A systematic review of the variability of ventilation defect percent generated from hyperpolarized noble gas pulmonary magnetic resonance imaging

Hyperpolarized (HP) gas pulmonary MR ventilation images are typically quantified using ventilation defect percent (VDP); however, the test-retest variability of VDP has not been systematically established in multi-center trials. Herein, we perform a systematic review of the test-retest literature on the variability of VDP, and similar metrics, generated from HP MRI. This review utilizes the Medline, EMBASE, and EBM Reviews databases and includes studies that assessed the variability of HP MRI VDP. The protocol was registered to PROSPERO: CRD42022328535. Imaging techniques and statistical analysis characteristics were extracted and used to group studies to evaluate the overall ability to pool data across grouped studies. The ability to pool data to provide systematic evidence was assessed using a modified COSMIN tool. A total of 22 studies with 37 distinct aims for repeated HP MRI acquisition or quantification were included. Studies were grouped into six categories based on HP gas and analysis type: repeated imaging (129Xe n = 13, 3He n = 12), interobserver repeated analysis (129Xe n = 4, 3He n = 4) or intraobserver repeated analysis (129Xe n = 1, 3He n = 2). Studies assessed variability using a variety of statistical tests including absolute difference, percent coefficient of variation, Bland-Altman limits of agreement, coefficient of reproducibility, or the intra-class correlation. Individual studies generally reported low variability of VDP (ICC range: 0.5–1.0; Bland-Altman bias range: −6.9–20%), but there was an overall inability to pool data and provide a meta-analysis due to methodological inconsistencies and small sample size. Overall, we found that VDP has low variability in most studies. However, inconsistent image acquisition and quantification methodologies between studies limits direct comparability and precludes grouping of study data for meta-analyses. Despite early efforts to standardize HP MRI acquisition, further work is necessary to standardize VDP quantification to allow broader validation and clinical implementation.