Backgrounds: Tendon healing after rotator cuff repair is often associated with greater functional improvement. Different imaging modalities are used for pre- and post-operative evaluation of the rotator cuff. Up to now, magnetic resonance imaging (MRI) is the most common choice. Many classifications have been reported for the evaluation of tendon integrity after surgical repair but there is still a lack of data regarding their reproducibility. The purpose of the present study was to estimate intra- and inter-observer reliability of postoperative evaluation of tendon integrity.
Forty post-operative shoulder MRI exams were selected from a series of patients, who underwent rotator cuff repair 6-12 months before. Revision surgery, partial-thickness tears and isolated subscapularis tendon injuries were excluded. Classifications of tendon integrity were selected from a previously published systematic review. MRI scans were observed separately by an orthopedic surgeon fellowship-trained in shoulder surgery and a radiologist expert in musculo-skeletal system. Each classification was evaluated on each possible sequence and plane. Intra-observer reliability was assessed by the orthopedic surgeon, who rated MRIs twice (at least two weeks between first and second observation). Inter- and intra-observer reliability were calculated with Cohen’s kappa coefficient for binary outcomes and with linear weighted kappa for ordinal variables, with coefficient value ranging from 0 (complete disagreement) to 1 (maximum agreement).
Inter-observer reliability was good (k=0.61 0.80) for the dichotomized Sugaya classification (type 1: Sugaya I-II; type 2: Sugaya III-V) in T2-weighted coronal views. All other classifications only showed a poor to moderate reliability regardless of sequence and planes. Modified Sugaya classification into four types showed excellent intra-observer reliability for T1-weighted coronal views and good reliability for T1-weighetd axial views. Good intra-observer reliability was found for binary outcome (healed vs re-torn) in T2-weighted coronal and sagittal, and PD sagittal views. Also dicothomized Sugaya classification showed a good reliability for T2-weighted coronal and sagittal views. Finally, Hayashida’s classification into 5 categories showed a good intra-observer reliability in T2-weighted sagittal view. All the other sequences and planes showed poor to moderate intra-observer reliability.
Dichotomic outcomes (healed vs re-torn), especially in T2-weigheted sequences were proved to be the most reliable classifications to assess the status of repaired rotator cuff tears. However, looking deeper into specific classifications, only few of them showed an acceptable reliability.