Frozen shoulder rehabilitation: if you have just received this diagnosis, you have most likely also been told the most common phrase about this condition: "it goes away on its own, it takes one or two years". It is a comfortable phrase, and it is also the one that holds up least when you look at what actually happens to people left without treatment.
This article takes both the good part and the uncomfortable part seriously. The good part is that movement does come back and that exercises have a proven role. The uncomfortable part is that the story of the "three phases" with automatic healing at the end has no evidence to support it, and that aggressive stretching, the thing that seems most logical to do when a shoulder is stuck, gave worse results than simply moving within the limit of pain in the only direct comparison.
⚠ Important: read before continuing
The exercises and advice in this article are general indications and may not suit your situation. Everyone has a unique clinical history: what works for others could be counterproductive for you.
Want a professional's opinion? Start with the questionnaire: you register and see who will follow you, a physiotherapist registered with the Albo chosen for your case, whom you write to in chat. If you would rather hear their voice, the first consultation is free and we look at your situation together.
Before the exercises: when a stiff shoulder is not a frozen shoulder
The technical name is adhesive capsulitis; the old term scapulohumeral periarthritis, which you still hear, was an umbrella label for very different shoulder pains, and that is why it has been abandoned.
This section comes first because it matters most. A shoulder that moves little and hurts is a symptom, not a diagnosis, and some causes need a doctor before a programme of exercises.
Go to the emergency room or call 118 (the Italian emergency number) if:
- shoulder or arm pain appears together with chest pain, breathlessness, cold sweating or nausea;
- the shoulder is hot, swollen and red and you have fever or shivers;
- after a fall, an accident or a violent tear you can no longer move the arm, or the shoulder looks deformed.
Book a medical visit before starting the exercises if:
- the stiffness appeared after an epileptic seizure, an electric shock or a major trauma: in these situations an unrecognised posterior dislocation imitates a frozen shoulder perfectly, and continuing to "free" it with stretching only wastes precious time;
- you have had cancer in the past, or you notice unexplained weight loss, night sweats or general malaise;
- you are a smoker or a former smoker and the shoulder is accompanied by persistent cough, tingling in the hand, or a drooping eyelid: the problem can originate at the apex of the lung and not in the joint;
- you have lost real strength or sensation in a part of the arm or hand;
- the pain came on suddenly, very severe, and was followed within days by marked weakness: this is the picture of a nerve injury, not of a retracted capsule;
- the arm is cold, pale or changes colour.
None of these points is a reason to give up movement. They are reasons to get checked first, and not to blame the capsule for a problem that lies elsewhere.
What frozen shoulder is, and why the name is debated
In frozen shoulder the capsule wrapped around the joint becomes inflamed, thickens and retracts. The result is a shoulder that moves little and in every direction, even when someone else moves it: this is the sign that sets it apart from tendinitis or a painful rotator cuff, where passive movement usually remains better than active movement. External rotation, the movement of bringing the forearm outwards while keeping the elbow at your side, is almost always the most restricted, and it is often the first to signal that something is wrong.
The term "adhesive capsulitis", probably the one on your report, is now considered inappropriate by several international study groups: a review published in EFORT Open Reviews in 2020 went through 1,559 articles, six position documents and all 24 systematic reviews from the previous five years, and recommends using "frozen shoulder" and adopting a definition that recognises how severe the pain can be. The ISAKOS upper-extremity working group is clear in judging "adhesive capsulitis" an inappropriate term, although 63% of the reviews examined still preferred it. The same review notes that the literature is divided even on classification: of the 24 systematic reviews examined, 9 considered only the primary form and 9 distinguished primary and secondary forms.
