Which lactate-threshold method is best?

There is no single best method. Modified Dmax has the strongest validity data (ICC = 0.96), but the best choice depends on data quality, sport and curve shape. LactateThreshold therefore compares all methods side by side.

What is the difference between LT1 and LT2?

LT1 marks the first measurable rise above resting lactate and guides easy aerobic endurance work. LT2 is closer to the highest sustainable controlled intensity and is central for tempo and threshold training.

Is the fixed 4 mmol/L value reliable?

The fixed 4 mmol/L value (Mader 1976) is a historical standard, but it can be far off for many athletes. Individualized methods such as Modified Dmax or Dickhuth IAT adapt to the individual curve shape and are usually more valid.

What is Modified Dmax and why is it recommended?

Modified Dmax (Jamnick 2018) identifies the maximum perpendicular distance from the lactate curve to the line between start and end point, beginning at the first rise above 0.4 mmol/L. It shows very high reproducibility (ICC = 0.96) and is therefore used as the recommended standard.

Scientific Documentation

Calculation Methods & Formulas

Name: LactateThreshold - lactate test analysis
Availability: InStock
Author: LactateThreshold

Complete documentation of all threshold calculations used, with exact formulas, scientific sources, and validation data.

Standard ⭐Recommended

★ New in V4Added in V4 for stronger scientific validation

LegacyLimited

ExperimentalUse with caution

Threshold Calculations (LT1/LT2)

Modified Dmax (Scientific)

Standard ⭐

Gold standard for threshold determination. Finds the point with the maximum perpendicular distance to the connecting line, starting before the first significant lactate rise.

Formula

1. Find start point S:
   If La[i] - La[i-1] ≥ 0.4 mmol/L → S = (P[i-1], La[i-1])

2. Calculate distance for each point P(x,y):
   d(x,y) = |A·x + B·y + C| / √(A² + B²)
   
   where: A = -(La_end - La_start) / (P_end - P_start)
          B = 1
          C = A·P_start - La_start

3. LT2 = Point with max(d)

✓ ICC = 0.96 with MLSS (Gold Standard)

Jamnick et al. (2018)

Dmax (Original)

Legacy

Original Dmax method by Cheng (1992). Uses the first data point as the start point, which can lead to underestimation.

Formula

Identical to Modified Dmax, but:
Start point S = (P[0], La[0])  ← First data point

⚠️ With elevated stress lactate at the start, this can
   underestimate the threshold by 10-20W.

Can underestimate MLSS (203 ± 53 W vs. MLSS 225 ± 60 W)

✓ r > 0.90 with MLSS, but larger deviation than Modified Dmax

Cheng et al. (1992)

Dickhuth (IAT)

Individual Anaerobic Threshold based on the lactate equivalent. The minimum of the equivalent indicates maximum metabolic efficiency.

Formula

1. Calculate lactate equivalent:
   LE(i) = La(i) / P(i)

2. Find minimum:
   i_min = argmin[LE(i)]
   La_min = La(i_min)

3. Thresholds:
   LT1 = Power at (La_min + 0.5 mmol/L)
   LT2 = Power at (La_min + 1.5 mmol/L)

✓ r = 0.97 Reproducibility

Dickhuth et al. (1999)

Delta (+1.5 mmol/L)

Simple and practical method. Adds a fixed offset to the baseline lactate value.

Formula

1. Determine baseline:
   La_baseline = min(La[0], La[1], La[2])

2. Thresholds:
   LT1 = Power at (La_baseline + 0.4 mmol/L)
   LT2 = Power at (La_baseline + 1.5 mmol/L)

3. Interpolate power:
   LT2_P = P[j] + (LT2_La - La[j])/(La[j+1] - La[j]) × (P[j+1] - P[j])

✓ Low mean deviation (6.7 ± 17.2 W)

Płoszczyca et al. (2020)

Mader (4.0 mmol/L)

Legacy

Fixed threshold at 4.0 mmol/L (OBLA). Historically widely used, but not individual.

Formula

LT2_Lactate = 4.0 mmol/L  (fixed value)

LT2_P = P[j] + (4.0 - La[j])/(La[j+1] - La[j]) × (P[j+1] - P[j])

⚠️ The individual MLSS varies between 2.0-7.0 mmol/L!

This method is NOT individual and systematically overestimates the anaerobic threshold in many athletes.

✓ Overestimates MLSS by 43.3 ± 17.8 W

Mader et al. (1976)

Log-Log

Experimental

Dmax calculation in logarithmic space. Originally developed for ventilatory thresholds.

