Friday, July 24, 2015

Overdiagnosis

Starting in the mid-1980s something terrible happened to about a million men. They were diagnosed with prostate cancer, which is pretty terrible, but what's worse is that they were treated for it. It's worse because these men did not need either the diagnosis or the treatment they got. While it is true that they had cancer, it was of a kind that would not develop into a medical condition before they died of some other cause. If they had left it untreated, it would never have bothered them. The treatment, meanwhile—either surgery or radiation—bothered many of them a great deal, leading to both incontinence and impotence. Indeed, about one third of them had complications as a result of treatment. Even just by getting the diagnosis, of course, they were subjected needless worry.

I first came across this very instructive episode in the history (1986-2005) of modern medicine through the work of Gilbert Welch. (His research is featured in this article at WebMD.) The concept of "overdiagnosis" has stuck with me ever since, and has applications, I think, also into the diagnosis of social ills. Indeed, perhaps those who have been reading along these last few days already know where I'm going with this. I think of Tim Hunt as a victim of the overdiagnosis of sexism in science today. In fact, the 2014 interview that has been cited to support a deeper charge of "ingrained" sexism (of which his "joke", then, was just an expression) seems to make the same point.

Before I make the connection, let me just point out a few more features of the PSA overdiagnosis problem. Treating a million men for a cancer that would never have caused them any health problems is, not just inconvenient to them, but very expensive for society. The treatment itself is costly and they work less efficiently while they are undergoing it. Also, we have to ask whether the effect of treating them does anything to bring down the overall level of mortality from cancers. Welch found that, using a conservative estimate, you have to treat around twenty people unnecessarily in order to save one additional person from death by prostate cancer. The reason for this is that mentioned before, namely, that not all cancers develop into what we can properly call a "disease", i.e., a health issue. He quotes (in this video) from George Crile's "Plea Against Blind Fear of Cancer" , who explains that "to say that a patient has a cancer gives as little information about the course of the disease as to say that he has an infection". That is, it may lead to his death if untreated, it may be treated and cured, or it may go away without any treatment whatsoever. Time, sometimes, is a healer.

Cancer is becoming increasingly less dangerous to men, prostate cancer included. This is the result of increased knowledge, of course, including better means of detection and, importantly, better means of treatment. But PSA screening, at least until its ill effects were discovered, proved to be a well-intentioned but ultimately harmful approach. It caused more harm than it prevented.

Consider also, the false positives. These are people whose PSA screenings come back positive, but who are later found not to have cancer. Their lives are turned upside down between the first and the second test. "These men have already taken a hit," as Welch puts it, "They've been told they have a diagnosis of prostate cancer. By the time you are told you have prostate cancer, you are all nervous, you have already lost some sense of well-being. The real issue is, do you want to play this game?" That is, the question is whether you want to get screened. Welch, and the American Cancer Society, no longer recommend it.

In my opinion, Tim Hunt was a false positive "sexist", caught in a regime of overdiagnosis of a very real, but smaller, problem, namely, gender inequality in science. When I say "false positive" I mean that Tim Hunt, in that fateful toast, failed a very sensitive test for sexism, namely, the habit of calling women "girls", that upon further testing would have revealed no actual, practical sexism underneath. He doesn't actually treat women different from men qua scientists and is mindful of their particular problems qua women, such as the possible need for a creche near a lab. Even if you don't quite buy that, then you might grant that his sexism was like a microcancer that would never grow into a full-blown malignancy, a microsexism (to use a suggestive term these days) that would never become the full-blown misogyny he was immediately accused of. Indeed, in all likelihood, Tim Hunt's views about women, like most men his age, had probably mellowed over the last thirty years, owing to increasing contact with intelligent (if sometimes "distractingly sexy") women in the lab. He was harmless and only getting better.

Welch points out that when we introduce new methods of early diagnosis we always discover that a lot more people had the disease than we thought. It's because we're looking much more closely. In what at first seems a paradox, however, a lot fewer people are at the same time dying of them. We certainly have that in the case of women in science. While it would seem that sexism is rampant and on the rise and proliferating in new forms, the amount women in science is increasing, and they are making increasingly successful careers there too. I don't know, but maybe this was also what was on Sir Tim's mind when he answered that interview question for Lab Times in 2014, for which he was accused of claiming that sexism isn't a problem in science. He was simply saying that it's a problem that is naturally going away, and that we may cause more harm than good by trying to "treat" the remaining microsexists that are, on the whole, unlikely to cause any trouble. If we look closely enough, there is a little sexism in everyone, but not only are there better ways to spend our time and resources, there is no need to treat a condition that will not otherwise cause anyone any harm.

To steal some sentences from Gilbert Welch, these men really take a hit. They've been outed as unreconstructed sexists. By the time everyone thinks you're a sexist monster, you are nervous and you have already lost some sense of well-being. That's when the BBC comes looking for you. The real issue is, do you want to play this game? Do we?

9 comments:

Jonathan said...

Here's a story:

http://www.cnn.com/2015/04/13/opinions/williams-ceci-women-in-science/

Steve Fuller said...

Yes and no. The basic idea that a policy ruled by false positives should be dropped is a good and deep one -- and it underscores my own dislike of the precautionary principle, which is way too casual on the false positives that result from its policies. However, your comparison of male prostate cancer and sexism in the academy is flawed -- at least in terms of how you gauge their relative importance -- because you're not comparing like with like. Sexism is a society-wide problem that pervades all male-female relations, as you yourself admit. Moreover, it is a problem that may be bred by the sheer tolerance of its existence. So, if you want to compare sexism to a medical condition, should you think more in terms of a contagious disease, in which the policy of false positive diagnoses seem more justified.

