The results of two lung cancer screening studies published in the New England Journal of Medicine are good news for CT lung cancer screening, showing the ability to differentiate malignant from benign nodules for many patients before invasive biopsy procedures are considered.
In one study, an analysis of National Lung Screening Trial (NLST) results led by Dr. Denise Aberle, from the University of California, Los Angeles (UCLA), found that by the third annual screening, CT can distinguish benign from malignant nodules detected on earlier screening rounds because the benign nodules hadn't grown or changed over time (NEJM, September 5, 2013, Vol. 369:10, pp. 920-931).
In a second report in the same NEJM issue (pp. 910-919), Canadian researchers examining two cohorts representing nearly 3,000 individuals screened with CT found that analyzing certain patient and nodule characteristics was a good way to determine whether nodules were malignant or benign.
The findings could bolster confidence in CT's ability to distinguish malignant nodules without the need for invasive procedures, making CT lung screening less invasive -- and less costly.
Malignancy on 1st screening
NLST made headlines in 2011 when it reported a 20% reduction in lung cancer mortality from a cohort of 53,454 participants at risk for lung cancer who were randomly assigned to three annual screenings with either low-dose CT or radiography. The results led the U.S. Preventive Services Task Force (USPSTF) to give a positive recommendation to CT lung cancer screening in July.
But managing the many lung nodules detected on CT scans is one of the most difficult aspects of screening. And it's a high-stakes game. More than one in five participants in low-dose CT lung cancer screening programs are found to have at least one nodule, and more than one in four patients with detected nodules who undergo an invasive procedure turn out to have a benign finding.
In the study from Aberle and colleagues, the group analyzed NLST data to determine the effectiveness of the third annual round of lung cancer screening in the context of previous rounds. The study team from UCLA and several other U.S. centers found that by the third year, low-dose CT screening detected three times as many cancers as chest radiography.
Also by the third screen, CT's ability to reduce mortality from lung cancer was accompanied by a reduction in the number of advanced-stage cancers and an increase in early-stage cancers, the authors wrote.
The third round of CT screening (T2) is particularly valuable because it weeds out many false positives found at baseline (T0) and in the second round (T1) -- which have a low likelihood of reducing mortality over time, said co-investigator Dr. William Black, a professor of radiology at Dartmouth Medical Center, in an interview with AuntMinnie.com.
By the third year's scan (T2), for example, the percentage of screening participants with a positive scan had dropped substantially, from 27.9% in the second year to 16.8%. "The reason that happened was because a lot of the nodules that had been called positive through the second screen are now all of a sudden called negative on the third screen when it was shown they didn't grow," Black said.
At the T2 screening, the sensitivity of low-dose CT was 93%, specificity was 83.9%, positive predictive value was 5.2%, and negative predictive value was 99.9%, the group reported. The sensitivity of radiography was 63.9%, specificity was 95.3%, positive predictive value was 6.7%, and negative predictive value was 99.8%
Notably, overall compliance was very high: Well over 90% of participants came back for all three screening rounds, Black said. And the positive predictive value for small nodules -- i.e., 4 mm to 6 mm -- was very low, at less than 1% for screening rounds T1 and T2, again validating the hypothesis that small nodules lead to few cancers and little mortality, despite their large numbers.
"More than half of the positive findings were 4 to 6 mm in size, but less than one in 100 of those turned out to be cancer," he said. "And there was a very low percentage of cancers for the youngest and largest age group, 55 to 59 years."
Overall, 43% of NLST participants who underwent baseline screening were 55 to 59 years of age at study entry, but the rates of lung cancer among participants in this age group were just 0.4% in the CT group (0.4% at T1 and 0.7% at T2), the authors wrote.
Start screening later?
The paucity of lung cancers among younger screening subjects led to some speculation by the NLST authors "that maybe screening would be more effective and efficient if we started screening at an older age group and were less aggressive with the very small nodules," Black said. That decision, of course, touches on a larger debate encompassing costs, politics, and societal values, he acknowledged.
"In addition to these highlighted findings, this study supports everything else that was written about NLST -- that CT finds more cancer than radiography and at an earlier stage, in later screening rounds as in the first round, and that there is a reduction in advanced-stage lung cancer," he said.
Importantly, even though this study definitely shows that screening works, the massive amount of information contained in NLST hasn't been fully analyzed, and that means that optimizing protocols and targeting screening subjects will require a lot more number crunching, according to Black.
"We'll have to figure out what is the best way to screen people going forward, knowing what we've learned from NLST," he said. "I think everyone's on board saying we should do screening. But exactly how we do it is something we're going to need to investigate in the future. Rather than repeating a lot of randomized trials, this process will largely consist of analyzing existing data."
Malignancy model
The second NEJM study focused on the question of whether a model based on nodule characteristics can be used to predict which suspicious lesions found on CT lung screening should be followed up. The study authors included Dr. Annette McWilliams from Vancouver General Hospital; Martin Tammemagi, PhD, from Brock University; and colleagues from several other Canadian institutions.
