REF-1704 · Lai et al., 2026 · PeerJ, · Paper · PubMed

Highly specialised youth athletes had about twice the odds of injury of low-specialisation peers (pooled OR 2.00, 95% CI 1.58-2.55; I² 65.9%) in a meta-analysis of 15 observational studies with 12,791 athletes aged mostly 12-18, 12 of them cross-sectional; age, sex (one male-only study), injury type and body region did not significantly change the association, which was weaker in studies sampling several sports (OR 1.61) than in single-sport samples (2.41-2.67); an association, not proof that specialising causes injury

Main caveat: The authors': English only; only studies using the Jayanthi framework, so other definitions of specialisation were left out.
moderate evidenceMeasured on non-athletes to trained athletesRead in full

Lai and colleagues pooled studies that classified youth athletes (10-19 years) by how specialised they were in one sport and recorded their injuries, and asked how strong the link is and whether study design, sport type, age, sex, how injury was measured, injury mechanism or body region change it.

From 4,049 records they kept 12 articles and added 3 from earlier reviews: 15 studies, 12,791 athletes (60.4% female), published 2015-2025 in the USA, New Zealand and Canada. All 15 reported odds ratios; where a study gave several, the high-versus-low specialisation comparison was taken.

Highly specialised athletes had about twice the odds of injury of low-specialisation athletes (OR 2.00, 1.58-2.55), with moderate heterogeneity (I² 65.94%), leave-one-out estimates of 1.97-2.12 and no sign of publication bias. Only sport type moderated the link: OR 2.67 in contact-sport samples, 2.41 in non-contact samples and 1.61 in studies that sampled several sports.

The odds were raised in early and middle-to-late adolescence, in girls and boys (one male-only study), for overuse and mixed injury outcomes, for the lower and upper limb, and whether injury was diagnosed by a clinician or self-reported.

The authors conclude that specialisation is positively associated with injury, that varied participation may play a role in lower risk, and that the observational designs and mixed injury definitions rule out causal claims.

Participants

15 studies; 12,791 athletes, 7,729 female (60.4%) and 5,062 male (39.6%), mostly 12-18 years (study mean ages 12.6 to 16.1).
Countries: USA 11, New Zealand 2, Canada 2.
Samples: high-school and junior-high athletes, single-sport samples in volleyball (girls), softball (girls), baseball (boys), basketball and soccer, and multi-sport samples; the two case-control studies are Jayanthi 2015, which compared 822 athletes seen in medical clinics for sports injuries with 368 seen in paediatric clinics for routine examinations, and Jayanthi 2020; the one cohort (Buser 2025) followed 485 injured athletes for 6 months.

Protocol

Search: eight databases from inception to 20 March 2026, English, peer-reviewed; Google Scholar screened by relevance until no new eligible studies appeared; two reviewers screened independently.
Exposure: specialisation level (high, moderate, low) on the Jayanthi three-item scale (quit other sports, train >8 months a year, main sport more important) or adapted four- or six-point versions; studies comparing only single-sport with multi-sport athletes were excluded.
Outcome: sports injury of any kind reported (overuse, acute, general musculoskeletal, stress fracture, concussion), pooled as one construct; mechanism and body region coded as moderators where reported.
Overlap: two overlapping datasets were reduced to one study each.
Quality: JBI checklist for cross-sectional studies (only "high", more than five yes ratings, kept); Newcastle-Ottawa Scale for case-control and cohort studies (only six or more stars kept).
Analysis: odds ratios with 95% CI, random effects; Q, I², τ²; leave-one-out; funnel plot, Egger's and Begg's tests; subgroup moderator analyses with post hoc pairwise tests for sport type.
Funding and interests: no funding; no competing interests declared.

