Exercise helps, modestlyStrong evidence
In a Cochrane review of 249 trials, exercise reduced pain by about 15 points on a 0 to 100 scale compared with no treatment, usual care or placebo (moderate certainty), and improved function by about 7 points, below the 10 points the authors set for a difference people would notice. Against other treatments the gains were smaller (about 9 points for pain and 4 for function). Side effects were mostly minor, such as muscle soreness.
In plain English: exercise takes a real but modest edge off long-term back pain. It won't usually make it vanish, and it works best when you keep it up.
Source 7
Every guideline puts exercise firstStrong evidence
NICE: consider a group exercise programme (biomechanical, aerobic, mind-body or a mix), chosen to suit your needs and preferences. The WHO (December 2023): a structured exercise programme may be offered, a conditional recommendation on low-certainty evidence. The American College of Physicians lists exercise, multidisciplinary rehabilitation, mindfulness, tai chi, yoga and motor control exercise among first choices. A US physiotherapy guideline says physiotherapists should use exercise, including trunk strengthening and endurance.
Sources 1, 2, 4, 5
No type of exercise is clearly bestStrong evidence
The WHO found too little evidence that one type of exercise is better than another and advises choosing by preference and circumstances. Cochrane reviews found motor control ("core") exercise and Pilates no better than other exercise, and yoga about the same as other back exercise.
Sources 2, 12, 13, 14, 16
The rankings disagreeMixed evidence
One network meta-analysis (217 trials) ranked Pilates, McKenzie therapy and functional restoration highest. Another (89 studies) ranked Pilates for pain and resistance and motor control training for function, and found stretching and McKenzie no better than control. A third (118 trials) found every type helped except stretching for pain and McKenzie for disability, with Pilates most likely best.
Pilates keeps coming out near the top, but the trials are small, and the differences between types are smaller than the difference between exercising and not.
Sources 8, 9, 10
Pick a lane you'll keep doing
The core moves in the plan suit everyone. Add one of these, about twice a week:
- Pilates: wall bridge, chest lift, swimming (or bird dog if arching hurts) and spine stretch forward, or follow the Pilates for back pain routine. Twice a week beat once a week in one trial.
- Strength: Romanian deadlifts, step-ups, chest-supported rows and suitcase carries.
- Walking: brisk intervals, adding hills or stairs once 30 minutes is easy, as the WalkBack programme progressed; rucking later.
- Mind-body: a beginner yoga or tai chi class; at home, cat-cow, child's pose, sphinx and a lying spinal twist.
- Water: swimming or aqua classes.
Sources 14, 15, 5, 9, 18, 19, 20, 16, 4, 21
How much, and howEmerging evidence
Programmes that were individually designed, supervised (or at least followed up) and higher in dose did better. The WHO adds that programmes with at least 20 hours of total programme time are generally more beneficial, and that group formats may help attendance.
In one network meta-analysis, the most helpful programmes included at least 1 to 2 sessions a week of Pilates or strength exercise, sessions under 60 minutes, and Pilates or core programmes of 3 to 9 weeks. In a trial of 296 people, Pilates twice a week beat once a week, and three times a week did no better than twice.
Sources 11, 2, 10, 15
Walking helps about as much as other exerciseEmerging evidence
Walking improved pain, disability and quality of life about as much as other exercise in a review of 5 trials. After recovery, a progressive walking and education programme (six physiotherapist sessions over 6 months, aiming for at least 30 minutes five times a week) delayed the next episode that limited activity: a median of 208 days against 112. The walking group reported more leg-related side effects (100 against 54).
Sources 18, 19, 20
Yoga: about the same as other exerciseEmerging evidence
Yoga gave small improvements over no exercise, too small to matter much clinically, and about the same as other back exercise. It caused more side effects than no exercise, mostly more back pain. In one trial, yoga classes did no better than stretching classes over 12 weeks.
Sources 16, 17
Water exerciseEmerging evidence
Exercise in water reduced pain and improved function in 8 small trials, and the NHS lists swimming among the activities that may help.
Sources 21, 23
Beyond exerciseEmerging evidence
Cognitive functional therapy (up to 7 physiotherapy sessions over 12 weeks plus a booster) reduced activity limitation by 4.6 points on a 24-point scale more than usual care at 13 weeks, with similar effects at a year, at lower cost. NICE withdrew its recommendations on psychological therapy and combined physical and psychological programmes in July 2026 because they may be out of date, without replacing them; the WHO conditionally supports CBT and operant therapy.
Sources 22, 1, 2
What not to rely onStrong evidence
NICE: no belts or corsets, foot orthotics, rocker-sole shoes, traction, acupuncture, ultrasound, TENS, PENS or interferential therapy; manual therapy only alongside exercise; no paracetamol on its own, opioids, antidepressants or gabapentinoids for chronic back pain; no spinal injections; spinal fusion only in trials.
The WHO advises against routine traction, ultrasound, TENS, lumbar braces, opioids and several other medicine groups, and conditionally supports anti-inflammatories (NSAIDs), spinal manipulation, massage and needling.
Sources 1, 2