REF-1203 · Bretonneau et al., 2024 · Frontiers in Sports and Active Living · Paper · PubMed · Free full text
Should I go into my taper overreached?
In 26 national swimmers, those only tired 3 weeks out raced 1.8% faster at the target meet; those flagged overreached raced 0.5% slower but also got smaller tapers. Observational.
Bretonneau et al. asked whether the level of fatigue a swimmer carries into the taper changes what the taper gives, and whether that fatigue goes with sleep.
Across three French national training centres in 2021 they followed 26 national-level swimmers (12 men, 14 women) through the last 13 weeks before each swimmer's main competition (French National 1, Junior or National 2 championships).
Heart-rate, mood (POMS) and tethered-swimming force profiles were measured at T0 (a two-week baseline 10 and 9 weeks out) and at T1 (a week 3 weeks out); from the size and number of negative changes between the two, 14 swimmers were classed afterwards as acutely fatigued (AF) and 12 as functionally overreached (F-OR).
Sleep was recorded every night of both periods with an EEG headband and a wrist accelerometer, and core temperature on one or two nights with an ingested pill. From T1 to the main competition, the race closest to each swimmer's career best improved by 1.80 ± 1.36% in AF and worsened by 0.49 ± 1.58% in F-OR (p < 0.05).
F-OR swimmers had a shorter total sleep time at both T0 and T1, more daytime sleepiness and, as a tendency, a more fragmented night. Their coaches also tapered them less, though no test of the difference is reported: overall load fell 23 ± 32% over 10 days in F-OR against 30 ± 20% over 12 days in AF, and dryland load rose in F-OR.
Our reading: not a replication of REF-1056 (no assigned overload, no control group, overreaching defined by markers rather than a performance drop), but an observational result in another sport, at a real championship, that points the same way, consistent with "hard, not overreached" before a taper (SYNTH-34); but the study cannot tell whether the overreached swimmers fell short because of their fatigue or because their taper was smaller, the performance difference held for only one of the two race-selection methods, and the sleep difference was already there ten weeks out.
Protocol
Design: prospective, multicentre, observational; the investigators described what the training groups did rather than imposing a taper. Bosquet's meta-analytic recommendations (REF-107) were presented to all coaches at the start. Approved by the Ile de France committee for the protection of persons (ID-RCB 2020-A03559-30). Three national training centres, January to December 2021.
Timeline: protocol from 13 weeks before the main competition to just after it. T0 = a two-week baseline 10 and 9 weeks out; T1 = one week, 3 weeks out. Official race times at T0 and T1 (± 7 days) and at the main competition.
Who: n = 30 recruited (15 men, 15 women; power analysis from Aubry's data asked for 8 per group, i.e. at least 16; the authors say they doubled this, yet recruited 30, not 32); 3 men withdrew and 1 woman was excluded for injury. Final 12 men (17.9 ± 2.3 y, 1.80 ± 0.08 m, 71.0 ± 8.4 kg) and 14 women (16.4 ± 1.3 y, 1.69 ± 0.05 m, 59.6 ± 9.3 kg). AF group n = 14 (16.8 ± 1.7 y), F-OR group n = 12 (17.5 ± 2.1 y).
Training load: session RPE (CR-10) × duration from a logbook, swimming and dryland separately; the taper change was the 14-day sum against the daily mean of the three weeks before. Coaches kept their usual approach to load, taper included.
Profiles at T0 and T1: physiological — resting heart rate (5 min lying, each morning), exercise heart rate in a daily 6-min swim over a fixed distance at RPE 3 (120-160 beats/min) and heart-rate recovery over 60 s, all from a wrist PPG monitor (Polar Unite); psychological — POMS (vigor, fatigue and the energy index, vigor minus fatigue) once per period, CD-RISC resilience once; biomechanical — mean force and impulse in two 10-s maximal tethered front-crawl efforts.
Classification: each change graded as trivial, small, moderate, large or very large by Cohen's d, a Z score or multiples of the smallest worthwhile change (0.2 × between-athlete SD). AF = several small to moderate negative changes and no more than one large one; F-OR = at least one large negative change in two or more of the three profiles (method of Vachon et al.).
Sleep: Dreem 1 EEG headband and ActiGraph wGT3X-BT wrist accelerometer every night of T0 and T1; chronotype (Horne-Ostberg), PSQI, Epworth sleepiness scale and a daily Spiegel questionnaire; core temperature from an ingested pill (e-Celsius) on at least one mid-week night per period.
Performance: official electronic race times; for each swimmer the race whose career personal best was closest to the world record ("PB Career", primary) and the race whose best time during the protocol was closest to it ("PB Protocol", secondary).
