Clinical trials are very expensive. Current funding from the NIH for clinical trials is : $7,327,474,432 (7 billion dollars) I do not think we are getting our money's worth.
When I treat a patient, I want to know the best treatment for my patient, not the best treatment for a population of patients who have (what the trialists have called) the same disease, and who fit the criteria for the trial ( i.e. no other major problems). I need to optimize the outcome of the patient as I see her. But the reporting of clinical trials obscures the information I need to make the best decision. The details are sacrificed to an illusion of statistical accuracy.
I have a patient who was treated for "multple myeloma." He has the characteristic abnormal antibody product ( monoclonal spike) resulting from a cancer of plasma cells( the immune cells that produce antibodies), Chromosome analysis demonstrates a translocation (4;14) This chromosomal abnormality identifies a disease that has an inferior outcome with (old) standard therapy. There is evidence that newer treatment ( boretzimib) produces a better outcome. The patient had standard therapy including a stem cell (autologous) transplant. What should the mainteneane therapy be?
The studes demonstrate an advatage for mainetenace therapy, But the results are expressed without regard to the chromosomal abnormality. At best, this chromosomal abnormality is lumped with others that define "risk" But risk is a function of treatment!. To make maters worse, the risk classification of his chromosomal abnormality varies among authors! Some call it high, other intermediate.
If the raw data were available, if the clinician could aggregate and analize the raw data, and then submit the data on his patient into the pool, we would have afar more efficient and productive process. Instead of basing decisions upon random data, we would base decisions upon outcomes that were also based upon constant correction, The search for optimization would be organic, ever improving.. Conceptually this approaches grows out of Bayes ( using the pretest assumption) and Hayek ( the wisdom of the market). This is an optimization algorithm, Why shouldn't each successive treatment attempt be corrected for the past treatment attempts?
