The decision this page serves: an older adult who has been training will, at some point, stop — illness, a fall, a hospitalization, a holiday, lost motivation. How much of the function they built survives the layoff, for how long, and does it matter what or how they trained? This is the retention / legacy half of the exercise-dose question; Resistance Training Prescription - Load Sets and Frequency and Power Training and Physical Function in Older Adults cover the acquisition half. (inferred from Buendía‐Romero et al., 2025)
Scope of the evidence — a PLANNED stop, not a forced one. Every included trial studied a planned cessation of structured training while ordinary life continued: «During this period, all par- ticipants were instructed to avoid any type of regular exercise and to carry on with their normal daily activities.» «No injuries, illnesses, surgeries, or physical restraints were reported in the included studies.» (Buendía‐Romero et al., 2025) A layoff forced by illness, a fall or a hospital stay adds bed rest, inflammation and catabolism — a harsher exposure the source has no data on, so the findings below do not transport to it; that is a named gap. (The MA’s mentions of falls and hospitalization are motivating rationale, not studied exposures.)
The headline: function does not evaporate when training stops — a medium residual advantage persists
- After a completed program (resistance or multicomponent), older adults retained a medium, significant functional-capacity advantage over non-exercising controls at the end of the cessation window: ES = 0.88 [95% CI 0.47 to 1.29], p < 0.01, I2 = 81% (15 RCTs, 21 arms, 787 adults >= 60 y; programs 8-43 wk, layoffs 4-36 wk, mean 11 wk, plus one 240-wk follow-up). (Buendía‐Romero et al., 2025)
- Plainly: «engaging in physical training twice a week for over 2 months before a hiatus of at least 1 month can significantly preserve functional capacity, including agility, walking ability, standing ability, and stair walking». (Buendía‐Romero et al., 2025)
- Decision-change: the common use it or lose it, so a layoff wipes the slate intuition is too strong for this stratum and window. A trained older adult who makes a planned stop from structured training (usual daily activity continuing) starts the re-entry from a higher functional floor than an untrained peer — the gains have inertia. Whether this holds after an illness- or hospitalization-forced layoff is untested here (see Scope above).
What the effect size IS — and the trap of reading it as a retention percentage
- The ES is a between-group standardized mean difference of the change from before training to after the layoff: «The effect sizes (ES) were calculated as the standardized mean differences between the intervention and control groups’ change scores.» (Buendía‐Romero et al., 2025)
- So ES = 0.88 means the trained group ended the layoff ~0.9 SD better than control, measured from baseline — it fuses how much was built with how much was kept. It is not 88% of gains retained, and the MA reports no retained-fraction or decay-rate at all. (inferred from Buendía‐Romero et al., 2025) — reading the stated estimand against what a maintenance decision needs.
- This matters because the moderator that most raises the residual is simply how much was gained in the first place (next section): part of a better retention signal is really trained harder, so more bank left after the same decay.
The two things that move the residual (study-level meta-regression — route-b hypotheses, not positive interaction evidence)
Only two of six meta-regression moderators reached significance:
- Bigger program gains -> bigger residual (beta = 0.66 [0.32 to 1.01], p = 0.003). «individuals with higher fitness levels may experience less functional decline compared to those with lower physical condition». (Buendía‐Romero et al., 2025) This is study-level effect-modification evidence, and partly definitional (more banked = more left) — treat it as “build a real adaptation and more survives,” not a separate retention mechanism. (inferred from Buendía‐Romero et al., 2025)
- Older age -> smaller residual (beta = -0.06 [-0.10 to -0.02], p = 0.006). Buendia reads this against prior work in which «the oldest par-ticipants (i.e., >= 74 years) reported no residual effect on walking ability, while younger participants (i.e., <= 73 years) showed long-lasting benefits». (Buendía‐Romero et al., 2025) (The >=74/<=73 split is cited prior literature, not Buendia’s own estimate.) The moderator is the arm’s mean sample age across 21 arms — an ecological, between-study association, not a within-trial interaction showing that older individuals lose more — and it is confounded with frailty and setting: «In seven studies, participants with chronic diseases such as type 2 diabetes or special physical conditions like prefrailty, insti- tutionalization, or a VO2max < 20 mL/kg/min were included [43, 44, 47, 51, 53, 55, 56].» (Buendía‐Romero et al., 2025) Decision-change: in the oldest-old, layoffs plausibly cost more and are plausibly worth shortening or bridging — a hypothesis for a route-(b) stratifier on age, not positive interaction evidence.
No modifier detected: modality and intensity (wide CIs — uninformative nulls, route-b)
- No modality difference detected (beta = -0.05, 95% CI -0.97 to 0.87, p = 0.660): multicomponent (ES 0.89 [0.33-1.46]) and resistance (0.84 [0.02-1.66]) both retained a significant residual. (Buendía‐Romero et al., 2025) The CI on the difference is wider than the pooled effect itself (0.88), so this is absence of evidence for a difference, not demonstrated equivalence — the same standard the duration null gets below.
