April 7 , 2008
Nodule Characterization — PET Proves More Reliable Than CT
By Dan Harvey
Vol. 9 No.7 P. 20
A solitary pulmonary nodule (SPN) presents physicians and patients with an unpleasant surprise. Small, round lesions measuring less than 3 centimeters, SPNs usually turn up as an incidental finding during a routine chest x-ray. While most are benign, SPNs can indicate early-stage lung cancer. Thus, prompt and accurate characterization is vital to determining subsequent strategies.
Clinicians accomplish this characterization with either CT or PET. However, researchers involved in a large, multisite study found that PET proved the more reliable tool in a head-to-head comparison of the two modalities for characterizing SPNs. The prospective study—called the Solitary Nodule Accuracy Project (SNAP)—took place over several years at 10 VA hospitals throughout the country. Researchers published their findings in the February issue of the Journal of Nuclear Medicine.
The study, “A Comparison of the Diagnostic Accuracy of 18F-FDG [fluorodeoxyglucose] PET and CT in the Characterization of Solitary Pulmonary Nodules,” proved remarkable not only for its findings but also for its scope and thoroughness. The SNAP project represents the largest and most rigorous evaluation of PET in patients with SPNs to date, and it provided one of the few head-to-head comparisons between PET and CT for that indication.
PET vs. CT
In developing this prospective study, the researchers sought to address limitations of previous studies. As they point out, most trials that assessed CT accuracy in lung nodule characterization were performed more than 15 years ago, with technology and methodology that have become outdated. Also, researchers who conducted previous and similar PET studies limited themselves by small sample size, incomplete masking, and biased patient selection.
SNAP researchers differentiated their study in several ways. One unique element was its multisite nature, explains lead investigator James W. Fletcher, MD, a professor of radiology and director of the Nuclear Medicine/PET Imaging Center at the Indiana University School of Medicine in Indianapolis. “That kind of setting was very important, as other institutions often can’t reproduce findings that come from a single-institutional study,” he says.
Also, the larger patient population provided a significant advantage. The researchers had more subjects to work with so they saw a greater number of benign nodules than were present in other studies, a factor that helped them better assess diagnostic specificity. “In earlier studies, as far as PET performance, researchers hit the mark pretty closely for sensitivity, but the specificity remained unclear,” says Fletcher. “These [earlier] studies saw many subjects with malignant nodules but not as many with benign nodules. As such, the researchers couldn’t really assess a test’s specificity.”
In addition, SNAP researchers were resolute in avoiding patient selection bias. Their enrollment protocol called for patients with nodules detected by chest x-rays. Conversely, previous studies enrolled patients with SPNs discovered by CT. SNAP researchers reasoned that if they enrolled patients with nodules first detected with CT, they would have biased the study by increasing the prevalence of malignant nodules. As it turned out, their protocol resulted in a significantly lower prevalence of malignant nodules (53%) than typically seen in SPN characterization studies. “[This study design feature] also allowed for an unbiased comparison of the accuracy of PET and CT,” the researchers wrote.
Further, SNAP project clinicians worked in a completely blind situation. PET readers didn’t know the CT readers’ results and vice versa, says Fletcher. The protocol blinded the study’s pathologists so they were unaware of the imaging results. “Other publications touched on some of those things and satisfied some of them but didn’t satisfy all of them,” says Fletcher.
Researchers initiated the SNAP project in January 1999 when they began recruiting trial participants. By June 2001, Fletcher and fellow researchers had recruited 532 patients with newly diagnosed, untreated SPNs ranging from 7 to 30 millimeters in size on a posteroanterior and lateral view of plain chest radiography. Nodules smaller than 7 millimeters are hard to characterize with PET, while those larger than 30 millimeters are considered malignant until proved otherwise, the researchers noted.
As this was a head-to-head study, all participants underwent CT exams and PET studies enhanced with 18F-FDG. A masked panel, including three PET and three CT experts, performed readings seven times at a central location during a three-year period. Patient images were reviewed twice: once by a CT reader and once by a PET reader. These readers rated patient studies with a five-point scale: definitely benign, probably benign, indeterminate, probably malignant, and definitely malignant. Final diagnosis was established by biopsy or a two-year follow-up. Researchers established a definitive diagnosis for 344 patients.
In characterizing SPNs, researchers determined that PET and CT were similar as far as sensitivity, but PET outpaced CT when it came to specificity. Specifically, they found the following:
• Likelihood ratios—a summary ratio that tests whether subjects with a particular finding have a particular disease, in this case a malignant SPN—for PET and CT results for combined ratings of either definitely benign (33% and 9% of patients, respectively) or probably benign (27% and 12%) were 0.1 and 0.11, respectively.
• Likelihood ratios for PET and CT results for combined ratings of indeterminate (1% and 25%), probably malignant (21% and 39%), or definitely malignant (35% and 15%) were 5.18 and 1.61, respectively.
