Just as we went to publish this review, we noticed that USA Today had published an updated version of the story one day after the original story was published. To be consistent with our reviews of the AP, New York Times and Reuters stories – all of whose first-day versions were reviewed, our comments pertain only to the original USA Today story.
Later versions did fill some gaps: adding absolute risk reduction, adding cost information, and adding cautious perspectives from Dr. Otis Brawley.
But we still haven’t seen as good an analysis as that done by Richard Knox of NPR – with his next day followups.
Cancer screening is a complicated and — as other stories about this study noted — controversial topic. Of course we all want better ways to identify cancer early, but screenings can be more of a costly comfort than they are a clinically proven lifesaver. Stories like this need to take extra care to quantify both the benefits and harms of screening and to provide the data in both relative and absolute terms to help readers understand the strength of the evidence.
The original USA Today story did not give the cost of such scans. It also didn’t project the potential cost of followup testing and treatment when false positives arise. It did note two important pieces of cost information: “Insurance companies don’t pay for lung screenings, according to the National Cancer Institute.” And, the “Medicare program plans to review the study results to decide whether to cover screenings for high-risk people, says Harold Varmus, National Cancer Institute director.” The study results suggest that about 300 people need to be screened to identify one person with early stage lung cancer. At $300 to $1,000 a scan, identifying who is best served by the screening and how often they should be screened is essential.
The original version of the USA Today story only quantifies the benefits in relative terms, and here we really needed to see the absolute numbers (something added in a version one day later). We are told, for example, that the study covered 53,000 people and that deaths were reduced by 20 percent. At first blush, one might think this means that the study saved 10,600 people. That’s because we are never told how many people actually died during the study period. Nor are we told how many people died in each study cohort. The study itself does not spell out how many people were in each cohort, but we did some rough calculations based on the study design report. The report showed that the total number of participants enrolled was 53,456. Assuming that 50 percent were put in each group, that would mean each arm had 26,728 participants. Based on the reported death numbers, this would mean that 354 people died in the CT scan group, or 1.32% and 442 died in the X-ray group, or 1.65%. That’s an absolute difference of 0.33%, not quite as exciting as a 20% reduction in deaths. Also, the number of people needed to treat to save one life (or NNT) is 304. That may be a justifiable number for added investment in spiral CT scans, but it should be part of the discussion.
The story does not quantify the potential harms, but it at least acknowledges them in several parts of the story, which is more than some of the other coverage did. “Patz says patients should think carefully before requesting a screening. While CT scans are painless, they expose patients to radiation that can potentially cause new cancers, Patz says. It’s possible that a person scanned frequently for lung cancer could develop breast cancer as a result.” The story also indicates that the screenings could lead to more potentially dangerous surgeries. “The screenings also don’t diagnose cancer, says Michael Unger, a professor at Philadelphia’s Fox Chase Cancer Center, who wasn’t involved in the study. While screenings can detect potential problems, doctors need to retrieve actual cells to confirm cancer.Making an incision through the ribs to perform a lung biopsy is a serious operation and poses significant risks of its own, Unger says.
The story made a good attempt to evaluate the quality of the evidence and to present readers with reasons to be cautious. It says, for example, “But Patz notes that doctors and healthcare leaders will have to look at the results carefully before recommending mass screenings.” But the story also was hampered by a number of factors, the main one being that it does not break down for readers all of the data necessary to evaluate the strength of these study results. For example, it never says how many people died during the study period, even though it offers the important context that “Significantly, 75% of deaths in the study were from causes other than lung cancer.” Nonethless, we’ll give it the benefit of the doubt.
The story avoids disease mongering and does a good job explaining how lung cancer fits into the universe of smoking-related deaths. It says, for example, “Nearly 160,000 Americans a year die from lung cancer, the leading cause of cancer deaths in the world, according to the American Cancer Society.” Later it says that “smoking-related illness, which kills more than 400,000 Americans a year, Johnson says.” And, the story added that “a screening may catch lung cancer early, but a patient could still die young from heart disease, emphysema, a stroke or other smoking-related condition.”
The story did a great job bringing in independent voices, and, interestingly, some of the most cautious comments in the story come from Ned Patz, a professor of radiology at Duke University who helped design the study.
The story did not adequately compare spiral CT scans with X-rays for many of the reasons cited above.
By saying that “Any hospital can perform the scan,” the story overstates the case a bit. The study focused not on all computed axial tomography (CT) scans but on what is known as a spiral CT scan. These are not as widely available as other CT scans, and other stories put their availability at 60%.
The story led with the fact that this was the first time a large study showed this kind of benefit from lung cancer CT scans.
The story did not rely on a news release.