REF-1135 · Buendía-Romero et al., 2025 · Scandinavian Journal of Medicine & Science in Sports · Paper · PubMed · Free full text
I'm over 60 — if I stop training for a few months, is the work wasted?
No, but it is not banked either: older adults who had trained still did better than never-trained peers after 4-36 weeks off (ES 0.88), less so the older they were (β −0.06 per year), and the comparison does not show how much of the gain was lost. The 15 trials were in general older adults aged 64-92, not athletes; the bigger the gain made, the more remained.
Buendía-Romero et al. pooled randomised trials in adults aged 60 and over that trained for a period, then stopped, and compared the change from before training to the end of the stop against a control group that never trained.
Fifteen RCTs (787 participants, 553 women; mean ages 64-92) gave 21 intervention arms: training lasted 8-43 weeks (mean 16), usually 2 sessions a week, and the stop lasted 4-36 weeks (mean 11; one trial followed up after 240 weeks).
Trained groups kept a large advantage in functional capacity after the stop (ES 0.88, 95% CI 0.47 to 1.29, I² 81%), significant for agility, walking ability, standing ability and stair walking but not significant for gait speed (ES 0.38, p = 0.056) or balance (ES 0.48, p = 0.051). Resistance and multicomponent training, high and low intensity all left a significant advantage.
In meta-regression the size of the gain made during training predicted the advantage left (β 0.66 per unit of training ES), older age reduced it (β −0.06 per year), and neither the length of the stop nor modality, intensity or training duration mattered.
The comparison is against people who never trained, so the review shows that a training advantage survives the break, not how much of the gain was lost.
Included studies
Search: Cochrane Library, PubMed/MEDLINE, Scopus, Web of Science to March 2024, plus reference and citation screening; 3383 records, 43 full texts, 15 RCTs in the qualitative and 14 in the quantitative analysis. Protocol registered (PROSPERO CRD42021235092).
Population: people 60 and over, any health status; seven trials included chronic disease or special conditions (type 2 diabetes, prefrailty, institutionalisation, VO2max below 20 mL/kg/min, dementia). 79.4% European, 11.8% American, 8.8% Asian.
Interventions: 11 multicomponent, 9 resistance, 1 aerobic (brisk walking) arm; 7 high-intensity, 12 low-intensity, 2 uncontrolled intensity. Intensity cut-offs: 77% HRmax, 60% HRR or RPE 15/20 for aerobic work; 69% 1RM or RPE 13/20 (7/10) for resistance.
Cessation: participants told to avoid regular exercise and carry on normal daily activities; no injuries, illnesses, surgeries or restraints reported — this is a planned stop, not hospitalisation.
Analysis: Hedges' g on change scores (pre-training to end of cessation) vs controls, robust variance estimation with small-sample correction; subgroups only with at least 5 interventions; RoB 2, GRADE, Egger's test.
Key Results
| Outcome (Table 2 and Table 3) | k | n | ES or β | 95% CI | p | I² |
|---|---|---|---|---|---|---|
| Overall functional capacity (results text) | — | — | 0.88 | 0.47 to 1.29 | < 0.01 | 81% |
| Multicomponent training | 11 | 455 | 0.89 | 0.33 to 1.46 | < 0.001 | 83.6% |
| Resistance training | 9 | 233 | 0.84 | 0.02 to 1.66 | 0.045 | 81.7% |
| High intensity | 7 | 325 | 0.67 | 0.10 to 1.23 | 0.027 | 73.4% |
| Low intensity | 12 | 329 | 1.06 | 0.37 to 1.75 | 0.006 | 84.0% |
| Agility | 11 | 328 | 0.61 | 0.13 to 1.09 | 0.018 | 68.8% |
| Balance | 8 | 237 | 0.48 | −0.01 to 0.97 | 0.051 | 55.0% |
| Gait speed | 8 | 350 | 0.38 | −0.01 to 0.77 | 0.056 | 43.0% |
| Stair walking | 6 | 181 | 1.26 | 0.09 to 2.42 | 0.039 | 85.0% |
| Walking ability | 5 | 120 | 0.88 | 0.01 to 1.75 | 0.049 | 0.0% |
| Standing ability | 16 | 562 | 1.35 | 0.77 to 1.94 | < 0.001 | 84.8% |
| Meta-regression: age (per year) | — | — | β −0.06 | −0.10 to −0.02 | 0.006 | — |
| Meta-regression: training effect (per unit ES) | — | — | β 0.66 | 0.32 to 1.01 | 0.003 | — |
| Meta-regression: detraining duration (per week) | — | — | β −0.01 | −0.03 to 0.02 | 0.381 | — |
| Meta-regression: training duration (per week) | — | — | β −0.02 | −0.09 to 0.05 | 0.408 | — |
| Meta-regression: intensity (low vs high) | — | — | β 0.29 | −0.53 to 1.10 | 0.461 | — |
The abstract reports ES 0.87, β 0.73 and β −0.07; the results section and Table 3 report 0.88, 0.66 and −0.06. The table values are used here.
Risk of bias: 11 trials "some concerns", 4 "low risk"; GRADE very low to moderate. Adherence was reported by 11 of 21 programmes (mean 77%; low intensity 83%, high intensity 71%).
Mechanisms
Measured: the larger the improvement a programme produced, the larger the advantage left after the stop (β 0.66); older samples kept less (β −0.06 per year of mean age).
