Beyond the Capsule: Frozen Shoulder and Stiff Shoulder as a Systemic, Immunometabolic and Neurobiological Condition

Frozen shoulder is not just a capsule problem

For decades, frozen shoulder (FS) has been taught as an idiopathic, self-limiting capsular disorder: it freezes, it stays frozen, and eventually it thaws. The "self-limiting" story does not hold up well. There is little evidence of complete recovery without supervised treatment, most improvement happens in the first three to four months, and 27–50% of people still report pain or stiffness up to ten years after onset (Wong et al., 2017; Brindisino et al., 2026). Treating the joint alone is often not enough.

A 2026 collaborative viewpoint led by Brindisino, Struyf, Mertens, Navarro-Ledesma and colleagues frames it differently. FS is best understood as the musculoskeletal expression of systemic dysregulation, with an inflammatory phase that shifts towards fibrosis (Brindisino et al., 2026). Metabolism, immunity, hormones, the nervous system and mood all play a part. This post briefly covers the classic clinical picture, then looks at what sits beyond the capsule, and finishes with what it means for everyday physiotherapy practice.

First, a word on the name

More than 15 different terms have been used for this condition. Following the ISAKOS Upper Extremity Committee, "shoulder stiffness" (stiff shoulder) is the umbrella term for any loss of shoulder motion, including post-traumatic and post-surgical stiffness, while "frozen shoulder" is reserved for the specific condition with no other identifiable cause (Itoi et al., 2016). "Adhesive capsulitis" should be abandoned: there are no true adhesions between the capsule and the humeral head, and the older label can encourage inappropriate treatment.

The old split into "primary" (idiopathic) and "secondary" FS is also being questioned. Labels such as "secondary", "related" or "associated" imply that conditions like diabetes cause FS, when often only a co-existing or temporal relationship has been shown, and FS can even precede the diagnosis of diabetes (Brindisino, Turgut & Struyf, 2024). Systemic and metabolic status is therefore best seen as a predisposing factor linked to low-grade inflammation, not as a separate category. The distinction is still clinically useful, as long as it is read with this caveat.

See full article for the comparison of primary and secondary frozen shoulder.

FS affects 2–5% of the general population, mostly between 40 and 65 years of age. Women account for 50–70% of cases, the non-dominant side is more often involved, about 14% of cases are bilateral, and 6–34% develop FS in the other shoulder within a few years. More than 35% of people with FS have three or more comorbidities (Brindisino et al., 2026).

The classic picture still matters

The systemic view does not replace the classic clinical features; they remain the foundation of diagnosis and must not be overlooked. There is no single gold-standard test, so FS is diagnosed by pattern recognition (Brindisino et al., 2026).

See full article for the classic clinical features of frozen shoulder.

Stage by irritability, not only by phase

The three-phase model (painful "freezing", stiff "frozen", resolving "thawing") is still useful for education. For choosing treatment, though, models based on the dominant symptom ("more pain than stiffness" versus "more stiffness than pain") and especially on tissue irritability are more useful (Kelley et al., 2013; Brindisino et al., 2026). High irritability means pain ≥7/10, frequent night or rest pain, pain through the whole range and an active range smaller than passive. Low irritability means pain ≤3/10, no night pain and pain only at end range. Treatment should be gentler and more pain-modulating when irritability is high, and more active and loading-based when it is low.

Person-level factors matter too. Advanced glycation end-products (AGEs) accumulate with age and inactivity and stiffen collagen, which partly explains the age profile. Sedentary work and reduced use of the upper limb are linked to FS and may explain why the non-dominant arm is more often affected. Coping style also counts: helplessness, kinesiophobia, catastrophizing and avoidance are associated with more pain and disability.

Injections, hydrodilatation and manipulation

Intra-articular corticosteroid injection has the strongest evidence for pain relief, especially early in the pain-predominant phase (Berner et al., 2024; Challoumas et al., 2020). Multi-site injection and doses of about 40 mg of triamcinolone or methylprednisolone appear more effective than lower doses, with low risk of glycaemic or joint complications (Liang et al., 2024), although transient rises in blood glucose should be discussed and monitored in people with diabetes. Many clinicians prefer the posterior approach partly because the patient does not see the needle, which is a reasonable argument but has not been tested as an outcome. The trial evidence on approach is mixed: one RCT of 60 patients favoured the anterior approach for pain and external rotation over 12 weeks (Deng et al., 2023), while another found no difference.

