Heavy lifting three times a week for 16 weeks raised knee-extensor strength 19% in 38 untrained men aged about 72 against no significant change (+1.4%) in 20 sedentary controls, and by a composite typical-error score 82% fell in the authors' 'Robust' or 'Excellent' responder classes and 5% (2 men) in 'Poor'; the extra spread beyond the controls' was estimated, but every 95% interval for it included zero
A secondary analysis of a randomised trial
Main caveat: A secondary analysis of a trial randomised for a drug question.
Soendenbroe and colleagues re-analysed a Copenhagen trial in which healthy, untrained men aged 64 and over were randomised to 16 weeks of supervised heavy lifting three times a week or to their usual sedentary life.
The original question was whether the blood-pressure drug losartan changes muscle adaptation; it did not, so the losartan and placebo training arms were pooled into one exercise group of 38 against 20 sedentary controls, all of whom took losartan.
Training raised isometric knee-extensor strength (MVC) by 19%, rate of force development (RFD) by 58%, quadriceps cross-sectional area by about 3% and type II fibre area by 14% at 16 weeks, with no group change in the controls.
To ask whether people truly differ in response, the authors compared the spread of change in the trained men with the spread in the controls (SDIR), and classified each trained man by a score averaged over up to four outcomes: +1 for a gain of at least one typical error (derived from the controls), −1 for a loss of at least one, 0 otherwise.
The trained group's spread exceeded the controls' for most outcomes, but every 95% interval for the SDIR included zero. Combining the four outcomes, 82% of trained men fell in the 'Robust' or 'Excellent' classes at 16 weeks and 2 men (5%) as poor; both of these improved strength and their 1RMs, and were classed poor because a muscle-size measure fell.
The authors avoid the word non-responder and conclude that heavy lifting should stay the first-line recommendation for older adults.
Participants
58 healthy men, age 72 ± 5 years (exercise 72 ± 5, range 65-83; control 72 ± 6, range 66-85); BMI 26-27 kg/m²; groups did not differ at baseline.
Normotensive, non-smoking, BMI 19-34, free of major disease, no blood-pressure or anticoagulant drugs.
Untrained: no structured strength training or regular strenuous exercise; walking or cycling for transport allowed.
Recruited 2015-2016 in the Copenhagen area; block-randomised on thigh lean mass, ACE genotype and age.
Protocol
Exercise: 48 scheduled sessions over 16 weeks, all supervised; leg extension, horizontal leg press and leg curl plus pulldown and machine shoulder press.
Leg press progression in six phases, from 3 × 12 at 15RM to 4 × 4-8 at 6-8RM; leg extension from 3 × 12 at 15RM to 4 × 6-8 at 8RM; leg curl similar with slightly fewer sets. Loads set from repeated 1RM tests (sessions 1, 7, 16, 25, 34, 43) and adjusted each session so sets ended at concentric failure.
Control: usual sedentary life, with losartan; participants were led to believe a fourth, placebo-only sedentary group existed.
Outcomes at baseline, 8 and 16 weeks: isometric MVC and peak RFD over the first 200 ms at 70° knee angle; quadriceps cross-sectional area by MRI at mid-femur; type I and type II fibre area from vastus lateralis biopsies (blinded analysis).
Analysis: SDIR = the square root of (exercise SD of change² − control SD of change²); typical error (TE) from the control group's SD of change, pooled over both intervals. Each outcome's change was positive (≥ +TE), negative (≤ −TE) or neutral; scores +1/0/−1 were averaged across MVC, RFD, quadriceps area and type II fibre area (type I left out for lack of a group change). Average 0 or below = Poor, 0.01-0.25 = Trivial, 0.26-0.74 = Robust, 0.75 or above = Excellent.
Key Results
Group changes at 16 weeks (mean ± SD; effect size with 95% CI):
| Outcome | Exercise (n = 38) | Control (n = 20) |
|---|---|---|
| MVC | +19.4 ± 14.0%, ES 0.83 (0.46 to 1.20) | +1.4 ± 7.9%, ES 0.08 (−0.38 to 0.53) |
| RFD | +58.4 ± 79.5%, ES 1.08 (0.68 to 1.48) | +8.1 ± 30.3%, ES 0.11 (−0.34 to 0.56) |
| Quadriceps area | +3.4 ± 3.6%, ES 0.18 (−0.14 to 0.51) | −0.9 ± 3.3%, ES −0.05 (−0.49 to 0.38) |
| Type I fibre area | +9.4 ± 19.2%, ES 0.43 (0.07 to 0.79) | −1.0 ± 19.0% |
| Type II fibre area | +14.2 ± 24.9%, ES 0.49 (0.12 to 0.85) | +0.6 ± 24.5% |
At 8 weeks MVC was already up 17.4 ± 12.6% and did not rise much further (19% at 16 weeks); type II fibre area rose significantly only by 16 weeks.The text calls the quadriceps-area gain significant although its effect-size interval includes zero; that need not conflict, since the effect sizes appear scaled by between-person spread, not by the spread of change (our arithmetic: the mean change, 2.10 cm², is close to its SD, 2.25 cm², in 38 men).It also says type I fibre area did not change at group level although its effect-size interval excludes zero (+385 ± 973 µm²); the paper does not explain this.
