Rehabilitation after lumbar spine surgery in adults: a systematic review with meta-analysis.

Only 8 of the 45 included studies had low risk of bias; 21 had high risk of bias, mainly because randomisation and allocation concealment were poorly described. Patients and therapists could not be blinded to the exercise intervention in 27 studies, so self-reported outcomes may be biased. The rehabilitation programmes were highly heterogeneous in type, duration, frequency, and intensity, making it impossible to identify which specific exercises or dosages drive the effect. Only the disc herniation subgroup could be meta-analysed; stenosis and spondylolisthesis rest on one and three studies respectively. Most comparisons involved fewer than 400 participants, and the supervised-versus-no-treatment comparison had only 2 trials with 166 participants. Grey literature was excluded, so unpublished trials with null or negative results may be missing. The authors note that in 10 studies, missing outcome data likely biased the results.

Do not read the supervised-exercise benefit as settled. The evidence is low to very low quality, 21 of 45 included trials had high risk of bias, and the pain difference of 1.14 points is small enough that a patient may not notice it. The spondylolisthesis and stenosis findings rest on one and three studies respectively, so they should not be generalised to a broader post-surgical population.