Baseline variables for the study groups are displayed in Table 1.
Before analysis, we assessed the baseline variables for the presence of multicollinearity.
The extent of missing data and drop out in the cancer group was analysed by examining baseline variables for the 'completers' and 'non-completers' to identify possible biases.
Predefined baseline variables for the univariate analysis were: sex, age ≥ 75 years, WHO performance status < 2, initial weight loss < 10%, Charlson score ≤ 1, histology, tumor stage and treatment.
Nine pre-defined baseline variables for the univariate analysis were sex, age, WHO performance status <2, initial weight loss <10%, albumine ⩾30 g l−1, Charlson score ⩽2, dose of cisplatin ⩾80%, dose of radiotherapy ⩾80%, and CCR to CRT.
This index also ranged from 0 to 6. Percentages or mean and standard deviations, as appropriate, were computed by group for baseline variables for the women interviewed, by arm, and checked for imbalance between groups.
To test the potential effect of dropouts, we compared the baseline variables for those who failed to complete the 6-month questionnaires with those who had completed questionnaires, utilizing t tests.
Moreover, multivariate Cox analyses were used to determine the significance of the baseline variables for predicting the primary endpoint during the study period.
The final model controlled for age, sex and baseline variable for the dependent outcome, regardless of the significance in the univariate analysis.
A comparison will be made of the baseline variables for each of the groups, to test the homogeneity obtained from the random allocation to groups, in terms of differences of the means and of proportions.
For the effect of baseline variables for time to death and on the risk of infection simple Cox models were calculated.
