Exercises for Lower Back Pain: What Works, What to Skip, and How a Trainer Approaches It

Roughly 4 in 10 American adults reported back pain in the previous three months, according to the CDC’s 2019 National Health Interview Survey. For nearly everyone with low back pain, no scan will ever pin the pain on one structure, and the treatment with the strongest evidence is not a device, a belt or a single magic stretch. It is exercise, done regularly, for months. This guide covers what the research says works, what to skip, and how a trainer actually builds a program for someone whose lower back hurts.

This is general information from published research and clinical guidelines, not medical advice. If anything in the red flag list below applies to you, see a doctor before you exercise.

The short version

• Most low back pain is “nonspecific”: the World Health Organization puts it at 9 in 10 cases seen in primary care. Acute episodes usually improve a lot within six weeks, but recurrence is the norm, not the exception.
• Exercise is the first-line treatment for persistent low back pain in the American College of Physicians, NICE, Lancet and WHO guidelines. The largest review, 249 trials, found it cuts pain by about 15 points on a 100-point scale versus no treatment or usual care.
• No single exercise style wins. Pilates, resistance training, motor control work and general exercise all beat doing nothing in the big reviews (McKenzie did in one and not the other); the differences between types are smaller than the difference between exercising and not. Pick what you will keep doing.
• A walking program nearly doubled the median time to the next flare-up (208 days versus 112) in a 701-person trial. Build toward 30 minutes, five days a week, over months.
• Skip belts, traction, TENS, bed rest and anything promising a fix in 30 seconds. Posture, sitting, a “weak core” and lifting with a rounded back are not the proven causes the internet says they are.

First: is it safe to exercise?

For most people, yes, and sooner than they think. Cochrane’s review of bed rest versus staying active found small but real advantages for staying active during an acute episode (Dahm and colleagues, Cochrane). The Lancet’s 2018 low back pain series lists bed rest among the things not to do (Foster and colleagues, Lancet, 2018).

The exception is the short list of warning signs that mean the pain needs a doctor first. The UK’s NHS list, in full (NHS, back pain):

• Emergency care now if back pain comes with pain, tingling, weakness or numbness in both legs; loss of feeling around the genitals or anus; changes in your bladder or bowels such as difficulty peeing or losing control; changes in how your penis or vagina feels during sex, including not being able to get or keep an erection or to orgasm; chest pain; or it started after a serious accident.
• Urgent appointment if you feel hot, cold, shivery or generally unwell, or the pain started suddenly and severely or is getting worse quickly.
• See a doctor if it is not improving after a few weeks of home care, it is stopping your day-to-day activities, you are worried about it or struggling to cope, you have lost weight without trying, there is a lump or change in the shape of your back, it is worse at night or does not ease with rest, it is worse when you sneeze, cough or go to the toilet, or the pain is between the shoulder blades rather than low down.

Outside that list, the guidelines are consistent: no routine imaging. NICE says do not routinely offer imaging in a non-specialist setting (NICE NG59), and the American College of Physicians’ Choosing Wisely statement is “don’t obtain imaging studies in patients with nonspecific low back pain” (Choosing Wisely compilation). Your doctor makes that call, not your trainer, but an MRI is not the first step for most people.

What the research says works

Exercise, in almost any form

The 2021 Cochrane review of exercise therapy for chronic low back pain pooled 249 trials and 24,486 participants. Against no treatment, usual care or placebo, exercise reduced pain by 15.2 points on a 0 to 100 scale, which met the review’s own threshold for a clinically important change; the gain in day-to-day function was 6.8 points, below the review’s 10-point threshold (Hayden and colleagues, Cochrane, 2021). Half the programs were one-on-one, half were group classes.

