REF-130 · Vaccari et al., 2020 · European Journal of Applied Physiology, · Paper · PubMed

HIDIT — High-Intensity Decreasing Interval Training increases time above 90% VO2peak; also: work-above-CP is NOT a valid HIT-effectiveness metric

Blimited evidenceMeasured on trained athletesWritten before October 2026, being re-checked

Participants

n=12 cyclists (abstract: "twelve cyclists"). Middle-aged amateur level, VO2peak 4.32 L/min (about 57 ml/kg/min per the discussion), Critical Power 212-320 W, age 29-62 (mean 41) (full text, checked 2026-09-27).

Protocol — three formats

Each protocol had the same work power, the same recovery power and the same work:recovery ratio (abstract). Work at ~117% CP and recovery at ~83% CP, ratio 3:2 (full text, checked 2026-09-27).

1

LIHIIT (long): 3 min work / 2 min recovery

2

SIHIIT (short): 30 s work / 20 s recovery

3

HIDIT (decreasing): work from 3 min down to 30 s, recovery from 2 min down to 20 s (abstract); the steps 3→2→1 min→45 s→30 s work with 2 min→80 s→40 s→30 s→20 s recovery (full text, checked 2026-09-27)

All protocols run to exhaustion (Tlim); time >90% VO2peak, Tlim, blood lactate and RPE measured.

Key Findings — full table (added 2026-06-15 per KIW Note #166)

ProtocolTime >90% VO2peak (s) — abstractTime to failure (s) — abstractMean %VO2max (%) — full text, checked 2026-09-27Work above CP (kJ) — full text, checked 2026-09-27
HIDIT (decreasing)312 ± 207798 ± 18584.2 ± 4.022.0 ± 10.4 (full text)
SIHIIT (short 30/20)182 ± 225 (P = 0.036 vs HIDIT)714 ± 26581.5 ± 6.618.7 ± 9.0 (full text)
LIHIIT (long 3/2)179 ± 145 (P = 0.027 vs HIDIT)664 ± 28279.6 ± 7.119.3 ± 11.1 (full text)

Two complementary findings from the same dataset:

1

HIDIT delivered about 1.7× the time >90% VO2peak of either SIHIIT or LIHIIT (312 vs 182 and 179 s), with no significant difference in Tlim, blood lactate or RPE at exhaustion (abstract). Inference, not tested: the long first interval drives VO2 up quickly; the subsequent shorter intervals keep it there as W' depletes.

2

Work above CP (kJ) does not track time >90% VO2peak across formats — they dissociate (full text, checked 2026-09-27): the format with the most time near VO2peak (HIDIT) had only slightly more work above CP than the long format, and the short format had slightly less work above CP than the long format (18.7 vs 19.3 kJ) while its time >90% VO2peak was the same as the long format's.

Mechanism for work-above-CP dissociation

Inference, not tested in this paper: VO2 during HIT is not determined solely by total time or work above CP but by the interaction of work intensity, recovery intensity and work:rest ratio. In short-interval formats VO2 stays elevated through the brief recoveries (slow VO2 off-kinetics), so %VO2max-minutes accumulate despite a modest work-above-CP integral; in long-interval formats the longer recoveries let VO2 fall between bouts even when the integral is comparable.

Practical Application

In n=12 cyclists riding three formats to exhaustion with the same work power, recovery power and 3:2 work:recovery ratio, the decreasing format (3 min → 30 s work, 2 min → 20 s recovery) accumulated 312 ± 207 s above 90% VO2peak vs 182 ± 225 s for 30 s/20 s and 179 ± 145 s for 3 min/2 min (P = 0.036 and 0.027), with no difference in time to failure (798 vs 714 vs 664 s), lactate or RPE. In VO2max blocks the decreasing format buys more time near VO2peak for the same perceived cost.
Do NOT use "work above CP / work above FTP" in kJ as a HIT-aerobic-effectiveness metric (dissociation from %VO2max and time >90% VO2peak: full text, checked 2026-09-27). It is widely available in WKO5 / TrainingPeaks / intervals.icu and tempting to read as a session-quality score — but for the same kJ above CP, different formats can deliver very different aerobic stimulus.
Not from this paper; the base's own practice: for evaluating a HIT session in the field, use time at or above the target power (e.g. time ≥ MAP, time ≥ 110% CP), not total kJ above CP/FTP.
This is the power-side analogue of the %HRmax caveat (REF-041 Odden 2024 / SYNTH-16) — a popular convenience metric that turns out not to track the underlying aerobic signal.

