| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Centennial dissertation |
1
Department of Radiology, Beth Israel Deaconess Medical Center and Harvard
Medical School, One Deaconess Rd., Boston, MA 02215.
2
Department of Pulmonary Medicine, Beth Israel Deaconess Medical Center and
Harvard Medical School, Boston, MA 02215.
3
Department of Medical Oncology, Beth Israel Deaconess Medical Center and
Harvard Medical School, Boston, MA 02215.
Received March 10, 2000;
accepted after revision April 14, 2000.
Honoring Arthur C. Christie, MD and William H. Stewart, MD
Introduction
 Top Introduction ScreeningEmerging Technologies: Methods...Future Directions: Where Do...References |
|---|
,
|
|
|
|
Current policies regarding lung cancer screening are based on decades-old studies that used chest radiography and sputum cytology [9,10,11,12]. Those studies failed to show a reduction in lung cancer-associated mortality. Several exciting new developments have occurred in the area of lung cancer detection, including low-dose helical CT, autofluorescence bronchoscopy, advanced sputum analysis, and molecular markers [5, 13,14,15,16,17,18,19,20,21,22,23,24,25,26]. Such developments, most notably low-dose helical CT, have encouraged many researchers, clinicians, health care policy officials, and advocates for patients with lung cancer to revisit the topic of lung cancer screening. Indeed, in the past year, this topic has catapulted from the pages of medical journals to the front page of the New York Times [27]. It has also been the subject of debate at multiple medical forums in the radiology community as well as within the scientific community at large.
In this article, we review the basic concepts of screening and describe several emerging methods for early lung cancer detection, with a special emphasis on low-dose helical CT. The purpose of this article is to familiarize the radiologist with the potential contributions and challenges of emerging technologies for the early detection of lung cancer.
|
TopIntroductionScreeningEmerging Technologies: Methods...Future Directions: Where Do...References |
|---|
The screening procedure itself does not usually yield a definitive diagnosis [28, 29]. Rather, it provides information that either excludes the disease or identifies a basis for the application of more sophisticated diagnostic procedures in persons who may harbor the disease.
To be effective, a screening study must be characterized by a high sensitivity [29]. In other words, a negative result must provide reliable evidence that the disease is not present. Moreover, a screening test must also have at least a moderate specificity for a particular disease [29]. In most screening models, a very specific test (often a more costly and invasive procedure, such as a biopsy) is applied to persons with a positive screening study. Because of patient care issues and economic considerations, a screening study should ideally be associated with few false-positive cases. Finally, the population on whom the screening test is applied should have a reasonable pretest probability of harboring the disease. When the pretest probability is low, even a highly sensitive and specific test will increase the posttest probability to only a moderate level [29].
Most screening outcome measures can be affected by several biases, including lead-time bias, length-time bias, and overdiagnosis [28, 30]. Importantly, mortality is not affected by these particular biases [28, 30]. Thus, mortality is generally considered the ultimate outcome measure of a screening study [28, 30]. Indeed, it has been a widely held belief that a screening study must show a significant reduction in mortality before it is adopted as standard care [28, 30].
All outcome measures, including mortality, may be affected by selection bias [28]. Selection bias refers to a bias in assignment that occurs when the study and control groups differ from one another by one or more factors that may affect the outcome of the study [28]. A properly designed randomized control trial will usually minimize the effect of selection bias. However, even when a study is properly designed so that selection bias is unlikely, random error due to chance alone may produce differences that might affect the results of the investigation [28].
Consensus exists among the American Cancer Society, National Cancer Institute, Food and Drug Administration, American College of Radiology, and the World Health Organization that traditional methods for lung cancer screening are not indicated [30,31,32,33]. In recent years, however, two major factors have fostered a renewed interest in screening for lung cancer. The first is a growing concern that studies performed to assess the role of screening have been flawed in terms of methodology and data analysis [16, 30, 34, 35]. The second is a growth of new technologies for early lung cancer detection, most notably low-dose helical CT [13,14,15,16].
