A positive correlation between steroid injections and cuff tendon tears: a cohort study using a clinical database

Key points

1. 01Rotator cuff tear incidence was 9.8% in the steroid-injection group vs 1.2% in controls (p < 0.001)
2. 02Adjusted hazard ratio for steroid injection was 7.44 (95% CI 3.45-16.00; p < 0.001) after controlling for sex, age, and comorbidities
3. 03Chronic liver disease raised risk 3.25-fold (CI 1.38-7.63; p = 0.007) and smoking raised risk 2.40-fold (CI 1.02-5.66; p = 0.046)
4. 04Mean time from injection to rotator cuff tear diagnosis was 39 +/- 22 months
5. 05Number of injections (one vs two vs three or more) did not significantly change the outcome

How it was conducted

Design

Retrospective cohort study using a single-centre clinical database (Kaohsiung Veterans General Hospital, Taiwan), January 2013 to December 2019

Participants

1025 patients with shoulder diseases; 205 in the steroid-injection case group and 820 matched controls (1:4 ratio, balanced by age, sex, and index date)

Intervention

Triamcinolone steroid injection (intra-articular, tendon, or trigger-point) identified by procedure and drug codes

Comparator

Patients with shoulder disease who never received steroid injections

Primary outcome

Occurrence of rotator cuff tendon tear or cuff tendon repair at any point during the study period, identified by ICD-9-CM/ICD-10-CM codes

Analysis

Cox proportional hazards regression (univariate then stepwise backward multivariate), adjusted for sex, age, and comorbidities; Kaplan-Meier survival curves

What they found

• Overall incidence of rotator cuff tear: 2.9% (30/1025)
• Incidence in steroid-injection group: 9.8% (20/205); incidence in control group: 1.2% (10/820); p < 0.001
• Adjusted HR for steroid injection: 7.44 (95% CI 3.45-16.00; p < 0.001)
• Adjusted HR for chronic liver disease: 3.25 (95% CI 1.38-7.63; p = 0.007)
• Adjusted HR for smoking: 2.40 (95% CI 1.02-5.66; p = 0.046)
• Univariate HR for steroid injection: 8.50 (95% CI 3.98-18.16; p < 0.001)
• Mean time from enrollment/injection to rotator cuff tear onset: 39 +/- 22 months
• Mean age of patients who developed rotator cuff tear: 62.2 years vs 59.4 years overall
• 76% of injection-group patients received only one injection; 16% received two; 8% received three or more

Limitations

• Single-centre retrospective database study limits generalisability to other populations and settings
• No access to injection dose, technique, exact steroid type beyond triamcinolone, or severity of underlying shoulder disease
• Asymptomatic rotator cuff tears before enrollment could not be fully excluded, potentially inflating tear incidence
• Relatively small number of outcome events (30 total tears) limits statistical power and precision of subgroup analyses

Why it matters

For patients

Patients receiving shoulder steroid injections should be informed of a potentially elevated risk of rotator cuff tear, especially if they also smoke or have liver disease, so they can make an informed choice about treatment.

For clinicians

Clinicians should carefully weigh the short-term pain relief benefit of shoulder steroid injections against the observed 7.44-fold increased hazard of rotator cuff tear, and consider additional caution in patients who smoke or have chronic liver disease.

For readers

This retrospective cohort adds human clinical evidence to animal and in-vitro data suggesting steroid injections are harmful to tendon integrity, but the single-centre design and small event count mean large prospective studies are still needed.