The diagnostic performance of cochlear endolymphatic hydrops and perilymphatic enhancement in stratifying Ménière's disease probabilities: A meta-analysis of semi-quantitative MRI-based grading systems

Abstract

Background: The diagnosis of Meniere's Disease (MD) presents significant challenges due to its complex symptomatology and the absence of definitive biomarkers. Advancements in MRI technology have spotlighted endolymphatic hydrops (EH) as a key pathological marker, necessitating a reevaluation of its diagnostic utility amidst the need for standardized and validated MRI-based grading scales.

Methods: Our meta-analysis scrutinized the diagnostic efficacy of semi-quantitative MRI-based cochlear endolymphatic hydrops (EH) and perilymphatic enhancement (PLE) grading systems in delineating clinically relevant discriminations: "Spotting" the shift from normal or asymptomatic ears to possible/probable MD (pMD), "Confirming" the progression to definite MD (dMD), and "Establishing" the presence of dMD. A thorough literature search up to October 2023 resulted in 35 pertinent studies, forming the basis of our analysis through a bivariate mixed-effects regression model.

Results: Using criteria from the American Academy of Otolaryngology-Head and Neck Surgery (AAO-HNS) and Barany Society, across varying thresholds and disease probabilities; the Establishment model at an EH grade 1 threshold revealed a sensitivity of 85.4% and a specificity of 82.7%. Adjusting the threshold to EH grade 2 results in a sensitivity increase to 92.1% (CI: 85.9-95.7) and a specificity decrease to 70.6% (CI: 64.5-76.1), with a DOR of 28.056 (CI: 14.917-52.770). The Confirmation model yields a DOR of 5.216, indicating a lower diagnostic accuracy. The Spotting model demonstrates a sensitivity of 48.3% (CI: 34.8-62.1) and a specificity of 88.0% (CI: 77.8-93.9), with a DOR of 6.882. The normal ears subgroup demonstrated a notably high specificity of 89.7%, while employing Nakashima's criteria resulted in a reduced sensitivity of 74.9%, significantly diverging from other systems (p-value < 0.001). The PLE grading system showcased exceptional sensitivity of 98.4% (CI: 93.7-99.6, p-value < 0.001).

Conclusion: Our meta-analysis supports a tailored diagnostic approach for MD, emphasizing the need for effective grading systems at each stage. For "Spotting," the model shows high specificity but requires improved sensitivity, suggesting additional criteria are needed. The "Confirming" stage highlights the need for refined, sensitive grading systems due to lower diagnostic accuracy. In the "Establishing" stage, an EH grade 1 threshold is effective, but grade 2 enhances sensitivity while reducing specificity, indicating a need for balance. The PLE grading system excels in sensitivity, making it highly reliable. High specificity in the normal ears subgroup confirms accurate non-pathological distinction, though Nakashima's criteria show reduced sensitivity, underscoring variability in grading systems. These findings advocate for a standardized, unified grading system balancing sensitivity and specificity across all MD stages to optimize diagnostics and clinical outcomes.

Fig 1. PRISMA flowchart of study selection process.

This flowchart illustrates the systematic process of study selection, adhering to the PRISMA guidelines.

Fig 2. MRI-based cochlear hydrops grading and PLE in MD.

Fig 3. HSROC analysis for MD diagnosis at grade 1 and 2 thresholds.

Part (a) examines diagnostic efficacy at Grade 1, showing sensitivity and specificity for various criteria and gadolinium routes. Part (b) compares diagnostic accuracy at Grade 2, assessing criteria performance at this elevated severity. The opening plot of each section shows the aggregate model performance including all subgroups.

Fig 4. HSROC analysis for MD at grade 1 threshold.

Section (a) contrasts definite Meniere’s Disease (dMD) against probable Meniere’s Disease (pMD), while section (b) differentiates probable MD (pMD) from control groups. The opening plot of each section shows the aggregate model performance including all subgroups.