuracy / reliability: Reliability issues are common. Many athletes do not show a clear deflection point, and protocol design can create artificial deflections. Often overestimates LT.
Key validation sources: Jones & Doust 1995; Vachon et al. 1999
Talk Test
Best aligns with: LT1 / VT1 proxy (aerobic threshold)
Protocol summary: During progressively harder exercise, the athlete recites a standard phrase or passage.
• Comfortable speech = below threshold
• Speech becoming difficult = around threshold
• Speech impossible = above threshold
Equipment: None required (HR monitor optional)
Typical time: ~20–40 minutes
Accuracy / reliability: Evidence supports talk-test stages approximating ventilatory threshold in both treadmill and cycling tests.
Key validation sources: Talk-test validation studies
RPE anchors (Borg scale)
Best aligns with: Practical LT1-LT2 training zones
Protocol summary: Use rating of perceived exertion during steady workouts to anchor intensity zones.
Equipment: None required (HR optional)
Typical time: Ongoing during training
Accuracy / reliability: Typical anchors in runners correspond roughly to CR-10 values around 4.3 (≈2 mmol/L lactate) and 6.5 (≈4 mmol/L), but individual variability is large.
Key validation sources: Dantas et al., 2015
HRV threshold using DFA-α1
Best aligns with: LT1 and sometimes LT2 proxies
Protocol summary: Perform an incremental test while recording heart-rate variability. Breakpoints in DFA-α1 values (often around 0.75) correspond to threshold transitions.
Equipment: Chest-strap HR monitor and HRV analysis software/app
Typical time: ~20–40 minutes
Accuracy / reliability: Meta-analyses show strong correlations with ventilatory or lactate thresholds, but results vary depending on devices and protocols.
Key validation sources: HRV threshold meta-analyses and field validation studies
Critical Power / Critical Speed
Best aligns with: Heavy-to-severe intensity boundary (LT2-adjacent concept)
Protocol summary: Perform several maximal trials over different durations and fit a power-duration or speed-duration model to determine critical power or speed.
Equipment: Power meter or accurate run timing
Typical time: Requires a block of testing sessions
Accuracy / reliability: Correlates well with several LT2 markers but limits of agreement can be wide. Results depend strongly on testing protocol.
Key validation sources: CP-MLSS comparison studies
FTP (20-minute test × 0.95)
Best aligns with: Practical LT2 proxy in cycling
Protocol summary: Perform a maximal 20-minute cycling effort and multiply average power by 0.95 to estimate FTP.
Equipment: Power meter or smart trainer
Typical time: ~45–60 minutes including warm-up
Accuracy / reliability: Evidence is mixed. FTP may correlate with MLSS in trained cyclists but can differ substantially for some athletes.
Key validation sources: Borszcz 2019; Inglis et al.
Wearable “LT estimate” algorithms
Best aligns with: Convenience-based estimate
Protocol summary: Smartwatch algorithms estimate LT from training data or guided tests.
Equipment: Smartwatch (often with chest strap HR monitor)
Typical time: Minimal if automated
Accuracy / reliability: HR estimates are often within ~9–11 bpm of lab thresholds, but pace estimates can be biased and device-specific.
