Further research on statistical methods for the validation of surrogate markers, as related to endpoint identification, is required. The evaluation of gene therapy or similar cell-mediated therapies provides a unique activity endpoint using gene transfer into the target cells under investigation. Although gene transfer may also be a surrogate marker for clinical effect, evidence of gene transfer may not translate into clinically significant change, although the absence of gene transfer might provide strong evidence against continuing with a proposed therapy. The use of biomarkers as surrogates for safety endpoints is another area for statistical validation. It would be highly desirable to identify biomarkers that could indicate at early stages whether an individual is prone to react adversely to a medical treatment. Genetic factors might provide such information, but phenotypic factors need to be evaluated to validate statistically such surrogate markers for treatment efficacy. The validation needs to be done causally without the no confounding assumptions that current statistical methods make for surrogate marker validation. There also is a need for biostatistics research aimed at evaluating safety endpoints and developing methods to identify potential problems better and more rapidly.

Design and Conduct of Early Stage Clinical Trials
Many currently available statistical methods for clinical trials are directly applicable to new therapeutic approaches in the genomics era but there are also some unique aspects, both ethical and statistical. For example, in a standard "three-plus-three" Phase I trial, a small number of subjects is treated at a "starting dose" and further cohorts treated at a higher, same, or lower dose based on the results in previous cohorts. Many authors have pointed out problems with the standard design, although it remains the most commonly used design in progressive disease trials. The standard design is conservative, inefficient, and imprecise, in that it poorly estimates the true maximum tolerated dose (MTD). Alternative Phase I designs can provide improvements in the number of patients per cohort, the number of cohorts, the probability of observing toxicity, and/or the precision with which the MTD is estimated. Despite these advantages, their use even in cancer research remains limited. There are a number of methodological issues related to design of early stage studies for biological agents, in which the dose-response curve may not be monotonic due to interactions with the immune system. The choice of doses to be used in dose-escalation studies can also be addressed through statistical methods, further improving study design. Finally, the dogma that Phase I clinical trials should focus only on tolerability as the primary endpoint is difficult to apply in rare populations, and methods for simultaneously evaluating safety and efficacy in early stage trials should be made more available in practice.