Frozen Shoulder and Diabetes: Risk, Weight and GLP-1 Evidence

Frozen shoulder, known medically as adhesive capsulitis, causes a painful and progressive loss of shoulder movement ^[1]. Diabetes is its best-documented risk factor: a 2026 meta-analysis put the odds at 3.69 times those of people without diabetes ^[2]. Higher blood sugar short of diabetes also carries a small added risk ^[5]. Body weight matters less than blood sugar does ^[7][8]. For GLP-1 drugs, the only evidence so far is two insurance-database studies, and both found more frozen shoulder diagnoses among users, not fewer ^[9][10].

The honest summary

• Diabetes is a strong, consistent risk factor. Pooled studies put the odds of frozen shoulder at roughly 3.7 to 5 times higher with diabetes ^[1][2], and about 30% of people with frozen shoulder have diabetes ^[1].
• Type 1 diabetes of long duration carries the highest burden. In one Norwegian study, 76% of people who had lived with type 1 diabetes for at least 45 years had had frozen shoulder at some point ^[4].
• Prediabetes adds a smaller step of risk. In 3.47 million Korean adults, the hazard was 8% higher with prediabetes and 47% higher with treated type 2 diabetes ^[5].
• Blood sugar looks causal; body weight looks less so. Genetic studies link higher long-term HbA1c to frozen shoulder, while genetically higher BMI showed no link ^[7]. People with obesity do get frozen shoulder more often in ordinary records ^[8].
• GLP-1 data point the wrong way, but it is weak data. Two retrospective database studies linked GLP-1 use to more frozen shoulder ^[9][10]. No randomized trial has tested it.
• A steroid injection early on has the best short-term evidence for treatment, and it temporarily raises blood sugar in people with diabetes ^[11][12].

How strongly diabetes is linked to frozen shoulder

The link has been reported for decades, and the pooled numbers are now large. A 2016 meta-analysis of 18 studies found people with diabetes were 5 times more likely than controls to have frozen shoulder (95% CI 3.2 to 7.7) ^[1]. It estimated that 13.4% of people with diabetes have frozen shoulder, and that 30% of people with frozen shoulder have diabetes. Insulin users and people on other diabetes treatments had similar rates. A 2026 meta-analysis reached a similar answer: 3.69 times the odds (95% CI 2.99 to 4.56) ^[2]. It also listed other risk factors seen across studies, including poor glycemic control, obesity, high cholesterol, high blood pressure, thyroid problems, age 40 to 65 and female sex. The authors were careful to note that unmeasured confounding could be shaping the result.

A UK primary care cohort helps with the cause-and-effect question. Researchers followed 43,977 people newly diagnosed with type 2 diabetes and 43,977 people without it ^[3]. Type 2 diabetes was linked to more than four times the hazard of developing frozen shoulder (HR 4.38, 95% CI 3.70 to 5.21). The team also tested whether the extra risk ran through other metabolic problems that developed later. It did not. Their conclusion was that type 2 diabetes may be a direct cause.

Type 1 diabetes

The most striking numbers come from people who have had type 1 diabetes for a very long time. The Dialong study examined 102 Norwegians who had lived with type 1 diabetes for at least 45 years and 73 people without diabetes ^[4]. On the day of the exam, 59% of the diabetes group had frozen shoulder, compared with none of the comparison group. Over their lifetimes, 76% of the diabetes group had had it, versus 14% of the comparison group. Shoulder disability scores also rose with higher long-term HbA1c.

Prediabetes and blood sugar below the diabetes line

The risk does not switch on only at a diabetes diagnosis. A Korean national cohort followed 3,471,745 adults from a 2009 to 2010 health checkup through 2018 ^[5]. Frozen shoulder occurred at 9.45 cases per 1,000 person-years in people with normal blood sugar and 11.91 in people with prediabetes. Compared with the normal group, the hazard was 8% higher with prediabetes (HR 1.084), 31% higher with newly found type 2 diabetes (HR 1.312) and 47% higher with treated type 2 diabetes (HR 1.473). The step from normal to prediabetes is small for any one person, but the trend was steady.

The link also runs the other way. In UK records, 31,226 adults newly diagnosed with frozen shoulder were matched with 31,226 without it, and nobody had known diabetes at the start ^[6]. Over follow-up of up to 15.8 years, 5% of the frozen shoulder group were diagnosed with type 2 diabetes, compared with 0.28% of the comparison group (HR 19.4). Adjusting for how often people saw a doctor and for other risk factors barely changed the result. One plausible reading is that many of these people already had high blood sugar that had not yet been diagnosed. The authors suggested future research test diabetes screening for people who develop frozen shoulder. For other body signs that can point toward insulin resistance, see skin tags, acanthosis nigricans and insulin resistance.

Does glycemic control matter?

