What is the Diagnostic Accuracy of MRI for Component Loosening in THA?

Abstract

Implant loosening is a common cause of reoperation after THA. Plain radiographs have been the default modality to evaluate loosening, although radiographs provide a relatively insensitive assessment of integration; cross-sectional modalities may provide a more detailed evaluation but traditionally have suffered from metal-related artifacts. We sought to determine whether MRI is capable of reliably detecting operatively confirmed component loosening in patients after hip arthroplasty. (1) Is assessing implant integration using MRI (with multiacquisition variable resonance image combination, [MAVRIC]) repeatable between readers? (2) What is the sensitivity and specificity of MRI with MAVRIC to evaluate component loosening, using intraoperative assessment as a gold standard? (3) How does the sensitivity and specificity of MRI with MAVRIC for surgically confirmed component loosening compare with those of radiographs? Between 2012 and 2017, 2582 THAs underwent revision at one institution. Of those, 219 had a preoperative MRI with MAVRIC. During that period, the most common indication for obtaining an MRI was evaluation of potential adverse local tissue reaction. The surgeons' decision to proceed with revision was based on their overall assessment of clinical, imaging, and laboratory findings, with MRI findings cited as contributing to the decision to revise commonly occurring in the setting of recalled implants. Of the THAs that underwent MRI, 212 were included in this study, while seven were excluded due to equivocal operative notes (5) and excessively poor quality MRI (2). MRI was performed at 1.5T using a standardized arthroplasty imaging protocol, including MARS (metal artifact reduction sequencing) and MAVRIC techniques. Two independent musculoskeletal fellowship-trained readers (one with 26 and one with 5 years of experience) blinded to operative findings scored a subset of 57 hips for implant integration based on Gruen zone and component loosening (defined as complete circumferential loss of integration around a component) to evaluate interobserver reliability. A third investigator blinded to imaging findings reviewed operative notes for details on the surgeon's assessment of intraoperative loosening. Gwet's agreement coefficients (AC) were used to describe interobserver agreement; these are similar to Cohen's kappa but are more resistant to certain paradoxes, such as unexpectedly low values in the setting of very high or low trait prevalence, or good agreement between readers on marginal counts. Almost perfect interobserver agreement (AC2 = 0.81-1.0) was demonstrated for all acetabular zones and all femoral Gruen zones on MRI, while perfect (AC1 = 1.0) agreement was demonstrated for the overall assessment of acetabular component loosening and near perfect agreement was shown for the assessment of femoral component loosening (AC1 = 0.98). MRI demonstrated a sensitivity and specificity of 83% (95% CI, 65-96) and 98% (95% CI, 97-100), respectively, for acetabular component loosening and 75% (95% CI, 55-94) and 100% (95% CI, 100-100), respectively, for femoral component loosening. Radiographs demonstrated a sensitivity and specificity of 26% (95% CI, 12-47) and 100% (95% CI, 96-100), respectively, for acetabular component loosening and 20% (95% CI, 9-47) and 100% (95% CI, 100-100), respectively, for femoral component loosening. MRI may provide a repeatable assessment of implant integration and demonstrated greater sensitivity than radiographs for surgically confirmed implant loosening in patients undergoing revision THA at a single institution. Additional multi-institutional studies may provide more insight into the generalizability of these findings. Level III, diagnostic study.