Shoulder injuries prevention programmes in handball: a systematic review with meta-analysis

Abstract

Objective

Handball is a popular team sport with injury risks, especially to the shoulder. This systematic review aims to assess the effectiveness of prevention programmes in reducing the incidence of shoulder injuries in handball.

Design

Systematic review with meta-analysis

Data sources

The systematic review was conducted using Embase, PubMed and Cochrane databases with specified keywords.

Eligibility criteria for selecting studies

Inclusion criteria encompassed randomised clinical trials focusing on adult and youth handball players, involving a preventive programme with specific exercises to reduce shoulder injuries, with a control group maintaining regular warm-up.

Results and summary

The initial database search identified 224 records (107 articles from Embase, 85 from PubMed and 32 from Cochrane). Five articles met inclusion criteria which recruited 1872 players (977 females/895 males). The programmes were varied but all included muscle strengthening exercises. Depending on the programme, they also incorporated exercises concerning range of motion, scapular, trunk and neuromuscular. Results from three studies included in the meta-analysis showed no significant difference in shoulder injury incidence rates between prevention programme and usual warm-up (OR: 0.73; 95% CI: 0.45 to 1.17; n=747).

Conclusion

This study did not demonstrate the effectiveness of prevention programmes for handball shoulder injuries and emphasises the need for more standardised research protocols to improve knowledge in this area.

PROSPERO registration number

CRD42022356497.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• The shoulder is one of the most affected regions by overuse injuries in handball.

• There is a lack of consensus in the literature regarding the effectiveness of prevention programmes for overuse injuries in handball.

WHAT ARE THE NEW FINDINGS

• This study demonstrates no effectiveness of prevention programmes for handball shoulder injuries.

• Further studies are needed, with standardisation particularly regarding the primary outcome measure (using the validated Oslo Sports Trauma Research Center overuse injury questionnaire) and the types of exercises included in prevention programmes, to reduce heterogeneity.

Introduction

Handball is a sport consisting of high-velocity overhead shots, contacts, changes of direction and jumps, thereby causing various injuries. In the epidemiological report by Giroto et al,¹ an injury incidence rate of 20.3 per 1000 matches and 3.7 per 1000 training hours was observed. Mashimo et al² show with 1017 Japanese university handball players that the body regions most affected by traumatic mechanisms are ankle (33.3%), knee (23.6%) and shoulder/clavicle (12.6%) while the body regions most affected by overuse mechanisms are lumbar spine/lower back (26.0%), knee (15.7%) and shoulder (15.0%). Indeed, the high frequency of shoulder problems has been found in prospective studies among elite handball players with a prevalence ranging from 17 to 44% of overuse injuries over a season using the validated Oslo Sports Trauma Research Center overuse injury questionnaire (OSTRC-O),3,7 and Møller et al reported an incidence of 1.4 shoulder injuries per 1000 hours in a prospective cohort of 679 elite adolescents.⁸

Shoulder overuse injuries are mostly secondary to the adaptation of shoulder anatomy with glenohumeral internal rotation deficit (GIRD) and the impingements.9,11 Burkhart et al⁹ described the adaptations of the thrower’s shoulder. He described a fibrotic change in the postero-inferior capsuloligamentous complex due to repeated micro-trauma in distraction, leading to a GIRD and a glenohumeral external rotation (ER) increase. These adaptations are thought to be responsible for superior labral antero-posterior (SLAP) lesions with peel-back mechanism and rotator cuff injuries with hypertwisting mechanism.^{12 13} While anatomical adaptations such as GIRD and related impingements are central to shoulder injury mechanisms, a substantial amount of research has also identified a myriad of internal and external risk factors influencing shoulder injury incidence among throwing athletes.

A lot of risk factors for shoulder injuries have been investigated among throwing athletes and particularly for handball players. Several external risk factors have been found such as back and to a lesser extent pivot position, match versus training for elite players, additional match per week and increased training load.18 14,18

For internal risk factors, some are unmodifiable like age (older), previous injury or surgery and sex (female athletes).13 6 19,21

Fortunately, some additional internal modifiable risk factors of overuse injury have also been identified in prospective cohort studies. The internal modifiable risk factor most often found in cohorts is weaker isometric ER strength.^{3 7 8 22} In terms of strength, one report identified weaker eccentric ER strength as a risk factor⁷ and another one weaker isometric internal rotation (IR) strength in women.²² ER/IR strength ratio seems also to be a risk factor as shown by Achenbach et al⁷ and Edouard et al with isokinetic muscular imbalance criteria in preseason (IRecc/ERcon at 60°/s>1.61 and ERcon/IRcon at 240°/s<0.69).²¹ The GIRD has been widely described in throwing sports, particularly in baseball.9,11 Regarding handball, three studies have reported three different significant modifications in the range of motion (ROM): a 5° reduction in total ROM,³ a 7.5° increase in ER and a 6–7.5° reduction in IR in women, and a 4° reduction in IR in men.^{4 7} The variability of these results can be explained by the possible dynamic nature of GIRD due to a muscular component, as noted by Kibler et al.¹⁰

