


Neurological rehabilitation
Neurology and geriatrics
Bryan Littré · Lucas Meilhou
Updated · published
Vertebral compression fracture: a quarter to a third are recognised, and 19.2 % recur within a year. Red flags, imaging, vertebroplasty and exercise.

An osteoporotic vertebral fracture is not rare, it is rarely named: only a quarter to a third are recognised at the time they occur.
19.2%of vertebral recurrence in the year following an incident fracture
What follows covers each of these points in detail, with sources. It is there if you need it.
This topic is taught in a course: Neurology and geriatrics, online with Bryan Littré.
1 / 14
13 article chapters · 56 min in total
Underdiagnosis◔ 4 min
The vertebral body gives way without an impact, under an ordinary load: carrying a shopping bag, coughing, bending.
Suspecting◔ 6 min
No isolated red flag is enough; it is their combination that informs.
Imaging◔ 5 min
The lateral X-ray shows the deformity, not whether it is recent: that is the job of MRI.
Differentials◔ 3 min
The real trap is not confusing the compression fracture: it is not thinking of it at all.
Recurrence risk◔ 4 min
A first vertebral fracture multiplies by 5.1 the risk of having another within the year.
Prevention◔ 5 min
Three levers, and none consists in treating the bone directly: spotting, reducing falls, strengthening.
Vertebroplasty◔ 6 min
Twenty years of vertebroplasty trials give results that contradict one another.
Brace◔ 2 min
The spinal brace is prescribed very widely, but the literature that supports it is thin.
Rehabilitation◔ 5 min
Physiotherapy after vertebral fracture has not demonstrated much, but remains to be conducted well.
Spinal flexion◔ 5 min
Loaded flexion is to be avoided in the recent phase, but lifelong prohibitions are counterproductive.
Clinical cases◔ 4 min
Four cases show the possible trajectories, from the good outcome to the missed diagnosis.
In practice◔ 3 min
Four decisions sum it all up: think of the fracture, know when to stop, know how to act, treat the bone.
Questions◔ 4 min
The useful benchmark is not a fixed duration but a direction: the pain should decrease.
Train on this



Neurological rehabilitation
Bryan Littré · Lucas Meilhou
Next comes practice: the course that teaches this topic, with Bryan Littré.
Our courses on this topic may be covered by DPC or FIFPL, the French continuing-education funding schemes, depending on your profession and situation.
A 76-year-old woman consults for back pain that appeared “just like that”, without a fall, without an awkward movement. It hurts when she gets up from a chair, and is better when she is lying down. The picture resembles ordinary mechanical thoracic spine pain, and when what has just happened is a vertebral fracture, it is recognised at the moment it occurs in only a quarter to a third of cases: the others, no one will name.1
This article treats the vertebral fracture as a reason for consultation: spotting it when nothing signals it, knowing what must interrupt the session, and what can be done afterwards. For the bone disease itself (densitometry, T-score thresholds, anti-osteoporotic treatments, primary prevention), see Osteoporosis and the prevention of fragility fractures.
Three figures that drive management
What the literature establishes on detection, recurrence and mortality.
Sources: clinical recognition, Schousboe 2016, review of the epidemiology of vertebral fractures;1 recurrence at one year, Lindsay 2001, analysis of 2,725 women from the placebo arms of four trials;4 mortality, Bliuc 2009, prospective Dubbo cohort6.
What to take away in one minute
Because they have neither the noise nor the scene of a fracture. There is no spectacular fall, no sudden loss of function, no deformity visible under the skin. There is a back that hurts in someone whose back hurting is considered normal.
The word “fracture” evokes an event. This one is not an event. The osteoporotic vertebral body does not break under an impact: it gives way under a load that healthy bone would have borne without flinching. Carrying a shopping bag, pulling a sheet, coughing, bending to tie a shoelace, sitting down a little heavily: the history often reports nothing more. In some patients there is not even that: the fracture is discovered months later on an X-ray taken for something else.
The reference figure comes from the review by Schousboe published in the Journal of Clinical Densitometry: only a quarter to a third of vertebral fractures are recognised clinically at the time they occur, the others requiring lateral imaging of the spine to be detected.1 This wording must be read exactly: it does not say that two thirds of fractures are asymptomatic, it says that they are not recognised. The nuance changes where the responsibility lies. Many of these patients had pain. They consulted. They were told it was age.
The osteoporotic vertebral fracture is not silent: it is inaudible. It speaks the same language as ordinary thoracic spine pain, in a population where ordinary thoracic spine pain is the norm.
Underdiagnosis does not stop at the clinic. It also runs through the imaging chain. An Irish study reviewed the thoracic CT angiograms performed in patients hospitalised for hip fracture between 2010 and 2017: of the 225 patients whose images were available, 40% had at least one vertebral fracture and 20% had several. Only one radiology report in five mentioned it. And only 24% of patients left hospital with anti-osteoporotic treatment, with no difference between those who had a vertebral fracture and those who did not.2
In other words, in patients already hospitalised for a fragility fracture, whose vertebrae had already been imaged, whose fracture was already on the image, the chain did not reach a conclusion. This is the chain that the physiotherapist can interrupt, because they are often the professional who sees these patients for the longest and most often.
Four measures of the gap between what happens and what is treated
Each stage of the chain loses patients: the fracture occurs, the image exists, the report is silent, the treatment does not start.
Source: Kelly et al. 2021, Journal of Clinical Densitometry, retrospective cohort of 2,122 hip fractures, 225 CT angiograms re-read blind and graded according to Genant.2
The European prospective EPOS study remains the reference measurement on this continent: 14,011 men and women aged 50 and over, recruited in 29 centres, X-rayed and then X-rayed again 3.8 years later on average. The age-standardised incidence of morphometric fractures was 10.7 per 1,000 person-years in women and 5.7 in men, with a marked rise with age in both sexes.3
Two practical lessons are hidden in it. The first is that men are not spared: the ratio is about two to one, not ten to one. A 78-year-old man who complains of back pain deserves exactly the same vigilance as a woman of the same age. The second is that incidence rises steeply with age, which shifts the pre-test probability with each decade: the same clinical picture does not mean the same thing at 60 and at 82.
Incidence of vertebral fractures in Europe
Two methods of reading the same radiographs, two sexes: the order of magnitude is stable.
Source: European Prospective Osteoporosis Study Group 2002, Journal of Bone and Mineral Research; 14,011 subjects aged 50 and over, 29 European centres, paired radiographs 3.8 years apart on average, fracture defined by a height loss of 20% and 4 mm.3
Key points
The honest answer is: not with a single sign. Research over the past twenty years has demolished the idea that an isolated red flag is enough. What remains, and is usable in the clinic, is a way of counting.