Authors' interpretation: people who end training fitter decline less in absolute terms; and older participants (by previous studies, about 74 and over) may keep little residual effect on walking, so stops should be minimised in the oldest.
Authors' interpretation: intensity did not moderate the effect possibly because 57% of programmes set intensity by self-rated RPE, which less experienced people under-report.
Inference: the advantage after cessation is a starting-point effect — the trained group starts the stop higher — so a residual effect over controls is compatible with losing much of the gain; the review does not report the fraction retained.
Practical Application
I'm over 60 — if I stop training for a few months, is the work wasted? No, but it is not banked either: older adults who had trained still did better than never-trained peers after 4-36 weeks off (ES 0.88), less so the older they were (β −0.06 per year), and the comparison does not show how much of the gain was lost. The 15 trials were in general older adults aged 64-92, not athletes; the bigger the gain made, the more remained.
Build as much as you can before a known break. The training gain predicted what was left after the stop (β 0.66 per unit of training ES; between programmes, not within people); the review cannot say whether modality or intensity matter, since neither moderated it.
In the oldest groups, keep breaks short rather than relying on the reserve. Age reduced the residual effect and the authors advise minimising cessation in the oldest; the length of the stop (4-36 weeks) did not reach significance, which with this heterogeneity is absence of evidence, not evidence of no loss (general older adults, many frail; whether this holds for masters athletes is untested).
Limitations
Population: general older adults, many frail, institutionalised or with chronic disease; mean ages 64-92; 70% women. Not athletes and not masters competitors — the age slope here is across cohorts, not within people, and mixes age with frailty and setting (institutionalised groups showed the smallest effects).
Comparator: never-trained controls. The "residual effect" is the difference in change from pre-training, so it cannot tell a retained gain from a large gain partly lost.
Heterogeneity and certainty: I² 81% overall and above 50% in most subgroups, GRADE very low to moderate, 11 of 15 trials with "some concerns"; the funnel was asymmetric (Egger p = 0.004), which the authors attribute to heterogeneity rather than publication bias — that cannot be shown from the data.
Small trials, estimated data: arms of 7-87 people; some results digitised from figures and some SDs estimated; intensity coded into two bins, often from RPE.
Reporting: the abstract and the results disagree (ES 0.87 vs 0.88; β 0.73 vs 0.66; β −0.07 vs −0.06), and the text calls ES 0.88 "medium" although the authors' own scale rates 0.80 and above as large.
Funding: Spanish public research grants (CIBERFES, Ministry of Science and Innovation, University of Granada); no conflicts declared.
How this entry was checked and changed
Why this level
PROSPERO-registered, RoB 2 and GRADE applied, robust-variance random-effects model; but heterogeneity is high (I² 81% overall), GRADE certainty very low to moderate, 11 of 15 trials have "some concerns", the funnel is asymmetric (Egger p = 0.004), and the trials are small (control arms of 7-83); read from the full text
Full text (checked 2026-10-15)
"15 RCTs (Table 1) involving 787 older adults (553 women) with a mean age ranging from 64 to 92 years old at baseline" (full text).
"The mean duration of the exercise programs was 16 weeks (range 8–43 weeks)"; "The exercise cessation period varied between 4 and 36 weeks, with a mean duration of 11 weeks"; "One study had a follow‐up period of 240 weeks" (full text).
Overall "ES = 0.88 [CI: 0.47 to 1.29], p < 0.01" with heterogeneity 81% (full text).
Abstract: "(ES = 0.87; p < 0.01)", "(β = 0.73; p < 0.01)", "(β = −0.07; p < 0.01)" (full text).
Results: training effect "(β = 0.66; p < 0.01)"; age "(β = −0.06; p < 0.01)" (full text).
Table 3: "Sample age (years) | −0.06 | −0.10 to −0.02 | 0.006*"; "Training effect (ES) | 0.66 | 0.32 to 1.01 | 0.003*"; "Detraining duration (weeks) | −0.01 | −0.03 to 0.02 | 0.381" (full text).
Table 2: "Multicomponent training | 11 | 455 | 0.89 | 0.33 to 1.46"; "Resistance training | 9 | 233 | 0.84 | 0.02 to 1.66 | 0.045*"; "High intensity | 7 | 325 | 0.67 | 0.10 to 1.23 | 0.027*"; "Low intensity | 12 | 329 | 1.06 | 0.37 to 1.75 | 0.006*" (full text).
Table 2: "Gait speed | 8 | 350 | 0.38 | −0.01 to 0.77 | 0.056 | 43.0"; "Balance | 8 | 237 | 0.48 | −0.01 to 0.97 | 0.051 | 55.0"; "Standing ability | 16 | 562 | 1.35 | 0.77 to 1.94" (full text).
"Eleven studies presented “some concerns” ... and four studies “low risk.”"; Egger's test "(p = 0.004)" (full text).
"only 11 out of 21 training programs (52%) reported adherence rates, with an average of 77%"; "low‐intensity training reported a higher adherence rate (83%) compared to high‐intensity training (71%)" (full text).
"a significant portion of the reviewed interventions (57%) relied on self‐rated perceived exertion scales" (full text).
Protocol "PROSPERO ... (CRD42021235092)"; "The authors declare no conflicts of interest" (full text).