Dr. Barry Kramer accepts plaque in memory of his mentor, Dr. Daniel Ihde

“A lot of things seem to make sense in lung cancer screening,” he said, “and that’s why a lot of people don’t want to wait for direct proof. There’s an assumption that things will work out, and that’s not always true.”

Kramer’s talk, “Lung Cancer Screening: The Clash of Science and Intuition,” was the third annual lecture in memory of Dr. Daniel Ihde, NCI deputy director from 1991 to 1994 and a renowned clinical researcher in the field of lung cancer. Special guests were Ihde’s widow Mary and son Steven.

The leading cause of cancer death in American men and women, lung cancer kills more people than all colon, breast and prostate cancers combined. And while there are two common ways of detecting lung cancer— spiral computed tomography (CT) and standard chest X-ray (CXR)—neither test has been shown to reduce a person’s chance of dying of the disease.

There is no recommended screening test to find lung cancer before symptoms begin.

Kramer, NIH associate director for disease prevention, is medical officer for the National Lung Screening Trial, which includes 50,000 men and women at more than 30 study sites across the country. The NLST, sponsored by the National Cancer Institute, is comparing spiral CT and standard CXR.

Joining Kramer at the Dr. Daniel C. Ihde Memorial Lecture are (from l) Ihde’s son Steven and widow Mary.

He is also part of the investigating team for the Prostate, Lung, Colon and Ovarian (PLCO) Cancer Screening Trial, which includes 37,000 men screened for prostate, lung and colorectal cancers and 37,000 women screened for lung, colorectal and ovarian cancers. PLCO is comparing CXR and usual care.

Although screening may yield an earlier diagnosis, Kramer explained, tests don’t necessarily help the person being screened. This is borne out by analyses of mortality and survival rates. And while these may sound like two sides of the same coin, they are actually different coins.

“Survival” looks at how long a group lives after being diagnosed. “Mortality” is the probability of death among the entire population at risk for a given disease or cause of death. The terms are defined within epidemiology, the methodological basis of public health, showing how health and disease occur across populations over time.

“Here’s a stark demonstration,” Kramer said, reviewing cancer statistics over the last half-century in the U.S., where changes in cancer survival rates and cancer mortality rates showed no relationship at all. “If I tell you what the trends are in cancer survival,” he said, “I’ve given you absolutely no information about what the trends are for the same cancer mortality, even if you think I have.”

The PSA test for prostate cancer, for example, increased survival rates simply because so many new cases were discovered—enough to create “a pseudo-epidemic.” Yet death rates did not drop substantially compared to 1973.

Could the same thing happen with lung cancer screening?

Kramer urged researchers to use an analytical framework, a “causal pathway,” because no matter how astute we are, we can be fooled by three kinds of bias: selection bias, length bias and lead-time bias.

For example, lead-time bias muddies the difference between survival and mortality. Here’s how. Imagine, Kramer said, a theoretical cancer that always kills on the fourth anniversary of its diagnosis, so the 5-year survival rate will be 0 percent. “Now along comes a screening tool, and the only thing it does is advance your date of diagnosis by 2 years, so you know about it earlier.”

And the 5-year survival rate? “It has gone from 0 percent to 100 percent...and you haven’t saved a single life,” said Kramer. “And yet what do you read about in the newspapers...and in prestigious journals about a new screening test? That is, the survival rate jumped dramatically from what would have been expected if you hadn’t been screening these people. Unfortunately, it gives you no information about what you can anticipate happening to the risk of actually dying of lung cancer. It’s just like Snidely Whiplash”—the mustachioed arch-villain who ties his victim to the tracks.

“When the 5 o’clock train comes by,” Kramer explained, “it will kill this victim. Now you develop a new screening tool, say binoculars. They can make the diagnosis of train a lot earlier, but it does not change the moment of impact.”

The goal of early detection is to reduce cancer mortality and morbidity—death and disease—while avoiding the added risks of the screening and related treatment.

“Every time we develop a new screening test, we not only change the natural history of the disease itself,” Kramer said, “we introduce new risks.” These are the minor risks of the tests themselves; overly aggressive treatment; treatment for one problem increasing risk for another; psychological changes; and overdiagnosis, “when you feel you have cured people that didn’t need a cure in the first place.”

He also pointed out screening’s societal outcomes and tradeoffs, including health care utilization and cost effectiveness. “Traditionally we’ve never had to ask the two completely separate questions,” he said. “Does it work? Should we do it? Countries with health systems know these are two separate questions.

“But we don’t have a health care system,” he quipped, “we have a health care situation.”

Any screening benefit has to overcome what he termed “the front-loaded harm of the mortality of surgery.” But can there be overdiagnosis in a fast-growing cancer like lung cancer? In fact, the Mayo Lung Cancer Screening Project of 9,200 smokers showed that patients were “heavily overdiagnosed,” Kramer noted. “The number of lung cancer deaths in the screened arm was slightly higher than in the control arm, although not enough to be statistically significant.

“This all justifies holding our fire until we get results,” he concluded. “The take-home point: recognize that survival and mortality in cancer have nothing to do with each other.” Randomized controlled trials are the most efficient way to get the information we need, Kramer said.