Not straightforward: modelling non-linearity in training load and injury research

Area of Science:

• Sports Medicine
• Biomechanics
• Exercise Physiology

Background:

• Understanding the relationship between training load and injury risk is crucial for athlete well-being.
• Previous studies have often assumed a linear relationship, potentially oversimplifying complex interactions.

Purpose of the Study:

• To investigate the non-linear relationship between training load and injury risk.
• To evaluate methods for accurately modeling non-linear training load-injury risk associations.

Main Methods:

• Analysis of daily training load (session rating of perceived exertion - sRPE) and injury data from elite youth handball and football players.
• Utilized restricted cubic splines in mixed-effects logistic regression to model the sRPE-injury probability relationship.
• Conducted simulations to compare seven different methods for modeling non-linear relationships.

Main Results:

• A J-shaped, non-linear relationship between sRPE and same-day injury probability was identified in elite youth handball players (p<0.001).
• No significant relationships were found in the studied football cohorts.
• Simulations indicated that quadratic models, fractional polynomials, and restricted cubic splines were effective for non-linear modeling.

Conclusions:

• The association between training load and injury risk should be considered non-linear.
• Fractional polynomials and restricted cubic splines are recommended for future research to account for non-linearity.
• A guide for selecting appropriate modeling methods is proposed.