Search for stretches to avoid with back pain and you will find a dozen confident lists, most of them the same five movements, none of them written by anyone who has seen you move. This page does not give you that list, and the reason is not caution — it is a randomised trial showing that the direction which helps differs between people, so a universal list of forbidden movements is wrong in principle rather than merely unhelpful. What follows is what can honestly be said, where the famous "never round your back" claim came from, and a gap in our own product that belongs on our own page.
The short answer
- Which movements provoke symptoms is individual. A trial showed matching the direction mattered and mismatching was worse than nothing.
- So a universal avoid-list is wrong in principle. Nobody can write one for a stranger.
- The famous "never flex your spine" evidence is pig spines in a materials-testing rig.
- What can be said is a set of situations, not a set of movements.
- Our own lower-back contraindication flag currently removes no movement at all. That is a gap and we would rather print it.
Read this first
Get examined rather than deciding which stretches to skip if you have loss of bladder or bowel control, difficulty passing urine, or numbness in the saddle area alongside back pain — that combination needs assessment the same day. Also: progressive leg weakness, back pain after significant trauma, back pain with fever or unexplained weight loss, a history of cancer, or pain that is constant and worse at night. Symptoms travelling below the knee, with or without numbness, deserve assessment before any routine — sciatica and stretching explains why that case is different.
Nothing here diagnoses anything and nothing here is treatment. If a specific movement is the question, a clinician who can examine you will answer it in one appointment.
Why a universal list is wrong
Start with the finding that settles it.
Long, Donelson and Fung, in Spine in 2004, examined patients with low back pain for a directional preference, then randomised them to exercises matching that direction, exercises opposing it, or non-specific exercises. The matched group did best on pain and medication use. The opposed group did worst — worse than the non-specific one.
Read that as a statement about avoid-lists. Direction is individual. The movement that eases one person's symptoms is the movement that provokes another's, and giving somebody the wrong direction is worse than giving them nothing in particular.
Grade: reasonably solid, and narrow. One well-designed randomised trial in a population selected for showing a preference, unblinded by necessity. It does not prove that every person has a direction. It does establish that direction is not universal, which is all an avoid-list needs to be wrong.
McKenzie method vs general stretching covers the broader approach that grew out of this, and the honest verdict on it: better than doing nothing, and no clear winner against other active care.
So when a page tells you never to touch your toes, never to twist, or never to do a seated forward fold, it is generalising from the writer's own back or from a mechanism. Which brings us to the mechanism.
The pig-spine study
The most-cited justification for "never round your lower back" is a real piece of research and it is not about you.
Callaghan and McGill, in Clinical Biomechanics in 2001, took porcine cervical spine segments, mounted them in a testing rig, applied a compressive load and cycled them through repeated flexion and extension — thousands of repetitions. Disc herniation was produced. It is a careful experiment and it is the empirical basis for a great deal of the "flexion is dangerous" advice in circulation.
Now the caveats, which are rarely quoted alongside it.
It is pig spine, cervical segments used as a lumbar analogue. It is a machine, applying loads and cycle counts no human produces in a stretching routine. It is thousands of loaded cycles, not one held position. And it is a tissue-failure experiment, which is a different question from whether a movement helps or hurts a person with pain.
Grade: solid as biomechanics, absent as a clinical recommendation. The experiment shows what it shows. The leap from it to "do not round your back to tie your shoes" is a leap somebody else made.
The clinical evidence does not support that leap. The trials of exercise for low back pain — Hayden and colleagues' 2021 Cochrane review, Saragiotto and colleagues' 2016 review of motor control exercise — find modest benefits and no consistent advantage for any type, including the flexion-avoiding ones.
The stretches to avoid with back pain
So what can be said? Situations, not movements.
When a movement reproduces symptoms travelling down a leg. Shooting, electric or burning sensations, or numbness following the limb, are not the pulling sensation the stretching literature is built on and do not respond to the same rule. Ease out rather than working through. This is the one situation where the advice is close to universal.
When you are in the first hour or so of the day, and the movement is loaded flexion. Adams, Dolan and Hutton found in Spine in 1987 that the lumbar spine carries higher bending stresses early in the day, when the discs are most hydrated. That is a laboratory finding rather than a demonstrated harm in living people, so it is a reason for care rather than a prohibition. Morning back stiffness covers it.
