Search results

The search retrieved 1286 records. After removing duplicates, 411 articles were excluded. An additional 801 articles were excluded following the screening of titles and abstracts for obvious irrelevance. The remaining 74 records, 11 were excluded due to the lack of target outcomes of this review, 14 were excluded due to no meta-analysis conducted, 21 were excluded due to not focus on breast cancer screening. Ultimately, 28 studies were included in this review (Nelson et al. 2009; Zhu et al. 2018; Iared et al. 2011; Vinnicombe et al. 2009; Giampietro et al. 2020; Hodgson et al. 2016; Souza et al. 2013; Yun et al. 2017; Song et al. 2019; Marinovich et al. 2018; Phi et al. 2018, 2017; Farber et al. 2020; Nickson et al. 2012; Magnus et al. 2011; Gøtzsche and Olsen 2000; Hendrick et al. 1997; Kerlikowske et al. 1995; Jacklyn et al. 2016; Posso et al. 2017; Gøtzsche and Jørgensen 2013; Smart et al. 1995). Details of the PRISMA flow chart of literature studies for this overview are presented in Fig. 1.

Fig. 1

Flow diagram of the study selection

Characteristics of included systematic reviews

The detailed characteristics of the included 28 systematic reviews with meta-analysis included primary studies ranging from 4 to 24 are presented in Table 1, with 24 reviews from developed countries, 4 from developing countries. Twenty-four systematic reviews reported the databases searched, only 2 study retrieved Chinese databases. Sixteen reviews assessed the methodological quality, of which 13 used QUADAS-2, 1 used QUADAS, 1 used Cochrane RoB Assessment tool, and 1 used quality score evaluation system for assessing the quality of RCTs published by Chalmers et al.

Table 1 Main characteristics of included studiesResults of methodological quality

Methodological quality using the ROBIS tool was assessed in the 28 systematic reviews. Among the reviews, 18 systematic reviews (64.3%) were classified as having a high risk of bias (RoBs) in Domain 1: Study Eligibility Criteria. In Domain 2: Identification and Selection of Studies, 75% (21 systematic reviews) were rated with high RoBs. A total of 9 systematic reviews (32.1%) were considered to have high RoBs in Domain 3: Data Collection and Study Characteristics. Domain 4: Statistical Analysis and Interpretation showed that 57.1% (16 systematic reviews) had high RoBs. When considering the overall risk of bias, 23 systematic reviews (82.1%) were judged to have high RoBs. The complete ROBIS assessment results are illustrated in Figs. 2 and 3.

Fig. 2

ROBIS result of 28 included studies

Fig. 3

The full compliance rate of each ROBIS item

In terms of study eligibility criteria, 36.4% of the systematic reviews were assessed with low RoBs. Notably, only nine systematic reviews (32.1%) reported prior study protocols or registration, consistent with previous findings. For Domain 2 (Identification and Selection of Studies), 18.2% of the systematic reviews demonstrated low risk for each signaling question. Sixteen systematic reviews conducted searches across three or more databases, while twelve reported a detailed search strategy, identified unpublished studies, and searched registry platforms. Eight studies did not report clear follow-up times, and seven did not involve at least two reviewers independently performing the process of selecting studies for inclusion. Half of the systematic reviews used appropriate tools to assess risk of bias or methodological quality, and only eight studies assessed the RoBs of included systematic reviews independently with at least two reviewers.

Association between mammography screening and breast cancer mortality

Eleven systematic reviews compared the association of mammography vs. no screening on breast cancer mortality. Pooled estimates for reduction in breast cancer mortality attributable to mammography screening stratified by study design ranged from 0.51 (OR, 95% CI 0.46–0.55) to 1.04 (RR, 95% CI 0.84–1.27). An systematic reviews of 10 case–control studies average a 49% reduction in breast cancer mortality for women who are screened (0.51 [OR, 95% CI 0.46–0.55]), which was similar to systematic reviews with RCTs conducted by Gøtzsche et al. (0.75 [RR, 95% CI 0.67–0.83]), Hendrick et al. (0.71 [RR, 95% CI 0.57–0.89] and Cochrane analysis (0.71 [RR, 95% CI 0.61–0.83]) (Gøtzsche and Olsen 2000; Hendrick et al. 1997; Gøtzsche and Jørgensen 2013). Pooled estimates of systematic reviews with RCTs only conducted by Magnus (0.83 [RR, 95% CI 0.72–0.97]), Gøtzschewere et al. (1.04 [RR, 95% CI 0.84–1.27]), and Cochrane analysis (0.93 [RR, 95% CI 0.79–1.09]) generally higher (mortality reduction lower) than with those observed with the other study design trials (Table 2) (Magnus et al. 2011; Gøtzsche and Olsen 2000; Gøtzsche and Jørgensen 2013).

