Joint hypermobility syndromes: HSD, hEDS and the dosing of exercise

A joint that goes too far is not a disease. Between ordinary hypermobility and hypermobile Ehlers-Danlos syndrome lies a spectrum that the 2017 international classification redrew, and that half the online content still describes with the vocabulary of before. This article sets out the current nomenclature, says what the Beighton score measures and what it does not tell you, and deals with the question physiotherapists actually ask: how do you load without triggering a flare?

What should you take away in ten lines?

Hypermobility, HSD, hEDS: what does the current nomenclature say?

This is the first value of this article, and probably the most useful in the clinic: knowing how to name what is in front of you, with the words the literature has used since 2017, and knowing that those words are going to change.

Until 2017, the reference nosology was that of Villefranche, published in 1998, which distinguished six subtypes of Ehlers-Danlos. Type III, known as “hypermobile”, sat alongside a competing rheumatological diagnosis, “benign joint hypermobility syndrome”, defined by the Brighton criteria. Two medical communities were therefore describing much the same patients with two distinct vocabularies, and the qualifier “benign” maintained the idea that there was not much to be done.

The International Consortium on the Ehlers-Danlos Syndromes reshuffled the cards in 2017. Its classification recognises thirteen subtypes, and sets out a principle that structures everything else: for twelve of them, the definitive diagnosis rests on molecular confirmation, that is, the identification of a causal genetic variant. hEDS is the only subtype with no known genetic marker, and therefore the only one whose diagnosis remains purely clinical¹. That exception is not an administrative detail: it explains both the severity of the clinical criteria adopted, since the absence of a test had to be compensated for, and the lasting discomfort they cause.

In the same move, the category of hypermobility spectrum disorders (HSD) was created to take in the symptomatic patients who do not meet the criteria for hEDS. It is not a second-class category: it is the recognition that between asymptomatic hypermobility and hEDS there is a large, suffering population that needs care.

The table of the 2017 nomenclature

The 2017 hEDS criteria, and why they cause trouble

The diagnosis of hEDS requires three criteria to be met simultaneously. The first is generalised joint hypermobility, measured by the Beighton score with thresholds that vary with age and puberty. The second combines systemic features, soft or slightly hyperextensible skin, unexplained striae, hernias, prolapse, dental involvement and many others, a first-degree family history, and musculoskeletal complications. The third is an exclusion: no other heritable connective tissue disorder, no autoimmune disease, no acquired cause of laxity.

These criteria were designed to improve specificity, and they succeeded. The problem is what they produced in practice. At the GoodHope clinic in Toronto, a retrospective review of 131 patients carrying a diagnosis of hEDS made under Villefranche showed that only 15 % (20 patients) met the 2017 criteria. More troubling still, most of the features supposed to distinguish hEDS were not significantly more frequent in those who crossed the threshold than in those who failed, and the usual comorbidities were distributed about equally in the two groups⁵. In other words: the boundary drawn separates poorly two populations that resemble each other.

An Italian study of 327 patients from 213 families reached a convergent conclusion and put it bluntly: the 2017 criteria should be relaxed so as to include some of the patients now classified as HSD, failing which diagnostic accuracy and quality of care suffer⁶.

A revision is on the way, and this article will be dated

Since 2024 the Ehlers-Danlos Society has been running an initiative called “Road to 2026”, intended to revise the classification, the diagnostic criteria and the care pathways. It draws on an international Delphi study, a real-world clinical practice study launched at the end of 2025, and the HEDGE genetic study, the largest ever conducted on hEDS. Publication of the new criteria is announced for the end of 2026.

Does the Beighton score really measure what it is asked to?

It is the test everyone knows and almost nobody interprets correctly. It deserves better than a figure out of nine copied into the assessment.

The Beighton score counts nine points, spread over five manoeuvres: passive extension of the fifth finger beyond 90°, passive apposition of the thumb to the forearm, hyperextension of the elbow beyond 10°, hyperextension of the knee beyond 10°, each scored right and left, that is eight points, and forward flexion of the trunk with the palms flat on the floor, knees straight, for the ninth.