An intra-articular injection of 3 ml of 1% lidocaine given 10–20 minutes before physiotherapy, used only when pain during capsular stretching reached 7/10 or more, gave better results than physiotherapy alone (Hsu et al., 2015). The aim is to make mobilisation tolerable, not to treat the disease, and it works best paired with active treatment. A suprascapular nerve block, targeting the nerve that supplies about 70% of the capsule, is an alternative.

The evidence on hydrodilatation is conflicting. A 2023 meta-analysis found at least transient, greater improvement in disability and passive external rotation than corticosteroid injection alone, with unclear clinical relevance (Poku et al., 2023), while an RCT of 108 patients in the frozen phase found steroid injection alone did better at 6 weeks and 3 months (Swaroop et al., 2023). Hydrodilatation is an option for selected stiffness-predominant cases, not a proven substitute for corticosteroid injection.

Manipulation under anaesthesia (MUA) is not recommended. It gains range by forcibly tearing the capsule: in a prospective arthroscopic study of 30 patients, every patient had haemarthrosis afterwards and 40% had new iatrogenic lesions such as SLAP lesions, labral detachments and partial subscapularis tears (Loew et al., 2005). In the UK FROST trial, MUA was no more effective at one year than early structured physiotherapy plus an intra-articular steroid injection, the least risky option (Rangan et al., 2020). With an equally effective and safer alternative available, MUA should not be offered routinely.

Frozen shoulder as an immunometabolic disorder

From inflammation to fibrosis

Capsular biopsies show an early inflammatory phase with subsynovial hypervascularisation, synovial hyperplasia and new vessel growth, mainly in the rotator interval. Alarmins such as HMGB1, IL-33 and S100 proteins activate fibroblasts, increase TNF-α, IL-1β, IL-6 and COX-2, and are linked to higher pain. Nerve growth factor and new nerve ingrowth (neo-innervation) help explain the disproportionate night pain (Brindisino et al., 2026).

Transforming growth factor-β1 (TGF-β1) then drives fibroblasts to become contractile myofibroblasts, concentrated in the rotator interval, CHL and axillary fold. This is the biological basis of anterior capsular stiffness and external rotation loss. Myofibroblasts are also mechanosensitive, so aggressive stretching during high irritability may feed the fibrotic loop. Single-cell studies show pro-fibrotic and pro-resolving cells co-existing without reaching resolution, alongside oxidative imbalance (Brindisino et al., 2026).

Glucose, AGEs and diabetes

Diabetes is the strongest known risk factor. FS prevalence rises to up to 40% in people with diabetes, and FS in diabetes tends to be more severe and longer lasting (Brindisino et al., 2026). AGEs form non-enzymatic cross-links in collagen, making it stiffer and resistant to breakdown, and bind RAGE receptors that further amplify TGF-β1 and inflammation, creating a vicious cycle.

A 2025 meta-analysis of more than 7,000 people found glycated haemoglobin (HbA1c) to be the most strongly associated metabolic biomarker, followed by total cholesterol, while triglycerides and thyroid function were less consistent (Hamed-Hamed et al., 2025). Liver enzymes and TSH have been linked to pain and disability, and leptin and insulin resistance to higher pain, greater disability and reduced flexion (Hamed Hamed et al., 2024; Pérez-Montilla et al., 2025). Leptin resistance may block the shift of macrophages to pro-resolving phenotypes and sustain JAK-STAT and TGF-β1 signalling (Navarro-Ledesma, 2025a).

The genetics point the same way. A genome-wide association study of 10,104 UK Biobank cases identified five loci for FS, the strongest near WNT7B, which is also linked to Dupuytren's disease. Type 1 diabetes (OR about 12) and type 2 diabetes (OR about 3) were associated with FS, and Mendelian randomisation provided evidence that type 1 diabetes is causal, with weaker evidence for type 2. Obesity was not causally associated once diabetes was accounted for, suggesting that diabetes acts through long-term glycaemic exposure rather than mechanical load (Green et al., 2021).

