Table 3.
Sensitivity analyses for outcome of coronary heart disease death and non-fatal myocardial infarction with different samples of trials*
| Regression model and predictor | Change in risk per 10 mg/dl increase in lipid subfraction—% (95% CI) | P value | R2† |
|---|---|---|---|
| Trials using interventions known to raise HDL cholesterol, excluding trials using agents associated with harmful effects‡ | |||
| Follow-up ≥6 months (n=70) | |||
| Univariable: | |||
| Change in LDL | 5.5 (3.9 to 7.0) | <0.001 | 0.41 |
| Change in HDL | −28.9 (−51.7 to −6.6) | 0.01 | 0.09 |
| Bivariable: | |||
| Change in LDL | 5.4 (3.6 to 7.2) | <0.001 | 0.41 |
| Change in HDL | −2.2 (−22.1 to 17.4) | 0.83 | |
| Multivariable§: | |||
| Change in LDL | 6.9 (4.2 to 9.6) | <0.001 | 0.51 |
| Change in HDL | 15.2 (−8.2 to 38.1) | 0.20 | |
| Follow-up >1 year (n=59) | |||
| Univariable: | |||
| Change in LDL | 5.5 (4.0 to 7.1) | <0.001 | 0.46 |
| Change in HDL | −29.0 (−52.7 to −5.9) | 0.02 | 0.10 |
| Bivariable: | |||
| Change in LDL | 5.5 (3.7 to 7.2) | <0.001 | 0.46 |
| Change in HDL | −2.0 (−22.0 to 17.6) | 0.84 | |
| Multivariable§: | |||
| Change in LDL | 7.1 (4.5 to 9.8) | <0.001 | 0.58 |
| Change in HDL | 16.2 (−7.0 to 38.8) | 0.17 | |
| Follow-up >2 years (n=49) | |||
| Univariable: | |||
| Change in LDL | 5.6 (4.0 to 7.2) | <0.001 | 0.51 |
| Change in HDL | −27.5 (−53.6 to −2.0) | 0.04 | 0.09 |
| Bivariable: | |||
| Change in LDL | 5.7 (3.9 to 7.5) | <0.001 | 0.52 |
| Change in HDL | 3.1 (−17.9 to 23.7) | 0.77 | |
| Multivariable§: | |||
| Change in LDL | 7.5 (4.8 to 10.1) | <0.001 | 0.63 |
| Change in HDL | 22.4 (−1.4 to 45.7) | 0.06 | |
| Trials using interventions specifically chosen to raise HDL cholesterol (niacin combinations and fibrates) (n=13) | |||
| Univariable: | |||
| Change in LDL | 5.7 (−102 to 20.8) | 0.44 | 0.07 |
| Change in HDL | −10.1 (−61.8 to 39.2) | 0.55 | 0.04 |
| Bivariable: | |||
| Change in LDL | 5.0 (−16.1 to 24.7) | 0.62 | 0.06 |
| Change in HDL | −5.0 (−74.7 to 60.1) | 0.89 | |
| Multivariable§: | |||
| Change in LDL | −2.3 (−23.7 to 18.6) | 0.81 | 0.30 |
| Change in HDL | −6.3 (−70.6 to 54.2) | 0.82 | |
| Trials with intention to treat data (n=78) | |||
| Univariable: | |||
| Change in LDL | 4.9 (3.4 to 6.5) | <0.001 | 0.35 |
| Change in HDL | −11.7 (−30.4 to 6.7) | 0.21 | 0.02 |
| Bivariable: | |||
| Change in LDL | 5.0 (3.4 to 6.6) | <0.001 | 0.35 |
| Change in HDL | 2.6 (−13.2 to 18.1) | 0.74 | |
| Multivariable§: | |||
| Change in LDL | 7.1 (4.1 to 10.0) | <0.001 | 0.47 |
| Change in HDL | 11.1 (−13.6 to 35.2) | 0.37 | |
HDL=high density lipoprotein; LDL=low density lipoprotein.
*I2 associated with risk ratios for coronary heart disease events was 41.1% (95% uncertainty interval 24.5% to 54.0%) in all trials (n=95), 28.3% (3.4% to 46.8%) in trials using interventions known to raise HDL cholesterol and excluding trials using agents associated with harmful effects (n=70), and 13.3% (0% to 52.5%) in trials using interventions specifically chosen to raise HDL cholesterol (n=13).
†Proportion of total variability in log risk ratio of outcome explained by model.
‡Excluding trials using n-3 fatty acids, low fat diets, probucol, torcetrapib, and hormones and trials focusing on patients with renal failure.
§Models include adjustment for drug class in addition to variables of lipid subfractions.