Lung CAD: A Closer Look

Can software boost early detection for lung cancer enough to reduce the death rate?  Probably, eventually, but the studies haven’t shown it just yet. And few want to pay extra for it — not hospitals or radiologists or insurers — until solid proof is in hand. But in its clinical capabilities, there are for sure believers.

Lung cancer is the No. 1 cancer killer in the U.S. — mainly because it’s usually diagnosed too late to offer effective treatment. The trick is to detect small lung nodules — 30 millimeters or smaller — and distinguish the benign ones from those that could be malignant. Then the suspicious ones have to be tracked to see whether they are growing, and how fast. All of these things are hard to do.
 
Only 16 percent of lung cancers are diagnosed at an early enough stage that the available therapies can give the patient a 50-percent shot at surviving for five years or more, according to the American Cancer Society.  

With those odds, a tool to increase the frequency of early diagnosis ought to be welcomed. Computer-aided detection for lung applications — lung CAD — has been approved by the FDA since 2001 for chest radiography, and since 2004 for computed tomography lung scans. Vendors large and small offer products either as stand-alone workstations or integrated into CT scanners or radiography systems (see sidebar on page 44). The applications flag suspicious areas on a scan for the radiologist to take a closer look. They can act as a check, or second read, on the radiologist’s initial evaluation. On follow-up scans, CAD can be used to track whether lung nodules have changed in size or shape.
 
Though both radiologists and market observers expect lung CAD to be part of the standard of treatment someday, it hasn’t generated a groundswell of enthusiasm so far, for a variety of reasons.


Solid potential

Melissa Ginsburg, MD, director of general radiology at Memorial Sloan Kettering Cancer Center, in New York, has looked at CAD for lung CT, and thinks it has the potential to be helpful, but solid research is lacking. “A couple of studies have come out, but we’d like to see a few more that prove accuracy,” she says.

Existing studies on CAD for lung CT aren’t adequate to draw any conclusions, says a recent study by ECRI of Plymouth Meeting, Pa. Most of the studies were too small, the authors said, and many used retrospective techniques, like having the CAD software analyze existing images where the correct interpretation was already known. A “gold standard” study — prospective, with large numbers of subjects and a control group — hasn’t been published yet. The ECRI report didn’t address the use of CAD with radiography images.     

There are fewer than 200 lung CAD units in use in the U.S., including both radiography and CT applications, according to Frost and Sullivan. About 40 percent of the units are used with CT, and the balance with radiography.  In contrast, the firm says that CAD for mammography — the earliest application of CAD, approved in 1998 — boasts almost 3,500 units. 
 

Eyeing growth

“More studies need to be done on lung CAD,” says Subha Basu, a Frost and Sullivan industry analyst for healthcare and medical imaging. “Mammography CAD has proved itself. We have quite a bit of data on picking up breast malignancies. With lung CAD, those studies are just being initiated. In the next three to four years, we’ll see how good the products are.” He predicts 20 to 30 percent growth per year for lung CAD until 2012, as studies are released and use becomes more widespread. 

Another reason for the large discrepancy in market penetration between mammography and lung CAD applications is that, unlike chest x-rays or lung CT, mammography is used for routine screening, so the procedure volumes are much larger. There’s considerable scientific debate as to whether lung screening would accomplish the desired goal — catching cancers while they’re still curable — in enough cases to make the enormous cost worthwhile.
 
A study of 31,000 current and former smokers published in the October New England Journal of Medicine suggested that lung CT indeed can help identify early-stage cancers, but the study lacked a control group and won’t settle the issue on its own. The National Cancer Institute’s National Lung Screening Trial, a prospective, randomized and controlled study of 50,000 current and former smokers, is intended to measure the value of lung screening, for both radiography and spiral computed tomography. But results aren’t due until 2009. Until then, the financial pressure not to do routine screening is likely to continue, and therefore the market for lung CAD will have limited room to expand.  

Jason Launders watches the lung CAD market for ECRI. He says that reimbursement will be the key to creating a market for the technology — and without it, there will be no market to speak of. “Under the current climate, asking [the Center for Medicare and Medicaid Services] to increase what they have to pay is just not going happen,” he says.


