Robust evidence exists in support of lung cancer (LC) screening with low-dose computed tomography in patients at high risk of developing LC; however, judicious patient selection is necessary to obtain optimal benefit while minimizing harm. Several professional societies have published recommendations regarding patient selection criteria for screening. Multiple risk prediction models that include additional patient-specific risk factors have since been developed to more accurately predict risk of developing LC. Implementation of a new screening program requires thorough multidisciplinary planning and maintenance. Multisociety guidelines highlight 9 principal components to implement and maintain a successful program.
Key points
- •
The goal of lung cancer (LC) screening is to detect early-stage LC in patients at high risk for LC who are healthy enough to undergo evaluation and successful treatment, while minimizing adverse effects of screening.
- •
Several randomized controlled trials have demonstrated lung-cancer mortality benefit of screening for LC with low-dose computed tomography in select patients.
- •
Professional societies and the US Preventive Services Task Force recommend LC screening in individuals based on age, smoking history, and ability to undergo curative treatment of a screen-detected LC.
- •
Patient selection for LC screening may be improved with the use of validated risk prediction calculators, which incorporate additional risk factors for LC.
- •
Implementation of LC screening requires multidisciplinary input to ensure that the essential components of a LC screening program are incorporated.
Introduction
The combined 5-year survival for lung cancer (LC) remains low, at 18%, because most patients present with advanced disease at the time of diagnosis. In those with early-stage disease, however, the 5-year survival is as high as 80%, making early detection ideal. The results of the National Lung Screening Trial (NLST) provided the evidence for screening, with annual low-dose computed tomography (LDCT) demonstrating a 20% reduction in LC mortality. Due to these results, the US Preventive Services Task Force (USPSTF) provided a grade B recommendation in favor of screening, and the Centers for Medicare and Medicaid Services (CMS) approved LC screening in their eligible beneficiaries. Both the USPSTF and CMS highlight the importance of proper patient selection and, in conjunction with professional societies, outlined components necessary for an effective LC screening program. This article focuses on patient selection for and implementation of LC screening.
Evidence for lung cancer screening
The Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial was the first large randomized trial to examine LC screening with the use of chest radiography versus usual care. Screening with chest x-ray did not result in a significant decrease in LC incidence or mortality. There were also similar rates of stage and histology between the 2 groups. This study provided definitive evidence that screening with chest x-ray is not effective.
Following the PLCO, there have been several cohort studies evaluating outcomes from screening with computed tomography (CT) that suggested a benefit to LDCT screening but were inconclusive in the absence of a comparator arm. In addition, there were several randomized controlled trials (RCTs) of LDCT that failed to demonstrate mortality benefit due to their lack of power and low enrollment. The largest and most often cited RCTs are highlighted.
The NLST randomized 53,454 patients at high risk for developing LC to chest x-ray versus LDCT annually for 3 years. Inclusion criteria to define individuals at high risk of developing LC included (1) ages 55 years to 74 years, (2) history of cigarette smoking of at least 30 pack-years, and (3) if former smokers, whether they had quit within the last 15 years. There was no usual care group in this study. The trial met its predetermined endpoint of a 20% reduction in LC-related mortality in the LDCT arm with a number needed to screen of 320 to prevent 1 LC death. , These findings provided the impetus for broad-based implementation of LC screening programs in the United States.
Simultaneous to the NLST, the Dutch-Belgian Randomized Lung Screening Trial (NELSON) was another large randomized trial in the Netherlands and Belgium that aimed to show that screening with LDCT would decrease 10-year mortality. Patient eligibility included analysis. The NELSON trial randomized 15,822 participants to LDCT or usual care and found a mortality benefit with a 26% reduction in LC mortality. This study was unique compared with the NLST and many other previous trials in that all pulmonary nodules were monitored with 3-dimensional volumetric analysis. Nodules were characterized by nodule size and volume doubling time, which was found to be more accurate than the 2-dimensional monitoring of nodules. The participants were also followed at longer intervals, at 1 year, 3 years, and 5.5 years from enrollment. This method of risk stratification for LC screening was novel in that it included patients’ CT findings as a part of their risk assessment for LC.