Secondary forms appear after a trauma, prolonged immobilisation, a surgical operation, or in the presence of general diseases. Diabetes is the most solid risk factor: a meta-analysis published in BMJ Open in 2023 estimated, across 6 case-control studies for a total of 5,388 people, an odds ratio of 3.69 compared with people without diabetes (95% confidence interval: from 2.99 to 4.56). Two cohort studies found a more modest association in the same direction, with hazard ratios of 1.32 (from 1.22 to 1.42) and 1.67 (from 1.46 to 1.91). The authors are explicit on one point: the risk of bias was judged high in seven of the eight studies and moderate in the eighth. That 3.69 should therefore be taken as an order of magnitude, not as a law, and it remains an association, not proof that diabetes directly causes the capsule to retract.
The "three phases": a comfortable story the evidence does not confirm
The idea that frozen shoulder goes through a painful phase, a stiff phase and a thawing phase until it returns like new dates back to the 1940s. It became the standard way of describing this condition, including in the frozen shoulder rehabilitation taught in courses.
In 2017 a systematic review published in Physiotherapy set out to verify it. The authors looked for studies with untreated groups followed for at least six months: from 508 citations, seven remained. The verdict is clear on two points. First: without treatment, movement improved partly, not completely, in follow-ups of 1 to 4 years, a result the authors themselves classify as low-quality evidence, so indicative and not definitive. Second, and more surprising: no study supported progression through the phases, and indeed three randomised trials with longitudinal data, moderate-quality evidence, showed that most of the improvement happens early, not late. Exactly the opposite of what the phase model leads you to expect.
Despite this, the 2020 EFORT review counted that 11 of the 12 systematic reviews that discussed phases still described three. The authors go as far as calling this repetition "confounding and potentially harmful", and the reason is practical: if you expect the thawing phase to arrive on its own, you wait. If you know that most of the recovery tends to concentrate in the first months, you act straight away.
What changes for you, concretely. The three phases remain useful as a description of what you feel: there really is a period in which pain dominates and one in which stiffness dominates, and exercises should be calibrated accordingly. They are not reliable as a calendar promise, and they do not justify waiting.
What the research really says about exercises
Here the honest answer has two sides, and it is worth keeping both in view.
Exercises work. A systematic review with meta-analysis published in Archives of Physical Medicine and Rehabilitation in 2022 included 33 studies (19 in the meta-analysis) and found that joint movement, function and pain improve both with exercises alone and with programmes that include them. There is preliminary evidence that supervised exercises give more benefit than those done alone at home, for mobility and function, and that adding stretches to a multimodal programme may increase movement. Between one programme and another, however, the differences in movement were minimal or nil, and for function the evidence was uncertain: the choice of protocol matters much less than people think, and this is good news, because it removes the anxiety of having chosen the wrong list of exercises. Two findings need careful reading because they dispel widespread habits: adding instrument-based physical therapies brought no further benefit, and the authors conclude that no conclusion can be drawn either on long-term results or on the best dosage of exercise.
But the wrong dose can cost you. In 2004, in the Journal of Shoulder and Elbow Surgery, a prospective study of 77 people compared two approaches: intensive rehabilitation with passive stretching and manual mobilisations, or supportive therapy and exercises within the limit of pain. At 24 months, 89% of the "within pain" group had a normal or nearly normal, painless shoulder (Constant score of 80 or above), and 64% had already reached this within 12 months. In the intensive group, the same threshold had been reached by only 63% at 24 months. It is a single study, not recent, and not described as randomised: the two groups were comparable for age, sex, time since onset and initial severity, but a comparison built this way carries less weight than a true randomised trial. It does not settle the question. But it is the direct comparison we have, and it points in the opposite direction to the instinct to "free it by force".
And exercises are not the only lever. The 2014 Cochrane review on manual therapy and exercise (32 studies, 1,836 participants) found that six weeks of manual therapy and exercise probably gives less improvement at seven weeks than a glucocorticoid injection: pain improved by 58 points on a 100-point scale with the injection against 32 points with manual therapy and exercise (a single randomised study, 107 participants), and 77% of those who received the injection (40 out of 52) reported treatment success, against 46% of those who did manual therapy and exercise (26 out of 56; a single randomised study, 108 participants). The number of people reporting adverse effects, by contrast, did not differ between the two groups. It must be said in full, however: the differences in pain and function at six and twelve months were not clinically important. The advantage of the injection is mainly a matter of how soon you feel better.
