The concept of early diagnosis of disease as a preventive health care measure has been developed over the last 35 years through clinical trials of screening strategies and a theoretical statistical literature. These data have aided clinicians to evaluate specific screening interventions, as and have demonstrated that more screening is not necessarily better care. The decision to recommend an intervention for an individual patient or for a population is dependent on the ratio of benefit to harm. Benefits are best quantified as an absolute risk reduction in adverse outcomes. Harms include complications arising from investigation, adverse effects of treatment and the costs and inconvenience incurred during investigation and treatment. These harms are particularly important in those who have positive results, but inconsequential disease – disease that would not become clinically manifest in the individual patient’s lifetime. More difficult to quantify are the anxiety generated by testing and treatment and the negative effects on insurability, career advancement and emotional well being of "labeling" someone with a disease.

The term "screening" refers to the testing of asymptomatic patients. We screen for early disease and for risk factors for disease or injury (such as smoking, hyperlipidemia, or non-use of seatbelts). Both are important, but this chapter will emphasize screening for disease. Four criteria are required to make a disease appropriate for screening:

1. The disease must have a detectable preclinical period, before it becomes clinically apparent and during which it can be detected by the screening test. Cervical intraepithelial neoplasia can be clinically silent for years, yet easily diagnosed by a screening test.
2. The detectable preclinical period must allow detection while cure is possible. Amyotrophic lateral sclerosis (ALS), for example, is incurable regardless of when it is detected and so does not meet this criterion.
3. Treatment must be more effective if given earlier (at the time of screen detection) than later (at the time the disease becomes clinically apparent).
4. The disease must be sufficiently common in the population. The prevalence of a disease, together with the sensitivity and specificity of the test, determine its positive and negative predictive values.

For a screening test to be effective, it must be sensitive, specific, have high predictive value, be feasible for broad use, and acceptable in terms of cost, risk and patient tolerability.

The current standard for evaluating screening strategies is the randomized clinical trial. The endpoint on which interpretation of these studies rests is the absolute reduction in age specific mortality rates. When reading the results of a screening trial, it is important to take careful note of the type of mortality rate the authors choose to report. The least useful is the case-fatality rate (the number of deaths from a disease divided by the total cases during a specified interval of time) because this can be lowered by an increase in the number of cases of inconsequential disease in the screened group, an artifact of screening. More useful rates are cause-specific mortality (probability of death due to a specific disease) and all cause mortality (probability of dying of any cause). Some experts advocate use of all cause mortality because it will help uncover adverse effects of screening or treatment. Others note the larger sample sizes needed to show effect on all cause mortality for relatively uncommon or non-fatal diseases and the difficulty of implementing such large trials.

Disease detected by screening is likely to be found at an earlier stage than that found in patients with signs or symptoms of illness. People with screen-detected disease may, therefore, live "longer" and have a lower case-fatality rate, even in the absence of effective treatment. This phenomenon is termed lead-time bias. In addition, screening detects a mix of aggressive and indolent disease. The screen-detected mix is a point prevalence sample and therefore overrepresents indolent disease compared to the overall mix among incident cases. This is termed length bias. Both lead-time and length bias favor screening even in the absence of true benefit unless cause-specific mortality is used as the study endpoint.

For some clinical situations, the evidence to recommend a screening strategy is strong; in others, it is weak or non-existent. Periodically, debates about specific screening tests, such as mammograms or PSA, become front-page news. In all cases, the active solicitation of informed patient preference is essential. In addition, recommendations should not be static – they will and should change as newer tests and treatments emerge. As with all patient care, screening must be individualized, but the basis of good practice is to understand current recommendations and the evidence on which they are based.

In 1989, the Preventive Services Task Force (USPSTF), commissioned by the Department of Health and Human Services, published a survey of strategies for the prevention of disease. The second edition of the Guide to Clinical Preventive Services was published in 1996; the third edition is expected shortly and parts are available on the Internet (www.ahrq.gov/clinic/uspstfix.htm) where individual chapters are published as they are completed. The Guide is a cornerstone of evidence-based medicine and the basis of many of our recommendations. However, multiple screening guidelines – including those from the National Cancer Institutes (NCI), the American Cancer Society (ACS), subspecialty organizations, and patient advocacy groups – exist and do not always agree with the Guide or with each other. Only familiarity with the evidence underlying recommendations and the principles of clinical epidemiology stressed above can assist us to evaluate competing guidelines.