The 2023 Cochrane review led by Han and Hancock reviewed fourteen diagnostic studies comparing the history and physical examination with reference-standard imaging. Its conclusion is frank: most red flags are not useful for screening for vertebral fracture in a patient with low back pain. Those that inform, in primary care, are trauma (positive likelihood ratio of 6.42; 95% CI 2.94 to 14.02 for osteoporotic fracture specifically) and corticosteroid therapy (2.46; 1.13 to 5.34). Advanced age informs mainly for unspecified vertebral fracture. And above all: combinations of red flags perform better than individual items.8
The Downie review published in the BMJ in 2013 had already given the measure in post-test probabilities, which are more telling than likelihood ratios: in a patient with low back pain, prolonged corticosteroid therapy raises the probability of fracture to 33% (CI 10 to 67), advanced age to 9% (3 to 25), severe trauma to 11% (8 to 16). And when several red flags are present simultaneously, the probability of spinal fracture rises to 90% (34-99).9 The intervals are wide (these are small studies) but the direction is consistent.
One tool is worth knowing because it was built for and by manual therapists, and because the 2023 Cochrane review retained it as the most informative combination of its review. Roman and Cook analysed more than 1,400 patients from a spine surgery centre and isolated a cluster of five characteristics:10
Two positive items or fewer give a sensitivity of 0.95 (CI 0.83 to 0.99) and a negative likelihood ratio of 0.16 (0.04 to 0.51): the fracture becomes unlikely, which is the most useful function in daily practice. Four items out of five give a positive likelihood ratio of 9.6 (3.7 to 14.9), recalculated as 9.62 (5.88 to 15.73) by the Cochrane review from the raw data.810
It must be said what this tool is not. It was derived in a specialised surgical centre population, where the prevalence of fracture is far higher than in a community practice, and the authors themselves write that prospective validation in a separate sample remains necessary. It has not taken place. The cluster is therefore a structured aide-memoire, not an approved test: it organises doubt, it does not replace it.
What the signs available in the clinic are really worth
Positive likelihood ratios. Above 5, the sign usefully shifts the probability; between 1 and 2, it changes almost nothing.
Sources: Roman's cluster, Roman et al. 2010;10 weight, wall-occiput and rib-pelvis, Green et al. 2004, JAMA, review of the clinical examination;11 trauma and corticosteroid therapy for osteoporotic fracture, 2023 Cochrane review;8 height loss, Mikula et al. 2017, 66,021 patients.12 The confidence intervals appear in the body text; they are wide for several of these signs.
The review of the clinical examination published in JAMA by Green and colleagues evaluated physical manoeuvres for the diagnosis of osteoporosis and occult spinal fracture. Three of them are performed without equipment:11
Documented height loss deserves a separate mention, because it is often quoted without its figures. The Mikula study of 66,021 patients shows a specificity that climbs quickly but a sensitivity that collapses: at 1, 2, 3 and 4 cm of height loss, the sensitivity for detecting a vertebral fracture is 42, 32, 19 and 14% and the specificity 70, 82, 92 and 95%.12 Clinical translation: a documented height loss of 3 or 4 cm justifies investigation, but the absence of height loss rules out nothing at all: even at the most sensitive threshold, 1 cm, nearly six fractures in ten escape detection.
Red flags: what should stop or change the session
None of these items requires “doing nothing any more”. Each requires not doing what was planned and obtaining a medical opinion before resuming.
⚠️ The safety rule fits in one sentence: as long as a recent fracture has not been ruled out, we do not manipulate, we do not mobilise forcefully, we do not apply posteroanterior pressure to the spinous processes, and we do not load the spine in flexion.
Key points

MRISagittal T2-weighted MRI of the lumbar spine: L1 compression fracture indicated by an arrow, with associated stenosis at L4-L5.
Source : Hatgis et al., Cureus, 2017, figure 1 · CC BY
The physiotherapist does not prescribe imaging, but is often the one who triggers the request. Knowing which examination answers which question avoids sending a patient away with an X-ray that will not say what we wanted to know.
“Is there a vertebral deformity?” The thoracic and lumbar spine X-ray in lateral view answers that one. The lateral view is not a detail: it is the projection on which loss of vertebral body height is read, and it is precisely the imaging that the literature identifies as necessary to recognise the fractures that escape clinical detection.1 Grading uses Genant's semi-quantitative method, which remains the reference thirty years after its publication and whose founding article mainly establishes inter- and intra-observer reproducibility.13 In its usual formulation, it grades the loss of vertebral body height: grade 1 for about 20 to 25%, grade 2 for 25 to 40%, grade 3 beyond 40%.
“Is this fracture recent?” The X-ray does not say. A grade 2 compression on a film carries no date: it may be six weeks old or six years old. It is MRI that answers, by the presence of bone marrow oedema (hyperintense signal on STIR or fat-saturated T2 sequences, hypointense on T1). It is the examination that distinguishes the vertebra that has just given way from the one that gave way long ago, and this distinction governs the whole management: the first requires caution, the second allows progressive loading.
CT can settle some cases when MRI is unavailable or contraindicated. A German study of 192 fractures measured the ability of multidetector CT alone to separate recent or subacute fractures from old ones: the area under the curve reaches 0.854 and 0.861 depending on the readers, with a sensitivity of 97.2 and 94.5% but a specificity of only 58.1 and 65.1% when doubtful cases are classified as recent.14 In other words, CT misses few recent fractures but overdiagnoses many: it is poorly reassuring. One sign is an exception, double trabecular compaction, present in about half of recent fractures but highly specific (93.2 and 88.6%).14
“Does this patient have untreated osteoporosis?” That is bone densitometry, and it is the general practitioner or the rheumatologist who requests it. The point to know for the physiotherapist is that in France, the updated recommendations of the French Society of Rheumatology and the GRIO recommend drug treatment in women with a severe fracture, vertebral fracture being in this category, without requiring a prior densitometric threshold.35 A patient who comes out of a vertebral fracture with no bone treatment at all is not a borderline case: it is a departure from the recommendation, and flagging it is part of the job.
| Examination | Answers | Does not answer | What it changes in the session |
|---|---|---|---|
| Lateral X-ray, thoracic and lumbar spine | Is there a deformity? Which Genant grade13? | Since when. An old compression and yesterday's compression look the same on the image. | Confirms the background and the number of levels affected; cannot on its own authorise loading. |
| MRI with a STIR or fat-saturated T2 sequence | Is the fracture recent? Is there bone marrow oedema? Is there posterior wall retropulsion, neurological compression? | Bone mineral density. | It is the examination that does or does not authorise progressive loading and the resumption of resisted exercises. |
| CT | Cortex, geometry, posterior wall; approximate dating. | Reliable dating: specificity of 58 and 65% depending on the reader for “recent”.14 | Useful as a second-line examination; does not remove the need for MRI when the question is the freshness of the fracture. |
| Bone densitometry | The T-score, hence the bone background and the drug indication. | The existence of a fracture, which it does not see outside a dedicated spine analysis module. | Nothing directly, but its absence in a fractured patient is a signal to pass on.35 |
Decision tree for spinal pain in the older person
The question is not “should everyone be imaged”, it is “does this particular patient cross a threshold”.
