EFFECTIVENESS OF ARTIFICIAL INTELLIGENCE-ASSISTED EXERCISE PRESCRIPTION COMPARED WITH CONVENTIONAL PHYSIOTHERAPY FOR CHRONIC LOW BACK PAIN: A RANDOMIZED CONTROLLED TRIAL
Abstract
Background: Chronic low back pain (CLBP) is a leading cause of years lived with disability worldwide, and individually tailored exercise therapy is a cornerstone of conservative management. Artificial intelligence (AI)-assisted exercise prescription platforms, which use algorithmic movement analysis and adaptive programming, have been proposed as a means of improving exercise precision, adherence, and outcomes relative to therapist-prescribed conventional physiotherapy, but head-to-head evidence remains limited.
Objective: To compare the effectiveness of a 12-week AI-assisted exercise prescription program with conventional physiotherapist-led exercise therapy in adults with chronic non-specific low back pain, in a real-world health and wellness rehabilitation center setting.
Methods: In this assessor-blinded, parallel-group randomized controlled trial, 120 adults aged 18–65 years with non-specific CLBP of at least three months' duration were randomized 1:1 to a 12-week AI-assisted exercise prescription program (n = 60) or conventional physiotherapy (n = 60). The primary outcome was change in the Oswestry Disability Index (ODI) from baseline to 12 weeks. Secondary outcomes included pain intensity (0–10 Numeric Pain Rating Scale), lumbar range of motion, exercise adherence, and patient satisfaction. Assessments were conducted at baseline, 6 weeks, and 12 weeks.
Results: In this illustrative dataset, both groups showed statistically and clinically meaningful improvement in disability and pain over 12 weeks. The AI-assisted group demonstrated a significantly greater reduction in ODI score (mean difference 6.8 points, 95% CI 3.9–9.7, p < 0.001) and higher exercise adherence (86.4% vs. 71.2% of prescribed sessions, p = 0.002) compared with the conventional physiotherapy group. Pain reduction and lumbar range-of-motion gains favored the AI-assisted group but did not consistently reach statistical significance for all secondary measures. No serious adverse events were recorded in either group.
Conclusion: In this trial template and its illustrative results, AI-assisted exercise prescription produced outcomes that were at least equivalent to, and on select measures superior to, conventional physiotherapist-led exercise therapy for chronic low back pain, with meaningfully higher adherence. These findings, once replicated with genuine trial data, would support AI-assisted prescription as a viable adjunct or alternative within multidisciplinary rehabilitation pathways, while highlighting the continued importance of clinician oversight.