Scapular dyskinesis seems to be a risk factor in several studies^{3 8 22} but is still discussed today because it was not significant in other prospective cohorts^{4 7}; two recent meta-analyses were carried out not specifically for handball but for throwing sports and found contradictory results: one found that scapular dyskinesis increased the risk of future shoulder pain by 43% in asymptomatic athletes,²³ while the other concluded that scapular dyskinesis should not be considered an isolated risk factor for shoulder injury.²⁴ Reduced kinetic chain function and limited thoracic mobility are still uncertain risk factors today.²⁵

Additionally, we can note that years of playing handball, level of play and body mass index (BMI) do not seem to be risk factors for overuse shoulder injury.^{3 6 16}

The identification of pathomechanics and modifiable internal risk factors for overuse injuries has opened up promising perspectives for prevention, particularly through the development of targeted prevention programmes aimed at addressing these risk factors directly, and thereby indirectly reducing the injury incidence in handball players. Given the absence of a clear consensus in the literature regarding the effectiveness of such programmes marked by contradictory results across studies and considerable variability in programme characteristics (eg, number of repetitions, exercise types and intensity), we considered it necessary to conduct a systematic review with meta-analysis to synthesise the current state of evidence and help inform future research directions.

Although there has been a growing interest in injury prevention across various sports, to our knowledge, no systematic review has yet explored the effectiveness of prevention programmes in reducing injury incidence within handball in contrast to football, where such evidence is well-established, particularly with regard to anterior cruciate ligament and lower limb injuries.^{26 27}

Given the high prevalence of shoulder injury in handball and the knowledge of modifiable internal risk factors, it is important to find a programme to prevent these injuries. This systematic review aims to explore the effectiveness of prevention programmes in reducing the incidence of shoulder injuries in handball players.

Materials and methods

Eligibility criteria

Inclusion criteria

Only records written in English or French and only randomised clinical trials (RCTs) and cohort studies were included if they met the PICO criteria:

• POPULATION: players of handball (adult and youth)

• INTERVENTION: application of a preventive programme, including specific exercises to reduce shoulder injuries.

• CONTROL: no change in warm-up or training routine

• OUTCOMES: shoulder injury (OSTRC-O and modified version of Fahlström questionnaire) or shoulder pain (visual analogue scale (VAS))

Exclusion criteria

All articles that did not analyse handball were excluded. Case reports, narrative reviews, systematic reviews, meta-analyses and studies that did not detail precisely the exercises making up the prevention programme were also excluded.

Information sources and search strategy

The present systematic review was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The study protocol was prospectively registered (PROSPERO ID: CRD42022356497, https://www.crd.york.ac.uk/PROSPERO/view/356497). No changes were made to the registered protocol; however, we specified the outcomes ‘pain’ and ‘shoulder injury’. An electronically assisted medical literature search was conducted on 29 April 2025 using three different databases (PubMed, Embase and Cochrane) and using keywords which included “(shoulder* OR arm) AND prevent* AND handball”. The search was intentionally broad to avoid losing studies. All articles published up to 29 April 2025 were included.

Study selection and data collection process

The studies (titles and authors) were extracted into an Excel file by one reviewer for each database and then sent to the other reviewers. The selection was based on the title content. If inclusion or exclusion of the record was not possible based on the title, the abstract was retrieved and read. The selected report was independently reviewed and assessed by two authors. In case of disagreement, it was jointly discussed. If no consensus could be reached regarding the inclusion or exclusion criteria, a third reviewer made the final decision. We did not contact authors of primary studies to clarify eligibility. In addition to the database search, we also performed backward citation tracking and checked the reference lists of relevant systematic reviews to identify additional studies.

One author extracted the following information from each of the included studies: (1) general characteristics of the publication: title, authors, year of publication and design of study; (2) characteristics of the participants: sex, age, country and BMI of athlete participants; (3) intervention/control: description of prevention exercises and frequency; (4) description and results of the primary and secondary outcomes; and (5) authors’ conclusion.

Study risk-of-bias assessment and reporting bias assessment

To evaluate the risk of bias, we used Cochrane’s Effective Practice and Organisation of Care (EPOC) criteria consisting of nine questions about random sequence generation, allocation concealment, baseline outcome measurements similar, baseline characteristics, knowledge of the allocated interventions adequately prevented during the study, protection against contamination, selective reporting and other bias. The EPOC criteria are chosen because they are specifically designed to assess the risk of bias in studies with non-individual randomisation and complex interventions, which better aligns with the design of the included prevention programmes in our review. The risk-of-bias assessment was conducted by a primary reviewer; in case of doubt, a consensus was reached with the assistance of a second reviewer. Each criterion was categorised as having a high, low or uncertain risk of bias. Due to the small number of included studies, the assessment of publication bias and small study bias using a funnel plot was not performed.

Effect measures and synthesis methods

The main summary measures included risk ratios and ORs, which were reported in the studies and combined into tables. All findings related to both primary and secondary outcomes were synthesised. The certainty of the evidence was assessed by one author using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach. Evidence was downgraded based on the following domains: risk of bias, inconsistency, indirectness, imprecision and publication bias.

