Non-linear transformations of age at diagnosis, tumor size, and number of positive lymph nodes in prediction of clinical outcome in breast cancer

Table 2 Hazard ratios from analysis of D-RFi for Cox-models with MFP- and RCStransformations of the predictors

For each method of covariate transformation, the prognostic index (PI) in the derivation set was categorized at the 16th, the 50th, and the 84th percentiles forming four risk groups named G1 to G4. The parameter estimates from the derivation set for MFP and RCS were used to calculate the PIs for each patient in the validation set. Each of these two indices was thereafter categorized into four groups using the percentile based cut-off values from the derivation set

ISSN: 1471-2407