Key validation sources: Wearable validation studies
A practical “method selection” heuristic for coaches
If you need LT1 (easy/steady boundary), prioritize Talk Test + RPE anchors (and optionally HRV/DFA-α1 if you have high-quality HRV data and consistent protocols). [38]
If you need LT2 / maximal steady-state proxy, prioritize 30-min TT (running), CP/CS modeling, or carefully interpreted FTP/TT metrics depending on modality-then cross-check with RPE and heart-rate stability. [39]
If you need convenience at scale, wearables can be adequate for broad HR zoning but are less trustworthy for pace at LT without validation against on-the-ground performance tests. [40]
Training implications and programming across athlete type, age, sex, and level
What improves lactate threshold in training terms
From a coaching perspective, “raising LT” generally means raising the speed/power you can sustain before lactate drifts upward and perceived exertion escalates quickly. That adaptation reflects increased oxidative capacity, better lactate transport/oxidation, and improved tolerance for high ATP turnover, while still maintaining mechanical efficiency. [41]
A widely used applied model anchors training distribution around LT1 and LT2 boundaries, defining “low intensity” as lactate stable below ~2 mmol, “threshold” as roughly between ~2-4 mmol, and “high intensity” above MLSS/≥4 mmol for many trained athletes-while acknowledging major individual variation. [42]
Training intensity distribution chart
The values below illustrate three-zone distributions drawn from published endurance training studies and observational work; they are not universal prescriptions, but they clarify the shape of different models. Polarized and pyramidal models commonly emphasize a large Zone 1 base, while threshold-heavy models shift more time into the moderate zone. [43]
These example ratios correspond to published distributions in trained cyclists (polarized vs threshold) and trained runners (pyramidal-like distributions in a 3-zone model), illustrating the conceptual difference between models rather than declaring a single optimal approach. [44]
Prescriptions by athlete type
The table below is written as actionable programming guidance with conservative, coach-friendly assumptions. When thresholds are estimated non-invasively, treat them as ranges (bands) rather than single “magic numbers,” and cross-check with symptoms, drift, and performance. [45]
New or returning athlete
Primary goals: Establish LT1 discipline and build aerobic base
Weekly structure: Mostly Zone 1, minimal Zone 3, cautious Zone 2
LT-focused training: 0–1 short tempo exposure per week
Monitoring: Talk test, RPE anchors, HR drift during steady workouts
Risk considerations: High risk of overuse or fatigue if moderate-hard days accumulate
Recreational endurance athlete
Primary goals: Increase LT2 pace or power and improve repeatability
Weekly structure: Pyramidal (large Zone 1 base, moderate Zone 2, small Zone 3)
LT-focused training: 1–2 threshold sessions per week totaling ~20–40 minutes near LT2
Monitoring: Periodic 30-minute time trials or CP/CS tests, RPE stability
Risk considerations: Environmental conditions can affect pacing, so RPE anchors help regulate effort
Competitive endurance athlete
Primary goals: Maximize LT2 while maintaining VO₂max and speed reserve
Weekly structure: Often polarized or mixed pyramidal/polarized
LT-focused training: 1–2 LT2 sessions per week plus occasional high-intensity work
Monitoring: Performance at fixed HR/RPE, fatigue trends, and repeated test blocks
Risk considerations: Training distribution should be individualized based on response and event demands
Masters athlete
Primary goals: Maintain LT speed/power while supporting recovery capacity
Weekly structure: Greater proportion of Zone 1 work
LT-focused training: ~1 LT2 session per week (often broken into intervals)
Monitoring: HR and RPE rather than percentages of HRmax
Risk considerations: Prioritize recovery, monitor sleep, stiffness, and soreness
Female endurance athlete
Primary goals: Develop thresholds while managing fueling and cycle-related variability
Weekly structure: Similar distribution to male athletes but adjusted for recovery
LT-focused training: Standard LT sessions with appropriate fueling
Monitoring: Perceived exertion and symptoms, with consistent testing conditions when possible
Risk considerations: Relative MLSS intensity can be similar between sexes, but substrate utilization differs
Hybrid athlete with strength focus
Primary goals: Maintain aerobic base while improving LT2 endurance without harming strength
Weekly structure: Base-heavy with 1–2 targeted quality days
LT-focused training: One LT2 session plus one hybrid-specific threshold circuit
Monitoring: Session RPE, strength performance, and running economy
Risk considerations: Concurrent endurance and strength training may interfere with hypertrophy or maximal strength if poorly sequenced
