Tuesday, April 03, 2007

April 3, 2007 - Lies, Damned Lies and Statistics

The Elizabeth Edwards and Tony Snow stories seem to be setting off an interesting debate, indeed. They’ve certainly catapulted cancer into the national consciousness.

A friend just sent me a link to a Washington Post article by one of that paper’s staff writers, David Brown: “For Cancer Patients, A Struggle to Prolong Hope as Well as Life.” Brown takes issue with some claims that have been made about Tony and Elizabeth’s respective prognoses.

President Bush said his press secretary “is not going to let this whip him.” John Edwards said of his wife that her disease – while incurable – may, in fact, be more like diabetes, in that it can successfully be managed for a long time. Brown questions the veracity of such comments, based on survival statistics for relapsed colon and breast cancer patients. When it comes to long-term survival, he claims, the odds are stacked heavily against both Tony and Elizabeth – although he also admits that statistics for relapsed patients are harder to find, in online and other reference sources, than statistics for patients receiving front-line treatments.

He seems to hint that the relative obscurity of the statistics is the result of some kind of dark cover-up on the part of modern medicine – that oncologists know perfectly well that most of their relapsed patients aren’t long for this world, but just aren’t telling. He goes on to cite the results of a research study that suggests doctors tend to be overly optimistic about their patients’ chances, but that patients are even more so: “A 2001 study asked 71 patients awaiting stem-cell transplants about their chances of dying from the treatment. On average they guessed 21 percent. Their doctors said 33 percent. Actual mortality was 42 percent.”

The diabetes analogy John Edwards used is a line I’ve actually used about my own prognosis. I got it from my own doctor, who deployed it just after saying – of recent, encouraging advances in Non-Hodgkin Lymphoma treatment – “If you had to choose a form of cancer to have, this is the one to choose.” So, Brown’s article hit me in the gut, in a very personal way (though I do need to remind myself that he’s talking about relapsed patients, a club to which I do not belong).

Brown’s article made me mad, in a way that few other things written about cancer have. It made me mad on behalf of Elizabeth and Tony, and anybody else who’s dealing with advanced cancer treatments. Using the image of a bell curve, he observes that a relatively small number of relapsed cancer patients are in the narrow “tail” of the curve, that stretches out to the farthest limits of the timeline. Most patients want to visualize themselves as being in that part of the curve, he says, and their doctors want to see them there, too.

I guess what bothers me about Brown’s article is that he sees that as a problem. Given the choice between a doctor who - consciously or unconsciously - bumps my chances of survival up a little, and another who reads the statistical averages like they’re a road map of Death Valley, I’ll go for the cockeyed optimist every time. (I’d have a sneaking suspicion that the pessimistic practitioner could give up on me too soon.)

Nineteenth-century British Prime Minister Benjamin Disraeli is noted for saying, “There are lies, damned lies and statistics” (sometimes people attribute this remark to Mark Twain, but Twain himself says he’s quoting Disraeli). I have a feeling that Brown’s use of statistics belongs to the “damned lies” variety. Sure, maybe his numbers are based on reliable averages, but the thing about averages is they’re just that: average.

Let’s say there are three relapsed patients sitting in a doctor’s waiting room. All three have the same form of cancer. Let’s also say that, for this form of cancer, statistics predict that only one relapsed patient out of three will live more than a year. As each of those patients, in turn, walks into the examining room, is the doctor going to prescribe a cutting-edge new therapy for only one of them? Of course not! The doctor doesn’t know which one of the three is going to make it – and, in fact, it’s possible that all three will do well (three being a rather small statistical sample).

By the same token, when President Bush approaches a microphone, he’s not likely to say, “Too bad about Tony, we’re going to miss him.” No, Tony is his friend, so he’s going to call him a fighter. The same is true of John Edwards, who’s not going to scrimp on his tough talk about standing side-by-side with his wife and doing whatever’s necessary.

Back in college, I wrote a senior thesis on Ludwig Wittgen- stein's philosophy of language. One thing that impressed me about this Austrian thinker – specifically, his later writings – was his remarkably flexible, even playful understanding of language and the ways we actually use it. Wittgenstein talks of “language games,” that each have their own rules. In interpreting the meaning of things people say, it’s hard to generalize from one game to the next - and, when we do, we tend to get into trouble.

When President Bush and John Edwards are speaking pugnacious words about the prospects of people they care about, they’re not reading papers at a medical convention. They’re speaking words of love – and love has no place on a bell curve, none whatsoever.


dwilton said...

The biologist Stephen Jay Gould wrote an excellent and very personal essay on this topic. It can be fould at http://cancerguide.org/median_not_msg.html.

Gould emphasizes that it is the statistic that is the abstraction and that the variation from the norm is the reality. We're all different and an average will only tell the general trend, not the outcome in any particular case.

Gould survived his battle with mesothelioma and beat the median 8-month survival rate by 20 years, eventually succombing to a different, unrelated cancer.

Carlos ("Carl") said...

Good article, Dave. He does a great job of reminding us that statistics ought not to be a self-fulfilling prophecy.