|Short Contents | Full Contents||Other books NCBI|
|HSTAT: Guide to Clinical Preventive Services, 3rd Edition: Recommendations and Systematic Evidence Reviews, Guide to Community Preventive Services Guide to Clinical Preventive Services, 2nd Edition, 1996 Guide to Clinical Preventive Services: Second Edition (1996) |
section one SCREENING
1. Screening for Asymptomatic Coronary Artery Disease
Burden of Suffering
Ischemic heart disease is the leading cause of death in the U.S., accounting for approximately 490,000 deaths in 1993. 1 The American Heart Association estimates that approximately 1.5 million Americans will suffer a myocardial infarction (MI) in 1995, and one third will not survive the event. 2 Atherosclerotic coronary artery disease (CAD) is the underlying cause of most ischemic cardiac events and can result in myocardial infarction, congestive heart failure, cardiac arrhythmias, and sudden cardiac death. Clinically significant CAD is uncommon in men under 40 and premenopausal women, but risk increases with advancing age and in the presence of risk factors such as smoking, hypertension, diabetes, high cholesterol, and family history of heart disease. Although mortality from heart disease has declined steadily over the past three decades in the U.S., 2 the total burden of coronary disease is predicted to increase substantially over the next 30 years due to the increasing size of the elderly population. 3 The cost of medical care and lost economic productivity due to heart disease in the U.S. has been projected to exceed $60 billion in 1995. 2
Angina is the most common presenting symptom of myocardial ischemia and underlying CAD, but in many persons the first evidence of CAD may be myocardial infarction or sudden death. 4 It has been estimated that 1-2 million middle-aged men have asymptomatic but physiologically significant coronary disease, also referred to as silent myocardial ischemia. 4,5Accuracy of Screening Tests
There are two screening strategies to reduce morbidity and mortality from CAD. The first involves screening for modifiable cardiac risk factors, such as hypertension, elevated serum cholesterol, cigarette smoking, physical inactivity, and diet (see Chapters 2,3, and 54-56). The second strategy is early detection of asymptomatic CAD. The principal tests for detecting asymptomatic CAD include resting and exercise ECGs, which can provide evidence of previous silent myocardial infarctions and silent or inducible myocardial ischemia. Thallium-201 scintigraphy, exercise echocardiography, and ambulatory ECG (Holter monitoring) are less commonly used for screening purposes. The efficacy of each of these tests may be evaluated by (a) its ability to detect atherosclerotic plaque, and (b) its ability to predict the occurrence of a serious clinical event in the future (acute MI, sudden cardiac death).
Several resting ECG findings (ST depression, T-wave inversion, Q waves, and left axis deviation) increase the likelihood of coronary atherosclerosis and of future coronary events. However, these findings are uncommon in asymptomatic persons, occurring in only 1-4% of middle-aged men without clinical evidence of CAD, 6,7 and they are not specific for CAD. One third to one half of patients with angiographically normal coronary arteries have Q waves, T-wave inversion, or ST-T changes on their resting ECG. 8-10 Conversely, a normal ECG does not rule out CAD. In the Coronary Artery Surgery Study, 29% of patients with symptomatic, angiographically proven CAD demonstrated a normal resting ECG. 11 Asymptomatic persons with baseline ECG abnormalities (Q waves, ST segment depression, T-wave inversion, left ventricular hypertrophy, and ventricular premature beats) have a higher risk of future coronary events. 6,12-19 However, prospective studies lasting between 5 and 30 years have found that symptomatic CAD develops in only 3-15% of persons with these ECG findings. 6,13,18,20 Furthermore, most coronary events occur in persons without resting ECG abnormalities. 6,7,18,21,22 Thus, routine ECG testing in asymptomatic persons, in whom the pretest probability of having CAD is relatively low, is not an efficient process for detecting CAD or for predicting future coronary events.
The exercise ECG is more accurate than the resting ECG for detecting clinically important CAD. Most patients with asymptomatic CAD do not have a positive exercise ECG, however. 23-26 ECG changes often do not become apparent until an atherosclerotic plaque has progressed to the point that it significantly impedes coronary blood flow. 24,27 In addition, most asymptomatic persons with an abnormal exercise ECG result (usually defined by a specific magnitude of ST-segment depression) do not have underlying CAD. 27,28 A 1989 meta-analysis found considerable variability in the accuracy of exercise-induced ST depression for predicting CAD (sensitivity 23-100%, specificity 17-100%). 29 Although several investigators reported that adjusting the ST segment for heart rate (ST/HR slope or ST/HR index) improves the ability to predict significant CAD 30-32 and future coronary events, 25 other studies have not shown an advantage. 33-37
The exercise ECG is also more accurate than the resting ECG in predicting future coronary events. While asymptomatic persons with a positive exercise ECG are more likely to experience an event than those with negative tests, 25,38-43 longitudinal studies following such patients from 4 to 13 years have shown that only 1-11% will suffer an acute MI or sudden death. 25,42,44,45 As with resting ECG, the majority of events will occur in those with a negative exercise test result. 24,26,44-47 The pathophysiology of acute coronary syndromes may explain the insensitivity of exercise ECG for subsequent coronary events. Unstable angina, MI, and sudden death often result from an acute, occluding thrombus precipitated by the rupture of a mild, non-flow-limiting plaque. 48-50 Among healthy men who subsequently developed symptomatic CAD after a negative screening test, 73% experienced a MI or sudden death as their initial manifestation. 24,45 In contrast, the majority of asymptomatic persons with a positive exercise ECG develop angina as their initial event. 5,24,45,51 Thus, while exercise ECG may predict the presence of more severe coronary stenosis and risk of angina in asymptomatic persons, it does not accurately predict risk of acute coronary events.
The addition of thallium-201 scintigraphy to conventional exercise testing improves its accuracy in detecting CAD, making it a useful diagnostic test in persons with symptoms of CAD. 52,53 However, the probability of CAD after a positive scan is low in asymptomatic persons, and most coronary events occur in those with a negative test result. 23,44 Because of these limitations and its expense, thallium-201 scintigraphy is not a practical screening test for asymptomatic persons. 23,44,52,54 The ambulatory ECG can detect episodes of ST-segment depression which may indicate silent ischemia in asymptomatic persons with CAD. These episodes, however, also occur commonly in healthy volunteers 55-57 and are not reliable predictors of future coronary events, even in asymptomatic or mildly symptomatic patients with documented CAD. 58,59 There have been no studies of exercise echocardiography in screening asymptomatic populations for CAD.
False-positive screening test results are undesirable for several reasons. Persons with abnormal results frequently undergo invasive diagnostic procedures such as coronary angiography. Abnormal test results may produce considerable anxiety. An abnormal ECG tracing may disqualify some patients from jobs, insurance eligibility, and other opportunities, although the extent of these problems is not known. Proposed strategies for reducing false-positive results include: performing workups in accordance with a Bayesian model 60 using discriminant functions to interpret the stress ECG 41 and targeting testing to high-risk groups.Effectiveness of Early Detection
Although case-control and cohort studies show that asymptomatic persons with selected ECG findings are at increased risk of MI and cardiac death, 5,7,22,25,38-43 there is little evidence that routine screening is an effective means to reduce the incidence of acute coronary events in asymptomatic persons. Antianginal drugs such as nitroglycerin, beta -adrenergic blockers, and calcium channel blockers reduce the frequency and the duration of silent ischemia. 61-63 In a recent study, atenolol reduced the incidence of cardiac events (MI, cardiac arrest, or worsening angina) in patients who had both silent ischemia and CAD documented by angiography or prior MI 64,65 extrapolating these benefits to completely asymptomatic patients with silent ischemia on routine screening may not be justified, given their much lower risk of acute events. 46
Both aspirin therapy and drug treatment for high cholesterol reduce the incidence of MI and cardiac mortality in patients with symptomatic coronary disease, but the balance of risks and benefits of these therapies in asymptomatic patients is not resolved (see Chapters 2 and 69). Benefits are more likely to exceed risks in asymptomatic patients with underlying coronary disease, however, due to their higher absolute risk of MI and coronary death. New diagnostic techniques may prove more sensitive than angiography in identifying the mild-to-moderate plaques that are a risk factor for developing an acute occlusive thrombus. 66,67 Their utility will remain in question, however, until appropriate trials demonstrate that early detection and treatment of small coronary plaques is more effective than treatment based on identifiable risk factors (e.g., high blood pressure or high cholesterol) in asymptomatic patients. 48,49
Among patients with symptomatic coronary disease, coronary artery bypass grafting prolongs life compared with medical therapy in patients with left main coronary or three-vessel disease with poor left ventricular function. 11 The prevalence of high-risk coronary disease among asymptomatic persons, however, is very low while some patients may suffer a MI or sudden cardiac death as their initial manifestation of CAD, most patients with severe coronary disease initially develop angina. 5,45 As a result, it is not clear that the benefit of identifying a small number of individuals with severe coronary disease before they develop symptoms is sufficient to justify routine screening of large populations of asymptomatic persons. Recent randomized trials have demonstrated that percutaneous transluminal coronary angioplasty (PTCA) reduces the frequency of angina in patients with symptomatic CAD, but it does not reduce the incidence of MI or cardiac death. 68,69 The value of coronary angioplasty for asymptomatic coronary stenoses is not known.
A screening ECG has been recommended to provide a "baseline" to help interpret changes in subsequent ECGs. 70 Even when important differences are noted between the baseline ECG and a subsequent tracing, these do not necessarily reflect ongoing or recent ischemia. Using the development of a new Q wave on serial ECG as a criterion, the Framingham Study reported an annual incidence of unrecognized MI of 5.4/1,000 men aged 65-74. 71 Less specific changes develop more commonly than Q waves. Baseline ECGs are often not available when needed for comparison, nor do they significantly contribute to decision making for patients being evaluated for chest pain, 72-75 especially in those with no history of cardiovascular disease. 76 One large study found that a baseline ECG was available in 55% of patients evaluated for acute chest pain. 73 The availability of a prior ECG was associated with small but significant reduction in hospitalization rates for those patients who had chest pain not due to acute MI. Only a small subset of the asymptomatic population is likely to benefit from having a baseline ECG, however: those with baseline ECG abnormalities suggestive of ischemia who subsequently develop acute noncardiac chest pain. Savings from preventing a few unnecessary hospitalizations among these patients must be weighed against the high costs of routine ECG screening in the large population of asymptomatic persons.
