Running with hypermobility

• About
• Latest Posts

Adam Foster

Director at The Fibro Guy

Adam Foster is the founder of The Fibro Guy and one of the leading voices in hypermobility, EDS, and chronic pain rehabilitation. A former Reconnaissance Soldier, he developed chronic pain after an IED blast in Afghanistan and built his approach from the ground up when the medical system had no answers. Since 2009 he has worked with thousands of people across eight countries, drawing on pain neuroscience, motor learning, and sensorimotor rehabilitation. He co-authored "No Pain, No Pain" with trauma psychologist Dr Mairi Harper, directed the documentary "Invisible Monster," and has been featured in Forbes, BBC Radio 4, ITV News, LiveScience, and The Telegraph.

Latest posts by Adam Foster (see all)

• Hypermobile hip exercises we use, and why clamshells aren’t one of them - 17 October 2026
• Is Fibromyalgia a Disability in the UK, and Can You Claim PIP? - 14 October 2026
• Fibromyalgia and Central Sensitisation, What Is It in Plain English? - 8 October 2026

Running with hypermobility tends to get answered in two ways, usually by people who are very sure of themselves. One camp says running is fine, crack on, movement is medicine, and the other says that impact and lax joints don’t mix and you should be in a pool. The truthful answer is that it depends, and in our opinion the thing it depends on most is what your leg actually does in the half second it spends on the ground.

The other thing worth saying early is just how thin the ground underneath all of this is, as nobody has published any research on hypermobile running gait ^[1]. Not a small literature, not a mixed one, none at all. So every running specific mechanical claim you’ll read about hypermobility, ours very much included, is extrapolated from general running biomechanics, from walking research, and from rehab work in related groups. That doesn’t make it useless, it does mean that any exact joint angle target or neat week by week timeline is an inference dressed up as a finding, and worth treating as one.

What has been looked at reasonably well is what a normal leg is supposed to do on landing, how a movement pattern gets built and then kept, and whether those with hypermobility generally do better moving or resting, and those three things between them get you a surprisingly long way.

Nobody Has Actually Studied This

What we know about hypermobility and the lower limb is genuinely useful, it just isn’t about running. Joint position sense in the lower limb is often poorer in those with hypermobility, particularly at the knee, and that’s about as consistent a finding as this field currently has ^[2][3]. It tends to arrive alongside differences in strength and in functional performance, rather than turning up on its own ^[4]. Walking has been measured properly, and it shows altered mechanics with less power absorbed and generated at the knee through the gait cycle ^[5]. The tissue behaves differently as well, as reduced tendon stiffness and altered muscle and tendon behaviour turn up fairly consistently in hEDS and HSD ^[6].

All the ingredients of a force absorption problem are sat right there then, what nobody has actually done is put a hypermobile runner on a force plate and measured what happens to the leg ^[1].

Now, the second thing worth knowing is which way the evidence has been moving, as it’s moved a long way in not very many years. The old position, that those with hypermobility should avoid sport, has quietly collapsed. The current position is risk stratified participation, meaning the question isn’t whether you’re allowed to move, it’s which movement, how much of it, and how it gets built up ^[7][8]. Inactivity is generally the worse option in symptomatic HSD and hEDS, as pain, fatigue, fear, deconditioning and poorer sensorimotor control all feed each other quite happily, while graded exercise and motor training improve function and quality of life ^[7][9][10].

That’s the backdrop, and running isn’t excluded from it, it simply hasn’t been studied.

Hypermobility Isn’t One Thing

“Hypermobile” covers an enormous amount of ground, as it covers people who are just hypermobile and have never had a day’s bother with it, people with hypermobility spectrum disorder, people with hypermobile Ehlers Danlos syndrome, people whose laxity arrived after an injury rather than being there from the start, people with skeletal dysplasias, and a small number of rarer connective tissue conditions where the stakes are considerably higher ^[11][12]. Those groups don’t share an exercise answer, which is why blanket advice about running is nearly always wrong for somebody reading it.

The genetics don’t tidy it up either. Most forms of Ehlers Danlos have an identified molecular cause, usually sitting in collagen or somewhere else in the extracellular matrix ^[13][14]. hEDS is the exception, and unfortunately it’s the big one, as there’s still no confirmed molecular test for it and the underlying cause remains genuinely unknown ^[12][15]. So “faulty collagen” as a blanket explanation is much too broad for the group most people reading this actually belong to, which is a bit awkward really, given how many times you’ll have been told exactly that in a ten minute appointment.

