Smoke screen

I apologize for the dreadful and pejorative pun in the title, but it’s either this or something dry and academic. Deal with it.

I spent some time at a private breast cancer screening clinic recently. I worked with a doctor who was clearly dedicated to unearthing every last breast cancer in his patients. A noble cause. Immediately after their digital mammograms, women sat down with this doctor to examine the images together. Most women—some more than others—have some areas where their breasts are naturally more dense. The images of these areas are not so clear; in all the women that morning, he proceeded to do an ultrasound.

I was there as an observer, and the main thing I observed was this: the quest for health was, for this doctor and these patients, a quest for certainty. Patients come in with one question: “could I have breast cancer?” But mammograms often cannot provide certainty. So he pressed on, refusing to definitively say “no” until all the tests had been done: the women that morning variously underwent ultrasound, MRI, and/or needle aspiration in addition to their mammogram. During one ultrasound, I saw a woman’s eyes go wide as this doctor casually mentioned that he could see a small cyst. His gaze was on the screen. “A cyst is normal”, I ventured. “Yes, it’s usually normal”, he half-agreed. The anxiety was, well, palpable. In the end, nothing was found. “Come back in six months.” Was this screening really helping this woman? Pull up a chair: the answer is going to take a while.

Living in a country where cancer is one of the leading causes of death, and which has the resources to screen people for cancer, I’ve grown up learning that checking for cancer is a crucial part of preserving your health. Cancer can strike anyone, and it starts with a single cell, but eventually it overtakes your body and kills you. So: the earlier you find it, the easier it will be to eliminate, the longer you will live. Everyone knows this, right? This is obvious. We are reminded regularly by ads like this one, which has been all over Montreal lately:

Ad campaign from the Quebec Breast Cancer Foundation: “If you love your breasts, raise your hand.”

Yes indeed, screening is a fine thing. Indeed, we doctors routinely recommend that everyone get their annual Pap smear (cervical cancer), mammogram (breast cancer), colonoscopy (colon cancer), chest X-ray (lung cancer), full skin examinations (skin cancer), blood counts (leukemia), thyroid ultrasounds (thyroid cancer), gastroscopy (stomach cancer), PSA levels (prostate cancer), CA-125 levels (ovarian cancer), endometrial biopsy (uterine cancer), brain scan (brain cancer), urine cytology (kidney and bladder cancer), abdominal ultrasounds (pancreatic cancer, liver cancer, biliary cancer) …

Wait a minute. We don’t? You mean we just sit around waiting until most of these cancers are so advanced that they’re causing obvious symptoms before we try to diagnose them?

By now, I’m sure you’ll have guessed what’s coming. The more I learn about screening, the more I suspect a lot of it has been oversold in our enthusiasm to do everything we can to keep healthy. Some of our screening tests are good, though not as good as they’re made out to be. Others—which are extremely popular—may actually be doing more harm than good.

How can that be? Isn’t early detection the key to prevention?

Looking back at that long list of cancers and supposed screening tests, there are a variety of reasons why we don’t use most of them. The simplest reason could be that they are just not very sensitive: they miss most of the early cancers, so they’re not very useful. Some don’t pick up any cancers until it’s too late, so they’re completely useless. Worst of all, many of these tests actually cause harm.

It is taken for granted that a new drug should be thoroughly tested for effectiveness, safety, and any harmful side effects. If so, maybe we should be even more cautious with screening tests than with treatments: we might tolerate a side effect from a drug if it helps people recover from an illness, but screening is something we recommend to large numbers of healthy people. And yes, screening—even though it “just provides information”—can do harm, because of the responses that information tends to provoke.

In medicine, we all want to find ways to prevent or relieve suffering, and we trust our peers when they tell us what works and what doesn’t. This desire and trust are very good things for the most part, but they can leave us vulnerable to putting our faith in tools that seem helpful but might not have been subjected to adequate scientific scrutiny—which we are not adequately trained to evaluate for ourselves anyway. The scarier the disease, the more willing we are to try anything that might help. However, we also profess to be a science, and the highest scientific virtue is a willingness to reconsider your most cherished beliefs in the light of objective evidence. So we are getting better at taking a sober second look at these tools, but for now, mainstream medicine remains jam-packed with dogma.

If you ask me, the ideal study of a supposed screening tool would look like this: recruit a big group of apparently healthy people, randomly divide them in two, offer half of them the test and the other half not, and then wait and see what happens to them—good and bad. If the screened group experiences more good than bad, then you have a good screening tool. Simple!

That’s the ideal. Is that really how we decide? Well, let’s look at two examples: breast and prostate cancer screening.

Breast cancer screening

Suppose that in one study of mammography for healthy women, researchers looked at the women who were found to have breast cancer: the five-year survival rate (from time of diagnosis) was 90% in the screened group, and 20% in the unscreened group. An amazing result! (Note to self: find an actual reference, don’t just assert that studies with numbers that look “something like this” actually exist. They do, though.)