You have to know where it comes from to understand its limits. It was designed as an epidemiological tool, intended to screen for hypermobility in large populations quickly and cheaply. Its later adoption as an individual diagnostic instrument was not intended by its designers, and it is from that slide that most of the misunderstandings come⁷.

Reliable does not mean valid

That is the distinction that settles most of the debates about this test, and it deserves to be set out properly, because two systematic reviews published in the same year appear to contradict each other while they are not talking about the same thing.

Reliability is good. A systematic review of 24 studies covering 1,333 patients, scored by examiners of varying experience, concludes that the Beighton score shows substantial to excellent inter- and intra-examiner reproducibility. Intra-examiner reliability was reported as excellent in all the studies using intraclass correlation coefficients⁸. Concretely: if two physiotherapists score the same patient, they most often agree.

Validity, on the other hand, is disputed. Agreeing does not guarantee measuring the right thing. The review by Malek et al. notes that the joints in the score are mostly in the upper limb, that they ignore major joints, and above all that one finding recurs consistently in the literature: the score fails to identify hypermobility present in the joints it does not test. It therefore works neither as a direct measure, nor as a reliable indirect indicator of generalised hypermobility. The authors explicitly call for a change in clinical thinking: the score must not serve as the main tool for distinguishing localised from generalised hypermobility, nor be used alone to rule out its presence⁷.

The score varies with age, sex and population

The Beighton score is not a biological constant. Laxity decreases with age, which means that a score of 5 does not have the same meaning at 12 and at 55, since a patient who is negative today may have been frankly hypermobile in their youth, hence the “historical” category of HSD and the existence of a five-item retrospective questionnaire used alongside it.

The meta-analysis by Sobhani-Eraghi et al., covering twenty studies and more than 21,000 children and adolescents, finds an overall prevalence of 34,1 % (95 % CI: 33.3-34.8), with 32.5 % in girls against 18.1 % in boys. Its authors themselves report considerable heterogeneity between studies and statistically significant publication bias: the overall figure is to be handled with care, but the gap between the sexes and the decrease with age are consistent across the literature⁴.

In young adults, the study by Reuter and Fichthorn in 654 American students using the threshold of ≥ 5 adopted by the 2017 criteria finds 12,5 % with generalised hypermobility. Notably and counter-intuitively, women did not have a higher overall rate than men, and scores did not differ by ethnic origin in that sample, even though women were more often hypermobile in the spine and in the right knee and elbow⁹.

The score is over-scored by untrained examiners

It is the most directly actionable result in this whole chapter, and it concerns our own practice. In the Toronto cohort, when the Beighton score had been scored by the primary care doctor and then rescored by a practitioner specialised in EDS, it was higher with the former in 81 % of cases (74 out of 91). And generalised hypermobility was ultimately confirmed in only 46 % of patients (51 out of 111) carrying a previous diagnosis of hEDS⁵.

The lesson is not that generalists score badly, but that this test looks simpler than it is. The sources of error are known: passive confused with active, absence of stabilisation, rough goniometry on the 10° thresholds, a prior warm-up that wins a few degrees, and above all the examiner's expectation seeking to confirm a hypothesis. A score dashed off at the end of an assessment is worth less than no score at all, because it will be taken over as it stands by the next person.

What should you examine beyond the Beighton?

Once the score is recorded, the bulk of the assessment remains to be done. Here is what, in practice, changes what you do.

Hypermobility of the joints the score ignores

Since the Beighton score leaves out the shoulder, the hip, the ankle, the wrist, the cervical spine and the temporomandibular joint, those joints are tested separately, particularly the ones the patient complains about. The shoulder deserves special attention: it is the site most studied in hypermobility rehabilitation trials, and multidirectional instability is frequent there¹⁴. For the lower limb, ankle instability has been the subject of recent work in dynamic imaging in hEDS patients².

The history counts as much as the measurement. The useful questions concern subluxations and dislocations, their number, their mechanism, whether the patient reduces them themselves, repeated sprains, night pain, fatigue on waking, and what the patient has already tried. Many of these patients have several interrupted courses of care behind them.

The systemic features and the family history

They serve not only to tick the 2017 criteria: they point towards connective tissue involvement rather than isolated laxity. You look at the skin, its softness, texture, widened scars, striae without weight gain or pregnancy; you ask about hernias, prolapse, dental and gum history, and above all first-degree family history, which is a criterion in its own right.

