If you like this blog, send it to 3 (or more) of your friends and encourage them to sign up. Let’s keep the conversation going!
The 21st Century seems to be, among other things, the Age of Health. Everybody’s concerned about health, everyone’s talking about health, and every cultural medium is flooded with ads about how to improve your health. The idea seems to be that, if you’re smart enough and take enough of the right drugs, you’ll live forever.
Whether you believe that myth or not, the ability to read and understand numbers in medical studies is becoming increasingly important. I’ve been talking about this for years, starting with the study of tamoxifen (a breast cancer treatment drug) for use in healthy women to reduce their risk of developing (not to be confused with “preventing”) breast cancer.
The “Tamoxifen for Prevention” Story
In 1998, when the study results were announced to great fanfare,the scientific journalsand press reports breathlessly reported that women taking tamoxifen reduced their chances of getting breast cancer by a stunning 50%. Reading this information, womennaturally concluded that, if they took tamoxifen, they could reduce their individual risk of developing breast cancer by almost 50%. From the hoopla that characterized and surrounded this news, you’d think we had finally cured cancer. Not really.
The 50% number led to the wrong conclusion. It turned out that, for individual women, the amount they would reduce their risk by taking tamoxifen was actually 2%. What’s the difference? The difference is between “relative” and “absolute” risk reduction.
A Brief Digression: Risk Reduction is Not Prevention
Before I get into the numbers and unravel the difference between relative and absolute risks or benefits, a few words about words. The tamoxifen study — like many subsequent studies of other drugs like raloxifene (Evista) and exemestane (Aromasin) — was described as a breast cancer prevention study. Its official name was BCPT-1: Breast Cancer Prevention Trial 1. And the results were described in terms of how much breast cancer was prevented. But some people in the study who took tamoxifen got breast cancer, just not as many as people who didn’t take the drug. So, for any individual woman, tamoxifen and other drugs used for the same purpose reduce the risk of getting breast cancer. They do not prevent the disease.The general understanding of the word “prevention” is that if you do step A, you will not get disease B. When scientists and health policy people use the word “prevention” to mean anything else, they mislead and confuse people. That shouldn’t be the goal, should it?
It’s Absolutely About You
The difference between “relative” and “absolute” benefit or risk is important to understand. I know that from my work as a breast cancer activist. I also know it because the New York Times, in an article published in the Science Section on May 30, 2011 reported that “Translation Matters in Choices On Data.” The article describes the different ways that the same medical data can be presented, and talks about how important it is to clearly present data so people can understand it. Since the New York Times reported it, this information must matter, right?
There are basically 3 different ways to convey information about a study of a drug to reduce the risk of a disease: relative risk reduction, absolute risk reduction, and number needed to treat.
In the simplest possible terms, here’s the difference among these 3 ways of describing the numbers, using a hypothetical example.
When someone says that taking a drug reduces your risk of disease by half, that’s a relative risk reduction number.
Another way to convey exactly the same information is to say that the risk of the disease is 2% for people who don’t take the drug, but your risk if you take the drug is reduced to 1%. Your risk has been reduced by half, but in absolute terms, that’s a reduction of only 1%. In other words, in this hypothetical, 98% of people won’t get cancer if they don’t take the drug; but if everyone takes the drug, that number changes to 99%.
A third way to convey exactly the same information is to say that 100 people would have to be treated with the drug for 1 person to get the benefit of the risk reduction from it. This number matters a lot if you’re concerned (or should be) about the other effects the drug might have, which are referred to medically as “side” effects.
For a real life example, we can look at these kinds of numbers from the tamoxifen prevention trial. In that study, as mentioned above, the relative risk reduction was 49%. The absolute risk reduction was 2.1%. The number needed to treat to prevent one case of breast cancer was 48.
So, when a medical professional recommends you start a drug, you should start by asking what your risk is if you don’t take the drug, and what your risk will become if you take the treatment. And what are the potential side effects?
“Other Effects”: The Devil’s in the Details
All drugs have side effects. Any drug powerful enough to prevent a serious disease like cancer is going to have other effects on your health besides potentially reducing your risk of getting cancer. With tamoxifen, the most serious risk is endometrial cancer. The drug also increases the risk of stroke, deep vein thrombosis (blood clots), and cataracts. Many women also experience hot flashes and vaginal discharge with tamoxifen.