The strongest evidence that blood sugar itself matters comes from Mendelian randomization, a method that uses inherited gene variants as a natural experiment. In 379,708 UK Biobank participants, each 10 mmol/mol rise in genetically predicted HbA1c was linked to 50% higher odds of frozen shoulder (OR 1.50, 95% CI 1.20 to 1.88) ^[7]. That is roughly a 0.9 percentage-point rise on the HbA1c scale most US labs report. The same study found the pattern across other hand and arm conditions, including carpal tunnel syndrome and trigger finger, and concluded these conditions should be considered diabetes complications. The weight side of that picture for the wrist is covered in carpal tunnel syndrome and body weight.

Because these genetic differences are present from birth, the result points to years of higher blood sugar, not a single bad reading. The Dialong study fits that pattern: people with higher long-term HbA1c had more shoulder disability ^[4]. Whether lowering HbA1c once frozen shoulder has started changes its course has not been tested in a trial.

Does weight or BMI matter?

In everyday medical records, yes. A German study followed 522,830 adults from 677 general practices for up to 10 years ^[8]. Frozen shoulder developed in 0.65% of people at a normal weight, 1.05% of those who were overweight and 1.25% of those with obesity. After adjustment, overweight carried a 52% higher hazard (HR 1.52) and obesity a 77% higher hazard (HR 1.77). Each extra BMI point added about 3%.

The genetic evidence tells a different story. In the UK Biobank analysis, a genetically predicted BMI 5 points higher showed no association with frozen shoulder, even though the same method linked higher BMI to carpal tunnel syndrome ^[7]. Taken together, the likeliest explanation is that weight is linked to frozen shoulder largely because it travels with higher blood sugar, not because extra weight strains the shoulder joint. The German authors also said more work is needed on what sits between BMI and frozen shoulder ^[8].

GLP-1 drugs and frozen shoulder

Given everything above, you might expect drugs that lower blood sugar and weight to protect the shoulder. The two studies that have looked found the opposite. Both are retrospective analyses of large US health-record databases.

A Johns Hopkins team used the PearlDiver database from 2010 to 2022 and randomly sampled 100,000 people with type 2 diabetes taking GLP-1 drugs and 100,000 not taking them ^[9]. After matching for age, sex, other illnesses, obesity, smoking, thyroid disease, metformin and insulin use and complicated diabetes, GLP-1 users had higher odds of developing frozen shoulder (OR 1.28). Among 253,717 people with type 2 diabetes and frozen shoulder, GLP-1 users also had higher odds of needing a procedure (OR 1.18). That was driven by manipulation under anesthesia (OR 1.20), with no difference in surgical capsular release.

A University of North Carolina team used the TriNetX database and followed adults prescribed GLP-1 drugs between 2017 and 2019 for five years, matched against nonusers ^[10]. GLP-1 use was linked to more frozen shoulder, shoulder osteoarthritis and rotator cuff tears in every group studied. The largest increase was for frozen shoulder in people with obesity but without diabetes, at about 2.5 times the hazard (HR 2.465). Shoulder fractures were less common among GLP-1 users with both obesity and type 2 diabetes.

These findings deserve attention, and they also have clear limits. Both studies rely on billing codes, not shoulder examinations. Neither assigned the drug at random, so people who were prescribed a GLP-1 may have differed from nonusers in ways matching could not capture, such as how long they had had diabetes or how often they saw a doctor. Neither study identifies a mechanism. Both groups of authors called for prospective research, and the second team wrote that GLP-1 drugs may have adverse effects in the shoulder ^[10]. No randomized trial on GLP-1 drugs and frozen shoulder has been published, so the honest status is an unconfirmed signal. For the broader picture of joint and muscle complaints on these drugs, see Ozempic joint and muscle pain.

Treatment: what the evidence supports

The largest comparison of treatments pooled 65 randomized studies with 4,097 participants ^[11]. Only a corticosteroid injection into the shoulder joint beat other options by a margin large enough to matter clinically in the short term (up to 12 weeks). Against placebo or no treatment, it lowered pain by about 1 point on a visual analog pain scale and improved function. Adding a home exercise program of simple stretches gave extra benefit over the following months. The authors concluded that an early injection, in frozen shoulder present for less than a year, paired with home exercise gives the best chance of recovery.

For people with diabetes, there is one practical catch: blood sugar. In a study of 40 patients with diabetes and 38 without, fasting glucose was significantly higher one and seven days after a shoulder steroid injection ^[12]. The rise was largest in the diabetes group, and levels returned to baseline by day 15. The authors judged the injection safe in well-controlled diabetes and recommended frequent glucose monitoring for two weeks afterward.

Practical guidance

• Have unexplained shoulder stiffness looked at early. The best-supported treatment works best in the first year.
• If you develop frozen shoulder without known diabetes, ask about an HbA1c test. People newly diagnosed with frozen shoulder are far more likely to be diagnosed with type 2 diabetes later.
• If you have diabetes, plan for a glucose bump after a steroid injection. Check more often for about two weeks.
• Keep long-term blood sugar in range. Years of higher HbA1c are the most consistent link in the evidence.
• Do the home exercises. Stretches added benefit on top of an injection.
• On a GLP-1 drug, report shoulder stiffness rather than stopping the medication yourself.