Editorial construction from published thresholds: red flags and combinations, 2023 Cochrane review8 and Downie 2013;9 five-item cluster, Roman 2010;10 Genant grading13. This algorithm has not been prospectively validated as such.
Key points
The differential diagnosis works in both directions. One must know how to recognise the compression fracture beneath thoracic spine pain that would have been called mechanical, and also know that a compression fracture visible on imaging does not necessarily explain the pain of the day.
The most common trap is not to confuse the compression fracture with another disease: it is not thinking of it at all, because the picture fits perfectly with something ordinary. Three entities of the spine, each dealt with in detail elsewhere on this site, present in the same patient and sometimes on the same day.
| Entity | What points towards it | What points against it | Management |
|---|---|---|---|
| Vertebral compression fracture | Clear, dated onset even without trauma; segmental pain on percussion; pain on moving from sitting to standing; relief when lying down; height loss; corticosteroid therapy. | Long-standing fluctuating pain over years; exclusively posterolateral pain reproduced by combined movements. | Lateral imaging; suspension of forceful techniques until the result. |
| Thoracic spine pain | Interscapular or paravertebral pain, often postural, aggravated by prolonged holding and improved by movement. | A dated sudden onset; pain that worsens week after week. | Active management; imaging not systematic. |
| Non-specific low back pain | Long history, repeated episodes, examination without a fixed painful point. | A first episode of life after 70 should prompt reflection before being labelled non-specific. | Classic biopsychosocial model; reassess if the trajectory departs from it. |
| Lumbar facet joint arthropathy | Pain on extension and rotation, relieved by flexion; non-radicular gluteal radiation. | Pain relieved by strict lying down and reproduced by spinous percussion. | Conservative treatment; lateral imaging remains useful if the background is osteoporotic. |
| Pathological fracture: metastasis, myeloma | Non-positional night pain, deterioration of general condition, history of neoplasia, involvement of the posterior wall or pedicle on imaging. | Strictly mechanical pain in an otherwise healthy patient. | Medical referral without delay. This is not a physiotherapist's diagnosis, it is a reason for referral. |
The reverse reasoning deserves as much attention. In a person of 80, finding a vertebral deformity on an X-ray is nothing exceptional: it is even expected. Automatically attributing the day's pain to this image is reproducing on the thoracic spine the error that physiotherapy took twenty years to unlearn on the lumbar spine. An old deformity, without marrow oedema on MRI, is not an active fracture: it is a scar.
A compression fracture on an X-ray has no date. It is the bone marrow oedema on MRI that distinguishes the vertebra that is suffering from the one that has suffered.
This distinction has a direct consequence on what we allow ourselves. With a recent fracture, caution is called for regarding loading and flexion. With an old, painless deformity in a patient whose pain comes from elsewhere, the same caution becomes an unjustified brake, and the international recommendations insist precisely on this point: not turning a history of fracture into a list of lifelong prohibitions.28
On the same spine, elsewhere on this site
Key points

MRISagittal MRI of the lumbar spine: severely collapsed vertebra with marked kyphotic angulation, a sequela of an untreated fracture.
Source : Jacobson et al., Cureus, 2017, figure 1 · CC BY
Because the fracture is not an accident, it is a symptom. It reveals a bone that gives way under ordinary stresses, and that bone does not become normal again because the pain has settled.
This is the figure to know by heart. Lindsay and colleagues analysed the 2,725 postmenopausal women randomised to the placebo arms of four large osteoporosis trials. Among the 381 participants who developed an incident vertebral fracture during the study, the incidence of a new vertebral fracture in the following year was 19.2% (95% CI 13.6 to 24.8). And the presence of at least one vertebral fracture at baseline multiplied the risk of fracturing during the first year of the study by 5.1.4
One woman in five. Within the year. This figure circulates a lot in distorted forms (“20%”, “one in five”, sometimes “25%”) without the primary source being cited. Here it is, with its population: postmenopausal women with a mean age of 74, on placebo, in trials where fractures were detected radiographically, including those that had never manifested clinically.4
The statistical synthesis by Klotzbuecher, published in the Journal of Bone and Mineral Research, gives the overall picture across all fracture sites. Its most striking result: the strongest association in the whole literature is that between prior vertebral fracture and subsequent vertebral fracture, with a risk about four times higher, which rises further with the number of fractures already present. For the other combinations of sites (hip, wrist, spine) the relative risk is around 2.5
The fracture cascade, in measured risks
The spine predicts the spine better than any other site predicts anything.
Sources: refracture at one year, Lindsay et al. 2001, JAMA, 381 patients who fractured out of 2,725 on placebo;4 relative risks, Klotzbuecher et al. 2000, JBMR, statistical synthesis of the literature: 2.0 (CI 1.8 to 2.1) in peri- and postmenopausal women, 2.2 (1.9 to 2.6) in the other studies.5
Vertebral fracture is not only a predictor of fracture. It is a marker of global frailty. The prospective Dubbo cohort, published in JAMA, followed women and men aged 60 and over for nearly twenty years: the standardised mortality rate after a clinical vertebral fracture is 1.82 in women (CI 1.52 to 2.17) and 2.12 in men (1.66 to 2.72), with an excess mortality that persists five years after the event.6
This figure must be read with the caution appropriate to an association: vertebral fracture does not directly kill in most cases. It occurs in people who are more frail, more undernourished, with more comorbidities, and it then aggravates what made it likely, through pain, immobilisation, loss of function, kyphosis, reduction of abdominal and thoracic volume. The chain is plausible and documented in its links, but overall causality is not demonstrated by this type of study, and the article will not claim that it is.