The Guide recommends some screening tests for all adult patients (Table 1) and others for those in specific risk groups (Table 2). Many of the screening procedures in these tables are discussed in other chapters in this syllabus. Although there are almost no data from randomized trials on patients over 75 years old, a useful framework has been developed by Walter et al.

Table 1: Screening tests recommended for all adults

Table 2: High risk groups (recommended screening tests in parentheses)

The process of individualizing and seeking informed patient preference is critical and complex. We do not ask permission to measure blood pressure in asymptomatic patients – we believe that benefits outweigh harms and assume that patients have given tacit permission to check vital signs. We do specifically ask permission to do a Pap smear, but strongly urge reticent patients to undergo the procedure because the benefits clearly outweigh the harms (see below). We should admit to our uncertainty about certain procedures (e.g., PSA) and discourage others for which we feel the harms outweigh the benefits. When there is more than one reasonable option, we should present them to the patient to help in decision making. The Guide’s website has several links to bilingual patient education materials, as does the ACS (www.cancer.org); more are available from the Medicine Clinic website (www.medicineclinic.org).

While virtually undetected in unscreened populations, the annual incidence of cervical carcinoma in situ (CIS) in screened populations is 130 per 100,000 with a lifetime incidence of 2,000 per 100,000. In populations where CIS is treated when it is detected, the lifetime incidence of invasive cervical cancer drops from the 2,500 per 100,000 seen in unscreened populations to 700 per 100,000 and a woman’s lifetime risk of dying from cervical cancer drops from approximately 1 percent to 0.3 percent. The standard screening procedure for cervical cancer is the Papanicolaou (Pap) test. Introduced before the advent of randomized trials, it has many of the characteristics of a good screening test. The test is safe, relatively inexpensive, and has a low false-positive rate (0.2 to 0.4 percent). It has a fairly high false-negative rate (5 to 50 percent), but given the relatively indolent nature of cervical intraepithelial neoplasia, the impact of a single false negative is diminished by periodic reexamination. Techniques to increase sensitivity are being developed.

Thus, despite the lack of a randomized trial, periodic Pap testing reduces mortality from cervical cancer by as much as 90 percent. Current debates concern only which age group to screen and optimal interval between screens. Annual testing appears unnecessary. Screening women more often than every three years does not increase survival in those in whom CIS is found, since finding the lesion in its first year confers little additional mortality benefit compared to finding it in its third year. Additionally, more frequent screening will detect a higher proportion of transitory dysplasia.

The developing consensus is to test women who are or have been sexually active beginning at the age of onset of sexual activity and continuing to the age of 65. Screening should be performed at least every three years. In women over 65 who have not been screened, screening should be instituted and continued until at least two high-quality Pap results are negative. For women who have been regularly screened prior to age 65 and who have had negative smears, the cost-effectiveness of continued screening is low, and the likelihood of a false positive result high.

Although cervical cancer is clearly associated with certain types of human papilloma virus infection (most notably HPV 16 and 18), routine screening for HPV infection itself is not yet recommended. As PCR technology is standardized, however, testing for specific oncogenic strains may be helpful in directing screening periodicity.

There are several large subgroups of women, to whom the recommendations above do not apply. HIV-infected women have higher rates of cervical cancer and greater risk of invasive cervical cancer; more frequent screening is required. Conversely, women with a history of total hysterectomy for benign disease have a very low probability of abnormal findings when Pap smears are done of the vaginal apex, and do not need further cervical cancer screening.

While lung cancer is now the leading cause of cancer death in women, breast cancer remains the most common non-skin cancer and one of the leading causes of death in women ages 35 to 55 years. Five-year survival is strongly dependent on the stage at which the disease is diagnosed. The effectiveness of early detection, specifically mammography, has again been recently questioned, making front page news. Controversies about breast cancer screening have been newsworthy for years, but the current debate stems from a meta-analysis intended for the Cochrane Collaboration and featured (without Cochrane-required modifications) in The Lancet with an accompanying supportive editorial, letters and longer responses.

