Figures that say cancer survival is better today than it was two decades ago for some cancer types may be little more than statistical mirages, Yale University researchers contend.

In a study published today, researchers say the outcome of treatment for lung cancer was about the same between 1953 and 1977 among patients treated at two Connecticut hospitals. The difference in survival rates between the two eras largely disappeared when statistical differences created by newer diagnostic procedures were eliminated.

"These results are distressing because they suggest that the contemporary improvement of survival rates, at least among patients with lung cancer, is a statistical artifact," Dr. Alvan R. Feinstein and his colleagues wrote in the current New England Journal of Medicine.

The result, they said, has been a "false sense of contemporary therapeutic accomplishment," stemming in part from new diagnostic technology that has allowed physicians to detect cancers at earlier stages than in the past, "extending the statistical length of a patient's survival without necessarily prolonging the duration of life."

There has been a debate in recent years between those who say significant treatment advances have improved cancer survival over time and those who charge that the end result -- cancer death rates -- has not changed.

The study is likely to provide ammunition to those charging that advances reported on paper have not changed the survival outlook for most patients. John Cairns of the Harvard School of Public Health called it a valuable "case history" demonstrating that apparent changes in cancer survival can be misleading. "Overall, I don't think the improvement in cancer survival is very large for the main categories of cancer," said Cairns.

However, Dr. Peter Greenwald of the National Cancer Institute questioned the validity of the Yale study. "I don't think the methods justify the conclusions," he said, adding that the sample was too small and had failed to separate out true advances in treating various forms of lung cancer -- including what he called "dramatic improvement" in treating one form of lung cancer that accounts for one-fifth of the total.

Lung cancer is the leading cancer killer in the United States, taking an estimated 125,000 lives each year. Federal figures overall suggest a slight improvement in five-year survival rates, from about 6 percent in the early 1950s to about 12 percent in the late '70s.

The researchers, however, said that new technologies for visualizing cancers before there are apparent symptoms had shifted the timing of diagnosis and treatment and had biased survival statistics without changing the medical consequences of the disease.

The Yale researchers dubbed this the "Will Rogers phenomenon," citing a quip about westward migration during the 1930s that is attributed to this well-known American humorist: "When the Okies left Oklahoma and moved to California, they raised the average intelligence level in both states." Rogers was suggesting that those moving were among the least smart in Oklahoma, but still smarter than Californians.

The researchers said a similar, statistical migration occurred in which cancer patients who previously would have been assigned to a less-advanced stage of disease were now assigned to a later stage. This would mean that advanced cases that were hidden to early diagnosis in the 1950s and '60s would now be moved out of the "early stage" category, so survival statistics in this early-stage group would improve in 1977. In addition, advanced cases would be moved into the more severe categories, but because these patients were not as sick as most patients in the "advanced" illness group, survival statistics here, too, would improve.