The same picture emerges from a systematic review with meta-analysis published in JAMA Network Open in 2020: 65 studies and 4,097 participants in the overall review, of which 34 studies and 2,402 participants went into the pairwise comparisons and 39 studies and 2,736 participants into the network meta-analyses. In the pairwise comparisons, despite numerous statistically significant results, intra-articular corticosteroid injection was the only treatment with superiority both statistical and clinical in the short term, the first twelve weeks as the authors define it, for pain and function, including in direct comparison with physiotherapy (a mean difference of 1.1 points on a 0 to 10 visual analogue scale for pain). The authors add two things that bear directly on this guide: early use, within the first year of the condition, is associated with better results, and the injection should be accompanied by a home exercise programme with simple movements and stretches, which in the subgroup analyses and the network meta-analysis seems to add benefit in the medium term.
Put simply: exercises are the part you carry out yourself, every day, and they are needed. But if the pain is strong and the diagnosis is recent, it makes sense to talk to your doctor about an injection instead of gritting your teeth for months, and then building the active work on top of it.
An important point: before trying these exercises, make sure you have received a correct diagnosis. People often assume they have one problem when the real cause of the pain is another. Doing the wrong exercises can make things worse rather than better.
The principles behind the programme
Pain is the dose regulator. Not an obstacle to overcome: it is the tool you use to decide how hard to push. Working to a tension that you feel but can breathe through is different from reaching sharp pain, and the 2004 comparison suggests the difference is not just one of comfort.
Little and often. Short sessions repeated several times a day are more tolerable than one long session, and they keep the shoulder moving without stirring up inflammation.
External rotation must be defended first. It is the direction that is lost most quickly and the one needed for everyday actions: opening a door, fastening a seatbelt, bringing your hand to the phone in a coat pocket.
Heat and ice are comfort, not treatment. A warm shower beforehand can make movement more bearable, and ice afterwards can calm an irritated shoulder: if they help, use them. But think of them as aids that let you do the exercises, not as active ingredients: in the 2022 meta-analysis, adding instrument-based physical therapies to an exercise programme produced no further benefit.
Consistency matters more than intensity. The weeks in which you do nothing are not made up for by a heroic session on Saturday.
Pendulum exercises
These are the starting point, because they ask very little of the shoulder muscles: gravity does the work. Lean with your healthy hand on a table, bend slightly forward and let the painful arm hang completely relaxed. From there let it swing: forwards and backwards, then from side to side, then in small circles in both directions. The movement comes from the body shifting slightly, not from the shoulder pushing.
The most common mistake is stiffening the arm and "doing" the exercise instead of letting it happen. If you feel the shoulder contracting, the swing is too wide: reduce it. Some people add a light weight in the hand to increase traction: this is a reasonable variation only if the shoulder tolerates it without increased pain, and it is worth having it assessed rather than deciding on your own. If you want to go into the technique in detail, we covered it in the article on the Codman pendulum.
Passive stretches
Here the painful shoulder is moved by someone else: the other hand, a stick, a wall. The shoulder muscles remain spectators, and that is what allows you to gain movement without triggering the painful reaction.
Posterior capsule. Bring the painful arm across the front of your chest, at shoulder height, and with the other hand gently pull the elbow towards the opposite shoulder. Hold the position while breathing, without bouncing.
Stick elevation, lying down. Lying on your back, hold a broomstick with both hands: the healthy arm pushes and guides the painful arm upwards, above the head. Lying down, the trunk cannot cheat by arching, and it is easier to see how much movement you are really gaining.
External rotation with a stick. Standing or seated, elbows bent and held at your sides, stick held with both hands. The healthy arm pushes the stick sideways, bringing the painful forearm outwards. The elbow must not leave your side: if it moves away, the gain is fake.
