REF-1713 · Liu et al., 2026 · BMC Sports Science, Medicine and Rehabilitation, · Paper · PubMed · Free full text
Core training in young team-sport athletes, against active controls without core work, gave a moderate gain in balance (SMD 0.67, 95% CI 0.30 to 1.04) and small gains in vertical jump (0.35), 10-40 m sprint time (−0.52) and change of direction (−0.40), in a meta-analysis of 22 small controlled trials with 588 athletes (median age 16.4) rated low to very low certainty by GRADE; between studies (exploratory subgroups, not head-to-head trials), balance gains were larger with more than two sessions a week, and no running-economy, cycling or endurance outcome was measured. Read in full
Main caveat: The authors': concerns about randomisation and selective reporting in many trials, which may inflate effects.
Liu and colleagues pooled controlled trials that compared core training with active control groups doing no extra core work, only in competitive team-sport athletes, and asked what core training does to balance, vertical jump, linear sprint and change-of-direction ability.
They searched four databases to November 2025 and included 22 studies with 588 players, mostly adolescent and male, mostly soccer. Core training was either a standalone programme or added to normal team practice, lasted 4-12 weeks at 2-7 sessions a week, and most programmes mixed static and dynamic exercises on stable surfaces.
Against controls, balance improved moderately (SMD 0.67, I² 66.6%), and vertical jump (0.35), sprint time (−0.52) and change of direction (−0.40) improved by small amounts with low heterogeneity.
Only balance was heterogeneous enough for subgroup analysis: more than two sessions a week went with a larger balance gain than two or fewer (0.97 against 0.20, P = 0.015), while age, programme length and standalone versus added-on structure did not differ significantly.
Risk of bias was mostly "some concerns" in the randomised trials and serious in the 4 non-randomised ones, and GRADE certainty was low to very low. The authors conclude that core training may improve these four measures in team-sport athletes, most for balance, and ask for larger, better-reported trials, especially in women and in sports other than soccer.
Participants
Pooled sample: 22 studies, 588 team-sport athletes; study sizes 12-40.
Sports: soccer 12 studies, basketball 4, handball 4, volleyball 1, field hockey 1.
Sex: 14 studies male only, 5 female only, 3 not reported, none clearly mixed (the Limitations section says 4 female-only studies; the Results and Table 2 list 5).
Age: study means 12.2 to 23.3 years, median 16.4.
Level: 19 studies Tier 2, 3 Tier 3, by the review's own classification against McKay 2022; healthy athletes only.
Season: 6 studies in season, 2 pre-season, 1 off-season, 13 not reported.
Protocol
Search: PubMed, the Cochrane Library, EBSCOhost and Web of Science Core Collection, each from inception to 30 November 2025; core-training terms combined with team-sport terms; English-language peer-reviewed studies. 445 records, 264 after duplicates, 60 full texts, 22 included.
Inclusion: randomised or non-randomised controlled trials; core training as the primary intervention for at least 4 weeks, alone or within team training; active controls without an added core component (regular sport training, mobility, strength training, small-sided games, recreational activity); inactive or untrained control groups excluded; at least one pre-post measure of balance, vertical jump, linear sprint or change of direction.
Interventions: 4-12 weeks, 2-7 sessions a week, session length reported as 14-60 min (Table 2 also lists a 6-min programme and an 8.3-20-min one, our reading); most used combined static and dynamic exercise, 2 static only by the text (Table 2 marks 3 as static); mostly stable surfaces, some unstable or mixed (Swiss ball, suspension, sling).
Outcomes: balance (mostly the Y-Balance and Star Excursion tests, also static balance tests); vertical jump (countermovement jump with or without arm swing, squat jump, jump height); linear sprint over 10-40 m; change of direction (T-test, Illinois, 5-10-5, 10 × 5 m shuttle, sport-specific tasks). One test per domain per study.
Analysis: Hedges' g on between-group differences in change scores, a pre-post correlation of 0.50 assumed when not reported; DerSimonian-Laird random effects; pooled only with at least 3 studies; leave-one-out for balance; a sensitivity analysis dropping 4 studies with atypical comparators; subgroups for balance only (frequency ≤2 vs >2 a week, age ≤16.4 vs >16.4 years, duration ≤8 vs >8 weeks, standalone vs added to sport training), treated as exploratory; Egger's test only for outcomes with 10 or more studies.
Risk of bias and certainty: RoB 2 and ROBINS-I by two reviewers with a third to settle; GRADE by the same two.
Funding and interests: no external funding; the authors declare no competing interests.
