REF-1275 · Rosenblat et al., 2021 · Sports Medicine · Paper · PubMed

How much faster will an interval block make me?

About 2% in a time trial if you are trained, up to ~6% if less trained (29 studies, 491 adults, mostly young men; before-after changes, no CIs in the abstract). A-, from the abstract.

Main caveat: Read from the abstract only. Which studies had controls, the pooled estimates with intervals, heterogeneity and the dose-response curves could not be checked.
fairly strong evidenceMeasured on non-athletes to highly trained athletes

Rosenblat, Lin, da Costa and Thomas asked how individual characteristics and training variables shape the time-trial gain from interval training.

They searched SPORTDiscus and Medline for prospective studies of adults aged 18-65 that ran a high-intensity interval (HIIT) or sprint-interval (SIT) programme for at least 2 weeks, programmed HIIT by power or velocity, and tested a time trial over a set distance.

Twenty-nine studies with 67 separate groups qualified: 400 men and 91 women, mean group age 25 (range 19-45) years and mean VO2max 52 (32-70) mL/kg/min; 75 participants were inactive, 146 active and 258 trained. Training status played a significant role: trained individuals improved their time trial by about 2%, while those of lower training status improved by as much as 6%.

With HIIT the change depended on the duration of the work bout but not on its intensity, and there was a dose-response relationship with the number of HIIT sessions, training weeks and total work; with SIT there was none.

The authors conclude that interval programmes should be built according to training status, that increasing the dose beyond minimal requirements may not augment the response, and that optimal dosing differs between HIIT and SIT.

Our reading: the honest expectation for a trained athlete from an interval block is a time trial about 2% faster on average — close to a laboratory test's noise and the first-retest learning effect if both are in the same unit (REF-1278: CV 0.9-2.0% of power for tests of 1 min to 3 h, +1.2% between the first two trials; the abstract does not say whether the 2% is time or power, and in cycling a time change is smaller than the power change behind it — inference) — so one retest may not settle it.

The abstract does not say how many studies had a control group, and gives no confidence intervals for the 2% and 6%.

Protocol

Sources: SPORTDiscus and Medline with Full Text; search dates not in the abstract.
Inclusion: prospective studies; ages 18-65 years; an interval training programme (HIIT or SIT) of at least 2 weeks; a time trial over a set distance; HIIT programmed by power or velocity. A control group is not among the criteria listed in the abstract.
Included: 29 studies, 67 groups; 400 men and 91 women (491 in total, our sum; the abstract gives the two counts); mean group age 25 (19-45) years; mean VO2max 52 (32-70) mL/kg/min; training status inactive 75, active 146, trained 258 (these three sum to 479, not 491 — our arithmetic).
Moderators analysed: training status; HIIT work-bout duration and intensity; number of sessions, training weeks and total work; HIIT vs SIT.
Not in the abstract: the definitions of the training-status groups, time-trial distances and modes, the number of controlled studies, the pooled estimates with intervals, heterogeneity, the shape of the dose-response.
Funding and interests: not stated in the abstract or the PubMed record.

Key Results

Factor (abstract)Time-trial change
Trained participants (n = 258)about 2% improvement
Lower training status (inactive 75, active 146)improvements as high as 6%
Training status as a moderatorsignificant
HIIT work-bout durationthe change depended on it
HIIT work-bout intensitythe change did not depend on it
HIIT dose: number of sessions, training weeks, total workdose-response relationship with the change
SIT doseno dose-response
Confidence intervals, heterogeneity, number of controlled studiesnot in the abstract
Authors' conclusion: optimise interval programmes by training status; "increasing interval training dose beyond minimal requirements may not augment the training response"; optimal dosing differs between HIIT and SIT.

Mechanisms

Measured: time-trial change by training status and by programming variables, pooled across groups.
Authors' interpretation: interval dose should be set by training status; beyond a minimal dose more may not add; HIIT and SIT need different dosing.
Inference: people of lower training status start further from their ceiling, so the same programme moves them more — the same pattern as larger VO2max gains at lower baseline fitness (REF-1276) and larger mitochondrial and capillary gains at lower initial fitness (REF-465). A beginner's first-block gain is not the slope for later blocks (REF-529, an inference in that entry).
Inference: a before-after time trial also improves from practice: across 101 reliability studies the second trial was 1.2% better than the first (REF-1278). If most included groups had no control, part of the trained group's ~2% may be learning.
Inference: the abstract reports a dose-response with HIIT sessions, weeks and total work, yet concludes that dose beyond minimal requirements may not help. Both can hold only if the gain flattens after a minimum, and the abstract shows no curve; why HIIT gains followed bout duration but not intensity is not explained there either.

Practical Application

How much faster will an interval block make me? About 2% in a time trial if you are trained, up to ~6% if less trained (29 studies, 491 adults, mostly young men; before-after changes, no CIs in the abstract). A-, from the abstract.
Say it as a range around a small number. For a trained amateur the expected payoff of one interval block is a time trial a couple of per cent faster on average, close to a retest's noise if both are in the same unit (REF-1278 gives power CVs; the abstract does not say whether the 2% is time or power); a gain is believable only with the same protocol (SYNTH-53) and after a familiarisation trial (REF-1278) (inference).
A beginner's first block is not the forecast for the next. Gains up to ~6% at lower training status shrink toward ~2% as the athlete becomes trained (inference from the training-status difference; REF-529).
More intervals are not automatically more gain. HIIT showed a dose-response with sessions, weeks and total work, but the authors conclude that dose beyond a minimum "may not augment" the response, and sprint intervals showed no dose-response at all. Not from this paper; the base's own practice (SYNTH-42): progress one lever per block and read it at the review date.
Against the starting bet: REF-999's 4.4-5.8% 40-km gain in 4 weeks for highly trained cyclists, quoted in SYNTH-42, sits above this pooled ~2% for trained people; quote the pooled figure as the expectation and the single trial as a good case (inference).

Limitations

Read from the abstract only. Which studies had controls, the pooled estimates with intervals, heterogeneity and the dose-response curves could not be checked.
Before-after, not necessarily controlled. The inclusion criteria ask for prospective studies; as far as the abstract shows, the gains include learning, season and other training, not only the intervals.
Who. 400 of 491 were men and the mean group age was 25; women and masters athletes are under-represented, and the training-status labels are not defined in the abstract.
Numbers that do not add up. The three training-status counts (75 + 146 + 258 = 479) fall 12 short of the 491 participants (our arithmetic). The Results report a dose-response for HIIT, while the Conclusion says more dose beyond a minimum may not help; the abstract gives nothing that reconciles the two.
Two databases, and the time-trial distances and modes are not described in the abstract.
Funding and interests: not stated in the abstract or the PubMed record. The first author's contact address in the PubMed affiliations is at evidencebasedcoaching.ca (as printed; no conflict-of-interest statement was found in the record).
How this entry was checked and changed
Why this level

a meta-analysis, held below A because the inclusion criteria asked for prospective studies, not controlled ones, so the pooled gains are before-after changes as far as the abstract shows; the studies are small (491 participants across 67 groups) and 81% of participants were men (400 of 491, our arithmetic); read from the abstract (closed access; no Europe PMC copy, and the University of Bern repository record listed by OpenAlex is not open access and sits behind a bot-check page)