One point deserves flagging because it is counter-intuitive. In the Toronto cohort, objective systemic manifestations, those the clinician observes, were found less often than the subjective manifestations reported by the patient, in both groups⁵. That does not invalidate the patient's account; it is a reminder that self-report and examination do not measure the same thing, and that confusing the two artificially inflates the picture.

Which comorbidities change the management?

A programme perfectly dosed in joint terms can fail entirely for reasons that are not joint-related. It is the chapter people skip and should not.

A meta-analysis published in 2026 in Alimentary Pharmacology & Therapeutics, covering 19 studies, 17,455 hEDS/HSD patients and more than 1.6 million controls, gives the broadest picture available to date. The results are of an order of magnitude that makes it impossible to treat these comorbidities as curiosities¹⁶.

Dysautonomia and orthostatic intolerance

This is the comorbidity that most directly changes how a session is run. The symptoms, palpitations, dizziness, a sense of impending faint, syncope, typically occur on standing and ease on sitting or lying down. They are worsened by anything that causes vasodilation: heat, effort, and digestion after a meal¹⁸.

Postural tachycardia syndrome (POTS) is its best documented form. It is defined by an increase in heart rate of more than 30 beats per minute in adults, and of more than 40 in adolescents, on standing, with no fall in blood pressure, that last point distinguishing it from orthostatic hypotension¹⁸. The relationship with hypermobility is strong in both directions: in 91 POTS patients assessed with the 2017 hEDS checklist, 31 % (28 patients) met the criteria for hEDS, and a further 24 % had generalised hypermobility without meeting the criteria¹⁹.

Be careful, however, not to turn that finding into systematic screening. The American Gastroenterological Association update is explicit on this point: the search for POTS or mast cell activation should be targeted at patients who show clinical manifestations of it, and universal screening in all hEDS/HSD patients is not supported by current data¹⁷.

Digestive problems

Two patients in three report at least one chronic digestive symptom, with an odds ratio of 4.29 compared with controls. Heartburn is the commonest (34.7 %), gastro-oesophageal reflux concerns 41.3 % of patients, and 44.2 % have a disorder of gut-brain interaction, functional dysphagia being the commonest among them at 34.2 %¹⁶.

For the physiotherapist, the consequence is practical more than theoretical: strict supine lying and trunk flexion exercises may be poorly tolerated with reflux, scheduling immediately after a meal is to be avoided, and a patient who eats little and poorly will have limited resources for loaded work. The AGA update in fact recommends the reverse of the usual reflex: that gastroenterologists following disorders of gut-brain interaction ask about hypermobility and include the Beighton score in their practice¹⁷.

Chronic pain and central sensitisation

The hypermobile patient's pain is not only mechanical. It combines nociceptive components, the real loading on tissues that defend themselves poorly, neuropathic components, and central sensitisation mechanisms that explain the frequent disproportion between observable lesions and reported intensity²⁰. Work conducted in adolescents with HSD or hEDS has explored these signs of central sensitisation specifically in this population²¹.

That dimension has a direct consequence for dosing: in a sensitised patient, the intensity of pain during exercise is a poor guide to the real tissue load. Relying on it leads to under-loading indefinitely. We return to it in the next chapter, because that is exactly the crux of the problem.

Fatigue

Almost one patient in two reports chronic fatigue¹⁶. It is multifactorial, poor quality sleep, night pain, dysautonomia, deconditioning, the increased energy cost of a poorly stabilised posture, and it is very often the first reason a programme is abandoned. A session that leaves the patient floored for two days will not be repeated, however good it is technically.

The diagnostic odyssey, and what it has left behind

Most of these patients arrive with a long history. The delay in recognising the symptoms, and then the underlying hEDS or HSD, is a recurring finding in the literature, including for the dysautonomic manifestations, whose nature often takes years to be identified, which delays treatment by as much¹⁸. A recent international survey of hEDS and HSD patients documented these diagnostic difficulties, the long-term burden of comorbidities and the unmet needs²². A qualitative review devoted to the primary care pathway analysed the same difficulties from the point of view of the patient and the practitioner²³.

