Research · Injury & Recovery

The Young Overhead Shoulder: Throwing Has Rules, Swimming Has 80 Theories

By Muhammad Dzulhisham · MSc Sports Coaching, NSCA-CSCS

Two young athletes turn up with shoulder pain. One is a thirteen-year-old pitcher; the other is a thirteen-year-old swimmer doing seventeen hours a week in the pool. Same joint, same age, roughly the same complaint. The conversations that follow tend to be nothing alike — and the reason is not clinical, it is that one sport has been studied far more successfully than the other.

It is worth laying the two evidence bases next to each other, because the contrast tells you something useful about which coaching decisions are actually supported.

Throwing: a long argument that mostly got settled

The foundational work here followed 481 youth pitchers aged 9 to 14 for ten years, interviewing them annually. Injury was defined severely — elbow surgery, shoulder surgery, or retirement from throwing injury — and the headline numbers are these.¹

That last tile is the one most people get wrong. The curveball was the villain of youth baseball for a generation, and this study set out to test it as one of three named hypotheses. It could not demonstrate the effect.¹ What it did demonstrate was a volume threshold.

A case-control study pointed the same direction from a different angle, comparing 95 adolescent pitchers who had undergone shoulder or elbow surgery against 45 who had never been significantly injured.² The injured group had pitched more months per year, more games per year, more innings per game, more pitches per game, more pitches per year, and even more warm-up pitches. They were more often starting pitchers, appeared in more showcases, threw harder, and — the detail that matters most — pitched more often while carrying arm pain and fatigue. The authors identified overuse and fatigue as the factors with the strongest associations.²

Notably, that study found no significant differences in pitch-type frequency or in the age at which different pitch types were first thrown.² Two independent designs, two acquittals for the curveball, and two convictions for volume and fatigue. This is about as close to convergence as youth sports-injury research gets, and it is why the sport now has innings and pitch-count rules rather than pitch-type rules.

Where the throwing picture is still moving

Settled is not the same as complete. A prospective study across the 2019–2023 seasons monitored 71 high-school pitchers aged 13 to 18 through 313 outings and 24,228 pitches, building a workload model that included velocity and intensity alongside raw pitch counts.³ Game pitch counts ranged from 19 to 219. The risk factors that reached significance were throwing at higher velocity in game, increased intensity — measured as the rise in mean velocity from preseason to in-season — and being an older pitcher. Workload itself did not separate injured from uninjured pitchers, and the authors are explicit that this analysis was underpowered.³

The practical reading is not that pitch counts were wrong. It is that a pitch count is a measure of quantity with no term for effort, and two pitchers with identical counts can have had very different afternoons.

There is also a quieter finding worth carrying into any sport. A cohort of 159 youth baseball players aged 9 to 12, physically assessed and then followed prospectively for about seven months, recorded an arm-injury incidence of 2.22 per 1,000 athlete-exposures. Injury frequency did not differ between pitchers and position players — 13.2% each.⁴ The arm is loaded by throwing, not by the job title.

The same study asked players and parents whether the athlete was specialised, then applied the standard research definition. Self-reported specialisation came out at 31%; by the research definition it was 83%, and 57.9% of parents and specialised players were unaware of the athlete's own status.⁴ A household can be deep into a risk factor while sincerely believing it does not apply to them — and the overuse picture in racquet sports and swimming runs on the same mechanism.

Swimming: a great deal of pain and very little proof

The scale of the problem is not in doubt. A systematic review and meta-analysis spanning 45 years and 94 studies, covering 10,973 swimmers, found the shoulder to be the most frequently reported injury region, appearing in 70.2% of included studies.⁵ Pooled shoulder-injury prevalence came out at 46.03% in elite swimmers (95% CI 35.85–56.22), 45.77% in amateurs and 33.96% in competitive swimmers. The reported injury prevalence across individual studies ranged from 0.3% to 91.2% — a spread that tells you as much about inconsistent definitions as about swimmers.⁵

A narrower review of 12 studies and 1,460 participants split swimmers by age and found adolescents worst affected, with shoulder pain rates of 91.3% against a range of 19.4% to 70.3% in the other age bands.⁶ Adolescent swimmers were training 17.27 ± 5.25 hours per week. In that group, weekly training volume and years active in competitive swimming both correlated significantly with supraspinatus tendon thickness — and every swimmer with tendon thickening had shoulder pain.⁶ The authors graded this a level II conclusion and stopped short of proposing cut-offs, because the data cannot support one yet.

