Establishing Clinically Distinct Patient Treatment Subgroups Following Anterior Cruciate Ligament Reconstruction: A...

Area of Science:

• Orthopedic surgery
• Biomedical data science
• Sports medicine

Background:

• Anterior cruciate ligament (ACL) injuries present complex treatment decisions influenced by return-to-sport desires and instability.
• Personalized treatment strategies are needed to optimize outcomes after ACL injury.

Purpose of the Study:

• To identify distinct patient subgroups following ACL injury using unsupervised machine learning.
• To evaluate the differential treatment effect of ACL reconstruction (ACLR) versus nonoperative management on patient outcomes, including posttraumatic osteoarthritis (PTOA).

Main Methods:

• A longitudinal registry study analyzed 923 patients with ACL injuries (1990-2016) with >7.5 years follow-up.
• An unsupervised random forest algorithm identified two distinct patient subgroups.
• Machine learning causal inference estimated treatment effects of ACLR versus nonoperative management.

Main Results:

• Two subgroups emerged: an optimal outcome group (n=653) and a suboptimal outcome group (n=270).
• The suboptimal group (older, higher BMI) had higher rates of secondary meniscal injury, PTOA, and total knee arthroplasty (TKA).
• ACLR significantly reduced PTOA and TKA risk in both groups, but was less effective for secondary meniscal and contralateral ACL injuries in the suboptimal group.

Conclusions:

• Two clinically relevant subgroups with differential responses to ACLR were identified.
• ACLR effectively reduces PTOA and TKA progression across subgroups.
• Patient age, BMI, and meniscal injury status influence ACLR effectiveness for preventing secondary injuries.