Analysis: t-test or Wilcoxon between groups, two-way ANOVA (group × period), Hedges' g, Pearson or Spearman correlations; p < .05.
Funding and interests: funded by the French government's third Future Investment Program (PIA3) through the National Agency of Research (ANR-19-STHP-0001); no commercial or financial conflicts declared. One author works for the French Swimming Federation.
Key Results
| Measure | AF (n = 14) | F-OR (n = 12) | Between groups |
|---|---|---|---|
| Race time change T0 → T1, PB Career race | −0.36 ± 1.79% | +0.25 ± 2.28% | not different |
| Performance change T1 → main competition, PB Career race | improved 1.80 ± 1.36% | worse by 0.49 ± 1.58% | p < 0.05 |
| Change T1 → main competition, PB Protocol race (sign convention not stated) | −2.10 ± 1.09% | −0.49 ± 1.86% | no difference |
| Overall training load in the taper | −30 ± 20%, linear, from 12 days out (r² = 0.97) | −23 ± 32%, exponential, from 10 days out (r² = 0.98) | no test reported |
| Training duration / session RPE in the taper | −26 ± 15% / −9 ± 26% | −18 ± 21% / −12 ± 28% | no difference |
| Dryland load in the taper | −30 ± 38% | +38 ± 71% | reported as contrasting, no p |
| Total sleep time, T0 / T1 (h:min) | 08:15 ± 00:39 / 07:55 ± 00:48 | 07:47 ± 00:37 / 07:08 ± 00:56 | group effect |
| Epworth sleepiness, T0 / T1 | 8.4 ± 4.6 / 7.9 ± 2.5 | 10.3 ± 2.6 / 10.9 ± 3.3 | group effect |
| Awakenings per hour, T0 / T1 | 3.4 ± 0.7 / 3.7 ± 0.5 | 2.9 ± 1.3 / 2.9 ± 1.0 | group effect (fewer in F-OR) |
| Fragmentation index, T0 / T1 | 11.5 ± 3.2% / 12.1 ± 4.4% | 14.0 ± 4.7% / 15.2 ± 3.0% | p = .06 (abstract) |
| Sleep efficiency, T0 / T1 | 91.2 ± 3.0% / 91.5 ± 2.4% | 91.6 ± 2.1% / 91.5 ± 4.5% | not different |
| Core temperature during sleep, T0 → T1 | 36.48 → 36.50 °C | 36.61 → 36.67 °C (rose) | — |
| Temperature nadir | 36.12 ± 0.15 °C | 36.31 ± 0.25 °C | lower in AF |
Swim load fell 28 ± 27% on average in the taper with no group difference; the authors note both tapers were smaller and shorter than the 40-60% over 14 days they had recommended.
213 official performances were recorded (about 8 per swimmer).
Correlations (T0 to T1, the training block, not the taper): the change in race time from T0 to T1 correlated with wake-up time at T0 (r = −0.95), with temperature 2 and 3 h after lights out (r = 0.78 and 0.68), with the temperature drop from lights out to the nadir (r = 0.70) and with the shift in lights-out time (r = 0.69).
Where the abstract and the body disagree: (1) the abstract and the Discussion say core temperature fell from wakefulness to sleep in AF "but not in F-OR"; the Results text gives mean falls in the hour before bedtime in both groups (0.23 ± 0.26 °C in AF, 0.14 ± 0.22 °C in F-OR; group effect p = .07) and leaves within-group significance to Figure 3, so this may be a significance statement rather than a contradiction.(2) The abstract's "before taper period" covers both T0 and T1; it does not make clear that the sleep difference was already present at T0, ten weeks out. (3) The abstract does not mention the PB Protocol race, for which the groups did not differ, nor that F-OR swimmers woke less often per hour.
Numbers that do not add up: the AF gain is 1.80% in the Results and 1.82% in the Discussion. For the PB Protocol race the paper gives −2.10% and −0.49% without saying whether these are race-time or performance changes; read as race times both groups swam faster, AF by more, so the same 0.49 appears for F-OR with opposite meanings in the two race selections (our reading).
Mechanisms
Measured: the F-OR swimmers had a shorter total sleep time and more daytime sleepiness in both periods, less N1 and N2 sleep, a tendency to a more fragmented night, and fewer awakenings per hour; their nocturnal core temperature was higher (nadir 36.31 vs 36.12 °C) and rose from T0 to T1, while the AF group's did not change.
Authors' interpretation: poorer pre-taper sleep could help explain why the two groups responded differently to the same training load, possibly through a weaker drop in internal temperature at the start of the night. They also read the result as a taper that did not match the fatigue: the more fatigued group got the smaller reduction, the opposite of what fitness-fatigue models prescribe.