What this figure does not say
A standardised mortality rate of 1.82 means that people who have had a clinical vertebral fracture die 1.82 times more than the general population of the same age and sex.6 It does not mean that the fracture causes these extra deaths, nor that preventing the fracture would avoid them. It is a signal of severity, to be treated as such: a patient who has just had a vertebral fracture deserves to have their overall condition looked at, not just their back.
Red flags: recognising the next fracture
A refracture does not present like the first: the patient is already known, already followed, and their back pain no longer surprises anyone. These are the signals that must reopen the file, in the middle of a course of care that was going well.
⚠️ In the face of one of these signals, forceful techniques and loading are suspended, and the patient is referred for imaging before resuming the programme where it had left off.
Key points
Three things, and none of them consists in treating the bone. Spotting that osteoporosis is untreated, reducing falls, and giving the patient back a spine that holds up. The first two are the least spectacular and the best demonstrated.
It is the simplest and most cost-effective contribution. It requires no new skill: it requires asking a question. After your fracture, have you had a densitometry? Are you taking any treatment for your bones?
The probability that the answer is no is high. In the Irish cohort already cited, only 24% of patients hospitalised for hip fracture left with anti-osteoporotic treatment, with no difference between those who had a vertebral fracture on CT and those who did not.2 And the French recommendations of the French Society of Rheumatology and the GRIO are explicit: in women with a severe fracture (a category that includes vertebral fracture), drug treatment is recommended, all available molecules being usable.35
What this detection produces is not hypothetical. Post-fracture care pathways, or fracture liaison services, have been the subject of a systematic review with meta-analysis covering 37 studies, 34 of them at low risk of bias: the risk of secondary fragility fracture is reduced, relative risk 0.68 (CI 0.55 to 0.83) beyond two years, with moderate certainty of evidence according to GRADE.7 A reduction of one third in the risk of refracture, obtained through organisation and not through a technical procedure.
The physiotherapist does not prescribe a bisphosphonate. But they are often the only caregiver to see this patient twenty times in three months, and the only one in a position to notice that no one has prescribed anything.
This is the domain where the evidence is the strongest in this whole article, and it does not specifically concern vertebral fracture. The Cochrane review by Sherrington, in its abridged version published in the British Journal of Sports Medicine, establishes, on 59 studies and 12,981 participants, that exercise reduces the rate of falls by 23% (rate ratio 0.77; CI 0.71 to 0.83), with high certainty (this mention is rare and deserves to be emphasised).30
The detail of the modalities matters as much as the overall result:
And a piece of data directly useful to the profession: the subgroup analysis shows a greater effect in trials where the intervention was delivered by a health professional, most often a physiotherapist.30
The logic of this argument must be named, because it is indirect. The evidence concerns falls, not vertebral fractures, and a significant share of vertebral fractures occur without a fall. Reducing falls by 23% therefore does not reduce vertebral fractures by 23%. But in a patient who has already fractured, falls prevention protects the hip, wrist and pelvis, whose risk is also doubled.5 It is a real benefit, on neighbouring targets, and it is better to present it as such than to over-interpret it.
This is the most specific lead, and the most fragile in evidence. It rests on a prospective study by Sinaki conducted at the Mayo Clinic: 50 postmenopausal women followed for ten years after a two-year randomised trial, 27 of whom had carried out progressive resistive strengthening of the spinal extensors. At ten years, the incidence of vertebral fracture was 1.6% of the vertebral bodies examined in the exercise group versus 4.3% in controls, a relative risk of 2.7 against the controls (p = 0.029). Extensor strength remained significantly higher eight years after the programme stopped.27
It is an encouraging result and it must be given its exact weight: fifty women, a single team, an endpoint measured per vertebral body and not per patient, and observational follow-up after the randomised phase. It is not the demonstration that strengthening prevents fractures; it is the best signal available in favour of this hypothesis, and it has never been replicated on a large scale.
What each lever reduces, and with what certainty
Three levers accessible to the physiotherapist, three different targets. None directly reduces vertebral fractures.
Sources: falls, Sherrington et al. 2020, abridged Cochrane review;30 post-fracture pathway, Danazumi et al. 2024, 37 studies, authors' GRADE assessment;7 extensors, Sinaki et al. 2002, 10-year follow-up of 50 women.27 The certainty ratings reproduce those of the authors for the first two rows; the third is an editorial judgement, no GRADE assessment having been published on this study.
What the physiotherapist does not do
The physiotherapist does not set the indication for an anti-osteoporotic treatment, does not interrupt it, and does not comment on it in terms of the choice of molecule. The role described here is a role of detection and transmission: noting the absence of a work-up or treatment, writing it in a letter, and continuing one's own work. A patient sent back to their doctor with a precise, dated observation obtains more than a patient to whom the pharmacology of bone is explained.
Key points
This is the subject on which it is most tempting to give a verdict, and the one on which giving a verdict would be dishonest. Twenty years of trials have produced results that contradict one another, and the way they contradict one another is instructive.
The principle is simple: injecting acrylic cement into the fractured vertebral body to stabilise it and relieve pain. Kyphoplasty adds a balloon inflated before the injection, supposed to restore part of the lost height. Both techniques are widely practised. Their evidence base is one of the most disputed in musculoskeletal medicine.
2010: VERTOS II. Dutch and Belgian open-label trial, 202 patients randomised, fractures less than six weeks old with oedema on MRI and pain of at least 5 out of 10. Vertebroplasty clearly outperforms conservative treatment: difference of 2.6 points on the visual analogue scale at one month and 2.0 at one year, with no serious complication.18
But the trial contains a piece of data that its conclusions do not put forward and that deserves to be read: of the 431 eligible patients identified, 229, or 53%, saw their pain resolve spontaneously during the assessment phase, before randomisation.18 More than one patient in two got better on their own, within a few days. No open trial can distinguish the effect of the cement from this natural course plus waiting.
