[MR imaging accuracy in the prediction of bone graft healing potential in scaphoid non-union]

To determine the accuracy of MR imaging in predicting bone graft healing in patients with scaphoid non-union.

21 patients with scaphoid non-union were examined with MR imaging prior to bone grafting (conventional bone graft in 14 cases and vascularized bone graft in 7 cases). The protocol included unenhanced and Gadolinium-enhanced sequences. Signal intensity and homogeneity of the proximal fragment was analysed by two independent radiologists. MRI findings were then correlated to the postoperative rate of union.

Healing occurred in 17 cases and failed in 4 cases with a mean follow up of 14 months. Intraobserver agreement in MR reading was respectively 0.92 et 0.86. Interobserver agreement was 0.88. On Tl-wi, the proximal fragment was hyperintense in 1 case (with positive surgical result), heterogeneous low signal intensity in 7 cases (healing in n = 7) and homogeneous low signal intensity in 13 cases (healing in n = 9). On T2-wi, the proximal fragment was hypointense in 4 cases (healing in n = 3), homogeneous high signal in 5 cases (healing in n = 4) and heterogeneous high signal intensity in 12 cases (healing in n = 10). After Gadolinium injection, enhancement was homogeneous in 4 cases (healing in n = 4), heterogeneous in 8 cases (healing in n = 7) and absent in 9 cases (healing in n = 6). In the group with no enhancement, 5 patients were treated with vascularized bone graft (healing in n = 4) and 4 with conventional bone graft (healing in n = 2).

The absence of enhancement of the proximal scaphoid fragment leads to poor surgical results except for vascularized bone graft.