and associates. These are papers that raise important questions in a developing field, and that put forth answers to those questions that will help us move forward in research on HIV/AIDS. I will proceed in reverse order. Kinley Larntz called some things to my remembrance from my graduate work in statistics in an agricultural setting. The work on design of experiments emphasized factorial designs followed by fractional factorials and lattices. It was an era when computing power was limited, so it was important to have balanced designs.
Later, as a faculty member, I taught similar courses in design and did many analyses of variance as a statistical consultant. As I became more involved with medical and public health studies, I changed the design course. Designs with several treatments and complex layouts were not being used in medicine. There are good reasons for this.
What is different about medical studies than agricultural field experiments that leads to simple designs, usually with just two treatment regimens? Although the designs in clinical trials may be simple in terms of the randomization of treatments to subjects, the responses are complex. Kinley mentioned that. He said that you may have many different endpoints. There is always the tendency in medical studies to collect a great deal of information on every subject or every specimen. The responses are multi-dimensional and complex. That is one reason why we want designs for the allocation of treatment regimens to patients to be simple.
Other reasons arise from the context in which many clinical trials are done. This is a context in which approval is sought for a new treatment, perhaps a drug therapy, where there has been considerable background work. A lot is known about the drug and its use in animals. There also have been dosing studies and studies of adverse effects in humans, so that at the point where a clinical trial starts, the question to be investigated has been narrowed to the efficacy of a single new treatment. Seldom is more than one new treatment under consideration in a major clinical trial since indifference between treatments is required. This ethical requirement is unlikely to be met for multiple new treatments in the same trial.
There is a need to restrict the subjects in a clinical trial to patients who are apt to benefit from treatment, who are not using concomitant medications, and do not have other medical complications. The reason is that a definitive result is sought; the purpose is to show unequivocally that the new treatment is efficacious. Afterwards there may be follow-up where the findings are extended to a larger patient group or to related drugs.
So in traditional clinical trials, there are good reasons to use simple designs and to not use factorials or more complex designs with multiple treatment combinations. However, the setting that Kinley Larntz and his colleagues are dealing with is quite different. The experimentation is