Objective measures of adenomyosis on MRI and their diagnostic accuracy - a systematic review & meta-analysis.

INTRODUCTION Magnetic resonance imaging (MRI) diagnosis of adenomyosis is considered the most accurate non-invasive technique, but remains subjective, with no consensus on which diagnostic parameters are most accurate. We aimed to systematically review literature on how adenomyosis can be objectively quantified on MRI in a scoping manner, and review the diagnostic performance of these characteristics compared to histopathological diagnosis. Additionally, to summarize correlations between measures of adenomyosis on MRI and clinical outcomes. MATERIAL AND METHODS We searched databases Pubmed, Embase and Cochrane for relevant literature up to April 2020 according to PRISMA guidelines. We included studies that objectively assessed adenomyosis on MRI, and separately assessed studies investigating the diagnostic performance of MRI versus histopathology for inclusion in a meta-analysis. The QUADAS-2 tool was used for risk of bias, with many studies showing an unclear or high risk of bias. RESULTS 80 studies were included, of which 14 assessed the diagnostic performance of individual MRI parameters, with four included in the meta-analysis of diagnostic accuracy. Common MRI parameters were: junctional zone (JZ) characteristics such as maximal JZ thickness; pooled sensitivity 71.6 (95% CI 46.0 - 88.2), specificity 85.5% (52.3 - 97.0) , JZ differential; pooled sensitivity 58.9% (95% CI 44.3 - 72.1), specificity 83.2% (95% CI 71.3 - 90.8) and JZ to myometrial ratio; pooled sensitivity 63.3% (95% CI 51.9 - 73.4), specificity 79.4% (95% CI 42.0 - 95.4), adenomyosis lesion size , uterine morphology (pooled sensitivity 42.9% (95% CI 15.9 - 74.9), specificity 87.7%, (95%CI 37.9 - 98.8)) and changes in signal intensity (e.g. presence of myometrium cysts; pooled 59.6% (95% CI 41.6 - 75.4) and specificity of 96.1% (95% CI 80.7 - 99.3). Other MRI parameters have been used for adenomyosis diagnosis however their diagnostic performance is unknown. Few studies attempted to correlate adenomyosis MRI phenotype to clinical outcomes. CONCLUSIONS A wide range of objective parameters for adenomyosis exist on MRI; however, in many cases their individual diagnostic performance remains uncertain. JZ characteristics remain the most widely used and investigated with acceptable diagnostic accuracy. Specific research is needed into how these objective measures of adenomyosis can be correlated to clinical outcomes.