Personal Science and Other Things
Ian Eslick of Vital Reactor presents a short talk about personal data and the scientific process. Access to the methods of science and the scientific process is an important piece of the puzzle, especially as personal data becomes easily captured and more readily understood. Too often, the world of science and research is held up on a pedestal, out of reach for individuals struggling to understand themselves. In this talk, Ian discusses about his personal journey of self-experimentation and how access to the “tools of science” can be highly impactful especially for those battling chronic conditions.
pen and paper
Gary Wolf: Ian Eslick is going to talk to us about personal science, and get us started.
Personal science and other things. So Gary and Nester has asked me to talk a little bit a personal story that got me into this landscape.
About a decade ago I started experiencing some symptoms, maybe not unlike Larry’s story and sadly I didn’t end up where Larry has ended up with a much deeper understanding of myself, but it’s a work in progress.
But about five years ago, Gary and I had good fortune and sit next to each other at a conferences, and I was talking about the kind of self-tracking that I did which involved never writing anything down. But what I developed was a good mental arithmetic for keeping track of variations in my daily life that deviated from baseline that I knew that kept me healthy and so we call these triggers often.
And so what I would do in a given week there’s only three or four things that I do and my discipline slips and I end up with a little natural experiment. So now I say I have a rise in symptoms. They might be GI symptoms, they might be fatigue, sometime blurry vision, all sort of interesting things that happen, all generally related to the kinds of things that people with autoimmune disease have. And I do have a very mild autoimmune disease.
And the side effects are something that’s understood, but not researches. People researched the primary manifestations with the disease but not the phenomenology of it. So I’ve had to learn this all on my own and it took about 10 year, and that 10 years and someday I’ll have my wife tell the story from her perspective, but it is a tremendous amount of suffering that you can hide day to day. Nobody knows, you can still function and you can perform, but life is not fun.
So one of the questions after talking to Gary that really got me into this is how can I turn that eight years of ad hoc discovery, which involved random conversations of a sand dune off the coast of Brisbane Australia with a woman with celiac, that mentioned something called fructose malabsorption, which in 10 years of internet surfing that I never encountered. And that was critical breakthroughs that lead me down a line to this investigation that why were eggplants bothering me, but not broccoli.
At the same time, about 2009 I got involved as a caretaker in the healthcare system; I’ve a relative with advanced cancer. And so for five years we’ve gone through 30 – 40 different consults. And at firsthand I’ve seen at the degree to which all of the wonderful knowledge generated in healthcare is so poorly translated into the decision making process’s that a patient is forced to go through. When you have to adjudicate 15, 20, 30 different encounters and there’s no longitudinal history maintained by the ecosystem as a whole.
So it started me thinking of the provenance of data and how decisions are made, and how we make personal decisions, how healthcare providers make decisions. And in the process of doing a mid-career doctorate during this period, I got focused on the question, well what can we do to improve the decisions we make on our own. And since I was working with a researchers, how can the data and the evidence of the experience that I have be shared with my peers and be shared with the healthcare system in ways that actually improve everybody’s participation in the generation of knowledge.
So I focused particularly on the idea of a self-experiment, and working with populations around, how might we design a self-experiment and what can people understand, not understand and what can they engage with.
And all of this goes down to a kind of a fundamental thinking of what is science. So when we talk about science in healthcare, I think there’s a major that this means double blind, placebo controlled, clinical trialed, peer reviewed published. And that is a wonderful mechanism for enabling us to generate a certain kind of knowledge.
But science is really about repeatability, about process, about discipline, about characterization about controlling noise and there’s lots of different mechanisms that we can pull together to tell a story or to inform a decision.
So in that case I don’t really care about causality. I just care about whether I reliably feel better, and that is a probabilistic decision. It doesn’t require cause or knowledge. It doesn’t require the bios and the odds in the right way, and I’ll talk a little bit more of this in a second.
So the science we’re all taught in school is about the hypothesis testing, “I’ve got a good idea, and now I’m going to test it. I’m going to spend millions of dollars, and months and years to try and improve this drug effects this outcome by a significant margin.”
But we don’t talk a lot about the design side of science. Where do these hypotheses come from in the first place? And that is something that when I started in this work is, how can patients participate in hypothesis generation.
And I would have really interesting discussions with researchers, and I remember one was like, Well, patients don’t really know. Because they will come in and say I’m doing horrible but their lab values are really great. Or they’ll come in and say gosh, I feel awful. But they are really improving. Or I have horrible symptoms, but I know you’re a stable patient.
So the definition of what the outcomes and objectives are I think is one area of real challenge. And as we start talking of this idea of patients being involved, that’s when I met Michael and Peter Margolis at Cincinnati, where they were developing something called the Collaborative Chronic Care Network. And in this network they were trying to take patient’s interests on the primary outcome, part of the triple bottom line.
And so a patient would come in and they’re stable, and they have horrible abdominal pains and they’re going to the bathroom seven or eight times a night. Well is that a successful outcome? Well from a clinical measured standpoint, they get the gold star because they’re in clinical remission, but they are not feeling good.
But the problem is there is no science on that particular phenomenology. There’s no literature, no journal results to tell you whether the probiotics are going to affect this, what lifestyle factors are going to affect it, and how to investigate it.
So one of the subprojects we’ve worked on over the last couple of years is called personalized learning. It’s how do we capture, characterize, and learn from and observe variations and symptoms of interest to it to a patient. And the idea is that by looking for special causation surprises, it’s the eggplant not the broccoli, what does that imply?
That we can systematically start to generate, and then test hypothesis for that particular patient, and we’re just a the cusp I think of starting to add all of these things up and saying, well know I’ve seen five, six, seven patients we can learn more about how to design the experiments better.
And that’s where I think it starts to get interesting and we can think about taking the different tools of science and adding them all up for example, controlling for confounders. Well if we have learned from prior studies that it’s a large affect if it’s going to work at all, and it’s going to be big and there are really no confounders that are significant. Then controlling for that noise is a waste of time. You either get better or you don’t, and move onto the next thing.
If you really know what the confounders are then you can control for them. So in my case I have a bad habit of all the way from college of being bad about my sleep habits. And so I was measuring fatigue as a function of diet, and what I discovered is I couldn’t differentiate for a long time, diets vary and induce fatigue and lack of sleep fatigue.
And so I actually did a controlled experiment where I spent about a month just trying to sleep eight hours every night, and it was profound realization of how much of my misery was bad habits.
So I ran a study and I had people describing similar things to me. So I said, I challenge you for one week, get eight hours a night as close as you can for one week. And three of the four people who had sleep issues who tried it, said you changed my life. I’ve been miserable for four year, large affects size.
And so when I think about personal science, we can fail soft and it’s okay to be wrong for a while because people aren’t dying right, we can learn. The question is are we being embedded in a process of peer review and replication and different components of the scientific apparatus that take that wrong initial assumption and the arch of science bends towards the truth.
And I think we can do that in the hinterland between the kind of ad hoc and hey, I tried this and it worked and hey I’ve got a double blind placebo controlled trial. So I would love to talk more about that continuum but I can see that the red light is about to hit me.