Exploration of the clinical course and longitudinal correlations in frozen shoulder

Key points

1. 01Pain intensity and shoulder disability improved significantly from baseline to all follow-up points, with the largest gains in the first three months.
2. 02Autonomic symptoms (dysautonomia) improved only from baseline to the 6- and 9-month follow-ups, suggesting a delayed trajectory.
3. 03Pain catastrophizing and hypervigilance showed fair longitudinal correlations with pain intensity (r = 0.301-0.397), indicating they track together over time.
4. 04Tactile sensitivity, local pressure pain thresholds, temporal summation, and conditioned pain modulation showed no significant time-related changes, suggesting impaired central pain processing is not a defining feature at the group level.
5. 05Correlations between psychological and central pain processing variables were weak to fair, and their clinical direction remains unclear.

How it was conducted

Design

Multicenter prospective longitudinal observational study (STROBE-reported)

Participants

149 adults with frozen shoulder recruited at hospital orthopaedic departments and general practitioner practices in Belgium and Spain

Follow-up

9 months with assessments at baseline, 3, 6, and 9 months

Primary outcomes

Pain intensity (VAS), shoulder pain and disability (SPADI), autonomic symptoms (COMPASS-31), pain catastrophizing (PCS), pain hypervigilance (PVAQ)

Quantitative sensory testing

Tactile sensitivity (Von Frey), pressure pain thresholds (PPT), temporal summation (TS), and conditioned pain modulation (CPM) at affected shoulder and remote sites

Analysis

Negative binomial regression and linear mixed models; longitudinal correlations via within-subjects multiple regression with Benjamini-Hochberg multiple-testing correction

What they found

• 149 participants enrolled at baseline; 121 completed 3-month, 97 completed 6-month, and 88 completed 9-month follow-up (59% completion).
• Pain intensity (VAS 0-100): baseline 48.68 +/- 27.77, declining to 31.89 at 3 months, 20.71 at 6 months, and 13.72 at 9 months; mean difference from baseline to final follow-up 34.96 [95% CI 28.39;41.53].
• Shoulder pain and disability (SPADI 0-100): baseline 60.75 +/- 21.16, declining to 40.60 at 3 months, 27.53 at 6 months, 22.00 at 9 months; mean difference 38.75 [95% CI 32.14;45.36].
• Autonomic symptoms (COMPASS-31 0-100): baseline 17.40 +/- 12.46, 15.85 at 3 months, 12.55 at 6 months, 12.73 at 9 months; significant improvement only from baseline to 6 and 9 months (p<0.05).
• Pain catastrophizing (PCS 0-52): baseline 17.21 +/- 11.29, declining to 13.02 at 3 months, 9.76 at 6 months, 10.43 at 9 months; significant improvement from baseline to all other time points (p<0.05).
• Pain hypervigilance (PVAQ 0-80): baseline 28.33 +/- 12.17, declining to 26.08 at 3 months, 24.04 at 6 months, 22.89 at 9 months; significant improvement from baseline to 6 and 9 months, and from 3 months to 9 months (p<0.05).
• Fair positive longitudinal correlation between pain intensity and pain catastrophizing (r = 0.397, p<0.001) and between pain intensity and hypervigilance (r = 0.301, p<0.001).
• Fair positive longitudinal correlation between catastrophizing and hypervigilance (r = 0.354, p<0.001).
• Poor longitudinal correlations between pain intensity and allodynia at shoulder (r = 0.193, p=0.002) and quadriceps (r = 0.171, p=0.006).
• Poor negative longitudinal correlations between pain intensity and PPT at affected shoulder (r = -0.180, p=0.004).
• Weak positive longitudinal correlation between catastrophizing and autonomic symptoms (r = 0.209, p<0.001).
• No time-related changes found for tactile sensitivity at either location, local PPT, temporal summation, or conditioned pain modulation (all p>0.05).
• No significant improvement in pain intensity or shoulder disability between 6- and 9-month assessments.

Limitations

• Low protocol adherence with only 59% of the sample completing all assessments, although additional participants were recruited to compensate.
• Observational design with no control group; various treatments were permitted and recorded but not controlled.
• Temporal summation reliability may be limited because the pressure stimulus is applied manually rather than with a computerized device.
• The COMPASS-31 reflects subjective autonomic symptoms rather than objective autonomic nerve activity, limiting interpretation of the dysautonomia findings.

Why it matters

For patients

People with frozen shoulder can expect the greatest reductions in pain and disability within the first three months, after which improvement slows, so early treatment initiation is important.

For clinicians

Psychological factors such as pain catastrophizing and hypervigilance track closely with pain intensity over time and may be useful treatment targets, while central pain sensitization measures are unlikely to be primary drivers at the group level.

For readers

This is the first longitudinal study to track autonomic function, central pain processing, and psychological variables together across nine months in frozen shoulder, providing a biopsychosocial picture of the condition's clinical course.