A meta-analysis on the diagnostic value of diffusion-weighted imaging on ovarian cancer.

PURPOSE

This study systematically evaluated the potential influences of diffusion- weighted imaging (DWI) on the initial diagnosis, clinical decision making and diagnostic accuracy of ovarian cancer in the follow-up period.

METHODS

Literature on the correlation between DWI and diagnosis of ovarian cancer were searched from PubMed, Embase, Cochrane Library, and Web of Science published before January 1, 2019. References in enrolled eligible literature were manually reviewed. Quality assessment on the diagnostic accuracy was performed using the QUADAS scale. Receiver operating characteristics (ROC) curve was depicted using STATA 12.0. Study heterogeneity and its sources were determined. Sensitivity (SEN), specificity (SPF), positive likelihood ratio (+LR), negative likelihood ratio (-LR) and diagnostic odds ratio (DOR) of eligible studies were calculated for depicting forest plot and summary of ROC curve (SROC). The area under the curve (AUC) was calculated.

RESULTS

A total of 15 articles involving 930 ovarian cancer cases and 832 control cases were enrolled. DWI was identified to exert a certain diagnostic value on ovarian cancer. The 95%CI of the merged SEN (91%, 95%CI=84-95%), SPF (85%, 95%CI=78-90%), +LR (6.18, 95%CI=4.17-9.15) and -LR (4.05, 95%CI=3.30-4.79) were calculated using the random-effects model due to the slight heterogeneity among these studies. AUC was 0.94 (95%CI=0.91-0.96). Subgroup analysis in Asian population obtained the following results: SEN was 85% (95%CI=78-91%), SPF 83% (95%CI=72-90%), +LR 0.18 (95%CI=0.11-0.27), -LR 3.34 (95%CI=2.60-4.09) and DOR 3.34 (95%CI=2.60-4.09); AUC was 0.91 (95%CI=0.88-0.93). In Caucasian population, SEN was 96% (95%CI=83-99%), SPF 89% (95%CI=84-93%), +LR 41.36 (95%CI=5.95-287.48), -LR 0.06 (95%CI=0.02-0.18) and DOR 5.31(95%CI=3.93-6.69); AUC was 0.94 (95%CI=0.91-0.96).

CONCLUSIONS

This meta-analysis proved that DWI exerted a relatively high sensitivity and specificity in diagnosing ovarian cancer, especially in the Caucasian population. This conclusion still needs to be further verified in a multi-center study with a large sample size.