2016: VAPOUR. Australian double-blind trial against a placebo procedure, 120 patients, restricted to fractures less than six weeks old with pain of at least 7 out of 10 and an optimised vertebral filling technique. Positive result: 44% of treated patients versus 21% of controls had pain below 4 out of 10 at fourteen days (difference 23 percentage points, CI 6 to 39; p = 0.011).16
2018: VERTOS IV. Dutch double-blind trial against a sham procedure, 180 patients, acute fractures. Negative result: no statistically significant difference over twelve months of follow-up, neither in pain, nor in quality of life, nor in disability, nor in analgesic consumption, both groups improving significantly.17
2018: The Cochrane review. Twenty-one trials, five of them against placebo. Verdict: evidence of high to moderate quality that there is no important benefit on pain, disability, quality of life or therapeutic success. Mean pain was 5 out of 10 under placebo and improved by 0.7 points with vertebroplasty, for a minimal clinically important difference set at 15%. The authors specify that subgroup analyses show no different effect according to the duration of the pain, acute or subacute, and that open trials comparing with usual care have probably overestimated the benefit.15
2023: VERTOS V. Change of terrain: this time the fractures are chronic, pain of more than three months with persistent oedema on MRI, and the comparator is not a placebo but an anaesthetic infiltration. 80 patients. Positive but modest result: difference of 1.3 points on the visual analogue scale at twelve months (CI 0.1 to 2.6), a significant difference also on quality of life, but not on disability.19
Twenty years of vertebroplasty trials
The result depends closely on the comparator and the population included. This is what makes the synthesis difficult, and the debate legitimate.
Sources: Klazen 201018, Clark 201616, Firanescu 201817, Buchbinder 201815, Carli 202319. The “positive” and “negative” labels summarise the primary endpoint of each trial as published by its authors.
Two readings clash, and neither is absurd.
The first: vertebroplasty does not work, the positive trials are positive for methodological reasons. This is the position of the Cochrane review. VERTOS II is open-label, hence inflated by waiting; VAPOUR is small, single-centre in its main recruitment, and its endpoint is a dichotomised threshold rather than a mean. Against placebo, across five trials, pain does not move in a clinically important way.15
The second: vertebroplasty works, but for a population that the negative trials diluted. This is the reading of the VAPOUR authors: fractures of less than six weeks in very painful patients, treated with sufficient filling, would be the target, and the negative trials would have included too many old or barely painful fractures, in whom spontaneous evolution erases the difference.16 To which the Cochrane review replies that its subgroup analyses according to the duration of pain do not find this differential effect.15
A third element has recently been added, and it concerns the survival argument. Several observational studies had reported lower mortality after kyphoplasty. An analysis of 235,317 Medicare beneficiaries took up the question with increasingly rigorous matching: the apparent advantage (hazard ratio 0.84) disappears then reverses when matching on medical complications and on age and comorbidities: hazard ratio 1.32 then 1.81 against kyphoplasty.20 The “survival gain” was an artefact of indication: the least ill patients were being operated on.
What the physiotherapist can say to a patient who asks
That an operation exists, that it consists of injecting cement, that it is sometimes useful in a specific situation (a recent, very painful fracture resistant to analgesics), and that the literature is divided on its benefit in the other cases. That the decision belongs to the doctor, and that the patient has the right to ask why it is being proposed to them in particular. What must not be said: neither “it's no use”, nor “it has to be done”.
Key points
The direct comparison exists. A meta-analysis of sixteen studies compared balloon kyphoplasty and percutaneous vertebroplasty: kyphoplasty reduces the wedging angle (standardised mean difference 0.98; CI 0.40 to 1.57), restores more vertebral height and reduces the risk of cement leakage (relative risk 0.62; 0.47 to 0.80). But these radiographic gains do not translate clinically: no significant difference on the visual analogue pain scale or on the Oswestry Disability Index.21
This is a textbook case of the gap between an intermediate endpoint and the endpoint that counts. The vertebra is straighter, cement leakage is less frequent, and the patient has neither less pain nor fewer difficulties. For the physiotherapist, the practical consequence is simple: the technique used upstream does not change the rehabilitation they then conduct.
The spinal brace is one of the most intuitive and least demonstrated measures of this management. It is prescribed very widely; the literature that supports it would fit in a pocket.
The systematic review by Newman, Minns Lowe and Barker gathered twelve studies (eight randomised or randomised pilot trials, four non-randomised studies) totalling 626 participants. Its result deserves to be quoted as it stands: only three studies, involving 153 people, evaluated bracing after acute vertebral fracture, and none was of high quality. There are not enough data to say that bracing modifies vertebral deformity, and it is with rigid braces that complications were most numerous.22
The review does, however, find more favourable signals in the subacute phase and in prolonged rehabilitation, across nine studies and 473 participants: three of them suggest that a semi-rigid thoracolumbar brace of the backpack type can improve strength, pain, posture and quality of life; one study finds a balance benefit with a weighted kyphosis-correcting brace. The authors conclude that these leads “should be explored in studies of sufficient size and quality, including men”.22
The useful question is not “is a brace needed”, it is “what is this one for, for how long, and what will replace its effect once it is removed”.
This formulation is the one that makes the debate operational. A brace worn for a few weeks to allow a patient to get upright and walk during the most painful phase does a job that nothing else does. A brace worn for six months “to hold the back” creates dependence and deprives the extensors of the stimulus they need, at the very moment when the literature suggests their strength matters.27
| Phase | What was studied | What can be concluded | Reasonable management |
|---|---|---|---|
| Acute fracture | 3 studies, 153 participants, none of high quality. | Inconclusive. No data showing an effect on the deformity. Maximum complications with rigid braces. | If a brace, the lightest one that allows getting upright, for a duration announced from the outset and short. |
| Subacute phase and prolonged rehabilitation | 9 studies, 473 participants, variable quality. | Favourable signal for the semi-rigid backpack-type thoracolumbar brace on strength, pain, posture, quality of life (3 studies). | Can be considered as a complement to an active programme, never in its place. |
| Balance and established kyphosis | 1 study on the weighted kyphosis-correcting brace. | Improvement in balance reported, in a single study. | A lead, not a recommendation. To be reassessed when trials of sufficient size are published. |
Key points
We must start with the bad news, because it is structuring: physiotherapy after vertebral fracture has not demonstrated much. The evidence has grown since, without becoming solid. Knowing this changes the way it is conducted, not the decision to conduct it.
The 2019 update of the Cochrane review, by Gibbs and Giangregorio, gathered nine trials and 749 participants, 68 of them men. Its conclusions, in order:23
The authors conclude that a high-quality trial remains necessary, and give the required sample size calculation: about 2,500 untreated participants, or 4,400 on anti-osteoporotic treatment.23 No trial of this size has been conducted since.
