Skip to Content
Cancer Imaging Program (CIP)
Contact CIP
Show menu
Search this site
Last Updated: 10/28/16

Screening and Early Detection

Charles Putman, M.D., Chair

Lung Cancer Section Prepared by:

Edward Patz, M.D.
Department of Radiology
Duke Medical Center
September 22, 1998


Lung cancer is the leading cause of cancer death for men and women in the United States, and more people die annually of lung cancer than of colon cancer, breast cancer, and prostate cancer combined. Screening trials with chest radiographs and/or sputum cytology, even targeted at high risk individuals, has so far been ineffectual. It is becoming increasingly clear that effective screening will depend not only on early detection, but a complimentary understanding of the biology of this disease. Even very small tumors can metastasize early and ultimately lead to death in most patients. This reflects the tremendous spectrum of biologic behavior of lung cancer, even among tumors with identical histologic classification. To believe that only larger and undifferentiated tumors metastasize is naive. Tumor cells are directly contiguous with "leaky" abnormal blood vessels and are probably continuously shed into the blood. The ability to seed distant sites and grow is likely dependent on genes which control cell adhesion, migration, and invasion, most of which are yet unknown or poorly characterized.

The most common early radiologic manifestation of lung cancer is a pulmonary nodule, and now with the requisite resolution of CT, very small lesions ( ~ 1mm; representing ~1,000,000 tumor cells) can be seen. This results in a significantly improvement over chest radiographs which typically identifies 1 cm lesions (~1 billion cells; much later in the biology of this disease).

Although CT is an extremely sensitive non-invasive study used to identify and localize lung cancer, the specificity is less than optimal. Differentiating benign from malignant lesions can be difficult because of the overlap in radiographic manifestations. Many small nodules remain indeterminate despite thorough radiological evaluation. In these cases, patients may be followed which might lead to progression of disease, or they may proceed to an invasive procedure for a tissue diagnosis because of the concern for cancer. These invasive procedures, including percutaneous, bronchoscopic or surgical biopsy, always have associated morbidity and expense. A non-invasive functional study which could compliment anatomic imaging would have tremendous clinical significance.


The Task Force discussed two basic issues confronting screening and early detection trials:

  1. the population to screen,
  2. the methods for screening.

If studies are to be constructed then both of these issues need to be addressed. It was decided that justification of a high-risk population was required. At this time it would probably include those who are smokers and possibly individuals with emphysema. Unfortunately, an increasing proportion of lung cancer patients are nonsmokers but until a better way of constructing trials and identifying high risk individuals is established, this appears to be the most logical group of individuals to screen.

The second issue discussed focused on the methods for screening. It is apparent that cost-effective, easy, noninvasive, readily available methods are essential for mass screening programs. Although some trials have included bronchoscopy particularly with the LIFE bronchoscope, this realistically will not be acceptable for mass screening. These invasive tools may be helpful for diagnostic purposes only.

At this time it appears there are three potential areas where noninvasive studies might be developed:

  1. radiologic imaging,
  2. blood analysis,
  3. sputum analysis.

As above, imaging with CT can detect very small lesions, but CT lacks specificity and alone probably will not be able to provide the requisite information for a screening test. CT needs to be complimented with other noninvasive studies. This may include the development of tumor specific imaging agents which is a difficult but promising area of research with tremendous potential. If the appropriate targets could be identified, then imaging agents could not only localize sites of tumor, but produce a "tumor profile", correlating to the biologic behavior.

Another potential method would be to use currently available techniques (CT), in combination with blood or sputum biomarkers to predict the probability and then the behavior of lung cancer. While there multiple biomarkers have been tested, none so far have been used in the appropriate trials demonstrating clinical applicability. However, integration of biomarkers with imaging studies may serve as a new model system for screening trials. This will not only provide us with increased specificity but hopefully we will be able to create a "tumor profile" thus determining some of the biologic activity and possible therapeutic strategies for treating these patients.

If we are to move forward with creating early detection screening trials we must integrate imaging with understanding of molecular medicine. There are studies which can be constructed at this time with the appropriate modeling systems and statistical design.

Methodology and Modeling Section Prepared by:

Constantine Gatsonis, PhD
Center for Biostatistics
Brown University
April 27, 1998

  1. Early detection of cancer provides the basis for a realistic strategy to improve cure rates and/or achieve longer survival and better patient functioning and quality of life. Recent advances in the development of diagnostic modalities based on imaging, biological markers, and genetics, provide a growing array of powerful new technologies with potential for screening. However, the methodologic aspects of designing screening interventions and evaluating their impact are not well understood.
  2. Published approaches to evaluating the effectiveness of screening interventions have included both randomized and observational studies. Traditional endpoints have been mortality and disease stage at diagnosis; more recent studies have also reported on costs and quality of life outcomes. Randomized studies of the effect of screening on mortality usually involve large numbers of participants, recruited and observed over long periods of time. The size of these studies is made necessary by the fact that screening interventions are applied to mostly disease-free individuals and the ultimate impact of screening on mortality is mediated by the available treatment strategies. A number of additional biases are often present even in randomized screening studies, leading to lengthy and often unresolvable discussions about the validity of the conclusions from such studies. In addition, the long duration of these studies may often diminish the relevance of the conclusions because of the evolution of diagnostic technology and treatment approaches.
  3. If the traditional, large randomized trial seems rather crude and inefficient an approach to evaluating relatively mature screening methods, it is clearly imprctical in the face of the proliferation of new diagnostic and treatment methods, which may have potential for screening. A flexible and efficient approach to the design and evaluation of screening interventions could be based on a combination of mathematical modeling and focused empirical studies. The mathematical model would be used to summarize existing knowledge about the course of the disease, the accuracy of the information provided by screening, and the patient outcomes following the screening test. The empirical studies would be directed, on the one hand, toward gathering information needed in speficic aspects of the mathematical model and, on the other, toward evaluating some of the predictions made by the model.
  4. A multi-disciplinary research effort is needed to address the methodologic questions involved in the construction and validation of mathematical models for disease progression and the effects of screening. In addition to physicians and statisticians, the endeavor would involve basic scientists, economists, psychologists, and health services researchers. Main areas of emphasis should include:
    1. developing mathematical models and corresponding estimation techniques for pre-clinical disease,
    2. understanding and modeling heterogeneity across individuals,
    3. designing screening interventions, including optimizing diagnostic performance and scheduling of screening examinations.
    4. understanding and properly quantifying uncertainty in models for decision and cost-effectiveness analysis.
  5. The area of genetic screening needs particular attention, with methodologic emphasis on:
    1. developing risk prediction models for major cancer causing genes, including models that can handle multiple genes and multiple related syndromes,
    2. deriving statistical methodology for estimating gene to gene interactions and their implications for risk prediction and screening, and
    3. developing quantitative models of the accuracy of genetic screening and its determinants.
    4. incorporating existing knowledge into computationally efficient and user friendly software to be used for "real time" counseling.