Another argument for ECG screening is that the early identification of persons at increased risk for CAD on the basis of ECG findings may help to modify other important cardiac risk factors such as cigarette smoking, hypertension, and elevated serum cholesterol. 70 While the efficacy of risk factor modification is well established, 22,77 no studies have evaluated whether identifying high-risk patients with abnormal ECGs improves efforts to modify risk factors or leads to better clinical outcomes.
Periodic ECG screening is often recommended for persons who might endanger public safety were they to experience an acute cardiac event at work (e.g., airline pilots, bus and truck drivers, railroad engineers). Cardiac events in such individuals are more likely to affect the safety of a large number of persons, and clinical intervention, either through medical treatment or work restrictions, might prevent such catastrophes. No studies have addressed the efficacy of ECG screening in these persons, however.
Preliminary exercise ECG testing has also been recommended for sedentary persons planning to begin vigorous exercise programs, based on evidence that strenuous exertion may increase the risk of sudden cardiac death. The usual underlying cause of sudden cardiac death during exercise is hypertrophic cardiomyopathy or congenital coronary anomalies in young persons and CAD in older persons. Cardiac events during exercise in persons without symptomatic heart disease are uncommon, however, and exercise ECG may not accurately predict those who are at risk. Among over 3,600 asymptomatic, hypercholesterolemic middle-aged men who underwent submaximal exercise ECG during the Lipid Research Clinics Coronary Primary Prevention Trial, 62 (2%) subsequently experienced an acute cardiac event during moderate or strenuous physical activity during follow-up (average 7.4 years). 78 Although men with exercise-induced ECG changes were at increased risk, only 11 of 62 events occurred in men with an abnormal baseline exercise test (sensitivity 18%). Moreover, few of the men with abnormal test results experienced an activity-related event during follow-up (positive predictive value 4%). Although the negative predictive value of baseline ECG was high (over 98%), it was no better than multivariate analysis based on clinical risk factors alone. Given the low incidence of activity-related events in middle-aged men, and the uncertain benefit of restricting activity in those with abnormal exercise tests, the potential benefits of pre-exercise testing appear small. In populations at low risk for heart disease, any benefits of detecting the rare individual with asymptomatic CAD may be offset by adverse effects of labeling and exercise restrictions for the larger number of persons with false-positive ECG results.Recommendations of Other Groups
The routine use of resting electrocardiogram to screen for CAD in asymptomatic adults is not recommended by the American College of Physicians (ACP) 79 or the Canadian Task Force on the Periodic Health Examination. 80 The American Academy of Family Physicians (AAFP) recommends a baseline electrocardiogram for men 40 years and older with two or more cardiac risk factors and sedentary men about to begin a vigorous exercise program this recommendation is under review. 81 A task force sponsored by the American College of Cardiology and the American Heart Association (ACC/AHA) recommends baseline testing for all persons over 40 years of age and for those about to have exercise stress testing. 82
The AAFP recommends exercise electrocardiography for those whose jobs are linked to public safety (e.g., pilots, air traffic controllers) or that require high cardiovascular performance (e.g., police officers, firefighters). 81 The American College of Sports Medicine recommends exercise ECG testing for men over age 40, women over age 50, and other asymptomatic persons with multiple cardiac risk factors, prior to beginning a vigorous exercise program. 83 The ACC/AHA recognize that the exercise ECG is frequently used to screen asymptomatic persons in some high-risk groups but concluded that there is divergence of opinion with respect to its usefulness. 84 The ACP does not recommend exercise testing with ECG or thallium scintigraphy as a routine screening procedure in asymptomatic adults. 79,85Discussion
Heart disease is the leading cause of death in the U.S., and interventions that produce even modest reductions in the incidence of acute ischemic events may have substantial public health benefits. Although the resting electrocardiogram can detect evidence of coronary heart disease in asymptomatic persons and identify individuals at increased risk of future coronary events, the ECG has important weaknesses as a screening test. The large majority of asymptomatic persons with abnormal ECG results do not have CAD and are at relatively low risk for developing symptomatic heart disease in the near future. Routine screening may subject many of them to the inconvenience, expense and potential risks of follow-up testing (i.e., cardiac catheterization or radionuclide imaging) to evaluate false-positive screening results. Although exercise testing is more sensitive and specific for high-grade coronary stenoses, the exercise ECG is too time-consuming and expensive for routine use in asymptomatic persons. Finally, neither resting nor exercise ECG reliably detects the mild to moderate atherosclerotic lesions which are often responsible for acute coronary events.
A second important problem with screening for asymptomatic CAD is the lack of evidence that earlier detection leads to better outcomes. The only interventions proven to reduce coronary events in asymptomatic persons are modifications of risk factors such as smoking, high cholesterol, and elevated blood pressure. These interventions, however, should be encouraged for all patients with modifiable risk factors, not only those with screening tests suggestive of CAD. The benefits of more invasive treatments for coronary stenosis (e.g., bypass surgery, angioplasty) are unproven in asymptomatic persons. For certain occupations, such as pilots and heavy equipment operators, where sudden death or incapacitation would endanger the safety of others, considerations other than benefit to the individual patient may favor screening. Although screening cannot reliably identify all persons at risk of an acute event, it may increase the margin of safety for the public.
To minimize the potential adverse effects of false-positive test results, routine screening with ECG should be avoided in populations where the prevalence of CAD is low, including most adults under 40, and middle-aged men and women without coronary risk factors. Even in high-risk individuals, the benefits of screening to identify asymptomatic CAD are unproven. For some persons, however, identifying those at high risk of coronary mortality may help guide treatment decisions (e.g., use of aspirin or cholesterol-lowering drugs).
There are major costs associated with widespread screening with resting ECG in asymptomatic adults, and use of other screening tests (ambulatory ECG, exercise testing, and echocardiography) would be substantially more expensive. 79 The inconvenience, expense, and potential risks of routine screening might be justified if it significantly reduced the incidence of MI and sudden cardiac death, but such evidence is not yet available. Until appropriate studies demonstrate a benefit of screening for CAD, identification and treatment of major cardiac risk factors such as hypertension, elevated serum cholesterol, and cigarette smoking remain the only proven measures for reducing coronary morbidity and mortality in asymptomatic persons.
CLINICAL INTERVENTION2. Screening for High Blood Cholesterol and Other Lipid Abnormalities
Burden of Suffering
Elevated blood cholesterol is one of the major modifiable risk factors for coronary heart disease (CHD), 1 the leading cause of death in the U.S. CHD accounts for approximately 490,000 deaths each year, 2 and angina and nonfatal myocardial infarction (MI) are a source of substantial morbidity. CHD is projected to cost over $60 billion in 1995 in the U.S. in medical expenses and lost productivity. 3 The incidence of CHD is low in men under age 35 and in premenopausal women (1-2/1,000 annually), 4 but climbs exponentially during middle age for both men and women. The onset of CHD is delayed approximately 10 years in women compared with men, probably due to effects of estrogen, 5 but women account for 49% of all CHD deaths in the U.S. 2 Clinical events are the result of a multifactorial process that begins years before the onset of symptoms. Autopsy studies detected early lesions of atherosclerosis in many adolescents and young adults. 6-10 The onset of atherosclerosis and symptomatic CHD is earlier among persons with inherited lipid disorders such as familial hypercholesterolemia (FH) 11 and familial combined hyperlipidemia (FCH). 12Serum Cholesterol and Risk of Coronary Heart Disease.
Epidemiologic, patho-logic, animal, genetic, and clinical studies support a causal relationship between blood lipids (usually measured as serum levels) and coronary atherosclerosis. 1,13-15 Extended follow-up of large cohorts (predominantly middle-aged men) 16-18 provides evidence that CHD risk increases in a continuous and graded fashion, beginning with cholesterol levels as low as 150-180 mg/dL [a] this association extends to cholesterol levels measured as early as age 20 in men. 14,19 During middle age, for each 1% increase in total cholesterol, CHD risk increases by an estimated 3%. 20 High cholesterol (>=240 mg/dL) is also a risk factor in middle-aged women, but most coronary events in women occur well after menopause. 5,17,21-24 Some studies report that cholesterol alone is a weak predictor of CHD mortality in the elderly, 24a,190 but an overview of 24 cohort studies indicates that high cholesterol remains a risk factor for CHD after age 65, 23 with the strongest associations among healthier elderly populations followed over longer periods. 25-27 The association is weaker in older women than in men 23 and is not consistent for cholesterol levels measured after age 75. 28-31
Expert panels have defined high and "borderline high" (200-239 mg/dL) cholesterol to simplify clinical decisions. 1 Because CHD is a multifactorial process, however, there is no definition of high cholesterol that discriminates well between individuals who will or will not develop CHD. 32,33 Due to nonlipid risk factors, persons with cholesterol below 240 mg/dL account for the majority of all CHD events. 34,35 Among middle-aged men, 9-12% of those with cholesterol 240 mg/dL or greater will develop symptomatic CHD over the next 7-9 years, 34,36 but most of them have multiple other risk factors for CHD. 35 The excess (i.e., absolute) risk due to high cholesterol (and the probable benefit of lowering cholesterol) increases with the underlying risk of CHD. In a 12-year study of over 316,000 men aged 35-57, the excess CHD mortality attributable to high cholesterol was greatest in men over age 45, and in those who smoked or had hypertension. 16 The increase in CHD mortality associated with a given increment in serum cholesterol was steepest at very high values (>300 mg/dL). 16 Excess risk from high cholesterol is smaller in women, who have less than half the CHD risk as do men at any given cholesterol level. 17,23,37 Although the relative risk associated with high serum cholesterol declines with age, 17,23,28 the excess risk generally does not, due to the much higher incidence of CHD in older persons. 31,38,39
[a] To convert values for serum total cholesterol, HDL-C, and LDL-C to mmol/L, multiply by 0.02586. Equivalent values for commonly used thresholds are 280 mg/dL = 7.2 mmol/L, 240 mg/dL = 6.2 mmol/L, 200 mg/dL = 5.2 mmol/L.Other Lipid Constituents and Risk of Coronary Disease.