The PROVE trial, conducted in Oxford, is the largest ever carried out in physiotherapy for vertebral fracture: 615 adults with back pain, osteoporosis and at least one vertebral fracture, randomised into three arms: seven individual manual therapy sessions over twelve weeks, seven home exercise sessions, or a single education session.24
At twelve months, on the two primary endpoints (quality of life measured by the QUALEFFO-41 and extensor endurance measured by the timed loaded standing), no statistically significant difference between the three groups. The endurance gain was 9.8 seconds in the exercise arm, 13.6 seconds in the manual therapy arm and 4.2 seconds in the single-session arm, with confidence intervals that cross zero for exercise.24
Two secondary results deserve attention. At four months, significant differences existed: endurance and balance in the manual therapy arm; endurance in those under 70, balance, mobility and walking in the exercise arm. And the authors write in black and white: “adherence was a problem”, and “the benefits at four months did not persist”, while concluding that “short-term benefits are probably valuable”.24
At the other end of the size spectrum, the Bennell pilot trial, conducted in Melbourne on twenty participants, is clearly positive: compared with a no-treatment group, ten weeks of individual physiotherapy and daily exercises reduce pain on movement by 1.8 points (CI 0.1 to 3.5) and at rest by 2.0 points, improve QUALEFFO function by 4.8 points and endurance measured by the timed loaded standing by 46.7 seconds (16.1 to 77.3). Nine participants out of eleven in the treated group rated their pain “much better” versus one out of nine in the controls.25
Twenty patients, a no-intervention comparator, no possibility of blinding on the experience of pain: this pilot does not demonstrate, it suggests. But the juxtaposition with PROVE is illuminating. Bennell compares physiotherapy to nothing; PROVE compares seven sessions to an education session. This is not the same question, and the fact that PROVE is negative means that seven sessions do no better than one well-done session, not that physiotherapy is useless.
PROVE does not say that rehabilitation is useless. It says that a well-conducted education session does almost as well as seven sessions, which is information about dosage, not about principle.
Moderate
Exercise for falls prevention. Reduction of the rate of falls by 23% in the general older population, high certainty in the Cochrane review, downgraded here to moderate because the population studied is not that of people with vertebral fracture and because the target is the fall, not the vertebral fracture.30
Moderate
Improvement in physical performance. The Timed Up and Go improves by 1.13 seconds, the only result of moderate quality in the 2019 Cochrane review, and judged by its authors to be not clinically important.23
Low
Strengthening of the spinal extensors to reduce fractures. One study, 50 women, 1.6% versus 4.3% of vertebral bodies fractured at ten years, relative risk 2.7. Never replicated.27
Low
Manual therapy and exercise for short-term pain and function. Real benefits at four months in PROVE, which do not persist at twelve months; positive Bennell pilot but on twenty patients and without an active comparator.2425 More recent syntheses strengthen the evidence without making it solid, and diverge on quality of life: a small benefit of exercise on pain, of low certainty, with no demonstrated gain in quality of life or activity, in a review of twenty trials of non-surgical and non-drug management (Bolton 2022);36 benefits on functional capacity, balance, trunk strength, pain and quality of life in a meta-analysis of nineteen exercise studies published in 2026, but without grading of certainty, including two quasi-experimental studies and a pilot, and with high heterogeneity on pain (I² of 72%).37
Low
Avoiding loaded spinal flexion in the recent phase. A single comparative study, very small numbers, but a massive result consistent with biomechanics, never contradicted since 1984.26
Very low
Semi-rigid thoracolumbar brace in the secondary phase. Three studies out of twelve in the Newman review suggest a benefit on strength, pain, posture and quality of life; variable methodological quality, men almost absent.22
Very low
Exercise to reduce fractures or falls in people with vertebral fracture. A single trial for each outcome, no difference, very low quality evidence according to the Cochrane review itself.23
Not demonstrated
Rigid brace after acute fracture. Three studies, 153 participants, none of high quality, no demonstrated effect on the deformity, and the highest complication rate of the review.22
The levels in this table are an editorial judgement applying GRADE principles, and not a published GRADE assessment. Two rows reproduce certainty judgements formulated by the authors of the cited reviews (exercise and falls; physical performance), and the short-term pain row is consistent with the Bolton 2022 judgement (low certainty); the others are derived from the number of studies, their size, their risk of bias and the consistency of the results. Apart from these ratings of isolated outcomes, no GRADE assessment covers the whole of rehabilitation of osteoporotic vertebral fracture.
Key points
This is the most frequently asked question and the worst handled. The answer comes in two parts that seem to contradict each other and do not: loaded flexion is to be avoided in the recent phase, and generic lifelong prohibitions are counterproductive.
From a single study, published in 1984 in the Archives of Physical Medicine and Rehabilitation. Sinaki and Mikkelsen followed 59 women with postmenopausal spinal osteoporosis and back pain, divided into four exercise programmes. After one to six years of follow-up with radiographs before and after, the new fractures or wedgings were distributed as follows:26
New fractures according to the type of exercise prescribed
The founding 1984 study: massive result, tiny numbers. Both deserve to be read together.
Source: Sinaki and Mikkelsen 1984, Archives of Physical Medicine and Rehabilitation; 59 postmenopausal osteoporotic women with back pain, follow-up of one to six years, radiographs before and after. Significant difference in favour of extension versus flexion (p < 0.001) and versus the combined programme (p < 0.01). The numbers per group range from 6 to 25 patients: the 89% in the flexion group represents eight patients out of nine.26
This graph must be looked at with both eyes. On one side, the gap is spectacular and the statistical significance is there: 16% versus 89%, p < 0.001.26 On the other, the flexion group comprised nine patients. A study of nine people does not normally establish anything. It nevertheless underpins a rule taught throughout the world for forty years, for three reasons: the effect is very large, it is consistent with the mechanics of the vertebral body (flexion transfers the load onto the anterior column, precisely where wedge compression occurs), and above all no study has ever contradicted it, for the obvious reason that no ethics committee would today authorise randomising osteoporotic patients to a flexion programme.
This is a case where the strongest evidence we will ever have is weak evidence, and where waiting for more would be unreasonable. The practical conclusion is therefore: in the recent phase, loaded spinal flexion is not taught, and is not used as an exercise. This does not mean the patient is immobilised in extension.
The international consensus Too Fit To Fracture, arising from a two-round Delphi process bringing together researchers and clinicians, is explicit on this point. Its key conclusions:28
The central argument of the consensus is that a list of prohibitions produces a patient who no longer moves, and that a patient who no longer moves loses strength, balance and bone mass, that is, the three things that were protecting them. The right answer is not to ban flexion: it is to learn to pick up an object by bending the hips and knees rather than the spine, and to repeat it until it becomes the default movement.
A prohibition protects today's vertebra. A technique protects all the vertebrae of the next ten years, because the patient takes it with them.
