Who should be screened?
The positive effect of lung cancer screening depends, to a great degree, on the prevalence of lung cancer in the screening population. In the NLST, only persons between 55 and 74 years of age and a smoking history of more than 30 years, or former smokers who quit smoking within the previous 15 years, were included (9). These inclusion criteria defined a study population with an estimated risk of developing lung cancer, ranging from 2% to more than 20%, within 10 years (10). The positive effect of lung cancer screening could be increased even further by adding additional inclusion criteria, such as gender, passive smoking history, history of pneumonia, history of non-pulmonary tumors, or occupational exposure to asbestos.
Using the data from the NLST, a risk prediction model for lung cancer death was recently published, which used the risk factors of age, body-mass index, family history of lung cancer, pack-years of smoking, years since smoking cessation, and emphysema diagnosis to estimate the five-year risk of lung-cancer death (11). This retrospective study confirmed that the number of prevented lung-cancer deaths increased with increasing risk quintiles (11). In the quintile with the lowest risk, only very few deaths (1%) would have been prevented. In fact, 88% of the prevented lung-cancer deaths were distributed among the three quintiles with the highest risk (11).
The impact of a more sophisticated risk model on the effectiveness of lung cancer screening is currently being investigated in the ongoing British UK lung screen (UKLS) trial. In this trial, only patients with an at least 5% risk for developing lung cancer within the next five years are included. The risk for developing lung cancer is estimated using a model developed in the Liverpool Lung Project (LLP) (12). The LLP risk prediction model includes age, sex, smoking duration, family history of lung cancer, history of non-pulmonary malignant tumor, history of pneumonia, and occupational exposure to asbestos to estimate the lung cancer risk (12). It is projected that, by using these inclusion criteria, the prevalence of lung cancer in the screening population will be twice as high as in the Dutch NELSON trial (13).
At what intervals should the screening be planned?
The screening interval has a direct impact on screening performance, as well as overall costs and the cumulative radiation dose. Long screening intervals carry the risk that, aggressively growing tumors, in which the interval between the origin of the tumor, its detectability by CT, and the point at which it manifests, is quite rapid, may not be detected in early stages. Thus, screening would detect mainly indolent, slowly growing tumors. However, short screening intervals increase the probability of detecting aggressive cancers with the shortcoming of increasing the overall costs and cumulative radiation dose.
To date, most of the prospective screening trials were designed with annual screenings for three or five years (1,6,8,14). However, although not yet investigated in a prospective trial, biennial (twice-yearly) screening could have the potential to be more cost-effective than annual screening. A current prediction model based on the UK lung cancer screening eligibility criteria and the NLST data suggests that the intervention effect of biennial screening could indeed justify the human costs (15). Prospective trials will be necessary to further investigate the effect of biennial screening on survival.
How should detected nodules be managed?
One of the major challenges of lung cancer screening is the high incidence of detection, coupled with a very low proportion of malignant nodules. In the NLST, a positive screening result was reported in 24% of all baseline LD-CT scans (1). A positive screening result was thereby defined as a non-calcified pulmonary nodule with a maximum diameter of more than 4 mm. Importantly, all but 3.6% of the detected pulmonary nodules eventually proved to be benign in nature (1). Most of the detected nodules were further evaluated with follow-up CT examinations, and only 2.6% of the nodules were surgically resected. Even though the majority of the nodules were investigated with follow-up examinations or minimally invasively, the added cost and cumulative radiation dose, the potential risks of these examinations, and, last but not least, the anxiety of the screened persons with a positive result are of major concern. As the probability of malignancy increases with increasing nodule diameter, using a threshold for nodule diameter, which would define positivity to 7 mm, would decrease the early recall rate by up to 70% (16). By considering additional data besides the size of the nodules, such as the location of the nodule, the number of detected nodules, the sex and age of the screened person, and the extent of emphysema, the risk of malignancy of the nodules could be even better predicted and the recall rate could be reduced even further (17).
Much has been done in the last few years to provide a reliable classification scheme for screening-detected nodules. Analogously to the Breast Imaging Reporting and Data System (BI-RADS) of the American College of Radiology (ACR), which is used worldwide in breast cancer screening, the ACR recently proposed a Lung Imaging Reporting and Data System (Lung-RADS) (18). A similar system, the Lung Reporting and Data System (LU-RADS), was published by another group (19). In both classification schemes, screening-detected nodules are categorized and managed according to their individual risk. Both classification schemes should be easy to apply in the clinical routine and allow standardized data collection and analysis.
How big is the risk of over-diagnosis?
One of the major uncertainties in lung cancer screening is the extent of over-diagnosis. Over-diagnosis is defined as the detection of cancer that otherwise would not become clinically apparent (20). Thus, the detection of lung cancer during screening does not necessarily result in improved lung cancer mortality, as a proportion of the detected cancers would have remained asymptomatic. Follow-up and treatment of such indolent cancers would add to the costs and potential risks of screening. Early reports concluded that the proportion of over-diagnosed cases could be as low as 5% (21). More recent data, however, indicate that the extent of over-diagnosis in the NLST could have been more than 18% (20). This estimation is almost as high as in a study based on data from an Italian cohort study, which estimated that over-diagnosis could be as high as 25% (22).
To date, there are no generally accepted criteria by which to differentiate indolent tumors from genuine ones. Strategies to reduce over-diagnosis focus on a reduction of the frequency of screening examinations, a better definition the screening population, and raising the threshold for follow-up examinations and invasive diagnosis (23).