Predictive tests in cancer chemotherapy. A reappraisal.

The evolution of medical oncology so far owes much to the preclinical and clinical development of antineoplastic agents. Prognostic factors and empiric treatment strategies have guided the clinician in his choice of drugs. In the light of increasing ethical restrictions met with phase I-II clinical trials and major advances in propagating human tumor cells outside the donor patient, a reappraisal of predictive tests in cancer chemotherapy is warranted. Among 'short-term assays' only the determination of steroid-hormone receptor content in tumor tissues has gained clinical acceptance, whereas other methods still suffer from theoretical or practical shortcomings. Both the human tumor stem cell assay and the xenograft model have revealed unique patterns of sensitivity for each individual tumor line. While interindividual heterogeneity among tumors sharing a common site of origin justifies efforts to develop predictive tests, microheterogeneity among tumor samples from the same donor patient limits the potential of this approach. Predictive tests should be performed in conjunction with clinical trials to ensure optimal extraction of information. As additional prognostic factors, they should in the near future accelerate drug development and reduce the hazard of unnecessary drug toxicity without therapeutic benefit.