Scope of the Problem: Why Lung Cancer Matters So Much to So Many
The American Cancer Society estimates that approximately 164,100 new cases
of lung cancer and 156,900 deaths due to lung cancer will occur in the United
States alone in 2000 [1]. Most
cases (85%) of lung cancer are associated with tobacco abuse
[2]. Carcinogenic compounds in
tobacco smoke include the polynuclear aromatic hydrocarbons, such as the
classic carcinogen benzo[a]pyrene and the nicotine-derived tobacco-specific
nitrosamine, 4-(methylnitrosamino)-1-(3-pyridil)-1-butanone
[36]. Because of the strong
association between tobacco use and lung cancer, smoking cessation is an
important factor in attempting to reduce the scope of this problem,
particularly among teenagers and young adults. However, up to 50% of lung
cancers are detected in former smokers
[37,
38].
Roughly 46 million current and 45 million former smokers are in the United States, together constituting approximately half the adult population [38]. Considering this large pool of at-risk individuals, lung cancer will undoubtedly remain a significant health problem in the coming decades, regardless of whether smoking cessation efforts are widely successful.
Given the large number of lung cancer deaths in the United States, even a modest reduction in mortality from screening has the potential to significantly affect our nation's public health. On the other hand, the large population at risk presents tremendous financial and logistic challenges for mass screening. To construct a viable mass screening program for lung cancer, it will be important to screen subgroups of smokers with a relatively high pretest probability for developing lung cancer.
Narrowing the Field: Which Smokers Are at Greatest Risk for
Developing Lung Cancer?
The causal pathway of lung cancer initiation by smoking includes the
duration of smoking, the depth of inhalation, the presence of emphysema, and
the deposition of particulate matter
[39]. A second branch of the
pathway includes chronic inflammation, release of proteolytic enzymes and
oxygen radicals, and macrophage secretion of polypeptide growth factors
[40]. A third component is
genetic susceptibility [39,
41].
Because only 15% of smokers ultimately develop lung cancer [42, 43], a variety of tests and models have been investigated to predict those at greatest risk for developing lung cancer. An association between the number of pack-years of cigarette smoking and the development of lung cancer is clear [2]; likewise, the correlation between advancing age and the risk of developing lung cancer is established [32]. The concept of individual variation in susceptibility to carcinogens is of great importance and is under investigation. For example, Krontiris et al. [44] reported an association between the variability in the HRAS1 minisatellite focus and the risk of developing several epithelial neoplasms, including lung cancer.
The effect of population selection on the prevalence of lung cancer in a screening study can be illustrated by comparing the low-dose helical CT screening study performed by Henschke et al. [16] with that of Sone et al. [15]. Henschke et al. screened persons older than 60 years with a positive history of smoking (>10 pack-years). In contrast, Sone et al. screened persons older than 40 years, including smokers and non-smokers. The prevalence rate of lung cancer in the study by Henschke et al. was fivefold that of the rate observed by Sone et al. (2.7% versus 0.48%, respectively). Population differences between the United States and Japan may have also played a role in this difference.
Emerging Technologies: Methods for Early Detection
|
TopIntroductionScreeningEmerging Technologies: Methods...Future Directions: Where Do...References |
|---|
The following algorithm was devised to guide the evaluation of noncalcified pulmonary nodules detected in the Early Lung Cancer Action Project [16]: nodules smaller than 5 mm in diameter were followed up using serial CT scans to assess interval growth over a 2-year period (3, 6, 12, and 24 months); nodules 5-10 mm were either followed up or biopsied; and nodules larger than 10 mm were biopsied.
In the study by Henschke et al. [16], 27 (2.7%) of 1000 subjects were found to have lung cancer on low-dose helical CT images, versus 7 (0.7%) on chest radiography. With regard to the cancers revealed on CT, 26 (96%) were resectable and 23 (85%) were stage I neoplasms. In contrast, lesions of only four (17%) of 23 patients with stage I disease were revealed on chest radiography.
Although the concept of using helical CT for the detection of early lung cancer is relatively new in the United States, this technique has been used in Japan on a wider scale. Sone et al. [15] reported their experience in screening 5483 Japanese adults between the ages of 40 and 74 years, including smokers and nonsmokers. Nineteen patients (prevalence, 0.48%) were diagnosed with lung cancer, including 84% with stage I disease.
