Editorial: Lung Cancer Field Cancerization: Implications for Screening by Low-Dose Computed Tomography

Lung cancer is the second most common cancer diagnosis and the leading cause of cancer-associated deaths in the United States (1). Most patients are diagnosed with locally advanced or metastatic disease, which carry a dismal five-year survival rate of 17% (1). Because up to 80% of lung cancer cases are related to exposure to cigarette smoke (2), research efforts have focused on developing screening methodologies that target individuals at risk. Results from the National Lung Screening Trial (NLST), published in 2011, demonstrated that repeated screening of high-risk individuals (defined by age and history of heavy smoking) using low-dose computed tomography (LDCT) identified lung tumors at an earlier stage compared with standard chest x-ray and reduced lung cancer–associated mortality by 20% (3). The Centers for Medicare and Medicaid Services (CMS) decided in early 2015 to cover the cost of LDCT screening according to NLST patient eligibility criteria (4). Consequently, it has become imperative to develop standards to prioritize the almost 10 million eligible individuals (5). A multivariable risk prediction model that includes socioeconomic and clinical parameters as well as smoking history has been proposed that may increase the efficacy and efficiency of screening (6). LDCT screening poses an additional diagnostic challenge as it identifies a high number of pulmonary nodules that prompt further testing but do not result in a lung cancer diagnosis. In the NLST, over 95% of detected pulmonary lesions were noncancer (3). The clinical management of pulmonary nodules frequently involves invasive procedures, including bronchoscopy and biopsy (7,8). Biomarkers that are highly specific and complementary to LDCT could help the oncologist assess the malignant potential of pulmonary nodules, resulting in fewer unnecessary surgeries (Figure 1). Biomarkers based on the quantitative analysis (“Radiomics”) of features extracted computationally from CT scans are now being evaluated for diagnostic performance using data and images from the NLST (9). Furthermore, noninvasive circulating or urinary biomarkers associated with lung cancer risk and diagnosis are being sought that better define the target high-risk population, help stratify patients, and guide screening and nodule management (10). A growing list of circulating biomarkers in various stages of clinical development includes microRNAs (11,12), methylated tumor DNA (13), proteins (14), and cytokines (15), as well as tumor-derived urine metabolites (16). In addition, minimally invasive biomarkers derived from tissue samples within the respiratory tract such as sputum (17) and bronchial epithelial cells (18,19) are being developed for the diagnosis of lung cancer. It has long been postulated that exposure of the respiratory epithelium to cigarette smoke generates a “field of injury” that can be measured molecularly (20) and serves as surrogate for less accessible lung tissue to allow detection of disease in smokers with lung cancer (21).

Figure 1.

Multivariable risk modeling and minimally invasive biomarkers can aid the screening and diagnosis of lung cancer in high-risk individuals. BMI = body mass index; LDCT = low-dose computed tomography.

In this issue of the Journal, Drs. Lenburg, Spira, and colleagues show that gene expression features measured in the more distant nasal epithelium of smokers reflect the presence of cancer in the lung (22). First, Perez-Rogers et al. obtained gene expression profiles from nasal epithelial brushings collected at the time of bronchoscopy in the Airway Epithelial Gene Expression in the Diagnosis of Lung Cancer (AEGIS)–1 and AEGIS-2 prospective trials (22). These trials accrued current or former smokers undergoing bronchoscopy for suspected lung cancer. In a subset of AEGIS patients with matched bronchial gene expression from an earlier study (18), they subsequently found concordant cancer-associated gene expression alterations between the two airway sites, suggesting that the molecular “field of injury” extends to the nose. Interestingly, a bronchial classifier score (23) computed on nasal samples showed high correlation to the score calculated on matched bronchial samples. However, some misclassifications were observed, indicating that there are important differences in gene expression profiles as the field of cancerization reaches the nose (24).

To be useful in a clinical setting, a diagnostic test should improve disease discrimination above clinical factors routinely available to the oncologist. Perez-Rogers et al. derived a nasal clinicogenomic classifier that combined clinical factors and the nasal expression of 30 genes using AEGIS-1 samples (22). This combined classifier resulted in higher disease discrimination (area under the curve [AUC] = 0.81, 95% confidence interval [CI] = 0.74 to 0.89) and sensitivity (AUC = 0.91, 95% CI = 0.81 to 0.97) than a clinical factor–only model in independent samples from AEGIS-2. Importantly, the clinicogenomic classifier had negative predictive values of 0.85 (95% CI = 0.65 to 0.96) and 0.93 (95% CI = 0.66 to 1.00) in patients with diagnostically challenging small (<3 cm) or peripheral lesions, respectively. Thus, the authors conclude that nasal gene expression can have clinical utility for lung cancer detection.

There are several strengths to this study, including the use of independent discovery and validation cohorts of patients prospectively collected during routine medical care at community and academic hospitals. The proposed clinical use of the classifier is narrowly defined and highly relevant, and appropriate parameters of test performance are extensively described. The authors provide evidence that gene expression features measurable in the nasal epithelium of smokers reflect the presence of cancer in the lung. The concept of field effect cancerization was first proposed by Slaughter et al. in 1953 (25), and its extension from the bronchial epithelium to the more accessible nasal epithelium is a substantial clinical advance for the screening of lung cancer.

Perez-Rogers et al. leverage molecular data from individuals already suspected to have lung cancer (22). It remains to be demonstrated that the nasal epithelium will be useful for prioritization of individuals for LDCT or for routine screening of asymptomatic individuals at high-risk for smoking-related lung cancer. From a more practical perspective, it is unclear how the specific genes included in the nasal gene expression classifier will be implemented as a test. Five features with no annotation to a known gene (“NA”) are found among the 30 included in the classifier. This may pose a challenge for classifier validation and implementation. The combination of the nasal epithelium classifier and other biomarkers that are statistically and mechanistically independent will also likely be required. Nevertheless, the work of Perez-Rogers et al. presents a notable advance in the understanding of smoking-related carcinogenesis as well as a promising tool to aid decision-making by physicians and bring the field of thoracic oncology much closer to the goal of precision medicine (26).

Funding

The authors are supported by the Intramural Research Program of the National Cancer Institute, National Institutes of Health.

Notes

The sponsor had no role in the writing of the editorial or decision to submit it for publication. The authors have no conflicts of interest to disclose.