- No intensity difference detected (beta = 0.29, 95% CI -0.53 to 1.10, p = 0.461): high (0.67 [0.10-1.23]) and low (1.06 [0.37-1.75]) both retained. (Buendía‐Romero et al., 2025) But 57% of trials set intensity by self-rated RPE, so the intensity null may be a measurement artifact, not a true absence of modification. (Buendía‐Romero et al., 2025)
- Decision-change: both modalities and both intensity bands retained a significant residual; the data cannot rank them, so for the retention goal there is no evidenced reason to prefer one effective program over another. This parallels the acquisition side, where Resistance Training Prescription - Load Sets and Frequency also finds route-b mostly quiet (the relative ranking of prescriptions does not shift by stratum).
What this source cannot tell you — the decay time-course and the maintenance floor (G-gaps)
- The decay curve could not be estimated. «due to substantial heterogeneity among interventions, a dose-response relationship could not be identified.» (Buendía‐Romero et al., 2025) Detraining duration was not a significant moderator (beta = -0.01, p = 0.381) over the studied 4-36-week range — but that is no gradient demonstrated, not a demonstrated plateau: a route-(b) null under low power and I2 ~ 81%, i.e. absence of evidence for a time-course, not evidence of a flat one. Do not read it as function holds indefinitely. This is The Underivable Optimum on the time axis — cessation windows averaged 11 weeks, so the data thin out beyond ~3 months, and any apparent no-decay is the sampling edge, not a curve feature. (inferred from Buendía‐Romero et al., 2025)
- This MA is about COMPLETE cessation, not a reduced maintenance dose. The cessation period was defined as a follow-up with no active voluntary physical activity, so the source answers how much persists when you stop entirely, NOT what is the smallest dose that holds the gain. The minimum maintenance dose for older-adult function is a named gap — no held source estimates it. (inferred from Buendía‐Romero et al., 2025) — a maintenance-dose (reduced-frequency retention) SR for strength/function in older adults.
- Balance and gait speed retention is unproven (p = 0.051 and 0.056) — the two subdomains closest to fall risk are exactly the ones the residual effect did not reach significance on. (Buendía‐Romero et al., 2025)
Decision relevance
- A planned layoff is not a reset. Reassure and plan from a higher floor: a trained older adult who stops structured training (usual daily activity continuing) keeps a meaningful functional advantage through a stop within the studied 4-36-week cessation range (mean ~11 weeks), so re-entry is a resumption, not a restart. Hold the duration loosely — most trials sit at the short end (mean 11 wk); one study (Englund 2009) followed to 240 weeks, on balance and gait speed; beyond ~3 months the evidence is sparse. Not evidenced for an illness-, fall- or hospitalization-forced layoff. (Buendía‐Romero et al., 2025) (corrected 2026-10-06: dense at 4-36 wk, silent beyond -> short-end-concentrated, one 240-wk study; self-critique)
- Consider protecting the oldest-old from layoffs. Mean sample age attenuated the residual across studies (ecological, frailty-confounded), so for the >= ~75 stratum it is plausibly worth keeping cessation windows short or bridging them with any movement — a precautionary hedge, not an evidenced age interaction. (inferred from Buendía‐Romero et al., 2025)
- Build a real adaptation while training — the bigger the gain, the more survives the layoff; a minimal-dose program that barely moves function also banks little to retain.
- Do not over-read the number. ES = 0.88 is heterogeneous (I2 = 81%), GRADE very-low-to-moderate, and is a net-change SMD, not a retention fraction. It supports gains persist, not any specific you can stop for N weeks rule, and says nothing about a reduced maintenance dose.
Limits
- Single gold MA;
confidence: low. Substantial heterogeneity (I2 ~ 81% overall), wide CIs, GRADE very low to moderate, Egger asymmetry (attributed to heterogeneity). Intensity set by RPE in 57% of trials. Function is a task-test surrogate one step from daily-life activity and from falls (never an endpoint here) — see Surrogate Outcomes. The provisional cluster-opening (question/structure induction) is single-source and expected to stay so until a maintenance-dose or second detraining SR lands.
References
Buendía‐Romero, Á., Vetrovsky, T., Hernández‐Belmonte, A., Izquierdo, M., & Courel‐Ibáñez, J. (2025). Residual Effects of Physical Exercise After Periods of Training Cessation in Older Adults: A Systematic Review With Meta‐Analysis and Meta‐Regression. Scandinavian Journal of Medicine & Science in Sports, 35(1). https://doi.org/10.1111/sms.70010