Interreader and intermodality reliability of standard anteroposterior radiograph and magnetic resonance imaging in detection and classification of lumbosacral transitional vertebra

Background context: Different types of lumbosacral transitional vertebra (LSTV) are classified based on the relationship of the transverse process of the last lumbar vertebra to the sacrum. The Ferguson view (30° angled anteroposterior [AP] radiograph) is supposed to have a sufficient interreader reliability in classification of LSTV, but is not routinely available. Standard AP radiographs and magnetic resonance imaging (MRI) are often available, but their reliability in detection and classification of LSTV is unknown.

Purpose: The purpose of this study was to evaluate the interreader reliability of detection and classification of LSTV with standard AP radiographs and report its accuracy by use of intermodality statistics compared with MRI as the gold standard.

Study design/setting: Retrospective case control study.

Patient sample: A total of 155 subjects (93 cases: LSTV type 2 or higher; 62 controls).

Outcome measures: Interreader reliability in detection and classification of LSTV using standard AP radiographs and coronal MRI as well as accuracy of radiographs compared with MRI.

Methods: After institutional review board approval, coronal MRI scans and conventional AP radiographs of 155 subjects (93 LSTV type 2 or higher and 62 controls) were retrospectively reviewed by two independent, blinded readers and classified according to the Castellvi classification. Interreader reliability was assessed using kappa statistics for detection of an LSTV and identification of all subtypes (six variants; 1: no LSTV or type I, 2: LSTV type 2a, 3: LSTV type 2b, 4: LSTV type 3a, 5: LSTV type 3b, 6: LSTV type 4) for MRI scans and standard AP radiographs. Further, accuracy and positive and negative predictive values were calculated for standard AP radiographs to detect and classify LSTV using MRI as the gold standard.

Results: The interreader reliability was at most moderate for the detection (k=0.53) and fair for classification (wk=0.39) of LSTV in standard AP radiograph. However, the interreader reliability was very good for detection (k=0.93) and classification (wk=0.83) of LSTV in MRI. The accuracy and positive and negative predictive values of standard AP radiograph were 76% to 84%, 72% to 86%, and 79% to 81% for the detection and 53% to 58%, 51% to 76%, and 49% to 55% for the classification of LSTV, respectively.

Conclusion: Standard AP radiographs are insufficient to detect or classify LSTV. Coronal MRI scans, however, are highly reliable for classification of LSTV.