That is the headline. The American College of Physicians lists exercise first for chronic low back pain, ahead of any medication (Qaseem and colleagues, Annals of Internal Medicine, 2017). NICE recommends a group exercise program and says manual therapy and talk therapy should only be offered alongside exercise (NICE NG59). The WHO’s first guideline on chronic low back pain, December 2023, recommends exercise programs and education (WHO, 2023). One nuance worth knowing: for a brand-new acute episode, the ACP lists heat, massage, acupuncture and manipulation rather than exercise, and the Lancet series says early supervised exercise is usually unnecessary. The trainer’s job is strongest once pain has lasted more than a few weeks, keeps coming back, or you want to stop the next episode.

Which kind of exercise

Two network meta-analyses have tried to rank exercise types, and they partly disagree. The 2021 analysis of 217 trials put Pilates, McKenzie therapy and functional restoration at the top for pain, and Pilates, McKenzie and flexibility work at the top for function (Hayden and colleagues, Journal of Physiotherapy, 2021). The 2020 analysis of 89 studies in the British Journal of Sports Medicine ranked Pilates first for pain, resistance training and motor control exercise first for function, and resistance and aerobic training first for mental health, and found stretching and McKenzie no different from control (Owen and colleagues, BJSM, 2020). The BJSM analysis rates its evidence low quality, and the first one’s conclusion is the one to remember: people “should be encouraged to perform the exercise that they enjoy.”

Three more Cochrane findings fill in the picture. Motor control (core stability) exercise beats minimal treatment but is no better than other exercise (Saragiotto and colleagues, Cochrane, 2016). Pilates beats minimal treatment and is not conclusively better than other exercise (Yamato and colleagues, Cochrane, 2015). Yoga is slightly better than no exercise, about the same as other back-focused exercise, with effects below the review’s thresholds for a meaningful change, and back pain was its most common side effect (Wieland and colleagues, Cochrane, 2022).

Walking

The single most useful trial for a desk worker is WalkBack, published in the Lancet in 2024. 701 adults in Australia who had just recovered from an episode of low back pain were randomized to an individualized, progressive walking and education program (six physiotherapist sessions over six months, mostly by phone or video) or to nothing. The walking group went a median 208 days before their next activity-limiting flare-up versus 112 days for the control group, a 28 percent reduction in risk (hazard ratio 0.72) (Pocovi and colleagues, Lancet, 2024). The goal was about five 30-minute walks a week by the end of six months, per the authors’ own summary (trial authors, The Conversation). Two things to note: participants averaged 54 years old and 81 percent were women, and this was prevention in people who had recovered, not treatment of pain that is there right now.

Supervision and design help, a little

A 2005 analysis of 43 trials found individually designed programs added about 5 points of pain relief over home exercise alone, supervision added about 6, and programs combining stretching and strengthening did best (Hayden and colleagues, Annals of Internal Medicine, 2005). Older data, modest numbers, and that is the honest size of what a trainer adds to the exercise itself: a program you will actually follow, scaled correctly, with someone watching.

How a trainer approaches it

A good first session for someone with back pain is mostly questions and a few careful movements, not a workout. In practice that means four things.

1. Screen. The red flag list above, plus what your doctor or physical therapist has already said. Trainers do not diagnose; they refer out when something does not fit.
2. Find what you can do today. Which positions ease the pain and which provoke it: bending forward, arching back, walking, sitting, lying down. The starting program is built from the tolerable list.
3. Set the pain rule. A review of seven trials in chronic musculoskeletal pain found that exercising into some pain produced slightly better short-term results than keeping everything pain-free, with no difference later on (Smith and colleagues, BJSM, 2017). The review set no numeric ceiling. The working rule most trainers use, and it is a rule of thumb rather than a research finding: mild discomfort during a set is fine; pain that spikes and stays worse for hours afterward means that exercise was too much, so scale it down.
4. Plan for months, not days. Every positive trial above ran for weeks to months with repeated sessions. Two sessions and a sheet of stretches is not a program.

A starting routine

This is the shape of what we build for clients with nonspecific low back pain, in an apartment, a building gym or a park, with little or no equipment. It has three parts and all three matter. Treat the sets and reps as a common starting point, not a prescription; a trainer will move them up or down from your first session.

