The observation period comprised one football season from August/September 2014 to June/July 2015.
Study Population and Recruitment
Between May and July 2014, 1094 officially registered clubs with teams in the age categories under 9 years, under 11 years and under 13 years (boys and girls, born 2002–7) were invited to participate in the study in Switzerland, Germany, the Czech Republic and the Netherlands (Fig. 1). Inclusion criteria were: (1) the club must be officially registered in the (regional) football association; (2) children must be between 7 and 12 years of age at the start of the study; and (3) regular training must take place at least twice per week. Teams were not eligible for inclusion if the coach already used an injury prevention programme or a structured warm-up focusing on neuromuscular control. Prior to the start of the study, information meetings were conducted to inform coaches about the aims and procedures of the study and, for INT teams only, to give detailed instructions and practical application on the ‘11+ Kids’ programme.
Ethics
The study complied with ethical standards and the Declaration of Helsinki and was approved by the lead ethics committee (Ethikkommission Nordwest- und Zentralschweiz, EKNZ, Approval number 2014–232) and by all other regional ethics committees (Saarbrücken, Prague and Amsterdam). All children and their parents received written information about the aim and the methodology of the project prior to the start of the study. Participation was voluntary. Passive informed consent was acquired to include children into the statistical analysis. In the case where children or parents declined participation, parents informed the researchers via e-mail or telephone. All parents of injured children gave their active consent.
Randomisation
Participating clubs were randomised into INT or CON. All teams of the same club were randomised into the same group (clustered allocation with the club serving as a cluster) to minimise the risk of contamination. Computer-generated cluster randomisation was conducted by one researcher (OF) who had no direct contact with the clubs or team officials and who was not involved in the intervention. Age group, country and number of participating teams per club served as the strata for the randomisation.
Intervention
‘11+ Kids’ is an exercise-based programme to prevent football injuries in 7- to 13-year-old children. It was developed by an international group of experts based on the findings of an epidemiological study on injury incidence and characteristics in children’s football [4, 5]. The structure of the programme refers to the established ‘11+’ programme, which has been shown to be efficacious in players aged older than 13 years [7,8,9, 16].
A study on the preliminary version of this programme showed slight improvements in motor performance [18]. Such improvements have been described as a prerequisite for successful injury reduction [16, 24]. In addition, the programme proved feasible, and its acceptance was high among coaches and players (unpublished data). The results of this pilot study provided input for final programme adaptations. Prior to the start of the study, two teams (that did not take part in the study) extensively pilot tested the final programme.
‘11+ Kids’ consists of seven different exercises and can be performed in about 15–20 min after familiarisation. Three exercises focus on unilateral, dynamic stability of the lower extremities (hopping, jumping and landing), three exercises on whole body and trunk strength/stability, and one exercise on falling technique. The difficulty of each exercise is progressively increased in five levels to account for the varying age- and maturity-related performance levels, as well as for general differences in motor skills of children aged 7–13 years [see ‘11+ Kids’ manual, Electronic Supplementary Material (ESM) 1 and 2]. Coaches were instructed to start with the first level of each exercise and to proceed with the next level when all players were able to perform the exercise according to the description in the manual. Hereby, specific attention was set on the body alignment during the exercises (e.g. leg alignment during single-leg jumps).
During the first weeks of the season, our study assistants visited INT clubs and gave the coaches an instruction session on how to apply the ‘11+ Kids’ programme correctly. Coaches received a detailed manual of the ‘11+ Kids’ and a two-page summary for the pitch, and were advised to use the ‘11+ Kids’ programme at the beginning of their training sessions as a replacement of their usual warm-up at least twice a week. The coaches of CON were instructed to perform their usual warm-up.
Injury Surveillance and Documentation of Football Exposure
Player-specific football exposure (in minutes), sustained injuries and session-based information about ‘11+ Kids’ utilisation (INT only) were collected using an Internet-based injury registration system. This online platform was developed for (and successfully applied in) a previous epidemiological study in children’s football [4, 5]. It had been adapted based on previous experiences to improve the usability of the system and, thereby, to increase the compliance of the coaches. Self-reported anthropometric baseline data of the children were provided by parents at the start of the study. One contact person for each team (preferably, but not necessarily the coach) was appointed and instructed to complete the injury and exposure entry into the online injury recording system. The documentation of exposure time and injuries in INT teams started after the instruction session.
In case no data were entered within a period of 1 week, an automated reminder e-mail was sent. If an injury occurred, trained study assistants contacted the coach, the player and the parents via telephone and/or e-mail to assess all relevant aspects of the injury based on a standardised injury registration form. If an injury received medical treatment, parents were instructed to obtain the exact diagnosis from the treating physician. All information on each injury was screened by two medically trained investigators (MB, KadF), who were blinded to group allocation, to ensure an objective and independent injury classification.
