The Culprit Behind Disease: It's More Tricky Than You Think

In my last post, I touched on a few of the dramatic differences - inequalities -  in the health status of the U.S. population. I promised that I'd begin shedding some light on the causes of these health inequalities. And I will.  But in order to do so properly, I'll need to step back for a moment and talk about a crucial epidemiologic principle: causality. 


A natural human tendency is to side with personal anecdote over data-driven automaticity.  The power of emotion comes through in a news clip depicting a mother with an autistic child, weeping as she describes taking him to routine immunizations like a responsible parent.

 Data-driven automaticity, on the other hand, is text in a journal article, sometimes translated on airwaves through the remarkably monotone voice of a bespectacled gray-haired gentleman in an imposing white coat, who uses phrases like "The hypothesis of association appears poorly justified, with reasonable certainty, given our existing body of evidence" to say things like "No."

In the eyes of the general population, it's no contest.*

(* The hyperlink opens to an article which covers a recent grand preventive medicine snafu in translating an "evidence-grounded" message for the general public:  the mammogram screening controversy. Communication problems partially furthered the general public outcry that followed new mammogram screening guidelines calling for raising the age when women should start getting screened.)

The power of emotion and anecdote is so great as to render former Playboy models reater authorities on neurobiology than the hyper-educated, bespectacled physician. Despite a lack of scientific connection between vaccines and autism, emotion fuels a sizeable and growing group of fervently anti-vaccine parents. (Whose impact, by the way, is already evident in places like Marin County, dealing with a sudden surge in kids coming down with preventable whooping cough and diphtheria.)


The science of epidemiology, unfortunately, may partially be responsible for the case of the doctor who seems to have a knee-jerk tendency to qualify and add caveats to every recommendation.  A truth of life is that the more you delve deeper into learning something, the less you realize is absolutely certain. (Besides death, I suppose, and taxes.) Epidemiology takes this to the hilt: take a few classes in the subject and you feel as though a joyriding teen just vandalized your brain. Suddenly, everything you thought you knew, and everything you thought you knew about knowing how to know, is suspect.  (See what I mean? I've just been studying epidemiology, and that's a classic post-epidemiology-studying statement.) Doctors' seemingly cautious, throat-clearing flourishes simply reflect belief in the limits of scientific analysis. (Or, for the cynics among us, fear of getting sued.)

This article takes it to the hilt, in true epidemiology style: can we really believe most of what we find in evidence based medicine?

As irritatingly exhausting as it is, epidemiology is also incredible. It combines everything you learned in grade school about the "scientific method" - forming and testing a hypothesis, performing experiments, analyzing results - with the thinking calisthenics of philosophy and logic, the liberal arts perspective of history and sociology, and the formula-based precision of calculus, accounting and statistics.  You can see why it's exhausting; almost as exhausting as writing that sentence.

At the core of epidemiology is piecing apart the puzzle of what causes disease and what prevents disease.

It's a puzzle that most of us have pondered at one time or another. You hear of the tragic situation of an avid cell-phone user who is diagnosed with brain cancer, for example, and wonder whether cell phones cause cancer. Or, you get a call from Aunt Mildred, who just got the flu shot and now feels horrible, like she's getting the flu. She's convinced the flu shot "caused" her to get even sicker.

A patient in San Francisco may buy a bottle of kombucha after a friend raves about its energy and health benefits (not to mention an enticing label boasting of potential benefits for everything from digestive health to warding off cancer). A Berkeley patient, on the other hand, might avoid peanuts on Tuesday, fast on kale juice on Wednesday and otherwise "is gluten/soy/dairy-free because everything else causes my system to be messed up."

We make associations all the time. But isolating whether X caused Y  is a trickier proposition. It respects the fact that sometimes, things just happen to coexist by random, non-causal chance. "Causality", though, requires more than one isolated person (or even a few people) happening to have a certain outcome when they have a specific "exposure" - whether a flu shot or cell phone.

 And even when you can show two things are associated  - ie, more cell phone users tend to have brain cancer - does not mean that the link is causal, i.e. that cell phone use causes brain cancer.

Our classic go-to example is matches and lung cancer. If you did a study where you tracked rates of matches and rates of lung cancer in a population, you could quickly see an association between numbers of matches and cancer. It could look quite convincing, with rising numbers of matches linked to higher cancer rates. You could probably repeat the study in many different settings and come up with the same result.