Doorframe stretch. Standing in a doorway, rest your forearm on the frame with the elbow bent at a right angle. Slowly rotate your trunk in the direction opposite to the arm. Small movements are enough, and it is easy to overdo it without noticing.
Active-assisted exercises
These are the intermediate step: the shoulder takes part in the movement, but with help.
Finger walk up the wall, facing it. Standing in front of a wall, place your fingers on it and "walk" them upwards, moving your body closer as you climb. Mark the highest point reached with a strip of tape: it is the simplest way to see your progress, and in the weeks when nothing seems to change it is also the most useful way to keep going.
Finger walk up the wall, sideways. Same logic, but with your side to the wall, to work on the lateral opening of the arm. Be careful not to bend your trunk the other way to gain centimetres: the compensation is almost automatic.
Towel behind the back. Hold a towel diagonally behind your back, the healthy hand high behind the neck and the painful one low behind the back. The upper hand pulls gently, bringing the lower one up along the spine. This is the internal rotation movement, the one you need to fasten a bra or put on a belt, and it is often the most unpleasant: proceed in very small increments.
Strengthening
Strengthening starts when mobility has returned to a usable level, not before. After months in which the arm has been used little, the muscles are deconditioned: recovering movement without recovering strength leaves a shoulder that moves but cannot cope with the demands of the day.
External rotation with a band. Elbow at your side bent at a right angle, band anchored in front of you: rotate the forearm outwards and return slowly. The controlled return phase is as important as the push.
Internal rotation with a band. Same position, band from the opposite side: bring the forearm towards your stomach.
Front and side raises. With a light weight, or even with the arm free, lift in front of you and to the side as far as the movement stays clean, without the shoulder hiking up towards the ear. It is better to stop early with good form than to climb higher while twisting everything.
Progress by increasing repetitions first and only afterwards the resistance. If the next day the shoulder is more irritated than usual, you have gone up too fast.
How to tell if you are progressing well
The best reference is not what you feel during the exercise, but how the shoulder feels afterwards. A discomfort that flares up during the movement and settles shortly after is acceptable. Pain that stays lit for the rest of the day, or that wakes you more than usual the following night, tells you the dose was excessive: the next session should be reduced, not repeated unchanged to "beat the resistance".
The second reference is measured movement. The tape on the wall, the distance your hand reaches behind your back, how far the forearm rotates with the elbow at your side: three concrete indicators, checked every one or two weeks. They are slow, and that is precisely why they work better than the feeling of the moment, which in frozen shoulder changes from one day to the next for no reason.
Exercises to avoid with a frozen shoulder (and other mistakes)
Before the list, the question as the person searching asks it: which exercises should I avoid? The honest answer is that there is no list of forbidden movements. There is a wrong way of doing them, and it is always the same: pushing into the pain.
In practice, with a frozen shoulder you avoid:
- forced stretches at the end of range, held with clenched teeth until they "pull": this is the comparison between the intensive approach and working within the pain described above, and the intensive approach loses;
- mobilisations done by someone else pulling the arm while you endure them: the same rule applies, pain is not the price of progress;
- overhead lifts with a load until the arm rises on its own: what rises is the shoulder, through compensation, and you train the compensation;
- the jerk to reach something up high, the movement with which many stiff shoulders flare up.
What follows are the remaining mistakes, which weigh as much as the choice of exercises.
Forcing. It is the most common and the one with the heaviest bill, as the comparison between the intensive approach and working within the pain shows.
Waiting for the thawing phase. The phase model is not confirmed, and the available evidence indicates that most of the improvement tends to arrive early. Waiting is not a strategy.
Doing everything for two weeks and then disappearing. Alternating between enthusiasm and giving up is the most effective way to get no results.
Treating the shoulder and ignoring the rest. If you have diabetes, blood sugar control is part of the clinical picture, not a separate chapter: it is the most solid risk factor we know of for this condition.
