Understanding Numbers in Medicine: Easier than You Might Think, and More Important Too
Dear Readers,
Here’s the fifth in our occasional series of newsletters on getting better medical care for yourself and your family.
Today, we’re talking numbers. But don’t be put off!!! If you know basic addition, subtraction, multiplication and division, you can understand enough about medical statistics to make intelligent choices about treatments and tests.
Pretty much all decision making in health care revolves around statistical estimates of the probabilities of bad things versus good things happening to you if you do or don’t try a particular drug or screening test or even surgery.
If you decide to tune out the numbers, rest assured that someone else will make the decision for you based on how they understand the numbers. And they could be wrong.
So it’s important for all consumers of health care who want to make informed choices to learn a little bit about how numbers are used — and misused — in health care.
Here’s one example: Many middle-aged Americans are now hearing that a statin drug for lowering cholesterol might reduce their risk of heart attack by as much as fifty percent, even if they don’t have high cholesterol.
Is that true? Not really.
Is it a fraudulent statement by the manufacturer? No, that’s not true either.
Read on to our first article to learn how you can figure out for yourself the real truth. Then go on to our next article on the breast cancer drug Tamoxifen to learn more about counting numbers.
Count the Patients If You Want to Know the Real Scoop
Should you take a cholesterol drug every day for the rest of your life even if you don’t have high cholesterol? Of course not, you say, but then you hear that the drug might cut your heart attack risk in half. So you think again. To understand enough to make an intelligent choice, you need to know some numbers.
If you have a high level of inflammation in your blood, but you have normal cholesterol levels, you are the focus of a new marketing campaign for Crestor, the statin drug. You will be told the drug can lower heart attack risk by as much as 50 percent. Should you swallow this advice?
The downside of statin pills is the possibility that they increase risk of diabetes and liver damage. The upside, of course, is reducing your risk of dying from the nation’s No. 1 killer, heart disease. So it’s important to know accurate numbers to make an intelligent decision.
The 50% reduction in heart attack is not technically inaccurate, but it’s a statistical sleight of hand. The correct risk reduction number is actually around one in 500, or 0.2%. To understand the number manipulation involved, read on for a gentle short course in medical statistics.
These numbers come from a big study that the FDA used to give the green light to Crestor’s new marketing campaign. The researchers found that around four in 1,000 patients with a high level of C-Reactive Protein (an inflammation marker in the blood) but with normal cholesterol had heart attacks if they took a dummy (placebo) pill in the study. Similar patients who took Crestor had a heart attack rate of around two in 1,000. These numbers come from the tail end of an article in the New York Times on the FDA’s approval of the Crestor campaign.
Now the difference between four and two is 50 percent, but that conveniently leaves out the denominator in the statistic. The real risk reduction is from four in 1,000 to two in 1,000, or a difference of two in 1,000, which is the same as 0.2 percent. The 50 percent number is called the relative risk, and the 0.2 percent number is the absolute risk.
Absolute risk is what measures real people. That’s the number we patients should focus on. But marketers like relative risk because it produces more dramatic numbers. Put another way, you can have statistically significant benefits from taking a drug on paper, but the clinically significant benefit, in real life, is much much smaller.
In my book, The Life You Save: Nine Steps to Finding the Best Medical Care — and Avoiding the Worst, I discuss the Crestor story because it’s just one example of how consumers can be misled by medical numbers. There’s a straightforward solution, which I call “counting the people,” when you are looking at any supposed medical breakthrough. Here’s what I said in the book:
The real number [of how many patients benefit from a drug] might be enough to persuade some patients to take the drug. But it’s a lot different than fifty percent. Bottom line: to make intelligent choices about treatments, patients need to understand how many patients like them are really expected to benefit from the treatment. You can get these answers by focusing on how many actual people are helped by the treatment. Do not focus on misleading vague numbers like “50 percent improvement.” Fifty percent of what? Focusing on real numbers of real people will give you the answer.
As a patient safety advocate, I want all smart consumers of health care to know at least a little bit about statistics — because the real numbers can be very eye-opening.
How to Read Numbers: The Tamoxifen story
Tamoxifen is another one of those 50 percent drugs. If you are at high risk for breast cancer, it can cut your risk of getting the cancer in half. That’s statistically “proven” by the studies. But the drug is proving to be a tough sell when women have a chance to fully understand the pros and cons of the medication. Again, counting the numbers of patients is very helpful, as you’ll see below.
This story gets a little more complicated than the story we just told of Crestor for patients with inflammation in their blood, but it’s more interesting in some ways too.