Part 1: daily walking

Start with what you can do without a flare-up, even if that is ten minutes, and add a few minutes a week. The WalkBack target, about 30 minutes five days a week, is also the general physical activity guideline, so it does double duty. In Manhattan this is the easiest prescription in the world: get off one stop early, take the long way to lunch, walk the park loop on the days you are home.

Part 2: trunk work, three to five days a week

Low-load, high-repetition work that teaches the trunk to hold position while the hips move. The three moves below are the widely taught “big three” from spine researcher Stuart McGill. They are well tolerated and reasonable as a starting point; the trials on these specific three exercises are small, so their support comes from the broader motor control evidence above rather than from a trial of their own.

• Modified curl-up. Lie on your back, one knee bent, hands under the low back to keep its natural curve. Lift the head and shoulders an inch, hold 8 to 10 seconds, lower. Start with 5 to 6 reps.
• Side bridge. On your side, propped on the forearm, knees bent to start (feet stacked once it is easy). Lift the hips so the body is straight, hold 8 to 10 seconds, lower. 4 to 6 per side.
• Bird dog. On hands and knees, reach one arm forward and the opposite leg back without letting the low back twist or sag. Hold 8 to 10 seconds, switch. 4 to 6 per side.
• Add when ready: dead bugs (on your back, opposite arm and leg lowered slowly while the ribs stay down) and glute bridges (8 to 12 slow reps).

Part 3: progressive strength, two days a week

This is the part most back pain routines leave out and one of the two the BJSM analysis ranked best for function. Legs and hips do the work, the trunk holds. Start with bodyweight or light load and add weight slowly.

• Box squat or goblet squat, 2 to 3 sets of 8 to 10. Sit back to a chair or bench if a full squat bothers you.
• Hip hinge, first with a dowel along the spine to learn the pattern, then a light kettlebell deadlift or Romanian deadlift, 2 to 3 sets of 8.
• Suitcase or farmer’s carry, one or two weights, 20 to 40 meters, 2 to 3 trips. Teaches the trunk to resist bending under load, which is what it does all day.
• Row, band or dumbbell, 2 to 3 sets of 10 to 12.
• Hip mobility at the end: a half-kneeling hip flexor stretch and a 90/90 hip rotation, a minute or so each side. Stretching feels good and is a fine finisher; it is not the treatment on its own.

If nothing has changed after six weeks of consistent work, that is worth a conversation with your doctor, not a new set of exercises.

What to skip

• Back belts and lumbar supports. Cochrane found moderate evidence they do not prevent low back pain and unclear evidence for treatment (van Duijvenbode and colleagues, Cochrane, 2008). NICE says do not offer belts or corsets. WHO says lumbar braces, belts and supports should not be routinely offered.
• Traction, TENS, ultrasound. All on NICE’s do-not-offer list; traction is on the WHO list too; the Lancet series calls passive modalities and back supports “generally ineffective.”
• Bed rest. See above.
• An MRI as step one. Without red flags, the guidelines say not to image routinely.
• Stretching as the whole program. In the BJSM analysis stretching alone did not separate from control. Keep it as the warm-up or finisher.
• Anything that fixes your back in 30 seconds. No trial in this article tested a single movement or a sub-minute intervention. The ones that worked ran for months.