Limitations

n=12 amateur cohort, single test occasion per format — between-individual variance is large (SDs comparable to means for time >90% VO2peak)
All formats run to exhaustion, not matched-volume; in a fixed-dose session the dissociation may differ
Critical Power estimated from 3-4 constant-load trials to exhaustion at 80-100% of ramp peak power (work-time model), not from a single all-out test (full text, checked 2026-09-27)
Acute responses only: no training study shows that more time >90% VO2peak per session translates into larger VO2max gains with this format
How this entry was checked and changed
Why this level

regraded 2026-09-27 from A: crossover, n=12

Content correction 2026-09-27
SIHIIT time >90% VO2peak 183 s -> 182 ± 225 s (abstract); P-values added.
Cohort descriptors (VO2max 56.8, CP 212-320 W, age 29-62), work/recovery intensities (117%/83% CP), the mean %VO2max and work-above-CP columns and the HIDIT step sequence are not in the abstract -> labelled "full text; not checked against the abstract".
Derived percentages "+70 %", "+10 %", "+3.3 %", "−3.2 %" -> deleted (not in the abstract and not consistent with the table values); the dissociation is stated qualitatively under the full-text label.
Mechanism and "time at or above target power" recommendation -> labelled inference / the base's own practice.
Lines labelled "full text; not checked against the abstract" were checked against the open-access full text (PMC7560936): every number in them appears there; relabelled "full text, checked 2026-09-27".
A separate set of lines labelled "full text; not checked" (without "against the abstract") was read against the open-access full text (PMC7560936): 1 corrected to what the paper says; 1 contradicted by the paper — corrected.
2026-09-28: the dissociation line said the short format "sat between" the others on work above CP; the table puts it lowest (18.7 kJ vs 19.3 long, 22.0 HIDIT) -> corrected.
Resync 2026-09-28: Finding 2 said the short format sat between the others on work above CP though its 18.7 kJ is below the long format's 19.3 kJ -> now says so; the two full-text-check log lines are distinguished.
Full text (checked 2026-10-12)
Participants were twelve middle-age amateur, non-smoking cyclists who reported at least three training sessions per week in the previous 6 months; the discussion refers to the sample as "well-trained male adults" (full text).
Group means were CP 254 ± 30 W, W' 12.8 ± 4.1 kJ, ramp load peak 356 ± 40 W, high-intensity bouts 297 ± 35 W and low-intensity bouts 212 ± 30 W (full text, Table 1).
High-intensity power was set as W'/300 s + CP (the power expected to cause exhaustion in 5 min); every test began with 10 min at 100 W plus 2 min rest, at a preferred cadence of ~ 90 rpm, and exhaustion was the inability to hold the cadence within 10 rpm for longer than 5 s (full text).
Time > 90% VO2peak was longer in HIDIT than in SIHIIT (P = 0.036; ES: 0.62) and LIHIIT (P = 0.027; ES: 0.64), while Table 3 lists SIHIIT at 183 ± 225 s against 182 ± 225 s in the abstract (full text, Table 3).
At the 3rd minute, %VO2peak was 90.75 ± 5.94 in HIDIT, 85.33 ± 7.11 in SIHIIT and 89.58 ± 6.52 in LIHIIT, with lactate at 8.03 ± 2.69, 5.69 ± 1.62 and 7.85 ± 3.01 mmol L−1 and CR10 at 6.67 ± 2.12, 5.29 ± 1.57 and 6.52 ± 2.03 (full text, Table 3).
At Tlim, lactate was 10.71 ± 4.72 (HIDIT), 10.75 ± 2.04 (SIHIIT) and 10.83 ± 3.58 mmol L−1 (LIHIIT), and CR10 was 9.25 ± 1.78, 9.48 ± 0.70 and 9.56 ± 1.08 (full text, Table 3).