An independent 2023 taper meta-analysis reproduces Bosquet's shape: the time trial improves and VO2max and economy do not
And the effect appears only in the subgroups that KEPT intensity and frequency
Key Findings
Abstract-level results: 14 studies; pre- to post-taper time trial (TT) improved (SMD −0.45, P < 0.05) and time to exhaustion (TTE) improved (SMD 1.28, P < 0.05); VO2max and economy of movement did not (P > 0.05). Pre-taper overload beat conventional taper on TT (P < 0.05).
Volume cut 41-60%, intensity and frequency maintained, duration ≤7, 8-14 or 15-21 days, progressive or step shape — each significantly improved TT (P < 0.05). The confidence intervals, heterogeneity values and non-significant subgroup estimates below are from the paper's full text (checked 2026-09-27 against the PMC full text; not in the abstract).
Time trial improves: SMD −0.45 [−0.68, −0.23], p<0.05, with low heterogeneity (I² = 8%). Against a control group the separation is larger: SMD −0.73 [−1.25, −0.21]. (full text; CIs and I² not in the abstract)
Time to exhaustion improves: SMD 1.28 [0.43, 2.12], p<0.05 — but I² = 64%, so treat the magnitude as unstable. (full text; CI and I² not in the abstract)
VO2max does not improve pre-to-post taper: SMD 0.20 [−0.93, 1.33], p>0.05, I² = 87%. It separates from control at SMD 0.77 [0.31, 1.23]; inference, not tested: that reflects the control group losing ground rather than the taper adding VO2max. (full text; not in the abstract)
Movement economy does not improve: SMD −0.47 [−1.06, 0.12], p>0.05. (full text; not in the abstract)
Intensity: the effect is present only when intensity was kept. Intensity maintained: SMD −0.55 [−0.79, −0.31], p<0.05. Intensity decreased: SMD 0.25 [−0.67, 1.17], p>0.05 — the effect disappears and the point estimate flips sign; the between-subgroup test itself was P = 0.10. (full text; not in the abstract)
Frequency behaves the same way. Frequency maintained: SMD −0.53 [−0.82, −0.25], p<0.05. Frequency decreased: SMD −0.32 [−0.76, 0.13], not significant; between-subgroup P = 0.42. (full text; not in the abstract)
Volume: only the 41-60% cut reached significance — SMD −0.77 [−1.23, −0.30], p<0.05. Cuts of ≤20%, 21-40% and ≥60% did not (p>0.05 each); between-subgroup P = 0.32. Too little and too much are both worse than the middle. (full text; CI and non-significant bands' P not in the abstract)
Duration: every band up to 21 days worked; past that it faded. ≤7 days significant; 8-14 days largest, SMD −1.47 [−2.75, −0.19]; 15-21 days significant; ≥22 days smallest and not significant, with a significant between-subgroup difference (P = 0.02). (full text; not in the abstract)
Taper shape: progressive SMD −0.51 [−0.81, −0.20] and step SMD −0.38 [−0.73, −0.04] — both work, progressive slightly larger; between-subgroup P = 0.60. (full text; not in the abstract)
Overload before the taper beat a conventional taper on time trial (P < 0.05; I² = 0% in both arms — full text; the I² not in the abstract).
Practical Application
Keep the taper inside three weeks. Meta-analysis of 14 taper studies in endurance athletes: a ≤21-day taper with volume progressively reduced by 41-60% and intensity and frequency unchanged improved time-trial performance (pooled SMD −0.45); every duration band up to 21 days worked, and the ≥22-day band was the smallest and not significant, with the duration bands differing significantly (P = 0.02 — full text; not in the abstract).A month of "peaking" is not a longer peak, it is a worse one.
Cut volume to the middle band — 41-60% off peak. That was the only reduction band that reached significance in this pooled data; the ≤20%, 21-40% and ≥60% cuts did not (P > 0.05 each — full text; not in the abstract).
Do not touch intensity or session frequency. The abstract's effective strategy keeps both unchanged; in the subgroup data (full text; not in the abstract) reducing intensity left a non-significant estimate that reversed sign, and reducing frequency lost significance. Same days, same hard sessions, far fewer hours.