Kaneko et al. [13] reported the use of biannual chest radiographs and helical CT scans in screening 1369 Japanese adults who were at high risk for developing lung cancer. Peripheral lung cancer was revealed in 15 subjects (1%) on CT but in only four (0.3%) on chest radiography. Most (93%) detected cancers were classified as stage I.
These studies have clearly shown that low-dose helical CT is superior to conventional chest radiography in the detection of early lung cancer. What has not been shown yet, however, is whether such early detection indeed results in a reduction in mortality. Although the authors of the Early Lung Cancer Action Project plan to follow up their patients to determine cure rates [16], the lack of a randomized control population limits the ability to distinguish a true survival benefit from lead-time bias. Lead-time bias refers to the concept that screening allows earlier diagnosis but may not necessarily delay the time of death [32]. Thus, although moving forward the time of diagnosis of a lung cancer will result in improved 5-year survival rates calculated from the time of diagnosis, it may not necessarily change the actual mortality [32].
These studies have also shown the consistent ability of low-dose helical CT to reveal lung cancer at an early, potentially curable stage. Although the results are promising, low-dose helical CT has several important potential limitations, including a relatively high false-positive rate, difficulty in accurately measuring the growth of small nodules, a bias toward detecting adenocarcinomas, and the potential for overdiagnosis of lung cancer.
First, the relatively high false-positive rate of the initial baseline helical CT scan in the Early Lung Cancer Action Project should be noted [16]. In that study, 233 (23.3%) of 1000 patients were found to have one to six noncalcified nodules on helical CT, but only 27 of these nodules proved to be malignant. Remarkably, however, when the prescribed guidelines for nodule assessment were followed, only 28 of the 233 patients with noncalcified nodules required biopsy. Of these nodules, nearly all (27) were proven malignant.
Although the lack of excessive invasive procedures is reassuring, the potential costs of performing serial follow-up CT in such a high percentage of patients may prove too expensive for mass screening. For example, a conservative cost estimate ($300 per scan) for a series of follow-up CT scans to confirm 2-year stability of a false-positive nodule is approximately $1800 (this estimate includes a screening CT scan, an immediate follow-up high-resolution nodule study, and four follow-up CT scans at 3, 6, 12, and 24 months).
False-positive rates for screening low-dose helical CT studies will likely
be higher in areas in which there is a high prevalence of granulomatous
infections. Indeed, preliminary data from the baseline screening study using
low-dose helical CT at the Mayo Clinic shows that 782 (51%) of 1520 screened
patients (
50 years old; 20 pack-years) revealed one or more
noncalcified nodules (Hartman TE, personal communication). Although the high
prevalence of granulomatous infections is likely the primary reason for the
high percentage of patients with one or more lung nodules, several technical
factors likely played a role in improved detection of nodules. Technical
differences from the Early Lung Cancer Action Project
[16] included the use of a
multidetector CT scanner, narrower collimation (5 mm versus 10 mm), and cine
viewing rather than film viewing. The prevalence of lung cancer in this study
was less than 1%, and four patients (0.3%) underwent resection for benign
granulomatous infection. These data suggest that important geographic barriers
to low-dose helical CT screening may be present in areas in which
granulomatous infections are endemic.
With regard to the specificity of low-dose helical CT, one would expect to see a lower false-positive rate at repeated yearly screenings than in the baseline prevalence study. Recently, Henschke et al. reported their findings from the first annual repeated low-dose helical CT examination in the Early Lung Cancer Action Program screening population (Henschke CI et al., presented at the Radiological Society of North America meeting, November 1999). At annual repeated low-dose helical CT, 31 (5%) of 623 patients had new or growing nodules compared with the findings on baseline screening. Nine of these nodules proved negative or demonstrated benign calcifications on additional high-resolution CT imaging. Of the remaining nodules, eight were larger than 5 mm in diameter and were biopsied. Malignant disease was found in seven of eight nodules. The overall detection rate of non-small cell lung cancer on first annual repeated helical CT was 1%. Importantly, 83% of these lesions were stage IA.
