Individual response to exercise training
A statistical perspective on what valid responder classification requires
Key Findings (the authors' argument)
"Individual response" = subject-by-training interaction. Standard approaches for quantifying and predicting it are not established, because interactive effects at the individual level need different handling than the usual main effects
Measurement error limits the accuracy of individual response estimates and itself contributes to the apparent variation. On the individual level this is critical: a low signal-to-noise ratio cannot be compensated by a larger sample — the individual is a case of one
Within-subject variation in training efficacy (the same person responding differently to the same program at different times) is a largely unstudied source of variability; it is not disclosed by comparison with a control group and calls for repeated interventions
Prediction of response: training response has many determinants, and indications of interaction between them suggest an interactive rather than additive predictive approach
The statistically rigorous requirements are beset with practical difficulties, so pragmatic alternatives are warranted
Hecksteden 2018 (not checked against an abstract — no identifier on file)
A follow-up in Sports Medicine on repeated testing for responder classification is cited by this base; its specific figures previously given here (≥3 baseline sessions, ~50% less classification error, non-responder rate inflated 30-50%) are not checked against that paper and are not quoted.
The resistance-training conference summary REF-1143 cites Hecksteden's 2018 repeated-testing paper as Hecksteden, Pitsch, Rosenberger & Meyer, J Appl Physiol 124:1567-1579, 2018, not Sports Medicine; its reference list gives DOI 10.1152/japplphysiol.00896.2017; the topic REF-441 gives its Sports Med citation matches this JAP paper's title, so a mis-citation is likely, but a separate Sports Med paper is not excluded (not checked offline).
Key Concept: Smallest Worthwhile Change (SWC)
Not from this paper; the base's own practice: a change smaller than the measurement's typical error (or than a pre-defined SWC) should not be read as a response; changes above it are potentially meaningful
The paper's own point is broader: even a change above the noise band in one block may not recur in the next, because training efficacy varies within a person
Practical Framework (the base's own practice, built on the paper's argument)
For a defensible responder classification you need:
1
Repeated baseline measurements to characterize the individual's measurement noise
2
A control group or repeatability data to estimate non-training variability (REF-430) — and, per this paper, repeated interventions to see whether the response itself is stable
3
A threshold tied to measurement error, not an arbitrary zero
4
Confidence intervals on the individual change, not point estimates
Implications for Phenotyping
Many "non-responder" claims in the literature are statistical artifacts of measurement error and single pre-post designs
Real-world coaches over-classify on insufficient data
Phenotyping requires longitudinal data — repeated tests AND repeated training blocks
For an individual athlete: repeated lab tests plus multi-year power data are far more reliable than a single lab test (the base's own practice)
Practical Application
Whether an athlete is a "responder" cannot be settled from one before/after test: measurement error blurs the individual estimate and cannot be averaged away in a single person, and the same person can respond differently to the same program on another occasion — so classification needs repeated tests and, ideally, repeated blocks — treat a one-off flat result as a question, not a label
A control-group comparison (REF-430) tells you whether true individual differences exist in a group; it does not tell you whether a given athlete's response is stable — only repeating the intervention does
Practical compromise (the paper itself calls for pragmatic alternatives): track the same benchmark across several blocks and judge the pattern, not the single change
The leading resistance-training groups now propose the same framework (REF-1143, B-, 2026 conference summary with Hecksteden as co-author): repeat validation trials as the theoretical standard (washout and muscle-memory carry-over are its weak points), repeated testing within one programme (baseline, 3, 6 and 9 months) as a practical stand-in, one-leg-trained within-subject designs (with the caveat that the untrained leg gains 10-15% strength by cross-education), MCID thresholds built on repeated pre-tests, and composite responder scores rather than one outcome.Their proposed starting thresholds for muscle — 0.5-1.0 kg of DXA appendicular lean mass over 12-16 weeks, 2-3% in ultrasound muscle thickness — are starting points, not norms.
Limitations
Methodological perspective, not empirical data or therapeutic guidance
Specific recommendations depend on measurement type
Practical limitations: athletes/coaches rarely have time/resources for repeated baselines and repeated interventions
Doesn't fully resolve the responder vs measurement-noise distinction
How this entry was checked and changed
Why this level
regraded 2026-09-27 from A: methodological perspective paper (no empirical cohort) from a leading group on individual response
Content correction 2026-09-27
"2015 study compared classification methods; only 11/20 subjects consistently classified" -> the 2015 paper is a statistical perspective with no cohort; this claim is not in the abstract and was removed.
2018 figures ("≥3 sessions", "~50% less error", "inflate 30-50%") and "SWC = 0.5 × CV", "≥2 lab tests" -> not verified (no identifier / not from this paper); removed or relabelled as the base's own practice.
Added the abstract's actual points: n-of-1 signal-to-noise cannot be fixed by sample size; within-subject variation in training efficacy requires repeated interventions, not just a control group; interactive prediction; pragmatic alternatives.
Level A -> B+ (methodological perspective); added Tier/Sport lines; cross-ref to REF-429 corrected (HERITAGE did not apply a control comparison).
2026-09-28: added Journal line for the DOI paper (Hecksteden 2015) from Europe PMC (PMID 25663672).
2026-10-01: cross-reference to SYNTH-13 no longer names an archetype — that model was retired as an unvalidated composite.
Revision 2026-10-16: added REF-1143 (Roberts 2026, B-, conference summary, full text): resistance training adopts this framework (repeat validation trials, repeated testing within a programme, unilateral within-subject designs, MCID, composite scores) — Practical bullet and cross-reference; noted that REF-1143 cites the 2018 repeated-testing paper as J Appl Physiol 124:1567-1579, not Sports Med; that DOI resolves on Crossref (Hecksteden, Pitsch, Rosenberger, Meyer), so a mis-citation in REF-441 is likely but a separate Sports Med paper is not excluded.Level unchanged.