Prognostic Performance of Ellipsoid Kidney Volume Measurement in a Computed Tomography–Heavy ADPKD Cohort: A Retrospective Observational Cohort Study
Marwan Abdrabou, Ayub Akbari, Wael Shabana, Mohan Biyani, Gregory L. Hundemer, Manish M. Sood, Arlene B. Chapman, Deena Fremont, Pierre Antoine BrownBackground
Accurate prognostic classification in autosomal dominant polycystic kidney disease is important for risk stratification and treatment decisions, including tolvaptan eligibility. Although the ellipsoid method has shown good performance for kidney volume measurement in magnetic resonance imaging cohorts, data from cohorts imaged predominantly with computed tomography remains limited.
Objective
To evaluate the agreement between the ellipsoid and manual segmentation methods for kidney volume measurement in a computed tomography-heavy cohort of patients with autosomal dominant polycystic kidney disease, with a focus on prognostic classification and identification of high-risk disease.
Design
Retrospective observational cohort study.
Setting
The Ottawa Hospital Cystic Kidney Disease Clinic, a tertiary academic center in Ottawa, Ontario, Canada.
Patients
One hundred fifty-one patients with autosomal dominant polycystic kidney disease who underwent cross-sectional abdominal imaging suitable for volumetric assessment between January 2018 and February 2022, including 107 computed tomography studies and 44 magnetic resonance imaging studies.
Measurements
Height-adjusted total kidney volume, Mayo Clinic Imaging Classification, and classification of high-risk disease (classes 1C-1E).
Methods
Height-adjusted total kidney volume was measured using both manual segmentation, the reference standard, and the ellipsoid method. Agreement in kidney volume measurement was assessed using intraclass correlation coefficients, Bland-Altman analysis, and mean percent difference. Agreement in Mayo Clinic Imaging Classification was assessed using weighted Cohen kappa. Diagnostic performance of the ellipsoid method for identifying high-risk disease was evaluated using sensitivity, specificity, positive predictive value, and McNemar testing.
Results
The ellipsoid method showed good concordance with manual segmentation height-adjusted total kidney volume measurement, with an intraclass correlation coefficient of 0.93 overall, 0.92 in the computed tomography subgroup, and 0.96 in the magnetic resonance imaging subgroup. The ellipsoid method underestimated kidney volume by a mean of 12.7%. Agreement in Mayo Clinic Imaging Classification was high, with a weighted Cohen kappa of 0.87, although 25.8% of cases were misclassified, most often to a lower risk class. For identification of high-risk disease, the ellipsoid method had a sensitivity of 83.0% and a specificity of 98.0% overall; in the computed tomography subgroup, sensitivity was 85.5% and specificity was 97.4%.
Limitations
Single-center design, use of a single reader, and cross-sectional analysis without longitudinal follow-up.
Discussion
In this computed tomography-heavy cohort, the ellipsoid method showed good concordance with manual segmentation for kidney volume measurement and high specificity for identifying high-risk disease. However, systematic underestimation may lead to downward misclassification, particularly in borderline cases.