I hope this is the first of a series of posts over the next while, examining how we learn to do medicine well. These issues are large and may take some time. No one understands them well.
I begin with a stunning experiment that has found certain skin cells, with their own unique pattern of gene expression, can be shifted to become an entirely different cell type, an embryonic-like stem cell, by the simultaneous activation of three transcription factors. This new embryonic-like stem cell can differentiate into most of the cell types in your body. The normal skin cell type cannot do this.
Three huge points: 1) We are learning, in systems biology, to control cell differentiation from one cell type, with its stable pattern of gene activities, to another normal, or perhaps not, cell type with a different pattern of gene activities. This must transform medicine. 2) Often, no single transcription factor suffices for these transitions, as in the three factors needed above to shift the skin cell to an embryonic-like stem cell. 3) Therefore, we must give up a simple minded view of linear causal pathways in cell biology, i.e. A -> B -> C ... , to think instead of a vast network of genes and other factors interacting.
Causality is often multifactorial. It is like throwing a small stone on bed springs and watching all the springs vibrate in different patterns as a result.
But how does this connect with the Food and Drug Administration, you ask?
We have been taught to do medical science well by varying a single factor at a time in double blind, random clinical trials across many patients, seeking strong statistical results showing that varying a single factor up and down, on and off, yields a clear clinical result. When random clinical trials work, they really work. We learn a lot.
But, if the causality is multifactorial, as in the three transcription factors needed above to go from skin cell to embryonic-like stem cells, no single factor variation alone will reveal this truth. In this case, random clinical trials — varying a single factor, as if all factors were independent — typically will fail.
The FDA's seemingly wise insistence on single factor, double blind, random clinical trials may result in the throwing away of vast amounts of critical medical and clinical data.
We will have to learn something very new. How do we learn to test multi-causal networks? How do we achieve results that are significant and also allow us to rule out dangerous treatments. This is a big problem, with potential implications for clinical medicine, FDA rules and the insurance industry.
I end by noting that surgeons do not use random clinical trials, but still learn to do surgery better over the years. So random clinical trials are not a prerequisite for medical learning.
We have much to learn.