Hypermobile Because of the Hips

Hip dysplasia and generalised hypermobility do turn up together more often than chance would explain, although how much more depends enormously on who’s being counted and how hypermobility is defined, and the published estimates sit a very long way apart from one another ^[16][17]. Which is worth knowing mostly so that you don’t take any single prevalence figure as gospel, including the ones that get quoted very confidently in hypermobility circles.

The blanket advice that running is off the table after hip dysplasia doesn’t really hold up though. Dysplasia does increase the risk of osteoarthritis and it does change the mechanics of the joint ^[18], but return to sport after a periacetabular osteotomy is often good, and that includes participation in high impact sport for a decent proportion of competitive athletes ^[19][20]. The evidence is limited and outcomes vary a great deal between individuals, so how much impact your hip tolerates is something to work out alongside your surgical team over time, rather than something anybody can read off a page.

Hypermobile After a Knee Injury

Most ACL injuries happen without anything touching the injured knee at all, and the best supported estimate is over half rather than the seventy per cent figure that gets repeated everywhere ^[21][22][23]. Which matters a great deal, as it means the mechanism is usually about how the leg is controlled during a landing or a change of direction, rather than about bad luck and a stray opponent.

Hypermobility and ACL risk is a great deal messier than it usually gets presented. Generalised joint hypermobility has been linked to increased ACL injury risk and to poorer outcomes after reconstruction ^[24], but the forward looking evidence for a first injury in women is genuinely mixed, with laxity, hypermobility and knee hyperextension not emerging as risk factors for a first tear in one long prospective follow up ^[25][26]. The signal gets a lot clearer on the second injury, as hypermobility raises the odds of a further ACL injury within a year of returning to sport after reconstruction, and excess knee hyperextension carries the same warning ^[27][26].

So running again after ACL surgery is normal and expected, the return itself is the risky window if you’re hypermobile though, and that’s an argument for building the landing properly before you get there, rather than an argument against running.

Hypermobile Because of a Skeletal Condition

Morquio syndrome sits at the other end of the range entirely. It’s a lysosomal storage disorder in which the body can’t break down certain glycosaminoglycans, and the consequence is severe skeletal dysplasia along with real cervical spine risk ^[28][29]. Treating running as high risk in this group is sensible, although that’s a cautious inference from the skeletal, spinal and airway picture, rather than something anybody has established by studying running in this population ^[28][30].

Down’s syndrome comes up a lot as well, and honestly, we can’t answer that one from the research, as the question hasn’t been looked at in terms that would help you decide anything. Our position there comes out of the studios rather than out of the research, and it’s the same as everywhere else on this page: graded, built up slowly, with the landing taught rather than assumed.

What Your Leg Is Meant to Do in the Half Second It’s on the Ground

Right, the mechanics. Running, in stance, behaves like a spring, as energy is absorbed during the first half of ground contact and a good chunk of it is handed back during the second half to help push you along ^[31][32]. Your centre of mass drops to its lowest point around the middle of stance and then rebounds out of it ^[31]. So a landing is a thing you use, rather than a thing you survive.

What that requires is a limb that’s compliant but organised, rather than one that meets the ground rigidly ^[33], which is where “bend your knees” stops being enough on its own, as the load in early stance is shared across the ankle, the knee and the hip together, with the whole leg behaving like a spring and a damper rather than a single hinge ^[34][32]. At lower running speeds the quadriceps do a lot of the braking and support early in contact, while the calf and ankle carry a lot of the vertical support and most of the propulsion ^[34].

The ankle is the real star of this, and it’s underrated by pretty much everybody, including a fair few physios. Across running speeds, the ankle plantar flexors handle the majority of the energy generation and absorption happening in the whole limb ^[35]. The Achilles stores a serious amount of elastic strain energy on the way in and returns it on the way out, and that storage and return accounts for a large share of the energy you save while running ^[36]. Joint contributions do shift with speed, with slope, with fatigue and with technique though, so none of this is a fixed allocation ^[35][37].

