After systematic retrieval in the databases described earlier, 3078 citations were obtained. After deleting the duplicates (n=903), citations categorized as reviews (n=244), conference papers (n=702), and case reports (n=34), 1195 records were remained for the title and abstract screening. An additional 1139 citations were removed by this step, and 56 articles qualified for full-text article assessment; we excluded the nonmatched documents that were reviews (n=8) or covered non-HER2-overexpression breast tumors (n=9) as well as those with no valid control cohort (n=12) or no prognosis (n=15). Ultimately, 12 satisfactory studies16,17,20–23 were involved in the meta-analysis. The PRISMA flow diagram of selecting qualified clinical trials is outlined in Figure 1.
Characteristics of included studies
The original nations of the included articles were USA (n=2), France (n=1), Canada (n=3), the Netherlands (n=1), Spain (n=1), the UK (n=1), India (n=1), China (n=1), and Japan (n=1). The sample size ranged from 43 to 748 (median: 81.5), with a total number of 2366 subjects. The year range of included studies was 2008 to 2018. The other details including study duration, whether trastuzumab was received or absent, chemotherapy strategy, and which cohort (study cohort or control cohort) received radiotherapy are presented in Table 1.
(To view a larger version of Table 1, click here.)
Four eligible trials with 960 subjects were included to analyze the OS between NAT and AT. As shown in Figure 2, the pooled data indicated that there was no significant difference in the OS of HER2-overexpressing breast tumors after receiving NAT compared to AT (pooled OR=1.04; 95% CI, 0.47–2.33).
There were 8 valid studies with 1406 patients who were collected to investigate the comparison of the LRR rate between mastectomy and BCT in treating HER2-amplified breast cancer. The data analysis suggested that women with this tumor undergoing mastectomy benefited from a lower LRR rate than those in the treatment of BCT (pooled OR=0.58; 95% CI, 0.38–0.89) (Figure 3A). We further divided the included trials into subgroups based upon whether trastuzumab administration to patients was documented. Interestingly, the subgroup with absent trastuzumab still showed a lower LRR rate with mastectomy than with BCT (pooled OR=0.52; 95% CI, 0.31–0.88), while the LRR rate between these two surgical interventions was not significantly different in the subgroup that received trastuzumab (pooled OR=0.71; 95% CI, 0.34–1.49) (Figure 3B).
Risk of bias in analyzed studies
All included trials referencing the OS of NAT versus AT and the LRR of mastectomy versus BCT were combined to judge each risk of bias domain. The risk of bias summary and the risk of bias graph are presented in Figure 4A and B, respectively.
In the Begg’s funnel plots of the OS of NAT versus AT and the LRR of mastectomy versus BCT, data were uniformly arranged above and below the axis. Consistently, their publication bias in light of Egg’s test was not statistically significant (p=0.844 and 0.515, respectively), suggesting that the moderate heterogeneity in the OS of NAT versus AT was not by virtue of bias (Figure S1 and S2 in Supplementary, page 1 and 2; see original article).
READ FULL ARTICLE From Dovepress