Compliance: 48 sessions (range 44-53), about 95%.
Spread of response beyond the controls (SDIR, 95% CI) and the typical error:
| Outcome | TE | SDIR 0-8 wk | SDIR 0-16 wk |
|---|---|---|---|
| MVC (Nm) | 10 | 14 (−5 to 21), above TE | 16 (−1 to 23), above TE |
| RFD (Nm/s) | 277 | 288 (−227 to 466), marked above TE, though the control SD of change was the larger (437 vs 329) | 109 (−293 to 331) |
| Quadriceps area (cm²) | 1.25 | 1.86 (−0.67 to 2.54), above TE | 0.96 (−1.61 to 2.10) |
| Type I fibre area (µm²) | 562 | 693 (−233 to 1008), above TE | 288 (−824 to 919) |
| Type II fibre area (µm²) | 582 | 377 (−597 to 800) | 217 (−788 to 846) |
At 16 weeks the control group's SD of change in type II fibre area was larger than the trained group's (894 vs 867 µm²); by the paper's method a larger control SD means the SDIR value is a reversed (negative) one, which the table prints without a sign.
How much the controls varied, outcome by outcome, at 16 weeks (share beyond the typical error):
| Outcome | Control positive | Control negative | Exercise positive | Exercise negative |
|---|---|---|---|---|
| MVC | 6 of 19 (32%) | 3 (16%) | 32 of 38 (84%) | not reported |
| RFD | 5 of 19 (26%) | 3 (16%) | 21 of 38 (55%) | 1 (3%) |
| Quadriceps area | 2 of 20 (10%) | 7 (35%) | 24 of 38 (63%) | 1 (3%) |
| Type II fibre area | 2 of 17 (12%) | 6 (35%) | 15 of 32 (47%) | 3 (9%) |
The composite responder classes were computed for the trained men only; how the controls would have been classed is not reported.
Composite classes, trained men: at 8 weeks Poor 3 (8%), Trivial 9 (24%), Robust 12 (32%), Excellent 14 (37%); at 16 weeks Poor 2 (5%), Trivial 5 (13%), Robust 16 (42%), Excellent 15 (39%). Poor plus Trivial fell from 32% to 18%.
Classed on muscle size alone, 32% would have been Poor; on strength alone 5%, but different men.
Both Poor responders (aged 69 and 72) gained MVC and 1RM; one had high baseline RFD and type II fibre area, the other showed a 5% loss of quadriceps area that the authors suspect was a baseline scan error.
1RM rose over sessions in all three exercises (p < 0.0001) with no interaction between session and responder class; a main effect of class appeared for the leg press only.
Lower baseline values went with larger % gains: MVC R² 0.29, RFD R² 0.40, type II fibre area R² 0.29; not quadriceps area. Poor and Trivial responders were spread across the whole baseline range.
Practical Application
An untrained older man who lifts heavy three times a week should expect to get stronger: 84% of the trained men cleared the test-noise threshold for strength by 16 weeks, and only 2 of 38 were classed poor overall, both of whom still gained strength. Measured in healthy untrained men aged 65-83, not in trained masters athletes or women.
Give it 16 weeks before judging it. The share of poor or trivial responders fell from 32% at 8 weeks to 18% at 16 weeks, mostly because muscle size catches up later than strength; early 1RM progress did not mark out who would end up a poor responder.
A single measure misleads. On muscle size alone a third of the men looked like poor responders; on strength alone 5% did, and they were different men. Judge a strength block on more than one outcome, and against the test's own noise (REF-1362).
How sure is "people truly differ"? Less sure than the authors' wording: at 16 weeks the spread beyond the controls exceeded the typical error only for strength (at 8 weeks for four outcomes, one of them a reversed value), every interval for it included zero, and a third of the sedentary controls cleared the strength threshold without training.This fits SYNTH-13's rule that much apparent responder variation is measurement noise; it does not show that true non-responders exist.
Inference, not tested here: for a masters endurance athlete who already trains, the size of these gains and the low share of poor responders may not transfer; the base's masters advice keeps heavy strength work in the week (SYNTH-35).
Limitations
A secondary analysis of a trial randomised for a drug question; the two training arms were pooled, and all 20 controls took losartan, so there was no placebo-only sedentary group (the authors and reviewers note this). Re-running the SDIR on losartan-only participants did not change the results, by the authors' report.
Modest size (38 vs 20); every SDIR interval includes zero, so the estimate of true between-person differences is imprecise; reviewers made the same point about the claim that non-responders are rare.
Responder classes rest on one arbitrary threshold (one typical error, pooled across both intervals); no smallest-worthwhile-change or other threshold was tested, and equal measurement error across groups and time was assumed, not checked.
The larger gains in men with low baseline values may partly be regression to the mean and ceiling effects, as the authors say.
Healthy, untrained men only; no women, no frail participants, no functional outcomes. 1RM was left out of the classification because of skill learning.
No trial registration or sample-size calculation is stated in the article. Funded by the Lundbeck Foundation and Nordea Fonden; the authors declare no competing interests.
How this entry was checked and changed
Why this level
read in full, but the claim rests on a post-hoc analysis of response spread in a trial randomised for a drug question, with about 19 per randomised arm and controls who all took losartan, and the spread estimates are imprecise (all SDIR intervals include zero)