Now, tendons also do something else, and it matters a great deal if your tissue is more compliant than average. A tendon sitting in series with a muscle can take energy in quickly and then release it more slowly, which drops the peak power the muscle fibres have to deal with and slows down how fast they get lengthened ^[38]. That’s a shock absorber in the proper engineering sense, and it shows up most obviously in hard deceleration tasks ^[39]. So the tendon is doing two jobs at once, storing energy for propulsion and protecting the muscle while it does it.

Good absorption also starts before you land, which does surprise people. Muscles are switched on ahead of foot contact, which reduces how far they get stretched during early absorption and tunes the stiffness of the limb for the impact that’s about to arrive ^[34][40]. Runners with high training mileage show adjustments in exactly that pre landing activity as speed goes up ^[40], and biomechanical strategies during running differ between people who have recently been injured and people who haven’t ^[41]. So the leg is set up in advance, and it’s set up using whatever information the nervous system currently has about where the ground is and where your joints are.

Then there’s damping, which is the part nobody mentions. Foot strike sets soft tissue vibrating, and muscle activity increases the damping of those vibrations, particularly as the input frequency gets close to the tissue’s own resonance ^[33][42]. Most runners adjust their damping when their footwear changes, although the footwear effects themselves are all over the place from one person to the next, so there’s no shoe answer hiding in there unfortunately ^[43][33].

The whole sequence, then, is pre activation, controlled yielding, rebound. Landing with more momentum increases pre stiffening and distal control, and can also bring in a compensatory yielding strategy that allows energy to be absorbed in a controlled way and protects the joint while it happens ^[44]. Greater calf activation goes along with greater leg and vertical stiffness during stance, especially at higher speeds ^[45], and how much energy the soft tissues dissipate is tied to leg stiffness across different step frequencies ^[46].

That’s a lot of machinery, and essentially none of it is conscious.

Where the Load Goes When the Landing Isn’t Organised

Force absorption is partly a question of routing, as the load is going somewhere regardless, and the landing decides where it ends up.

Foot strike is the obvious lever and it’s easily the most oversold one. Landing on the heel tends to produce higher vertical loading rates and more load through the knee and the kneecap, while landing on the forefoot lowers the impact loading rate and moves load onto the ankle and the Achilles instead ^[47][48]. In joint contact terms that’s a real shift rather than a rounding error, as forefoot landing raises early stance contact forces substantially at the ankle and to a lesser degree at the knee ^[49]. So switching your foot strike is a redistribution rather than a fix, and if the ankle and the Achilles are the parts already complaining, it’s a redistribution in precisely the wrong direction. There’s no evidence at all for prescribing forefoot running to those with hypermobility ^[1].

The angle at contact matters more than the label does anyway. Among heel strikers, the angle of the foot at initial contact related strongly to how quickly load came on, and an atypical heel strike pattern produced the highest loading rates of the lot, which fits with not making much use of the available strategies for absorbing impact ^[50]. So it’s the organisation of the landing that counts, rather than which bit of the foot arrives first.

Surface plays a part too, as stiffer ground raises vertical ground reaction forces and increases knee and hip moments late in stance, while softer surfaces reduce some joint loads ^[51]. It isn’t an enormous effect, but if you’re sitting right at the edge of what you tolerate, the towpath is a kinder place to be than the pavement.

Cadence is the lever with the best return for the least fuss. Taking your step rate up by a small amount at the same speed reduces the mechanical energy absorbed at the knee, and pushing it up further starts reducing hip absorption as well ^[52], and it works because shorter steps put your foot closer to underneath you, which cuts the braking impulse and reduces how far your centre of mass travels up and down on each step ^[52]. Retraining cadence over a few months shifts impact peak and loading rates too, so it isn’t only a within session trick ^[53].

Impact Isn’t the Villain It Gets Made Out to Be

External force measurements don’t map cleanly onto what the bone is actually experiencing. Ground reaction force metrics correlate poorly with tibial bone load across different speeds and slopes, because most of the force going through the tibia while you run comes from your own muscle contraction rather than from the ground ^[54]. So a big external impact signal doesn’t automatically mean a big internal load, which has some fairly obvious implications for the impact score your watch keeps showing you ^[54].

