These results coincide with univariate analyses (which were not shown here) based on linear mixed effects models when a single covariate effect is considered, adjusted for time trends and baseline scores.
Our results, therefore, suggest that studies of air pollution and health should consider adjusting for SES, especially NSES.
If differences are found in baseline characteristics, we will consider adjusting for these variables in the efficacy analysis.
We did not consider adjusting for baseline covariates in the analysis, which may improve the estimation of the treatment effect.
If there is an imbalance in the two treatments groups we shall further consider adjusting for gender, age, cancer stage, diagnosis, co-morbidity, treatment status and education which may have a prognostic value [ 17, 18].
We did not consider adjusting for variables such as time since diagnosis, previous surgery, occurrence of metastasis, or chemotherapy regimen, as these variables reflect genuine characteristics of the chemotherapy line.
In multivariable analyses, several models were considered – adjusted only for age and gender and adjusted for all the variables significantly associated with weight gain.
Prior analyses found small to moderate increases in risk among highly exposed women with exposures above the 75th and 90th percentile when 0-15 years latency was considered (adjusted ORs 1.6-1.9 1.6-1.9th and adjusted ORs 1.3-1.9 for >75th, Tande 5).
The study sponsors should consider adjusting their design for that purpose.