Thomas said...

That's an interesting point. I was mainly just being upfront about where my understanding of the concept of overdiagnosis came from. It may be a bit strained with PSA (but I'm not willing to give it up without thinking a bit more about it.)

I guess you're proposing to compare sexism to something like influenza. But here the point holds even stronger. In an important sense, we have already been "vaccinated" against sexism. Dealing with an isolated "case" like Tim Hunt by quarantining him is excessive and unnecessary. In this case, it was also a false diagnosis, as we now know, which makes interfering with his enjoyment of science an injustice with no benefit.

In the case of an ebola outbreak, of course, you have do something drastic. But not just to one man. And not in the slipshod ad hoc manner that happened here.

Debbie Kennett said...

Thomas

You always have an interesting take on a subject! In this particular case I think your analogy is perhaps not appropriate. The problem with the overdiagnosos of prostate cancer relates to the fact that it is "detected" as a result of screening. It seems counterintuitive but screening is an intervention which has the potential to cause harm as well benefit. If a condition is detected the patient might undergo harmful surgery to combat the problem whereas left untreated, as you say, he might not develop the disease at all and would die of some other unrelated cause. There is a similar problem with breast cancer screening. While it can save lives it also means that many women undergo unnecessary and disfiguring surgery.

The paradox of screening is explained in this Sense About Science publication:

http://www.senseaboutscience.org/resources.php/7/making-sense-of-screening

This article explains the problem with screening for prostate cancer:

http://www.testingtreatments.org/book/why-are-fair-tests-of-treatments-needed/earlier-is-not-necessarily-better/weighing-benefits-and-harms/prostate-cancer-screening-clear-harms-with-uncertain-benefits/

The difficulty with "diagnosing" sexism is that we all have different tolerance thresholds so there is no clear diagnosis. Take the word "girls" as an example. At one end of the spectrum we have Deborah Blum for whom "the word “girl” tends to be a signal flare, a red light warning of problems ahead".

http://www.thedailybeast.com/articles/2015/06/16/sexist-scientist-i-was-being-honest.html

I'm at the other end of spectrum because I hadn't realised that anyone could possibly be offended by the use of the word "girl" by a 72-year-old man.

The question is how we deal with these differing sensibilities when language is constantly evolving and the "rules" are a moving goalpost.

I don't know what the answer is but it's an interesting question!

Thomas said...

Welch also talks about treat of high blood pressure. In the beginning, the research showed a clear need to intervene, just based on numbers, not symptoms. But that was in cases of extremely high blood pressure. Now, we're intervening in cases where there's actually a risk of doing harm (bringing people's blood pressure down too low) and -- here's the crucial thing -- the potential benefit is decidedly marginal. So we take a 1000 people and intervene in 100 cases, harming 20, and helping 1 in addition to the 4 that we'd have saved anyway. (I'm making those numbers up.)

With sexism, I think by brining our tolerance thresholds down, women are more often getting held back by their own perception of sexism (their "diagnosis" that they are being treated unfairly because they are women) than by any actual sexism in their environment. It may even technically be sexism. The male professor may be condescending, for example. It's just not going to affect their career in any way. Unless, of course, they decide to "treat" his condescension as a problem. Again, 1 out of 20 women may actually be hitting a real pig, and may actually be improving their career by dealing with him as a sexist (on the basis of some condescending remark). But the 19 will embroil themselves either in internalized resentment or a hopeless (and baseless) disciplinary procedure that will only harm their careers. Instead of just moving on because it was "just a joke".

Thomas said...

should add that there are, of course, other ways of getting that pig. I'm not saying we let him off. I'm saying that we don't ask twenty women to sacrifice their careers to get him.

Debbie Kennett said...

By lowering the thresholds you will indeed increase the perceived incidence. Zero tolerance is supposed to work with reducing crime. Is it the same with sexism? I suspect it probably has the opposite effect and that pointing out microagressions is counter-productive and creates a feeling of resentment. There need to be effective ways of dealing with bullies and sexual harassment in the workplace but it doesn't help to think of it only as a women's problem. Women can sometimes be the worst bullies, and men can also be the victims of sexual abuse. Any interventions to deal with this issues should be evidence based and properly tested.

I don't know if you saw this article about bias training which suggests that bias is exaggerated by creating awareness:

http://uk.businessinsider.com/companies-are-spending-millions-on-bias-training-that-isnt-working-2015-7

Thomas said...

Yes, I did see that article. The general problem, I think, is the idea that "biases" and "attitudes" can be corrected directly. As if we can talk people out of their prejudices. There is a very simple (but slow) process that will correct men's misperceptions about women: continuous unavoidable contact with them in professional life. This will cause them some "trouble", as Tim Hunt tried to suggest, but it will be worth it.

I don't think zero tolerance works, even for crime, and, like the precautionary principle, it is sort of nonsense. True zero tolerance would be the death penalty for any infraction. We don't want to live in a world that doesn't measure the punishment against the crime. And as soon as you measure, you get into some interesting subtleties. You want to treat first time offenders differently than repeat offenders. You want to consider intent. Etc.

You don't want have some automatic "trigger" like a man calling women "girls" to make you fly into action.

Debbie Kennett said...

I would hope that it is possible to change people's misperceptions and to reduce bias. I think it becomes harder to do this as people get older. There needs be more tolerance and understanding of people from older generations.

I agree that punishment needs to fit the crime. People should be allowed to learn from their mistakes, and punishment for a first offence is inappropriate.