Now, the injury link is patchier than the running internet suggests. Runners who had been injured did show higher vertical loading rates and higher vertical stiffness at initial loading than uninjured runners, but that was driven mainly by particular injuries, kneecap pain and plantar fasciitis especially, rather than by injuries across the board ^[55]. Pooled across studies, the evidence is stronger for loading rate in people with a history of stress fracture than it is for running injury in general ^[56]. And looking forwards rather than backwards, in a large group of recreational runners followed over time, vertical impact peak and loading rate weren’t risk factors for a future injury at all, while longer steps, shorter contact time and a lower duty factor were ^[57].

The shape of the loading curve may well matter more than its height. Early and erratic maximal loading or unloading rates look like they reflect a poor transition from swing into stance, instability, or poor impact attenuation, rather than simply too much force arriving ^[58]. Whether impacts cause injury at all, in the straightforward way the idea usually gets stated, is genuinely still open ^[59].

Which brings it back round to the same place, as the size of the force is a great deal less interesting than whether the leg is organised to receive it.

What Fatigue Does to a Landing

Fatigue does something specific and fairly predictable to a landing. Leg stiffness drops, contact time gets longer, movement variability goes up, and work shifts away from the ankle and onto the knee and the hip ^[60][61]. Muscle coordination reorganises while it happens, with the synergies driving the leg changing measurably through a run ^[62]. The body holds onto the speed and pays for it with a less efficient and probably more injurious strategy further up the chain ^[61].

Now, read that back with hypermobility in mind.

Fatigue is already a large part of the picture for a lot of those with hEDS and HSD, and it feeds the same loop that pain, fear and deconditioning do ^[7][9][10]. If the ankle stops doing its share and the knee and the hip take it on instead, then the joints doing the extra work, are the same ones already relying on a noisier signal to know where they are in the first place ^[2][3].

We can’t prove that chain end to end, as nobody has measured it in hypermobile runners ^[1]. It’s a guess though, like a great deal of things in this space, and it’s a guess that fits what’s been measured in everybody else.

There’s a related finding worth borrowing from the ankle instability literature, with the usual caveat that it’s a different population. People with chronic ankle instability run with altered knee joint loading, so a problem at the ankle doesn’t politely stay at the ankle ^[63]. The mechanical differences show up in a task specific way rather than as one global change across every movement ^[64], and the neuromuscular response to repeated perturbation follows distinct patterns compared with people without instability ^[65]. That isn’t hypermobility though, so it’s a reason to suspect that a poorly controlled ankle sends its problems upstairs, rather than a reason to believe it.

Proprioception, and Why More Balance Work Isn’t the Whole Answer

Joint position sense in the lower limb is often reduced in those with hypermobility, especially at the knee ^[2][3], and it comes packaged with differences in strength and functional performance rather than sitting on its own ^[4]. Reactive balance behaves differently as well, as the way balance reactions adapt after being pushed forwards differs in joint hypermobility compared with controls ^[66].

So the sensory side is real, and it does matter. Where it goes wrong is in the strong version of the argument, the one where ordinary exercise supposedly doesn’t work and only clever neural work does. Therapeutic exercise and motor function training come out as effective overall across hEDS and generalised HSD, even though the specific protocols vary enormously from one study to the next ^[9][7]. Dismissing exercise as a category isn’t supported by any of that.

The brain against muscle framing doesn’t hold up either, and it’s a framing this corner of the internet is incredibly fond of. Reduced tendon stiffness and altered muscle and tendon behaviour are documented in hEDS and HSD, and resistance training is one of the very few interventions shown to increase tendon stiffness at all ^[6]. Neural control and tissue capacity both matter, and the useful question in our opinion is the order you build them in, rather than which one of them is the real problem.

Perturbation training gets oversold too. People with symptomatic hypermobility do respond differently to perturbation, which is a decent reason to think that reactive balance control is worth attention ^[66], but nobody has shown that perturbation training prevents dislocations, and nobody has shown that it’s central for hypermobile runners specifically. In change of direction work aimed at reducing ACL loading, adding perturbation to plyometric training wasn’t clearly effective at reducing hazardous knee loads ^[67]. Feedback and feedforward responses during perturbed running have been mapped carefully in people without symptoms ^[68], which tells you the mechanism is there, and tells you nothing at all about whether training it transfers to a hypermobile runner on a wet pavement.