Non-lab monitoring and recovery strategies that directly matter for LT work
Active recovery tends to clear blood lactate faster than passive recovery, and some data suggest that intensities in a band around a fraction of LT are effective for clearance kinetics-useful in competitions with repeated bouts and in training when you want to maintain session quality. [13]
Programming-level fatigue management matters as much as physiology. In concurrent endurance + strength plans, a large evidence base indicates that interference effects depend on endurance modality, training frequency/volume, and recovery structure; separating hard endurance and heavy strength, and prioritizing the most event-relevant quality, is often more important than chasing a precise lactate number. [51]
HYROX-specific demands and how LT-focused training applies
Physiological and station-level demands
A simulated HYROX “Individual Open” format consists of 8 × 1 km runs alternating with 8 standardized stations (SkiErg, sled push, sled pull, burpee broad jumps, row, farmers carry, sandbag lunges, wall balls), with prescribed loads/distances by sex/division. In one detailed physiological study, the median completion time was ~86.5 minutes, with the majority of the event performed at hard to very hard heart-rate intensities and peak blood lactate around ~8.5 mmol/L, highest at the final wall-ball station. Faster completion correlated with higher VO₂max, greater endurance training volume, and lower body fat percentage. [52]
The same study provides station prescriptions (Individual Open) that matter for metabolic specificity: sled push and pull distances and loads; carries and lunges; and wall-ball volume and load. These stations create repeated spikes in glycolytic demand layered onto sustained running, which makes “threshold management” (production + clearance + pacing) a central performance constraint even when maximal strength is not the dominant limiter. [53]
How LT training transfers to HYROX performance
HYROX appears to punish two common errors:
1) Run too hard early (above LT2) → lactate and ventilation drift upward; stations become slower and more “expensive,” and the athlete loses the ability to “reset” between efforts. This aligns with the general physiology of heavy vs severe domains: once the athlete is repeatedly above the maximal sustainable steady-state boundary, drift accumulates. [54]
2) Insufficient LT1 base → the athlete cannot clear metabolites effectively between stations at a pace that still moves the race forward; the “easy” pace is not easy enough. [55]
A HYROX-ready LT profile is therefore not only “high LT2,” but also a wide base (high LT1 speed) so that the athlete can recover while still running. [56]
Coaching techniques and session examples for HYROX
Below are session templates that target the lactate dynamics obvious in HYROX: produce lactate at stations, then clear it while running and transitioning.
Threshold intervals with station “lactate spikes” (LT2 + clearance)
Warm up well. Then repeat 3-5 times:
- Run 6-10 min at LT2 band (controlled hard; RPE stable)
- Immediately perform 60-120 s of a HYROX station element at moderate-hard effort (e.g., SkiErg, row, wall balls, lunges)
- Recover 2-3 min easy (enough to reduce ventilation, not necessarily to “fully recover”)
The goal is pace control: the run segments should not fall apart across sets; if they do, the athlete is above their sustainable threshold band. This structure is consistent with the idea that small changes above maximal steady state can sharply reduce tolerance, so “precision” matters. [57]
Broken tempo for adults with limited tolerance (LT2 accumulation without collapse)
- 3 × (10 min “comfortably hard” + 3 min easy)
Use talk test/RPE to keep the athlete out of the red. This is a practical way to accumulate LT time without inducing excessive fatigue. [58]
Run-carry threshold progression (LT1→LT2 bridge + HYROX specificity)
- 30-50 min continuous in Zone 1-low Zone 2, inserting every 6-8 minutes: 30-60 s farmers carry or sandbag carry at controlled form
This targets economy under load and teaches the athlete to keep lactate and breathing controlled while the upper body is taxed. The rationale matches the observed HYROX pattern: station intensity spikes occur repeatedly, but overall success correlates strongly with endurance capacity and volume. [59]
Taper-week LT maintenance (keep the engine, drop fatigue)
In the final 7-10 days: keep one short LT2 session (e.g., 2 × 8-10 min) and one brief high-intensity touch, while reducing total volume. This aligns with endurance periodization logic: maintain intensity anchors while shedding fatigue load. [60]
Periodization flow for a HYROX build
This structure prioritizes the two demands supported by current HYROX physiological profiling: sustained high cardiovascular load plus repeated station-driven lactate peaks. [61]
[1] [8] Lactate threshold concepts: how valid are they?