Estrogen, thyroid and the endothelium

The peak of FS in women around perimenopause and menopause points to the estrogen axis. Estrogen deficiency or resistance weakens anti-inflammatory, antifibrotic and antioxidant defences and disturbs lipid handling, favouring fibroblast activation; hypothyroidism adds to fibrosis and pain sensitisation (Navarro-Ledesma, 2025b). Endothelial dysfunction, driven by high-glycaemic diets, AGEs and oxidative stress, impairs microvascular health in the capsule, and chronic capsular hypoxia, worsened by sedentary behaviour and underuse of the non-dominant arm, may favour AGE accumulation further (Brindisino et al., 2026).

Stress and the autonomic nervous system

The autonomic, endocrine and immune systems act as one regulatory macro-system. Chronic psychosocial stress can sustain HPA-axis activation and sympathetic dominance, raise pro-inflammatory cytokines and promote low-grade inflammation (Slavich & Irwin, 2014). People with FS report more autonomic symptoms than controls (Mertens et al., 2022a), although the stress pathway in FS remains largely inferred.

Night pain and the circadian loop

Night pain is one of the most typical and disabling features of FS, and circadian biology helps explain it. Cortisol, the body's main endogenous anti-inflammatory hormone, falls to its lowest level around midnight; during this low the brake on the immune system is released, and IL-6 and TNF-α rise, so symptoms peak in the late night and early morning, a pattern well described in rheumatoid arthritis (Straub & Cutolo, 2007). In an inflamed, newly innervated capsule, the same rhythm is likely to amplify nociception at night.

This creates a negative loop. Night pain fragments sleep, poor sleep lowers pain thresholds and weakens pain inhibition the next day (Finan et al., 2013), and short or disturbed sleep is associated with higher systemic inflammation (Irwin et al., 2016). More inflammation then produces more night pain, and circadian disruption from late light, late caffeine or shift work also worsens insulin resistance. The mechanism has not been measured directly in FS, but clinically it means that protecting sleep is a treatment target in its own right, not an afterthought.

The gut microbiota connection

The brain, the gut and its microbes form one bidirectional system. Microbes signal to the brain through neural (vagal), endocrine and immune routes, and the brain shapes the microbiota through the autonomic nervous system (Dong & Mayer, 2024). Microbial products such as short-chain fatty acids (SCFAs) can sensitise sensory neurons, influence microglia and modulate cytokines, which links dysbiosis to pain processing (Pak et al., 2024). In FS, the most plausible routes are systemic: dysbiosis and a leaky gut barrier release lipopolysaccharide (LPS) and other pro-inflammatory products into the circulation, feeding the low-grade inflammation and endothelial dysfunction described above (Navarro-Ledesma, 2025b).

A two-step Mendelian randomisation study found 27 gut microbial taxa genetically associated with FS, 15 with higher risk and 12 with lower risk (Shen et al., 2026). A butyrate-producing Eubacterium taxon and Blautia hansenii were protective, the latter partly through a higher glycine-to-serine ratio (serine is needed for TGF-β-driven collagen production), while Bacteroides faecis increased risk. The authors propose that butyrate supports regulatory T cells and may block the fibroblast-to-myofibroblast transition. However, the results did not survive strict Bonferroni correction and the study was explicitly exploratory. A larger Mendelian randomisation study of 26 musculoskeletal disorders found only one method showing a protective association for adhesive capsulitis, not confirmed by the main analysis (Wang et al., 2025).

So the gut–shoulder link is biologically plausible and genetically suggestive, but there are no clinical trials of microbiome-targeted treatment in FS. Microbiome testing or specific probiotics cannot be recommended at present; probiotic results in other pain conditions are mixed and strain-specific (Pak et al., 2024). What clinicians can support are the lifestyle factors that shape the microbiota and also act on metabolism: dietary fibre diversity, fewer ultra-processed and high-glycaemic foods, regular physical activity, adequate sleep, stress regulation and avoiding unnecessary antibiotics (Pak et al., 2024; Molska et al., 2024).

