Thursday, November 4, 2010

$180,000 To Save One Life In the National Cancer Institute's National Lung Screening Trial on Helical CT Scanning For Early Detection of Lung Cancer

In its many years of clinical practice, the Disease Management Care Blog watched many ex and active smokers succumb to lung cancer. So, when a flurry of news reports today announced the "huge" finding that "spiral CT" screening saves lives, it was motivated to look at the story behind the story.

Lung cancer typically starts out as a "silent" tumor. Generally, by the time it is big enough (about 1 centimeter in diameter) to be seen on regular chest x-ray, a year has passed and the tumor may have already spread. Persons typically don't come to medical attention until the cancer is much larger, which makes the prognosis even more grim: after about 5 years, only 5% to 14% of victims are still alive.

Enter "Computed Tomographic (CT) Scanning," which uses computers to array x-ray imaging that, in turn, construct highly detailed and two (and lately three) dimensional images of the human body. Because lung tumors are solid and, in CT images, contrast nicely with the surrounding air-filled chest tissues, this imaging approach has the potential to find cancers earlier and before they've spread. Recall, however, that CT scans use a much higher dose of x-rays, which could theoretically cause cancer in a small fraction of patients. That risk was decreased by the advent of fast (within a single deep breath), low dose "helical" or spiral scans, where the patient quickly moves through the circling x-ray detectors. There's a good picture of what happens here.

The hope of earlier detection translating into chest surgery and saved lives was the basis of the National Institutes of Health (NIH) funding a large randomized trial (how the study was set up is described here) that recruited former (within the last 15 years) as well as current smokers (with at least 30 "pack-years") who were cancer free and between the ages of 55 to 74 years.

They were assigned to one of two treatment arms: 1) three yearly helical low dose scans or 2) three yearly chest x-rays. The "National Lung Screening Trial" began in August of 2002 and by April of 2004, 53,456 had been randomized in 33 U.S. research centers. The latest round of screening CT or chest x-rays was completed in the summer of 2007.

Once that was completed, it was a matter of counting death certificates.

So, after ALL that, is the counting done? Not exactly. The story behind the story is that preliminary summary results are available in a National Cancer Institute press release. A complete and detailed analysis will not be made available to the DMCB or its readers until....

"A fuller analysis, with more detailed results, [is] prepared for publication in a peer-reviewed journal within the next few months."

So exactly what does the press release say? In the first paragraph, it says there was....

"20 percent fewer lung cancer deaths among trial participants screened with low-dose helical CT."

This "20%" does seem "huge," doesn't it? Fortunately, the DMCB has been trained to look past this seductive relative risk reduction (i.e., comparing one group to the other group) for the absolute risk reduction. To the New York Times' credit, they reported the actual numbers from the press release, which noted that 354 CT scan patients died of lung cancer vs. 442 of the chest x-ray patients.

But what is the denominator? The DMCB can't find that number in the press release. Assuming, however, that the random allocation was about equal, that's 26,728 assigned CT scanning and 26,728 assigned chest x-rays. If that's true, the absolute cancer death percentages were 1.3% in the CT scan group and 1.7% in the chest x-ray group, for an absolute difference of about .4% or four in a thousand. More than 98% of both groups hadn't died of lung cancer.

At first glance, that seems like an awful lot of CT scans for a less than 1% absolute risk reduction. To put the numbers in perspective, the DMCB likes to rely on the "number needed to treat" statistic. If the numbers are inputted here, about 300 yearly CT scans will be needed to prevent one cancer death.

According to this NCI web site, helical CTs can cost between $300 and $1000. The DMCB assumes that not all of the patients assigned to the CT arm of the study got three scans (which would cost between $900 and $3000), so it'll make another assumption: the average number of CT scans was two and that the average charge was on the low side at $300 for a total of $600. If the NNT was 300 and each patient got $600 worth of CT scanning, that back of the envelope blogging calculation suggest $180,000 has to be spent to save one life from cancer.

The DMCB suggests this puts things into perspective, which so far has been missed by the national mainstream media. While there was a "20% reduction," a first time review of the press release suggests:

1) The vast majority (more than 98%) of participants did not die from lung cancer.

2) By going through yearly helical CT scanning, smokers can reduce their absolute risk by about .4% over 3 years. This is risk reduction at the margins of being meaningful.

3) The NNT is 300 and the corresponding cost of saving one life is $180,000. It's difficult for the DMCB to translate that into QALYs, but at first glance and given our budget travails, that may not be a wise use of our national treasure. To put things into perspective, consider how many long term (30 pack year) smokers there are among the over 90 million former and current tobacco users in the U.S. today, and multiply that times $600.

4) Based on these results, it's unlikely mainstream health insurers will think this is a wise use of their customers' premium dollars. It might be smarter public policy to let smokers pay for their own scans out of pocket. If they can afford a year's worth of tobacco products.......

One last DMCB point: there may be "lead time bias." Just because a cancer is found earlier in the course of disease, doesn't mean the patient is destined to live longer. It may take additional years of follow-up to see if the absolute death rate in the CT cancer group catches up to chest x-ray group.

Was the national news media beguiled by a press release that spun relative risk without the necessary information to make an informed judgment? The DMCB will find out when the complete study goes through peer review and gets published.

Stay tuned.

November 9 addendum: Thanks to a heads-up from a smart colleague, the DMCB became aware of this blogger Gary Schwitzer of HealthNewsReview also spotted the problems and has a good link to an useful NPR story.


GlassHospital said...

This post will be required reading for all medical students, especially the ones that tell me I now have to order CTs on all the smokers and ex-smokers that I see.

HeyBeNice said...

What's $180,000 to a specialist anyway? A drop in the bucket - that's what. You would make more headway with a less-shallow approach.

Glenn Laffel, MD, PhD said...


Thanks for adding a rare bit of sound reasoning to the coverage of this new finding about Chest CT scans and lung cancer.

Actually, I think you've understated the down-side risk associated with the new screening tool, perhaps by quite a bit.

Apparently in this study, the CT scans were associated with a whopping 25% false positive rate, which necessitated who knows how many repeat CT scans, ancillary diagnostic tests and even, perhaps, thoracotomies to track down what turned out to be benign lesions.

Adding the indirect costs associated with these false positive results would drive up your calculation for costs per life saved quite substantially.

Glenn Laffel, MD, PhD
CEO, Pizaazz

Anonymous said...

Maybe you should also stop putting people on statins then... Crestor, for example costs over $500,000 per life saved. Mammography is between $100,000 and $200,000. Screening is expensive. If you want to save money in healthcare, stop screening and let more people die. If you value extending life, then you need to pay the costs of doing so.

Jaan Sidorov said...

Hey... Be nice BeNice; that $180K is spread out among multiple patient encounters involving 300 patients. No single doc sees that money. However, Glenn Laffel points out that the false positive rate (nodules that LOOK cancerous, but aren't) is leading many other patients to undergo potentially dangerous - and individually pricey - invasive biopsies.

Anonymous makes a good point and the DMCB has addressed lipid screening (sort of) here YES, lipid lowering agents ARE oveprescribed; in fact they should only be used inpatients with established heart disease or if the chance of a heart attack is greater than 10%. Ultimately, the only way to reconcile dollars and years saved will be a metric like $ per QALY. Politicians don't like that, the medicalindustrial complex fear that and most persons don't understand it. But its time is coming....