Toward the establishment of optimal computed tomographic parameters for the assessment of lumbar spinal fusion.
Academic Article
Overview
abstract
BACKGROUND CONTEXT: The accurate detection of the extent of bony fusion after attempted lumbar arthrodesis is important given that subsequent efforts-such as decisions regarding need for continued external bracing, use of enhancing modalities (electrical stimulation and pulsed ultrasound), recommended activity levels, return to employment, early surgical intervention, and others-may be needed to reduce the risk of late failure, especially in light of the fact that late revisions for failed fusions often result in poor outcomes and significant costs. Thin-cut computed tomography (CT) has emerged as the study of choice for this purpose. PURPOSE: To delineate the optimal CT parameters for determining fusion versus pseudarthosis after attempted lumbar fusion. STUDY DESIGN: Blinded CT assessment with cadaveric specimen as a gold standard. METHODS: A human cadaveric spine specimen with a T10 to S1 thoracolumbar posterolateral fusion augmented by instrumentation and anterior lumbar interbody fusions was used as a gold standard. Two experienced spine surgeons and one musculoskeletal radiologist-all blinded to the pathology results-assessed a series of CT scans of the specimen, each CT using one of six predefined sets of parameters. RESULTS: Predictive values and sensitivity generally improved with decreasing slice thickness and slice spacing, but only modestly. All sets of parameters had higher negative predictive value (NPV) than positive predictive value (PPV). Computed tomographic parameters of 0.9-mm thick sections with 50% overlap showed the highest PPV and NPV, where NPV was 90, but PPV was only 59. CONCLUSIONS: In this study, using the best widely available CT technologies and the ideal gold standard, thin-cut CT remained less than ideal for the assessment of lumbar arthrodesis/pseudarthrosis. Tuning slice thickness and slice spacing down generally improves detail, but marginally. We have successfully defined "optimal" as "best available," but "optimal" as "nearly perfect" awaits further technological advances.
