Limited reproducibility of individual physiological adaptations to repeated endurance exercise training
Participants
n=42 middle-aged untrained men and women completed both periods. Sex split (21/21), menopausal status, age range 30-65, medication, ancestry and the 53 enrolled with reasons for withdrawal (full text; not checked against the abstract).
Protocol
Two identical 8-week training periods (TP₁ and TP₂) separated by an 8-week detraining period.
Each period: supervised cycling interval training; session count 24 per period is implied by the compliance figures. Trainer model, 2-weekly intensity adjustment from a 4×5-min test, and the session structure (6×6-min MIT at RPE 13-14, 4×8-min HIT as 30 s at 108% / 15 s at 52% at RPE 16-17, 4×5-min maximal intervals) are full text; not checked against the abstract.
Hbmass restored to TP₁ pre-values by phlebotomy before TP₂ (full text; not checked against the abstract); the abstract states that baseline values were highly consistent between periods, "indicating effective detraining".
Outcomes named in the abstract: haemoglobin mass (Hbmass), skeletal-muscle citrate synthase (CS) activity, capillaries per fibre, VO₂max and 15-min maximal mean power output (MMPO15-min). GE, lactate at 25% and 40% MMPO, HAD and fibre type are full text; not checked against the abstract.
Compliance: 23.9 (0.4) and 23.8 (0.7) sessions completed in TP₁ and TP₂ — minimal within-individual variation in training dose.
Key Findings — within-individual reproducibility (ICC [95% CI])
| Variable | ICC | Interpretation |
|---|---|---|
| Baseline VO₂max | 0.98 [0.97, 0.99] | Near-perfect baseline reproducibility — phenotype stable, test reliable |
| Baseline MMPO15-min | 0.96 [0.93, 0.97] | Near-perfect baseline reproducibility |
| ΔVO₂max response | 0.19 [0.00, 0.37] | *Poor* within-individual reproducibility of adaptation |
| ΔMMPO15-min response | 0.20 [0.00, 0.37] | *Poor* |
| ΔHbmass response | 0.00 [0.00, 0.30] | *Essentially no reproducibility* |
| ΔCS activity response | 0.15 [0.00, 0.33] | *Poor* |
| Δcapillaries/fibre response | 0.00 [0.00, 0.36] | *Essentially no reproducibility* |
Group-mean adaptations were consistent between TP₁ and TP₂ — population averages reproduce. Individual adaptation patterns did not.
Within-individual variability in haematological and skeletal-muscle adaptations explained little of the within-individual variability in ΔVO2max and ΔMMPO15-min responses — i.e. the inconsistency at the performance level is not just downstream of inconsistency at one or two specific upstream tissues.
Headline conclusion (paraphrased)
> The authors conclude that individual responses to endurance training reproduce poorly across many physiological and performance outcomes, and that the capacity to adapt does "not reflect a stable, individual-specific attribute" but arises from dynamic, context-dependent biological processes.
Mechanism / why this happens
Inference, not tested — the abstract says only that adaptability "emerges from dynamic and context-dependent biological processes"; the list below is the base's reading of what those contexts could be.
Adaptation is the integrated output of many context-dependent inputs: prior training history at start of block, sleep + nutrition + stress at the time, hormonal cycle phase, infections/sub-clinical illness, day-to-day RPE drift, recent injury / soreness, even season and weather. Most of these were not stabilised between TP₁ and TP₂.
The two training periods were separated by 8 weeks — long enough that the participant arriving at TP₂ was in a different physiological context than at TP₁, even after detraining.
Phlebotomy controlled for Hbmass starting point but could not control for the more subtle context shifts.
Suggests that "responsiveness" is state-dependent, not trait-like.
Practical Application — major reframe
Odden et al. 2026 (n=42 untrained middle-aged adults, the same 8-week programme done twice with 8 weeks of detraining between): baseline VO₂max and 15-min power were near-identical between periods (ICC 0.98 and 0.96), but each person's response was not — ICC 0.19 for ΔVO₂max, 0.20 for ΔMMPO, 0.15 for citrate synthase and 0.00 for Hbmass and capillaries. How much an athlete adapts to a block is not a fixed personal trait; one block's response does not predict the next.
The "non-responder" label is unstable across blocks. A given individual may show low ΔVO2max in one 8-week block and a substantial ΔVO2max in the next identical block, with no clear reason.
This rescues the rescue-framework (REF-432 Bonafiglia/Vollaard): switching stimulus is one way to rescue non-response, but just running the same stimulus a second time may rescue too, simply because context shifts.
Argues against locking athletes into "their optimal stimulus" based on past response patterns. Past response is descriptive snapshot, not predictive lock-in.
Argues for varied stimuli + periodic re-assessment rather than over-engineered hyper-personalisation.
Argues for measuring more than one block before drawing strong individual-pattern conclusions. A single block's adaptation magnitude is not a reliable estimate of the athlete's "trainability."
Baseline performance phenotype IS reproducible across blocks (ICC 0.98 for VO₂max, 0.96 for MMPO15-min, "and most other outcomes"). What is *not* reproducible is the magnitude / pattern of response to a given stimulus.
Limitations
Untrained middle-aged cohort — applicability to highly trained masters / elite athletes is uncertain. Trained populations may have less room for adaptation (so variance is compressed) or may have more stable patterns (more crystallised physiological habits) — unknown.
8 weeks per training period is moderate; longer interventions might reveal more stable response components masked by 8-week noise.
Phlebotomy controlled Hbmass starting point but could not control all context variables.
The wide 95% CIs (e.g. ICC 0.19 [0.00, 0.37]) mean "poor reproducibility" is firm at the upper bound — true reproducibility could plausibly be even worse than the point estimate suggests.
Cycling-only stimulus — modality-specific findings; running, rowing, swimming may differ.
No DOI line is kept in this entry (PMID 42241643 identifies the paper); the abstract was checked via PubMed by PMID.
How this entry was checked and changed
Why this level
regraded 2026-09-27 from A: single-arm repeated pre-post design, n=42, highly standardised and supervised; the rubric has no A for a non-randomised design, and a reproducibility study is graded as a measurement study
Content correction 2026-09-27
Level A -> B+: single-arm repeated intervention, n=42, no randomised comparison; graded as a rigorous measurement/reproducibility study. Tier "N/A (preclinical/animal)" -> Tier 0-1 (untrained adults). Sport "cycling, running, swimming" -> cycling.
Full-text details (sex split, menopausal status, age range, withdrawals, trainer model, session structure, phlebotomy, GE/lactate/HAD/fibre-type outcomes) marked "full text; not checked against the abstract"; ICCs, compliance and n=42 verified against the PubMed abstract (fetched by the entry's PMID).
Added a lead Practical bullet with the verified numbers; Mechanism section marked as inference.
Cross-reference descriptions checked against the cited entries' current content; wrong years, wrong entries and claims the cited entry does not make were corrected.
2026-10-01: cross-reference to SYNTH-13 no longer names an archetype — that model was retired as an unvalidated composite.