The New York Times

The five-year survival rate of lung cancer patients whose tumors were detected early by a CT scan is 80%, compared with 15% for patients who did not receive the CT scan. One problem with this statistic is that a CT group patient's 5 year clock starts earlier because his or her tumor is diagnosed earlier with the with the scan. So that while the tumors of a CT-scan patient and a non-CT-scan patient may start at the same time, and while the patients may lapse into cancer and die at the same time, the CT-scan patient will be said to survive more years from diagnosis. The other "numbers lie" problem is that about 4 out of 5 in the CT scan group will actually have inactive tumors that will not progress to cancer, and yet they are likely to undergo the dangers chemotherapy and/or surgery. The death rate for screened patients is actually higher than for non-screened, according to some researchers.

For a different article on this issue click here.



The New York Times, August 22, 2005 pA13(L)


Warned, But Worse Off.

by Steven Woloshin; Lisa Schwartz; H. Gilbert Welch.

Full Text: COPYRIGHT 2005 The New York Times Company

LUNG cancer seems to be the disease of the moment. The announcement that Dana Reeve, the widow of Christopher Reeve, has the disease -- coming just two days after Peter Jennings died from it -- has many Americans wondering if they should get tested with a CT scan. They should think twice. Wait, isn't CT screening one of those ''quick and painless tests'' that could save your life? And hasn't it been reported that people whose lung cancer is found early by such scans have a five-year survival rate of 80 percent, as opposed to 15 percent for the typical lung-cancer patient whose condition is detected later? Why not get scanned today?

The answer may surprise most Americans: we just don't know if lung cancer screening does more good than harm. While the benefits of screening are unproven, the harms -- one familiar, the other less so -- are certain.

The familiar harm is caused by false alarms. CT scans are great at finding abnormal areas of the lung. But while relatively few people have lung cancer, many have other lung abnormalities. After a positive CT scan, many are biopsied, and most will turn out not to have cancer. A lung biopsy is not a trivial procedure. Although serious complications are rare, the procedure may result in hospitalization (largely for a collapsed lung), and there have been deaths.

The less familiar, but more worrisome, harm comes from overdiagnosis and overtreatment. In the largest study to date, Japanese researchers using CT scans found almost 10 times the amount of lung cancer they had detected in a similar group of patients using X-rays. Amazingly, with CT screening, almost as many nonsmokers were found to have lung cancer as smokers.

Given that smokers are 15 times as likely to die from lung cancer, the CT scans had to be finding abnormalities that were technically cancer (based on their microscopic appearance), but that did not behave in the way most people think of cancer behaving -- as a progressive disease that ultimately kills. So here's the problem. Because we can't distinguish a progressive cancer from a nonprogressive cancer on the CT scan, we tend to treat everybody who tests positive. Obviously, the patients with indolent cancers cannot benefit from treatment; they can only experience its side effects. Treatment -- usually surgery, but sometimes chemotherapy or radiation therapy -- is painful and risky. Some 5 percent of patients older than 65 die following partial lung removal, and nearly 14 percent die with complete removal.

But wait a minute. Don't those compelling five-year survival statistics of 80 percent vs. 15 percent prove that CT screening works? The short answer is no. You have to consider exactly how a five-year survival rate is figured. It is a fraction. Imagine 1,000 people diagnosed with lung cancer five years ago. If 150 are alive today, the five year survival is 150/1000, or 15 percent. Yet even if CT screening raised the five-year survival rate to 80 percent, it is entirely possible that no one gets an extra day of life.

The best way to understand this paradox is to work through a thought experiment. First, consider a group of people with lung cancer who will all die at age 70. If they first receive the diagnosis when they are 67, their five-year survival rate would be zero percent. But if these same people had received their diagnoses earlier -- at, say, age 63 -- the five-year survival rate would be 100 percent. Yet death would still come at 70 for all of them. Earlier diagnosis always increases the five-year survival statistic, but it doesn't necessarily mean that death is postponed.

A second thought experiment helps further understand why CT scans, which find so many minute, nonprogressive tumors, inflate survival rates. Imagine a city in which 1,000 people are found to have progressive lung cancer following evaluation for cough and weight loss. At five years after diagnosis, 150 are alive and 850 have died: a five-year survival rate of 15 percent. However, if everyone in the city were screened with CT scans, perhaps 5,000 would be given a cancer diagnosis, although 4,000 would actually have indolent forms. These 4,000 would not die from lung cancer in 5 years, and the five-year survival rate would increase dramatically -- to 83 percent -- because these healthy people would appear in both parts of the fraction: 4150/5000. But what has really changed? Some people have been unnecessarily told they have cancer (and may have experienced the harms of therapy), and the same number of people (850) still died.

This is exactly what was found in a randomized trial of chest X-ray screening at the Mayo Clinic -- five year survival was higher for those who were screened (35 percent vs. 19 percent) but death rates were in fact slightly higher in the screened group. Consequently lung cancer screening with chest X-rays is not recommended.

Someday we will know if CT lung cancer screens help more than they hurt (the results of a major National Cancer Institute trial will be available in about five years). But until then, everyone should know that screening is a two-edged sword.