Introduction
Methods
Data sources
Statistical analysis
IVW | Weighted-median | MR-Egger | |
---|---|---|---|
Assumption | All genetic instrumental variables are valid or any horizontal pleiotropic effects of instruments are balanced | No more than 50% of the weight of the estimate is from invalid genetic instrumental variables No single instrumental variable contributes >50% of the weight | InSIDE (instrument strength independent of direct effect) assumption, which states that the effect of the instrument on the exposure is not correlated with any direct effect of the instrument on the outcome |
Equation |
\( {\widehat{\upbeta}}_{IVW}=\frac{\sum_{K=1}^K{E}_k{D}_k{\upsigma}_{Dk}^{-2}}{\sum_{K=1}^K{E}_k^2{\upsigma}_{Dk}^{-2}} \)
\( {SE}_{{\widehat{\upbeta}}_{IVW}}=\sqrt{\frac{1}{\sum_{k=1}^k{E}_K^2{\upsigma}_{Dk}^{-2}}} \)
E
K
is the mean change in exposure level per additional effect allele of SNP k; D
k
is the mean change in disease outcomes (e.g. log odds of CHD or levels of other cardiovascular disease risk factors) per additional effect allele of SNP k with SE σ
Dk
| Weighted-median estimator is the median of a distribution having estimate βj as: Pj = 100(S
j
− W
j
/2)th percentile P is the percentile for the jth ordered ratio estimate; W
j
is the weight given to the jth ordered ratio estimate, proportional to the inverse of the instrumental variable variance, and S
j
is the sum of weights up to and including the weight of the jth ordered ratio estimates, calculated using the following equation:
\( Sj={\sum}_{K=1}^j{W}_k \)
| MR-Egger uses a weighted linear regression of the gene–outcome coefficients θj on the gene–exposure coefficients δj: θj = β0E + βE × δj
All the δj associations are orientated to be positive. If β0E is truly zero (or were constrained to be zero) the MR-Egger slope estimate βE is the same as the β from IVW |
Application | The IVW estimate is a statistically efficient method but it can be biased even if just one genetic variant is invalid (i.e. if just one variant has horizontal pleiotropic effects) | The weighted-median estimator is a modification of the simple median approach and takes account of the variance of the individual genetic instruments | The MR-Egger method is used to test for directional horizontal pleiotropy and correct for this in MR analyses |
Results
Relationships between potential mediators and CHD
Outcome | Exposure | ||||||||
---|---|---|---|---|---|---|---|---|---|
BMI, SD | T2DM | FPG, mmol/l | HbA1c, % | FI, log-pmol/l | LDL-C, SD | HDL-C, SD | TG, SD | CHD | |
BMI, SD | 0 | −0.07 (0.009)*** | 0.007 (0.04) | 0.04 (0.05) | −0.51 (0.14)*** | −0.04 (0.01)* | 0.0008 (0.01) | −0.004 (0.02) | −0.015 (0.01) |
T2DM | 0.67 (0.19)*** | 0 | 1.23 (0.14)*** | 0.36 (0.2) | 0.53 (0.45) | −0.09 (0.06) | −0.03 (0.05) | 0.05 (0.08) | 0.10 (0.05)* |
