Exercise eases stubborn low back pain, and desk workers may gain the most

Low back pain is one of the leading causes of disability worldwide, and exercise therapy is the treatment guidelines recommend most consistently for the persistent, non-specific kind. The awkward truth is that exercise is only moderately effective on average, and averages hide a lot: some people improve substantially while others barely move. If researchers could work out in advance who responds best, treatment could be matched to the person rather than prescribed uniformly. This project set out to answer that question at a scale no single trial can manage.

Working in parallel with a Cochrane review update, the team requested the original individual participant data from high-quality randomised trials of exercise for persistent low back pain. They assembled a master dataset of 27 trials and 3514 adults, covering baseline characteristics, the details of each exercise and comparison programme, and outcomes for pain, functional limitation and global recovery at short, moderate and long-term follow-up. They then ran a one-stage meta-analysis using multilevel regression, testing both the overall treatment effect and a prespecified list of characteristics that might modify it.

Compared with no treatment or usual care, exercise therapy reduced pain by 10.7 points on a 100-point scale, a result compatible with a clinically important 20% smallest worthwhile effect. Functional limitations fell by 10.2 points per 100 at short-term follow-up, a clinically important 23% improvement. Three characteristics stood out as potential effect modifiers: not having heavy physical demands at work, using medication for low back pain, and a lower body mass index were each associated with better outcomes from exercise relative to non-exercise comparisons. The authors flag that the availability and measurement of participant characteristics was inconsistent across trials.

The headline result confirms what the field already believed: exercise works for persistent low back pain, moderately and reliably, rather than dramatically. The novel contribution is the subgroup analysis, which is among the first serious attempts to identify who benefits most using raw data from thousands of participants rather than trial-level summaries. The modifiers found here are described as potential, not proven, because the underlying trials did not measure participant traits consistently. Whether matching exercise programmes to these subgroups actually improves outcomes is the obvious next trial to run.

If you have back pain that will not shift, this is a reason to keep exercise on the table: across 27 trials it beat doing nothing on both pain and day-to-day function. The sizing is honest rather than thrilling, so expect meaningful improvement, not a cure. The tentative finding that people without heavy physical jobs did better is worth knowing if you work at a desk, though it should not put anyone with a manual job off. The right programme and progression are conversations for a qualified professional.