How a Landing Pattern Actually Gets Built

Motor learning has two things going on at once rather than one after the other. There’s an explicit side, the conscious change you make because somebody told you to shorten your stride or look further up the road, and it works quickly ^[69][70][71]. And there’s an implicit side, which keeps movement calibrated in the background by updating an internal model from the difference between what the nervous system predicted and what it actually got ^[69][72]. The old fast and slow model maps onto that roughly, with the fast part looking like conscious strategy and the slow part looking like implicit adaptation, although implicit learning contains several sub processes of its own ^[73][74].

A more useful modern framing splits it three ways: working out what your action does to the outcome, refining the movement itself, and then retrieving the right control policy at the right moment ^[72]. That last one is the part rehab usually forgets about entirely, as learning something in a session isn’t the same thing as having it available on a hill, in the rain, at mile four, when you’re already tired.

Implicit change is also a great deal faster than it used to be given credit for, with some of it appearing within the first few attempts ^[74]. So the idea that nothing has changed until six weeks have gone by is wrong, although having something change and having it stay are two different problems and only one of them is quick.

Running makes all of this harder than a reaching task does, for a structural reason. Gait is a continuous stability problem, as every step changes the starting conditions for the next one, so the nervous system has to improve the movement without losing balance while it does it ^[75]. You can’t take a stride out, have a look at it, and put it back.

What happens when gait gets deliberately disrupted is well described. People make fast reactive corrections first, then gradually recalibrate the underlying locomotor commands over repeated exposure ^[76][77]. The cerebellum sits at the centre of that error driven recalibration, particularly for updating the forward models that predict what a movement is going to do ^[76][77]. Proprioceptive feedback drives the early adjustment in limb muscle activity, while proprioceptive and visual information together feed the planning and the storage of the new pattern higher up ^[78].

And note the framing there, as it isn’t the one you usually get sold. It’s sensorimotor integration and internal models being updated ^[78], rather than a map somewhere being redrawn.

What practice actually changes is genuinely interesting. Repeated disturbance produces rapid balance adaptation, with most of the gain arriving after only a handful of exposures ^[79]. Early on the nervous system recruits extra and somewhat redundant muscle synergies, and then prunes them down into something more efficient as it goes ^[79]. Most of the underlying coordination patterns stay stable through adaptation, while a smaller subset gets re timed or selectively recruited to meet the new demand ^[80]. And with practice, implicit adaptation starts to stabilise a noisy conscious strategy, which is a decent chunk of why skilled movement gradually stops feeling like hard work ^[81].

So the arc is attention heavy corrections first, then a solution that’s faster, more efficient and considerably cheaper to run ^[81][71]. A new cue can help you immediately, and durable change needs repetition under the constraints you’ll actually meet ^[70][82].

Learning is also context specific, and that’s the part that quietly ruins a lot of rehab programmes. What you learn on a treadmill doesn’t fully transfer to running overground, most likely because the sensory environment differs and the nervous system files the error against a context specific model rather than a shared one ^[76][82]. Drills, treadmill work and perturbation exercises all have a place, and transfer is strongest where the training resembles the real task ^[82]. Different interventions also load you differently while you’re doing them, as rhythmic cueing carries a bigger cognitive burden while split belt adaptation runs more autonomously on proprioceptive feedback ^[82].

Practice has to target the actual coordination problem, in other words, rather than isolated strength or isolated instructions ^[83]. And people differ, genuinely and measurably, as faster and slower adapters show different cortical dynamics in the parietal, visual and sensorimotor areas during gait adaptation ^[84][85]. So if yours has taken longer than somebody else’s doing the same thing, that’s one of the more likely reasons.

What We Do in the Studios

Everything above is the evidence. This bit is us, as it’s what we’ve built up in the studios over the years rather than a tested protocol, and there’s a real difference between the two.

The order we work in is sensory clarity first, then the movement pattern, then load. Not because load doesn’t matter, it clearly does ^[6], but because loading a pattern the nervous system can’t reliably reproduce mostly just makes that pattern more available.

Sensory first means giving the joint and the skin something clearer to report before you ask the brain to make fine decisions off the back of it. Laxer tissue degrades the quality of the signal before it ever reaches the brain, and a brain making movement decisions on a noisier signal, makes worse ones. That’s our reading of why joint position sense comes out poorer at the knee so consistently ^[2][3][4], and it’s a reading rather than a demonstrated mechanism, so please do take it as such.