What should be taken from this is not a matter of compassion but of clinical efficiency: a patient who has been sent away several times with “your tests are normal” arrives with justified mistrust and often with the conviction that movement damages them. The first task is sometimes to undo that conviction, and the next chapter shows that the recent data help us do so.

How do you dose exercise without triggering a flare?

This is the central question of this article, the one that decides the patient's trajectory. Well dosed, exercise transforms; badly dosed, it costs months. And the most solid trial we have says something many practitioners do not believe.

What the best available trial says

The Danish trial by Liaghat et al. is the current reference on this question. One hundred adult patients recruited in primary care, with a hypermobility spectrum disorder and shoulder pain or instability for more than three months, were randomised into two groups for sixteen weeks of exercise at three sessions a week. The HEAVY group worked through full range, at high progressive load, supervised twice a week. The LIGHT group worked at low load, in neutral to mid range, with only three supervision sessions in total¹².

First result, at one year: high load is not superior. The between-group difference in self-reported function measured by the WOSI was −92.9 points in favour of the high-load group, but with a 95 % confidence interval running from −257.4 to 71.5, and a p of 0.268: the result is not statistically significant. The secondary outcomes were largely inconclusive, with one exception: the patients in the high-load group improved more on the shoulder-related emotions subscale (−36.3; 95 % CI −65.4 to −7.3; p = 0.014), and the effect perceived by the patient pointed the same way on the emotions and lifestyle dimensions¹².

Second result, and this is the one that changes practice: high load does not hurt more. A secondary analysis of the same trial measured pain before and after each session over sixteen weeks. Exercise-induced pain remained below 0.5 points on the numerical scale in both groups, throughout the sixteen weeks, with no difference between them. The pain trajectories were similar too. The authors conclude in so many words that these results challenge previous beliefs¹³.

How to translate that into progression

If high load is neither superior nor more painful, what should you do? The honest answer is that the choice of load is decided on criteria other than fear: the patient's goals, their preferences, the supervision time available, and their tolerance measured rather than assumed.

Two further pieces of data help set the frame. First, the trial by Spanhove et al. in 21 hEDS/HSD patients with multidirectional shoulder instability: two different home programmes, six months, and significant improvement in both: 240 WOSI points at twelve weeks and 325 at twenty-four weeks, with no difference between the programmes. Kinesiophobia, by contrast, is the only outcome that did not improve, which led the authors to suggest that a supervised, multidisciplinary approach would be more effective on that particular dimension¹⁴. Second, an analysis of effect modifiers in the Danish trial showed that psychological factors and symptom duration are associated with the response to treatment²⁴.

In other words: the variable that decides is probably not the number of kilos, but the support given and the patient's beliefs.

A workable progression

• Start where the patient can hold, not where the protocol begins. The starting point is determined by trial: a load held without deterioration of control and without worsening within 24 to 48 hours is an acceptable load, whatever its absolute value.
• Progress one variable at a time. Load, range, speed, weight bearing, volume: moving them together makes any flare uninterpretable, and it is the leading cause of programmes being abandoned.
• Use an explicit tolerance window. Pain that rises during the exercise but comes back down to its usual level within 24 hours is acceptable; pain that persists beyond 48 hours marks an overshoot and calls for a return to the previous step, without stopping.
• Do not suspend at the first symptom. Complete stopping is the most damaging response in the long term: it confirms to the patient that movement damages them, when the data show the opposite.
• Count in months. The programmes in the trials last from four to sixteen weeks, and the benefits in Spanhove's trial were still growing between the twelfth and the twenty-fourth week¹⁴. Announcing that horizon from the outset avoids abandonment at six weeks.

The question of inner range

The idea of strengthening in inner range, in the portion of the range where the muscle is most shortened, is widely taught in hypermobility, with attractive reasoning: since the joint lacks a passive brake at end of range, you may as well build active control where the muscle is in a position to provide it, and avoid the extreme positions where the capsuloligamentous apparatus is most loaded.

Proprioception

A deficit in joint position sense is a documented feature of these populations, and it is one of the rare areas where the systematic reviews report a consistent signal. The review by Reychler et al., covering six randomised trials of 20 to 57 patients and programmes of four to eight weeks, finds a significant improvement in pain or in proprioception in the intervention group whatever the type of intervention, and a systematic improvement in quality of life¹¹. The scoping review by Brittain et al., on 28 studies and 630 participants, concludes for its part that therapeutic exercise and motor control training are the effective modalities, the others resting on weak evidence¹⁵.