Meta-analysis

Statistical analysis was performed using Review Manager software (RevMan, V.5.4.1). We calculated the OR for the Achenbach and Asker studies using the published data and used the generic inverse variance method for Andersson to directly input the original OR. Effect sizes for outcomes were expressed in terms of OR with 95% CIs. Data extraction was cross-checked by two independent reviewers to ensure accuracy in reporting measures of spread. A random-effects model was used to account for anticipated clinical and methodological heterogeneity among studies, providing a more conservative and generalisable estimate of the pooled effect. Heterogeneity was assessed using Cochran’s Q test and the I² statistic. No subgroup analyses were performed to explore potential sources of heterogeneity. High-risk-of-bias studies were defined as those presenting more than two domains assessed as high risk according to the EPOC guidelines. Sensitivity analyses excluding these studies were conducted to assess the robustness of the meta-analytic findings.

Results

Study selection

During the literature search, we found 107 articles from Embase, 85 from PubMed and 32 from Cochrane, for a total of 224 articles. After removing 83 duplicates, 141 records remained for the title screening. A total of 116 records were excluded. Of the 25 remaining abstracts, only 5 met the inclusion criteria and were analysed in the review (figure 1). The five studies consisted of five RCTs involving 1872 patients. Using the keywords “shoulder injury prevention exercises”, we did not identify any discrepancies between the registered protocols and the published studies, limiting the risk of publication bias.

Figure 1. Flowchart of literature search.

Study characteristics

The five studies included were recent (2015–2022) interventional controlled studies using a random selection of teams; three studies were stratified, one by gender and competition level,²⁸ one by age, sex and league level²⁹ and another by the number of players enrolled at each school.³⁰

Four studies included only elite players2830,32 and one included elite/amateur/recreational level.²⁹ One study included U19 and senior league,²⁹ one only senior league,²⁸ two U16 leagues^{31 32} and one U15, U17 and U19.³⁰

Data collection was done differently for all studies both qualitatively and quantitatively. Østerås et al³² give little information. We only know that they gave a questionnaire before and after the season, but unfortunately, they did not explain how they collected compliance, and they did not calculate the response rate.

Sommervold et al³¹ opted for a monthly collection through coaches. Andersson et al,²⁸ Achenbach et al²⁹ and Asker et al³⁰ used an online questionnaire via email, SMS or application. Only Andersson et al²⁸ and Asker et al³⁰ sent automatic reminders to non-responders per email and SMS, and after that, the teams were called or visited in person to complete the questionnaire on paper.

All authors have chosen a follow-up of one season, but one decided to evaluate outcomes at the start and end of the season,³² one every month,^{28 31} one every 2 weeks²⁹ and 1 weekly.³⁰

Finally, Andersson et al²⁸ have chosen to exclude acute injuries from analysis and Sommervold et al³¹ to exclude previously undergone shoulder surgery.

A total of 1872 handball players were included in the review, with 977 females (52%) and 895 males (48%).

The mean BMI was 22.3 kg/m², with a range of 18.0–23.7 kg/m². The mean BMI values were calculated with the mean weight and mean height for Andersson et al²⁸ and Achenbach et al²⁹ because the BMI was not provided.

Regarding the mean age and the mean years played, only three studies listed the data, making it difficult to analyse and compare; the range of mean age was 15–24 years, and the range for years played was 8–16 years (table 1).

Table 1. Baseline demographic and clinical characteristics of players of each study.

Reference:	Østerås et al32	Sommervold et al31	Andersson et al28	Achenbach et al29	Asker et al30	Total
Players (F/M)	109 (F)	106 (F)	321 (F)/339 (M)	259 (F)/320 (M)	182 (F)/236 (M)	977 (F)/895 (M)
IG	53 (F)	53 (F)	160 (F)/171 (M)	123 (F)/161 (M)	81 (F)/120 (M)	470 (F)/452 (M)
CG	56 (F)	53 (F)	161 (F)/168 (M)	136 (F)/159 (M)	101 (F)/116 (M)	507 (F)/443 (M)
Mean age (years)
IG	15	NR	23 (F)/23 (M)	NR	17
CG	15	NR	22 (F)/24 (M)	NR	17
Mean BMI (kg/m2)
IG	18.5	NR	23.2 (F)/25.2 (M)*	23.6 (F)/24.6 (M)*	NR
CG	18.0	NR	23.4 (F)/25.6 (M)*	23.7 (F)/24.5 (M)*	NR
Mean years played
IG	8	NR	NR	16 (F)/14 (M)	10
CG	8	NR	NR	13 (F)/14 (M)	9

^*Calculated with mean weight and mean height in each respective article.

BMI, body mass index; CG, control group; F, female; IG, intervention group; M, male; NR, not reported.

For playing level, it was impossible to calculate a precise number of professional and amateur players because in the study by Achenbach et al²⁹ only the number of professional and amateur teams was reported. However, we can notice that the Achenbach et al report²⁹ represented 30% of all the players, so at least 70% are high level.

