Dark energy and epidemiologic theory

Supernova, type 1a CREDIT: NASA, ESA, E. Jullo (JPL/LAM),
P. Natarajan (Yale) and J-P. Kneib (LAM) (space,com)

“When you choose one way out of many, all the ways you don't take are snuffed out like candles, as if they'd never existed.” 
― Philip Pullman, The Amber Spyglass

Start with a basic notion, an essential fact: if I throw a ball up into the air, gravity will cause it to move more slowly and return to my hand (if I’m a good catch…). Apply the theory of gravity to the universe, and you may infer that expansion of the universe (set in motion by the Big Bang) is increasing, but will eventually slow and contract.

Adam Reiss, speaking at Johns Hopkins University

But physicists, like Dr. Adam Reiss at Johns Hopkins -- one of the three 2011 Nobel Physics prizewinners who discovered “dark energy” – studied the light emitted by type 1a supernovae using the Hubble telescope at Hopkins. Instead of proving that expansion is decelerating, they found the opposite. Using two pages from his lab notebook, Dr. Reiss shows the key calculations that led to a great “aha” moment. He and his lab measured the cosmic distances to these star explosions and found that the light they could see through the Hubble did not indicate a slowing of expansion.

Reiss’ team and a separate team of phsyicists found in 1998 that cosmic expansion is accelerating because they could observe that light emitted from supernovae was weaker than expected. In order to account for that acceleration, they theorized that about 75% of the mass-energy content of the universe was made up of a gravitationally repulsive substance that no one had ever identified before – “dark energy.” This left only about 23% of the universe as it exists now (i.e., planets, stars, and gases) as “dark matter.” 

In awarding the Nobel Prize in physics to Reiss and his colleagues, the Nobel Committee stated: “The findings of the 2011 Nobel laureates in Physics have helped to unveil a universe that to a large extent is unknown to science,” noted the Academy, “And everything is possible again.”

What a concept: “…everything is possible again.”

So, Anne: what has this got to do with EPIDEMIOLOGY?

ANSWER: What fascinates me are the unknowns. The creation of new unknowns due to the development and testing of new theory. Epidemiology is a science that has lacked the theory to understand the knowns and the sophistication in theory to discover new unknowns.

The point I am making, however belabored, is that we need to develop better and more novel theories to explain causes of disease and other conditions. We need to stop saying that we don't have enough data to analyze key factors or that we don't have the analytic methods. There are problems, yes, but we need to face the challenges.

In other words: does the supernova (poverty at birth) decelerate over time, remain the same, or does it accelerate? If it decelerates or accelerates, then at what speed does it do so? What variables confound or mediate the relationship between poverty exposure and time?

We know a few things: mosquitos transmit malaria, smoking causes cancer, and poverty is directly associated with poor health. That final association is generally considered to be linear and usually considered to be stable in time. Put simply, the poorer you are, the worse your health outcomes (or greater your health risks, if you prefer). ) I say this association is stable in time because we have not done enough research to know if poverty at birth (or in utero) retains its power as a risk factor at the same level throughout a lifetime, or if it dissipates. If it dissipates –that is, if you are born poor and maybe grow up poor, but you are educated and have income or wealth resources by mid-life, does the early effect of poverty retain it’s power as a risk factor?

So, we could come up with a theoretical equation to test a hypothesis. Living in poverty at a critical developmental period (e.g., ages 0-5) increases risk of preterm delivery, while controlling for demographic and other confounders.  OR, a slightly more difficult hypothesis: As poverty status changes over time, risk for heart disease changes with it, so that there is an inverse relationship between the accumulation of wealth and risk for disease, controlling for environmental and other confounders. What we have real trouble with is observational data. What numbers can we plug in when we don’t do a very good job of measuring socioeconomic status (SES)? Where can we get strong socioeconomic and other demographic data for individuals and a calendar of health. We would have a lot of trouble getting data on outcomes, including degree/intensity of outcomes over time. We do not usually have data available to plug a number into the SES term, the rate of change in SES, whether it is positive or negative, and we don’t have the data to measure the degradation of, say, the cardiovascular system over time.

Hypothetical chart showing rise in wealth over time (blue)
and reduction in disease risk (red)

Why not model it? We have the computing capacity to model a lot of complex environmental, mechanical, and physical problems. We know that if a goose hits a wing engine on a plane, that the engine may seize up, less energy will be delivered, the plane will stop accelerating against gravity, and will begin to decelerate. Soon it will crash into the East River. Because we know the values to insert into all the terms, we also know what will happen if two geese hit the engine, etc. There is still error – we don’t know how the pilot will react – but that error becomes a smaller and smaller part of unknown portion of the equation.

Ω = health outcome

Δt = time elapsed

P = poverty or better yet, SES

ΔP = change in poverty

It would be possible to conduct a retrospective study using interviews with individuals about their lifetime exposure to poverty and their recollection of disease outcomes. Obviously, there is bias in asking people to recall past experiences; medical charts, as available, would provide additional data (although it would be rare to find lifetime charts). 

One possibility (although there are others) would be to create a multilevel, multivariable model using points in time as the first level and individuals as the second level. Obviously, explanatory variables and potential confounders would be identified at the time or individual level, and according to Barbosa & Goldstein, points in time do not have to be equidistant. Working closely with biostatisticians should result in more accurate models than the one posited above. We need to make EVERYTHING POSSIBLE AGAIN in order to figure out the real conundrums of our time.

“Iorek Byrnison: Can is not the same as must.
Lyra Silvertongue: But if you must and you can, then there's no excuse.” 
― Philip Pullman, The Amber Spyglass