Giving up on using the arm in daily life. The exercises take twenty minutes; the other sixteen waking hours count too. Continuing to use the arm for the activities it tolerates is part of the treatment.
How long it really takes
This is the question everyone asks, and it is where honesty costs the most.
We do not have a reliable number to give you. The 2022 meta-analysis explicitly states it cannot draw conclusions on long-term results. Studies that followed untreated people observe them for up to four years and find a partial improvement in movement, not a complete return. The most usable data we have comes from the 2004 study: in the group treated within the limit of pain, 64% had a normal or nearly normal shoulder already within 12 months, and 89% within 24.
What these numbers really say is that this is a matter of months, not weeks, and that a share of people keep some residual limitation. Knowing this from the start avoids two opposite mistakes: giving up at the third month because "it is not working", and forcing it at the sixth because "it should be over by now".
The role of telerehabilitation
Frozen shoulder is a condition in which the decisive part is done by you, at home, every day, and it is also one where the evidence suggests that supervision matters: in the 2022 review, supervised exercises were more effective than unsupervised home exercises, for mobility and function, although as preliminary evidence. For accuracy, it should be said that in those studies the supervision was in person: no one has directly compared remote supervision with supervision in a clinic, so the reasoning is a plausible extension, not a measured result.
At a distance you do exactly this well: watch how you perform the movements and correct them, measure together the movement you really have, decide when it is time to add strengthening, and understand, in the off weeks, whether the shoulder is irritated because you are pushing too hard or because the programme has been stuck for too long. A telerehabilitation session with Ri-Hub costs 29,90 euro, and in the frozen shoulder rehabilitation pathway the value lies in the frequency of check-ins more than in the length of a single session.
Have you been prescribed a sling and do not know how long to wear it? We discuss it in Shoulder sling: when it is needed and how to wear it.
If the shoulder no longer moves in one particular direction, before forcing it it is worth understanding why: Stuck shoulder: what to do to free it.
Sources
- Natural history of frozen shoulder: fact or fiction? A systematic review, Physiotherapy, 2017
- 'Frozen shoulder' is ill-defined. How can it be described better? EFORT Open Reviews, 2020
- Gentle thawing of the frozen shoulder: a prospective study of supervised neglect versus intensive physical therapy in seventy-seven patients with frozen shoulder syndrome followed up for two years, Journal of Shoulder and Elbow Surgery, 2004
- Exercise Therapy Is Effective for Improvement in Range of Motion, Function, and Pain in Patients With Frozen Shoulder: A Systematic Review and Meta-analysis, Archives of Physical Medicine and Rehabilitation, 2022
- Manual therapy and exercise for adhesive capsulitis (frozen shoulder), Cochrane Database of Systematic Reviews, 2014
- Comparison of Treatments for Frozen Shoulder: A Systematic Review and Meta-analysis, JAMA Network Open, 2020
- Diabetes as a risk factor for the onset of frozen shoulder: a systematic review and meta-analysis, BMJ Open, 2023
Last source check: August 2026.
You might also be interested in:
Strengthening Shoulders at Home: 12 Effective Exercises Without Equipment
Disc Protrusion Exercises: An Effective Routine to Reduce Pain
Glute Exercises: 15 Effective Movements for Strong, Toned Glutes
Stay in touch with us
Join our dedicated Facebook group: we answer all patients, give practical advice and clear up the most common doubts.
Wondering how much it would cost to be followed? The cost depends on how many sessions are needed, not on the single rate: you will find the figures in How much physiotherapy for the shoulder costs.
Need a consultation?
Start with the questionnaire on Ri-Hub: a few questions about your case, you register, and you see who will follow you: a physiotherapist registered with the Albo, whom you write to in chat whenever you like. If you prefer to talk it through by voice, the first consultation is free, lasts 20 minutes and commits you to nothing: they arrive having already read what you wrote, and together you decide whether and how to go on.