Gaining reimbursement

As of the beginning of 2006, lung CAD for radiography has taken the first step toward becoming a reimbursable service. Through action by Riverain Medical, the pioneering vendor of lung CAD software for radiography, the service has been assigned a category III CPT reimbursement code by the American Medical Association, which administers the coding program. Category III codes are reserved for new technologies, and are a “tracking code” used to measure whether and how a procedure or service is being used by clinician, and they expire after five years.  They’re not as good as category I codes, which are permanent, but can often get a provider reimbursed to some extent, says reimbursement consultant Michael Longacre, who assisted Riverain in getting the code approved. 

“Payers don’t pay for things that don’t exist,” Longacre says. “A lot of providers think a Category III code is a scarlet letter for an experimental treatment and that payers will automatically deny it, but you can’t get reimbursed at all until you have a code.”   The other crucial step is to get a value assigned through the AMA’s Resource-based Relative Value Scale (RBRVS), a system that calculates the resources used to provide a service and comes up with a base value that providers can use in computing their charge.  The value assigned to the lung CAD code is $20, Longacre says.  

He’s working with several providers who are trying out the new code. Only one so far has filed enough claims to establish a track record, but its results are promising.  When the code was used correctly, the claims resulted in reimbursement at the requested rate about a third of the time. The other two-thirds were evenly split between denials (either outright denials or contractual adjustments) and claims that are still pending.
 
Even with the question of direct reimbursement unsettled, CAD still could gain a foothold if it were shown to save time for each reading. Some studies show that it cuts the time 70 to 80 percent for each reading, says Frost and Sullivan’s Basu.
  
It’s likely that CAD will take off as a feature bundled with a CT scanner or radiography system, especially if it also integrates with PACS products, says Jason Launders of ECRI. While customers may resist paying $150,000 for a stand-alone workstation application, they might pay somewhat for a system that incorporates CAD as a feature. Most of the stand-alone CAD vendors either have or are seeking partnerships with imaging system vendors.


Clinical believers

EDDA Technologies, whose IQQA-Chest CAD application for digital radiography was approved by the FDA in 2004, has almost intentionally held back its sales as a standalone system. “The real benefit is found with an enterprise solution,” says Tim Ketchmark, vice president of marketing and sales. “A standalone system doesn’t make any sense in a PACS environment.” Philips Medical Systems licensed the company’s technology earlier this year for integration with its digital radiography products.  
 
Even at this early stage, lung CAD has its fans. David Mendelson, MD, associate professor of radiology at Mt Sinai School of Medicine, uses ImageChecker CT Lung CAD from R2 Technology, now part of Hologic, and relies on it as his second read. “I know it definitely catches things I’ve missed on occasion,” he says. “There has to be some equivalent of a second read, and CAD is as good as another human being.” The sheer volume of images produced by a CT scan makes it virtually impossible for a human reader to find every region of interest, Mendelson says. “There’s a big fatigue issue on reading any CT study.  The R2 tool is not perfect — it has a few false positives — but it’s effective and easy to use.” He believes it should be part of the standard of care even now, though he concedes that reimbursement issues will have to be resolved first. 

The most useful aspect of the technology is not the initial detection of nodules, but the ability to track their growth, says Joanne Shepard, MD, director of thoracic radiology at Massachusetts General Hospital in Boston. She also uses ImageChecker. “Most nodules are not malignant, and we need to follow them from scan to scan to see if they’re stable or growing,” she says. The measurements that we can attain in the axial plane are not necessarily completely representative of size, because the nodules are three-dimensional. We need a way to measure their volume, and these CAD systems provide that. I think that as the programs are refined and are available on the user interface on PACS workstations, that they will be used as a routine double-read eventually.”


Conclusion

Despite her current hesitations, Ginsburg of Memorial Sloan Kettering is optimistic about the future of lung CAD. “The technology is out there and it’s pretty imminent,” she says. “I think it will be like spell-check.”