Italian researchers believe they have struck a good balance in CT lung cancer screening by following up suspicious nodules at a slightly larger size threshold. By only following up nodules 5 mm and larger rather than 4 mm, they achieved good sensitivity with fewer false positives.
Several lung cancer screening trials have now shown that low-dose CT (LDCT) can reliably diagnose individuals at high risk of lung cancer, and that it can save lives. But a relatively high rate of intervention for suspicious nodules is usually required to demonstrate whether suspicious nodules are malignant or benign, leading to increased costs and risks of complications, wrote the authors from the European Institute of Oncology in Milan.
"This poor specificity may constitute an obstacle to the large-scale implementation of screening," wrote lead author Dr. Giulia Veronesi, along with Patrick Maisonneuve, Dr. Lorenzo Spaggiari, and colleagues (J Thorac Oncol, July 2014, Vol. 9:7, pp. 935-939).
CT screening sites can improve the situation by changing their nodule follow-up protocols, which govern which nodules get worked up and which ones don't. But few LDCT screening studies have looked at the performance of such protocols, possibly because they aggregate data from multiple centers, leading to "considerable heterogeneity" in nodule management.
The Italian researchers didn't have that problem in the Continuous Observation of Smoking Subjects (COSMOS) study, which performed LDCT screening on high-risk smokers at a single center. Therefore, they decided to assess the performance, invasiveness, and side effects of their follow-up protocol over a five-year follow-up period.
Annual LDCT screening
In 2004 and 2005, the study team screened high-risk smokers who had a minimum of 20 pack-years' smoking history and a minimum age of 50 with low-dose CT; individuals with negative results were screened annually for five years, the authors wrote. The mean age of the 5,203 participants was 57 years, and median smoking history was 44 pack-years.
Their follow-up protocol differed slightly from the protocol recommended by the Fleischer Society and which was used in the landmark National Lung Screening Trial (NLST) of following up individuals with nodules 4 mm and larger. This led to a very high recall rate, Veronesi and colleagues noted.
Instead, the COSMOS researchers used a 5-mm threshold, and individuals with nodules 5 mm or smaller were scheduled for a repeat CT scan a year later. Individuals with nodules larger than 5 mm but smaller than 8 mm were rescanned within three to six months, and those with nodules larger than 8 mm were referred for PET/CT follow-up.
The researchers evaluated their diagnostic protocol based on at least 12 months of follow-up by determining true positives, growing nodules diagnosed at stage I or greater, localized multifocal tumors, and new nodules at any stage. They counted false positives only when benign nodules were resected surgically; true negatives were represented by nongrowing nodules, both at the baseline scan and subsequently, the authors wrote.
"We always used PET/CT in the workup, and after the first years also assessed the growth rate of nodules," Veronesi told AuntMinnieEurope.com in an email.
In all, the study team performed 204 invasive diagnostic procedures for suspected cancer. Primary lung cancer was diagnosed in 175 patients, including 55 at the baseline scan, 117 in follow-up scanning rounds, and three as symptom-detected interval cancers (all small cell lung cancer). Just over 6% of patients had a second primary cancer. Surgery showed benign lung disease in 29 (14.2%) of 204 cases undergoing invasive diagnostic procedures.
Cancer types were as follows: 129 (73.7%) adenocarcinomas, 20 (11.4%) squamous cell carcinomas, 14 (8.0%) other non-small cell lung cancers, and 12 (6.9%) small cell lung cancers, the authors wrote.
High accuracy for biopsied lesions
The researchers then compared the COSMOS protocol to that of a subgroup of patients from their study who met the inclusion criteria for NLST (patients could be included in COSMOS with only a 20 pack-year smoking history, compared with a 30 pack-year history for NLST). They assessed the accuracy of each protocol for nodules that underwent invasive management.
| NLST vs. COSMOS for suspicious nodule management | ||||
| Protocol | Sensitivity | Specificity | Positive predictive value | Negative predictive value |
| NLST (≥ 30 pack-years, 4-mm nodule minimum) | 88.5% | 99.2% | 83.9% | 99.5% |
| COSMOS (≥ 20 pack-years, 5-mm nodule minimum) | 90.3% | 99.4% | 84.5% | 99.7% |
Only 6.4% of those screened required a procedure beyond low-dose CT, the authors reported. False negatives represented 8% of cases, and a few false positives were diagnosed at intervention.
Nearly all operable
Surgeons were able to resect tumors in 153 of 175 cases, or 87% of patients, the study team wrote. Three other subjects underwent pneumonectomy, 135 had lobectomy following their CT results, and 15 had sublobar resection. Forty-three of the lobectomies and five segmentectomies were assisted by robot.
The researchers found minor postoperative or postbiopsy complications in 19% of patients and major complications in 7%. One patient died of acute respiratory stress syndrome.
The COSMOS protocol earned good marks for sensitivity and specificity of LDCT for lung cancer screening, Veronesi told AuntMinnie.com, adding that only 1.3 individuals per 1,000 (6.4%) underwent invasive biopsy for benign disease.
"The routine use of PET/CT in the diagnostic protocol instead of CT-guided biopsy contributed to reduced invasiveness of the workup," Veronesi said. "The rates of delayed diagnosis and interval cancers were also very low."
And only 14 cases were misdiagnosed, meaning that the same lesions were present on the previous year's CT scans. These included five centrally located cancers and three that were fast-growing tumors.
Veronesi attributed the positive numbers to the COSMOS protocol's higher workup threshold.
"If we had lowered the cutoff to 4 mm, as in NLST, the recall rate would have been excessively high." Veronesi told AuntMinnieEurope.com. Although certainly at the 5-mm level, "some smaller tumors can be missed, especially aggressive tumors with a very fast growth rate," Veronesi said.
In fact, the results showed six cases where nodules 5 mm and smaller turned out to be malignant. Some researchers advocate even larger lesion size cutoffs to reduce false positives, Veronesi said. She cited a recent analysis of International Early Lung Cancer Action Program (I-ELCAP) results that showed a nodule size threshold of 6 mm can reduce the workup rate by 36% without any delay in diagnosis.
"This is strong evidence from reliable investigators," Veronesi wrote in her email.
The new study's main drawback is its lack of a control group, which the researchers felt would be unethical given the knowledge that CT screening is more effective than either x-ray-based screening or no screening at all.
One way to further improve upon the results is to use a more tailored approach to screening based on patient risk.
"One way of doing this would be to tailor the screening interval to the risk of an individual subject developing cancer, as determined by a risk evaluation algorithm," the authors wrote. Also unexplored in this study is whether more conservative surgical approaches might also be effective.
The use of serum markers for early detection of lung cancer in heavy smokers might also help reduce the number of false positives, Veronesi said.
"Around 269,000 deaths from lung cancer are expected in Europe in 2014, and although antismoking campaigns are having an effect and must be continued, it is now clear that screening can contribute to reducing lung cancer mortality," Veronesi and colleagues concluded.