The risk associated with high total cholesterol is primarily due to high levels of low-density lipoprotein cholesterol (LDL-C), 1 but there is a strong, independent, and inverse association between high-density lipoprotein cholesterol (HDL-C) levels and CHD risk. 40-42 Low HDL-C increases risk even when cholesterol is below 200 mg/dL, 41 a pattern present in up to 20% of men with confirmed CHD. 43 In many studies, measures of HDL-C or the ratio of total cholesterol to HDL-C are better predictors of CHD risk than is serum cholesterol alone. 5,22,23,24a,41,44 High total cholesterol in association with high HDL-C (>=60 mg/dL) is common in older women (especially those taking estrogen) but is not associated with an increased risk for CHD. 1,41 The importance of triglycerides as an independent risk factor for CHD remains uncertain. 40,45 Three large studies reported strong associations between triglyceride levels over 200-300 mg/dL (2.26-3.39 mmol/L) and cardiovascular mortality in women, 21,22,24 but other analyses found no association after controlling for obesity, fasting glucose, or low HDL-C. 46 The combination of high triglycerides and low HDL-C often occurs in association with other CHD risk factors such as hypertension and diabetes and is associated with a high risk of CHD. 46aPrevalence of High Cholesterol and Low HDL-C.
Serum total cholesterol and LDL-C increase 1-2 mg/dL per year in men from ages 20-40, 2 mg/dL per year in women from ages 40-60, 47 and an average 18% during the perimenopausal period, due in part to age-related increases in weight. 48 The prevalence of serum cholesterol 240 mg/dL or higher increases from 8-9% in adults under age 35 to nearly 25% for men age 55 and nearly 40% for women over 65. 49 Approximately 11% of men and 3% of women over age 20 have low HDL-C (<35 mg/dL) with desirable or borderline-high total cholesterol. 49Accuracy of Screening Tests
Both total cholesterol and HDL-C can be measured in venipuncture or finger-stick specimens from fasting or nonfasting individuals. Due to normal physiologic variation and measurement error, a single measurement may not reflect the patient's true (or average) cholesterol level. Stress, minor illness, posture, and seasonal fluctuations may cause serum cholesterol to vary 4-11% within an individual. 50 Laboratory assays are subject to random errors, due to variation in sample collection, handling, and reagents, and to systematic errors (bias), due to methods that consistently overestimate or underestimate cholesterol values. 51 In a survey of 5,000 clinical laboratories, 93% of the measurements were within 9% of a reference standard. 52 Desktop analyzers can produce reliable results, but some devices may not meet standards for accuracy. 53 Variation in training and operating technique can introduce additional error when instruments are used outside clinical laboratories. 54 Average bias for measurements based on capillary specimens compared to venous specimens was +4-7%. 55
As a result of these considerations, a single measure of serum cholesterol could vary as much as 14% from an individual's average value under acceptable laboratory conditions. 50 For an individual with a "true" cholesterol level of 200 mg/dL, the 95% range of expected values is 172-228 mg/dL. 56 Some authorities therefore recommend advising patients of their "cholesterol range," rather than a single value. 56 Where more precise estimates are necessary, an average of at least two measurements on two occasions has been recommended, and a third if the first two values differ by more than 16%. 50Screening Children by Family History.
Although cholesterol levels in child-hood correlate moderately well with levels in adulthood (correlation coefficient 0.4-0.6), many children with elevated serum cholesterol (defined as serum cholesterol >=200 mg/dL or LDL-C >=130 mg/dL, the 90-95th percentile in U.S. children under 19 years) 57 do not have high cholesterol as adults. 58-60 Furthermore, the association between childhood cholesterol levels and CHD in adults has not been studied. Because of the familial aggregation of CHD and hypercholesterolemia, 57,61,62 some experts recommend screening for family history of either premature cardiovascular disease (age 55 or younger) or parental hypercholesterolemia (>=240 mg/dL) to identify a subset of children who are more likely to be at risk from hypercholesterolemia as adults. 57 Under this definition, only 25% of all children would be screened, but the predictive value of family history is limited: 81-90% of children with such histories have normal cholesterol. 63-66 Even when parental cholesterol has been measured and found to be elevated, most children have normal cholesterol values. 57,67,68
Parental and childhood cholesterol levels are highest in heterozygous FH (estimated prevalence 1 in 500), which is strongly associated with premature CHD. Up to 50% of men with FH develop clinical CHD by age 50. 69,70 Screening based on family history, as defined above, does not appear to be an efficient strategy for detecting FH, however. Many children would be screened, and few of those identified and treated for high cholesterol would have FH. 71 By itself, a parental history of premature CHD is likely to detect less than half of all children with FH. 70 Tracing and screening families of index cases with FH may be more cost-effective than population screening for FH. 72Screening for Other Lipid Abnormalities.
Measurements of HDL-C and triglyc-erides are less reliable than measurement of total cholesterol due to greater biologic and analytic variability. 73,74 The 95% range of expected values for an individual with HDL-C of 37 mg/dL is 29-45 mg/dL. 75 A survey of 250 laboratories found that one third of all HDL-C measurements varied more than 10% from a reference value. 76 Triglycerides must be measured on fasting specimens. Even then, intraindividual variation is greater than 20%, and a single measure is inadequate to categorize levels as high or normal. 73,74 Measurement of apolipoproteins (e.g., apoB) has been evaluated as a screening test for FH, familial coronary disease, and high LDL-C, but these assays are not yet widely available or adequately standardized. 57Effectiveness of Early Detection
No long-term study has compared routine cholesterol screening to alternate strategies (selective case-finding or universal dietary advice without screening) with change in cholesterol levels or CHD incidence as an outcome. The increase in cholesterol screening over the past decade in the U.S. has been accompanied by significant improvements in dietary knowledge, 77 fat consumption, 78 average cholesterol levels, 79 and CHD mortality, 14 but it is difficult to isolate the contribution of screening from other factors (e.g., public education, changes in food supply) that may account for these trends. In community- or practice-based trials, patients receiving risk-factor screening and targeted dietary advice had slightly lower average cholesterol levels (1-3%) than did unscreened controls at 1-3-year follow-up, but dietary interventions were limited. 80-82 Whether screening improves the effectiveness of routine dietary advice has been examined in two short-term studies where all subjects received counseling about diet cholesterol screening modestly improved mean cholesterol levels in one study but had no effect in the other. 83,84 In a school-based study in which all children received similar health education, cardiovascular risk-factor screening (including cholesterol measurement) was associated with improved dietary knowledge and self-reported behavior, but changes in lipid levels were not assessed. 85
The primary evidence to support cholesterol screening is the ability of cholesterol-lowering interventions to reduce the risk of CHD in patients with high cholesterol. These benefits are now well established for persons with preexisting atherosclerotic vascular disease. In individual trials and overviews of studies enrolling persons with angina or prior myocardial infarction (MI), cholesterol-lowering treatments slowed the progression of atherosclerosis, 86 reduced the incidence of CHD, 87,88 and reduced overall mortality. 89 In the first long-term trial of newer cholesterol-lowering drugs, treatment with simvastatin over 5.4 years reduced coronary mortality 42% and all-cause mortality 30% in 4,444 men and women with coronary disease. 90
The absolute benefit of treating high cholesterol in persons without cardiovascular disease, however, is much smaller due to the much lower risk of death or MI (annual CHD mortality 0.1-0.3% in middle-aged men with asymptomatic high cholesterol vs. 2-10% per year in patients with symptomatic CHD). 91 The risks and benefits of lowering cholesterol in asymptomatic persons -- primarily middle-aged men with very high cholesterol -- have been examined in trials using medications, modified diets in institutional patients, or outpatient dietary counseling, and in overviews of these trials.Trials of Cholesterol-Lowering Drugs in Asymptomatic Men.
Three large, multicenter, placebo-controlled trials of lipid-lowering medications provide the best evidence that lowering cholesterol can reduce combined CHD incidence (fatal and nonfatal events) in asymptomatic persons. These trials enrolled hypercholesterolemic middle-aged men (age 30-59, mean cholesterol 246-289 mg/dL) and lowered total cholesterol 9-10% (and LDL-C 10-13%) over periods of 5-7 years. In the World Health Organization Cooperative Trial, treatment with clofibrate significantly reduced the incidence of nonfatal MI by 25%, 92 but this benefit was offset by significant increases in noncardiac and total mortality (40% and 30% respectively, p = 0.01). 93 The Lipid Research Clinics (LRC) Coronary Primary Prevention Trial reported a significant 19% reduction in cumulative incidence of MI and sudden cardiac death in patients treated with cholestyramine over 7 years (7.0% vs. 8.6%). 36 In the Helsinki Heart Study, treatment with gemfibrozil significantly reduced the 5-year cumulative incidence of cardiac events by 34% (2.7% vs. 4.1%). 94 Most of the benefit of gemfibrozil was confined to men with a high ratio of LDL-C to HDL-C (>=5) and triglycerides >200 mg/dL. 95 Effects on CHD mortality were not statistically significant in any of these trials. Two additional drug trials reported 1-3-year results in largely asymptomatic populations. 96,97 Roughly 30% of subjects had CHD at entry, however, and these patients accounted for most of the coronary events during follow-up.Trials of Diet in Institutionalized Persons.