To ensure good compliance regarding entry of exposure and injury data, four scientific assistants (one in each country) and nine study assistants supported coaches during data collection and injury recording. Each study assistant was responsible for 10–15 clubs. Study assistants were continuously in touch with the coaches via telephone and e-mail, and visited unannounced two training sessions of each team during the study period.
Evaluation of Feasibility and Acceptance Among Coaches
Study participants (coaches and players) were involved in the design of the intervention programme. In a previous (pilot) study [18], the intervention programme has been evaluated by coaches and players. Their valuable feedback has been considered during the development of the final version of the intervention programme, which was used in the study at hand.
Furthermore, in this study, the intervention programme has been evaluated by the coaches. The INT coaches were asked to complete an online questionnaire about their general rating of the programme (13 questions on e.g. quality of the manual, time requirement) and each of the seven exercises (five questions per exercise) at the end of the study (see ESM 3). A five-level Likert scale was used as described earlier [25].
Sample Size
For the primary outcome (overall injuries), sample size estimation revealed that a total of 1935 children are needed to detect a hypothetical hazard ratio (HR) of 0.66 with a power of 80%, and an alpha level of 0.05. This is based on the assumption that the club (cluster) contains 40 players on average, the intraclass correlation coefficient is 0.05, that 7% of the players in the control group will sustain an injury during the season and taking into account a design effect (inflation factor of 2.95) [4, 9, 26, 27].
Statistical Procedure
Player-specific time-to-injury data were analysed using extended Cox models. Uninjured children contributed their right censored ‘survival times’ to the analysis. The models contained mixed (random and fixed) effects. To acknowledge the clustered data structure, we included the variable ‘club’ and ‘team’ as a random effect. Further, to acknowledge that multiple injuries of one player are not independent, we also included the variable ‘id’ (player-specific identifier) as a random effect. The models were fit to reflect that “teams are nested in clubs”, “players are nested in teams” and “multiple injuries of one player are nested in the player”. This approach has been used previously [5]. During model building, we also fitted models to include the ‘country level’ to account for the hierarchical structure (i.e. “clubs located in countries”). To decide whether to include the additional level, we compared the integrated log-likelihood value with the less complex model (i.e. containing ‘team’ and ‘id’ level only). The Chi square tests showed large p values (p ≥ 0.58) for the comparisons between models. We therefore decided to use these less complex models (including ‘team’ and ‘id’ level) [28,29,30].
The proportional hazard assumption was tested during model building [31]. The intervention variable (INT vs. CON) was used as a fixed effect. Further, we entered variables (age, age-independent body height, age-independent body mass and match-training ratio) that had p < 0.2 in the univariate analysis into the multivariate model [5, 32]. When multicollinearity between two variables was present, we included the one with the smaller p value into the multivariate analysis.
The analyses were performed using R (Version 3.2.2) in combination with RStudio (Version 0.99.484) in a cloud computing environment on multiple servers. We used the ‘coxme’ package (Version 2.2-5) to fit the models. Kaplan–Meier curves were plotted for overall, severe and lower extremity injuries.
To conduct a compliance analysis, we carried out a tertile split of the INT players according to their weekly ‘11+ Kids’ completion rate [8]. “Completion” was defined as the full utilisation of the ‘11+ Kids’ warm-up programme (with all of its seven exercises as described in the manual) at the beginning of a training session. We compared the three INT groups (high/middle/low compliance) against each other as well as against CON using extended Cox models. We used player-specific ‘11+ Kids’ completion data (rather than team-based information). Therefore, the actual individual exposure to the intervention programme was taken into account.
In addition, we used the aforementioned extended Cox models in a different approach to investigate the influence of compliance (completed ‘11+ Kids’ training sessions per week) on the reduction of injury rate. Thereby, a compliance threshold was increased stepwise by 0.01 increments, removing all players with a compliance below this threshold. Our intention was to evaluate the benefit from additional weekly sessions.
Finally, we applied the magnitude-based inference approach to investigate the intervention effect regarding specific types, locations and mechanisms of injuries. It has to be mentioned that the study was not powered for these subgroup analyses. However, it might provide useful information about the clinical relevance of the effects found. We used an open source spreadsheet to run the analyses [33]. Hazard ratios and the associated p values were used to get 90% confidence limits for, and inferences about, the true value of an effect statistic. Threshold values for “benefit” HR < 0.77 and “harm” > HR 1.30 were used. Qualitative descriptors were assigned to quantitative chances of intervention effects as follows: 0.5–5%: “very unlikely”; > 5–25%: “unlikely”; > 25–75%: “possibly”; > 75–95% “likely”; > 95–99.5%: “very likely”; > 99.5%: “almost certainly” [34].