And that's exactly what he found. "I do observe essentially that the cancers where there has been the most innovation have seen the largest reductions in mortality," Dr. Lichtenberg said.

Of the nearly 13 percent drop in age-adjusted mortality rates between 1996 and 2006, Dr. Lichtenberg discovered that the lagged effects of advanced imaging accounted for nearly 40 percent of the decline, with drug innovation accounting for around 27 percent of the drop. Although the incidence of cancer declined in this period, too, this only accounted for around 7 percent of the decline in mortality, Dr. Lichtenberg observed.

Intriguingly, by his estimates, Dr. Lichtenberg suggests that advances in imaging have so greatly impacted cancer survival, they could be responsible for nearly 9 percent of the decline in mortality from all causes in the period studied.

OVERCOMING BIAS

Dr. Lichtenberg's research contradicts earlier findings that suggested new drug advances and the adoption of new imaging modalities made only a small dent on cancer death rates. But Dr. Lichtenberg said by counting up all deaths from cancer at the sites instead of relying on a 5-year survival rate, which some earlier studies used, he was able to avoid one of the great plagues of long-term cancer research, the so-called lead time bias.

"There's this issue that if you have advances in diagnostic techniques, people might appear to live longer after diagnosis, but only because they're diagnosed earlier. I avoid that problem by looking at the unconditional mortality rate -- simply looking at how many people die from each type of cancer, rather than the probability of dying within 5 years of diagnosis," Dr. Lichtenberg said.

Still, he admitted the data might not be entirely free from error. The incidence of cancer was determined using NCI's data set, which only represents about one-quarter of the U.S. population, and might be subject to sampling errors, he said. And the data on diagnostic imaging and pharmaceutical usage come from MEDSTAT, which excludes Medicare patients, the most likely to require advanced imaging and come down with cancers. Nonetheless, Dr. Lichtenberg does not believe this would have skewed the results.

"Even if there tended to be a difference in survival trends between elderly and non-elderly, that would not result in any bias, unless that difference varied across cancer sites in a way that was correlated with treatment innovation," he said. "I don't think there's any compelling reason to think there would be such a bias."

More Industry Headlines