Demonstrating a clinical benefit of modern cholesterol-lowering diets in asymptomatic persons has proven difficult. In three controlled trials in institutionalized patients, fat-modified diets reduced serum cholesterol 12-14% with generally favorable effects on CHD over periods of up to 8 years. 98-100 Each of these studies used diets high in polyunsaturated fat, which have been associated with adverse effects, 15 and none excluded patients with CHD. As a result, their findings may not be applicable to currently recommended low-fat diets in asymptomatic persons.Trials of Dietary Advice in Outpatients.
The only trials to examine the clinical benefits of a diet low in total and saturated fat in persons without CHD are multifactorial intervention trials, which offered dietary counseling, smoking cessation advice, and/or treatment of high blood pressure to middle-aged men. 101-105 Among Norwegian smokers with very high cholesterol levels (mean 320 mg/dL) and fat consumption (44% calories), dietary advice lowered cholesterol 13% and, in conjunction with smoking cessation, reduced CHD incidence by 47%. 103 The remaining trials achieved much smaller (0-5%) reductions in cholesterol and insignificant effects on CHD the benefits of intervention in some studies may have been limited by ineffective counseling and follow-up, 101,104 lower cholesterol levels at baseline, 101 or adverse effects of other therapies. 102,105 In the most systematic test of dietary counseling in adults, 10 weekly group sessions and periodic individual counseling were provided over 6 years to over 6,000 men (mean cholesterol 253 mg/dL). 102 Average cholesterol level declined 5% in men receiving counseling, but only 2% compared to controls. Greater changes were observed in men who lost at least 5 pounds and those with higher serum cholesterol at baseline, 106 but there was no significant reduction in CHD mortality or incidence in the intervention group. 102,107
Short-term metabolic studies and selected trials in patients with CHD indicate that reducing dietary saturated fat and/or increasing polyunsaturated fat intake can reduce elevated total and LDL-C as much as 10-20%. 108-110 Due to variable compliance, trials of diet counseling in the primary care setting have achieved much smaller and inconsistent average reductions in serum cholesterol in asymptomatic persons (0-4%). 80-82,111-116 Although larger changes have been reported in uncontrolled follow-up studies after cholesterol screening, 117,118 these results may be biased by selective or short-term follow-up and regression to the mean in persons with high cholesterol. Ongoing studies are examining the efficacy of cholesterol screening and intervention in primary care settings in the U.S. 119 More stringent diets can produce larger reductions in cholesterol, 120 but long-term data in asymptomatic persons are limited. Two trials in women at risk for breast cancer lowered total fat intake to 20% of calories and reduced total cholesterol 6-7% over 1-2 years. 121,122Overviews of Cholesterol-Lowering Trials.
At least 10 quantitative overviews (meta-analyses) of randomized trials have attempted to resolve uncertainties about the risks and benefits of lowering cholesterol, including effects on mortality. 18,88,89,91,123-128 Three recent overviews provide the most comprehensive analyses of long-term cholesterol-lowering trials published through 1993 35 diet and drug trials were included in one analysis, 91 28 in the second (excluding trials that used estrogens or thyroxine), 18,89 and 22 in the third (all trials achieving at least a 4% reduction in cholesterol for at least 3 years). 128 These overviews support a dose-response relationship between change in serum cholesterol and reduction in CHD incidence (fatal and nonfatal events combined) comparable to that predicted from epidemiologic studies: after 2-5 years of treatment, each 1% reduction in serum cholesterol yields a 2-3% reduction in total CHD, for both diet and drug interventions, and in patients with or without CHD at entry. 18,89,128
When only trials enrolling asymptomatic persons were analyzed, however, neither CHD mortality nor total mortality was significantly reduced by cholesterol lowering: 128 difference in total mortality among treated versus control subjects = +6%, 95% confidence interval (CI) -3% to +17%. 89 Moreover, noncardiac mortality was increased 20-24% among patients treated with lipid-lowering medications. 89,91,128 While observing similar effects of treatment, each overview offered distinct interpretations of these findings. Law et al. concluded that the increase in noncoronary mortality was most likely due to chance: the finding was of borderline statistical significance (p = 0.02), did not reflect any consistent cause of excess mortality among trials, and was independent of compliance with therapy. 89 Gordon attributed adverse effects to trials employing hormones or fibrate medications. 128 Davey Smith et al. concluded that lipid-lowering drugs reduced overall mortality in high-risk persons (i.e., persons with CHD), but were harmful in those at lower risk. 91 When trials were stratified by the observed CHD mortality in the control group, drug treatment was associated with a significant 20% increase in all-cause mortality in 10 trials enrolling low-risk subjects (CHD mortality <1% per year), including the WHO, LRC, and Helsinki studies. 91 A single trial (the WHO clofibrate study) accounts for nearly half of all patient-years of treatment in persons without CHD 128 and has a strong influence on results of any meta-analysis.
Due to methodologic concerns about combining results from trials employing different cholesterol-lowering drugs and diets, meta-analysis cannot prove or disprove possible harms from lipid-lowering medications. 129 These analyses, however, illustrate the importance of underlying CHD risk in determining whether expected benefits are likely to justify possible risks of treatment. Even if drugs are safe, the margin of benefit may be small for many persons with asymptomatic hypercholesterolemia. In the LRC and Helsinki trials, preventing one coronary death required treating 300-400 middle-aged men for 5-7 years. 36,94 The benefits of lipid-lowering medications on nonfatal CHD are more pronounced but must be weighed against the unpleasant and occasionally serious side effects of some drugs (see below). 92,127,130 The newest class of lipid-lowering drugs, HMG-CoA reductase inhibitors or "statins," lowers cholesterol more effectively and appears to be well-tolerated in trials lasting up to 6 years. 90,97 These drugs are more likely to have significant effects on mortality in patients without CHD, but long-term trials of these agents in asymptomatic persons have yet to report results. 131Cholesterol Reduction in Women.
Lipid-lowering medications and diet effectively lower cholesterol in women, 132 but no trial has specifically examined the benefits of cholesterol reduction in asymptomatic women. 133 Trials that included female subjects with CHD observed qualitatively similar benefits of cholesterol reduction on angiographic or clinical endpoints in women and men. 90,99,133 In the 4S trial, simvastatin significantly reduced CHD incidence, but not mortality, in women with CHD. 90 Two trials in women without CHD, with a cumulative enrollment of more than 6,000 women, observed no effect of drug or diet treatment on CHD incidence or mortality after 1-3 years 96,100 the short duration of follow-up may have limited the power of these studies to detect a difference. 133 Long-term data on drug therapy in women are limited, with the exception of estrogen therapy (see Chapter 68).Cholesterol Reduction in Older Adults.
The benefit of lowering cholesterol in older persons has been questioned due to the weak association between serum cholesterol and all-cause mortality after age 60. 17,28,30 Associations between cholesterol and mortality in unselected elderly populations, however, are likely to be confounded by the increasing prevalence of chronic illnesses which increase mortality and independently lower serum cholesterol. 26,134,135 Direct evidence that cholesterol reduction is beneficial in asymptomatic older persons is not yet available, but cholesterol-lowering diets and medications reduced overall mortality 26-30% in persons over 60 with clinical CHD. 90,136,137 In two trials in patients without CHD that included older subjects, however, cholesterol reduction produced significant benefits in younger but not in older patients (over age 60 or 65). 96,98 Newer cholesterol-lowering agents are efficacious and well-tolerated in older patients. 90,138 A large multicenter trial is under way to examine the effectiveness of pravastatin and various antihypertensive medications in asymptomatic persons over age 60 with hypertension and high cholesterol. 138 There are few controlled trials of dietary counseling to lower cholesterol in older patients no significant change in cholesterol levels was observed among rural Medicare recipients offered diet counseling 139 or older patients receiving diet counseling and placebo medication. 138Cholesterol Reduction in Adolescents and Young Adults.
Determining the benefits of lowering cholesterol in children, adolescents, and young adults is difficult, due to their low near-term risk of clinical coronary disease. The assumption that early treatment is more effective than treatment begun later in life 57 rests on observations that early atherosclerosis is present in many adolescents and young adults, is associated with lipid levels, progresses with age, 6 and is difficult to reverse in middle age. 86 New evidence, however, suggests that much of the clinical benefit of lowering cholesterol can be achieved within 2-5 years of initiating therapy. 18 These benefits have been attributed to stabilizing "lipid-rich" lesions 87 and improving endothelial function, 140 and they suggest that the additional benefits of early drug therapy for hypercholesterolemia (i.e., before middle age) may not justify the added expense and possible risks of longer treatment. Intensive diet or drug intervention for adolescents and young adults with FH, although never tested in a prospective trial, has become standard treatment due to the very high levels of LDL-C and dramatically increased risk of premature CHD in persons with FH. 11,71 Even in FH, however, most clinical events occur in middle age (i.e., after age 40), and risk is variable: MI was rare before age 30 in men in one study, and onset of CHD is later in women and nonsmokers with FH. 69,141
Modified diets lower cholesterol in young adults, but the contribution of universal screening in motivating risk reduction in young persons is uncertain. Neither a multiple-intervention trial in Australian workers 142 nor a study of risk assessment in a general practice in the U.K. 116 demonstrated that screening and dietary advice led to long-term reduction in cholesterol levels in younger men (under age 35-40). The effectiveness of screening and dietary counseling has not been adequately studied in young adults and cannot be predicted reliably from studies in middle-aged men.Cholesterol Reduction in Children.
