The antitumor activity of most anticancer agents is limited by a number of different factors, such as their cellular targets and activating enzymes, while constitutive genetic polymorphisms may limit drug bioavailability and influence either antitumor efficacy or toxic side effects. An example of a drug for which clear predictive parameters have been identified is 5-fluorouracil (5FU): its antitumor activity is limited by either a high activity of the target enzyme thymidylate synthase (TS) and/or a high activity of its degrading enzyme, dihydropyrimidine dehydrogenase (DPD). Retrospective studies showed a clear correlation between a high expression of TS and a poor response, which was stronger when DPD was included in the evaluation (high DPD, poor response). Therefore we initiated a clinical prospective study in which we treated previously untreated patients with advanced colorectal cancer with tailored chemotherapy: at a low TS-mRNA and low DPD-mRNA patients were stratified to receive a standard weekly 5FU-leucovorin regimen. At a high TS and/or DPD, patients were stratified to receive a combination of oxaliplatin and irinotecan. Up to now this proof-of-principle study demonstrated that selection of patients is possible and can clearly improve the clinical outcome. The next step is to develop algorithms to select patients for combination chemotherapy with 5FU-leucovorin and new compounds, such as oxaliplatin or irinotecan, or novel targeted agents such as bevacizumab or cetuximab. For these combination schedules the optimal combination of predictive factors has to be explored.