Formula

1. Log transformation:
   x' = log₁₀(P)
   y' = log₁₀(La)

2. Dmax in Log space:
   d'(x',y') = |A'·x' + B'·y' + C'| / √(A'² + B'²)

3. Back transformation:
   LT2_P = 10^(x'_max)
   LT2_La = 10^(y'_max)

✓ Limited concordance with other methods

Beaver et al. (1985)

Interactive method comparison

LAB TIER

Load a saved test and see all 6 threshold methods side by side on the same lactate curve. This visualizes how strongly methods can diverge on your data.

★ Unlock Lab

Uncertainty & Validation

★ V4

The core innovation: honest measurement uncertainty instead of false precision. Every threshold receives a CI, every curve gets coverage, and every progression comparison is checked against measurement tolerance.

Bootstrap confidence intervals

★ New in V4

Bootstrap procedure for estimating LT1/LT2 measurement uncertainty. Two variants are used: non-parametric resampling with replacement for n ≥ 6, and parametric heteroscedastic Gaussian noise σ(y) = 0.10 + 0.03·y for n = 4-5. Runs 500 deterministic seeded iterations for stable UI output.

Formula

σ(y) = σ_base + σ_cv · y
σ_base = 0.10 mmol/L (device floor)
σ_cv = 0.03 (3% CV)
CI₆₈ = [P₁₆, P₈₄] → ±Half-width

✓ Returns ±watt/pace uncertainty for every threshold.

Efron (1979) + Pyne et al. (2001), J Sci Med Sport

Goodness of fit (R²)

★ New in V4

Polynomial fit (cubic for n ≥ 5, quadratic for n = 4) used to quantify lactate-curve noise. Low R² can indicate noisy measurements, stress lactate or a non-monotone curve.

Formula

R² = 1 − (Σ(y_i − ŷ_i)²) / (Σ(y_i − ȳ)²)

Categories:
≥ 95% → Excellent
≥ 90% → Good
≥ 80% → Acceptable
< 80% → Noisy

✓ Noise metric that complements the coverage score.

Standard statistics

Coverage score

★ New in V4

Checks whether the test actually reached LT2. R² only measures noise: a 4-stage test at 40-60% of LT2 can have R² ≈ 1.0 and still miss LT2 completely. Coverage answers whether the curve truly brackets LT2.

Formula

Complete ⟺ all 3 criteria are met:
1. Peak lactate ≥ 4.0 mmol/L
2. ≥ 2 stages > 2.5 mmol/L
3. Δ(last 2 stages) ≥ 0.8 mmol/L

✓ Detects incomplete tests (incomplete/borderline/complete).

Internal criterion (V4)

Progression significance (error propagation)

★ New in V4

Compares two tests with error propagation: is the improvement significant relative to measurement uncertainty? This prevents athletes from interpreting noise as progress. Three levels are used: 95%, 68% and not significant.

Formula

σ_combined = √(CI_a² + CI_b²)

|Δ| > 1.96·σ → sig. 95%
|Δ| > 1.00·σ → sig. 68%
otherwise   → n.s.

✓ Prevents false progress signals caused by measurement uncertainty.

Gaussian error propagation

LT1-anchored training zones

★ New in V4

The upper boundary of Z2 is the measured LT1, not a fixed Coggan percentage. According to San Millán & Brooks 2018, LT1 is the transition from fat-dominant to mixed substrate use, making it the correct Z2 boundary. The middle of Z3 is set between LT1 and LT2. No heuristic is needed.

Formula

lt1Frac = lt1 / lt2  (bike)
lt1Frac = lt2 / lt1  (run, speed ratio)

Z1_upper = 0.85 · lt1Frac
Z2_upper = lt1Frac           ← LT1
Z3_upper = (lt1Frac + 1) / 2
Z4_upper = 1.05              ← +5% LT2

✓ Derived empirically from measurements, not from a calibration coefficient.

San Millán & Brooks (2018), Int J Sports Physiol Perform

MLSS calibration & method ranking

★ New in V4

Gold-standard validation: the user uploads MLSS constant-test values, the app compares all 7 methods against this anchor, and ranks them by absolute error. Across multiple comparison pairs this creates a per-athlete method recommendation. Lab tier.

Formula

For each method M:
  error_M = |LT2_M − MLSS|

Ranking: ascending by mean(error_M)
across all (stepTest, mlssTest) pairs.