Dietary fat intake in children is associated with total cholesterol and LDL-C levels, 143,144 but controlled trials have not consistently demonstrated that individual dietary counseling is effective in children. 145-147 Results from the largest trial reported that children with elevated LDL-C who received intensive family-oriented dietary counseling (30 sessions over 3 years) experienced a significant but modest (3.2 mg/dL) reduction in mean LDL-C compared with controls. 148 Uncontrolled studies of dietician counseling for hyperlipidemic children and adolescents have reported larger short-term reductions in mean cholesterol and LDL-C, 149-153 but such studies are prone to bias from regression to the mean and selective follow-up. Physical activity and fitness are associated with higher levels of HDL-C in children and adolescents, but controlled and uncontrolled trials 154-159 have reported inconsistent effects of exercise interventions on lipids. Drug therapy effectively lowers cholesterol in children, but side effects limit compliance with bile-acid resins, the only therapy currently recommended for routine use in children. 57 In one study of 80 children with FH or FCH, only 13% were still compliant with resin therapy after 3 years. 160 Ongoing studies are examining the safety and efficacy of newer agents in children.Potential Adverse Effects of Screening and Intervention.
Measurement of serum cholesterol is safe and relatively inexpensive, but widespread screening may have some undesirable consequences. In populations in which the potential benefits of early detection may be small (e.g., low-risk young persons), the possibility of harm may influence decisions about universal screening. 161 Anecdotal reports have reported decreased well-being in persons diagnosed with high cholesterol (i.e., "labeling"), 162 but a prospective study did not confirm this effect. 163 Other possible adverse effects of screening include inconvenience and expense of screening and follow-up, opportunity costs to the busy clinician, misinformation due to inaccurate results, and reduced attention to diet in persons with "desirable" cholesterol levels. 164 The importance of possible adverse effects of screening has not been systematically studied.
The safety of cholesterol-lowering interventions is especially important in children and young persons. Dietary restrictions may reduce intake of calories, calcium, vitamins, and iron in children, 165-167 and failure-to-thrive due to excessively fat-restricted diets has been reported, albeit rarely, in children. 168,169 In the most comprehensive trial of dietary intervention in children, however, no adverse effects on growth, sexual development, psychological measures, iron status, or blood micronutrients were detected at 3-year follow-up. 148 Other controlled studies also support the safety of properly performed dietary intervention in children. 147,166,170 The elderly may also be at risk from modified diets if adequate intake of calories, calcium, and essential vitamins is not maintained, but these effects have not been directly examined.
The inappropriate use of drug therapy is of greater concern, especially in young persons in whom the benefit of early drug treatment may not justify the costs and possible risks. 18,161 According to a national survey of pediatricians and family physicians, one in six regularly prescribed drugs for hypercholesterolemic children, and a substantial number did so based on inappropriate criteria, or used drugs not routinely recommended for children. 171 Persons under age 40 accounted for over 1 million prescriptions for lipid-lowering drugs in 1992 172 gemfibrozil was the second most commonly prescribed lipid-lowering drug in the U.S. in 1992, 172 despite limited indications for its use 1 and important safety concerns. Fibrate medications (e.g., clofibrate and gemfibrozil) have been associated with an increase in gallstone disease, 92 adverse trends in CHD mortality 93,173 and cancer mortality in individual trials, 93,174 and a significant increase in noncoronary mortality in a recent overview of long-term trials. 128 HMG-CoA reductase inhibitors have not been associated with important adverse effects in trials lasting up to 6 years. 90 The safety of lifelong therapy with these agents cannot yet be determined several medications in this class have been reported to cause liver tumors in animal studies.Early Detection of Other Lipid Abnormalities.
The importance of detecting low HDL-C or high triglycerides remains unproven, especially in persons with normal serum cholesterol. Weight loss in obese subjects, 132,175 smoking cessation, exercise, 176,176a and moderate alcohol consumption 177 can raise HDL-C and/or lower triglyceride levels. Some of these lifestyle interventions have only small effects, however, and most can be recommended independent of lipid levels. Most importantly, no trial has directly examined the benefit of raising HDL-C or lowering triglycerides. 40,45 Secondary analyses of several trials have attributed varying proportions of the clinical benefit of drug therapy to increases in HDL-C, 40,94,95 or reductions in triglycerides, 136 but all of the subjects had high total or LDL cholesterol. The benefit of drug treatment for low HDL-C and normal cholesterol has not been determined but is being studied in men with CHD. 43Recommendations of Other Groups
The National Cholesterol Education Program Adult Treatment Panel II, convened by the National Heart, Lung, and Blood Institute, recommends routine measurement of nonfasting total cholesterol and HDL-C in all adults age 20 or older at least once every 5 years. 1,178 The Canadian Task Force on the Periodic Health Examination concluded there was insufficient evidence to recommend routine cholesterol screening but endorsed case-finding in men 30-59 years old. 127 The American Academy of Family Physicians 179 recommends measurement of total cholesterol at least every 5 years in adults age 19 and older these recommendations are under review. The American College of Obstetricians and Gynecologists recommends periodic screening of cholesterol in all women over age 20, and in selected high-risk adolescents. 180 In guidelines revised in 1995, the American College of Physicians (ACP) concluded that screening serum cholesterol was appropriate but not mandatory for asymptomatic men aged 35-65 and women aged 45-65 screening is not recommended for younger persons unless they are suspected of having a familial lipoprotein disorder or have multiple cardiac risk factors. The ACP concluded that evidence was not sufficient to recommend for or against screening asymptomatic persons between the ages of 65 and 75, but it recommends against screening after age 75. 181
Selective screening of children and adolescents is recommended by the National Cholesterol Education Program Expert Panel on Blood Cholesterol Levels in Children and Adolescents, 57 the American Academy of Pediatrics (AAP), 182 the Bright Futures guidelines, 183 the American Medical Association Guidelines for Adolescent and Preventive Services (GAPS), 184 and the American Academy of Family Physicians. 179 Screening with nonfasting cholesterol in all children and adolescents who have a parental history of hypercholesterolemia, and with fasting lipid profile in those with a family history of premature cardiovascular disease, is recommended. These organizations recommend that children who have multiple risk factors for CHD (such as smoking or obesity) and whose family history cannot be ascertained be screened at the discretion of the physician.Discussion
Elevated serum cholesterol is an important risk factor for CHD in men and women in the U.S., and there is now good evidence that lowering serum cholesterol can reduce the risk of CHD. Whereas measures that lower serum cholesterol and provide other health benefits (e.g., regular physical activity, reducing dietary fat, and maintaining a healthy weight) should be encouraged in all persons, cholesterol screening can identify high-risk individuals who are most likely to benefit from individualized dietary counseling or drug treatment. In addition, screening may help clinicians and patients identify priorities for risk factor modification and reinforce public awareness of the importance of a healthy diet.
Some important questions remain, however, about routine lipid screening in asymptomatic and low-risk persons, including when to begin screening and which constituents to measure. Overall, evidence is strongest for screening for high serum cholesterol in middle-aged men (ages 35-65), based on the reduction in coronary morbidity in trials enrolling asymptomatic men with very high cholesterol (mean 280 mg/dL). The epidemiology and pathophysiology of CHD is similar in men and women, suggesting that reducing high cholesterol levels will also reduce CHD in asymptomatic women. Extrapolations to premenopausal women may not be appropriate, given their low risk of CHD and the apparent protective effects of estrogen on CHD incidence. The optimal age to screen women is not known the later onset of hypercholesterolemia and CHD suggests that routine screening should begin around age 45.
Direct evidence that screening and intervention is effective in persons over age 65 is not yet available, but epidemiologic studies indicate that the risks of high cholesterol extend up to age 75. Given the high risk of CHD in the elderly, and the benefits of lowering cholesterol in symptomatic older men and women, screening may be reasonable in older persons who do not have major comorbid illnesses. Since individual cholesterol levels usually plateau by age 65 in women (and earlier in men), continued screening is less important in patients who have had desirable cholesterol levels throughout middle age.
There is not yet evidence that routine lipid screening is effective in reducing cholesterol levels or CHD risk in younger populations. Universal screening is an inefficient way to identify the small number of hypercholesterolemic young persons at risk for premature CHD, most of whom have multiple nonlipid risk factors or a history suggestive of familial dyslipidemia. Most "high-risk" young persons (excluding young men with FH) have a near-term risk of CHD well below that of hypercholesterolemic middle-aged men, 185 and are not appropriate candidates for early drug therapy. Screening young persons can provide information to help stimulate lifestyle changes, but promoting a healthy lifestyle (e.g., healthy diet, regular physical activity, etc.) is important for all young persons, including the majority with "desirable" cholesterol levels. International comparisons suggest that cholesterol levels explain only part of the strong association between diet and heart disease. 186a As a result, it is uncertain whether routine cholesterol screening in low-risk younger populations is of sufficient benefit to justify the inconvenience, costs, and possible risks of screening and treatment. In a study modeling benefits of cholesterol screening, a conservative strategy of screening only middle-aged men and others with multiple CHD risk factors produced benefits comparable to screening all adults over age 20 if interventions had adverse effects on quality of life, the more conservative strategy was preferable. 186 Should future studies demonstrate that routine screening and targeted interventions are more effective in the primary care setting than universal dietary advice in young persons, this would provide some additional justification for early screening.
The benefits of screening children are even less certain. Progression of atherosclerosis in childhood is limited, many children with high cholesterol are not hypercholesterolemic as adults, and it is uncertain whether or not reducing cholesterol levels in childhood will significantly alter the risk of CHD many years later. Given the limited effectiveness of dietary counseling, poor compliance with currently recommended drug therapy, and the potential for adverse reactions in children, widespread pediatric screening might result in more harm than good.
The benefit of measuring HDL-C or triglycerides at initial screening is unproven. Measures to lower high triglycerides or raise HDL-C (e.g., weight reduction in obese persons and exercise) have relatively modest effects and should be encouraged regardless of lipid levels. Measures of HDL-C and lipoprotein analysis improve the estimation of coronary risk and should be obtained to guide treatment decisions in patients with high total cholesterol. There is, however, no evidence that they significantly improve the management of patients who do not have high total cholesterol.