FPG, mmol/l | 0.07 (0.02)*** | 0.08 (0.007)*** | 0 | 0.07 (0.07) | 0.05 (0.07) | −0.04 (0.01)** | −0.002 (0.01) | −0.01 (0.02) | 0.01 (0.01) |
HbA1c, % | 0.05 (0.03)*** | 0.03 (0.01)** | 0.45 (0.04)*** | 0 | −0.09 (0.1) | −0.01 (0.02) | −0.02 (0.02) | −0.03 (0.02) | 0.01 (0.02) |
HbA1c, mmol/mol | 0.55 (0.33)*** | 0.33 (0.11)** | 4.92 (0.44)*** | 0 | −0.98 (1.09) | −0.11 (0.22) | −0.22 (0.22) | −0.33 (0.22) | 0.11 (0.22) |
FI, log-pmol/l | 0.18 (0.03)*** | 0.06 (0.009)*** | 0.05 (0.05) | −0.06 (0.06) | 0 | −0.03 (0.02) | 0.007 (0.02) | 0.02 (0.02) | −0.008 (0.01) |
LDL-C, SD | −0.05 (0.07) | 0.02 (0.01) | 0.02 (0.04) | 0.09 (0.05) | 0.27 (0.14)* | 0 | −0.21 (0.02)*** | 0.19 (0.03)*** | −0.03 (0.02) |
HDL-C, SD | −0.23 (0.04)*** | −0.003 (0.01) | 0.04 (0.04) | 0.11 (0.05)* | 0.52 (0.16)*** | −0.18 (0.02)*** | 0 | −0.47 (0.03)*** | −0.02 (0.02) |
TG, SD | 0.20 (0.03)*** | 0.02 (0.01) | 0.03 (0.03) | −0.02 (0.05) | −0.21 (0.16) | 0.07 (0.03)* | −0.16 (0.01)*** | 0 | 0.001 (0.02) |
CHD | 0.37 (0.07)*** | 0.10 (0.03)*** | 0.19 (0.09)* | 0.31 (0.12)* | −0.49 (0.31) | 0.49 (0.05)*** | −0.13 (0.04)** | 0.21 (0.05)*** | 0 |
Effects of BMI on CHD and glycaemic and lipid traits
Exposure: BMI (n = 322,154) | Effect estimate | 95% CI |
p value |
---|---|---|---|
CHD (n = 60,801 case and 123,504 control participants)a
| |||
IVW | 1.45 | 1.27, 1.66 | < 0.001 |
Weighted-median | 1.44 | 1.24, 1.67 | < 0.001 |
MR-Egger regression | |||
Slope | 1.55 | 1.26, 1.91 | < 0.001 |
Intercept (directional pleiotropy) | 1.00 | 0.99, 1.00 | 0.50 |
Type 2 diabetes mellitus (n = 34,840 case and 114,981 control participants)a
| |||
IVW | 1.96 | 1.35, 2.83 | < 0.001 |
Weighted-median | 2.63 | 2.16, 3.21 | < 0.001 |
MR-Egger regression | |||
Slope | 3.42 | 2.63, 4.46 | < 0.001 |
Intercept (directional pleiotropy) | 0.98 | 0.98, 0.99 | < 0.001 |
Fasting glucose, mmol/l (n = 46,186)b
| |||
IVW | 0.07 | 0.03, 0.11 | < 0.001 |
Weighted-median | 0.08 | 0.05, 0.12 | < 0.001 |
MR-Egger regression | |||
Slope | 0.09 | 0.036, 0.15 | < 0.001 |
Intercept | −0.0007 | −0.002, 0.001 | 0.37 |
HbA1c, % (n = 46,368)b
| |||
IVW | 0.05 | 0.01, 0.08 | 0.005 |
Weighted-median | 0.09 | 0.04, 0.14 | < 0.001 |
MR-Egger regression | |||
Slope | 0.09 | 0.008, 0.16 | 0.03 |
Intercept | −0.001 | −0.003, 0.001 | 0.31 |
Fasting insulin, log-pmol/l (n = 38,238)b
| |||
IVW | 0.18 | 0.14, 0.22 | < 0.001 |
Weighted-median | 0.18 | 0.12, 0.24 | < 0.001 |
MR-Egger regression | |||
Slope | 0.16 | 0.07, 0.25 | < 0.001 |
Intercept | 0.0007 | −0.002, 0.003 | 0.60 |
LDL-cholesterol, SD (1 SD = 1.0 mmol/l) (n = 188,577)b
| |||
IVW | −0.05 | −0.19, 0.09 | 0.50 |
Weighted-median | −0.01 | −0.08, 0.05 | 0.66 |
MR-Egger regression | |||
Slope | −0.10 | −0.184, − 0.02 | 0.02 |