The Detection and Screening of Early Lung Cancer by Novel Imaging Technology and Molecular Essays trial was an RCT comparing usual care with LDCT annually for 5 years. This Italian study, which was not powered to detect a difference between the 2 groups, randomized 2472 men to LDCT or usual care. Eligible participants ages 60 years to 74 years with at least a 20 pack-year smoking history were followed for a median of 8 years. Although more early-stage and advanced-stage LCs were discovered in the LDCT arm, there was no significant stage shift compared with the usual care arm and there was no difference in LC or all-cause mortality.
Finally, the Multicentric Italian Lung Detection trial was a single-center trial that included 4099 smokers, ages 49 years and older, with a greater than 20 pack-year smoking history, and randomized them to annual CT, biennial CT, or usual care. The relative risk (RR) for dying of LC was lower in the biennial CT and annual CT groups compared with the usual care group, but all-cause mortality did not significantly differ when comparing the combined screening groups with the usual care group.
Patient selection for lung cancer screening
Current Lung Cancer Screening Recommendations
Several professional societies have endorsed LC screening in the United States, each having slightly different age criteria for patient selection and some including other risk factors. Table 1 summarizes the current recommendations for patient selection for LC screening based on different professional societies.
| Guideline | Inclusion Criteria | Exclusion Criteria | When to Stop Screening |
|---|---|---|---|
| NLST |
|
| |
| USPSTF |
|
|
|
| CMS |
|
|
|
| NCCN |
| ||
| CHEST |
|
| |
| AATS |
|
US Preventive Screening Task Force
In 2013, largely based on the results of the NLST, the USPSTF recommended screening for LC with LDCT in a high-risk population. The criteria recommended for screening remained mostly true to inclusion criteria for the NLST. As a result of the Cancer Intervention and Surveillance Modeling Network for health care research, however, the age criterion for inclusion was increased from 55 years to 74 years to 55 years to 80 years to balance the benefits of screening with the risk of false-positive results.
National Comprehensive Cancer Network
The National Comprehensive Cancer Network (NCCN) was the first major organization to recommend and develop official guidelines for LC screening. The NCCN recommends screening with LDCT for 2 separate groups of individuals felt to be at high risk for developing LC. The first group are those meeting the age and smoking history criteria for NLST inclusion. The second group, given a category 2A recommendation, includes younger individuals (ages 50 years and older) with lighter smoking histories (minimum 20 pack years) and an additional risk factor for LC. Additional risk factors include a personal history of cancer or lung disease, family history of LC, radon exposure, or occupational exposure to carcinogens. The inclusion of these additional risk factors was based on previous studies, which showed association with higher risk for LC.
Occupational carcinogens, such as arsenic, chromium, asbestos, nickel, cadmium, beryllium, silica, diesel fumes, coal smoke, and soot, have a calculated mean RR of 1.59 for development of LC. Among these patients with occupational exposures, smokers have an even higher risk for developing LC. A 2005 meta-analysis showed that the amount of radon exposure had a linear relationship with the risk of development of LC, which again was even higher in smokers. Patients with a personal history of cancer, whether lung primary, head and neck, lymphoma, or other smoking-related cancers, also have increased risk of developing LC due to both genetic susceptibility and treatment, including radiation and alkylating chemotherapy agents. Although there is no specific genetic syndrome associated with LC, a family history of a first-degree relative with LC portends an RR of 1.8 (95% CI, 1.6–2.0) of developing LC. Finally, underlying lung disease, specifically COPD and pulmonary fibrosis, have been associated with higher risk for developing LC.