While a single cholesterol test is relatively inexpensive, the cumulative costs of screening can be substantial under protocols calling for measurement of HDL-C, periodic screening, and detailed evaluation and treatment of the large population with high cholesterol. To be effective, dietary interventions require regular follow-up and reinforcement. Under optimistic assumptions, tailored dietary therapy in middle-aged men is estimated to cost more than $20,000 per year of life gained, when costs of screening and follow-up are included. 187 Drug treatment of asymptomatic middle-aged men (assuming no important adverse effects) has been estimated to cost at least $50,000-90,000 per year of life saved. 35,188 HMG-CoA reductase inhibitors are substantially more expensive than earlier medications, but they lower LDL-C more effectively and also raise HDL-C. These agents may improve the cost-effectiveness of drug therapy for asymptomatic hypercholesterolemia, especially in high-risk men, 189 but the long-term safety and effectiveness of these agents in persons without CHD have not yet been established.
CLINICAL INTERVENTION3. Screening for Hypertension
Burden of Suffering
Hypertension is usually defined as a diastolic blood pressure of 90 mm Hg or higher or a systolic pressure of 140 mm Hg or higher. 1 It is present in an estimated 43 million Americans and is more common in blacks and older adults. 1a Hypertension is a leading risk factor for coronary heart disease, congestive heart failure, stroke, ruptured aortic aneurysm, renal disease, and retinopathy. These complications of hypertension are among the most common and serious diseases in the U.S., and successful efforts to lower blood pressure could thus have substantial impact on population morbidity and mortality. Heart disease is the leading cause of death in the U.S., accounting for nearly 740,000 deaths each year (287 deaths per 100,000 population), and cerebrovascular disease, the third leading cause of death, accounts for about 150,000 deaths each year (58/100,000). 2 Milder forms of hypertension predict progression to more severe elevations and development of cardiovascular disease. 1,3,4 Coronary heart disease mortality begins to increase at systolic blood pressures above 110 mm Hg and at diastolic pressures above 70 mm Hg. 5 The prevalence of unrecognized and uncontrolled hypertension, and the mortality from cardiovascular disease, have declined substantially in the U.S. in the past several decades. 1
Treatable (also known as secondary) causes of hypertension such as aortic coarctation or renovascular disease also may be associated with severe consequences, including congestive heart failure, aortic rupture, or stroke. 6-9 There are no population data available for estimating the true prevalence of secondary hypertension. The incidence of coarctation of the aorta has been estimated at 0.2-0.6/1,000 live births and the prevalence at 0.1-0.5/1,000 children. 10-12Accuracy of Screening Tests
The most accurate devices for measuring blood pressure (e.g., intra-arterial catheters) are not appropriate for routine screening because of their invasiveness, technical limitations, and cost. Office sphygmomanometry (the blood pressure cuff) remains the most appropriate screening test for hypertension in the asymptomatic population. Although this test is highly accurate when performed correctly, false-positive and false-negative results (i.e., recording a blood pressure that is not representative of the patient's average blood pressure) do occur in clinical practice. 13 One study found that 21% of persons diagnosed as mildly hypertensive based on office sphygmomanometry had no evidence of hypertension when 24-hour ambulatory recordings were obtained. 14
Errors in measuring blood pressure may result from instrument, observer, and/or patient factors. 15 Examples of instrument error include manometer dysfunction, pressure leaks, stethoscope defects, and cuffs of incorrect width or length for the patient's arm size. The observer can introduce errors due to sensory impairment (difficulty hearing Korotkoff sounds or reading the manometer), inattention, inconsistency in recording Korotkoff sounds (e.g., Phase IV vs. Phase V), and subconscious bias (e.g., "digit preference" for numbers ending with zero or preconceived notions of "normal" pressures). The patient can be the source of misleading readings due to posture and biologic factors. Posture (i.e., lying, standing, sitting) and arm position in relation to the heart can affect results by as much as 10 mm Hg. 15 Biologic factors include anxiety, meals, tobacco, alcohol, temperature changes, exertion, and pain. Due to these limitations in the test-retest reliability of blood pressure measurement, it is commonly recommended that hypertension be diagnosed only after more than one elevated reading is obtained on each of three separate visits over a period of one to several weeks. 1
Additional factors affect accuracy when performing sphygmomanometry on children these difficulties are especially common when testing infants and toddlers under 3 years of age. 16-18 First, there is increased variation in arm circumference, requiring greater care in the selection of cuff sizes. 19 Second, the examination is more frequently complicated by the anxiety and restlessness of the patient. Third, the disappearance of Korotkoff sounds (Phase V) is often difficult to hear in children and Phase IV values are often substituted. Fourth, erroneous Korotkoff sounds can be produced inadvertently by the pressure of the stethoscope diaphragm against the antecubital fossa. Finally, the definition of pediatric hypertension has itself been uncertain because of confusion over normal values during childhood. The definition of hypertension in childhood is essentially arbitrary, based on age-specific percentile. 18 Age-, sex-, and height-specific blood pressure nomograms for U.S. children and adolescents have been published more recently, based on data from 56,108 children aged 1-17 years. 20
Self-measured (home) blood pressure and ambulatory blood pressure monitoring may provide useful information in special circumstances (e.g., research, persistent "white-coat" hypertension), but there is insufficient evidence at present to warrant their routine use in screening. 1,21-28Effectiveness of Early Detection
There is a direct relationship between the magnitude of blood pressure elevation and the benefit of lowering pressure. In persons with malignant hypertension, the benefits of intervention are most dramatic treatment increases 5-year survival from near zero (data from historical controls) to 75%. 29 Over the past 30 years, the results of many randomized clinical trials of the effects of antihypertensive drug therapy on morbidity and mortality in adult patients (>=21 years of age) with less severe hypertension have been published. 30-32 The efficacy of treating hypertension is clear, as demonstrated in a number of older randomized controlled trials in adults with diastolic blood pressures ranging from 90 to 129 mm Hg. 33-38 For example, in the Veterans Administration Cooperative Study on Antihypertensive Agents, middle-aged men with diastolic blood pressure averaging 90 through 114 mm Hg experienced a significant reduction in "morbid" events (e.g., cerebrovascular hemorrhage, congestive heart failure) after treatment with antihypertensive medication. 34
Persons with mild (Stage 1) to moderate (Stage 2) 1 diastolic hypertension (90-109 mm Hg) also benefit from treatment. 30,39-41 This was confirmed in the Hypertension Detection and Follow-Up Program, a randomized controlled trial involving nearly 11,000 hypertensive men and women, of whom 40% were black. 39 The intervention group received standardized pharmacologic treatment ("stepped care") while the control group was referred for community medical care. There was a statistically significant 17% reduction in 5-year all-cause mortality in the group receiving standardized drug therapy the subset with diastolic blood pressure 90-104 mm Hg experienced a 20% reduction in mortality. 39 Deaths due to cerebrovascular disease, ischemic heart disease, and other causes were also significantly reduced in the stepped care group. 42 Similar effects on all-cause mortality and cardiovascular events have been reported in other randomized controlled trials, such as the Australian National Blood Pressure Study (initial diastolic blood pressure 95-109 mm Hg) 40 and the Medical Research Council (MRC) trial (diastolic blood pressure 90-109 mm Hg). 41 In these two trials, the relative reduction in rates of stroke or other trial endpoints with treatment was similar in those with diastolic blood pressures <95 or 95-99 mm Hg and those with higher diastolic blood pressures, although the absolute benefit was less due to smaller initial risk of stroke and other diseases at lower blood pressures. Both trials included untreated control groups and did not report a significant reduction in deaths from noncardiovascular causes in the actively treated groups, confirming that the benefit was due to antihypertensive treatment rather than to other medical care.
Earlier studies included some subjects over age 65 years, but in insufficient numbers to permit firm conclusions. Four large, randomized placebo-controlled trials have since demonstrated conclusively the benefit of antihypertensive treatment in elderly subjects (aged 60-97 years). 43-48 Three of these studies included persons with diastolic blood pressures of 90-120 mm Hg, and among them reported significant reductions in all-cause mortality, 46 cardiovascular mortality, 43,46 cardiovascular events, 47 and strokes. 46,47 The Systolic Hypertension in the Elderly Program (SHEP) trial included over 4,000 subjects >=60 years of age with isolated systolic hypertension (systolic blood pressure >= 160 mm Hg, with diastolic blood pressure < 90 mm Hg), and reported significant reductions in the incidence of stroke, myocardial infarction, and left ventricular failure. 48 A meta-analysis combining these and other trials that included persons aged >=60 years demonstrated that antihypertensive treatment in elderly persons significantly reduced mortality from all causes (-12%), stroke (-36%), and coronary heart disease (-25%), as well as stroke and coronary heart disease morbidity. 49 This meta-analysis suggested reduced benefit with increasing age, although differences were not statistically significant. A second meta-analysis of randomized controlled trials in persons over age 60 years concluded that absolute 5-year morbidity and mortality benefits derived from trials were greater for older than for younger subjects. 50 This meta-analysis calculated that 18 (95% CI, 14-25) elderly hypertensive subjects needed to be treated for 5 years to prevent one cardiovascular event.
