Bespoke labs

A graduate student complained to me about his advisor. "He treats me and another graduate student totally differently! With her he's hands off, but with me he micromanages!"

A faculty member commented to me, "One of the hardest things I had to learn as a PI was how to recognize that everyone in my lab needs different things from me because of their own strengths and weaknesses, so I need to tailor my style to their needs."

Who's right?

I'm sympathetic to the faculty member, because it's quite true that different trainees have differing strengths. Some are good at experiments but bad at writing; others are good at thinking about science but bad at benchwork; others can't give a presentation worth a damn, regardless how good their data are. So the various lab members really do need different kinds of guidance from their mentor.

However, what I think I hear in the graduate student's complaint is not, in fact, that he is angry about differential treatment. He is angry that he perceives the differential treatment to reveal something about his advisor's opinion: namely, that the advisor respects the other graduate student more (leading to hands-off mentorship), and doesn't trust the male graduate student enough to let him conduct his own experiments without significant input from the PI.

(Of course, it's possible that the female grad student in this example is peeved that the male grad student gets so much attention, while she is neglected. Grass is always greener etc.)

The point is, not only should the PI ideally tailor his or her mentoring strategies to meet the different needs of the trainees, it must be clear that this tailoring is not favoritism, which is the tailoring to the benefit of one trainee over the others.

This is one of the many reasons that management looks hard to me.

They are carbon-based

Our mouse colony has expanded so rapidly that we've hit a space crunch despite getting one increase from the space committee already. Each person in lab "owns" a certain set of cages, and so we sat down to come up with a system to keep cage usage under certain limits.

The result? We each have a fixed maximum number of cages for our mice, but we can loan those spaces to others if we're not using them.

What do we call this system? Cap and trade.

Best practices

Dr Hyde's PI attended grad school ~40 years ago. He was taught a list of ten or so "fundamental concepts" in neuroscience. It ran something like:

1. Dale's Law--each neuron releases only one type of neurotransmitter.
2. No new neurons are born in the adult brain.
3. Inputs to the neuron occur only on the dendrites and soma.
4. Outputs from the neuron occur only from axonal boutons.
5. Action potentials are always initiated in the soma or axon.
6. Synapses are unidirectional; there is no retrograde signaling.
7. The brain is immune privileged.

Those of you who are neuroscientists are laughing, because none of these things are true. They were all reasonable approximations to the truth, based on the best available evidence at the time--but they've all been convincingly overturned.

With this context, how to think about the move towards more evidence-based medicine? The Obama administration is making an effort to decide upon "best practices" in medicine as evidenced by research, and then to nudge doctors towards using those "best practices" by a series of governmental carrots or sticks.

Dr. Jerome Groopman discusses the potential pitfalls to this idea in this month's New York Review of Books. Not only is the literature often (usually?) wrong, it is also inadequate. Clinical studies are frequently restricted to patients with only one condition, or to selected subsets of patients (most glaringly, men); whereas in reality, many people have multiple ailments at a given time, and women, or poor people, or underrepresented minorities, have an irritating propensity to respond differently to treatments than do the more homogeneous clinical populations.

I'm not a doctor, but as a researcher I know how much literature is going to be disproven a few years down the road. Furthermore, basic science research is usually far superior in quality to clinical trials, in large part due to the laboratory setting where one can actually hold all variables constant except for the one under study--far from true in a clinical setting.

But at the same time, why do we even bother trying to figure out the best way to treat a disease if we're not going to use that information? I support Groopman's plea that doctors should have significant leniency to pursue the course of treatment that seems best to them, even if it deviates from "best practices" as currently evaluated.

But I'll go one step further (and this is a naive suggestion from someone who knows next to nothing about medicine, really): the number and type of clinical trials should be vastly expanded, by virtue of keeping tabs on the success of procedures at clinics and hospitals all over the country. We already do this with some types of procedures (close to my heart is sart.org, which tracks IVF success at different locations), but we aren't going far enough. We should be tagging all these outcomes with detailed patient information, ideally using a unique identifier for each patient that allows us to track patients wherever they go.

Then, put the army of supernumerary biomedical Ph.D.s to work. They can easily learn how to sift through these massive datasets to identify what treatments work for 35-year-old white men with elevated blood pressure but not for 60-year-old Asian women with fallen arches.

Effectively, clinical trials designed by one or a few doctors, occurring in a small number of patients, would be a thing of the past. The dreadful meta-analysis of twelve crappy published studies, which is what we currently rely on to devise most "best practices," would vanish.

There are probably a lot of reasons, including patient privacy, why this can't ever happen. But it would completely transform biomedical research. Basic scientists mostly look down their noses, and rightfully so, at the quality of clinical trials--but this approach would give clinical trials a sample size far beyond even the most ambitious basic science research. Rigorous researchers might actually want to be involved with medical science.

Especially now that we've come up with a whole new set of fundamental principles in neuroscience, and this time they're definitely right.