Fig. 1

Summarised results obtained by nuclear magnetic resonance (NMR) metabolomics: a area under the receiver operating characteristics curve (AUC) for the Random Forest (RF) model discriminating 26 early breast cancer (eBC) patients free from cancer recurrence at follow-up and 28 metastatic breast cancer (mBC)-matched patients (training set). The score plotted is the RF risk score that expresses the probability that each sample included in the model has been classified correctly as eBC, or misclassified as mBC. b AUC for RF risk score of the validation set constituted by 7 eBC patients who developed recurrent disease and 54 eBC patients without recurrence. High RF risk score is deemed to represent a high risk of recurrence, as it means that the metabolomic fingerprint of an eBC patient is more closely resembles that of mBC. c Overall eBC patients, plotting actual disease-free survival over time (measured in years) according to estimated metabolomic risk score (Kaplan–Meier curves). “Low” (LR) and “High” (HR) RF risk patients are significantly clustered with a P value of 0.001 (calculated with log-rank test) and a hazard ratio of 14.3. Censored events represent either the time of last recorded clinical follow-up or time of disease recurrence. Timing of recurrent disease events is separately presented in the lower-most risk table