Treatment of hypertension is associated with multiple benefits, including reduced coronary heart disease and vascular deaths, but meta-analyses suggest it produces the largest reductions in cerebrovascular morbidity and mortality. 30-32,49,50 Improved treatment of high blood pressure has been credited with a substantial portion of the greater than 50% reduction in age-adjusted stroke mortality that has been observed since 1972. 1,51,52
Although the efficacy of antihypertensive treatment for essential (also called primary) hypertension has been well established in clinical research, certain factors may influence the magnitude of benefit from hypertension screening achieved in actual practice. Compliance with drug therapy may be limited by the inconvenience, side effects, and cost of these agents. 53,54 Serious or life-threatening drug reactions in the clinical trials were rare, but less serious side effects were common, resulting in discontinuation of randomized treatment (almost 20% by the fifth year in the MRC trial, 41 for example) or a substantial increase in patient discomfort. 34 Higher incidences of mild hypokalemia, hyperuricemia, and elevated fasting blood sugar have also been reported in treated individuals. 35 A population-based case-control study suggested an increased risk of primary cardiac arrest with certain diuretic regimens (e.g., higher doses, use without potassium-sparing therapy). 55 However, current drug regimens, including low-dose diuretics, are associated with fewer adverse effects and with favorable effects on quality of life. 55a Newer classes of drugs (e.g., calcium channel blockers, angiotensin-converting enzyme inhibitors) have not been assessed in long-term trials with clinical endpoints. Their effects on cardiovascular morbidity and mortality may differ from the effects reported in the clinical trials cited above, which used diuretics or beta-blockers.
Whether hypertension screening is equally effective for other populations or with treatments other than drugs is less clear. The benefits of hypertension treatment are less well studied in certain population groups, such as children (see below), Native Americans, Asians and Pacific Islanders, and Hispanics. The effects of nonpharmacologic first-line therapy (i.e., weight reduction in overweight patients, increased physical activity, sodium restriction, and decreased alcohol intake) on cardiovascular morbidity and mortality are unstudied. Although these nonpharmacologic therapies can sometimes lower blood pressure in the short term, 1,56-62a the magnitude of blood pressure reduction achieved is generally smaller than that achieved with drug therapy, and both the magnitude and duration of reduction in actual practice may be limited by biologic factors (e.g., varying responsiveness to sodium restriction) and the difficulties of maintaining behavioral changes (e.g., weight loss). Some of these interventions, such as sodium restriction, may also have adverse effects on quality of life. 63
The detection of high blood pressure during childhood is of potential value in identifying those children who are at increased risk of primary hypertension as adults and who might benefit from earlier intervention and follow-up. Hypertensive vascular and end-organ damage may begin in childhood, 64-69 although it is unclear how strongly these pathophysiologic changes are associated with subsequent cardiovascular disease. Prospective cohort studies have shown that children with high blood pressure are more likely than other children to have hypertension as adults. 70-78 Correlation coefficients from these studies were generally less than 0.5, however, suggesting a limited role for high blood pressure in childhood as a predictor of adult hypertension. Although controlled trials in children show that short-term (up to 3 years) effects on blood pressure can be achieved with changes in diet and activity, 79-82 studies demonstrating long-term changes in blood pressure are lacking. There are no trials showing that lowering blood pressure in childhood results in reduced blood pressure in adulthood. A relationship between lowering blood pressure during childhood and improved morbidity and mortality in later life is unlikely to be demonstrated, given the difficulty of performing such studies.
A relatively high proportion of children with hypertension have secondary, potentially curable forms. Among children and adolescents whose hypertension was evaluated in primary care centers, an estimated 28% had secondary hypertension (e.g., renal parenchymal disease, coarctation of the aorta). 69 This contrasts with hypertensive adults seen in primary care settings, of whom only 7% are estimated to have secondary hypertension. 83 Screening children and adolescents may be justifiable if the morbidity of these conditions is improved by early detection and treatment. Many causes of secondary hypertension in childhood are detectable by careful history-taking (e.g., preterm birth, umbilical artery catheter, chronic pyelonephritis, renal disease, bronchopulmonary dysplasia symptoms of cardiac, renal, endocrinologic, or neurologic disease) or physical examination (e.g., murmur, decreased femoral pulses, abdominal bruit). 69,84 Characteristic symptoms and signs, such as those of aortic coarctation, are often overlooked, however. 85-87 Numerous surgical case series suggest that delay in surgical repair of aortic coarctation increases the likelihood of irreversible hypertension, 88-94 although none of these series controlled for other differences between persons presenting early versus late in life. Uncontrolled studies indicate that some important causes of hypertension for which definitive cures are available, including coarctation and renovascular disease, may not be diagnosed until complications such as congestive heart failure, aortic rupture, or stroke occur. 6-9 Prognosis with early surgical intervention is improved compared with historical controls. 88,95Recommendations of Other Groups
Recommendations for adults have been issued by the Joint National Committee on Detection, Evaluation, and Treatment of High Blood Pressure, 1 and similar recommendations have been issued by the American Heart Association. 96 These call for routine blood pressure measurement at least once every 2 years for adults with a diastolic blood pressure below 85 mm Hg and a systolic pressure below 130 mm Hg. Measurements are recommended annually for persons with a diastolic blood pressure of 85-89 mm Hg or systolic blood pressure of 130-139 mm Hg. Persons with higher blood pressures require more frequent measurements. The American College of Physicians (ACP) 97 and the American Academy of Family Physicians (AAFP) 98 recommend that all adults 18 years and older be screened for hypertension every 1-2 years. The AAFP policy is currently under review. The ACP also recommends screening at every physician visit for other reasons, and that those in high-risk groups (e.g., diastolic 85-89 mm Hg, previous history of hypertension) be screened on an annual basis. The Canadian Task Force on the Periodic Health Examination recommends that all persons aged 21 years and over receive a blood pressure measurement during any visit to a physician ("case finding"). 99
The American Academy of Pediatrics (AAP), 100 the National Heart, Lung, and Blood Institute, 18 the AAFP, 98 Bright Futures, 101 the American Medical Association, 102 and the American Heart Association 103 recommend that children and adolescents receive blood pressure measurements every 1 or 2 years during regular office visits. The Canadian Task Force found insufficient evidence to recommend for or against routine blood pressure measurement in persons under age 21 years. 99 The AAP recommends against universal neonatal blood pressure screening. 104Discussion
It is clear from several large randomized clinical trials that lowering blood pressure in hypertensive adults is beneficial and that death from several common diseases can be reduced through the detection and treatment of high blood pressure. Estimates suggest that an average diastolic blood pressure reduction of 5-6 mm Hg in everyone with hypertension could reduce the incidence of coronary heart disease by 14% and the incidence of strokes by 42%. 30,31 At the same time, it is important for clinicians to minimize the potential harmful effects of detection and treatment. For example, if performed incorrectly, sphygmomanometry can produce misleading results. Some hypertensive patients thereby escape detection (false negatives) and some normotensive persons receive inappropriate labeling (false positives), which may have certain psychological, behavioral, and even financial consequences. 105 Treatment of hypertension can also be harmful as a result of medical complications, especially related to drugs. Clinicians can minimize these effects by using proper technique when performing sphygmomanometry, making appropriate use of nonpharmacologic methods, and prescribing antihypertensive drugs with careful adherence to published guidelines. 1,106-108
The diastolic blood pressure above which therapy has been proven effective (i.e., diastolic blood pressure > 90 mm Hg) is to a large extent based on the artificial cutpoints chosen for study purposes rather than on a specific biologic cutpoint defining increased risk. The coronary heart disease mortality risk associated with blood pressure occurs on a continuum that extends well below the arbitrarily defined level for abnormal blood pressure, beginning for systolic blood pressure above 110 mm Hg and for diastolic pressure above 70 mm Hg. 5 Nevertheless, many organizations outside the U.S. have been reluctant to recommend drug therapy for persons with diastolic blood pressures below 100 mm Hg who lack additional risk factors. 106,108-111 Drug treatment of mild hypertension is of particular concern for young adults: the evidence for therapeutic benefit comes primarily from several older trials 34,36,38 that included only a few individuals in their 20s, the potential adverse effects of decades of antihypertensive therapy are undefined, and the absolute benefits in young adults are likely to be limited given their small risk of stroke and coronary heart disease. For persons with mild hypertension, most recommendations suggest including age and/or the presence of other cardiovascular disease risk factors or concomitant diseases (e.g., smoking, obesity, renal disease, peripheral vascular disease) to modify treatment decisions. 1,106,108-111
Tracking studies and pathophysiologic evidence suggest there may be some benefit from early detection of primary hypertension in childhood, but there is insufficient evidence to support routine screening solely for this purpose. The lack of evidence is of concern because it is unclear whether a policy of routinely screening all children and adolescents to detect primary hypertension would achieve sufficient clinical benefit later in life to justify the costs and potential adverse effects of widespread testing and treatment. Potentially curable causes of hypertension, which account for a relatively large proportion of cases in young children, are often overlooked on history and physical examination, with rare but potentially catastrophic consequences. Evidence from case series and multiple time series indicate that early detection of secondary hypertension in childhood is of substantial benefit to the small number of patients affected.
The draft update of this chapter was prepared for the U.S. Preventive Services Task Force by Carolyn DiGuiseppi, MD, MPH, based in part on material prepared for the Canadian Task Force on the Periodic Health Examination by Alexander G. Logan, MD, FRCPC, and Christopher Patterson, MD, FRCPC.4. Screening for Asymptomatic Carotid Artery Stenosis
Burden of Suffering
Cerebrovascular disease is the third leading cause of death in the U.S., accounting for over 149,000 deaths in 1993. 1 Most stroke-related morbidity and mortality occur in older adults: 87% of all deaths and 74% of all hospitalizations occur in persons age 65 years or older. 2 Strokes can result in substantial neurologic deficits as well as serious medical and psychological complications. With an estimated prevalence of 3 million stroke survivors, 3 this illness places enormous burdens on family members and caretakers, often necessitating skilled care in an institutional setting. The direct and indirect costs of stroke in the U.S. have been estimated at $30 billion annually. 4 The principal risk factors for ischemic stroke are increased age, hypertension, smoking, coronary artery disease, atrial fibrillation, and diabetes. 4-7 Of these, the most important modifiable risk factors are hypertension and smoking. 8 Improved treatment of high blood pressure has been credited with the greater than 50% reduction in age-adjusted stroke mortality that has been observed since 1972 (see Chapter 3).
