Bilateral magnetic resonance imaging findings in individuals with unilateral shoulder pain

Key points

1. 01123 people with pain in only one shoulder had both shoulders scanned, giving 246 shoulders to compare.
2. 02Abnormal MRI findings were highly prevalent on the painless side too, so most findings did not match where the pain was.
3. 03Only full-thickness supraspinatus tears and glenohumeral osteoarthritis were significantly more common on the painful shoulder.
4. 04Two expert readers, a shoulder surgeon and a radiologist, often disagreed, with agreement ranging from slight to moderate.
5. 05The authors caution that MRI findings alone should not drive shoulder treatment decisions.

How it was conducted

Design

Prospective cross-sectional epidemiology study (Level III)

Participants

123 community adults (246 shoulders) with chronic intermittent unilateral shoulder pain, average age 39.4 years, 66 men

Comparison

Symptomatic shoulder vs. the same person's asymptomatic shoulder

Imaging

1.5 T MRI (T1, T2, proton density, T2 gradient echo) in coronal, sagittal, and axial planes, no contrast

Readers

A fellowship-trained orthopedic shoulder surgeon and a musculoskeletal radiologist, reading anonymized randomized images independently

Primary outcome

Frequency of each MRI abnormality compared between symptomatic and asymptomatic shoulders

What they found

• Rotator cuff tendinopathy was found in 92.7% of symptomatic and 88.6% of asymptomatic shoulders by the radiologist (P = .38), and 74.8% vs. 73.0% by the surgeon (P = .85).
• Full-thickness tears (any tendon) were higher on the symptomatic side per the surgeon: 20.5% vs. 8.1% (chi-square = 6.66, P = .01); per the radiologist 5.7% vs. 0.8% was not significant (P = .06).
• Full-thickness supraspinatus tears per the surgeon were 16.3% symptomatic vs. 3.3% asymptomatic (chi-square = 10.38, P < .01).
• Glenohumeral osteoarthritis per the surgeon was 10.7% symptomatic vs. 3.3% asymptomatic (chi-square = 4.11, P = .04).
• Partial-thickness infraspinatus tears per the radiologist were 10.6% vs. 1.6% (chi-square = 7.10, P < .01).
• Supraspinatus retraction per the surgeon was 13.0% vs. 2.4% (chi-square = 8.21, P < .01).
• Inter-rater agreement between the radiologist and surgeon (Cohen kappa) ranged from 0.00 to 0.51, with observed agreement of 44.71% to 98.14%.

Limitations

• MRI captures a single moment and cannot tell whether asymptomatic findings will later become painful.
• No intra-observer reliability analysis was performed.
• The study may have been underpowered to detect small side-to-side differences (between 5% and 10%) in some findings.
• Cross-sectional design cannot establish that any finding caused the pain.

Why it matters

For patients

An MRI showing a shoulder abnormality does not prove that finding is causing your pain, because the same changes are often present in your pain-free shoulder.

For clinicians

Interpret most shoulder MRI findings cautiously, weight findings like full-thickness tears and osteoarthritis more, and base intervention on functional loss rather than imaging alone.

For readers

This adds a sizable community-based dataset to evidence that shoulder MRI abnormalities are frequently incidental rather than the cause of symptoms.