REF-1613 · Barth et al., 2022 · Sports Medicine · Paper · PubMed · Free full text

Adult world-class athletes had done more coached practice in other sports as children and teenagers, started their main sport later, accumulated less main-sport practice over their careers and progressed more slowly than national-class peers, the reverse of what went with junior success; peer-led play had only negligible links with either, in a meta-analysis of 71 reports (9241 athletes, GRADE moderate to high for these predictors) built on retrospective sport histories. Read in full

Main caveat: Retrospective, self-reported sport histories; confounding and survivor bias cannot be excluded (the authors).
fairly strong evidenceMeasured on highly trained to world-class athletesRead in full

Barth and colleagues pooled studies that compared higher- and lower-performing athletes within an age category, sport and sex on six participation variables: age at performance milestones, age at starting the main sport, and accumulated coach-led practice and peer-led play in the main sport and in other sports.

They kept 262 athlete samples and 685 effect sizes from 71 reports and excluded comparisons between extreme levels (for example Olympic medallists against local competitors), which produced inflated effects, and 10 outliers.

Among adults, world-class athletes differed from national-class athletes by more childhood and adolescent coached practice in other sports, a later start, less main-sport practice and later milestones. Among juniors, the better performers showed the opposite: earlier start, more main-sport practice, less other-sports practice and earlier milestones.

Peer-led play, in the main sport or others, had negligible effects. The patterns were similar across types of sport, and the predictors' effects were correlated with one another.

The authors argue that heavy specialised practice in youth may cost long-term development through overuse injury, burnout, a poorer fit between athlete and sport, and less learning capital, and frame the world-class pattern as resource-preserving and risk-buffering.

Participants

9241 athletes in 262 samples (221 independent, 41 partly dependent): 5690 juniors and 3551 seniors; world class 1003 (191 junior, 812 senior), national class 4818 (3322 junior, 1496 senior), regional and below the rest. Including 219 Olympic, world or continental champions and 474 medallists. 67% men, 33% women; Europe 78%, North America 9%, Asia 6%, Oceania 5%. Sport histories by interview (761) or questionnaire (8115).

Protocol

Search: SPORTDiscus, ERIC, ProQuest, PsycINFO, PubMed, Scopus, WorldCat and Google Scholar to 28 February 2021, added to searches run to 2014 and 2019; authors of 19 reports contacted, all replied.
Inclusion: original studies comparing higher- and lower-performing athletes within defined age categories, sports and sex, with effect sizes or data to compute them; 157 of 685 effect sizes from unpublished reports.
Effect size: meta-analytic mean Cohen's d; a positive d means higher performance went with a later age or a larger amount of the activity.
Analysis: random-effects meta-analyses; moderators junior vs senior, type of sport and, for seniors, world vs national class; GRADE per predictor.

Key Results

Predictor (d, 95% CI)Senior world class vs national classAll seniorsJuniors
Age at milestones0.420 (0.22 to 0.62)0.362 (0.15 to 0.58)−0.493 (−0.66 to −0.32)
Starting age, main sport0.409 (0.19 to 0.63)0.278 (0.15 to 0.41)−0.330 (−0.51 to −0.15)
Main-sport coach-led practice−0.232 (−0.38 to −0.08)0.200 (0.03 to 0.38)0.610 (0.40 to 0.82)
Main-sport peer-led play−0.033 (−0.30 to 0.23)0.167 (−0.13 to 0.47)0.238 (−0.00 to 0.48)
Other-sports coach-led practice0.497 (0.37 to 0.63)0.465 (0.35 to 0.58)−0.232 (−0.35 to −0.12)
Other-sports peer-led play0.114 (−0.02 to 0.25)0.130 (0.01 to 0.25)−0.121 (−0.22 to −0.02)
Early amounts (to age 15): in seniors, early other-sports practice d = 0.505 (0.36 to 0.65) and early main-sport practice d = −0.099 (−0.26 to 0.06), the latter not predictive.
Main-sport practice by level: world-class seniors had less than national-class ones, while in lower-level senior comparisons the better athletes had more (d = 0.474, 0.27 to 0.68); the authors describe the relation as rising up to national level and reversing beyond it.
Type of sport: no type showed a contrary pattern; the one significant difference was starting age among world- vs national-class seniors, d = 0.65 (0.32 to 0.99) in cgs sports against 0.00 (−0.43 to 0.43) in game sports (p = 0.028).
Linked predictors: effects of milestones, starting age, main-sport and other-sports practice correlated with one another (|r| 0.64 to 0.79).
Play: the abstract says all play effects had p > 0.05, but the full-text table shows two small significant ones for other-sports play (juniors −0.121, p = 0.021; all seniors 0.130, p = 0.040); the authors call play effects negligible because each is under |0.20| or not significant; GRADE low.

Practical Application

Among adult athletes, those who reached world class had on average done more coached practice in other sports as children and teenagers, started their main sport later and done less of it over their careers than national-class peers (d about 0.4-0.5 for other-sports practice and starting age, -0.23 for main-sport practice), while peer-led play such as pick-up football showed no clear link either way. For an adult who played a lot of football and ran as a child, this says a varied childhood is the usual history of the very best, not that it gives a recreational adult a measured head start: the comparison is between elite levels, the histories are recalled, and nothing here is causal.
If the football was coached club football, it counts as coached practice in another sport, the activity that went with adult world class; kick-abouts count as play, whose links were negligible.
If running is now the main sport, childhood running is an early start in it; in these elite samples a later start went with world rather than national class, most clearly in timed and measured sports (d 0.65), as an association only.
The best juniors and the best adults are largely different people (the authors, citing earlier studies), so early success says little about an adult ceiling.
The authors' explanations (less overuse injury and burnout, a better match of athlete and sport, broader learning) are hypotheses; they cite earlier work suggesting the benefit runs through later skill learning rather than physical development. None was tested here.
For a young athlete now, the base's youth guide SYNTH-81 and the injury data in REF-602 carry the practical rules; this review adds the long-term performance side.

Limitations

Retrospective, self-reported sport histories; confounding and survivor bias cannot be excluded (the authors).
About performance level, not health, fitness or enjoyment; recreational athletes are not the subject.
67% men; few samples from Africa, Central and South America, combat, artistic and Paralympic sport; junior game samples dominated by male soccer.
The micro-structure of practice and interactions with genes, family support or psychology were not analysed.
Sample sizes and power varied across models; the supplementary heterogeneity and GRADE details were not read.
How this entry was checked and changed
Why this level

a systematic review and meta-analysis whose certainty was read: GRADE moderate for milestones, starting age, main-sport and other-sports practice, moderate to high for the world-class vs national-class comparisons, low for peer-led play; not A because the pooled studies compare recalled sport histories between performance levels, which the authors say cannot exclude confounding or survivor bias, so no causal reading is possible