The risks of the other effects, and their seriousness, will depend a lot on the individual to whom the drug is offered and that person’s current medical condition. But, in almost every case where we’re giving strong drugs to people to reduce their risk of getting sick, the “other effects” can loom very large, indeed.
Pay Attention to How Numbers are Reported
The take-home message in the New York Times article on how reporting data matters, is this: “Journalists have to be careful about press releases with ‘new’ or ‘groundbreaking’ studies presented with relative risk reductions.” I would add that scientists should be required by the journals that publish their studies to report the absolute risk reduction numbers, as well as the number needed to treat.
Tamoxifen Redux: Aromatase Inhibitors for Breast Cancer “Prevention”
In the early 2000’s, I was attending a San Antonio Breast Cancer Symposium where the first data on an aromatase inhibitor (AI) for treating breast cancer were presented. There was a great deal of excitement about the data. It was the kind of excitement that meant that medical practice was about to change, and did it ever. Almost overnight, oncologists began prescribing an AI called Arimidex (anastrozole) for post-menopausal women with hormone receptor positive breast cancer.
Up to this point, the standard treatment in this setting had been tamoxifen. But the news of the AI study was so striking that doctors came up to me at the San Antonio meeting to assure me that tamoxifen was dead.
Tamoxifen’s obituary hadn’t really been written at the time, and still hasn’t. But the other thing that I heard at that San Antonio was from advocates and activists who were concerned that this new class of drugs would be tested in healthy women for breast cancer risk reduction, just as tamoxifen had been.
Those advocates were right. At a huge cancer meeting recently in Chicago, the results of the first study of an AI in women who did not have breast cancer were released. While sources like Medscape and the National Institutes of Health heralded the news about exemestane (Aromasin) as breast cancer “prevention,” the June 4, 2011 New York Times was more cautious, with a headline that read “Drug Can Reduce Breast Cancer Risk, Study Says.”
And how were the results presented in the study? As relative risk reduction numbers, of course. All the press stories led with the news that women taking the drug in the trial reduced their risk of breast cancer by 65%.
The New York Times followed the 65% relative risk reduction numbers with the absolute risk reduction number, which is far different: about 0.9%. Imagine what the response to the story would have been different if it had led with this number instead?
According to the study, 94 women would need to be treated for 3 years with exemestane to prevent one case of breast cancer. That means that 93 women would be exposed to other effects that include bone pain, joint pain, and — noticeably missing from the list in the New York Times — osteoporosis. Most women on AI’s for breast cancer develop osteoporosis, which means they get to take another drug to reduce the risk caused by the AI.
Know Your Numbers
In the Age of Health, we need to understand health numbers and their significance. Only with this information can we make intelligent decisions about treatment options and their risks. A great deal depends on understanding health numbers– maybe even your life.
© Barbara A. Brenner 2011
Another wonderful post, Barbara. You continue to write about these complex issues with such clarity and force. You have made such a difference for so many people.
love,
Amy
I finally really understand these terms! Thank you so much for your concise and simple explanation of something that is often difficult to grasp. You are amazing as always!
Jill
Once again Barbara Brenner has unraveled and explained a complicated issue that literally effects us all. How does she do it?
Nice explanation of the use and misuse of numbers. Aside from the human impact of “treating” so many people with drugs of limited efficacy, the economic/financial impact on the system is a disaster. Those are terrible risk/reward numbers.
Brilliant.
Thank you, Barbara. I paid attention during stats class & chemistry class, before I got to grad school for PT. And thank goodness I did. Oncs especially just love to quote stats out of context & scare us with them. I had to do my own digging to find out how much difference their proposed treatments would really make in the scheme of things. And for some of them, it was decidedly ‘not much,’ especially at the expense of quality of life. The way that the media hypes new study results does not help clarify things either, but tends to heap more confusion on results that are already quoted out of context. Not to mention that many studies are done and reported with an unconscious bias embedded in them by the researchers themselves. Plus, if you wait an average of two years after a study is published, it will likely be refuted by then.
