Table 3. Mediation analysis results using data from a lung cancer genome-wide association study.* .
| Standard Approach | Our Approach | |||
| Estimates | 95% CIs | Estimates | 95% CIs | |
| a1 | 0.2477 | (0.1270, 0.3684) | 0.2231 | (0.1024, 0.3438) |
| a2 | 0.2395 | (0.0741, 0.4049) | 0.2092 | (0.0438, 0.3746) |
| d | 0.6482 | (0.4148, 0.8816) | 0.6397 | (0.4063, 0.8731) |
| b1 | 0.1036 | (−0.0667, 0.2739) | − | − |
| b2 | 1.0919 | (0.8422, 1.3416) | − | − |
| c′ | 0.2350 | (0.1101, 0.3599) | − | − |
| IE1 ( PM1 ) | 0.0257 (3.7%) | (−0.0181, 0.0752) | 0.1385 (18.3%) | (0.0601, 0.2195) |
| IE2 ( PM2 ) | 0.2615 (37.5%) | (0.0283, 0.4059) | 0.2284 (30.2%) | (0.0398, 0.4624) |
| IE3 ( PM3 ) | 0.1753 (25.1%) | (0.0733, 0.2995) | 0.1558 (20.6%) | (0.0566, 0.3123) |
| IEt ( PMt ) | 0.4625 (66.3%) | (0.1890, 0.6481) | 0.5227 (69.1%) | (0.2722, 0.8476) |
| TE | 0.6975 | (0.3864, 0.9079) | 0.7577 | (0.4846, 1.0987) |
CI: Confidence interval
*
Both daily smoking quantity and COPD were used as the mediators in the multiple-mediator model. The 95% CIs for the indirect effects were estimated based on 10,000 bootstraps. Includes the indirect effect through smoking (i.e., IE1 = a1b1), the indirect effect through COPD (i.e., IE2 = a2b2), the three-path indirect effect through smoking and COPD (i.e., IE3 = a1db2), the total indirect effect (i.e., IEt = IE1+IE2+IE3), the total effect (i.e., TE = IEt+c′), and the percentages of the SNP-lung cancer association explained by different paths (i.e., PM1 = IE1/TE, PM2 = IE2/TE, PM3 = IE3/TE, and PMt = IEt/TE).