Screening procedures include breast self-examination (BSE), clinical breast examination (CBE), and mammography. BSE is safe and inexpensive but ineffective. Sensitivity is low and the false positive rate is currently unknown. A very large prospective randomized trial of BSE in 266,064 female factory workers in China was recently completed and showed no difference in breast cancer mortality at 10 years. Women in the BSE arm found more benign lesions than those in the control group, but there was no evidence that cancers were discovered at an earlier stage; one characteristic response to the results of this study is an editorial titled "Routinely teaching breast self-examination is dead: what does this mean?" Based on these findings, we view the teaching of BSE as discretionary.

A systematic review of the literature on CBE concluded that sensitivity is 54 percent and specificity is 94 percent. While there have been no clinical trials of CBE alone, in studies of breast cancer screening a small proportion of cancers that were not seen on mammography were detected by clinical examination. Like SBE, CBE is safe, inexpensive and an opportunity to provide patient education about breast cancer screening; we recommend annual clinical breast exam for all adult women.

Mammography provides a substantial improvement over manual examination. Sensitivity is 50 to 87 percent (depending on patient age) and specificity is 94 to 99 percent. However, the risk of a false-positive mammogram is still relatively high for women who obtain regular mammography and some ductal carcinoma in situ (DCIS) may be non-progressive and therefore inconsequential. One retrospective cohort study showed that the risk of a false-positive mammogram over a ten-year period approached 24 percent. Some authors suggest that this number may approach 50 percent, but the definition of false positivity has varied in different studies and often included patients in whom a simple repeat view was requested.

There have been 11 large clinical trials of mammography, eight of which were randomized, controlled studies. Although the recent Danish meta-analysis questioned the methodologic rigor of six of these eight trials, the pooled results show a decrease in age-adjusted mortality from breast cancer of approximately 30 percent. The controversy over the methodologic rigor of the RCTs highlights differences in randomization techniques; change in quality of radiologic equipment and interpretation over the past 40 years; definitions of false positivity; length of follow up and types of mortality rates used (disease specific or all-cause) and can be found in the referenced articles, letters and editorials. It is extremely unlikely that there will ever be another RCT of mammography. However, the raw data from some of the prior RCTs have been reanalyzed and continue to show significantly decreased breast cancer mortality.

Routine mammography continues to be endorsed by every organization that publishes screening guidelines. However, there are differences in recommendations on when to start screening, how frequently to screen and when to stop screening.

The question of when to initiate screening mammography is complicated by the lack of data. Only one clinical trial was designed to accrue enough patients to address this question, and this study (the Canadian National Breast Screening Study) is considered by some to be fatally flawed by design weaknesses. Only one trial (Gothenberg) shows a statistically significant difference in mortality between women who began screening at 40 and those who began screening at 50. Meta-analyses that include all clinical trials of screening mammography show a 17 percent advantage for mammography which is not significant, although a mortality benefit can be shown after 12-14 years. The Gothenberg trial does report than the significant benefit observed resulted predominantly from tumors diagnosed before age 50. They also kept a tally of numbers of tumors detected on each arm of the study and found them to be the same, suggesting that the DCIS’s detected in the mammography arm became clinically important tumors that contributed to the higher mortality in the unscreened arm. In contrast, meta-analyses that exclude the Canadian trial do show a statistically significant difference.

Because the prevalence of breast cancer among young women is lower, and because higher breast density (such as that seen in premenopausal women) decreases the sensitivity of mammograms, the positive predictive value of an abnormal mammogram is lower in a 40-year old than in a 50-year old. In addition, starting to screen at the age of 40 means that each woman may have 35 mammograms in her lifetime. (In contrast, the randomized studies screened women with a total of only 2 to 5 mammograms). The topic is controversial, but the National Cancer Institute and the American Cancer Society now recommend mammography for all women 40 and older. The USPSTF Guide had not endorsed mammography for women under 50 until it released revised recommendations to initiate testing at 40 at a highly publicized news conference in February 2002.