Then comes the awkward part. A 2023 systematic review searched for risk factors for shoulder pain and injury in competitive swimmers, screened 1,356 studies, and carried 22 through to a best-evidence synthesis.⁷

Read the middle two tiles together and the position is stark: after decades of work, not one variable in competitive swimming has strong evidence behind it. Only competitive level and shoulder-muscle recruitment profiles reached moderate evidence for an association.⁷

More usefully for anyone running a squad, several popular screening targets reached moderate evidence against an association — internal and external rotation range of motion, the back-scratch test, training frequency, specialty stroke, height and weight, sex, and age.⁷ Swimming's equivalent of the curveball myth is the belief that you can find the at-risk shoulder by measuring its flexibility. The review does not support it.

What did survive, in both sports

The highest-quality study in that review followed 201 pain-free competitive swimmers across two seasons — 96 male and 105 female, mean age 13.9 — and recorded 42 new cases of shoulder pain interfering with training.⁸

Treat that first number with the caution its own confidence interval demands: a lower bound sitting exactly on 1.00 is the weakest form of a positive finding, and the model's discrimination is fair rather than good. But the direction is consistent with the throwing literature. What predicted pain was not the shape of the shoulder. It was how much work had recently been done relative to what the athlete was conditioned for, and how much endurance the posterior shoulder had.

Strip both literatures back and the overlap is small but real: volume, recent change in volume, and fatigue. Everything else — pitch selection in baseball, joint flexibility in swimming — has been tested more thoroughly than people assume, and has mostly failed to earn its reputation.

What this changes in practice

Count the exposure, not the position. Arm-injury frequency was identical in pitchers and position players. Any athlete throwing or swimming a lot is carrying the load, whatever the team sheet says.

Watch the rate of change, not just the total. The single strongest modifiable signal in the swimming model was acute workload relative to chronic. A jump into a training camp or a new squad is the event to plan around.

Treat "tired" and "sore" as data. Pitching while fatigued and while in pain separated the surgical group from the uninjured group. An athlete who reports arm tiredness has given you a usable measurement, and the response to it is a coaching decision, not a medical one.

Build posterior shoulder endurance. It is one of very few intrinsic variables with support in the swimming model, and it is trainable — unlike height, sex, age or stroke preference, all of which have moderate evidence against them.

Do not screen your way to safety with a goniometer. Range-of-motion measures have moderate evidence against an association with swimmers' shoulder pain. Measuring them is fine; treating the numbers as a risk score is not supported.

Ask the specialisation question with the definition in hand. More than half of the families in the baseball cohort misclassified their own child. Do not rely on the household's self-assessment.

Where this evidence stops

The throwing studies are strongest where their endpoint is most severe. A 5% ten-year incidence refers to surgery or retirement, not to the far more common experience of an arm that hurts for a fortnight, and the 3.5× finding carries a confidence interval running from 1.16 to 10.44 — real, but imprecise. The case-control design cannot establish direction: pitching in pain may be a marker of an arm already failing rather than the cause of the failure.

The swimming picture is weaker still, and the 2023 review says so plainly. Prevalence estimates ranging from 0.3% to 91.2% across studies mean the field has not agreed what counts as a shoulder injury, which alone would prevent strong conclusions. The prediction model was internally validated only — no external cohort has tested it.

None of this substitutes for assessment. Persistent shoulder pain in a growing overhead athlete can involve the proximal humeral growth plate, and that is a clinical diagnosis. What the evidence does support is a shift in where coaching attention goes: away from technique folklore and screening batteries, and toward the training load and the athlete's own report of fatigue.