But matches aren't the cause of lung cancer. The culprit is what the matches are used for - i.e., smoking cigarettes, which we commonly accept promotes lung cancer. (Bear with me here and pretend that cigarette lighters haven't been invented yet, so people use matches to light up. You could do the same thing with "lighter use" if you wanted.)

It's quickly apparent how carefully teasing apart the cause of an outcome is crucial to effective policy. Even though matches were associated with cancer in our study, calling for a policy that bans matches would hardly target the root cause of cancer. And it would negatively affect people who use matches for other things, like moms lighting candles on birthday cakes, or overachieving eighth-graders reproducing that authentic yellowed hue of 18th-century parchment paper for a history project on the Declaration of Indepdence.

Epidemiology lays out how you can take data that tracks exposures (ie, matches) and outcomes (ie, cancer) for a large number of people and analyze it to see if 1) there actually is a "real", non-chance kind of link and 2) if  that link could be causal. It's the holy grail of the science: trying to figure out if a curious pattern we perceive or wonder about is actually real on a population scale. In other words - does Aunt Millie's experience with that flu shot translate to a "real" side effect seen by the population? Is there evidence that the flu shot can "cause" the flu in people?  (For the record, the answer is no.  Get  your flu shot!)

An inherent conflict between science and our mind: human beings naturally "want" the world to follow intuitive observations. And sometimes, intuition is in fact the way the world works. But sometimes, science doesn't follow our logic.  After all, the earth is round and rotates around the sun - even though the sun seems to circle above us, rising and setting on a fixed, flat horizon.

That our observations may not necessarily explain the full picture holds with the realm of causality as well. Take the following example. The circle is what epidemiologists call a "causal pie": it represents the component causes of a disease. Each thing in the circle has to happen for a person to get the disease. If just one piece is "prevented", the person won't get the disease.

But there's a catch: for most diseases, no-one knows what all the pieces of the pie actually are. Our best guess at the pie's structure come from our basic knowledge, scientific experiments and observations - a collection of inputs based on theories and the patterns we see.

For example, if the disease is being overweight, a very simple model might say that A represents a gene for overweight, B is overeating, and C is not exercising.

Now, let's say that in Population 1, everyone has a gene for overweight (A) and everyone overeats (B). About 10% of people don't exercise (C) - and are therefore overweight, since they have A, B, and C and complete the pie. Thus, it's observed that the 10% of the population who are overweight ahappen to be non-exercisers. It's the only thing that seems to differentiate the overweight group from the rest.

 So what would an observer logically conclude about this population? Being overweight a behavioral thing - it's caused mainly by not exercising. Exercise seems like the "magic switch" that "prevents" and "cures" overweight.

Now take another population. In this population, everyone is a non-exerciser (they have C) and everyone overeats (B). 10% of the population has a gene for being overweight - and thus, this 10% ends up becoming overweight, with pieces A,B, and C completed.

 Then, scientists come up with a miracle drug that blocks the gene and give it to overweight people, and voila!  The drug cures weight gain in 10% of the population.

In this population, the magic switch seems to be a gene - the only thing that seems to differentiate the overweight group from the rest. What do people in this population conclude? Being overweight is genetic - it's caused by a gene, not behavior.

In reality, preventing any one of the pie pieces could have prevented overweight effectively. In fact, if a smart scientist somehow knew that overeating - something that no-one thought -  caused overweight, and somehow was able to prevent people from overeating  - then he could prevent everyone in both populations from being overweight. This intervention would have a larger scale impact in both populations.

Obviously, real life is much more complex. But the point of all this holds: When thinking about what causes a disease, the pie pieces that are the most common causes in a population are often the least recognized.   On the other hand, the pieces that are present in a relatively small percent of people are quickly identified as the "magic switch", and are often the focus of prevention efforts.

Sometimes, these observably different pieces indeed represent the best - and most "do-able" - approach to prevent disease, given our incomplete knowledge of the pie's full structure. But sometimes, the prevention strategy with largest potential impact rests on undiscovered, commonly present pie pieces - those that affect most of the population. Finding these pieces requires thinking outside the box.

 For example, some scientists are starting to wonder whether being overweight is linked to a commonly present virus. (To see how this could happen, replace the "overeating" pie piece B with "virus".) Mind-boggling, no?

There you have it - a quick intro to causality. Next post- as promised - applying causality to a practical situation: health disparities and health outcomes.