Table S3. Jackknife sensitivity analyses of studies.

Study	RR (95% CI)
All-cause death
Omitting Arslan, 2006	0.40 (0.34, 0.48)
Omitting Ladwiniec, 2015	0.41 (0.34, 0.49)
Omitting Tomasello, 2015	0.42 (0.35, 0.50)
Omitting Choi, 2016	0.41 (0.34, 0.49)
Omitting Yang, 2016	0.42 (0.34, 0.52)
Subtotal	0.41 (0.35, 0.48)
Cardiac death
Omitting Arslan, 2006	0.39 (0.30, 0.51)
Omitting Ladwiniec, 2015	0.37 (0.28, 0.48)
Omitting Tomasello, 2015	0.42 (0.32, 0.55)
Omitting Choi, 2016	0.41 (0.31, 0.53)
Omitting Yang, 2016	0.40 (0.29, 0.54)
Subtotal	0.40 (0.31, 0.51)
MI
Omitting Arslan, 2006	0.57 (0.28, 1.19)
Omitting Ladwiniec, 2015	0.60 (0.25, 1.39)
Omitting Tomasello, 2015	0.94 (0.53, 1.66)
Omitting Choi, 2016	0.84 (0.45, 1.59)
Omitting Yang, 2016	0.71 (0.31, 1.60)
Subtotal	0.72 (0.38, 1.38)
Stroke
Omitting Arslan, 2006	0.20 (0.06, 0.71)
Omitting Tomasello,2015	0.46 (0.16, 1.35)
Omitting Choi, 2016	0.50 (0.15, 1.71)
Subtotal	0.36 (0.14, 0.92)
MACE
Omitting Arslan, 2006	0.61 (0.34, 1.09)
Omitting Ladwiniec, 2015	0.54 (0.39, 0.75)
Omitting Tomasello, 2015	0.73 (0.48, 1.11)
Omitting Choi, 2016	0.71 (0.46, 1.10)
Omitting Yang, 2016	0.64 (0.38, 1.10)
Subtotal	0.64 (0.43, 0.97)

Pooled estimates for all cause death, cardiac death, MI, stroke, and MACE were recalculated, using a random-effect model, by omitting one study at a time. Each line represents a re-analysis of the data with exclusion of one study at a time to assess the influence of this particular study on the overall result. Numbers in bold show the statistically significant differences. RR, risk ratio; CI, confidence interval; MI, myocardial infarction; MACE, major adverse cardiac events.