Recent studies show that when the statistics are laid out for women to make an informed choice, only one in 100 actually fill the prescription for tamoxifen. Are the women who decline being illogical? Or just making their own personal choices about what is important to them?
Tamoxifen interferes with the body’s use of estrogen. That can lower the risk of estrogen-dependent breast cancers but can increase the risk of other estrogen-related side effects like cancer of the endometrium (the lining of the uterus), blood clots and sexual dysfunction.
Here is how the numbers were spelled out to women in a recent study at the University of Michigan (as first reported last winter in the New York Times).
The risks of breast cancer vary with age, family history, and age of first childbirth. So a 52-year-old woman who had her first baby after age 30 and whose mother had breast cancer, has about a 1.9% risk of developing breast cancer over the next five years. If 1,000 women just like this 52-year-old took tamoxifen over those five years, the research says that here is what would happen:
* Of the nineteen women (same as 1.9%) who otherwise would have developed breast cancer, nine will not develop breast cancer. (Thus the statistic about lowering the odds by half.)* Thirteen women would avoid broken bones from osteoporosis, another benefit of tamoxifen.
* Twenty-one women would develop endometrial cancer (typically more treatable and less deadly than breast cancer if caught early).
* Twenty-one women would develop blood clots.
* Thirty-one women would develop cataracts.
* Twelve women would experience sexual problems.
* One hundred twenty extra women would get hot flashes and other menopausal symptoms (in addition to those who would get such symptoms anyway).
Behavioral economists might say this is an example of “omission bias,” where we are more worried about a small risk from doing something new (taking a pill) than we worry about a larger risk from doing nothing. Put another way, we often see the status quo of doing nothing as safer when it really isn’t.
At least that’s how the researchers quoted in the NYT article explained the unpopularity of tamoxifen. But for readers who posted comments on the newspaper’s blog, they tended to see the women voting against tamoxifen as being quite sensible. It just doesn’t sound worth the downside.
Part of the problem is the apples-to-oranges comparisons involved when a fatal condition is compared to a non-fatal one. As one commenter posted:
What might make women make better choices is if they had data on whether the pill reduced the risk of DEATH from all causes. If only some of the breast cancers avoided would have resulted in death but all of the endometrial cancers aquired resulted in death, women might make the choice to avoid tamoxifen. If the risks were reversed, they might choose to take tamoxifen. We have to move beyond a discussion of risk of cancer and towards a discussion of risk of cancer DEATH. https://www.medpie.com – Barbara Lock, MD
If we had the same careful discussion about mammograms, women likely would opt for far fewer of these tests, which save lives on a similar scale as tamoxifen, with plenty of downside.
My conclusions from this debate:
* Patients need to know there are no magic bullet drugs that are all gain, no pain. Tamoxifen interferes with estrogen, which is good for some diseases, not so good for others. This is typical. Each drug must be carefully weighed for its pros and cons. There are no magic bullets that are all gain, no pain.
* We’re all better off with a full exploration of the odds and then make our own decisions. There is no right or wrong.
* The best way to understand risk is the way it’s spelled out here: with numbers of actual people in a given standard-sized group. It’s too confusing when we talk about percent this and percent that.
A New Approach: The “Drug Facts” Box
Wouldn’t it be great if the numbers we need about drugs and other treatment choices in medicine were laid out in some easy to grasp format? One possible wave of the future is the “Drug Facts” box, which is modeled after the “Food Facts” box that we see everywhere in the grocery store.
Click here to see a “Drug Facts” box on Tamoxifen (also called nolvadex). Note the clear presentation of the statistical likelihoods of being helped and harmed by Tamoxifen. This was developed by some researchers at Dartmouth medical school. They’ve proven in scientific studies that consumers can make smart choices if they’re given the information in a straightforward way. (Isn’t it comforting to know that we’re not so dumb after all?)
You don’t see these Drug Facts boxes anywhere now. Instead, what we have now are ads to consumers that first show happy, healthy, bouncy people, presumably after they’ve taken the drug, along with a few simplistic sentences pushing the drug’s benefits, but next these same ads show acres of fine print with all the downsides of the drug. The subliminal message is that all patients need to do is “ask your doctor” if Drug X is right for you. Meantime, your doctor has been sold a message on Drug X — usually by a well-tailored, attractive sales rep — that is only slightly less simplistic than the “direct to consumer” advertisement.
So really, both patients AND doctors could benefit from a requirement that the drug information be reorganized and presented in an intelligible way. The FDA is now considering such a rule. Let’s hope it happens.
To your continued health!
Patrick Malone
Patrick Malone & Associates
