Figure3 Sensitivity analysis for OS(A) and PFS(B).

We use funnel plot and Egger’s test as well as Begg’s test to detect publication bias. For OS, the conventional funnel plot seems symmetric (Figure 4A), however, the result of Egger’s test and Begg’s test (Egger’s test, P = 0.0407; Begg’s test, P = 0.0487) implied that publication bias may exist. Furthermore, we made funnel plot of precision by log hazard ratio (Figure 4B), and the result showed some asymmetries. Hence, we suspect publication bias did exist for OS.

For PFS, the publication bias was not detected. Both conventional funnel plot and precise funnel plot were symmetric (Figure 4C and 4D). The result of Egger’s test and Begg’s test also support that (Egger’s test, P = 0.272; Begg’s test, P = 0.348).

We also made cumulative analysis presented in Figure5A and 5B, the result demonstrated the negative influence of PPIs use with the passage of time and appearance of more evidence.