BDNF, neuroplasticity and pain

Brain-derived neurotrophic factor (BDNF) supports neuronal survival, growth, synaptic plasticity and learning, mainly through the TrkB receptor. In pain, it is a double-edged neurotrophin whose effects depend on where and how it acts (Mazzitelli et al., 2025). In the spinal cord, BDNF released after nerve injury or inflammation down-regulates the chloride transporter KCC2 and weakens GABA and glycine inhibition, resulting in disinhibition and central sensitisation. In FS, sustained nociceptive input from the inflamed capsule is thought to increase central glutamate, substance P and BDNF, which may support the transition from peripheral to central sensitisation in some patients (Brindisino et al., 2026). In the brain, by contrast, lower BDNF in the hippocampus and prefrontal cortex is linked to depression, anxiety and reduced resilience. Put simply, BDNF can amplify pain locally in the spinal cord while supporting mood and adaptive learning in the brain.

BDNF also sits at the centre of a microbiota–exercise triangle. Germ-free and antibiotic-treated animals show reduced hippocampal BDNF, butyrate from fibre-fermenting bacteria raises it, and prebiotic fibre increases BDNF while lowering neuroinflammatory cytokines, although most of this evidence is from animals (Molska et al., 2024). In humans, aerobic exercise is one of the most potent lifestyle factors raising BDNF, with resistance training showing less consistent effects, and a year of aerobic training in older adults increased hippocampal volume alongside higher serum BDNF (Erickson et al., 2011).

To our knowledge, no study has measured BDNF directly in people with FS, so its role remains inferential. The rationale rests on capsular neo-innervation, the central sensitisation pathway in a subgroup of patients, and the high burden of anxiety and depression. Serum BDNF is largely platelet-derived and does not reliably reflect brain levels, so it is not a useful clinical marker (Naegelin et al., 2018). The clinical message is therefore not to "measure BDNF" but to prescribe its known enhancers: regular aerobic activity adapted to irritability, a fibre-rich diet, sleep and stress regulation. These also target metabolism, inflammation and mood.

Anxiety, depression and the nervous system

FS pain is mostly peripheral, but there is a vulnerable subgroup. Case-control work by the Antwerp group found local allodynia and primary hyperalgesia, but no secondary hyperalgesia and no group-level differences in remote sensitivity, pain facilitation or inhibition, although effect sizes were medium to large with wide individual variation (Mertens et al., 2022a). Small studies also report impaired tactile acuity and slower left–right judgements, suggesting altered body representation in some patients (Mena-Del Horno et al., 2020; Breckenridge et al., 2020). In a nine-month multicentre cohort, latent profile analysis identified two phenotypes, and the more pain-sensitive, psychologically burdened group had higher pain throughout follow-up despite improving (Mertens et al., 2023). FS pain is predominantly nociceptive and inflammatory at group level, while nociplastic features develop in a subset that needs to be identified early.

Depression and anxiety are associated with worse self-reported and functional baseline scores (Brindisino et al., 2022). Catastrophizing is linked to a wider perceived pain area, mood, catastrophizing and perceived social support are associated with both night and day pain (Brindisino et al., 2025a), and fear, catastrophizing, anxiety and depression relate to disability and quality of life (Brindisino et al., 2025b). Patients describe sadness, helplessness, loss of roles and of control, and a "hidden disability" that others do not see; many fear they will never recover, and a lack of clinician empathy fuels uncertainty and poor adherence (Crestani et al., 2025). Survey data suggest that psychological distress is more often a consequence of FS than its cause, but once present it shapes pain and recovery (Brindisino et al., 2024).

Depression and inflammation feed each other. Psychological stress raises pro-inflammatory cytokines through the HPA axis and sympathetic system, and inflammation in turn promotes low mood, fatigue and pain sensitivity (Slavich & Irwin, 2014), with low brain BDNF and dysbiosis as additional shared pathways. In the language of Article 1, a person with FS who is anxious, sleep-deprived and catastrophizing is likely to give high precision to threat-related predictions, with reduced vagal regulation and amplified pain. The interventions from Article 1, including slow breathing, graded exposure, acceptance, compassion and savouring, are therefore directly relevant to FS, alongside shoulder-specific care.