Risk of bias in studies

In the assessment of the risk of bias with EPOC (figure 2), three out of nine criteria were estimated to have a low risk of bias. Two studies had a high risk of bias for ‘baseline characteristics similar’, with the Sommervold et al report³¹ not reporting on the characteristics of the population and the Andersson et al report²⁸ showing a statistically significant difference in age between the two groups. All the studies experienced a significant loss of players between randomisation and analysis due to a lack of completed questionnaires, leading to potential attrition bias. Only Andersson et al²⁸ conducted a multiple imputation analysis, limiting the risk of bias, and Østerås et al³² had an identical distribution of losses between the two groups (28% vs 27%). Additionally, all the studies exhibited significant bias due to ‘knowledge of the allocated interventions adequately prevented during the study’ as the players were not blinded to the intervention.

Figure 2. Risk of bias with the Effective Practice and Organisation of Care.

Results of individual studies

The five studies had their own prevention programme, each of which was different. All the programmes were based on strength training; one prevention programme had only overall shoulder muscle strengthening exercises with push-ups and an exercise consisting of moving the body from elbow standing to hand standing with extended elbows and then down to elbow standing again which introduced instability.³¹ Others studies introduce specific shoulder muscle strengthening exercises for ER strength28,3032 and scapular muscle strength.28,30 Some prevention programmes have added exercises to improve glenohumeral rotation internal ROM (ROM IR),28,30 trunk mobility and kinetic chain^{28 30} (table 2, onlinesupplemental Annex 15).

Table 2. Summary of studies evaluating a programme to prevent shoulder injuries in handball.

Studies (design)	Population	Intervention group/control group	Outcomes	Results	Authors’ conclusion
Effects of a strength-training program for shoulder complaint prevention in female team handball athletes. A pilot study. Havard Østerås et al32 2015 (interventional controlled study using a random selection of teams)	Junior female handball players from highest club of Norwegian junior league U16 season 2012–2013	Intervention group: prolonged part of the warm up (online supplemental annex 1): Push-ups 3 ×12 repetitions Shoulder internal rotation (90° abducted upper arm) 3 × 20 repetitions with elastic Shoulder external rotation (90° abducted upper arm) 3 × 20 repetitions with elastic Control group: classical warm-up	Primary outcome: number of players reported shoulder complaints defined by a player who dropped training sessions and/or matches (modified version of Fahlström questionnaire was used) Secondary outcomes: Pain intensity (VAS 0–100 mm) Muscular strength of upper bodies by performing a maximal repetition test of push-ups	Seven teams included (109 players) Shoulder complaints at pre-test 34% in intervention group and 23% in control group. At post-test 11% in the intervention group and 36% in the control group. High significance (p<0.001) in the increase in shoulder strength between 2 groups, as measured by the number of push-ups (+300% in the intervention group). No results for pain intensity.	Positive effects of a strength-training preventive programme of shoulder complaints by additional shoulder girdle strength Necessary to investigate optimal exercises and dosage
What is the effect of a shoulder-strengthening program to prevent shoulder pain among junior female team handball players? Maria Sommervold et al31 2017 (interventional controlled study using a random selection of teams)	Junior female handball players from highest club of Trondela region in Norwegian junior league U16 season 2014–2015	Intervention group: prolonged part of the warm up, three times a week (online supplemental annex 2): Push-ups, 3 × 12 repetitions Moving the body from elbow standing to hand standing with extended elbows and then down to elbow standing again, 3 × 12 repetitions Control group: classical warm-up	Primary outcome: pain intensity with VAS (0–100 mm) Secondary outcomes: Quick-DASH Three physical tests at baseline and post-test: Maximal repetition test of push-ups Maximal throwing distance Strength of the internal and external rotators of the throwing shoulder measured with a handheld dynamometer	Six teams included (106 players). No significant difference in or between the groups at baseline or post-test for mean pain intensity (VAS). No difference significant for Quick-DASH. Difference significant for Quick-DASH sport. Functional test significant for push-ups and no significant for throwing distance and strength of internal and external rotators.	A strength-training programme had no effect on preventing shoulder pain among junior female team handball players
Preventing overuse shoulder injuries among throwing athletes: a cluster-randomised controlled trial in 660 elite handball players. Stig Haugsboe Andersson et al 201628 (interventional controlled study using a random selection of teams stratified by gender and competition level)	Players in the two top seniors divisions in Norway season 2014–2015	Intervention group: five groups of three exercises three times a week aiming at increasing the glenohumeral internal range of motion, external rotation strength, scapular muscle strength, improving the kinetic chain and thoracic mobility (online supplemental annex 3) Control group: classical warm-up	Primary outcome: Prevalence of shoulder problems using OSTRC-O and substantial shoulder problems (= moderate or major effect on training or performance in OSTRC-O) Secondary outcomes: Severity score of injuries (0–100) calculated with OSTRC-O Exposure to handball training, match play and additional strength training six times during the season	46 teams included (660 players). Lower risk of shoulder problems in the intervention group compared with the control group (OR 0.72, 95% CI 0.52 to 0.98, p=0.038). No significant difference for substantial shoulder problems between the intervention and control groups (OR: 0.78, 95% CI 0.53 to 1.16 p= 0.23). Average severity score of the shoulder problems reported: 29 in the intervention group and 35 (95% CI 32 to 37) in the control group. Exposure: control group 17 more minutes of strength training per week (p=0.004).	The OSTRC shoulder injury prevention programme should be included as a part of the warm-up in throwing sports
Multicomponent stretching and rubber band strengthening exercises do not reduce overuse shoulder injuries: a cluster randomised controlled trial with 579 handball athletes. Leonard Achenbach et al 202129 (interventional controlled study using a block-stratified cluster randomisation 1:1)	Players aged over 16 years in elite or amateur league during season 2019–2020 in Germany	Intervention group: five exercise blocks from three steps three times a week for work scapular activation/strength/control, external rotation strength and rotational internal range of motion (online supplemental annex 4) Control group: usual training modules	Primary outcome: prevalence of overuse injury using OSTRC-O Secondary outcomes: Substantial shoulder injury (= moderate or major effect on training or performance in OSTRC-O) Intensity of symptoms with short handball WOSI	61 teams (579 players). No significant difference in the prevalence of overuse shoulder injury between the intervention group 38.4% vs 35.9% (p = 0.542). No significant difference for substantial overuse shoulder injury: intervention group 31.7% vs 2 control group 6.4% (p=0.164). No differences for the intensity of symptoms.	A multicomponent exercise programme using rubber band, stretching and partner exercises did not significantly reduce prevalence or symptoms of overuse injury to the throwing shoulder in primarily amateur and recreational handball athletes of both sexes
The effect of shoulder and knee exercise programmes on the risk of shoulder and knee injuries in adolescent elite handball players. Martin Asker et al30 2022 (interventional controlled study using a three-armed cluster randomisation)	Players aged 14–19 years in elite level during season 2018–2019 in Sweden	Intervention group: five exercises with four levels three times a week and consisting of shoulder and trunk strength, control and mobility It includes a throwing programme to be performed during the off-season and pre-season (online supplemental annex 5) Control group: usual training modules	Primary outcome: number of first-time shoulder injury events during the study follow-up per 1000 hours of handball play using OSTRC-O Secondary outcomes: Time-loss shoulder and knee injury with OSTRC-O Substantial shoulder problems (= moderate reduction in training volume or performance with the OSTRC-O) Any shoulder: reporting anything but ‘full participation without any shoulder/knee problem’ with the OSTRC-O Weekly prevalence of substantial shoulder problems Weekly prevalence of any shoulder problems	12 teams (418 players). 56% lower shoulder injury rate in intervention group than the control group (HRR 0.44; 95% CI 0.29 to 0.68 with p<0.001). ARR between the intervention group and the control group: 11% (95% CI 4 to 18). NNT: 9 (95% CI 6 to 24). In the intervention group: 56% lower rate of time-loss shoulder injury (p < 0.001), 49% lower rate of substantial shoulder problems (p < 0.01) and 51% lower rate of any shoulder problems (p < 0.01). Weekly prevalence of any shoulder problem and substantial shoulder problems during the trial period was 3% and 1% in the intervention group and 8% and 2% in the control group.	The IPEPs shoulder control reduced the risk of shoulder injuries in adolescent elite handball players The shoulder group reduced also the total weeks of shoulder injuries and the substantial shoulder injuries compared with the control group