American College of s Physicians (CHEST)
American College of Chest Physicians (CHEST) guidelines for patient selection for LC are in line with the NLST entry criteria, including patients ages 55 years to 77 years, patients who have smoked at least 30 pack-years or more, and patients who are current smokers or have quit within the past 15 years. This differs from the age cutoff of 80 years recommended by the USPSTF but is reflective of what is covered by the CMS. The guidelines do remark on the improved efficiency of identifying high-risk patients using risk prediction calculators; however, they do not currently recommend using these calculators to qualify high-risk patients who do not meet the NLST criteria for LC screening. This is attributed to the idea that the risk factors included in many of these calculators also portend a higher risk of death from competing comorbidities or morbidity from evaluation of the nodules, mitigating the benefit and increasing the harm of LC screening in this population. Additionally, for patients who meet these criteria, but have comorbidities that disallow them to tolerate evaluation or treatment of early-stage cancer, or substantially decrease life expectancy, the guidelines recommend against screening.
American Association for Thoracic Surgery
The American Association for Thoracic Surgery (AATS) recommendations for inclusion in LC screening also reflect the inclusion criteria for the NLST, with the main difference of age cutoff from ages 55 years to 79 years. Although the NLST screened patients with LDCT annually for 3 years, the risk of developing LC after 3 years does not decrease. By the end of follow-up in the trial, 5 years after the third annual screen, the percentage of stage I LCs detected had decreased from 63% to 50% whereas the rate of diagnosis of stage IIIB/IV LC had increased from 21% to 33%. This suggests that continued surveillance with annual LDCT after the third scan may lead to greater mortality reduction if said cancers had been diagnosed at earlier stages. Thus, the AATS recommends the higher age cutoff of 79 years old, because risk of LC increases linearly with the age and the average life expectancy in the United States is 78.6 years, with an additional 9 years for Americans who reach age 79 years. The AATS also specifically recommends annual screening with LDCT for LC survivors starting 5 years after treatment, because these patients maintain a high risk for recurrence or secondary LC and were excluded from most trials. Additionally, they recommend screening for patients ages 50 years to 79 years with a 20 pack-year smoking history and an additional risk factor that produces at least a 5% risk of developing LC over the next 5 years. They do recommend the use of clinical risk calculators to assist in determining patient risk.
Risk Prediction Models for Patient Selection
Following the publication of the NLST, there have been several investigations into developing and validating risk prediction calculators to be more efficient (eg, find more cancers while screening less people) than the selection criteria of age and smoking history. By enriching the pool of patients screened for LC, there is the potential to both reduce the number of false positives and the number needed to screen. Furthermore, providing a person with an individual risk of developing LC can be beneficial in facilitating informed decision making around LC screening.
In the United States, the number of screen-eligible patients from 2010 to 2015 decreased by 1.5 million. The decrease in number of patients with a 5-year LC risk of at least 2%, however, was only 0.8 million, suggesting there are patients at high risk of developing LC who are not being screened. Beyond age and smoking history, other risk factors identified to increase risk of LC include family history, ethnicity, level of education, socioeconomic status, body mass index (BMI), chronic obstructive pulmonary disease (COPD), personal history of cancer, and smoking intensity. Although many models have been developed, external validation and comparison of these models to each other are somewhat limited. A study from Ten Haaf and colleagues, published in 2017, compared the performance of 9 of the more prevalent risk models on the NLST and PLCO trial populations.
Table 2 identifies the 9 different risk models, their inclusion risk factors, and the prediction time frame. All these models outperformed the NLST eligibility criteria with higher sensitivity for all models and higher specificity for some models. Fig. 1 compares the sensitivity and specificity of the different models to the NLST criteria. The PLCO m2012 , Bach, and two-stage clonal expansion (TSCE) incidence models had the best overall performance in that order with highest sensitivity and specificity for prediction 6-year LC incidence. These 3 models had the best discriminative performance (based on areas under the curve >0.68–0.77) when coupled with specific risk thresholds. For example, the PLCOm2012 risk threshold for screening is at least a 1.5% 6-year risk. The study concluded that LC risk prediction models, when considering their specific riskh thresholds, outperform current recommended LC screening criteria.