Table 2 Main results of the included meta-analysis evaluating association between mammography screening and breast cancer mortality

In the systematic reviews of RCTs that stratified by age, a significantly reduced breast cancer mortality among women with the latest follow-up data (aged 40–49) invited to screening mammography was observed by Hendrick et al. (0.82[RR, 95% CI 0.71–0.95], 7 RCTs) and Magnus et al. (0.83 [RR, 95% CI 0.72–0.97)], 8 RCTs) (Magnus et al. 2011 Jun; Hendrick et al. 1997); however, pooled estimates are not statistically significant for systematic reviews conducted by Nelson et al. (0.92, [RR, 95% CI 0.75–1.02], 9 trials), and Smart et al. (0.84 [RR, 95% CI 0.69–1.02], 8 RCTs) (Nelson et al. 2009; Smart et al. 1995). The Cochrane analysis with adequately randomized trials found no significant breast cancer mortality reduction among women below 50 years for follow up after 7 years (0.94 [RR, 95% CI 0.78–1.14]) and 13 years (0.87 [RR, 95% CI 0.73–1.03]). The Cochrane analysis reported that the pooled estimates for mortality reduction in screening women aged 50 years or older was lower when combined sub-optimally randomized trials with 7 and 13 years follow up (0.88 [RR, 95% CI 0.64–1.20] and 0.94 [RR, 95% CI 0.77–1.15]) than adequately randomized trials with 7 and 13 years (0.67 [RR, 95% CI 0.56–0.81] and 0.70 [RR, 95% CI 0.62–0.80]) (Gøtzsche and Jørgensen 2013). An systematic reviews conducted by Nelson et al. observed a statistically significant difference among women aged 50 to 59 (0.86 [RR, 95% CI 0.68–0.97]) and aged 60–69 years (0.67 [RR, 95% CI 0.54–0.83]), the combined RR is not statistically significant for women aged 70–74 (P > 0.05) (Nelson et al. 2009). Canelo-Aybar et al. found high certainty evidence that mammography screening reduces breast cancer mortality risk for women 50–69 (0.77 [RR, 95% CI 0.66–0.90]).

Performance of mammography screening technology

Nine systematic reviews reported the diagnostic accuracy of breast cancer screening conducted using mammography (Table 3). Sensitivity of different types of mammography was ranged from 55% to 90.77%, specificity of different types of mammography ranged from 84% to 97%. The sensitivity of digital breast tomosynthesis + full-field digital mammography, contrast-enhanced spectral mammography, digital mammography, film mammography, and full-field digital mammography, were 90.77% (95% CI 80.7–96.51%), 89% (95% CI 88–91%), 76% (95% CI 70–81%), 76% (95% CI 70–81%), and 60.00% (95% CI 47.10–71.96%), respectively. The specificity of film mammography, digital breast tomosynthesis + full-field digital mammography, digital mammography, full-field digital mammography, and contrast-enhanced spectral mammography were 97% (95% CI 94–98%), 96.49% (95% CI 96.04–96.90%), 96% (95% CI 94–97%), 95.55% (95% CI 95.04–96.01%), and 84% (95% CI 82–85%). The DOR of breast cancer screening conducted using the contrast-enhanced spectral mammography, digital mammography, and film mammography were 71.36(95% CI 36.28–140.39), 72(95% CI 44–118), and 91(95% CI 52–157). However, Alabousi et al. found that synthetic 2D mammography and synthetic 2D mammography plus digital breast tomosynthesis showed comparable diagnostic accuracy to digital mammography and digital mammography plus digital breast tomosynthesis, respectively.

Table 3 Main results and subgroup analysis of the included meta-analysis evaluating the accuracy of screening mammography

Seven systematic reviews compared the recall rate and CDR of screening mammography of different mammography (Table 4). The pooled estimates for CDR of digital breast tomosynthesis vs. digital mammography, digital breast tomosynthesis + synthetic 2D mammography vs. digital mammography, and digital breast tomosynthesis + full-field digital mammography vs. full-field digital mammography were 0.0016 (RD, 95% CI 0.0011–0.002), 1.38(RR, 95% CI 1.24–1.54), and 1.29 (RR, 95% CI 1.164–1.429). The CDR of film mammography was similar to full-field digital mammography (0.93[RR, 95% CI 0.83–1.03]), two studies evaluated the CDR of digital breast tomosynthesis + digital mammography and digital mammography and statistically significant were both observed (1.36 [RR, 95% CI 1.18–1.58] and 1.52 [RR, 95% CI 1.08–2.12]). Moreover, the pooled estimates for CDR of digital mammography vs. film mammography in two systematic reviews found a statistically significant (1.17 [RR, 95% CI 1.06–1.29] and 0.00051 [RD, 95% CI 0.00019–0.00083]). The risk ratios and the respective 95% CI of recall rate of digital mammography vs. film mammography in two systematic reviews were inconsistent (1.07 [RR, 95% CI 0.94–1.22] and 0.00695 [RD, 95% CI 0.00347–0.01042]). Moreover, the risk ratios and the respective 95% CI of recall rate of digital breast tomosynthesis + digital mammography vs. digital mammography were also inconsistent (1.13 [RR, 95% CI 0.96–1.32] and 0.72[RR, 95% CI 0.64–0.80]). The recall rate of film mammography vs. full-field digital mammography and digital breast tomosynthesis + synthetic 2D mammography vs. digital mammography were not statistically significant (0.95 [RR, 95% CI 0.71–1.26] and 1.08 [RR, 95% CI 0.92–1.26]). The RD of recall rate of digital breast tomosynthesis was statistically significant lower than digital mammography (− 0.0219 [− 0.0298, − 0.014]). Screening using synthetic 2D mammography + digital breast tomosynthesis has similar breast cancer detection but reduces recall and biopsy when compared with digital mammography + digital breast tomosynthesis.

Table 4 Main results and subgroup analysis of the included meta-analysis evaluating the recall rate and CDR of screening mammography