In explaining these findings, the researchers wrote that definitely and probably benign results on PET and CT strongly predict a benign SPN. But they pointed out that definitely malignant results on PET were more predictive of malignancy than were these results on CT. “A malignant final diagnosis was approximately 10 times more likely than a benign final diagnosis in participants with PET results rated definitely malignant,” they wrote.
Researchers also underscored how PET resulted in far fewer indeterminate test results based on the five-point scale. They pointed out that CT readers characterized 25% of nodules as indeterminate, whereas PET readers only characterized 1% of nodules in this way. Further, nodules classified as indeterminate on CT were correctly characterized on PET in more than 80% of cases (sensitivity, 83%; specificity, 89%).
“In other words, it was easier for the PET readers to be much more definitive about what they thought of a nodule,” explains Fletcher. “This was a very important finding because when you have an ‘indeterminate’ classification by CT alone, then you’re a bit stuck. It’s like you have to consider the nodule malignant until you prove it otherwise. Thus, indeterminate classifications push you toward a more aggressive management strategy than you’d ordinarily have to adopt.”
Researchers also found that interobserver and intraobserver reliability for PET was superior to that of CT. “This was another important point,” Fletcher says. “If your readers can’t agree, then you’re going to encounter difficulty in reproducing the results, no matter how well your test performs.”
Rank and File Readers
Researchers also compared results of the expert panel readers with those of the VA site readers, a study element that yielded interesting information. They found that neither the PET site reader results nor the CT site reader results differed significantly from panel reader results. “This was another important and unique aspect of our study,” says Fletcher. “We wanted to see how the performance of your day-to-day nuclear medicine and CT radiologists would match up against a body of experts. Sometimes, in some studies, experts are gathered and they perform quite well. However, when people in clinical settings try and emulate what these experts accomplish, they come up short. But with this comparison and its results, we demonstrated that our results are obtainable in a variety of settings and not just in a high-powered academic setting staffed with world class experts.”
SPNs represent a potentially curable form of lung cancer, which kills more patients than any other in the United States. Reiterating the importance of timely characterization, Fletcher says that when identified early, malignant SPNs can be treated more aggressively, thus improving a patient’s chances of survival. “The survival rate is much higher in individuals when their disease is detected at an early stage. Physicians can begin appropriate therapy before it metastasizes,” he says.
As the American Cancer Society (ACS) points out, lung cancer is the leading cause of cancer deaths in the United States. But the survival rate is as high as 49% in cases detected when the disease is still localized. However, only 16% of lung cancer cases are diagnosed that early, according to the ACS.
Fletcher says that for patients with untreated and undiagnosed SPNs ranging in size from 7 to 30 millimeters, PET provides more accurate identification of malignant nodules that would require the more aggressive approach.
During the study’s course, combined PET/CT imaging emerged. On the surface, this would seem to date the SNAP study. The researchers acknowledge the fusion imaging’s impact. “There is some evidence that integrated PET/CT is more accurate than dedicated PET for lung cancer staging,” they wrote. “In this regard, and in uncommon circumstances, the CT component might help to improve the performance of PET by identifying rapid growth consistent with an infectious process, providing alternate diagnostic hypotheses, demonstrating a typical pattern of bronchioloalveolar carcinoma, and similar advantages.”
Considering Hybrid Imaging
However, the emergence of PET/CT also validates much of what the researchers reported. For instance, they indicated that PET combined with CT could provide accurate identification of nodules that are most likely benign. As such, a watch-and-wait strategy can be adopted instead of biopsy or surgical procedures, which can be invasive and expensive. “PET is typically used as an adjunct to CT in the evaluation of suggestive nodules,” the researchers wrote. “Our findings support this approach.”
“I think PET/CT does validate what we’re saying,” says Fletcher. “People now realize that the two modalities are complementary. While we found that while PET is superior to CT in some aspects, we also found that the combination of the two modalities is far superior to either one used alone. PET information coupled with the CT characteristics enable us to be much more accurate as far as malignancy. Conversely, PET and CT tests enable us to better determine if a nodule is benign. In either case, PET and CT together help physicians develop appropriate management strategies.”
Fletcher notes that reaction to the study findings has been positive and generated a great deal of interest. He anticipates that additional publications will arise from other researchers using the SNAP study’s data set. Authors of these papers can even come from within the project participants, he adds. “We had a mix of people involved in this—biostatisticians, pulmonologists, nuclear medicine experts, and others—and different groups are quite interested in the various aspects of the data as it relates to their respective fields. We anticipate different publications that would look at the methodology, or the study design, or the economic aspects.”
One of the study’s coauthors, Michael Gould, MD, from the Department of Veterans Affairs in the Palo Alto Health Care System in California, is working with colleagues at the Stanford School of Medicine on the economics of PET in its role of managing SPNs. “They will be using our results, in terms of sensitivity, specificity, and accuracy, combined with an economic analysis,” reports Fletcher.
— Dan Harvey is a freelance writer based in Wilmington, Del., and a regular contributor to Radiology Today.