[Meta-analysis of ultrasonography in diagnosis of deeply infiltrating endometriosis]

Abstract

Objective: To evaluate the quality of literatures and the accuracy of ultrasonography in diagnosis of deeply infiltrating endometriosis (DIE).

Methods: The database of Medline (1966 to 2009), the excerpta medica database (EMBASE, 1980 to 2009), Chinese biological medicine on disc (CBMdisc, 1978 to 2009), China national knowledge infrastructure (CNKI, 1979 to 2009) and VIP for Chinese technology periodical database (VIP, 1989 to 2009) Cochrane library of studies about the diagnosis of ultrasound for DIE were searched and analyzed. Quality assessment of diagnostic accuracy studies (QUADAS) items were used to evaluate the quality of literatures. The sensitivity, specificity, positive likelihood ratio (+LR), negative likelihood ratio (-LR), diagnostic test odds ratio (DOR) for the pooled analysis and heterogeneity test were analyzed for transvaginal ultrasonography (TVUS), transrectal ultrasonography (TRUS) and rectal endoscopic sonography (EUS) by Meta-disc software, and drew the summary receiver operating characteristic (SROC) curves for those without heterogeneity.

Results: Totally 15 literatures in English were enrolled into this study. The positive rate of 10 items of QUADAS were above 60%, whereas that of the 11th item "Were the reference standard results interpreted without knowledge of the results of the index test" was 46.7%; none of studies had mentioned the 13th item "Were uninterpretable/intermediate test results reported". All researches had no heterogeneity by explored threshold effect. The results of pooled sensitivity, specificity, +LR, -LR, DOR were 0.925, 0.986, 30.036, 0.107, 299.25 for TRUS, 0.799, 0.944, 11.972, 0.187, 69.126 for transvaginal ultrasonography (TVUS), and 0.635, 0.928, 8.022, 0.320, 39.606 for EUS, respectively. Area under the curve of EUS was 0.9479, and that of TVUS was 0.9246.

Conclusions: TRUS, TVUS and EUS all showed optimal value in diagnosis of DIE. The bias identified from the 15 studies might be mainly resulted from reference standard review bias.