Key Results
| Outcome (core vs non-core control) | Studies (n) | SMD (95% CI) | I² | GRADE |
|---|---|---|---|---|
| Balance | 15 (392) | 0.67 (0.30 to 1.04), moderate | 66.6% (Q = 41.95, P < 0.001) | very low |
| Vertical jump | 8 (217) | 0.35 (0.07 to 0.62), small | 0.0% (P = 0.543) | low |
| Linear sprint time (10-40 m) | 10 (272) | −0.52 (−0.79 to −0.25), small; negative = faster | 15.2% (P = 0.303) | low |
| Change-of-direction time | 10 (280) | −0.40 (−0.66 to −0.14), small; negative = faster | 13.9% (P = 0.315) | low |
Balance subgroups (exploratory, between studies):
| Subgroup | SMD (95% CI) | P for difference |
|---|---|---|
| More than 2 sessions a week | 0.97 (0.45 to 1.49), I² 72.9% | 0.015 |
| 2 or fewer sessions a week | 0.20 (−0.14 to 0.54), I² 0.0%, not significant | |
| Age ≤16.4 vs >16.4 years | larger in older players, not significant (numbers in a figure only) | 0.175 |
| ≤8 vs >8 weeks | larger with longer programmes, not significant | 0.316 |
| Standalone vs added to sport training | larger when added to sport training, not significant | 0.211 |
Risk of bias: most randomised trials "some concerns" overall, mainly the randomisation process and selection of the reported result; deviations, missing data and outcome measurement mostly low risk; all included studies raised some concerns about selective reporting (the authors). The 4 non-randomised trials: serious overall, mainly confounding from group allocation.
GRADE downgrades: risk of bias and imprecision (small pooled samples) for all four outcomes; inconsistency also for balance.
Sensitivity: dropping the 4 studies with atypical comparators (leaving 11, 6, 9 and 7 studies) gave results "generally consistent" with the main analyses; the balance leave-one-out is described without numbers.
Publication bias: Egger's test P = 0.360 (balance), 0.118 (sprint), 0.635 (change of direction); not assessed for vertical jump (fewer than 10 studies).
Practical Application
For young team-sport players, core work buys a moderate balance gain and small gains in short sprints, change of direction and jumping, at low to very low certainty. Across 22 small controlled trials, against teammates doing their usual training without core work, balance rose by SMD 0.67, sprint and change-of-direction times fell by 0.52 and 0.40, and vertical jump rose by 0.35. Nothing in this review tests whether core work makes a runner, cyclist or triathlete faster.
Against [REF-1632] (Saeterbakken 2022, A-, 31 trials in athletes of many sports, active controls): the direction agrees and the sizes are close or smaller — sprint −0.52 here against 0.66 there, change of direction −0.40 against 0.70 (there significant only in children), lower-limb power 0.30 there and vertical jump 0.35 here.This review adds balance as a pooled outcome, GRADE certainty (low to very low) that REF-1632's abstract did not give, and a team-sport-only sample; how many trials the two share is not reported.
Against [SYNTH-83]'s core row: that row says a speed gain from core work is not shown for the bike or the run, and REF-1018 pooled balance (ES 1.17) and jump (0.69) in healthy people over 12.This review keeps the endurance answer unchanged — no economy, time-trial or endurance outcome was measured — but it does show small, low-certainty gains in 10-40 m sprint and change-of-direction times in team-sport players, so "never promise speed" holds for endurance speed, while short-sprint speed in team sports has some low-certainty support.The balance and jump gains here are smaller than REF-1018's.
Frequency: more than two sessions a week went with a larger balance gain (0.97 against 0.20, the latter not significant). This is a between-study split, and the over-two group includes programmes of 6-7 sessions a week of 40-60 min, so it does not set a dose.
Not from this review; the base's own practice: for an endurance athlete who wants trunk control or balance, core work after heavy lifting is kept as an optional add-on (SYNTH-83); this review neither tests nor supports that practice. In many trials core work was added on top of the same team training the controls did, and matching of total training time is not reported, so part of the gain may be extra training time rather than core work as such.
Limitations
The authors': concerns about randomisation and selective reporting in many trials, which may inflate effects; mostly male samples, so the effects mainly describe men; protocols varied widely in type, frequency, length and setting, and session length or adherence was often not reported, so no dose can be named; comparators varied (sport training, mobility, strength training, small-sided games, recreation); training season unreported in 13 of 22 studies; different tests within each domain (dynamic and static balance, sprints of 10-40 m); short interventions with no follow-up; too few trials for moderators such as competition level; small trials, so the low I² for jump, sprint and change of direction is not proof of consistency.
Certainty: low (jump, sprint, change of direction) to very low (balance) by GRADE; 4 of 22 trials non-randomised with serious risk of bias.
Design (our reading): in many trials the core group did core work in addition to the same team training as the controls, so added training time and core content are not separated; subgroup results are between-study comparisons, run for balance only.
Reporting: the female-only count is given as 5 in the Results and 4 in the Limitations; the stated session range (14-60 min) leaves out the 6-min and 8.3-20-min programmes listed in Table 2; the risk-of-bias figures and the subgroup forest plots were not available as text, so per-study judgements and the age, duration and structure subgroup estimates are not reported here.
Transfer: adolescent team-sport players (median age 16.4), mostly soccer; no adult endurance athletes, no runners or cyclists, no endurance, economy or injury outcome.
Revision 2026-11-10 (headline): the headline now says that the frequency finding is an exploratory comparison between studies, not head-to-head trials.
How this entry was checked and changed
Why this level
a meta-analysis of controlled trials whose risk of bias and GRADE certainty we read in the full text; held at A- by the rubric's cap for serious certainty problems: most randomised trials carry some concerns, the 4 non-randomised ones serious risk, GRADE certainty is low for jump, sprint and change of direction and very low for balance, and the trials are small (12-40 participants each)