Frozen shoulder: the movement that identifies it, the two phases, and what actually helps in each

Nobody can tell you how long this lasts, and the honest reason why

You will read that it resolves on its own in one to three years. Hold that number loosely. The care people actually receive is so inconsistent that the "natural history" everyone quotes is really an average over wildly different treatments.

A 2026 review pulled the patient information leaflets that thirty-eight NHS trusts hand out for this condition. Not one of them contained everything national guidance recommends, and they disagreed with each other about painkillers, activity, exercise and injections.

That is the same health system, the same condition, the same year. If the written advice varies that much, the treatment varies more — and the "one to three years" figure is a smear across all of it.

The two phases, and why the order matters more than the labels

This condition changes character partway through, and the treatment that helps in the first half is close to useless in the second.

Phase one is pain. The shoulder aches at rest, wakes you at night, and hurts before it stiffens. The capsule is angry. Stiffness is arriving but pain is the complaint.

Phase two is stiffness. The pain settles. What is left is a shoulder that will not move. Now the capsule is not inflamed so much as short.

Almost every disagreement about how to treat this comes from people applying a phase-one treatment in phase two, or the reverse. An injection aimed at an angry capsule does much less for a merely short one. Aggressive stretching of an angry capsule makes it angrier.

When practitioners around the world were surveyed about steroid injections for this in 2025, they agreed strongly about the timing: 191 of 233 said the right moment to inject is during the pain-predominant phase.

The same survey shows what agreement is worth. They agreed on when. They did not agree on what: preparations varied enough that the authors concluded many of the choices come down to personal preference or whatever the local guideline happens to say.

Most people are diagnosed too late to be in the trial that would have helped them

A British trial tried to test an anti-inflammatory injection specifically in early, pain-predominant frozen shoulder. It could not recruit enough people. The reasons the researchers give are the useful part: most patients who came through physiotherapy had already moved past the early stage, or had already decided what treatment they wanted.

Read that as a fact about how this condition is actually experienced. By the time frozen shoulder has a name attached to it, the window in which the early treatments were designed to work has usually closed. If your shoulder is in the painful phase now, you are in the position most of the research wishes it could study, and you will not be there long.

What the evidence supports, strongest first

A steroid injection into the joint, during the painful phase. This is the best-supported active treatment and the professional consensus is clear about the timing. It is not a cure — it buys a better few months and makes the movement work possible.

Adding fluid distension to that injection buys you nothing. This one is worth knowing because it is offered constantly. A double-blind randomised trial compared a cortisone injection alone against the same injection followed by saline pushed in until the capsule stretched or resisted. Both groups improved on every measure. There was no significant difference between them at eight weeks.

So if you are offered hydrodilatation as an upgrade, the trial evidence says it is a more elaborate version of the injection you were getting anyway.

Whether that changes if your rotator cuff is torn. A 2025 cohort found distension relieved pain in every subgroup, but the movement people got back depended on the state of the cuff — partial and full-thickness tears recovered less range. The authors also saw smaller gains in the diabetic patients, and were careful to say that group was small.

That is the argument for imaging before an injection rather than after a disappointing result.

Platelet-rich plasma has better numbers than steroid at three and six months, from weaker studies. A pooled analysis favoured it on pain, function and range, and the effect held or grew to six months.

Take the effect sizes with the design in mind. These are small studies with varied preparation methods, and the comparison group is a treatment already known to fade. Better than something that wears off is a lower bar than better than nothing at a year.

Guidance itself is built on thin ground, and says so. The British Elbow and Shoulder Society's care pathway searched the literature to 2023, graded what it found, and states plainly that where high-quality trials were missing it had to write narrative reviews instead.

A national body admitting the evidence ran out is more useful to you than a confident website. It tells you which parts of your treatment plan are established and which are convention.

If you have diabetes, this is a different conversation

Frozen shoulder turns up far more often in people with diabetes, it tends to be more stubborn, and the movement recovered after treatment tends to be less. None of that means treatment does not work. It means the timeline you should expect is longer than the one a friend without diabetes reports, and that a steroid injection will move your blood sugar for several days — worth planning for rather than discovering.

Ten people with a frozen shoulder

Ten people lose outward rotation and get the diagnosis. Most are already past the painful phase by the time somebody names it, which is why the trial above could not fill. Of the ten, the ones who present while it still hurts at night are the ones an injection has most to offer, and they are the minority. Everyone gets told one to three years, from evidence that is really an average over thirty-eight different sets of advice. The ones with diabetes take longer and get back less range. The ones with a torn rotator cuff get pain relief from an injection but less movement.

What none of them can be told is which they are. That is not a gap in this page; it is a gap in the field, and the honest version of it is the one the specialist society wrote down.

Why no compound has been tested on a shrinking capsule

Nothing sold as a repair peptide has been tested in frozen shoulder in a randomised human trial. There is no result to report, and there is no point discussing mechanism at length on a page about the condition — that argument belongs on a page whose title names both the compound and this condition, with its evidence tier stated. The operator of this site has a commercial interest in compounds of that kind, which is why the sentence here is the absence of the trials rather than a case built around it.

The order these things need doing in

1. Check the passive movement yourself. Elbow at your side, someone else turns your forearm outward. If it stops early no matter who is moving it, you are probably in the right article.
2. Work out which phase you are in. Waking at night with an aching shoulder is phase one. A quiet but immovable shoulder is phase two. This decides what is worth doing.
3. If you are in the painful phase, do not wait it out. That is the window the injection evidence is built on, and most people miss it.
4. Ask for imaging before an injection, not after. A torn cuff changes what range you should expect back.
5. Decline the upgrade unless someone can say why. Distension on top of cortisone showed no advantage at eight weeks in a double-blind trial.
6. If you have diabetes, set the expectation now. Slower, less range back, and a few days of higher readings after a steroid injection.