Skip to main content

Don't slam Canada for mammogram study

By H. Gilbert Welch
updated 7:42 AM EST, Wed February 19, 2014
STORY HIGHLIGHTS
  • A recent Canadian study found that mammogram screenings don't reduce cancer deaths
  • Mammographers slammed it, saying researchers cheated and screening machines were old
  • Gilbert Welch: Cheating charges unfounded, and pro-screening studies used older machines
  • Welch: The real problem is that mammography results in women getting treatment they don't need

Editor's note: H. Gilbert Welch is a professor of medicine at the Dartmouth Institute for Health Policy and Clinical Practice and author of "Should I Be Tested for Cancer? Maybe Not and Here's Why" (University of California Press) and co-author of "Overdiagnosed: Making People Sick in the Pursuit of Health."

(CNN) -- We all like to think medical care is about science, but too often it's about professional interests.

Last week, a 25-year follow-up of the Canadian National Breast Screening Study was published -- one of the eight major randomized trials of screening mammography. The headline was simple: Mammogram screenings don't reduce cancer death rates.

The reaction by some American mammographers was predictable -- discredit the study. It's predictable because it is exactly what they did when they didn't like the first findings of the study published more than 20 years ago.

The effort by the American College of Radiology to discredit the Canadian trial relies on two allegations:

H. Gilbert Welch
H. Gilbert Welch

The investigators were cheating: Let's look at the background on this. Randomized trials are a critical tool for clinical researchers. Study participants are placed at random in either one group (in this case those who get mammograms) or the other (those who do not). Who is in which group is solely based on the play of chance -- a flip of a coin.

The allegation of cheating -- purposely putting women whom researchers knew had advanced cancers in the mammography group -- is an incredibly serious one. It sure was taken seriously by Canada's National Cancer Institute. It launched a two-year independent review of the entire randomization process. In 1997, the review found no credible evidence of cheating.

But that didn't stop the allegation from being trotted out last week. The new study provides evidence that randomization did exactly what it is supposed to do: It created two identical groups of women. The rate of death in the two groups was exactly the same, every year, for 25 years. That can't happen by cheating -- that can only happen when the groups are formed solely by chance.

It sounded as if Canada were Botswana. Like the nation had only recently gotten electric power and was still struggling to train doctors.
Gilbert Welch

Canada is a Third World country: More specifically, the Canadians used old mammography machines that produced inadequate images, interpreted by inadequate mammographers.

It sounded as if Canada were Botswana, as if the nation had only recently gotten electric power and was still struggling to train doctors.

For mammographers who need to point to the benefits demonstrated by earlier trials of mammography, this is an odd allegation. Why? Because the other trials used even older technology. In fact, one of the trials most favorable to screening -- the Health Insurance Plan of New York's -- dates from two decades before Canada's, in the early 1960s, when mammography technologies were primitive.

Yet, in both New York and Canada, outmoded technology did exactly what it was supposed to do. It found small breast cancers.

Armed with these two allegations, the mammographers followed a well-worn strategy: Make the allegations often and loudly enough and maybe they will stick.

To be clear, not all mammographers share this view. A new generation has openly acknowledged the problems of mammography. But many in the old guard are more likely to attack any suggestion that screening doesn't work as well as advertised, characterizing researchers who raise the possibility as "malicious" or "dangerous" and questioning the editorial policies of the journals that publish their work.

It's time to stop the unfounded allegations. It might be standard procedure for politics but not for science. Too much energy has been devoted to discrediting the Canadian study and not enough to understanding it.

To make sense of information the Canadian trial offers, you need to understand its unique design -- specifically what is being compared with what. Most of the other randomized trials simply compared regular mammography with doing nothing.

In the Canadian trial, one group received a regular physical exam of the breast -- a very careful exam performed by specially trained nurses. The other group received the same regular physical exam plus regular mammography. In other words, the trial tested the usefulness of adding mammography to a physical exam in an effort to detect abnormalities that are too small to feel.

And the trial showed that finding these "too small to feel" abnormalities doesn't help women live longer.

That's really important information. It doesn't mean that mammography can't help at all -- it is extremely challenging to standardize a physical exam of the breast across an entire population. It does mean, however, that if we are going to do mammography, we should be using it to find big, important things -- not small, unimportant things.

Further, the Canadian trial confirms that the harm of being overdiagnosed by screening mammography is real: One in five invasive cancers found by screening represents overdiagnosis. Overdiagnosis happens when cellular abnormalities meet the pathologic definition of "cancer," yet never progress to cause clinical disease. Overdiagnosed women are told they have cancer, are treated for cancer, yet their "cancer" is not destined to cause them any problems.

It is important to recognize this estimate doesn't include the smallest, earliest form of breast cancer found only by mammography -- ductal carcinoma in situ or DCIS.

