Table 3.
Random-effects meta-analysis with robust variance estimation for associations of a 5 μg/m3 increase in PM2.5 and 10 μg/m3 increase in NO2 with urological cancer risk: main analyses, sensitivity analyses (SA), and population attributable fractions (PAF)
| Pollutant | Meta-analysis | n association estimates | RR (95%CI) | I2(%) | Heterogeneity p-value ^ |
|---|---|---|---|---|---|
| PM2.5 | Main analysis uncorrected for publication bias | 41 | 1.06 (1.03, 1.10) | 52.36 | <0.001 |
| Main analysis corrected for publication bias¶ | 46 | 1.06 (1.02, 1.09) | 51.88 | <0.001 | |
| Sensitivity analyses (SA) | |||||
| SA.1 Leave-one-out meta-analysis$ | 40 | 1.07 (1.04, 1.10) | 44.70 | <0.001 | |
| SA.2 Restricted to populations with smoking adjustment | 25 | 1.06 (1.02, 1.10) | 44.43 | 0.012 | |
| SA.3 Restricted to quality assessment score ≥6 | 30 | 1.05 (1.02, 1.08) | 52.26 | 0.002 | |
| SA.4 Restricted to studies with exposure assessment based on LUR modelling | 22 | 1.05 (1.01, 1.09) | 47.89 | 0.014 | |
| SA.5 Restricted to studies published in 2020 or later | 28 | 1.06 (1.01, 1.11) | 52.13 | <0.001 | |
| PAF, % (95%CI)* | |||||
| k = 100% | 41 | 5.91 (3.61, 8.16) | -- | -- | |
| NO2 | Main analysis uncorrected for publication bias | 28 | 1.03 (1.00, 1.07) | 22.26 | 0.039 |
| Main analysis corrected for publication bias¶ | 30 | 1.03 (1.01, 1.05) | 20.51 | 0.026 | |
| Sensitivity analyses (SA) | |||||
| SA.1 Leave-one-out meta-analysis$ | 27 | 1.02 (1.00,1.05) | 8.46 | 0.168 | |
| SA.2 Restricted to populations with smoking adjustment | 23 | 1.05 (0.98, 1.12) | 19.81 | 0.070 | |
| SA.3 Restricted to quality assessment score ≥6 | 28 | 1.03 (1.00, 1.07) | 22.26 | 0.039 | |
| SA 4. Restricted to studies with exposure assessment based on LUR modelling | 24 | 1.02 (0.98,1.05) | 0.06 | 0.316 | |
| SA.5 Restricted to studies published in 2020 or later | 7 | 1.02 (0.98,1.05) | 0.00 | 0.448 | |
| PAF, % (95%CI) * | |||||
| k = 100% | 28 | 3.05 (0.51, 5.50) | -- | -- | |
Notes:
¶ Estimates are from trim-and-fill analysis without robust variance.
$ Estimates are from the meta-analysis that excluded the study that contributed most to heterogeneity (PM2.5: Taj 2022 testicular cancer, NO2: Gandini 2018 kidney cancer) by leave-one-out meta-analyses.
*PAF quantified the proportion of all urologic cancers that are attributable to a 5 μg/m3 increase in PM2.5 or a 10 μg/m3 increase in NO2. We assumed the prevalence of air pollution k = 100% and PAF = (RR-1)/RR. 95%CI was calculated by bootstrap method.
^All statistical tests are two-sided.
CI confidence interval, KCa kidney cancer, NO2 nitrogen dioxide, RR relative risk, PM2.5, fine inhalable particles, with diameters that are generally 2.5 micrometers and smaller.