At present, it is fair to say that initiating screening mammography at 40 instead of at 50 may improve mortality from breast cancer. As Drs. Shea and Antman point out, age, risk of breast cancer and cost-effectiveness are all continua –no abrupt change occurs at age 50. Given this uncertainty (described as a "toss-up" by some), it is reasonable to leave the decision to the patient and provider. Women at higher risk may choose to be screened earlier, and it is now standard practice to screen women with known BrCa1 and 2 much earlier. There is no substitute for informed patient participation in this sort of decision-making.

No clinical trial has been designed to answer the question of how often women should undergo screening mammography, and the optimal schedule is thus controversial. Enthusiasm for annual mammography is tempered by the low prior probability of new disease (between 0.02 percent and 0.18 percent depending on age), and the positive predictive value of only 1 to 13 percent (depending on test performance). In contrast, subgroup analyses in the Gothenberg trial suggest that the shorter the interval between mammograms, the larger the mortality benefit.¹⁷ Current recommendations from the ACS, NCI and the Division of Oncology at New York Presbyterian Hospital are to screen with regular mammography after the age of 40.

The USPSTF concluded that the existing evidence is generalizable to women aged 70 and older whose life expectancy is not compromised by comorbid disease. The absolute probability of benefits of regular mammography increase along a continuum with age, whereas the likelihood of harms from screening (results which require additional evaluations, anxiety, biopsies, and cost) diminish from ages 40-70.

Women over the age of 75 were not included in any of the randomized mammography trials. Thus the usefulness of mammography in women over the age of 74 who have previously been screened is unclear. Many practitioners use functional status rather than age as an indication to stop screening; if a patient has a limited life expectancy or comorbid conditions that would preclude lumpectomy and tamoxifen, few would recommend continuing screening mammography. In contrast, an active 85 year-old willing and able to undergo surgery and radiation therapy if breast cancer is detected might be a good candidate for continued screening.

Like breast cancer, five-year survival for colorectal cancer (CRC) is strongly dependent on the stage at which the disease is diagnosed, ranging from 90 percent for localized disease to six percent for cases with distant metastases. Unlike breast cancer, there is no controversy over the utility of screening for colon cancer, which has been clearly shown to decrease age-adjusted disease-specific mortality.

Screening procedures for CRC include fecal occult blood testing (FOBT), flexible sigmoidoscopy (FOS), and colonoscopy. Digital rectal examination is a useful method for conducting FOBT, but palpation is inadequate as a primary screening method; more than 85 percent of lesions are beyond reach. Biologic markers, such as carcinoembryonic antigen (CEA), are useful to monitor disease relapse but are insufficiently sensitive and specific for use in screening.^, However, a recent report demonstrates the feasibility of detecting mutations in the adenomatous polyposis coli (APC) gene in fecal DNA, an approach to the diagnosis of early colorectal cancer that is likely to be developed in the near future.

FOBT can detect blood from a wide variety of conditions, including colorectal cancer, benign or premalignant adenomata, gastric carcinoma, peptic ulcer and gastric irritation due to nonsteroidal anti-inflammatory agents. False positives can be caused by the ingestion of a variety of medications. Nevertheless, in an asymptomatic population over the age of 50, FOBT has a positive predictive value of 5 to 10 percent for cancer and 30 percent for adenomata.⁹ The sensitivity of FOBT has been as high as 92 percent in some studies of patients with known malignancy, but is probably much lower in asymptomatic populations (hence lowering the negative predictive value). The sensitivity is even lower for small adenomata than for malignant lesions, a disadvantage given the goal of CRC screening to detect and remove premalignant adenomata before they become malignant. In one study of 248 patients with guaiac-positive stool but without iron deficiency or active GI bleeding, upper gastrointestinal lesions were found more frequently than colonic lesions (21.8 percent vs. 28.6 percent).

The best and largest randomized trial conducted in persons 50 to 80 years of age compared the outcomes of three strategies: annual screening (three rehydrated Hemoccult cards), biannual screening, and control (no screening). Patients in the first two groups with at least one positive Hemoccult test were evaluated with colonoscopy. The incidence of cancer was comparable in the three groups. Mortality from colorectal cancer was 33 percent lower in the annually-screened group when compared to the other two groups after 13 years (each group had approximately 15,500 patients with 82, 117 and 121 colorectal cancer deaths respectively), with no difference in mortality between the biannually screened group and the controls. Other large trials have replicated these results and support the conclusion that annual FOBT screening is superior to biannual screening. We recommend annual use of the protocol followed in this study. Patients should be instructed to avoid red meat and NSAIDS for several days prior to testing and to collect the samples on three consecutive days. Compliance with this strategy may be lower in our setting than in the study population and therefore make colonoscopy the preferred option (see below).