A second potential limitation of low-dose helical CT is the difficulty encountered in reliably detecting a malignant growth rate in small (<1 cm) nodules [45]. For example, a 5-mm nodule can double in volume during a 6-month period, but its diameter will increase by only 1.25 mm [45]. The accurate assessment of the growth of small nodules will require more sophisticated methods of nodule measurement. In recent years, promising work has been done in computer-aided three-dimensional nodule measurement using sophisticated software programs [46,47,48] (Fig. 3A,3B). Preliminary studies by Zhao et al. [47] and Yankelevitz et al. [48] using computer-aided volumetric measurement methods to assess phantom models and lung nodules in human subjects suggest that this technique is highly accurate in detecting a malignant growth rate in small nodules. To gain widespread application, however, such methods will need to become more widely accessible and less labor intensive.
|
|
A third potential limitation of low-dose helical CT screening is its bias toward detecting adenocarcinomas, which constitute most peripheral lung cancers [2]. In the Early Lung Cancer Action Project baseline study [16], 26 (93%) of 28 neoplasms were characterized by an adenocarcinoma cell type (18 adenocarcinomas, 4 mixed squamous adenocarcinomas, and 4 bronchoalveolar cell carcinomas; 1 patient had 2 neoplasms in 1 lobe) [16]. For detected cancers to more accurately reflect the entire spectrum of non-small cell lung cancer, low-dose helical CT should ideally be combined with a tool such as advanced sputum analysis for detecting central neoplasms. These techniques are discussed later in this article.
Finally, the possible overdiagnosis of lung cancer is a concern. Overdiagnosis occurs when a lesion is detected and labeled as clinically important, but, in fact, might not be cancer at all [32]. With regard to lung cancer screening, the detection of bronchioloalveolar cell adenomas—a benign lesion that may have malignant potential [49, 50]—is an example of potential overdiagnosis. This is a controversial subject that requires further study.
Recent advances in technology, including multidetector CT scanners, cine viewing, computerized detection methods, and three-dimensional reconstruction techniques, will improve the ability of low-dose helical CT to detect and accurately characterize lung nodules [46,47,48, 51,52,53,54]. Moreover, the addition of more specific noninvasive methods of imaging evaluation such as CT nodule enhancement [55] and 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET) imaging [56,57,58,59] may help to reduce the number of cases requiring close follow-up or biopsy. FDG PET imaging, which relies on physiologic (glucose metabolism) rather than morphologic features to distinguish malignant from benign nodules, has a relatively high sensitivity and specificity for malignant nodules 1 cm or larger in diameter [58, 59]. However, this technique is not without limitations. False-positive diagnoses can be encountered in active granulomatous infections, and false-negative diagnoses are commonly observed in bronchoalveolar cell carcinomas and carcinoids [56,57,58,59].
Important questions to answer before proceeding to mass screening with low-dose helical CT include: Will the outstanding results reported by the Early Lung Cancer Action Project investigators and Japanese researchers be equaled by those at other institutions? Does early detection on helical CT result in a measurable reduction in the mortality of lung cancer? Can mass screening with helical CT be performed in a cost-effective manner? National studies are being planned to address these questions.
Conventional and Digital Chest Radiography
Early studies, including the Mayo Lung Project
[10] and the Memorial
Sloan-Kettering Project [9],
showed no mortality benefit of screening with chest radiography and sputum
cytology. Moreover, the results of recent studies by Henschke et al.
[16] and Kaneko et al.
[13] have clearly shown the
limited ability of chest radiography to reveal early stage lung cancer.
An ongoing trial by the National Cancer Institute Early Detection Branch will likely provide a definitive answer to the question of whether conventional radiography has any role to play in screening for lung cancer [60]. This 16-year, randomized screening study for the early detection of prostate, lung, colorectal, and ovarian cancer involves 74,000 participants. This study is designed to assess whether screening studies such as chest radiography, flexible sigmoidoscopy, digital rectal examination, serum prostate-specific antigen, CA-125, and transvaginal sonography can reduce cancer mortality in men and women 60-74 years old [60].