Their goal was to find a way to reduce the number of biopsies that might be required from CT lung screening, reducing the risk of morbidity and mortality in screening programs, they wrote.
"In the NLST study, over 20% of people have an abnormal CT, some will need follow-up CT, and some will even get a biopsy performed," said co-author Dr. Stephen Lam. "So we tried to reduce the number of false-positive scans by using this prediction model to reduce it, potentially, to less than half the people that need follow-up."
The investigators used the Pan-Canadian Early Detection of Lung Cancer Study (PanCan) to develop a prediction model to determine whether a nodule detected on the first CT scan was cancerous or not, Lam said. There were 1,871 CT scans acquired in the study on MDCT scanners.
To validate the model, they applied it to a dataset from a study performed by the British Columbia Cancer Agency (BCCA), which involved 1,090 participants in trials sponsored by the U.S. National Cancer Institute.
The study tracked the outcomes of all nodules of any size that were detected at low-dose CT, and correlated nodule characteristics to whether they turned out to be malignant or benign.
"We looked at several clinical parameters like age, gender -- the female sex has a higher risk, and whether they have a family history of lung cancer," Lam said. The authors estimated risk by creating two models for analysis, one limited to malignancy predictors that were significant (model 1), and a fuller multivariable logistic-regression model that included additional variables thought to be associated with a higher risk of malignancy (model 2).
For example, the results showed higher odds of malignancy in women in both models (odds ratio [OR] of 1.79 in model 1, OR of 1.76 in model 2). Also elevated were the odds of malignancy in patients with a family history of lung cancer (OR of 1.35 in model 2) and emphysema (OR of 1.41 in model 2).
The analysis of nodule characteristics included nodule type (solid, nonsolid, or part solid), whether nodule margins were smooth or spiculated, and where the nodule was located in the lung, as well as whether there was any emphysema adjacent to the nodule.
Predictors of malignancy in the model included older age, female sex, family history of lung cancer, emphysema, larger nodule size, location of the nodule in the upper lobe, part-solid nodule type, lower nodule count, and spiculation, the authors concluded. Associations between nodule characteristics and malignancy are detailed in the chart below.
| Nodule characteristics and likelihood of malignancy | ||||
| PanCan | BCCA | |||
| Benign nodule | Lung cancer | Benign nodule | Lung cancer | |
| Mean nodule size (mm) | 4.1 | 15.7 | 3.6 | 13.9 |
| Nodule type | ||||
| Nonsolid or ground-glass opacity | 98.1% | 1.9% | 98.7% | 1.3% |
| Perifissural | 100% | 0 | 100% | 0 |
| Part solid | 93.4% | 6.6% | 77.8% | 22.2% |
| Solid | 98.9% | 1.1% | 99.4% | 0.6% |
| Nodule location (lobe) | ||||
| Left lower | 98.9% | 1.1% | 98.7% | 1.3% |
| Left upper | 98.1% | 1.9% | 99.4% | 0.6% |
| Right middle | 98.7% | 0.3% | 99.8% | 0.2% |
| Right upper | 98.5% | 1.5% | 98.6% | 1.4% |
| Mean nodules at baseline | 5 | 4 | 7 | 4 |
| Spiculation | ||||
| No | 98.9% | 1.1% | -- | -- |
| Yes | 84.8% | 15.2% | -- | -- |
Overall, the PanCan data revealed that 1,871 persons had 7,008 nodules, including 102 that were malignant, for a 5.5% rate of malignancy. In the BCCA data, 1,090 individuals had 5,021 nodules, including 42 that were malignant, for a 3.7% rate of malignancy.
Malignancy: More than just size
CT lung cancer screening should look at more than just nodule size when trying to determine which nodules to follow up, according to Lam.
"It's not size alone," he said. "Up until now, all of the guidelines used the size of the nodule to decide whether we needed to take action in terms of follow-up CT or biopsy. What we found is that size is a major parameter, but it's not the only parameter. In fact, 20% of the lung cancers develop from nodules that are not the largest one; it actually can be the second or third largest one. That's why it's important to calculate nodule risk individually and not just look at one nodule and assume that's the one that will develop cancer."
Although variables such as smoking history, body mass index, and lung expiratory studies can identify smokers at risk for lung cancer, these factors weren't independently associated with lung cancer in the fully adjusted model, the authors noted. They cautioned that the study cannot address individuals at very low risk not currently recommended for screening.
The group's next study will apply the model prospectively in a screening study to add another level of validation, Lam said, and the results will be used to tailor further investigation.
"People who have very low risk probably don't need a repeat scan for one or two years," he said. "For people who have very high risk, even though the nodule may not look that big, we may decide to move up the biopsy rather than repeat the scan in three months."