In practice, proprioceptive work does not have to be sophisticated: control of range without going to the end, visual feedback then removal of visual feedback, mid positions held, slow transitions, then integration into functional movements. The important point is that it combines with strengthening rather than preceding it: the two progress together.

Fatigue and post-exertional pain

This is where most programmes fail, and the mechanism is well identified. The patient has a good session, feels capable, does a bit more, and pays two days later. The experience repeats, and they conclude that exercise does not suit them.

Three principles limit that scenario. The first is to cap the good days : the instruction is not to exceed the planned volume even when everything is going well, which means explaining it, because it is counter-intuitive. The second is to spread rather than concentrate : three short weekly sessions, as in the trials, rather than one long session. The third is to distinguish explicitly, with the patient, overshoot pain from sensitisation pain: the first is proportionate to what was done and eases within one to two days, the second is disproportionate, diffuse, and does not follow the load. Confusing the two leads to under-loading a patient who did not need it.

What is the evidence actually worth, modality by modality?

This chapter is the most uncomfortable to write, because it forces an admission of how thin the foundations are on which we work. Saying so is nevertheless the condition of honest practice.

The state of the corpus, in figures

Three recent systematic reviews make it possible to measure precisely what we have, and the picture they draw converges.

Buryk-Iggers et al. counted, up to November 2020, ten eligible studies totalling 330 participants : five randomised trials, one cohort, two single-arm interventions, one retrospective study and one feasibility study. All reported improvement in at least one physical or psychological outcome, and only one non-serious adverse effect possibly related to the intervention was reported. But of the five randomised trials, only two were judged to be of high quality and at low risk of bias on the PEDro scale, and most of the non-randomised studies were classified at critical risk of bias. Their conclusion explicitly calls for rigorous, adequately powered randomised trials¹⁰.

Reychler et al., searching up to April 2020, retained six randomised trials, with sample sizes between 20 and 57 patients, programmes of four to eight weeks and heterogeneity described as enormous between the interventions. They conclude to a benefit on proprioception and pain, while noting that robust trials are lacking¹¹.

Brittain et al., with a search extended to April 2023, included 28 studies and 630 participants, mostly women, of mean age 26.2. They identify therapeutic exercise and motor control training as effective methods, and describe as weak the evidence supporting adaptive equipment, patient advice, manual therapy and functional training. They stress that further research is still needed to determine efficacy and above all the dosing of interventions¹⁵.

A word on what that means. Thirty-three years after the description of the hypermobility syndrome, the whole of the interventional physiotherapy literature fits into a few hundred patients, almost all studied at the shoulder, over periods of a few weeks. No high-level clinical practice guideline, of the kind we have for low back pain or tendinopathy, exists for this population. An editorial published in 2026 in the Journal of Orthopaedic & Sports Physical Therapy took note of that situation by challenging the profession directly on its ability to recognise these patients²⁵.

The table of modalities

What do published clinical cases teach us?

Three real, published and identifiable cases, each showing one face of the problem: what sustained rehabilitation can achieve, what the absence of active care produces, and what caution manual therapy demands.

Case 1: An adolescent who goes from a wheelchair back to walking

Published in BMJ Case Reports in 2024, this case concerns a patient in her late teens with hEDS, who had experienced a rapid functional decline to the point of depending on a wheelchair. She followed a hippotherapy programme of thirty hours. The authors report substantial improvement in fatigue and chronic pain, and a recovery of walking ability: the patient went from wheelchair dependence to walking with elbow crutches, then markedly improved the quality and speed of her gait. The authors attribute these gains to work on postural balance, motor skills, proprioception, muscle function and endurance, as well as to an effect on cognitive and emotional regulation³¹.

What to take from it, and what not to. A single case does not demonstrate the efficacy of hippotherapy, and it would be dishonest to present it that way. What it does illustrate solidly is that severe deconditioning in this population is not inevitable: an active, prolonged and sufficiently dosed intervention can reverse a trajectory that seemed established. The volume is the notable element here: thirty hours, when so many programmes stop at six sessions.