ARR, absolute risk reduction; HRR, HR ratio; IPEP, Injury Prevention Exercise Programme; NNT, number needed to treat; OSTRC-O, Oslo Sports Trauma Research Center overuse injury questionnaire; U16, under 16; VAS, visual analogue scale; WOSI, Western Ontario Shoulder Instability Index.

Results of syntheses

Outcomes

Shoulder injury

Østerås et al³² have chosen the number of players who reported shoulder complaints defined by a player who dropped training sessions and/or matches (a modified version of the Fahlström questionnaire originally for badminton players to identify the rate of shoulder complaints³¹ was used). Three studies used the Overuse Injury Questionnaire, a validated questionnaire for registration of overuse injuries (OSTRC-O).³³

In addition, one included study used the VAS to assess shoulder pain as its primary outcome; we report this for completeness.

Overall, two studies found a lower risk of shoulder complaints or problems associated with the prevention programme, while three studies reported no significant effect. The results of all primary and secondary outcomes are presented in table 2.

Meta-analysis

We found it more appropriate to include only the studies by Andersson et al,²⁸ Achenbach et al²⁹ and Asker et al³⁰ because the three studies use the OSTRC-O questionnaire as the primary outcome, especially to try to reduce heterogeneity. Additionally, the two studies by Østerås et al³² and Sommervold et al³¹ presented more bias, did not use validated criteria for shoulder overuse injuries and incidentally used smaller sample sizes.

Results from the three studies included in the meta-analysis, presented in figure 3, showed no significant difference in shoulder injury incidence rates between prevention programme and usual warm-up (OR: 0.73; 95% CI: 0.45 to 1.17; n=747).

Figure 3. Forest plot showing the effects of prevention programmes on shoulder injuries assessed using the Oslo Sports Trauma Research Center overuse injury questionnaire compared with control intervention. IV, inverse variance.