Then the landing, taught in sequence, as a skill rather than as a cue. What we’re after is the same thing happening every time: the foot arrives, the ankle takes its share, the knee and the hip take theirs, and the leg gives the energy back rather than swallowing it. Variability in the practice, consistency in the outcome. That’s a motor learning principle rather than a running one, which is why we build patterns through varied repetition rather than the same drill done two hundred times in a row (nobody has ever got better at anything doing that, in our experience, though plenty of people have got very bored).

And then load, and only then. Tendon stiffness responds to resistance training ^[6], and a stiffer, better organised tendon is both a better shock absorber and a better spring ^[36][38]. Doing that work before the landing is reliable mostly just loads a landing you don’t particularly want repeating.

We don’t give week numbers for any of this. Individual differences in how fast people adapt are large and at least partly built in ^[84], so a fixed timeline is a guess rather than a plan, and we’d rather not hand you a guess with a number on it.

The Practical Bits

– Cadence before foot strike: If you’re going to change one thing about how you run, make it a small increase in step rate at the same speed, as it reduces the energy your knee has to absorb and puts your foot closer to underneath you ^[52][53]. Changing which part of your foot lands first mostly moves load from the knee to the ankle, which is only an improvement if the knee was the problem in the first place ^[47][48][49].

– Stop before the form goes: Fatigue reliably takes work away from the ankle and hands it to the knee and hip, drops leg stiffness and lengthens contact time ^[60][61]. The end of a run is where a landing pattern degrades, so finishing while it still looks like the thing you practised is worth considerably more than the extra kilometre.

– Surface, within reason: Stiffer ground raises ground reaction forces and increases knee and hip moments late in stance ^[51]. It’s a small lever rather than a solution, but if you’re near your limit anyway, softer ground is the easier place to be.

– Don’t read too much into an impact score: The external force your watch estimates doesn’t track the load inside the bone particularly well, as most of the force through the tibia comes from your own muscles ^[54]. And in a large group of recreational runners followed forwards in time, impact peak and loading rate didn’t predict who ended up injured ^[57].

– Stretching before you run: Nothing in the evidence establishes that stretching before a run raises injury risk in those with hypermobility, and it doesn’t establish the opposite either. Activity specific preparation is standard practice, and the effects of stretching depend on the type, the duration and the context, so treat confident advice in either direction with a bit of suspicion.

– Build the landing where it counts: What gets learned on a treadmill doesn’t fully transfer to the pavement, as the nervous system files it against the context it happened in ^[76][82]. Practise in conditions that resemble the running you actually want to be doing.

– Coming back after an ACL reconstruction: Returning to running is normal and expected, and the first year back in sport is the window where hypermobility raises the odds of a second injury ^[27][26]. That’s an argument about how carefully the return gets built, rather than an argument against running.

What Nobody Knows Yet

Nobody has run the study, as there’s still no published research on hypermobile running gait, which means there are no validated running specific mechanics and no validated rehabilitation model for hEDS or HSD, however confidently anyone presents one ^[1][7]. Everything in the mechanics sections above is drawn from runners without hypermobility, and everything about how learning works is drawn from general motor learning research, some of it in other populations entirely.

The direction of travel is at least clear, as the evidence has moved away from telling those with hypermobility to avoid sport and towards working out which activity, in what dose, with what built underneath it ^[7][11].

The bigger open question is which measurable features of absorption actually predict injury or good adaptation in real runners, and that one isn’t settled in anybody yet, let alone in the groups nobody has got round to studying ^[59][1].

The Fibro Guy

References

[1] Hook, J., Kendall, A. and Chappell, A. (2025) ‘Walking and running in people who are hypermobile: A scoping review’, Gait & Posture. https://doi.org/10.1016/j.gaitpost.2025.06.013

[2] Smith, T.O., Jerman, E., Easton, V., Bacon, H., Armon, K., Poland, F. et al. (2013) ‘Do people with benign joint hypermobility syndrome (BJHS) have reduced joint proprioception? A systematic review and meta-analysis’, Rheumatology International. https://doi.org/10.1007/s00296-013-2790-4

[3] Rombaut, L., De Paepe, A., Malfait, F., Cools, A. and Calders, P. (2010) ‘Joint position sense and vibratory perception sense in patients with Ehlers–Danlos syndrome type III (hypermobility type)’, Clinical Rheumatology. https://doi.org/10.1007/s10067-009-1320-y