The Theory of a Screening Test

Screening, sometimes termed “secondary prevention,” is a major component of disease control, the others comprising primary prevention, diagnosis, treatment, rehabilitation after treatment or disability, and palliative care. Ideally, the control of a disease should be achievable either by preventing the disease from occurring or, if it does occur, by curing those who develop it by appropriate treatment. Completely successful prevention would render treatment obsolete. Completely successful treatment would not make prevention obsolete, however, because there are costs and undesirable sequelae of the disease itself (especially diseases such as cancer, diabetes and hypertension) and of the treatment that patients and society would like to avoid if possible. At present, neither prevention nor treatment is completely successful for most diseases; they will continue to complement each other for many diseases, and screening can be regarded as complementary to either of these approaches.

Because of a strong belief among physicians that “early diagnosis” of disease is beneficial, many regard screening as necessarily effective. This belief, however, is not necessarily true, as explained later. This article describes important epidemiologic considerations relevant to screening for colon cancer.

Definition of Screening

Screening, as defined by the Commission on Chronic Illness, is “the presumptive identification of unrecognized disease or defect by the application of tests, examinations or other procedures that can be applied rapidly” . A screening test is not intended to be diagnostic. Rather, a positive finding is confirmed by special diagnostic procedures. Screening, by definition, is offered to those who do not suspect that they have a disease. This lack of suspicion is subtly different from being asymptomatic. Symptoms may be revealed by careful questioning related to the organ of interest, not regarded by the individual attending for screening as being related to a possible disease.

An ideal screening test is simple, inexpensive, accurate in terms of being positive if disease is present (sensitivity) or negative if disease is absent (specificity), relatively safe, and acceptable to the patient. Some of these desirable features of an ideal screening test are lacking in the currently available screening tests for colon cancer. Two of these features, sensitivity and specificity, measure a screening test’s validity. Clinicians, however, also are interested in another feature, the predictive value of the screening test, as a practical measure of test utility. The positive predictive value, the proportion of those who test positive who actually have the disease, is of great interest. False positives, persons who test positive but who are found on further investigation not to have the disease, suffer anxiety and the costs and risks of investigation. The negative predictive value indicates the extent to which a negative test truly indicates the absence of disease rather than being a false-negative test. These attributes of a screening test are interrelated, as illustrated in Table 1 .

There is a reciprocal interrelationship between sensitivity and specificity if the test limits for positive and negative can be changed, when a semiquantitative test is used. Maneuvers that increase sensitivity tend to reduce specificity, and vice versa. This phenomenon occurred when rehydration of the FOBT was introduced in the Minnesota screening trial. Rehydration substantially increased sensitivity but substantially reduced specificity, resulting in an enormous increase in the load on the diagnostic services . Consequently, the European trials of FOBT used nonrehydrated slides, and nearly all organizations that have made recommendations for population-based FOBT screening have followed suit. This example illustrates that in population-based screening specificity should be optimized, even at the cost of some small reduction in sensitivity to reduce costs from screening tests with low specificity . A solution to this dilemma would be to develop a better test with higher sensitivity without poorer specificity—hence the interest in immunologically based FOBTs as successors to guaiac-based tests.

Definition of Screening

Screening, as defined by the Commission on Chronic Illness, is “the presumptive identification of unrecognized disease or defect by the application of tests, examinations or other procedures that can be applied rapidly” . A screening test is not intended to be diagnostic. Rather, a positive finding is confirmed by special diagnostic procedures. Screening, by definition, is offered to those who do not suspect that they have a disease. This lack of suspicion is subtly different from being asymptomatic. Symptoms may be revealed by careful questioning related to the organ of interest, not regarded by the individual attending for screening as being related to a possible disease.

An ideal screening test is simple, inexpensive, accurate in terms of being positive if disease is present (sensitivity) or negative if disease is absent (specificity), relatively safe, and acceptable to the patient. Some of these desirable features of an ideal screening test are lacking in the currently available screening tests for colon cancer. Two of these features, sensitivity and specificity, measure a screening test’s validity. Clinicians, however, also are interested in another feature, the predictive value of the screening test, as a practical measure of test utility. The positive predictive value, the proportion of those who test positive who actually have the disease, is of great interest. False positives, persons who test positive but who are found on further investigation not to have the disease, suffer anxiety and the costs and risks of investigation. The negative predictive value indicates the extent to which a negative test truly indicates the absence of disease rather than being a false-negative test. These attributes of a screening test are interrelated, as illustrated in Table 1 .

There is a reciprocal interrelationship between sensitivity and specificity if the test limits for positive and negative can be changed, when a semiquantitative test is used. Maneuvers that increase sensitivity tend to reduce specificity, and vice versa. This phenomenon occurred when rehydration of the FOBT was introduced in the Minnesota screening trial. Rehydration substantially increased sensitivity but substantially reduced specificity, resulting in an enormous increase in the load on the diagnostic services . Consequently, the European trials of FOBT used nonrehydrated slides, and nearly all organizations that have made recommendations for population-based FOBT screening have followed suit. This example illustrates that in population-based screening specificity should be optimized, even at the cost of some small reduction in sensitivity to reduce costs from screening tests with low specificity . A solution to this dilemma would be to develop a better test with higher sensitivity without poorer specificity—hence the interest in immunologically based FOBTs as successors to guaiac-based tests.