A sensitivity analysis including only high-quality studies was feasible in this meta-analysis. As the Achenbach study presents a significant bias in statistical power attributed to the COVID-19 pandemic (acknowledged by the authors), we performed a sensitivity analysis excluding this study (figure 4). This sensitivity analysis shows a change in the overall effect with an OR of 0.58 (95% CI 0.35 to 0.97) and a decrease in a substantial heterogeneity with an I² of 62%. Results showed consistency with the main results.

Figure 4. Forest plot from sensitivity analysis excluding the Achenbach study using Oslo Sports Trauma Research Center overuse injury questionnaire (OSTRC-O) as the primary outcome measure. IV, inverse variance.

Compliance with the prevention programme

Compliance varied between studies from 53% to 71%. First, in the Summervold et al study,³¹ no players filled in the compliance form; therefore, no data are available. All other studies had moderate compliance. The study by Andersson et al²⁸ reported a compliance of 53%, equivalent to 1.6 completed sessions per week, which is comparable to the study by Asker et al,³⁰ where players completed a mean of 1.3 prevention programme sessions per week out of an average of 2.2 training sessions per week. Finally, the study by Østerås et al³² reported a compliance of 72%, while in the study by Achenbach et al,²⁹ 64% of players completed at least two out of three sessions.

Response rate

All the studies have a high response rate, with a range of 61–86% of questionnaires returned by players for post-test. There was no difference between the interventional group and the control group except in Sommervold et al,³¹ with a response rate of 70% in the control group and 92% in the intervention group, with no explanation provided.

Otherwise, two out of four studies that have several questionnaires during the season note an important decrease in response rate. In the Achenbach et al study,²⁹ there was a response rate of <45% in the last two months in both groups, and for the Summervold et al study³¹ the response rate for the intervention group in the 2 months preceding the post-test was only 47% and 28%. Note that the response rates were not calculated in the Østerås et al study.³²

Certainty of evidence

The certainty of evidence was assessed using GRADE. Initially rated high RCT, it was downgraded to low due to high heterogeneity and imprecision (OR 0.73, 95% CI 0.45 to 1.17), suggesting that further high-quality studies may significantly impact the estimated effect.

Discussion

Summary of the main findings

Our study presents a systematic review and meta-analysis, which showed a non-statistically significant overall effect of prevention programmes on the incidence of overuse injuries. First, we tried to understand what differentiates significant studies from non-significant ones by dissecting prevention programmes. We decided to begin by discussing each study individually before making comparisons.

Østerås et al³² showed that shoulder complaints increased in the control group and decreased in the prevention programme group. We have to be careful with these results because no significance of the results is mentioned, and this pilot study had some bias. There is a low number of players at baseline n=109 and only 72 players at the end of the season without the use of a sample size calculator, only young females were included, a fourth control team was added to get approximately equal number of players in the two groups and the questionnaire was not validated.

In the Sommervold et al study,³¹ the mean pain intensity measured using the VAS showed no significant difference in or between the groups at baseline or post-test. The two exercises included in the prevention programme were based on push-ups movement. As shown by Borreani et al³⁴ and Calatayud et al³⁵ in the EMG report, the push-ups increase muscle strength in the triceps brachii, upper trapezius, anterior deltoid and clavicular pectoralis. These exercises did not strengthen the muscles of the rotator cuff and in particular the ER strength, which is one of the most frequently occurring risk factors found in the literature.^{3 7 16} In addition, one of the secondary outcomes of this study was the strength of the internal and external rotators measured with a handheld dynamometer, and the difference at baseline and post-test was not significant. Thus, it is likely that the lack of significant difference is due to exercises not acting on the muscle groups which have been identified as injury risk factors. Additionally, there are some limitations with no data on personal characteristics and for compliance.

The Achenbach et al study²⁹ did not find a significant difference in the prevalence of overuse shoulder injury. This study had a good methodology with randomised controlled design, a high number of participants and a high compliance; however, no difference in overuse injuries was seen between the intervention group and the control group. Several factors may explain this result, for example, the target sample size was not reached (n=825 instead of 924) and a quarter of the participants were lost due to insufficient reporting; so, the power was probably not big enough and a premature ending of the season due to the COVID-19 pandemic shortened the intervention time. The premature termination of this intervention period study highlights the potential critical role of intervention duration in achieving preventive effects. Shorter intervention periods may not allow sufficient time for meaningful physiological adaptations, which could partially explain the failure to observe significant effects in some studies. Unfortunately, all included studies assessed the effectiveness of prevention programmes over a single season, making it difficult to fully support this hypothesis. Future research should assess the effectiveness of prevention programmes over multiple seasons to better capture the variability and long-term sustainability of outcomes, an issue already raised in the systematic review by Herman et al.³⁶

Andersson et al²⁸ showed reduced risk of overuse injuries in the intervention group, but no reduction was observed in substantial overuse injuries. A subanalysis of compliance showed a significantly lower risk of substantial overuse injury if the player did the prevention programme at least once per week. However, this is the result of a subgroup analysis, so it has low statistical power. Additionally, the lack of compliance and more strength training in the group control may have underestimated the results on substantial overuse injuries. This study has many players and a good methodology, but limitations have been raised like the inability to include player exposure as a potential confounder due to a lower response rate for the exposure data. Furthermore, if we focus on the extent of effectiveness, we can note that the OR of this study is lower than those of the report on prevention of injuries to the lower limb, for example.³⁷