Excellent post. I wish everyone who panics over the release of the latest research would read this article.
I wish that you were in charge of the world….those that are do not belong in charge. a
Just caught up on the posts I had missed from our travels, etc. I agree with Abby–being in charge of the world sounds like a good job for you (and for all of us) now. Amazing wisdom and brilliant, clear writing. Thanks so. Love you, liz
Thanks, Barb. So straightforward and helpful.
Thank you.
perfectly clear and so important! as someone who got a second case of breast cancer (a whole new one, not a metastatic one) even though I took my tamoxifen dutifully for 5 years as they said, clearly it doesn’t always work! Also, now they want me to take AI– not so fast, what is the point? thanks for this post, very timely for me! xoJ
I have just read this and it’s incredible to me. I too have had a second (new) breast cancer after taking tamoxifen for 5 years for the first one I had. My doctors have me back on Tamoxifen as they state that it’s a new cancer and is treated from scratch. Like JoAnn I feel that – for me – tamoxifen did not work in decreasing my risk of breast cancer. I am now in a real quandary as to what to do. I see my oncologist on Tuesday so will discuss this. Thanks for writing this post and for the responses from women like JoAnn – I don’t feel so alone.
Thanks, Barb. This should be required reading for anyone diagnosed with breast cancer. I am currently on Arimidex but will be balancing the side effects with the risk reduction analysis you spoke about. Your clear voice comes through loud and strong.
-Ellen
Pingback: NEIN zur „Brustkrebsvorbeugung“ mit Aromasin (Exemestan): Die Breast Cancer Action-Position – infoblog!
From my vantage point, dear Barbara, your voice is stronger than ever! This is a great post, disconcerting as the contents is. Not surprising info, but surprisingly clear. Thanks!
Wonderful analysis. I think that the NNT (number needed to treat) and benefits vs. side effects are the essential elements in evaluating treatment (or “prevention”) options.
As was also mentioned recently in the New York Times, the side effect list as proffered by the drug companies is getting out of hand (long lists of everything ever reported), making it more difficult to figure out what the relevant side effects might be.
Good to hear your voice again, Barbara! (Just found your blog.) I was watching the media reporting on Aromasin and shaking my head. Not only are the absolute numbers small, but they totally downplayed the side effects. Having been on the drug, I know they are real and affect quality of life.
Thanks for your blog.
Barbara, I am very grateful to be one of your blog followers. Your views on healthcare, research, and the whole picture have taught me a lot… Bob
I agree that numbers really do need to be kept in perspective because percentages look so much bigger than they actually are. That being said, for some reason – unlike most other drugs out there, including tamoxifen, – the AIs do not worry me too much, both in and of themselves, and also in terms of how they could potentially be marketed for cancer “prevention.” For one – the pool of patients who might even consider it based on risk is so much smaller than say half the entire population for whom they would like to force feed statins!
But for AIs as a drug class and for those currently taking them to ward off a recurrence of cancer, I am not terribly concerned, with the exception of bone loss. I think the dogma of estrogen protecting against everything under the sun has really been shot down, both by the disastrous result of HRT trials and the mere fact that longitudinal studies have not shown an increased risk of any disease at menopause – if anything, high estrogen in later life is risky. As an avid reader of clinical trial data, I have been quite amused by editorialists’ commentary expressing their surprise at how AIs have not turned up cardiac and cognition risks over placebo, but it never really was proven that estrogen protected women from heart disease or dementia. So the AI studies, FOR ME, have been vindicating because they add more evidence to the premise that the paradigm of women’s health (ie “you’re only as good and healthy as your estrogen levels”) is long overdue for a major change! The estrogen shills have been relentless in trying to discredit the WHI and it has harmed women.
Believe me, I , in no way, espouse taking pills for prevention (obviously because it puts money-making over inidividuals’ wellbeing), but these data, for the reasons stated above, have had a secondary advantage in debunking the estrogen myth.
Pingback: NEIN zur „Brustkrebsvorbeugung“ mit Aromasin (Exemestan): Die Breast Cancer Action-Position | Breast Cancer Action Germany