The question has arisen since the LIFTMOR trial, conducted in Australia in 101 postmenopausal women with low bone mass (T-score below −1.0), randomised between eight months of high-intensity resistance and impact training (five sets of five repetitions at more than 85% of the maximum, twice a week for thirty minutes, supervised) and a low-intensity home programme. The results are clearly in favour of high intensity: lumbar spine bone mineral density +2.9% versus −1.2%, femoral neck +0.3% versus −1.9%, cortical thickness, height, and all functional tests. Only one adverse event was reported.29
This trial has been widely circulated as proof that “you can lift heavy even with osteoporosis”. It has to be read for what it is. LIFTMOR included women with low bone mass, selected and screened for conditions and treatments interfering with bone and physical function; it was not a population of people with recent vertebral fracture. The training was supervised in the gym by professionals, not prescribed for independent practice. Nothing in this trial allows heavy loading of a vertebra that still has marrow oedema on MRI.
What LIFTMOR does establish, however, is worth retaining: in a well-supervised osteopenic and osteoporotic population, high intensity did not produce the wave of fractures that traditional caution led people to fear. This shifts the cursor of what is reasonable, not the management in the acute phase.
The rule of conduct in one line per phase
Key points
Four observations, chosen because they show the four trajectories this diagnosis can take: the one that goes well, the one discovered too late, the one that turns around, and the one that begins as ordinary low back pain.
Case 1: The ordinary trajectory, and it is a good one
Papa 2012, Journal of the Canadian Chiropractic Association 56(1):29-39: PMID 22457539
A 74-year-old man consults for acute pain in the thoracolumbar region that occurred after lifting an object. The X-ray shows generalised demineralisation and a moderate wedge fracture of L1.
Management is entirely conservative: postural education, adaptation of activities, interferential currents, taping in extension, instrument-assisted soft tissue technique, and prescription of rehabilitation exercises. The follow-up criteria are a verbal pain scale, medication consumption and return to activities of daily living. Lasting resolution of symptoms, with no painful recurrence at the twelve-month check.
What this case teaches. The most frequent scenario is this one: a moderate-grade fracture in an independent patient, which progresses well without invasive intervention. It is the reminder that is often missing when one has only complications in mind. What this case does not teach: it is a single observation, without a comparator, and the improvement at twelve months is also what natural history does; VERTOS II showed that more than one eligible patient in two improved spontaneously within a few days.1831
Case 2: The one that went unseen for two years
Massoud and Alashkar 2026, Cureus 18(3):e104527: PMID 41783551
A 78-year-old woman presents after a fall. The work-up finds an acute vertebral fracture, and, alongside it, a chronic vertebral fracture and a chronic diaphragmatic hernia that had gone unnoticed for more than two years.
The authors draw from their observation a conclusion that also holds for the physiotherapy clinic: falls are frequent in older people, the prevalence of osteoporosis is high there, and the resulting lesions can be subtle and overlooked by the patient as well as by the doctor. Discovering them makes it possible not only to treat them early, but to put in place appropriate preventive measures.
What this case teaches. It is the clinical illustration of Schousboe's figure.1 Two years of unnamed fracture in a patient who certainly crossed paths with caregivers during that period. The physiotherapist who sees an older patient for back pain is, statistically, seeing this kind of situation.32
Case 3: The low back pain that becomes an emergency, Kümmell's disease
Kuppan et al. 2022, Journal of Orthopaedic Case Reports 12(10):34-38: PMID 36874888
A 65-year-old woman consults for low back pain evolving for four weeks. Then progressive weakness and bladder and sphincter disturbances appear. The X-rays show a compression fracture of T12 with an intravertebral vacuum sign. MRI finds intravertebral fluid and significant spinal cord compression. Posterior decompression, stabilisation and transpedicular graft; histology confirms Kümmell's disease. The patient recovers her strength, bladder control and independent walking.
The authors recall the mechanism: Kümmell's disease is an osteoporotic vertebral fracture progressing towards avascular necrosis of the vertebral body, with an initially asymptomatic period then progressive pain, kyphosis and neurological deficit after trivial trauma. They emphasise that osteoporotic fractures are more exposed to pseudarthrosis because of poor vascular and mechanical support.
What this case teaches. Spinal pain that worsens over the weeks in an osteoporotic patient is not a fracture that is consolidating poorly: it is a signal. The physiotherapist is the professional best placed to spot this reversal of trajectory, because they are the one who sees the patient again.33
Case 4: Late collapse in the very old
Picazo et al. 2014, European Spine Journal 23(12):2696-2702: PMID 23778750
The authors report the case of an older man with an osteoporotic fracture progressing towards a severe late collapse with neurological involvement. Treatment consisted of bilateral posterior transpedicular decompression and instrumented arthrodesis with cemented pedicle screws; the course was uncomplicated and the patient recovered his deficit and his walking capacity.
The text frames it in a sentence that can be taken up as it stands: vertebral fracture is common in people with osteoporosis and does not cause clinical problems in the majority of cases; only a small number of patients suffer serious neurological complications related to late collapse.
What this case teaches. The proportion matters as much as the event. The aim is not to make every compression fracture a neurological threat (that would be false and anxiety-provoking) but to know that the complication exists, that it is late, and that its first signs pass through the physiotherapy appointment before passing through the emergency department.34
Why these four
These are real published observations, with their PubMed identifier, and not reconstructed vignettes. They have the drawback of their kind: a clinical case demonstrates nothing, it illustrates. Three of them describe situations that went badly, which is the natural bias of case publication (what surprises is published). Case 1 is there to restore the proportion: the most frequent trajectory is the one that goes well.
Key points
Everything above comes down to four decisions: thinking of the fracture, knowing when to stop, knowing what to do next, and not letting a patient leave whose bone is being treated by no one.