Emerging technologic advancements in digital radiography, including computer-aided diagnosis and dual-energy subtraction methods, may significantly improve the ability of chest radiography to detect small lung nodules [61,62,63]. Once this technology matures, studies will be necessary to address the ability of digital chest radiography to detect early lung cancer.
Sputum Analysis
Sputum cytology.—The sensitivity of sputum cytology for
detecting lung cancer in screening studies is approximately 20-30%
[5,
9,
10,
64]. This yield may be
improved by adherence to proper technique for collection, processing, and
interpretation [20]. Sputum
analysis shows the highest sensitivity for squamous cell carcinoma and the
lowest yield for adenocarcinoma. Thus, this technique is complementary to
low-dose helical CT, which has a high yield for adenocarcinoma
[16]. In general, centrally
located lesions, lower lobe lesions, and lesions larger than 2 cm are best
suited to detection with sputum cytology
[5]. This technique has a
false-positive rate of less than 2%
[5].
Sputum immunostaining.—A nuclear ribonucleoprotein, hnRNP A2/B1, has been found to be overexpressed in most lung cancer cell types and transformed epithelial cells [5, 26, 65]. In recent studies, sensitivity ranged from 77% to 91% and specificity ranged from 65% to 88% [26, 65]. Overexpression of hnRNP A2/B1 appears to be a feature of preneoplastic cellular transformation and does not seem to be histology-dependent. Preliminary data suggest that hnRNP may be a potentially useful lung cancer biomarker. This ribonucleoprotein could eventually be used in a panel in combination with other markers such as PGP9.5 to improve sensitivity and specificity [21].
Sputum polymerase chain reaction-based assays for detecting oncogene mutations.—Polymerase chain reaction-based assays can be used to detect tumor-specific molecular alterations, such as the K-ras and p53 mutations [5, 66, 67]. However, to our knowledge, no prospective study has been performed to determine the risk of lung cancer if one or more of these mutations is found [5].
A recent study by Ahrendt et al. [19] suggests that molecular detection of lung cancer using polymerase chain reaction-based assays is limited by a poor sensitivity. In their study, four promising molecular assays (including K-ras and p53) were used to evaluate bronchoalveolar lavage fluid obtained from 50 patients with resectable non-small cell lung cancer. Oncogene mutations or microsatellite instability was detected in the fluid of 23 (53%) of 43 patients with tumors carrying a genetic alteration. Future technologic advances will likely increase the yield of this technique [68]. At present, polymerase chain reaction-based assays for detecting oncogene mutations are strictly research tools.
Sputum computer-assisted analysis of malignancy-associated changes.—Malignancy-associated changes refer to the nongenetic changes in normal cells that are induced by the presence of malignant cells in the vicinity [5]. Unlike sputum immunostaining and polymerase chain reaction-based assays, this technique does not require the presence of premalignant or malignant cells for diagnosis [5, 22, 23]. The presence of malignancy-associated changes may precede the clinical diagnosis of lung cancer by 12 months or more [23].
Bronchoscopy
Conventional and autofluorescence bronchoscopy.—Conventional
bronchoscopy is valuable for localizing preinvasive lung cancer. In general,
it can detect nodular or polypoid lesions larger than 2 mm and flat or
superficially spreading lesions greater than 2 cm
[5]. With regard to carcinoma
in situ, 75% of lesions are superficial or flat and 25% are nodular or
polypoid [5].
Autofluorescence bronchoscopy is an optical imaging method designed to improve our ability to localize small preinvasive lesions that are not visible by conventional white light bronchoscopy [5, 24, 69]. Autofluorescence bronchoscopy involves illuminating the bronchial surface with violet or blue light (400-440 nm) to distinguish normal from abnormal tissue. Dysplastic lesions and carcinoma in situ will result in a diminution in the intensity of autofluorescence. The light-induced fluorescence endoscopy device (LIFE-lung; Xillix Technologies, Richmond, B.C., Canada) was designed to capitalize on these differences in autofluorescence properties to aid in the detection and localization of preinvasive lung cancer [5]. This device is approved by the Food and Drug Administration for the detection of early lung cancer [5] and is similar to conventional bronchoscopy except for differences in the illuminating light and the addition of a special camera [5]. After a bronchoscopist has received extensive training in using this device, autofluorescence bronchoscopy adds a few minutes to a conventional bronchoscopic procedure. A recent multicenter trial using the light-induced fluorescence endoscopy device showed that it improved the detection rate of preinvasive lung cancer by several-fold compared with conventional bronchoscopy alone [69].