Key Results

Comparison (high vs low specialisation)Pooled OR (95% CI)Notes
All 15 studies2.00 (1.58-2.55), p < 0.001I² 65.94%, τ² 0.13, Q 41.11; leave-one-out 1.97-2.12
Contact-sport samples2.67 (2.05-3.48)sport type moderated, Q 6.17, p = 0.049
Non-contact-sport samples2.41 (1.78-3.26)contact vs non-contact p = 0.55
Studies sampling several sports1.61 (1.17-2.20)lower than contact (p = 0.004) and non-contact (p = 0.037)
Cross-sectional / case-control / cohort (1 study)2.01 (1.49-2.70) / 1.70 (1.13-2.55) / 3.19 (1.69-6.03)design not a moderator, p = 0.26
Age 14-18 / 10-132.18 (1.69-2.80) / 1.75 (1.09-2.80)p = 0.42
Female studies / one male study2.39 (1.78-3.20) / 3.77 (1.39-10.20)p = 0.39; subset pooled 2.48
Clinical diagnosis / mixed / self-report2.06 (1.29-3.29) / 1.89 (1.32-2.69) / 2.58 (1.66-4.00)p = 0.55
Overuse / mixed injury outcomes1.82 (1.33-2.48) / 2.48 (2.01-3.05)p = 0.11
Lower / upper limb2.61 (1.94-3.53) / 2.87 (1.49-5.55)p = 0.80; subset pooled 2.66
Publication bias: Egger β 1.47, p = 0.17; Begg z 0.94, p = 0.35.
Single studies (Table 1): odds ratios ran from 0.75 (0.49-1.15; gradual-onset injury in 914 New Zealand children aged about 12.6) to 4.17 (1.48-11.74; gradual-onset injury in 340 youth football players); 4 of 15 intervals included 1.
Quality as rated: all cross-sectional studies met the JBI threshold for "high quality" and all case-control and cohort studies scored six or more stars (item scores in a supplement not read).

Practical Application

Youth athletes who specialise heavily in one sport report about twice the odds of injury of those who do not (OR 2.00 across 15 studies and 12,791 athletes). Specialised here means quitting other sports, training more than 8 months a year in one, and ranking it above everything else. The studies are observational and mostly asked about past injuries, so this is a link, not proof that specialising causes injury.
It agrees with [SYNTH-81], which keeps a teenage cyclist in varied riding and at least one other sport, and broadens the evidence under it: the guide's basis was one case-control study (REF-602), which this review includes. The review took that study's high-versus-low estimate for serious overuse injury (OR 2.25, 1.27-3.99), not the OR of 1.27 for injury that the review's own introduction quotes from it and the base also cites.
Do not read the sport-type result as "multi-sport athletes are protected". The "multiple sports" group is a set of studies that drew their athletes from several sports, and its OR of 1.61 still compares high with low specialisation within those samples; the review did not compare multi-sport athletes with specialists. The authors themselves say this needs testing directly.
Inference, not tested here: for a teenage cyclist the three flags translate to riding year-round with no other sport, no break months and the bike ranked above everything; no study here included cyclists, whose injury pattern may differ.
This is not treatment advice: an injured young athlete, or pain at a growth site, goes with a parent to a physician or physiotherapist.

Limitations

The authors': English only; only studies using the Jayanthi framework, so other definitions of specialisation were left out; few studies per moderator (one male-only and one cohort study), little from high-contact sports such as wrestling or rugby; the "multiple sports" category mixes exposure contexts; clinically different injury outcomes pooled into one; chronological age rather than maturity. No formal certainty rating (not one of the authors' listed limitations).
Design: 12 of 15 studies cross-sectional, mostly with recalled injury history, so injury may have shaped later training rather than the reverse; the case-control design relies on recalled specialisation (the authors); odds ratios were taken as reported, with adjustment differing between studies (not described in the review).
Reporting: the abstract says athletes in contact and non-contact sports had higher odds than those "involved in multiple sports", which the review's own coding of that category does not support; Bell 2016 is a cross-sectional study whose outcome is labelled "prospective".
Transfer: US school and club team and throwing sports; no cycling, running or endurance sport, and no hours or volume threshold analysed.
Not read: the supplementary quality scores, search strings, audit table and moderator plots.
How this entry was checked and changed
Why this level

the quality appraisal was read in the full text (all included studies met the review's thresholds for "high quality"), but the rubric's A rows are for reviews of controlled trials and these inputs are observational, 12 of 15 cross-sectional with injury history and specialisation measured at the same time, so it is graded one step below A-, as the base grades other syntheses of cross-sectional inputs; heterogeneity is moderate (I² 65.9%) and certainty was not graded