Nothing spectacular is asked. Three minutes added to the history and two measurements are enough to change the probability.
| Phase | Main objective | What is done | What is not done |
|---|---|---|---|
| Phase 1: recent fracture or suspicion not ruled out weeks 0 to 4 approximately |
Avoid aggravating; maintain getting upright and walking; do not allow bed rest to set in. | Education and explanation of the diagnosis; transfers without spinal flexion (log-roll technique); daily walking split into bouts; breathing work; gentle isometrics of the extensors in the pain-free range; adaptation of the home. | Manipulation. Forceful mobilisation. Posteroanterior pressure on the spinous processes. Loaded flexion. Added axial load. Stretching into global flexion. |
| Phase 2: consolidation weeks 4 to 12 approximately |
Restore extensor endurance and balance; establish spine-sparing techniques. | Progressive strengthening of the spinal extensors;27 daily balance training;2830 systematic teaching of the hip hinge applied to real everyday movements; gradual weaning from the brace if one has been prescribed. | Going straight to heavy loading. Prolonging the brace with no end date. Settling for unloaded exercises. |
| Phase 3: maintenance beyond 3 months |
Reduce the risk of falls and of new fracture; make the patient independent. | Programme combining balance, functional exercises and strengthening: it is the combination that reduces falls the most;30 moderate intensity preferred after vertebral fracture;28 periodic reassessment of height and posture. | Maintaining generic prohibitions that no longer serve a purpose. Stopping follow-up without having checked that bone treatment is in place. |
This is the least technical part and the most useful. A letter to the general practitioner containing three precise lines does more than one more session. What it must contain:
It is this third point that matters most in the long run. Three quarters of patients hospitalised for fragility fracture leave without bone treatment2, whereas the French recommendations provide for treating severe fractures35, and the organisation of post-fracture pathways reduces the risk of refracture by about a third.7 The physiotherapist prescribes nothing. They can point out that something is missing.
Key points
There is no published standard timeframe, and the round figures that circulate should be treated with caution. What can be said from the data: in VERTOS II, 53% of eligible patients saw their pain resolve spontaneously during the assessment phase, that is, within a few days to a few weeks.18 The useful benchmark is therefore not a duration but a direction: the pain of a recent fracture should decrease. Pain that increases over the weeks raises suspicion of pseudarthrosis or progressive collapse and requires an opinion.33
Not until a recent fracture has been ruled out: that is the position of caution this article supports. We must, however, be honest about the nature of this recommendation: it does not rest on comparative trials, which do not exist and never will, but on mechanics and on potential severity. The reasoning is asymmetric: the expected benefit of manipulation in this setting is modest, the cost of an error is one more vertebra.
That the decision rests with the doctor, and that the literature is divided. Against placebo, the 2018 Cochrane review finds no clinically important benefit;15 the VAPOUR trial, for its part, is positive, but in very specific patients: a fracture less than six weeks old and pain of at least 7 out of 10.16 The question the patient can legitimately put to their doctor is: “am I in that situation?”
The evidence after an acute fracture is inconclusive, and it is with rigid braces that complications are most numerous.22 When a brace is prescribed, two questions must have an answer from the first day: for how long, and what will take over. A brace that accompanies the return to an upright position for three weeks is nothing like a brace worn for six months.
By distinguishing what is risky from what is not, rather than by giving a list of prohibitions. That is exactly the position of the Too Fit To Fracture international consensus: providing spine-sparing techniques (the hip hinge for picking things up, the log roll for getting up) rather than generic restrictions such as “do not carry anything over five kilos”.28 The patient who knows how to do it keeps moving; the patient who has been given a list stops.
Good for confirming, poor for excluding. In 66,021 patients, a height loss of 3 cm has a specificity of 92% but a sensitivity of 19%; at 4 cm, specificity rises to 95% and sensitivity falls to 14%.12 In other words: marked height loss justifies looking further, but the absence of height loss lets the great majority of fractures through.
Yes, to a degree that is often surprising. The European EPOS study measures an incidence of 5.7 per 1,000 person-years in men versus 10.7 in women: the ratio is about two to one.3 And the excess mortality after clinical vertebral fracture is higher in men (standardised rate 2.12) than in women (1.82).6 An older man with back pain must not be given less vigilance.
The available data are reassuring without being definitive. The 2019 Cochrane review reports four adverse effects related to exercise across nine trials and 749 participants: a costal cartilage fracture, a rib fracture, knee pain, skin irritation, and rates the certainty of this information as very low.23 A 2026 meta-analysis estimates the probability of a minor adverse event during exercise at about 6.3% (pain or discomfort that does not require stopping) and that of a major event at about 1.65% (an injury or accident leading to discontinuation of the programme), with no statistically significant excess compared with control groups (usual care or daily life; 2 points more on the exercise side, p = 0.07); these percentages do not measure a risk of fracture.37 The LIFTMOR trial, with training at more than 85% of the maximum, reports only one adverse event among 101 women with low bone mass, but in a supervised setting and with no recent fracture.29
It is essential for everything else (independence, mood, bowel transit, prevention of deconditioning) but it must not be credited with what it does not have. In the Cochrane review on falls prevention, the authors explicitly declare themselves uncertain about the effect of programmes based mainly on walking, on dancing or on strengthening alone; it is the balance and functional programmes, and the combined programmes, that carry the effect.30
The question arises mainly if the pain does not follow the expected course, or if a new episode occurs: the risk of a new vertebral fracture in the following year is 19.2%4, which makes a new painful episode anything but trivial. It is a medical decision, but the physiotherapist is often the one who notices the change in course and who must report it.
Thirty-eight references, each resolved by its PubMed identifier, and whose abstract was read before citation. The figures in the text refer to the source that establishes them, at the place where they are written.
Method and limitations
Sources searched in PubMed via the NCBI E-utilities services, crossing osteoporotic vertebral fracture with epidemiology, underdiagnosis, red flags, imaging, vertebroplasty and kyphoplasty, braces, exercise, falls prevention and reported cases. Each identifier was resolved and each abstract read before citation; the four clinical cases are real published observations, cited with their identifier.
Main limitation of this article: the strongest evidence it contains does not concern the question asked. Certainty is high for exercise against falls30, high to moderate for the absence of benefit of vertebroplasty against placebo,15 but it is low or very low for almost everything that directly concerns rehabilitation after vertebral fracture, the 2019 Cochrane review stopping explicitly at a finding of insufficient evidence.23 Two central claims of practice rest on tiny studies that have never been replicated: the protective effect of extensor strengthening (50 women)27 and the danger of loaded flexion (9 patients in the group concerned)26. They are retained here because they are consistent with mechanics and because no data contradict them, not because they are demonstrated.
Second limitation: the levels in the table of modalities are an editorial judgement applying GRADE principles, and not a published GRADE assessment; apart from the ratings of isolated outcomes (Cochrane 2019, Bolton 2022), no assessment of this type covers the whole of rehabilitation of osteoporotic vertebral fracture. Finally, Roman's diagnostic cluster has never undergone the prospective validation its authors called for.10
The 2023 international consensus on the non-drug and non-surgical management of osteoporotic vertebral fracture (Delphi method) concerns pain, early satiety, weight loss, bracing, safe movement and exercise; its full text was not read, and this article draws no statement from it.38
Article written on 15 August 2026.