The Antidote: Lung cancer screening trial - how the news reports stack up

Today's New England Journal of Medicine presents the results of a study on lung cancer screening. This study purports to test the value of spiral computerized tomography (spiral CT) against lung cancer in smokers and former smokers, and reports a 92% 10-year survival rate, compared to 70% 5-year survival. Sounds pretty good, right? Well, not exactly. First of all, this study has no control groups - everyone in the study underwent screening, so the 70% number is a commonly accepted figure based on other populations. That's a problem because their diagnostic patterns differ.

Also, survival sounds intuitively like the flip side of mortality, which is the population-appropriate measure of the number or proportion of people who die from a disease, but it is not. According to a too-often-overlooked paper by HG Welch and colleagues, survival is not even a predictor of mortality. Instead, it is likely to reflect lead-time bias inherent in many screening tests - screening picks up smaller tumors early in their development, without necessarily changing the actual date of death. It's also likely that some screen-detected tumors would never go on to cause disease, and that has to be acknowledged as a risk of screening, because a false positive finding that leads to a lung biopsy can be dangerous.

I knew this study would get a lot of coverage; here's a look at a non-random, convenience sample of articles from big newspapers and wire services, and I'm not giving reporters' names because I know that editorial constraints often result in the loss of crucial details. So does mainstream coverage acknowledge the lack of control group and the 5-year survival problem?

A very good job of explaining the controversy from the NY Times:

...everyone in Dr. Henschke’s study had CT scans. And so, researchers say, with no comparison group of people who did not have scans, they are left with a pressing question: Does screening, in the end, save lives or cost lives?

“Intuitively, it makes sense —if you have a cancer, take it out,” says Dr. Stephen Swensen, a professor of radiology at the Mayo Clinic who conducted a study that was similar to Dr Henschke’s but smaller. “It makes sense that if you find a cancer earlier you will save lives,” he added. But Dr. Swensen said, “The science hasn’t backed that up yet.”

Cancer specialists have long known that all cancers - and lung cancers are no exception - include ones that stop growing and never kill or that grow so slowly that they never cause problems if they are simply left alone. So, some of them ask, how many of the people said to be cured were never in danger? And how often will people have operations that can involve removing part of a lung, and that can themselves kill a patient, when their cancer was not lethal? The problem, as with other cancer scans, is that science today cannot always tell the difference between cancers that will stop and those that will go on to kill.

The Washington Post also covered the lead-time bias and lack of control group issues well:

"We have shown that when you diagnose it early that you can cure it," said Claudia I. Henschke of Weill Medical College of Cornell University, who headed the study. "We think this provides compelling evidence that you can save lives."

Some who disagree say the study had no controls -- no group randomly assigned to have no screening or to have some other kind of screening. Without a comparison group, it is impossible to say whether the outcomes of the smokers who received CAT scans was better.

It is not surprising that people whose lung cancers are found by a CAT scan survive longer than people whose tumors are found the usual way, after they cause pain or breathing problems. Screened patients receive diagnoses earlier and therefore will have the diagnosis for a longer time than someone whose tumor is found later, even if the two people die at the same time.

from Reuters...:

But other experts immediately questioned the results and said the report was not the last word.

"What the data don't show us is that there been an actual decline in lung-cancer mortality. This study can't show us that because it is not a random comparison," said Dr. David Johnson, of Vanderbilt University in Tennessee, a former president of the American Society of Clinical Oncology.

It is possible the screening simply turned up more slow-growing tumors that would have caused little trouble for years, he said.

...and the Associated Press:

Henschke [the study's lead author] said the general population can be the comparison group, because lung cancer is so common and its survival odds are so well known. But many scientists disagreed, and said her study falls short for this reason.

"It raises great hope for CT screening," but it doesn't prove a benefit, said Dr. Denise Aberle of the University of California, Los Angeles, who is helping conduct a government-funded study that should give more definitive answers. It is screening 53,000 current and former smokers with CT scans or regular chest X-rays to see whether either can cut lung cancer deaths. The Mayo Clinic also is leading a screening study, and others are under way in Europe.

As an aside, I chuckled at the defensive wiggle-language in the AP report from the American Cancer Society's Bob Smith (ACS funded the study):

Even though the study lacked a comparison group, he said, "it's highly unlikely that this completely invalidates the observation of a favorable benefit from early diagnosis."

The Wall Street Journal acknowledged the controversy, but it was framed only by a quote from a lung cancer advocate worried that smokers were perceived to be to blame for their disease:

"It's hard to demand the same level of compassion and support as other cancers," Ms. Fenton said. "Mammographies, colonoscopies - you don't see the same doubts about these tests as you do for CT scans of the lungs."