Although this technique is promising, its invasive nature and cost are limiting factors for widespread screening. As this technology matures, the cost will likely decrease. At present, for widespread screening, autofluorescence bronchoscopy should ideally be coupled with a noninvasive, first-line study that selects patients with a high pretest probability of harboring early lung cancer. For example, a recent study by Phillips et al. [70] described the use of a breathalyzer to identify volatile organic compounds that may serve as potential markers for lung cancer. Future studies are needed to determine the precise role of this exciting new technology in the detection of early lung cancer.
Virtual bronchoscopy.—Virtual bronchoscopy is an advanced visualization technique in which helical CT data and virtual reality computing are used to create three-dimensional endobronchial simulations [71,72,73] (Fig. 4). A preliminary investigation by Summers et al. [72], which was published in 1998, assessed the computer-assisted detection of polypoid airway lesions on virtual bronchoscopy images. This technique was associated with a relatively high sensitivity (90%) for lesions larger than 5 mm but was limited by a high number of false-positive diagnoses.
|
Major limitations of virtual bronchoscopy include its labor-intensive nature, its relatively high radiation dose, the limited experience of most radiologists with this technique, and its inability to differentiate malignant from benign lesions [71,72,73]. Future technologic advances will hopefully overcome many of these obstacles.
Future Directions: Where Do We Go from Here?
|
TopIntroductionScreeningEmerging Technologies: Methods...Future Directions: Where Do...References |
|---|
Low-dose helical CT appears to be one of the most promising techniques for early lung cancer detection. However, important challenges must be considered before proceeding to mass screening of lung cancer with this or other emerging technologies. Both noncomparative and randomized trials will be important. Although the former allow a relatively rapid assessment of an emerging technology, the latter are important for determining the effect of screening on disease-specific mortality. A proven reduction in lung cancer mortality at an acceptable cost will likely be an important prerequisite before a new technology is used for mass screening [30]. National trials are being planned for this purpose.
Acknowledgments
We thank Morris Simon and Ellen C. Boiselle for their thoughtful review of
this manuscript. We also thank Nancy Williams for administrative
assistance.
|
TopIntroductionScreeningEmerging Technologies: Methods...Future Directions: Where Do...References |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  B. E. Chapman, D. F. Yankelevitz, C. I. Henschke, and D. Gur Lung Cancer Screening: Simulations of Effects of Imperfect Detection on Temporal Dynamics Radiology, February 1, 2005; 234(2): 582 - 590. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. P. Ko, H. Rusinek, E. L. Jacobs, J. S. Babb, M. Betke, G. McGuinness, and D. P. Naidich Small Pulmonary Nodules: Volume Measurement at Chest CT--Phantom Study Radiology, September 1, 2003; 228(3): 864 - 870. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. Li, S. Sone, H. Abe, H. MacMahon, S. G. Armato III, and K. Doi Lung Cancers Missed at Low-Dose Helical CT Screening in a General Population: Comparison of Clinical, Histopathologic, and Imaging Findings Radiology, December 1, 2002; 225(3): 673 - 683. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Garg, R. L. Keith, T. Byers, K. Kelly, A. L. Kerzner, D. A. Lynch, and Y. E. Miller Randomized Controlled Trial with Low-Dose Spiral CT for Lung Cancer Screening: Feasibility Study and Preliminary Results Radiology, November 1, 2002; 225(2): 506 - 510. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. M. Boiselle, K. F. Reynolds, and A. Ernst Multiplanar and Three-Dimensional Imaging of the Central Airways with Multidetector CT Am. J. Roentgenol., August 1, 2002; 179(2): 301 - 308. [Full Text] [PDF]
|
|
|
|
 |
|
 P. M. Boiselle and A. Ernst Recent Advances in Central Airway Imaging* Chest, May 1, 2002; 121(5): 1651 - 1660. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