✓ Empirical per-athlete ranking, unique among freemium tools.

Beneke (2003), Eur J Appl Physiol + internal ranking

VLamax Calculation

VLamax Sprint Test (Gold Standard)

Standard ⭐

Precise determination of the maximal glycolytic rate through an all-out sprint.

Formula

VLamax = ΔLa / t_eff

where:
   ΔLa = La_peak - La_rest  [mmol/L]
   t_eff = t_sprint - 5.5s  [s]
   5.5s = Alactic time constant (PCr depletion)

Example:
   Sprint: 15s, Rest-La: 1.2, Peak-La: 12.8
   t_eff = 15 - 5.5 = 9.5s
   VLamax = (12.8 - 1.2) / 9.5 = 1.22 mmol/L/s

✓ Gold standard for VLamax determination

Mader & Heck (1986)

VLamax Step Test Estimation

Experimental

Approximation from step test data. Accuracy: ±15-25%.

Formula

VLamax ≈ 0.3 + (ΔLa/ΔP × 50) × 1.2

where:
   ΔLa = La[end] - La[mid]     [mmol/L]
   ΔP  = P[end] - P[mid]       [W]
   50  = Normalization factor (per 50W)

Range: 0.2 - 1.5 mmol/L/s

This is an estimate! For precise values use the Sprint Test.

✓ Estimation with ±15-25% accuracy

Approximation

Mathematical Foundations

Linear Interpolation

Simple interpolation between two known points.

Formula

y = y₁ + (x - x₁)/(x₂ - x₁) × (y₂ - y₁)

For threshold calculation (Y → X):
x = x₁ + (y - y₁)/(y₂ - y₁) × (x₂ - x₁)

Standard Mathematics

Fritsch-Carlson Monotone Spline

★ New in V4

Monotone cubic Hermite interpolation. It guarantees that no new extrema appear between data points, unlike Catmull-Rom splines, which can oscillate on steep or non-monotone curves. This matters for Dmax geometry because overshoot can create false inflection points.

Formula

1. δ_k = (y_{k+1} − y_k) / (x_{k+1} − x_k)
2. m_k = (δ_{k-1} + δ_k) / 2
3. if δ_{k-1}·δ_k ≤ 0 → m_k = 0
4. if α² + β² > 9 → scale m_k, m_{k+1}
5. Hermite basis: h₀₀, h₁₀, h₀₁, h₁₁

✓ Guaranteed monotone on every segment, with no new extrema.

Fritsch & Carlson (1980), SIAM J Numer Anal

Point-to-Line Distance (Dmax)

Core formula for all Dmax-based methods.

Formula

Perpendicular distance of a point (x,y) to the line:

          |(y₂-y₁)·x - (x₂-x₁)·y + x₂·y₁ - y₂·x₁|
d(x,y) = ─────────────────────────────────────────
                  √[(y₂-y₁)² + (x₂-x₁)²]

where (x₁,y₁) = Start point, (x₂,y₂) = End point

Analytical Geometry

References

1. Jamnick NA et al. (2018). Manipulating graded exercise test variables affects the validity of the lactate threshold.PLoS ONE

2. San Millán I & Brooks GA (2018). Assessment of Metabolic Flexibility by Means of Measuring Blood Lactate, Fat, and Carbohydrate Oxidation Responses to Exercise in Professional Endurance Athletes and Less-Fit Individuals.Sports Med 48:467–479

3. Allen H & Coggan A (2019). Training and Racing with a Power Meter (3rd ed.). VeloPress.

4. Jeukendrup A (2014). A step towards personalized sports nutrition: carbohydrate intake during exercise.Sports Med 44(Suppl 1):S25–33

5. Fritsch FN & Carlson RE (1980). Monotone Piecewise Cubic Interpolation.SIAM J Numer Anal 17(2):238–246

6. Pyne DB et al. (2001). Relationships between repeated sprint testing, speed, and endurance.J Sci Med Sport 4(1):17–27

7. Bishop D et al. (1998). The validity of the modified Dmax method for determining the lactate threshold.Med Sci Sports Exerc 30(2):272–278

8. Dickhuth HH et al. (1999). Ventilatory, lactate-derived and catecholamine thresholds.Int J Sports Med

9. Mader A, Heck H (1986). A theory of the metabolic origin of anaerobic threshold.Int J Sports Med

10. Beneke R (2003). Methodological aspects of maximal lactate steady state — implications for performance testing.Eur J Appl Physiol 89(1):95–99