Unfortunately, it’s extremely misleading: it looks at the wrong thing. If you diagnose a disease earlier, of course your short-term survival will look better. Say you have a woman who would have noticed a big breast lump at age 55, but you screen her and find it at age 50. Five years later, she’s still alive, and you congratulate yourself, but she would have been alive even without screening. Congratulations: you’ve increased the amount of time with apparent disease, and you’ve decreased the amount of time with apparent health, and you’ve told yourself you’ve done a great thing. Well… we don’t know that yet. (This is called “lead time bias”.)

So maybe you do find that if you screen, the breast cancers that get diagnosed end up being less fatal. Great! That must be because earlier treatment is more effective! Again, maybe not. Unfortunately, screening is most effective at picking up those very slow-growing cancers that would not have killed you anyway. The unscreened group will have more serious cancers; the screened group will have more “benign cancers” that didn’t need to be found at all, but they make your stats look good. Not really a fair comparison either. (This is called “overdiagnosis”: making a diagnosis that is actually irrelevant.) Plus, here we find one of the harms of screening: people who would have lived out their whole lives never being aware of their small, localized, non-lethal cancer have now been told that they have a scary disease, and the discomfort, side effects, and complications of the investigation and treatments.

Okay, maybe those are the wrong calculations, but isn’t it still true that treatment is more effective when cancer is detected earlier? Maybe not. Breast cancer treatment is getting better and better, and ironically, the better our treatments are, the less important it is to get screened, because your chances are still good even if you find the cancer relatively “late”.

When we look at the right thing—how many people in each group die of breast cancer, and indeed how many people die at all, regardless of whether they got diagnosed with breast cancer (which might have been an artifact of screening)—we do indeed find that mammography decreases death rates. Somewhat.

The best guidelines on this subject—best in my view, at least—are from the exquisitely tough Canadian Task Force on Preventive Health Care. They calculate that for women aged 50–69, for every life saved from breast cancer, 721 women need regular mammograms for over ten years, of which 204 will have false positives, 26 will have unnecessary biopsies, and 3 or 4 will have unnecessary surgery. For women aged 40–49, the numbers are even less inspiring: 2108 women screened for ten years, with 690 false positives, 75 unnecessary biopsies, and 11 unnecessary surgeries to save one life. Is this worth it? Well, that’s a matter of personal choice.

Basically, the benefit is small, but it looked big on older inappropriately-designed studies. Guess which studies were used to inform government policy and promotional campaigns?

Now we have a legion of people who’ve spent their careers promoting mammograms, doing mammograms, interpreting mammograms, or investigating and treating women with suspicious-looking mammograms. We also have a lot of women who’ve had suspicious mammograms, and suffered through various amounts of treatment for various degrees of cancer. Most of these people feel that mammograms save lives.

The breast cancer doctor I mentioned at the start of this article told me how arrogant he thought it was for these guideline-writers to dismiss mammography, when the studies have really only examined film mammography, and we have digital mammography now. This is possible. Still—and I may be just a hard-nosed amateur epidemiologist—I find it even more arrogant to promote an unproven tool which might be harmful. I didn’t say so at the time.

The guideline-writers have further been accused of hyperbole about these harms, and insensitivity to the plight of women with breast cancer found too late. They meekly respond that they are still for mammography, but with a full discussion of benefits and risks. They (and I) do not see what is so controversial about this.

As for other screening methods:

• Teaching self-exams mostly leads to a lot of benign lumps and anxiety, and it has never been shown to save lives in the scientific sense described above. We now think teaching breast exams does more harm than good. This hasn’t stopped private charities from spending donated money promoting this harmful practice, including via the image I included at the beginning of this post.
• Physical examination of the breasts by doctors: also unproven.
• Breast ultrasound: unproven.
• Breast MRI: unproven.
• Digital mammography: unproven to be meaningfully different from film mammography.

This despite the assertions of their proponents. Breast exams and higher-tech imaging may pick up more abnormalities, but the more you dig for abnormalities, the more you’ll tend to pick up things that turn out to be benign or minor.

Prostate cancer screening

Prostate cancer screening is frequently performed in men over 50 with rectal exams (the back of the prostate can be felt via the rectum) and PSA testing. PSA is a protein made by the prostate which can be measured in the blood. Some prostate cancers make lots of it, so a high PSA might be a sign of prostate cancer. Lots and lots of men have had this screening in the US; fewer have in Canada.

Prostate cancer is even more slow-growing than breast cancer. If a bus full of 80-year-old men falls off a cliff, you’ll find prostate cancer in 80% of them when you do the autopsy. Prostate cancer can be aggressive, but it is most often indolent.

I’ll cut to the punchline: unlike with mammography, researchers have not been able to show any mortality benefit when PSA screening is offered to asymptomatic men, regardless even of risk factors such as age or family history. PSA screening has, however, been shown to lead to transrectal biopsies (painful and carrying a risk of infection), surgery (high rates of incontinence and impotence), hormone and radiation therapy (side effects galore), and anxiety. (It might be useful for diagnosis of symptoms that are “highly suggestive” of prostate cancer, but that’s a separate question.)

The other time-honoured method for prostate cancer screening is the rectal exam. Sadly, there is no evidence here either.

