These results are shown in Tables 2 and 3 (univariable models), and Table 4 (multivariable models).
To decide which variables should be included in the multivariable models, an exploratory analysis was performed by fitting univariable models and considering as candidates for the multivariable model all variables significant at the 0.10 significance level in these univariable models.
The variables tested in univariable models and variables inserted into the multivariable models are listed in Additional file 1.
First, we tested plausible factors in univariable models and, second, we entered significant variables (P <0.20) into a multivariable model (logistic regression).
To decide which variables should be included in this multivariable model, an exploratory analysis was performed by fitting univariable models and considering as candidates for the multivariable model, all variables significant at the 0.15 significance level.
When the number of trials was ≥20, predictors were initially screened in univariable models and then added in multivariable models using a forward-selection process, up to a maximum of one predictor per 10 trials.
For the analyses, we used three models: (1) crude model, (2) univariable model, and (3) multivariable model.
A test for interaction between EGFR status and the treatment arms (Arm A vs Arm C) did not achieve statistical significance in both the univariable model (P-value=0.15) and the multivariable model (P-value=0.15), which could be due to the limited power to detect an interaction.
Univariable models were fit and any variable with a p-value < 0.20 using the likelihood ratio test was included in a full model.
The association was statistically significant for Arm C in both univariable and multivariable models, but not for Arms A and B in either model.