When the movement is a test rather than a stretch, and today is a bad day. A toe touch is a loaded forward fold. If bending forward is what provokes your symptoms, taking a reading is not worth what it costs. A reading you had to hurt for is not a reading worth having.
When a direction reliably makes things worse over hours rather than seconds. The useful signal is not what a movement feels like during it, but what your symptoms do in the hours after. Record that over a couple of weeks and you have something a clinician can work with, which is more valuable than any list.
Notice that none of those is a named movement. That is the point.
The flag that removes nothing
A page criticising other people's avoid-lists should audit its own product, so here is that audit.
Limber's engine carries a hard contraindication filter. Every movement in the catalogue can carry flags; every flag you have active removes the matching movements from the pool before selection rather than ranking them lower. A test generates 500 routines with each flag set and asserts that no flagged movement appears in any of them. That is a real mechanism and it is the feature reviewers of other apps ask for most often.
The onboarding asks which areas are sore and offers three: the lower back, the neck and the knees. Each writes the matching flag.
And here is the gap. Across the 56-movement catalogue, the neck flag is carried by two movements and the knee flag by two. The lower-back flag is carried by none. Setting it removes nothing, because no movement in the catalogue has been marked as one to withhold from someone with an irritable lower back.
That is not a safety failure — the separate red-flag gate in onboarding stops the flow entirely for anyone reporting sharp pain, numbness or tingling, a recent injury or post-surgery recovery, and it is not dismiss-to-continue. But a flag that filters nothing is a promise with nothing behind it, and this is the page where saying so belongs.
There is a second, related honesty. The app offers "ease an achy area" as a goal, and that goal steers the plan toward no score group in particular — it falls through to whatever reads lowest. It does not bias the session toward the area you named. Both of these are ours to fix and neither is something a reader should have to discover.
If you want a reading, the toe touch test reads the posterior chain in five bands with the forward-fold caveat above, and the deep squat test reads the whole lower body over a 30-second hold. Together they are the entire lower-back score in Limber, half each, both labelled as proxies. The positions themselves are in lower back stretches, and the cluster sits under lower back.
Questions
Which stretches should I avoid with back pain?
There is no list that applies to everybody, and that is a finding rather than a hedge. In a randomised trial, people given exercises matching their own directional preference did best and people given the opposing direction did worse than those given non-specific exercises. Since the helpful direction differs between people, any universal avoid-list will be wrong for a substantial share of readers.
Is touching your toes bad for your back?
Not as a general rule, and it is a genuine test of one direction. A forward fold loads spinal flexion and puts the sciatic nerve on tension, so it is a poor idea on a day when bending forward provokes your symptoms or when there are leg symptoms. The blanket prohibition usually rests on a laboratory experiment on pig spines cycled thousands of times in a machine, which does not transfer to a person holding a position.
Should I avoid rounding my lower back?
The evidence for that advice is thinner than its confidence suggests. It comes largely from a biomechanics experiment in which porcine spinal segments were compressed and cycled through repeated flexion until failure — careful work, and a long way from a person doing a stretch. The clinical trials of exercise for low back pain find no consistent advantage for flexion-avoiding approaches over any other kind.
Are twists dangerous for the lower back?
There is no good evidence that a gentle unloaded twist is dangerous, and one is in our own back track — the supine twist, held for 45 seconds a side, appears on eight of the twelve days. What matters is the same thing that always matters: whether it reproduces symptoms, particularly anything travelling down a leg, and what your symptoms do in the hours afterwards.
How do I know if a stretch is making things worse?
Watch the hours rather than the seconds. A pulling sensation during a stretch that settles when you come out of it is the ordinary experience the whole stretching literature is built on. Symptoms that are worse for hours afterwards, symptoms that travel further down the leg than before, or any new numbness or weakness are different signals, and the last two are reasons to see a clinician who can examine you.
Does an app know which movements to skip for me?
Only what you tell it, and less than you might assume. Limber removes flagged movements from a session before it is built and proves it in test, which is a real mechanism. But the lower-back flag it offers in onboarding currently matches no movement in the catalogue, so setting it removes nothing. We would rather print that here than have it found by somebody else.