Table 2 Results of empirical analyses with different IV estimators derived from various MR methods.

From: Exploiting horizontal pleiotropy to search for causal pathways within a Mendelian randomization framework

Methods

All variants

Estimates (95% CIs)

Removing outliers

Removing candidate outliers

Adjustment for candidate outliers

Empirical analysis 1: systolic blood pressure (mmHg) on coronary heart disease (Odds ratio) 

 Heterogeneity (Q)a

682.7 (N SNPs = 157)

312.1 (N SNPs = 150)

448.7 (N SNPs = 155)

567.6 (N SNPs = 157)

IVW random effects

1.761 (1.474, 2.104)

1.876 (1.655, 2.125)

1.797 (0.558, 5.789)

1.706 (1.449, 2.008)

Egger random effects

2.641 (1.490, 4.679)

2.951 (1.970, 4.419)

2.206 (0.314, 15.472)

–

Intercept

0.980 (0.969, 0.992)

0.990 (0.982, 0.998)

0.996 (0.988, 1.004)

–

Weighted median

1.770 (1.528, 2.050)

1.782 (1.539, 2.065)

1.765 (0.576, 5.403)

–

Weighted mode

1.770 (1.264, 2.479)

1.726 (1.218, 2.447)

1.740 (0.600, 5.043)

–

Empirical analysis 2: urate (mg/dl) and coronary heart disease (Odds ratio)

Heterogeneity (Q)

81.6 (N SNPs = 24)

20.7 (N SNPs = 21)

33.4 (N SNPs = 22)

44.1 (N SNPs = 24)

IVW random effects

1.081 (0.996, 1.174)

1.054 (1.008, 1.103)

1.062 (1.057, 1.122)

1.070 (0.992, 1.155)

Egger random effects

0.952 (0.846, 1.071)

1.008 (0.937, 1.084)

0.990 (0.910, 1.077)

–

Intercept

1.015 (1.003, 1.027)

1.006 (0.998, 1.014)

0.992 (0.984, 1.000)

–

Weighted median

1.019 (0.961, 1.081)

1.016 (0.958, 1.078)

1.017 (0.961, 1.077)

–

Weighted mode

1.028 (0.975, 1.084)

1.022 (0.966, 1.082)

1.025 (0.970, 1.083)

–

Empirical analysis 3: sleep duration (hour/night) and schizophrenia (Odds ratio)

Heterogeneity (Q)

204.8 (N SNPs = 36)

54.1 (N SNPs = 30)

121.4 (N SNPs = 34)

147.7 (N SNPs = 36)

IVW random effects

1.184 (0.573, 2.445)

1.289 (0.828, 2.008)

1.215 (0.674, 2.192)

1.181 (0.634, 2.197)

Egger random effects

0.866 (0.056, 13.383)

2.428 (0.485, 12.158)

2.363 (0.254, 21.955)

–

Intercept

1.004 (0.968, 1.042)

0.991 (0.969, 1.013)

0.991 (0.963, 1.020)

–

Weighted median

1.276 (0.774, 2.104)

1.249 (0.746, 2.090)

1.250 (0.761, 2.052)

–

Weighted mode

1.327 (0.679, 2.593)

1.504 (0.728, 3.105)

1.428 (0.702, 2.904)

–

Empirical analysis 4: years of schooling (years) and body mass index (kg/m2)

Heterogeneity (Q)

211.9 (N SNPs = 59)

101.9 (N SNPs = 56)

101.9 (N SNPs = 56)

197.8 (N SNPs = 59)

IVW random effects

−0.272 (−0.386, −0.158)

−0.232 (−0.314, −0.150)

−0.232 (−0.314, −0.150)

−0.265 (−0.377, −0.153)

Egger random effects

0.013 (−0.677, 0.703)

−0.404 (−0.910, 0.102)

−0.404 (−0.910, 0.102)

–

Intercept

−0.005 (−0.017, 0.007)

0.003 (−0.005, 0.011)

0.003 (−0.005, 0.011)

–

Weighted median

−0.209 (−0.307, −0.111)

−0.217 (−0.315, −0.119)

−0.217 (−0.315, −0.119)

–

Weighted mode

−0.141 (−0.413, 0.131)

−0.127 (−0.405, 0.151)

−0.127 (−0.405, 0.151)

–

  1. N SNPs number of single nucleotide polymorphisms, 95% CIs 95% confidence intervals, IVW inverse variance weighted. Empirical analysis 1: systolic blood pressure (mmHg) and coronary heart disease (log odds); Empirical analysis 2: urate (mg/dl) and coronary heart disease (log odds); Empirical analysis 3: sleep duration (hour/night) and schizophrenia (log odds); Empirical analysis 4: years of schooling (years) and body mass index (kg/m2).
  2. aHeterogeneity amongst the estimates were assessed based on contribution of individual variant to Cochran’s statistic.
Back to article page