Assessing the cumulative effect of long-term training load on the risk of injury in team sports

Abstract

ObjectivesDetermine how to assess the cumulative effect of training load on the risk of injury or health problems in team sports.MethodsFirst, we performed a simulation based on a Norwegian Premier League male football dataset (n players=36). Training load was sampled from daily session rating of perceived exertion (sRPE). Different scenarios of the effect of sRPE on injury risk and the effect of relative sRPE on injury risk were simulated. These scenarios assumed that the probability of injury was the result of training load exposures over the previous 4 weeks. We compared seven different methods of modelling training load in their ability to model the simulated relationship. We then used the most accurate method, the distributed lag non-linear model (DLNM), to analyse data from Norwegian youth elite handball players (no. of players=205, no. of health problems=471) to illustrate how assessing the cumulative effect of training load can be done in practice.ResultsDLNM was the only method that accurately modelled the simulated relationships between training load and injury risk. In the handball example, DLNM could show the cumulative effect of training load and how much training load affected health problem risk depending on the distance in time since the training load exposure.ConclusionDLNM can be used to assess the cumulative effect of training load on injury risk.

Causal inference did not detect any effect of jump load on knee complaints in elite men's volleyball

“…We initially calculated a cumulative effect of past jump load with the distributed lag nonlinear model, 28 but terms were inestimable, and we settled for the second-best option: the exponentially weighted moving average. 28…”

“…To handle the many assumptions underlying the complex relationship between jump load and knee pain, 29 we applied a rigid causal inference approach 30 and the latest recommendations on statistical methodology 26,28,31 . The weekly prevalence of knee complaints in the studied cohort was high (31%), 3 and comparable to 36% in high school volleyball players, 1 and 45% of jumper's knee in another study on elite volleyball players 2 .…”

“…Accounting for nonlinearity using the distributed lag nonlinear model caused inestimable models, and we had to use the exponentially weighted moving average. This problem would be resolved with greater sample sizes which allow the use of statistical approaches that can handle nonlinearity 28 …”

Improving statistical methodology in training load and injury risk research (PhD Academy Award)

“…In three studies,4–6 I simulated hypothetical relationships between training load and the probability of injury, using 1-season cohort data from Norwegian Men’s Premier League football (n=39 players) and Norwegian academy U-19 football (n=81 players). Ideally, a study uses multiple measures to capture different aspects of training load.…”

“…In the simulated relationships, I construed various scenarios based on current hypotheses circulating in the field; scenarios of (1) missing data,4 (2) non-linearity5 and (3) time-lagged associations of long-term training load 6. I compared different statistical approaches in their ability to detect the simulated relationships under different conditions.…”