Factors Impacting the Performance and Heterogeneity of Studies Assessing the Reliability of 2-Dimensional Shear Wave Elastography for Thyroid Malignancy Diagnosis

Abstract 2-dimensional shear wave elastography (2D-SWE) is a quantitative technique used in the evaluation of thyroid nodule stiffness to diagnose malignancy. High heterogeneity has been reported among the available 2D-SWE studies but the possible causes have remained unsubstantiated. Objective : This study aims to analyze the factors causing the high heterogeneity in diagnostic performance between studies and whether these factors impact on reliability of 2D-SWE for thyroid malignancy classification. Materials and methods : The literature search was carried out in PubMed, Scopus and Web of Science databases. The final 14 studies were reviewed and information on study design was retrieved including level of pre-compression (LoP) [none-minimal, unclear], plane of measurement (PoM) [sagittal, transverse, unclear], region of interest placement (ROIP) [whole lesion, stiffest, unclear], region of interest (ROI) size (ROIS) [largest, 2mm/less, unclear, variable], elasticity index (EI) parameters used [mean, maximum, minimum] and optimal EI parameter (OEIP). Pooled sensitivity and specificity estimates plus area under receiver operating characteristic curve and their respective 95% confidence intervals (CI) were used to assess 2D-SWE diagnostic performance among subgroups. Subgroup heterogeneity was significant at I 2 =50% and marked at I 2 =75%. A difference of ≥5 units in subgroup diagnostic performance and I 2 ≥ 25% in heterogeneity was deemed considerable. Univariate meta-regression of subgroup models and multivariate regression of methodological [LoP, ROIP, ROIS, OEIP] and spectrum composition [country, patient inclusion bias, malignancy prevalence] models was performed, p-value 50%) for both sensitivity and specificity estimates. ROIP, ROIS and study population were found to be potential independent predictors of DOR on univariate meta-regression (p=0.002, <0.001 and <0.001 respectively). On multivariate meta-regression, ROIS emerged as the only significant covariate affecting 2D-SWE diagnostic performance (p=0.022). There was a moderate threshold effect noted (r=0.494, p=0.032). Conclusion : ROIS was a consistently predominant covariate for 2D-SWE diagnostic performance in this study. However, the influence of LoP cannot be understated as the disparity in applied LoP was ubiquitous. Broadly, 2D-SWE study design standardization is critical in pursuit of a reference thyroid malignancy threshold EI. Otherwise the accuracy, reliability and reproducibility of 2D-SWE in the thyroid gland shall remain a medical quandary.