Table 2 Sensitivity analyses of NAFLD and all-cause and cardiovascular mortality.

From: Nonalcoholic fatty liver disease and mortality from all causes, cardiovascular disease, and cancer: a meta-analysis

Categories

All-cause mortality

Cardiovascular mortality

n

HR (95% CI)

I2 (%)

n

HR (95% CI)

I2 (%)

Using a random-effects model to pool risk estimates of included studies

12

1.34 (1.17–1.54)

80.0

7

1.13 (0.92–1.38)

57.5

Using a fixed-effects model to pool risk estimates of included studies

12

1.28 (1.21–1.35)

80.0

7

1.09 (0.96–1.23)

57.5

Excluding studies using the general population as the reference group

9

1.34 (1.11–1.61)

82.7

6

1.05 (0.87–1.27)

44.7

Excluding studies using liver biopsy to diagnose NAFLD

8

1.28 (1.07–1.54)

73.2

4

1.05 (0.81–1.37)

64.3

Excluding studies conducted in the population with comorbidity

8

1.27 (1.10–1.48)

78.8

5

1.15 (0.89–1.49)

71.1

Excluding studies with sample size more than 10000

8

1.40 (1.22–1.61)

25.0

4

1.24 (0.78–1.97)

74.7

Excluding studies with sample size less than 1000

8

1.32 (1.12–1.55)

84.8

6

1.14 (0.92–1.41)

64.3

Excluding studies without full adjustment for confoundersa

7

1.42 (1.17–1.73)

78.6

4

1.01 (0.85–1.20)

41.6

  1. aFull adjustment refers to adjustment for age, sex and ≥3 important confounders (cardiovascular diseases, diabetes, smoking, alcohol, body mass index, physical activity and socioeconomic status as well as factors associated with these confounders.