In the context of biomarkers, antibodies fall into four main categories, each of which carries a different level of risk from an economic development point of view (Fig. 1) . Those markers used for routine diagnostic purposes entail little or no risk, and prognostic markers involve high risk. Predictive markers associated with established therapies are low risk whereas those for new therapies involve high risk. The focus of the presentation was on development of antibodies for detecting cancer

more »

... iomarkers. It is crucial to use prospectively defined criteria to select patients who are most likely to respond to a specific molecularly targeted therapy. Proper patient selection enables efficient clinical trial design for targeted therapies and ensures that the number of individuals exposed to the risks of anticancer therapy is minimised. Patient selection can be facilitated through the use of systems, such as pharmDx, Dako's complete diagnostic assays that enable selection of patients more likely to benefit from targeted therapy. Herceptest Ò was the first such system developed. It is used to identify patients whose tumours overexpress Her-2/ERB2 and, therefore, who would be mostly likely to respond to treatment with trastuzumab (Herceptin Ò ), a humanised antibody targeting the HER-2 receptor. By screening with the pharmDx system, the response rate is greater than if the general patient population were treated with trastuzumab. Semiquantitative scaling was used for registration of the pharmDx technique and is the basis for its labelling. The quality of antibodies under development in terms of sensitivity and specificity is extremely important. Antibodies can be 12 E J C S U P P L E M E N T S 5 ( 2 0 0 7 ) 1 0 -1 7 non-small-cell lung cancer (NSCLC). Data from early-phase trials did not show a clear correlation between patient outcome and EGFR expression in archived tissue. 1 Subsequently, however, data emerged indicating that EGFR mutations and increased gene copy number, as measured by fluorescence in situ hybridisation (FISH) are associated with clinical response to gefitinib treatment. [2][3][4][5] Other potential biomarkers of gefitinib outcome have also been identified. One of the challenges in the development of gefitinib was that knowledge of potential biomarkers emerged during the conduct of the pivotal trials. Indeed, increased EGFR gene copy number measured by FISH was shown in 2003 to be a prognostic biomarker for outcome after surgery in patients with NSCLC, 3 and subsequently shown to be predictive of response to gefitinib. 4 In 2004, EGFR mutations also emerged as predictors of response to EGFR TKIs in patients with advanced NSCLC. 5 Evolution of biomarkers during the conduct of large randomised trials might become the rule rather than the exception. Although initial candidate biomarkers are evaluated early in development, knowledge increases exponentially as research and clinical experience become more widespread and increased clinical data with which to correlate the translational work becomes available. THE CRITICAL IMPORTANCE OF TISSUE SAMPLES: The IRESSA Survival Evaluation in Lung Cancer (ISEL) phase III trial highlighted many of the challenges in acquiring tissue samples in large multinational randomised phase III trials. The phase I studies of gefitinib involved collaboration of just a few academic centres that were very devoted to collecting tissues. In phase II, 40 centres were involved, but the sample acquisition rate dropped to 80%. The ISEL phase III study accrued very quickly worldwide, but only 33% of patients' samples were available (Fig. 1 ). Of these, 177 samples were evaluable for all three of the following biomarkers: FISH, EGFR expression and EGFR mutations. Dr. Botwood outlined the challenges encountered in collecting tissues for such studies. In ISEL, more than 25% of tissue samples were inadequate (insufficient quantity or fixation) for any sort of analysis, more than 60% were inadequate for mutational studies, and 80% required remounting. Documentation of samples also proved to be quite challenging as many were incorrectly labelled and could not be validated. The informed consent process also presented some challenges with fewer than 40% of patients consenting to tissue sampling overall. Local changes to, and interpretation of, the consent by ethics committees internationally also meant that it was not possible to analyse all available samples from all countries.