The Conditions Sought by Screening

For appropriate screening, the disease should be an important health problem. Colorectal cancer fulfills this prerequisite. It is the third most important cancer site, with a lifetime incidence of about 5% in both men and women . The goal of screening, however, is not merely early detection; the main goal is reduced mortality. In epidemiologic parlance, this goal is a reduction in disease-specific mortality, that is, a reduction in the numbers of deaths from the disease in a defined population. This reduction is not equivalent to improving the survival of screen-detected cases: the diagnosis is made earlier by screening, resulting in lead-time bias introduced by screening that automatically will increase survival. In length bias, rapidly progressive cancers tend to present clinically in the intervals between screens when screening is applied intermittently, so that cancers detected by screening will have a better prognosis than those that are not. In selection bias, those who volunteer for screening tend to be more health conscious than those who do not and therefore would have a better prognosis even without screening. Thus even if detection improves observed survival, it is not necessarily beneficial in terms of reduced disease mortality.

In addition, some cancers detected by screening will be fatal despite earlier detection. Some cancers detected by screening would have been cured even if they had presented clinically later in the normal way. Some cancers identified by screening never would have presented in the subject’s lifetime, because the subject was destined to die earlier from another cause of death. The latter phenomenon, called “overdiagnosis,” has been demonstrated for breast, lung, and prostate screening but not yet definitively shown for colorectal screening.

Colorectal cancer is, however, different from breast, lung, and prostate cancer in that screening, with the appropriate test, can detect cancer precursors, that is, adenomatous polyps. Detection of cancer precursors reduces the incidence of the relevant cancer, as has been well demonstrated for cervical cancer screening and for colorectal cancer screening using the FOBT . Endoscopic screening also is expected to reduce the incidence of colorectal cancer, but this reduction has not yet been demonstrated in randomized screening trials. The disadvantage of screening for cancer precursors is that most detected lesions will not develop into cancer during the subject’s lifetime (eg, about 75% in the case of colorectal cancer). Persons found to have nonprogressive polyps therefore will be subjected to unnecessary diagnostic tests and treatment. Furthermore, colorectal cancers may arise from “flat” adenomas that are not detectable by currently available screening tests. As yet it is not known whether or how often this phenomenon occurs. Much of this information will come from ongoing randomized trials of flexible sigmoidoscopy in the United States and Europe for cancers within reach of the sigmoidoscope.

The Ethics of Screening

The interaction between a patient and a physician is governed by ethical principles. Patients generally consult physicians for symptoms or for reassurance that all is well. Because screening is promoted by professional and sometimes by lay organizations, patients increasingly initiate the screening in the expectation of personal benefit. This places the onus on ensuring that patients benefit from the screening on the organizations that promote screening and on the physicians who administer the screening tests. It is unethical to promote screening unless patients benefit from these tests. Benefit is demonstrated unequivocally only by randomized screening trials . A major problem is that only a small proportion of those screened will benefit from the screening. In North America, the average man or woman has a lifetime risk of dying from colorectal cancer of 2% to 3% . The randomized trials of FOBT screening on average showed only a 20% reduction in mortality, or less than 1 in 100 deaths. The same computation related to incidence indicates a benefit of 1 to 2 per 100; that is, 98% of those screened by FOBT derive no benefit, and many of them incur risks of harm from diagnostic interventions and inconvenience and expense of the screening and the diagnosis that follows. The physician recommending the screening therefore must ensure that the potential benefits and harms are described accurately. Further, the physician should ensure that those patients in whom an abnormal screening result is reported do attend for diagnosis—and treatment if necessary—and are negotiated through the difficulties in reaching the correct conclusion with the accepted treatment.

Implementation of Colorectal Screening

In Europe and Canada, organizations concerned with cancer control established mechanisms to implement screening, and many now are applying these mechanisms to colorectal cancer. They generally follow the principles established many years ago for organized screening programs for cervical cancer :

• •
  

  The target population is identifiable.

  
  
• •
  

  Measures are available to guarantee insurance coverage and patient attendance.

  
  
• •
  

  Enough facilities are available that perform high-quality screening tests.

  
  
• •
  

  The referral system for diagnosis and treatment of abnormalities is effective.

  
  
• •
  

  There are adequate facilities for diagnosis and treatment.

Although the Centers for Disease Control has established some organized programs for disadvantaged groups, especially for breast and cervical cancer screening, screening in the United States generally is not implemented in an organized way. This lack of a unified strategy results in part from the absence of the health care organizations that are active in other countries, which often are pejoratively called “socialized medicine” in the United States.

Expecting all subjects in the target group to volunteer for screening, labeled “opportunistic screening,” has been shown in cervical cancer to be largely ineffective on a population and disease-control basis and to lead to major overexpenditures from unnecessarily frequent screening and investigations . Despite this experience, gastroenterologists in the United States probably will continue to have the implicit responsibility of ensuring screening of referred patients. Clearly, gastroenterologists cannot be expected to recruit people who do not consult them. It is, however, reasonable for gastroenterologists to promulgate screening actively to the public in their catchment area, if opportunities arise (either through the local division of the American Cancer Society, or through the media), and to promote screening actively to family physicians or internists in their area. Gastroenterologists should promote the accepted guidelines and contribute to revisions of these guidelines as new data emerge.

For patients who fall into the target age group for screening, including those referred for other reasons, it is appropriate for gastroenterologists to ensure that their computer system is set up to remind patients automatically of their next screening test date. Some physicians have expressed doubt as to whether mailing screening reminders to healthy people is in accordance with established principles of medical practice. In Canada medical and legal authorities have confirmed that such reminders are appropriate, ethical, and legal. The author doubts that different opinions would be rendered in the United States.