If overdiagnosis is a problem in invasive cancer, you can imagine it might be a greater problem for a smaller, earlier form of cancer. Even mammography's old guard objects to having DCIS included in estimates of the amount of overdiagnosis.

Why do they object? Because so much of DCIS represents overdiagnosis, including it makes the estimate even higher.

It's time to get the science back in screening mammography and to recognize that mammographers may not be the ideal source for balanced information. It's too much like asking the dentists for balanced information about routine dental X-rays.

Follow us on Twitter @CNNOpinion.

Join us on Facebook/CNNOpinion.

The opinions expressed in this commentary are solely those of H. Gilbert Welch.

ADVERTISEMENT
Part of complete coverage on
updated 11:17 AM EST, Fri December 26, 2014
Today's politicians should follow Ronald Reagan's advice and invest in science, research and development, Fareed Zakaria says.
updated 8:19 AM EST, Fri December 26, 2014
Artificial intelligence does not need to be malevolent to be catastrophically dangerous to humanity, writes Greg Scoblete.
updated 10:05 AM EST, Fri December 26, 2014
Historian Douglas Brinkley says a showing of Sony's film in Austin helped keep the city weird -- and spotlighted the heroes who stood up for free expression
updated 8:03 AM EST, Fri December 26, 2014
Tanya Odom that by calling only on women at his press conference, the President made clear why women and people of color should be more visible in boardrooms and conferences
updated 8:12 AM EST, Fri December 26, 2014
When oil spills happen, researchers are faced with the difficult choice of whether to use chemical dispersants, authors say
updated 1:33 AM EST, Thu December 25, 2014
Danny Cevallos says the legislature didn't have to get involved in regulating how people greet each other
updated 6:12 PM EST, Tue December 23, 2014
Marc Harrold suggests a way to move forward after the deaths of NYPD officers Wenjian Liu and Rafael Ramos.
updated 8:36 AM EST, Wed December 24, 2014
Simon Moya-Smith says Mah-hi-vist Goodblanket, who was killed by law enforcement officers, deserves justice.
updated 2:14 PM EST, Wed December 24, 2014
Val Lauder says that for 1,700 years, people have been debating when, and how, to celebrate Christmas
updated 3:27 PM EST, Tue December 23, 2014
Raphael Sperry says architects should change their ethics code to ban involvement in designing torture chambers
updated 10:35 PM EST, Tue December 23, 2014
Paul Callan says Sony is right to call for blocking the tweeting of private emails stolen by hackers
updated 7:57 AM EST, Tue December 23, 2014
As Christmas arrives, eyes turn naturally toward Bethlehem. But have we got our history of Christmas right? Jay Parini explores.
updated 11:29 PM EST, Mon December 22, 2014
The late Joe Cocker somehow found himself among the rock 'n' roll aristocracy who showed up in Woodstock to help administer a collective blessing upon a generation.
updated 4:15 PM EST, Tue December 23, 2014
History may not judge Obama kindly on Syria or even Iraq. But for a lame duck president, he seems to have quacking left to do, says Aaron Miller.
updated 1:11 PM EST, Tue December 23, 2014
Terrorism and WMD -- it's easy to understand why these consistently make the headlines. But small arms can be devastating too, says Rachel Stohl.
updated 1:08 PM EST, Mon December 22, 2014
Ever since "Bridge-gate" threatened to derail Chris Christie's chances for 2016, Jeb Bush has been hinting he might run. Julian Zelizer looks at why he could win.
updated 1:53 PM EST, Sat December 20, 2014
New York's decision to ban hydraulic fracturing was more about politics than good environmental policy, argues Jeremy Carl.
updated 3:19 PM EST, Sat December 20, 2014
On perhaps this year's most compelling drama, the credits have yet to roll. But we still need to learn some cyber lessons to protect America, suggest John McCain.
updated 5:39 PM EST, Mon December 22, 2014
Conservatives know easing the trade embargo with Cuba is good for America. They should just admit it, says Fareed Zakaria.
updated 8:12 PM EST, Fri December 19, 2014
We're a world away from Pakistan in geography, but not in sentiment, writes Donna Brazile.
updated 12:09 PM EST, Fri December 19, 2014
How about a world where we have murderers but no murders? The police still chase down criminals who commit murder, we have trials and justice is handed out...but no one dies.
updated 6:45 PM EST, Thu December 18, 2014
The U.S. must respond to North Korea's alleged hacking of Sony, says Christian Whiton. Failing to do so will only embolden it.
updated 4:34 PM EST, Fri December 19, 2014
President Obama has been flexing his executive muscles lately despite Democrat's losses, writes Gloria Borger
ADVERTISEMENT
ADVERTISEMENT
ADVERTISEMENT
ADVERTISEMENT