Which specific modes of exercise training are most effective for treating low back pain? Network meta-analysis

Key points

1. 01Pilates had the highest probability of being the best exercise for reducing pain (SUCRA 100%).
2. 02Stabilisation/motor control and resistance training ranked highest for improving physical function (both SUCRA 80%).
3. 03Resistance and aerobic training ranked highest for improving mental health (both SUCRA 80%).
4. 04Doing no specific intervention (true control) was consistently the least effective option across all outcomes.
5. 05Stretching and McKenzie exercises did not clearly outperform doing nothing for pain or function.

How it was conducted

Design

Systematic review and network meta-analysis of randomised controlled/clinical trials (PRISMA-NMA, PROSPERO CRD42017068668)

Participants

Adults aged 18 and over with non-specific chronic (12 weeks or longer) low back pain

Data sources

MEDLINE, CINAHL, SPORTDiscus, EMBASE, CENTRAL

Interventions compared

Exercise modes including Pilates, resistance, stabilisation/motor control, aerobic, yoga, McKenzie, stretching, water-based and multimodal, versus true control, therapist hands-on and therapist hands-off controls

Outcomes

Pain, physical function, mental health and trunk muscle strength, ranked by standardised mean difference and SUCRA

Evidence rating

Quality of evidence graded using GRADE

What they found

• Pain: Pilates ranked best (SUCRA 100%; pooled SMD -1.86, 95% CI -2.54 to -1.19).
• Physical function: resistance training (SUCRA 80%; SMD -1.14, 95% CI -1.71 to -0.56) and stabilisation/motor control (SUCRA 80%; SMD -1.13, 95% CI -1.53 to -0.74) ranked best.
• Mental health: resistance training (SUCRA 80%; SMD -1.26, 95% CI -2.10 to -0.41) and aerobic training (SUCRA 80%; SMD -1.18, 95% CI -2.20 to -0.15) ranked best.
• For pain, Pilates (p<0.001), aerobic (p=0.006), stabilisation/motor control (p<0.001), multimodal (p<0.001), resistance (p=0.002) and 'other' (p<0.001) all reduced pain versus true control.
• True control was most likely to be worst across all outcomes (SUCRA 10% or lower), followed by therapist hands-off control for pain and function.
• Stretching and McKenzie exercise effect sizes did not differ from true control for pain or function (p>0.095; SUCRA below 40%).

Limitations

• The quality of the synthesised evidence was rated low according to GRADE criteria.
• There was substantial heterogeneity among the included studies.
• Few of the included studies had a low risk of bias.
• Some outcomes were based on small numbers of studies (for example only 16 studies for mental health and 4 for trunk muscle strength), and network meta-analysis was not possible for trunk muscle endurance or analgesic medication.

Why it matters

For patients

If you have ongoing back pain, structured exercise such as Pilates, core stabilisation or strength training is likely to help more than passive treatment or doing nothing.

For clinicians

Exercise prescription should be tailored to the primary goal, with Pilates favoured for pain, stabilisation and resistance for function, and resistance or aerobic work for mental health, while recognising the low certainty of evidence.

For readers

This network meta-analysis ranks exercise modes against each other rather than just against no treatment, but the low evidence quality means rankings should be read as guidance, not proof of one best exercise.