Case 2: Ten years of opioids instead of active care

Published in Cureus in 2023, this case describes a 67-year-old woman with hypermobile Ehlers-Danlos, with associated osteoarthritis and anxiety, treated with opioids, oxycodone and slow-release morphine, for a decade, until a change of general practitioner led to a reassessment. Morphine equivalents were calculated and a progressive withdrawal begun in order to reduce the risk of overdose. The authors are a reminder that chronic pain in EDS has both nociceptive and neuropathic components, and that its management must rest on a global approach combining non-opioid treatments, physiotherapy and psychological support³².

What to take from it. It is the exact counter-example of the previous case, and it is more frequent in practice. Ten years of analgesic escalation reflect first of all a failure of recognition: without a clear diagnosis, pain becomes a symptom to suppress rather than a problem to treat. The physiotherapist who receives these patients often arrives after those years, and their first task is to offer a credible alternative to a strategy that has failed slowly.

Case 3: Three patients, and a warning about manual therapy

A series of three cases published in 2020 in the Journal of the Canadian Chiropractic Association describes the presentation, assessment, management and course of patients with joint hypermobility syndrome. The authors note that recognising hypermobility as a significant contributing factor is often difficult, and that failing to know this framework delays both diagnosis and the setting up of effective care. Their practical conclusion is doubly explicit: manual therapy must be used with discernment, and active exercise is an essential element of care³⁰.

What to take from it. In a patient whose joints already lack passive restraint, techniques that remove still more have no obvious rationale, and may bring a misleading immediate relief that delays active work. That does not mean forbidding all manual contact, treating secondary muscle tension has its place, but that high-velocity manipulation at end of range requires a particularly solid justification in this population.

How do you apply all this from the next session?

An operational summary of the five previous chapters, in the order in which the questions actually arise in the clinic.

The initial assessment, in four stages

1. Score the Beighton correctly, and put it in perspective straight away. Passive, stabilised, measured, with no prior warm-up. Then record it for what it is: one item of data among others, over-scored by non-specialists in 81 % of cases⁵.
2. Examine the joints the score ignores, starting with those the patient complains about: shoulder, hip, ankle, wrist, cervical spine, temporomandibular joint.
3. Look for the red flags, and first of all the signs suggesting a vascular type. It is the only point in this assessment where an error can be serious.
4. Screen for the comorbidities that will shape the programme (orthostatic intolerance, fatigue, digestive problems, signs of sensitisation), in a targetedway, in the patients who show manifestations of them, and not systematically¹⁷.

A useful framework for organising that assessment is the one proposed by the international group of physiotherapists convened in 2017, which adopts the International Classification of Functioning to structure the assessment around functions, activities and participation, without forgetting environmental and personal factors. That text remains, nearly ten years later, the reference statement for our profession on this subject, and it already noted that many physiotherapists knew neither the criteria, nor the prevalence, nor the usual presentation³³. In children, a specific diagnostic framework was proposed in 2023³⁴.

What to say to the patient

The content of what is said counts as much as that of the programme, because psychological factors are associated with the response to treatment²⁴ and because kinesiophobia is the outcome that resists most in the home-based trials¹⁴. Four messages deserve to be put explicitly.

• “Your joints go further than average; that is not a disease in itself.” About nine carriers in ten develop no symptoms³.
• “Loading does not damage you.” Over sixteen weeks of full-range, high-load strengthening, the pain induced by the sessions stayed below 0.5 out of 10¹³. It is probably the message that changes the most.
• “A flare does not cancel the work done.” You drop back one step, you do not start from zero, and you do not stop.
• “This is counted in months.” Setting the horizon from the first session avoids abandonment at six weeks, when the benefits are not yet established.

When to refer

To clinical genetics or a reference centre when the picture suggests a heritable connective tissue disorder needing confirmation, and without delay for any suspicion of a vascular type. To a cardiologist or neurologist for disabling orthostatic intolerance symptoms that have never been investigated. To a gastroenterologist for chronic digestive symptoms affecting nutritional status. To chronic pain management when signs of sensitisation dominate the picture and block progression despite correct dosing.

Frequently asked questions