Inter- and intraobserver reliability of non-weight-bearing foot radiographs compared with CT in Lisfranc injuries

BACKGROUND: Injury of the tarsometatarsal (TMT) joint complex, known as Lisfranc injury, covers a wide range of injuries from subtle ligamentous injuries to severely displaced crush injuries. Although it is known that these injuries are commonly missed, the literature on the accuracy of the diagnostics is limited. The diagnostic accuracy of non-weight-bearing radiography (inter- or intraobserver reliability), however, has not previously been assessed among patients with Lisfranc injury. METHODS: One hundred sets of foot radiographs acquired due to acute foot injury were collected and anonymised. The diagnosis of these patients was confirmed with a CT scan. In one-third of the radiographs, there was no Lisfranc injury; in one-third, a nondisplaced (< 2 mm) injury; and in one-third, a displaced injury. The radiographs were assessed independently by three senior orthopaedic surgeons and three orthopaedic surgery residents. RESULTS: Fleiss kappa (κ) coefficient for interobserver reliability resulted in moderate correlation κ = 0.50 (95% CI: 0.45– 0.55) (first evaluation) and κ = 0.58 (95% CI: 0.52–0.63) (second evaluation). After three months, the evaluation was repeated and the Cohen’s kappa (κ) coefficient for intraobserver reliability showed substantial correlation κ = 0.71 (from 0.64 to 0.85). The mean (range) sensitivity was 76.1% (60.6–92.4) and specificity was 85.3% (52.9–100). The sensitivity of subtle injuries was lower than severe injuries (65.4% vs 87.1% p = 0.003). CONCLUSIONS: Diagnosis of Lisfranc injury based on non-weight-bearing radiographs has moderate agreement between observers and substantial agreement between the same observer in different moments. A substantial number (24%) of injuries are missed if only non-weight-bearing radiographs are used. Nondisplaced injuries were more commonly missed than displaced injuries, and therefore, special caution should be used when the clinical signs are subtle. LEVEL OF EVIDENCE: III.