The performance of FS is limited by the amount of colon it can reach. Longer scopes can visualize just under half of the colon and, not surprisingly, sensitivity of sigmoidoscopy is approximately 45 percent that of colonoscopy. A combined strategy of one time sigmoidoscopy plus FOBT missed advanced colonic neoplasia in the proximal colon of 24 percent of patients studied. Specificity of the test depends on what is being sought. Detection of benign polyps is frequent and is often considered a false positive result. At the time of discovery, of course, there is no way to know the malignant potential and all polyps must be removed. Removing adenomas appears to decrease the incidence of later malignant transformation.

There are no randomized clinical trials of screening sigmoidoscopy, although case-control studies have produced suggestive results. One case-control study showed that endoscopic procedures of the large bowel (including sigmoidoscopy and colonoscopy) were associated with a 50 percent reduction in the risk of developing colorectal cancer. Another case-control study compared patients who had had a rigid sigmoidoscopy with those who had not; the former had an odds ratio for distal colorectal cancer of 0.41 (0.25-0.69). There was no difference in the rate of proximal colorectal cancer – i.e. cancer diagnoses dropped significantly only for the area of the colon visualized by the sigmoidoscope. The American Cancer Society recommends flexible sigmoidoscopy at age 50 for patients with no risk factors. At CPMC, sigmoidoscopy has been replaced by colonoscopy for colorectal cancer screening (see below).

There are no published trials of screening colonoscopy. Although clearly more sensitive than sigmoidoscopy, even colonoscopy is not 100 percent sensitive. A 1997 study of 183 patients undergoing two colonoscopies on the same day showed miss rates of 6 percent for adenomas >1 cm, 13 percent for adenomas 6-9 mm and 27 percent for adenomas smaller than 5 mm. The overall complication rate of colonoscopy is 1-2 percent and the risk of colonic perforation has been estimated to be 1-2 per thousand as compared with 1 in ten thousand in published trials of sigmoidoscopy. Colonoscopy is also considerably more expensive than FOBT or FOS. The most important factors in determining cost effectiveness for colonoscopy relative to FOBT are the charge for the colonoscopy (cutting the charge increases cost effectiveness of colonoscopy) and the compliance rate of patients (FOBT loses cost effectiveness if patients get fewer than ten).^,

Although colonoscopy has not been extensively evaluated as a screening strategy, it has continued to gain adherents as a good option in the 14 years since two Columbia P&S faculty first suggested its widespread use.^, The procedure is now covered by Medicare once every 10 years for patients at average risk and is being recommended by the American Cancer Society as part of its "Get the test. Get the polyp. Get the cure." campaign.

As Dr Ransohoff states in his recent review, "the variation in recommendations should not obscure the larger message that screening can reduce the rate of death from colorectal cancer...at present, physicians should not be dogmatic about which test to use."⁴⁸ Like other screening tests, the active solicitation of informed patient preference is essential and the decision about a particular strategy will depend on patient’s preferences and other individual characteristics such as compliance. Poorly compliant patients may be better able to make one appointment for colonoscopy every ten years than to perform fecal occult blood testing annually.

Prostate cancer is the leading cause of non-skin cancer in American men, and the second leading cause of cancer death. It meets almost all of the criteria for a "screenable" disease – it is prevalent, serious, and treatable and has a preclinical detectable period. In addition, mortality from prostate cancer has fallen seven percent since 1992, which some attribute to the introduction of PSA screening. Why, then, is there so much controversy about the utility of prostate cancer screening? The two answers are closely related – first, a mortality benefit from early treatment of asymptomatic low-grade disease has not been proven, and second, no randomized controlled studies have been completed.