Intercept | 0.0016 | −0.001, 0.004 | 0.19 |
HDL-cholesterol, SD (1 SD = 0.40 mmol/l) (n = 188,577)b
| |||
IVW | −0.23 | −0.32, −0.15 | < 0.001 |
Weighted-median | −0.21 | −0.27, −0.16 | < 0.001 |
MR-Egger regression | |||
Slope | −0.23 | −0.307, −0.15 | < 0.001 |
Intercept | −0.0001 | −0.002, 0.002 | 0.90 |
Triacylglycerol, SD (1 SD = 1.024 mmol/l) (n = 188,577)b
| |||
IVW | 0.20 | 0.14, 0.26 | < 0.001 |
Weighted-median | 0.21 | 0.15, 0.27 | < 0.001 |
MR-Egger regression | |||
Slope | 0.17 | 0.09, 0.24 | < 0.001 |
Intercept | 0.001 | −0.001, 0.003 | 0.37 |
Effects of potential mediators on CHD
Risk factor | OR | 95% CI |
p value |
---|---|---|---|
Type 2 diabetes mellitus | |||
IVW | 1.12 | 1.06, 1.18 | < 0.001 |
Weighted-median | 1.11 | 1.05, 1.17 | < 0.001 |
MR-Egger regression | |||
Slope | 1.07 | 0.99, 1.15 | 0.10 |
Intercept | 1.01 | 1.00, 1.01 | 0.17 |
Fasting glucose, mmol/l | |||
IVW | 1.31 | 1.09, 1.58 | < 0.001 |
Weighted-median | 1.21 | 1.01, 1.44 | 0.03 |
MR-Egger regression | |||
Slope | 1.08 | 0.87, 1.35 | 0.50 |
Intercept | 1.01 | 1.00, 1.01 | 0.04 |
HbA1c, % | |||
IVW | 1.30 | 1.08, 1.56 | 0.01 |
Weighted-median | 1.36 | 1.07, 1.74 | 0.01 |
MR-Egger regression | |||
Slope | 1.66 | 1.03, 2.68 | 0.04 |
Intercept | 0.99 | 0.97, 1.01 | 0.27 |
Fasting insulin, log-pmol/l | |||
IVW | 2.80 | 1.89, 4.16 | < 0.001 |
Weighted-median | 2.61 | 1.61, 4.23 | < 0.001 |
MR-Egger regression | |||
Slope | 0.49 | 0.09, 2.59 | 0.40 |
Intercept | 1.03 | 1.00, 1.05 | 0.04 |
LDL-cholesterol, SD (1 SD = 1.0 mmol/l) | |||
IVW | 1.58 | 1.43, 1.75 | < 0.001 |
Weighted-median | 1.63 | 1.48, 1.80 | < 0.001 |
MR-Egger regression | |||
Slope | 1.74 | 1.59, 1.90 | < 0.001 |
Intercept | 0.99 | 0.98, 0.99 | 0.01 |
HDL-cholesterol, SD (1 SD = 0.4 mmol/l) | |||
IVW | 0.86 | 0.78, 0.95 | < 0.001 |
Weighted-median | 0.88 | 0.81, 0.95 | < 0.001 |
MR-Egger regression | |||
Slope | 1.03 | 0.95, 1.12 | 0.46 |
Intercept | 0.99 | 0.98, 0.99 | < 0.001 |
Triacylglycerol, SD (1 SD = 1.024 mmol/l) | |||
IVW | 1.24 | 1.10, 1.41 | < 0.001 |
Weighted-median | 1.23 | 1.11, 1.36 | < 0.001 |
MR-Egger regression | |||
Slope | 1.13 | 1.03, 1.24 | 0.01 |
Intercept | 1.01 | 1.003, 1.01 | < 0.001 |
Mediating effects of lipids and glycaemic traits on BMI–CHD effects
OR | 95% CI |
p value | Mediation effect (%) | |
---|---|---|---|---|
MR-IVW regression, crude | 1.45 | 1.27, 1.66 | < 0.001 | |
Multivariate model | ||||
(1) Adjusted for triacylglycerol | 1.16 | 0.99, 1.36 | 0.06 | 22 |
(2) Adjusted for HbA1c
| 1.36 | 1.19, 1.57 | 0.001 | 4 |
(3) Adjusted for type 2 diabetes | 1.35 | 1.17, 1.56 | 0.001 | – |
(4) Adjusted for triacylglycerol + HbA1c
| 1.09 | 0.94, 1.27 | 0.25 | 38 |
(5) Adjusted for triacylglycerol + type 2 diabetes | 1.10 | 0.94, 1.29 | 0.22 | – |