https://pubmed.ncbi.nlm.nih.gov/19453206/?utm_source=chatgpt.com
[2] https://pmc.ncbi.nlm.nih.gov/articles/PMC5037729/
https://pmc.ncbi.nlm.nih.gov/articles/PMC5037729/
[3] [19] [20] [22] [39] [47] https://www.researchgate.net/profile/Charles-Tanner/publication/7665807_A_Comparison_of_Methods_for_Estimating_the_Lactate_Threshold/links/5a3d20b40f7e9ba8688e6b7e/A-Comparison-of-Methods-for-Estimating-the-Lactate-Threshold.pdf
[4] [50] https://pmc.ncbi.nlm.nih.gov/articles/PMC6316329/
https://pmc.ncbi.nlm.nih.gov/articles/PMC6316329/
[5] [7] https://pubmed.ncbi.nlm.nih.gov/15308499/
https://pubmed.ncbi.nlm.nih.gov/15308499/
[6] [9] [14] [41] [55] https://www.sciencedirect.com/science/article/pii/S1550413118301864
https://www.sciencedirect.com/science/article/pii/S1550413118301864
[10] https://pmc.ncbi.nlm.nih.gov/articles/PMC12619971/
https://pmc.ncbi.nlm.nih.gov/articles/PMC12619971/
[11] https://journals.physiology.org/doi/full/10.1152/ajpendo.1999.276.2.E255
https://journals.physiology.org/doi/full/10.1152/ajpendo.1999.276.2.E255
[12] https://pmc.ncbi.nlm.nih.gov/articles/PMC9505902/
https://pmc.ncbi.nlm.nih.gov/articles/PMC9505902/
[13] https://pubmed.ncbi.nlm.nih.gov/24739289/
https://pubmed.ncbi.nlm.nih.gov/24739289/
[15] [18] [42] [48] [56] [60] https://umh1617.umh.es/files/2016/05/02-IJSPP_2010_Seiler_0009.pdf
https://umh1617.umh.es/files/2016/05/02-IJSPP_2010_Seiler_0009.pdf
[16] [17] [32] [54] https://kar.kent.ac.uk/id/document/3182611
https://kar.kent.ac.uk/id/document/3182611
[21] https://pubmed.ncbi.nlm.nih.gov/3429086/
https://pubmed.ncbi.nlm.nih.gov/3429086/
[23] https://www.thieme-connect.com/products/ejournals/pdf/10.1055/s-2007-973051.pdf
https://www.thieme-connect.com/products/ejournals/pdf/10.1055/s-2007-973051.pdf
[24] https://pubmed.ncbi.nlm.nih.gov/10409607/
https://pubmed.ncbi.nlm.nih.gov/10409607/
[25] [38] [46] [58] paulogentil.com
https://paulogentil.com/pdf/Consistency%20of%20the%20talk%20test%20for%20exercise%20prescription.pdf
[26] https://pubmed.ncbi.nlm.nih.gov/25187249/
https://pubmed.ncbi.nlm.nih.gov/25187249/
[27] https://pubmed.ncbi.nlm.nih.gov/22615009/
https://pubmed.ncbi.nlm.nih.gov/22615009/
[28] [45] https://pmc.ncbi.nlm.nih.gov/articles/PMC11461412/
https://pmc.ncbi.nlm.nih.gov/articles/PMC11461412/
[29] Borg Rating Of Perceived Exertion
https://www.physio-pedia.com/Borg_Rating_Of_Perceived_Exertion?utm_source=chatgpt.com
[30] https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2021.676484/full
https://www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2021.676484/full
[31] https://pmc.ncbi.nlm.nih.gov/articles/PMC3737850/
https://pmc.ncbi.nlm.nih.gov/articles/PMC3737850/
[33] https://pubmed.ncbi.nlm.nih.gov/30676826/
https://pubmed.ncbi.nlm.nih.gov/30676826/
[34] [57] https://kar.kent.ac.uk/id/document/3194378
https://kar.kent.ac.uk/id/document/3194378
[35] https://kar.kent.ac.uk/100046/1/Final%20Accepted%20FTP%20JSS.pdf
https://kar.kent.ac.uk/100046/1/Final%20Accepted%20FTP%20JSS.pdf
[36] [40] https://pmc.ncbi.nlm.nih.gov/articles/PMC12309276/
https://pmc.ncbi.nlm.nih.gov/articles/PMC12309276/
[37] https://pmc.ncbi.nlm.nih.gov/articles/PMC11090120/
https://pmc.ncbi.nlm.nih.gov/articles/PMC11090120/
[43] https://pmc.ncbi.nlm.nih.gov/articles/PMC5187238/
https://pmc.ncbi.nlm.nih.gov/articles/PMC5187238/
[44] https://www.researchgate.net/publication/233977574_Six_weeks_of_a_polarized_training-intensity_distribution_leads_to_greater_physiological_and_performance_adaptations_than_a_threshold_model_in_trained_cyclists
[49] https://journals.physiology.org/doi/10.1152/japplphysiol.00298.2003
https://journals.physiology.org/doi/10.1152/japplphysiol.00298.2003
[51] https://pubmed.ncbi.nlm.nih.gov/22002517/
https://pubmed.ncbi.nlm.nih.gov/22002517/
[52] [53] [59] [61] https://pmc.ncbi.nlm.nih.gov/articles/PMC11994925/