A single randomized trial has been published, and suggests a mortality benefit – analysis was not based on intent to treat, however, and its critics await further studies. Such trials are underway, and recommendations about screening may change over the next few years, as interim data become available. The two largest screening trials – the National Cancer Institute’s prostate, lung, colorectal and ovarian (PLCO) trial and the European Randomized Study of Screening for Prostate Cancer (ERSSPC) – include 37,000 and 135,000 men, respectively and final data are expected starting in 2006. An intervention trial, the U.S. Prostate Cancer Intervention Versus Observation Trial (PIVOT) has enrolled over half of its target 1,050 patients.

Until these data are available, clinicians must decide what to tell patients about prostate cancer screening. Because indolent disease is extremely common and the treatment of advanced disease is rarely curative, the utility of screening for this disease has been challenged. As Whitmore has asked, "When cure is possible, is it necessary? When cure is necessary, is it possible?" Others counter that indolent disease in a younger man may progress to advanced disease over time, and that early stage but intermediate or high grade lesions progress rapidly and should be treated.

Screening for prostate cancer has been complicated by the inability of screening tests to distinguish between three clinical settings: detection of an asymptomatic cancer that would never have caused illness, detection of an asymptomatic cancer that has already escaped from cure, and detection of an asymptomatic cancer which can be cured with therapy. It is clear that many cancers identified by prostate cancer screening fall into one of the first two groups and equally clear that the goal of screening should be to detect those in the third – those with clinically significant and still-treatable disease.

The Guide to Clinical Preventive Services, last published in 1996, summarized studies of the effectiveness of digital rectal exam, transrectal ultrasonography and serum tumor markers (such as prostate specific antigen) for detection of and reducing mortality from prostate cancer. The sensitivity and specificity of these tests is variable, and there was no evidence to support the hypothesis that early detection and aggressive treatment of prostate cancer leads to improved mortality (after taking into account lead time and length bias). The Task Force concluded that there was no evidence to support screening for prostate cancer and, in fact, recommended against screening.

Since the Guide’s most recent edition, additional studies of prostate cancer screening,^, have fueled the controversy surrounding this intervention continues.^,,, PSA testing, always extremely sensitive, has become somewhat more specific with the addition of PSA isoenzyme testing,^, PSA density corrections, age-specific and race-specific PSA cutoffs and PSA velocity observations. Treatment of screen-detected prostate cancer has advanced, with some data suggesting a mortality benefit for selected patients. We recognize the significant controversy about prostate cancer screening and that the American Urological Association and the American Cancer Society advocate screening at this time, however, we do not recommend routine screening of all asymptomatic men. We endorse the guideline of the American College of Physicians, which suggest that "as a matter of routine, physicians should describe the potential benefits and known harms of screening, diagnosis, and treatment; listen to the patient’s concerns; and then individualize the decision to screen."

Even when physicians agree that patients should have preventive health care screening, such screening is often underused. Physicians do not refer patients for screening tests as often as they should, and when they do, the tests are often not completed. In addition, physicians significantly overestimate their own performance of screening tests. Low income, Hispanic and African-American ethnicity, low educational attainment and age over 65 are associated with underuse of mammography. Similarly, there are marked racial differences in mortality from cervical cancer – African American women are twice as likely to be diagnosed and two to three times as likely to die of the disease as are white women – suggesting differences in access to screening as well as to treatment.

There are multiple studies of interventions aimed at improving the utilization of screening tests. A trial of multiple outreach programs in public housing projects increased screening mammography among Hispanic women in Los Angeles from 12 to 27 percent. The use of lay health advisers and a nurse practitioner increased screening mammography and Pap smears in a clinic serving low-income women. Computerized reminder systems for providers increases rates of screening tests, as do reminders by nurses or other trained personnel; even the use of screening flow sheets in charts has been shown to improve provider adherence to screening recommendations. Entrepreneurs have increasingly marketed expensive high technology screening tests directly to patients who are willing to pay cash – no doctor referral necessary. The clinical, financial, ethical and professional issues involved are nicely discussed in a recent article.

It is an auditable standard of care to offer preventive services to appropriate patients, to document the counseling provided and the actions taken, and to follow up with results of screening tests.

We thank Dr. Karen Antman for helpful comments and suggestions.