Effective treatment and good prognosis of breast cancer are due to early identification, but there are pros and cons of various screening methods, such as: mammography, MRI, biopsy, etc., which have high sensitivity but poor specificity, which often leads to over-diagnosis and unnecessary biopsies for patients [3, 6]. Liquid biopsy provides a new landscape for cancer identification and detection, and studies have found that a variety of miRNAs play an important role in the progression of breast cancer, especially miR-21 is significantly upregulated[29] which has been most extensively studied. To assess the value of miR-21 as a clinical application for breast cancer diagnosis, we conducted this meta-analysis with the aim of providing evidence-based clinical support through a comprehensive analysis.
It was concluded that miR-21 showed good diagnostic efficacy for identifying breast cancer. Using a fixed effects model, the sensitivity of the combination was 0.91 [95% CI (0.86, 0.95)], the specificity of the combination was 0.85 [ 95CI (0.77, 0.91)], and the DOR was 56.62 [ 95% CI (21.00, 184.83)].This is consistent with the findings of Shichao Li et al[30], but in their study only serum samples were included, whereas we included not only serum sources, but also plasma as well as saliva sources, and the sample size was much larger. In addition, we plotted SROC curves and crosshair-plots to assess the overall study distribution, and the point estimates of most studies and the combined point estimates were concentrated in the upper left corner of the graph, suggesting that miR-21 could provide a more intuitive diagnostic efficacy for clinical purposes. Subgroup analysis revealed that serum was more sensitive and less specific than plasma, which could be due to the way the blood samples were processed. In serum, during clot formation, blood cells may release markers, such as miR-21 and miR-16, which can lead to higher serum marker concentrations, and hemolysis can also affect the determination of results[31, 32]. Therefore, plasma may be a better assay for miR-21. miRNAs can be extracted from multiple pathways, such as saliva, serum, plasma, and urine, and we emphasize standardization in the marker extraction process to reduce errors. In addition, the limitations of miR-21 in liquid biopsies need further elucidation, including the overlap of disease spectrum, optimization of extraction methods, and criteria for threshold values. In addition, the findings suggest that miR-21 in combination with other miRNAs can further improve the diagnostic accuracy, especially the combination of three miRNAs has the best diagnostic performance; In a study by Seema Thakur et al[26], the AUC value of three miRNAs (miR-21 + miR-210 + miR-221) was 1, and the sensitivity and specificity of diagnosing breast cancer were significantly higher, which is consistent with our study; miR-21 was shown to be overexpressed in triple-negative breast cancer in several studies[22, 26, 33]. The above studies suggest that miR-21 combined with multiple miRNAs will further improve the diagnostic accuracy of breast cancer and has great potential in subtype prediction.
Evidence level grading by GRADE grading, miR-21 may have higher diagnostic efficacy compared with other included diagnostic methods (GRADE A). First, in this study, after evaluation by QUADAS-2 (Table 2), the risk of bias of the original study is small, so it is not downgraded in the risk of bias column; since the included original study, index test and the gold standard are consistent with our study, it is not downgraded in the indirectness column. By calculation, I2 < 50%, indicating that heterogeneity is small, so it is not degraded in the inconsistency column.Our study meets the minimum sample size required to test the efficacy of diagnostic tests, and the confidence interval for sensitivity meets the absolute width of the acceptable confidence interval, the confidence interval for specificity does not, but we need markers with high sensitivity but not necessarily high specificity, so we do not downgrade in the imprecision column; Finally, by Deek's funnel plot test, t = 0.16,P = 0.88, no significant publication bias was seen, so it was not downgraded in the column of publication bias.
There are also certain limitations in this meta-analysis. First, our literature search was limited to English databases, which would likely lead to publication bias and language bias; Second, it was difficult to be blinded because the studies we included were all retrospective studies; Third, breast cancer stage and subtype can affect the levels of circulating markers, but this meta-analysis did not conduct a discussion of breast cancer subtype and stage. Fourth, the miR-21 thresholds and sample sources of the included studies were not uniform, which may also be a source of heterogeneity.