Abstract
Background:
Return-to-sport decisions following anterior cruciate ligament (ACL) reconstruction in adolescents are often based on time since surgery or gross symmetry metrics, which may overlook residual neuromuscular deficits. The Kinetic Asymmetry Index (KAI) provides a phase-specific assessment of lower limb asymmetries during dynamic tasks and has been applied in adult populations. Generalized additive mixed models (GAMMs) have been used to model nonlinear recovery trajectories in adults, enabling individualized recovery profiling (Jordan et al., 2022). However, the use of KAI and GAMM-based modeling in adolescents remains limited. This study adapts Jordan et al.’s framework to adolescent countermovement jump (CMJ) data to support development of clinically useful tools for tracking neuromuscular recovery over time.
Hypothesis:
We hypothesized that KAI values derived from CMJ testing would improve over time following ACL reconstruction in adolescents. We further hypothesized that a generalized additive mixed model (GAMM) would effectively capture the nonlinear recovery trajectory of the eccentric (KAIecc) and concentric (KAIcon) phases, and that patient-specific factors such as gender or graft type might influence outcomes.
Methods:
This retrospective study included a convenience sample of 130 adolescent patients who completed post-operative CMJ testing as part of return-to-play rehabilitation, contributing 157 data sets. Ground reaction forces were collected using in-ground force plates (1200 Hz), and vertical impulse was calculated during the eccentric (braking) and concentric (propulsive) phases of take-off. GAMMs were used to model KAIcon and KAIecc trajectories over time, with fixed effects for gender and graft type, a smooth term for days post-operation, and a random effect for subject. Four models per outcome were compared using AIC and likelihood ratio tests.
Results:
The final KAIcon model included gender (non-significant, p = 0.326) and a significant nonlinear time effect (p = 0.004), explaining 47.2% of the deviance (R²adj = 0.291). For KAIecc, the baseline model with only time and subject effects performed best, with a significant linear time effect (p = 0.041) and higher explanatory power (65.9% deviance explained, R²adj = 0.438). Neither gender nor graft type improved model fit for KAIecc. Substantial inter-individual variability was present in both models.
Conclusion:
KAIcon and KAIecc improved over time following ACL reconstruction in adolescents but exhibited distinct temporal patterns, with nonlinear concentric and linear eccentric recovery. High between-subject variability reinforces the importance of individualized monitoring. Phase-specific KAI indices represent a promising tool for guiding return-to-sport decisions in youth athletes and may enhance current rehabilitation benchmarks by enabling data-driven, time-sensitive recovery tracking.