Differentiation of benign and malignant lymph nodes with ADC value on diffusion MR: a Metaanalysis / 中华放射学杂志

ABSTRACT

Objective To summarize the diagnostic efficacy of ADC value for differentiation of benign and malignant lymph nodes on diffusion MRI with Meta-analysis. Methods Published papers on differentiation of benign and malignant lymph nodes with ADC value were searched and reviewed.Quality evaluation was performed for the eligible papers before data extraction.Test for heterogeneity was performed first,then appropriate model was selected to calculate the weighted mean difference,sensitivity,specificity,positive likelihood ratio,negative likelihood ratio,diagnostic odds ratio,pretest and posttest probability.The potential of ADC value for differentiation of benign and malignant lymph nodes was assessed qualitatively and quantitatively.Results Fifteen papers including 735 cases and 1963 lymph nodes were selected.According to Meta-regression analysis,subgroup analysis and robust analysis,two studies with benign lymph nodes in patients with benign lesion and one study using chemical shift saturation technique were excluded because of their impact on the robustness of the pooled results. The weighted mean difference (WMD) between malignant and benign lymph nodes was -0.355 × 10-3mm2/s[95％ confidence interval (CI)-0.423 ×10-3- -0.288 × 10-3 mm2/s].Although the cutoff of ADC value for differentiation in each study was different,the diagnostic efficacy was stable,the pooled sensitivity,specificity,positive likelihood ratio,negative likelihood ratio,diagnostic odds ratio and area under summarized receiver operator's curve were 0.87 (95％ CI0.79-0.92),0.87 (95％ CI0.82-0.90),6.5 (95％ CI4.7-9.2),0.15 (95％ CI0.09-0.25 ),43 ( 95％ CI21-87 ),0.93 ( 95 ％ CI0.90-0.95 ).The posttest malignancy probability of benign lymph node indicated by ADC was 6％,while that of malignant lymph node was 72％.Conclusion The ADC value can be used to differentiate benign and malignant lymph nodes with good sensitivity and specificity noninvasively.