Five things people blame that the evidence does not support

• Posture. A review of 41 systematic reviews found an association between some postures and back pain but “no consensus regarding causality” (Swain and colleagues, Journal of Biomechanics, 2020). Sit however you like; change position often. Our posture guide covers what posture work can and cannot do.
• Lifting with a rounded back. A review of 12 studies, 697 people, found no evidence that lifting with a flexed spine raises back pain risk compared with a straight one, for loads up to 12 kg (Saraceni and colleagues, JOSPT, 2020). Heavier loads were not studied, so a trainer still coaches the hinge for a heavy deadlift. The point is to stop fearing normal bending.
• A weak core. Core work helps about as much as any other structured exercise (Cochrane, above), which is the opposite of what you would expect if a weak core were the cause.
• Sitting. A 2022 meta-analysis of 16 longitudinal studies and 100,002 people found sitting time was not associated with developing low back pain, though sitting three or more hours a day was associated with more disability in people who already had it (Alzahrani and colleagues, PeerJ, 2022). And in one cohort of 250 people who had just recovered, sitting more than five hours a day predicted recurrence within a year (da Silva and colleagues, Journal of Physiotherapy, 2019). Sitting does not appear to cause it; once you have it, moving more matters.
• A firm mattress. A 2021 review of 39 papers favored medium-firm over firm, on thin evidence (Caggiari and colleagues, Journal of Orthopaedics and Traumatology, 2021). A mattress is not a treatment.

If you sit at a desk in Manhattan

Three findings worth acting on. First, recurrence: in that Australian cohort, 69 percent of people had another episode within 12 months of recovering, and the older figure of about a third came from a single lower-quality study identified in a 2017 systematic review by the same group. Plan on prevention, not a one-time fix. Second, movement breaks: in a six-month trial of 193 office workers, prompted active breaks or posture changes through the day cut new low back pain to 9 and 7 percent, versus 33 percent in the control group (Waongenngarm and colleagues, Scandinavian Journal of Work, Environment and Health, 2021). One trial, small arms, but the direction is clear: change position often; the exact interval is not established. Third, standing desks: Cochrane found, on very low quality evidence, that they cut sitting by 84 to 116 minutes a day, with no harms seen and no evidence either way on back pain (Shrestha and colleagues, Cochrane, 2018). Use one to move more, not as treatment.

What happened to the old posts

This guide replaces two earlier articles, “Lower Back Pain and Exercise: A Personal Trainer’s Perspective on Safety” and “How to Fix Lower Back Pain: 7 Tips From a Personal Trainer,” which now redirect here. Both leaned on the posture, core weakness and “exercises to avoid” framing that the research above does not support, so the advice has changed, not just the layout.

Working with a trainer on this

Type A Training’s corrective exercise personal training is built for exactly this: nonspecific low back pain that has lasted or keeps returning, after your doctor has ruled out the red flags. Trainers come to your apartment, your building’s gym or a nearby park in Manhattan and parts of Brooklyn and Queens, and the program looks like the one above, scaled to you and progressed every few weeks. For what corrective exercise is and how it differs from physical therapy, see our corrective exercise guide; for more trunk work, see our core exercise guide.

Frequently asked questions

What is the best exercise for lower back pain?

There is no single winner. In the largest reviews, Pilates, resistance training, motor control work and general exercise all beat doing nothing, McKenzie did in one analysis and not the other, and the differences between types are small and uncertain. Walking, separately, cut recurrences. The exercise you will keep doing for months is the best one.

Should I exercise if my back hurts right now?

If none of the red flags apply, staying active beats bed rest in Cochrane’s review. For a fresh acute episode, guidelines favor heat, staying active and normal daily movement over a structured program; the structured program is for pain that persists past a few weeks or keeps returning. Some discomfort during exercise is acceptable; pain that spikes and stays worse afterward is the signal to scale down.

How long until exercise helps?

The trials that showed benefit ran for weeks to months with repeated sessions. Acute episodes improve a lot in the first six weeks on their own (pain scores fell by more than half in a meta-analysis of 33 cohorts), so judge a program by whether episodes get shorter and rarer over months, not by one week.

Is walking enough?

For preventing the next episode, a progressive walking program with a few coaching sessions cut recurrence risk by 28 percent in the 701-person WalkBack trial. For function and strength, add trunk work and progressive loading; the BJSM analysis ranked resistance training and motor control exercise joint highest for function.

Do I need an MRI first?

Not usually. NICE, the American College of Physicians and the Lancet series all say not to image routinely without red flags. Your doctor decides, but it is not the normal first step.