Table 1

 Selected predictors of 12 month survival for patients with kidney cancer. Estimated mutually adjusted regression coefficient (standard error) for three multivariable models obtained using different strategies to select variables (see text)

Predictor*Full modelAkaike information criterion5% significance level
WHO performance status 1 (versus 0)0.62 (0.30)0.55 (0.29)0.50 (0.2)
WHO performance status 2 (versus 0)1.69 (0.42) 1.62 (0.41)1.55 (0.40)
Haemoglobin (g/l)−0.45 (0.08)−0.44 (0.08)−0.39 (0.08)
White cell count (×109/l)0.12 (0.05)0.13 (0.05)0.13 (0.05)
Transformed time from diagnosis of metastatic disease to randomisation†−0.29 (0.10)−0.30 (0.10)−0.27 (0.09)
Interferon treatment−0.61 (0.27)−0.61 (0.26)−0.58 (0.26)
Nephrectomy0.39 (0.29)0.44 (0.28)
Female sex−0.57 (0.29)−0.56 (0.28)
Lung metastasis−0.36 (0.28)
Age (per 10 year)−0.07 (0.13)
Multiple sites of metastasis−0.09 (0.36)
Intercept6.54 (1.63)5.70 (1.29)4.99 (1.22)
C index0.80 (0.02)0.80 (0.02)0.79 (0.02)

We assumed linear effects of continuous predictors. Details of the distribution of each candidate predictor have been omitted to save space.

*Binary variables are coded 0 for no, 1 for yes.

†log(days from metastasis to randomisation + 1).