Table 6.
Mediational effects of health measurements on the associations between ever WS exposure and all-cause mortality
| Potential mediator | Ever WS exposurea | Mediational effect size (%)b | Ppermc | |
|---|---|---|---|---|
| C (P = 0.0003) | C' (all Ps < 0.005) | |||
| SGRQ score | ||||
| Symptom | 0.42 (0.12) | 0.37 (0.12) | 12.0 | 0.03 |
| Activity | 0.42 (0.12) | 0.35 (0.12) | 17.0 | < 0.005 |
| Impact | 0.42 (0.12) | 0.35 (0.12) | 17.9 | < 0.005 |
| Total | 0.42 (0.12) | 0.33 (0.12) | 21.1 | < 0.005 |
| SF-36 score | ||||
| Physical functioning | 0.42 (0.12) | 0.33 (0.12) | 20.9 | < 0.005 |
| Role physical | 0.42 (0.12) | 0.39 (0.12) | 7.4 | 0.03 |
| Role emotional | 0.42 (0.12) | 0.40 (0.12) | 4.1 | 0.03 |
| Social functioning | 0.42 (0.12) | 0.41 (0.12) | 2.5 | 0.21 |
| Mental health | 0.42 (0.12) | 0.39 (0.12) | 6.8 | 0.02 |
| Vitality | 0.42 (0.12) | 0.40 (0.12) | 5.5 | 0.055 |
| General health perceptions | 0.42 (0.12) | 0.36 (0.12) | 15.5 | < 0.005 |
| Bodily pain | 0.42 (0.12) | 0.39 (0.12) | 6.6 | 0.015 |
| Spirometry | ||||
| FEV1 | 0.42 (0.12) | 0.34 (0.12) | 18.8 | < 0.005 |
| FEV1/FVC ratio | 0.42 (0.12) | 0.38 (0.12) | 10.2 | < 0.005 |
WS woodsmoke
aCox proportional hazards model assessed the impact of WS exposure on all-cause mortality. Baseline values of age, smoking status, packyears, and annual income, education, sex, and ethnicity were included in Cox proportional hazards model for covariate adjustment. C was the estimate for WS exposure in model without individual potential mediators. C' was the estimate for WS exposure in model with individual potential mediators.
bMediational effect size (%) was calculated as ([C–C'] × 100)/C
cPperm was calculated using permutation based method. The relationship between survival data (survival time and censor status) and the vector of independent variables was permuted for 200 times. Each permutated database allowed the association analysis of all-cause mortality with ever WS exposure and other covariates without and with including individual potential mediators to calculate the C and C'. Permutation was conducted for 200 times to generate the distribution of C–C' under null hypothesis of no mediation. Value of C–C' calculated using observed data was compared to the distribution generated by permutation and Pperm was calculated as the number of permuted databases generating a C–C' that exceeded the observed value divided by 500