Asker et al³⁰ found a significant lower risk of shoulder injuries in adolescent elite handball players but also a reduction in the secondary outcomes: total weeks of shoulder injuries and substantial shoulder injuries. Even if we cannot draw conclusions with secondary outcomes, this is the first study to find significant difference for substantial shoulder injuries. Furthermore, we can identify two potential biases in this study. First, like all the studies of this systematic review that use self-reported questionnaire, there is a classification bias. Indeed, self-reported questionnaires may introduce classification bias as subjective pain reports can be influenced by individual variability (such as pain tolerance, emotional state or social pressures), potentially leading to an underestimation of injury incidence or prevention effects, as discussed by Clarsen et al.³³ Second, there is a difference in the baseline characteristics with more players at the national level in the shoulder programme group than the control group. However, this bias probably has a little impact, knowing that several studies find no or a modest influence of the level of play.^{3 6 16}

Finally, we were able to isolate two programme models: on the one hand, Østerås et al³² and Sommervold et al³¹ integrated only a few exercises and focused mainly on muscle strengthening exercises; on the other hand, Andersson et al,²⁸ Achenbach et al²⁹ and Asker et al³⁰ had a more global work around the shoulder. If we compare the programme of Andersson et al,²⁸ Achenbach et al²⁹ and Asker et al,³⁰ we can see that they all work IR ROM, ER strength and scapular strength plus control. However, Andersson et al²⁸ and Asker et al³⁰ have added trunk mobility and strengthening exercises. These exercises are potentially non-negligible as Machado et al’s meta-analysis found that throwing athletes with shoulder injuries had worse trunk stability and endurance tests.³⁸

The players also performed more repetitions per exercise in the Andersson et al and Asker et al programme than the Achenbach et al programme (3 sets of 8–16 repetitions/2–3 sets of 30 s vs 2 or 3 sets of 8–10 repetitions). Most studies did not clearly define the targeted training intensity, understood as the relative load of the exercise. Although frequency and volume were generally reported, the absence of information regarding intensity may partly explain the lack of significant effects as it is a key variable to induce muscle adaptations.^{39 40} Furthermore, when we analyse the programme in detail, we can see that Andersson et al²⁸ and Asker et al³⁰ have used more complex exercises. These exercises include kinetic chain work aimed at enhancing the transfer of energy from the feet to the hand during shooting. That could potentially improve both performance and injury risk, as demonstrated by Cope with lumbopelvic control.⁴¹ They also integrated exercises with a rubber band or ball that can be assimilated to plyometric exercises and so neuromuscular exercises. Swanik et al⁴² performed a study on swimmers who show a 6-week plyometric training with elastic tubing and pitchback system improves proprioception, kinesthesia, activation time, amortisation time and so potentially functional stability of the shoulder. It will be interesting to study if a specific neuromuscular exercise with, for example, the pitchback system by reproducing handball passing or shooting movement decreases handball overuse injuries. Another element that may have influenced effectiveness is the teaching of the programme to the players. Indeed, at the beginning of the season, Andersson et al²⁸ and Asker et al³⁰ organised sessions for each team to learn the different exercises making up prevention programmes while Achenbach et al²⁹ sent written instructions by email and a DVD with the exercises. All these arguments support the sensitivity analysis, which showed a significant overall effect with an OR of 0.58 (CI 0.35 to 0.97, p 0.04).

Strengths and limitations

The strength of our study is the use of PRISMA 2020 guidelines, prior registration in PROSPERO, rigorous risk-of-bias assessment using validated tools such as EPOC and a meta-analysis incorporating sensitivity testing to assess the robustness of our findings.

The first limitation of this systematic review, and particularly of this meta-analysis, is the small number of included studies, which prevented the conduction of subgroup analyses despite initially broad inclusion criteria designed to ensure no report was omitted. This led to a second limitation: the inclusion of multiple primary outcomes, which complicated the interpretation of results. We are aware that the inclusion of the Sommervold et al study³¹ which uses VAS as the primary outcomes is a third limitation since it is not specific to an overuse injury. This was a voluntary choice to have the most possible exhaustive study. Finally, a fourth limitation was the high heterogeneity of the meta-analysis, even though these three studies had strong similarities and a common primary outcome. Given the high heterogeneity (I²: 78%), we attempted to identify potential points of divergence. However, we observed discrepancies in the initially included populations. Achenbach et al²⁹ included both amateur and professional players, unlike Asker et al³⁰ and Andersson et al,²⁸ who included only high-level players; the population might also be older based on the inclusion criteria, but in the absence of an average age provided in this study, this remains hypothetical. As we observed in the sensitivity analysis, the Achenbach et al²⁹ study appears to have a notable impact on both the overall effect size and heterogeneity. This indicates that the meta-analysis lacks robustness, and further research is necessary to confirm any conclusions. In fact, the study experienced a substantial participant loss due to insufficient reporting and was prematurely terminated because of the COVID-19 pandemic, potentially leading to underpowered results. Differences in the design of the prevention programme may also have contributed to the lack of observed effect; the absence of trunk mobility exercises and kinetic chain strengthening exercises, which were present in other more effective programmes, could have reduced the preventive impact. Additionally, the learning of exercises was delivered remotely via email and DVD rather than through supervised in-person sessions, potentially affecting participants’ understanding, execution quality and adherence to the programme.