Promise a better race, not a better test. Time-trial and time-to-exhaustion performance moved; VO2max and movement economy did not (P > 0.05). If an athlete tests VO2max after a taper and it is unchanged, that is the expected result and not a failed taper.
A hard block immediately before the taper is part of the taper. Taper combined with pre-taper overload had a more significant effect on TT than a conventional taper (P < 0.05), which SYNTH-34 carries as "finish the build hard".
Limitations
Small evidence base: 14 studies, 174 athletes (full text; not in the abstract). Several subgroups rest on a handful of trials, which is why the 8-14 day estimate carries a confidence interval from −2.75 to −0.19 (full text).
Predominantly male samples; the authors name limited female representation as a gap (full text; not in the abstract).
Included sports were mainly cycling and running; speed skating, cross-country skiing and kayaking are underrepresented (full text; not in the abstract).
No stratification by training status — the authors did not analyse athletes of different training statuses separately (full text; not in the abstract), so the numbers cannot be tiered.
High heterogeneity for time to exhaustion (I² = 64%) and VO2max (I² = 87%) (full text; not in the abstract).
Subgroup findings are observational comparisons across studies, not randomised contrasts: "intensity maintained" studies may differ from "intensity reduced" studies in more than intensity. Only the duration subgroups differed significantly between bands; intensity (P = 0.10), volume, frequency and shape did not (full text; not in the abstract).
How this entry was checked and changed
Content correction 2026-09-27
Confidence intervals, I² values, non-significant subgroup estimates, n=174 and ages 17-32 are not in the abstract; each such line is now labelled "(full text; not in the abstract)" after checking them against the PMC full text (all present).
Added the between-subgroup P values from the paper's Table 2: only the duration comparison was significant (P = 0.02); intensity P = 0.10, volume P = 0.32, frequency P = 0.42, shape P = 0.60 — the entry's "intensity decides whether the taper works at all" softened accordingly.
"≥22 days smallest, with a significant between-subgroup difference" -> clarified: the ≥22-day estimate itself was not significant. "Sport: cycling, running, triathlon, swimming" -> "cycling, running" (the included studies were mainly running and cycling; no swimming or triathlon studies are named).
Lead bullet now opens with the abstract's finding (≤21-day taper, −41-60% volume, intensity/frequency unchanged, TT SMD −0.45).
Resync 2026-09-28: Attributed 'detraining within two weeks' to REF-110, which removed it -> risk of detraining, no time course.
2026-10-01: cross-reference repointed from a merged framework to the one that replaced it (SYNTH-01→31, 05→34, 09→70, 12/15→32, 19→26).
2026-10-01: citation of the superseded pointer SYNTH-05 repointed to SYNTH-34 (the two-day mini-overload wording is not in SYNTH-34 and was dropped), which carries the cited content.
Full text (checked 2026-10-12)
Duration subgroups on TT: ≤7 days SMD −0.36 (−0.63, −0.08), 15–21 days SMD −0.78 (−1.43, −0.14), ≥22 days SMD 0.69 (−0.22, 1.60), the last based on N = 10 (full text, Table 2).
Volume subgroups on TT: ≤20% SMD −0.25 (−1.27, 0.77), 21–40% SMD −0.45 (−0.90, −0.00) with P = 0.05, ≥60% SMD −0.21 (−0.58, 0.16) (full text, Table 2).
In the original studies the pre-taper overload training had a mean intensity of 85–95% of maximum heart rate (HRmax), an increase of 23–26% over the mean intensity of normal training (full text).
Only two included studies had female participants (M/F 12/4 and 8/3); all others were male only (full text, Table 1).
Inclusion required endurance athletes with VO2max > 55 ml/kg/min (or at least well-trained), a taper of ≥6 days, and PEDro > 4 points; included studies scored 5 to 6 points (full text).
TT distances ranged from 800 m to 40 km, tested on track, treadmill and cycle ergometer; taper durations ranged from 6 to 28 days and volume reductions up to 85% (full text).
Economy of movement heterogeneity was I2 = 57% across four studies (full text).