What this means in the clinic

The systemic view does not replace classic shoulder care; it widens it. Physiotherapy, built on education, exercise and manual therapy, remains first-line, individualised to the person's history, goals and tissue irritability (Brindisino et al., 2026). What changes is the assessment, which now looks beyond the shoulder to metabolic and endocrine status, lifestyle, sleep, psychological factors and the pain phenotype.

See full article for the integrated assessment.

Management then follows irritability. With high irritability (pain-predominant), the focus is on education about the natural course and good prognosis with treatment, pain-free or low-pain active-assisted movement, sleep positioning, avoiding aggressive stretching that may stimulate myofibroblasts, and early referral for an intra-articular corticosteroid injection when pain blocks function and sleep. With moderate irritability, active range and functional, task-oriented exercise are gradually increased, with manual therapy for pain modulation. With low irritability (stiffness-predominant), progressive end-range stretching and loading and strengthening of the shoulder girdle and scapula take over, and hydrodilatation or surgical options are considered only after a failed structured programme.

Promoting an active, healthy lifestyle should be part of the core treatment, not an add-on (Brindisino et al., 2026):

• Motivational interviewing: a collaborative conversation that explores the person's own reasons for change rather than prescribing. It has a small but significant effect on physical activity in people with chronic health conditions (O'Halloran et al., 2014) and is the vehicle for every change below.

• Activity: regular aerobic activity for glycaemic control, inflammation, mood and BDNF, and brief movement breaks every 30 minutes of sitting.

• Diet: an anti-inflammatory, fibre-rich, low-AGE pattern with fewer ultra-processed and high-glycaemic foods.

• Circadian rhythm and sleep: regular sleep timing, less late caffeine and night-time light, the arm supported on a pillow and a short wind-down routine to break the night pain–sleep loop.

• Stress regulation: mindfulness, cognitive strategies, yoga or Tai Chi to lower sympathetic drive and systemic inflammation.

• Communication: empathy, validation of the "hidden disability", involving family, and Socratic questioning to build self-efficacy.

The daily micro-dose framework from Article 1 fits naturally here: one-minute breathing resets, movement snacks, graded exposure to feared arm movements, acceptance and savouring can all be built into the home programme alongside shoulder exercises. People with FS often also need care beyond physiotherapy, so close links with general practice, endocrinology, nutrition, psychology and orthopaedics allow metabolic screening, better diabetes control, support for distress and a timely injection or surgical opinion.

What we still don't know

The systemic model is well supported for diabetes and glycaemic control, but much of the remaining evidence is early. Most metabolic, hormonal and psychological findings come from cross-sectional or observational studies, and only diabetes has genetic causal support (Green et al., 2021). The microbiota evidence is genetic and exploratory, based on European populations and not yet tested in clinical trials, BDNF in FS is inferred rather than measured, and no randomised trial has yet tested a combined metabolic, dietary, sleep and stress programme in FS. Future research should stratify patients into metabolic, inflammatory and psychosocial subgroups, follow pain phenotypes over time, and test multimodal lifestyle and physiotherapy programmes against usual care.

Key messages

Frozen shoulder is more than a stiff capsule. It is an inflammatory-then-fibrotic condition that sits at the meeting point of metabolism, immunity, hormones, the nervous system and mood. Clinicians should keep the classic examination, staging and evidence-based injections, while treating the whole person.

• Diagnose by pattern: two-plane ROM loss with external rotation reduced >50%, capsular pattern, anterior capsular pain and supportive imaging such as CHL thickening.

• Stage by irritability, not only by the three classic phases, and dose loading accordingly.

• Screen for diabetes, glycaemic control, lipids, thyroid function and menopausal status; HbA1c is the most consistent biomarker, and type 1 diabetes has genetic evidence for a causal role.

• Gut microbiota and BDNF are promising mechanistic links, but current evidence is genetic, animal or inferential.

• Anxiety, depression, catastrophizing and poor sleep worsen pain and disability and identify a vulnerable subgroup.

• Make lifestyle medicine, stress regulation and empathetic communication part of the core treatment.