Comparison with previous studies

In this systematic review, the studies focused on overuse injuries as opposed to traumatic injuries because the identified risk factor mostly led to overuse injuries. Wilk et al⁴³ declared in a clinical commentary that overuse injuries are more frequent in throwing sports, and Østerås et al³² found 55% of shoulder overuse injury; however, the epidemiological report by Mashimo et al² among Japanese university handball players found 57 traumatic shoulder injuries and 19 overuse shoulder injuries: that is, 25% of overuse injuries. This discordance probably reflects both the difficulties in defining overuse injuries and the variability in study populations, depending on whether amateur or elite players were involved, given the large differences in training volume. The definition of overuse injury is a non-traumatic injury which is very subjective.⁴⁴ Indeed, should a simple pain be considered an overuse injury? This problematic definition could explain the multitude of different primary outcomes in this analysis. The validated OSTRC-O used by the latest studies Andersson et al,²⁸ Achenbach et al²⁹ and Asker et al³⁰ seems to be a good compromise by combining reports of pain with difficulties in participating in normal training and competition, reduction in training volume and affection of performances. 

In the existing literature, we identified two systematic reviews that addressed a research question closely related to ours. Moiroux-Sahraoui et al⁴⁵ investigated the impact of prevention programmes in throwing athletes, focusing on functional tests, ROM, strength and injury incidence. Wright et al⁴⁶ assessed the effect of prevention programmes on injury incidence across a broader spectrum of overhead sports, including baseball players, swimmers and elite wheelchair basketball athletes, resulting in a slightly different study population. Furthermore, they did not perform a meta-analysis, thereby limiting the ability to directly compare our results. To date, meaningful comparison with previously published studies remains challenging due to differences in study populations and primary outcome measures.

One of the most important issues of prevention programmes is the compliance. Soligard et al,⁴⁷ Steffen et al⁴⁸ and Silvers-Granelli et al⁴⁹ showed that the compliance of the FIFA or FIFA 11+ programme, which is one of the most famous prevention programme today for football injury prevention, is significantly linked with the reduction of injury risk. Soligard et al⁴⁷ also noted that teams with a coach who used a prevention programme in the past have a lower risk of injury. The compliance of the studies in this review (between 53% and 71%) is lower than the studies in Thorborg et al’s²⁶ systematic review of FIFA/FIFA+; for the FIFA programme composed of 10 exercises, compliance was between 52% and 71%; and for the FIFA+ programme composed of 15 exercises, compliance was between 74% and 77%. These results are surprising because as we saw in this study the more complex and numerous the exercise in the programme is, the more the compliance decreases. Proof of the effectiveness of these programmes and therefore better communication regarding the importance of carrying out these programmes surely has a significant role in this increase in compliance. We have not found any studies exploring the link between the number of exercises in a programme and compliance, which could assist in the development of future programmes.

Suggestions for future research

Future research should aim to reduce the heterogeneity observed in current studies by standardising both the design of prevention programmes and the primary outcome measures. The inconsistency in the definition and measurement of overuse injuries across studies, such as the use of pain intensity scales (VAS in Sommervold et al³¹) instead of validated overuse injury questionnaires, likely contributed to the variability in outcomes. This heterogeneity complicates the interpretation of pooled results and highlights the need for use of validated instrument, such as OSTRC-O, to ensure comparability across studies. Furthermore, prevention programmes should be based on exercises targeting modifiable risk factors, including external rotator strength, IR ROM and scapular control, with an emphasis on integrating kinetic chain and neuromuscular training components. The complexity of prevention programmes appears to influence participant compliance, suggesting the need to balance exercise effectiveness with practical feasibility. Future programmes should consider incorporating simpler exercises that still address modifiable risk factors. The ideal approach would be to identify simple exercises that simultaneously target multiple risk factors; for example, by combining ER strengthening, scapular control and trunk mobility within the same movement pattern. High-quality RCTs with sufficient sample sizes and standardised reporting of adherence are necessary to better understand the true effectiveness of shoulder injury prevention programmes in handball players. Due to the lack of clear evidence on the effectiveness of a programme to prevent shoulder injuries in handball players, we believe that future programmes should be based on expert opinion, as with the DELPHI study that Fredriksen et al⁵⁰ and Lau et al⁵¹ began to do. It would also be interesting to include in the secondary outcomes the impact that the interventions have on the risk factors of overuse injuries to be certain of their desired effectiveness. One of the challenges in the future will also be to succeed in combining the different prevention programmes (eg, prevention of the shoulder and the knee anterior cruciate ligament) while keeping the completion time reasonable.

Conclusion

This study demonstrates no effectiveness of prevention programmes for handball shoulder injuries and the need to conduct further studies and harmonise study, particularly regarding the primary outcome measure using the validated OSTRC-O and the types of exercises included in the prevention programmes to reduce heterogeneity.

Data availability statement

Data are available upon reasonable request.