No single winner in chronic low back pain — dose and adherence may matter more

The question every clinician with a chronic low back pain caseload eventually asks: which exercise type actually works, and how much of it do I prescribe?

Mind-body exercise beats control, but not each other

Xiang et al. (2026) pooled trials of mind-body exercise — yoga, Pilates, Tai Chi, qigong — in adults with chronic nonspecific low back pain. Compared with control conditions, MBEs reduced pain on a 0–10 cm VAS with a mean difference of −1.56 cm (95% CI −1.98 to −1.15). In the network meta-analysis, all four modalities estimated lower pain than control, but no between-modality differences were detected.

Before you conclude that any of the four is interchangeable, note the authors' own caveats: heterogeneity was substantial (I² 92.7%), the 95% prediction interval ranged from −3.69 to 0.56 cm, and the certainty of evidence was low for all network comparisons assessed. The direction is consistent; the comparative precision is not.

Walking has a dose–response — and it's U-shaped

Núñez-Cortés et al. (2026) ran a meta-analysis of 13 randomised trials of walking programs in patients with chronic low back pain. Walking reduced pain intensity (SMD −0.59, 95% CI −0.95 to −0.24) and disability (SMD −0.69, 95% CI −1.14 to −0.23), with low and moderate certainty of evidence respectively.

The interesting part is the spline model: a U-shaped association between total minutes of walking and reductions in both outcomes. The largest effect for pain came at a total of 1000 minutes distributed over a mean of 8 weeks (SMD −0.67, 95% CI −0.76 to −0.59); for disability, the peak was at 900 minutes (SMD −0.61, 95% CI −0.69 to −0.53). More was not automatically better — the association was U-shaped, not linear.

A dose prescription, with heavy caveats

Arora et al. (2026) attacked the dose question directly with a dose–response network meta-analysis across adults with chronic low back pain disorders. The model-predicted optimal dose for short-term disability was 4 sessions per week of 25 minutes over 13 weeks; for short-term pain intensity, 3 sessions per week of 30 minutes over 8 weeks. Exercise at low to moderate doses was associated with short-term improvements in disability and pain that may be clinically meaningful, with limited additional benefit at higher exposure.

The authors' limitations deserve equal billing: credible intervals of most dose levels overlapped, few dose combinations were clinically meaningful beyond the short term, 84% of trials were at high risk of bias, and certainty was very low to low. Treat those numbers as a starting prescription to titrate, not a target to defend.

Head-to-head trials: adding things often doesn't add much

Where trials do compare variants head-to-head, the differences are rarely dramatic. In a pilot randomised trial of 28 adults with chronic nonspecific low back pain, Alsobayel et al. (2026) found a significant between-group difference in pain reduction favouring VR exercise over conventional home exercise at two weeks (p = 0.044) — but no significant differences for disability, fear-avoidance beliefs, or physical performance outcomes. The authors themselves note it had little impact on fear-avoidance beliefs or disability, highlighting the need for multimodal approaches addressing broader biopsychosocial factors.

In a related line of work, Tomazoni et al. (2026) randomised 148 people with chronic nonspecific low back pain to photobiomodulation plus motor control exercise versus sham plus the same exercise: the intention-to-treat analysis showed a negligible effect at 6 weeks for pain, disability or PGE2, with negligible between-group differences across 12 months of follow-up.

Baek et al. (2026) compared treadmill-based programs in 45 patients with chronic mechanical low back pain: both treadmill groups improved more than controls in TUG performance, gait speed, pain and depressive symptoms, and the vertical-unloading-plus-treadmill group beat treadmill exercise alone on TUG, gait speed, disability and repositioning error. Pain and depressive symptom improvements were maintained at follow-up in both treadmill groups. The authors caution the sample is small and they could not distinguish vertical unloading from harness support.

Five modalities beat exercise plus medication — in one study

Koshti et al. (2026) randomised 123 patients with chronic nonspecific low back pain to a two-modality intervention (exercise and medication) versus a five-modality holistic program (behavioural therapy, exercise, nutritional optimisation, medication and posture training). At six months, the between-group difference in mean VAS improvement (3.65 vs 1.55) was statistically significant (p < 0.001). Subgroup analysis showed patients with higher baseline anxiety improved significantly more with the holistic approach. One study, six-month follow-up, exploratory subgroups — worth noting, not worth rebuilding your clinic around.

The 45% problem

Here is the number that reframes everything above. In a cross-sectional study of 491 adults with chronic nonspecific low back pain, Giraudo-Guglielmi et al. (2026) found adherence to home-based exercise therapy was 45%. In adjusted multivariate analysis, higher pre-cLBP physical activity, lower pain intensity and lower treatment burden were associated with higher adherence (OR 2.66 [1.77–4.00], 1.01 [1.00–1.02] and 0.95 [0.94–0.96] respectively). Notably, they were unable to demonstrate a significant influence of supervision on adherence.

That is an association from cross-sectional data — greater treatment burden and higher pain travel with lower adherence, not proof they cause it. Adherence to home-based exercise therapy was reported at only 45% in that cohort; separately, Arora et al.'s model-predicted optimal dose for short-term disability was 4 sessions per week of 25 minutes.

What to do with this on Monday

A reasonable synthesis for your own practice:

1. Stop agonising over modality choice. For mind-body exercise, the review found reduced pain versus control conditions, with no statistically significant differences between modalities. Pick what the patient will sustain.

2. Prescribe a dose, not a program. In the dose–response network meta-analysis, low-to-moderate doses captured most of the short-term effect — the model-predicted optimum was 3–4 sessions per week of 25–30 minutes — with limited additional benefit at higher exposure. Separately, a walking meta-analysis found the largest effects at around 900 (disability) to 1000 (pain) total minutes over about 8 weeks. Overlap in the intervals means these are anchors, not laws.

3. Budget adherence before you budget volume. Adherence to home-based exercise therapy was only 45%, and a lower burden of treatment was associated with better adherence.

4. If the exercise plan isn't moving pain and disability together, look beyond it — the multimodal and biopsychosocial signals appear in both the Koshti study and the VR trial's authors' own conclusions.

A recent qualitative synthesis of 150 RCTs in chronic low back pain notes that most management strategies share common elements and that combining key elements may prove crucial, while stating that this result is not decisive for treatment, and a recent consensus statement lands on individualised, stepwise decision-making tailored to the patient in front of you.

If you want a practical way to track actual session completion and dose delivered between visits rather than relying on recall at review, that is precisely the gap we built PhysioSync to fill — the evidence above tells you the dose matters; measuring it is on you.