Diagnostic value of seven biomarkers for breast cancer: an overview with evidence mapping and indirect comparisons of diagnostic test accuracy.

Several meta-analyses have evaluated the value of biomarkers in diagnosing breast cancer, but which biomarker has the optimal diagnostic value remains unclear. This overview aimed to compare the accuracy of different biomarkers in diagnosing breast cancer. PubMed, Embase.com, the Cochrane Library of Systematic Reviews, and Web of Science were searched. The assessment of multiple systematic reviews-2 (AMSTAR-2) was used to assess the methodological quality and preferred reporting items for a systematic review and meta-analysis of diagnostic test accuracy (PRISMA-DTA) for reporting quality. Pairwise meta-analyses were performed to estimate the pooled results for each biomarker, and indirect comparisons were conducted to compare diagnostic accuracy between biomarkers. Eleven systematic reviews (SRs) involving 218 original studies were included. All SRs were of critically low methodological quality, 3 SRs had minimal reporting flaws and 8 SRs had minor flaws. The pooled sensitivity and specificity were 0.77 and 0.87 for miRNA, 0.70 and 0.87 for circulating cell-free DNA, 0.29 and 0.96 for APC gene promoter methylation, 0.69 and 0.99 for 14-3-3σ promoter methylation, 0.63 and 0.82 for CA153, 0.58 and 0.87 for CEA, and 0.73 and 0.56 for PSA. Compared with CA153 and PSA, miRNA had a higher sensitivity and specificity. The sensitivity of miRNA was higher than circulating cell-free DNA and CEA, although they had the same specificities. APC gene promoter methylation and 14-3-3σ promoter methylation were more specific than miRNA, but they had unacceptably low sensitivity. In conclusion, miRNA had better diagnostic accuracy than the other six biomarkers. But due to the low quality of included SRs, the results need to be interpreted with caution. Further study should investigate the diagnostic accuracy of different biomarkers in direct comparisons and focus on the value of combined biomarkers.