Now, these studies of harm were mostly done in the USA, where urologists treat screened prostate cancer more aggressively than they do here in Canada. The Canadian Urological Association recommends at least serial PSA testing. This is a promising technique but, sadly, unproven to make a difference in the general population, at least for the kind of “proof” I expect for this sort of thing. The preventive services task force in Canada, whose methods I consider more scientifically objective, has recommended against routine PSA screening since 1994, and its American counterpart agreed, tentatively, in 2011.

In medicine in general, one finds that specialists (including those who write guidelines) are more enthusiastic about their tools than generalists. Who to believe? Specialists do have substantial clinical expertise, and nobody is perfectly objective, but there are methods for reducing bias… and with all due respect to specialists, when your only tool is a hammer, you see nails everywhere.

There is a poster promoting prostate cancer screening right near the hospital where I work. I don’t know who funded it, but I’d like to, because I think it’s a bit of a menace.

Cancer screening in general

Basically, a perfect screening test requires the following:

• the disease is serious and easily diagnosed as such, but with a significant early asymptomatic period during which our treatments are substantially better than the later symptomatic period;
• the test is easy, acceptable, comfortable, and cheap;
• the test picks up most serious disease, and either does not pick up benign disease, or only picks up benign disease which is easily distinguishable from serious disease.

The difficulty with all cancers is that the serious ones grow quicker, any screening test disproportionately picks up the slow-growers, and our predictions about whether a given cancer will grow slowly or quickly are not all that refined. And again, the better our treatments get, the less benefit there is from early detection.

This whole idea—that such a seemingly obvious and widely-promoted idea could actually be wrong—has been quite a revelation for me this year. And you know me: there’s nothing I like better than dismantling dogma.

Cynical yet?

Can you use statistics to prove anything? No. You can misuse statistics to make misleading and incorrect claims. There are ways to get it right.

Cynicism also says that science is constantly changing its mind, so why should we even listen to it? Last year they said PSA was good, this year it’s bad, maybe next year it’ll be good again. They’ve done the same with aspirin, cholesterol drugs, and moderate alcohol to prevent heart disease… what’s the point? Well, I might ask, are we going back and forth on penicillin for syphilis? Pap smears for cervical cancer? Appendectomy for appendicitis? Smoking for lung cancer? There are degrees of certainty in science, and often we in medicine get a bit too enthusiastic about promoting something based on encouraging but not definitive results. Any such claim ought to be accompanied by a sense of its certainty.

The medical profession itself teaches us to pride ourselves on our independence and decisiveness. Scientific humility, on the other hand, is harder to come by. Understandably so: people who are sick or worried are reassured by certainty. But when we give people certainty without justification, we risk making things worse.

Did I go through this for nothing?

I know people, including some very close to me, who’ve had cancer screening, and who’ve then been treated for cancers that were found early. They may be living with the enduring side effects of that treatment, and they probably feel that their lives were saved. They may be right. Am I saying they suffered through all this for nothing?

Well, no. I can’t say for sure what would have happened to any one individual. Maybe your life was saved. Especially if you had any symptoms or were at particularly high risk, in which case the entire above discussion doesn’t even apply to you.

But if you’re thinking of using your good outcome to encourage more people to get screened, I hope you’ll take the above information into account.

So my doctor is incompetent?

Well… this whole screening business is controversial. Doctors disagree. Doctors are also influenced by their cultures and their personal experiences, and while they’re generally your best bet, lots of their beliefs are not totally justifiable from scientific literature, they’re not going to know everything, and they’re going to make mistakes. The very notion that medicine should be so self-critical and evaluate our interventions with hard scientific evidence is relatively new, and we’re still working on the many, many details that that entails. Expecting perfection is unrealistic. Opting for a screening method which seems promising but is unproven is a valid choice. (Whether the health care system should encourage and finance it is another question entirely.) And again, your case might have been different from the “average” case discussed above, if you had suggestive symptoms or were otherwise at higher-than-normal risk.

The bottom line

Screening has under-recognized harms which you may prefer to avoid, but some screening methods are helpful (depending on your risk), and others are oversold or useless.

So I’m a public health doctor discouraging people from getting preventive health care. Ironic, right? Not at all. Cancer prevention is a fine thing. It’s just that some of the things that get called “prevention” are not all they’re cracked up to be. I’ll still tell you to not smoke, get plenty of exercise, eat plenty of fresh fruits and vegetables and whole grains, not eat plenty of meat or fat or sugar or processed foods or salt, and demand of your elected officials that their economic and regulatory policies make these things easier and not harder.

Well—probably. We’re not totally sure about this stuff.

(*And for the record: I do believe in screening for sexually transmitted infections, cervical cancer, breast cancer, colon cancer, diabetes, hypertension, high cholesterol, osteoporosis, and an array of genetic and congenital disorders, with varying degrees of confidence. I’m not against medical screening. I’m for proven medical screening, and I’m against unproven screening, especially when it’s harmful.)

Correction (August 4, 2012): I apologize for my repeated implication in the post above that only women have breasts and only men have prostates. I got caught up in standard medical oversimplification. Gender is not so narrowly defined.