Inference: the study cannot separate the two explanations. The F-OR group differed in fatigue state and in taper dose (−23% over 10 days, with dryland load up 38%, against −30% over 12 days), and both point the same way. (Authors' interpretation: because pre-taper load, session RPE × duration, did not differ between groups, the overreaching was a different response to the same load, not a heavier block.)
Inference: the sleep difference was present ten weeks out, before the classification period ended, so it may be a trait of these swimmers (or of their centres) rather than a product of the training block.
Practical Application
Should I go into my taper overreached? In 26 national swimmers, those only tired 3 weeks out raced 1.8% faster at the target meet; those flagged overreached raced 0.5% slower but also got smaller tapers. Observational.
Same direction as [REF-1056]: in trained triathletes whose overload was assigned, the acutely fatigued gained more over the taper than the functionally overreached. Two small studies in two sports now point the same way: the block before the taper should tire, not dig a hole (SYNTH-34, "Finish the build hard"; B, observational).
What flagged "overreached" here: a large negative change in at least two of three profiles — morning, exercise or recovery heart rate; POMS vigor and fatigue; and maximal tethered force — while race times had not yet dropped. Inference for coaching: several markers worsening together across a block matter more than one bad number; how to read daily readiness is in SYNTH-30.
Sleep as a warning sign (association, not tested as an intervention): the overreached swimmers slept 28 min less a night ten weeks out and 47 min less three weeks out (our arithmetic from the table) and were sleepier by day. The authors' advice is to assess and improve sleep before an overload, not only in the taper (REF-203 for sleep hygiene, C+).
If an athlete arrives at the taper overreached: the overreached swimmers here got the smaller, shorter taper, and the authors call for a larger load reduction and added recovery for them. Inference, not tested; SYNTH-34 carries it as the base's practice: treat the first days of the taper as recovery, cut volume more, and do not test fitness until the signs settle.
Limitations
Observational and post hoc. The groups were formed after the fact from marker changes; there was no control group and no randomisation, and the taper was each coach's choice. The groups then differed in taper dose and length, so fatigue state and taper size are confounded.
Small. 14 and 12 swimmers; the power analysis (8 per group) was based on the triathlon study it set out to replicate.
One of two race selections. The headline difference holds for the race closest to the career best; for the race closest to the protocol's best the groups did not differ.
Classification. "Functional overreaching" is a composite of grading thresholds for heart rate, mood and maximal tethered force, with no between-group difference in race times before the taper; another rule could split the 26 differently.
Measurement. Sleep from a consumer EEG headband and wrist accelerometer over one or two weeks; core temperature on one night (two in 18% of cases), which the authors themselves say calls for caution.
Many correlations in 26 swimmers without correction (our observation); an r of −0.95 between wake-up time and a change in race time is implausibly tight for field data and should not be quoted as a rule.
Who. National-level adolescent and young adult swimmers, three different championships; not adults, not cyclists or runners.
Funding and interests: public funding (ANR-19-STHP-0001); no conflicts declared.
How this entry was checked and changed
Why this level
small cohort (n = 26), read from the full text (open access, CC BY 4.0). No control group and no randomisation, the taper was not standardised and differed between the groups, and the comparison of interest is a post-hoc split
Full text (checked 2026-10-26)
Fourteen swimmers (17 ± 2 years) were allocated to the acute fatigue group and 12 swimmers (18 ± 2 years) to the functional overreaching group (full text)
From T1 to the main competition performance improved in AF (+1.80 ± 1.36%) and was impaired in F-OR (−0.49 ± 1.58%, p < 0.05 vs. AF) (full text)
All swimmers were considered Tier 3 according to McKay et al. (highly trained/national) (full text)
Overall training load fell 30 ± 20% linearly 12 days before the competition in AF and 23 ± 32% exponentially 10 days before in F-OR; dryland-training load decreased by 30 ± 38% in AF and increased by 38 ± 71% in F-OR (full text)
No difference in performance changes for the race associated with PB Protocol (−2.10 ± 1.09% vs. −0.49 ± 1.86%) (full text)
Total sleep time at T0 was 08:15 ± 00:39 in AF and 07:47 ± 00:37 in F-OR; at T1 07:55 ± 00:48 and 07:08 ± 00:56 (full text)
During the hour preceding bedtime temperature decreased by 0.23 ± 0.26°C in AF and by 0.14 ± 0.22°C in F-OR (p = .07 for the group effect) (full text)
The taper of AF led to an average improvement of 1.82% (Discussion; full text)
Funded by the French government through the third Future Investment Program (PIA3) of the National Agency of Research, reference ANR-19-STHP-0001 (full text)