I have had frozen shoulder twice, consecutively, one on each side. Both recovered over about two years while I did a great many sensible things. I cannot tell you that any of them helped, and neither can anyone else.
I have had frozen shoulder twice. Consecutively. Once on each side.
The first took about two years to settle. The second followed a similar course. Through both, I was exercising, managing stress, sleeping better than I had been, and eating well — because that is what I do, and because when something hurts for two years you try things.
Both shoulders recovered.
And I cannot tell you that any of it helped.
Frozen shoulder — adhesive capsulitis — is unusual among musculoskeletal conditions in that it very often resolves on its own. Not always, not completely, but the natural course is toward recovery over a period measured in years.
Hand and colleagues followed 269 shoulders for a mean of 4.4 years. Around 41% had ongoing symptoms at follow-up — but of those, 94% were mild. Vastamäki and colleagues reported that 94% of untreated patients recovered to the range of motion of their other shoulder, with a mean duration of about 15 months.
Read that again, because it is the crux. Most people get better without treatment, over roughly the timescale that any treatment would be running.
Which means: if you do something for two years while a condition that resolves in two years resolves, you cannot distinguish your intervention from the passage of time. Not by feel, not by conviction, not by how carefully you tracked it.
I have a strong intuition that the work I did helped. I also know that my intuition is exactly what this confounder is designed to fool.
Because frozen shoulder is a treatment-claim factory.
Two years of pain and disability, resolving eventually, in a condition where almost anything applied during that window gets the credit. Manual therapy, dry needling, supplements, specific exercise protocols, hydrodilatation, injections, devices, diets. Every one of them can point to people who got better while using it.
They did get better. That part is true. The question nobody asks is what would have happened anyway.
Any claim about frozen shoulder treatment that does not address natural history is uninterpretable. Including mine about my own shoulders.
The honest position for any practitioner — and I include the ones I have paid — is: the natural history is favourable, our treatments may accelerate it or improve the experience of it, and separating those requires controlled comparison rather than testimony.
The more interesting development is not about treatment at all.
Ng and colleagues published a single-cell atlas of frozen shoulder tissue in Nature Communications in 2024, and the picture is genuinely unusual. The condition involves fibrosis — the laying down of fibrous tissue — but unlike most fibrotic conditions, this one resolves.
Their work implicates crosstalk between MERTK-positive macrophages and DKK3-positive, POSTN-positive fibroblasts. Macrophages of that type are associated with the resolution of inflammation and with tissue repair rather than ongoing damage.
Which frames frozen shoulder as something quite specific: a self-limiting fibrosis, in a body that appears to know how to switch it off again.
That is scientifically interesting well beyond the shoulder. Fibrosis in the lung, liver and kidney is generally progressive and a major cause of illness. A tissue that fibroses and then reverses it is worth understanding in detail.
A caution: this is mechanistic work, not a treatment. It does not tell you what to do on a Tuesday morning with a painful shoulder. But it does explain why the natural history is what it is, which makes the natural history harder to ignore.
Nothing here says do nothing. It says be honest about what you can claim.
Range of motion work matters — not because it demonstrably shortens the condition, but because the shoulder you end up with at the end depends partly on what you asked of it along the way.
Pain management matters in its own right. Two years is a long time to hurt. Improving that is worth doing even if it changes nothing about the timeline.
Thoracic mobility and scapular control are worth attention, because the shoulder does not work in isolation and stiffness elsewhere increases demand on it.
And there is one genuinely actionable clinical association, which is the most useful thing in this article. Frozen shoulder is strongly associated with diabetes and with thyroid disease. If you have developed one and nobody has checked your HbA1c or your thyroid function, that is worth raising with your GP. Not as an explanation for the shoulder, but because a condition that flags an undiagnosed metabolic problem has done you a favour.
I had two frozen shoulders. They both recovered. I did a lot of things during that time.
I do not know whether any of it helped, and I am not going to imply otherwise, because doing so would be the same move I criticise in others — and my sample size is two, uncontrolled, in a condition that resolves on its own.
That is a less satisfying story than the one I could tell. It has the advantage of being true.
If you have a frozen shoulder: it will very probably get better. It will take longer than anyone wants. And you should be sceptical of anybody — me included — who tells you they know what fixed it.
This article is general information and not medical advice. Persistent shoulder pain, particularly with night pain or restricted movement, deserves proper assessment.
Sources
Hand C, Clipsham K, Rees JL, Carr AJ. Long-term outcome of frozen shoulder. Journal of Shoulder and Elbow Surgery 2008;17(2):231-236.
Vastamäki H, Kettunen J, Vastamäki M. The natural history of idiopathic frozen shoulder: a 2- to 27-year followup study. Clinical Orthopaedics and Related Research 2012;470(4):1133-1143.
Ng J, et al. Single-cell transcriptomic atlas of frozen shoulder. Nature Communications 2024.
Frozen shoulder is strongly associated with diabetes and thyroid disease. If nobody has checked yours, raise it.
Ask AIdan → See the TDG Programme