Population-based cohort studies have established that persons with carotid artery stenosis are at increased risk for subsequent stroke, myocardial infarction (MI), and death. 9,10 The risk of stroke is greatest for persons with neurologic symptoms such as transient ischemic attacks (TIAs), but is also increased in patients with asymptomatic lesions. The prevalence of hemodynamically significant carotid stenosis varies with age and other risk factors: population-based studies estimate that 0.5% of persons in their 50s and about 10% of those over age 80 have carotid stenosis greater than 50%. 11 The proportion of all strokes attributable to previously asymptomatic carotid stenosis seems to be small, however. In a study of 250 patients over age 60 with cerebral infarction, only 13% had ipsilateral carotid stenosis of 70% or greater. 12Accuracy of Screening Tests
Two methods are used to screen for carotid artery stenosis: clinical auscultation for carotid bruits and noninvasive studies of the carotid artery. Neck auscultation is an imperfect screening test for carotid stenosis. There is considerable interobserver variation among clinicians in the interpretation of the key auditory characteristics -- intensity, pitch, and duration -- of importance in predicting stenosis. 13 In addition, a cervical bruit can be heard in 4% of the population over age 40, but the finding is not specific for significant carotid artery stenosis. Between 40% and 75% of arteries with asymptomatic bruits do not have significant compromise in blood flow 14 similar sounds can also be produced by anatomic variation and tortuosity, venous hum, goiter, and transmitted cardiac murmur. 13,15-17 Finally, hemodynamically significant stenotic lesions may exist in the absence of an audible bruit. 13,15,18 Using 70-99% stenosis on a carotid angiogram as a reliable standard, a carotid bruit has a sensitivity of 63-76% and a specificity of 61-76% for clinically significant stenosis. 19
Persons with cervical bruits can be evaluated further with greater accuracy by noninvasive study of the carotid arteries. Older techniques (e.g., spectral analysis phonoangiography, continuous-wave or pulsed Doppler ultrasound, B-mode real-time ultrasound, oculoplethysmography, ophthalmodynamometry, periorbital directional Doppler ultrasound, and thermography) have been replaced largely by carotid duplex sonography which combines the capabilities of B-mode and Doppler ultrasound. A 1995 meta-analysis of 70 studies comparing the accuracy of noninvasive diagnostic tests to carotid angiography (the reference standard) concluded that carotid duplex, carotid Doppler, and magnetic resonance angiography (MRA) were equally effective in diagnosing carotid stenosis of 70% or greater: estimated sensitivity ranged from 83% to 86%, specificity from 89% to 94%. 20 Depending on the underlying population characteristics, the positive predictive value of carotid duplex ranges from 82% to 97%. 21 The performance of noninvasive tests for screening asymptomatic persons, however, has not been assessed in a prospective study. Although MRA seems to be quite sensitive and spares patients the risks of conventional angiography, it is unlikely to be a useful screening test due to costs (over $400) and inconvenience. 22Effectiveness of Early Detection
The rationale for testing for carotid artery stenosis is that persons with asymptomatic stenoses are not only at increased risk for cerebrovascular disease, 11,12 but that early detection can reduce morbidity due to cerebrovascular disease. According to this rationale there are several benefits to early detection of asymptomatic carotid stenosis. An awareness of the diagnosis might motivate patients to modify other risk factors (e.g., high blood pressure, smoking, physical inactivity). Performing carotid endarterectomy in some individuals might prevent subsequent cerebral infarction distal to the obstruction. Finally, antiplatelet drugs (aspirin and ticlopidine) might reduce stroke risk in asymptomatic individuals with carotid artery stenosis. No study has specifically compared a strategy of screening and early intervention in asymptomatic persons to intervening only in symptomatic patients (e.g., those with TIAs). The first symptom of carotid stenosis in some patients may be an irreversible stroke, however. A number of studies have examined whether interventions in asymptomatic persons can reduce the subsequent incidence of fatal and nonfatal stroke.
A bruit over the carotid artery is a fair indicator of vascular disease but a poor predictor that ischemic stroke will occur in its arterial distribution. The proportion of persons with asymptomatic bruits who will experience stroke is small: the annual incidence of stroke ipsilateral to a bruit and unheralded by TIAs is only 1-3%. 9,10,16,23-25 Higher grades of stenosis (assessed by sonography) are associated with increasing risk of neurologic events, rising to 5-7% per year with high-grade stenosis or total occlusion. 26,27 However, in those persons who will suffer a stroke, the degree of carotid stenosis does not always predict the risk of cerebral infarction, 16,23,28 or its location. 11,12 Carotid artery lesions may be less a predictor of atherothrombotic strokes than of generalized atherosclerotic disease persons with carotid artery disease are considerably more likely to die from ischemic heart disease than from a cerebrovascular event. 9,10
One of the major justifications for screening is the belief that carotid endarterectomy for high-grade, asymptomatic lesions detected through screening can prevent stroke. Three studies published before 1987 reported improved outcomes after endarterectomy. These studies provide poor quality evidence because they included previously symptomatic patients, were convenience samples derived from surgeons' practices, had inconsistent measurement criteria, or did not randomly assign patients. 17,29,30 Four more recent randomized trials have compared aspirin with endarterectomy in patients with asymptomatic carotid artery stenosis. The first study comparing aspirin alone with endarterectomy alone 31 enrolled only 71 patients before it was terminated due to excess MI in the surgical group no conclusions could be drawn regarding the effectiveness of endarterectomy for preventing stroke. 32 The second study, the 1991 European CASANOVA trial, randomized 410 patients with moderately severe stenosis (> 50% but < 90%) to treatment with aspirin/dipyridamole or aspirin/dipyridamole plus surgery. 33 The protocol was complex: some patients in both groups had contralateral symptoms, patients with stenosis greater than 90% were excluded, and 72 patients received therapy appropriate for the other group (more in the group randomized to surgery). There were no differences in the numbers of neurologic events and deaths between the two groups. The power of the study, however, was insufficient to exclude a clinically important benefit in the surgical group. 33 A third study, published in 1993, randomized 444 older veterans (mean age 64) with 50% or greater carotid stenosis to aspirin plus carotid endarterectomy or aspirin therapy alone. Patients who underwent carotid endarterectomy had lower rates of ipsilateral neurologic events, the primary endpoint: the combined incidence of TIAs, transient monocular blindness, and stroke was 8% in the surgery group versus 21% with aspirin only (p < 0.001), during an average follow-up of 48 months. The two groups had similar outcomes, however, using a combined endpoint of stroke or death from any cause. The power of the study was insufficient to exclude up to a 20% reduction in stroke in the surgically treated group. 34 The generalizability of this study was limited by the lack of female subjects and by the excessive morbidity and mortality in both groups (over 40% incidence of stroke or death in both groups over the 4-year follow-up).
The Asymptomatic Carotid Artery Study (ACAS), 35 funded by the National Institutes of Health, recently reported final results that provide stronger evidence of the benefit of endarterectomy for asymptomatic stenoses. 36 This multicenter study randomized 1662 patients with asymptomatic stenoses greater than 60% (mean stenosis 73%) to endarterectomy plus aspirin or to aspirin alone. Most patients (87%) were over age 60, and more than two thirds had coronary heart disease. The trial was stopped after a median follow-up of 2.7 years. The estimated 5-year risk for ipsilateral stroke or perioperative stroke or death was 5.1% for surgical patients and 11% for medically treated patients, a reduction in cumulative risk of 53% (95% confidence interval, 22 to 72). The absolute reduction in the combined incidence of major ipsilateral stroke, major perioperative stroke, or perioperative death, however, was considerably smaller (estimated 5-year risk of 3.4% in the surgery group vs. 6% in the medical group), not statistically significant (p = 0.13), and evident only in the fifth year of follow-up. Subgroup analyses suggest that endarterectomy may be less effective in women than in men (17% vs. 66% reduction in 5-year event rate), possibly due to higher perioperative complication rates (3.6% in women vs. 1.7% in men) neither of these differences between genders was statistically significant, however. The medical centers participating in this trial had been rigorously evaluated for the quality of patient management, and only surgeons with a perioperative complication rate of less than 3% among asymptomatic patients were allowed to participate. 37 Published studies have reported a perioperative mortality ranging from 1% to 3%, 33,34,38-40 and a perioperative stroke rate ranging between 2% and 10%, depending on patient characteristics and surgical expertise. 13,33,34,39-44 In six prospective trials of endarterectomy published after 1990, perioperative complication rates (stroke and death combined) range from 3% to 8%. 38 Complication rates seem to be lower in asymptomatic patients than in symptomatic patients, however. 38,40 A fifth trial of surgery versus medical management for asymptomatic carotid artery stenosis is still in progress. 11
Antiplatelet therapy with aspirin or ticlopidine offers a second possible intervention to reduce the risk of stroke in patients with asymptomatic carotid artery stenosis. Clinical trials have demonstrated a benefit of aspirin in reducing stroke among symptomatic patients (i.e., in persons with TIAs or stroke), 45-48 but observed no benefit on stroke in a large trial in asymptomatic physicians (prevalence of carotid disease unknown). 49 Among patients with asymptomatic carotid disease, who have a lower risk of ischemic events than do symptomatic patients, chronic aspirin therapy may not provide sufficient benefits to justify the documented risks of hemorrhagic complications (see Chapter 69). A multicenter prospective study comparing aspirin to placebo in asymptomatic patients with >50% carotid stenosis found no difference in stroke rates. 50 Ticlopidine is an alternative to aspirin in patients with risk factors for gastrointestinal hemorrhage, aspirin intolerance, and in patients who continue to have vascular events despite aspirin therapy, but its use is limited by high cost and small risk of neutropenia (approximately 1%). 51,52 The efficacy of ticlopidine in patients with asymptomatic carotid artery stenosis is not known.
Reducing serum lipids may slow the progression of carotid atherosclerosis and reduce clinical events. In a randomized trial enrolling patients with moderately elevated levels of LDL cholesterol (130-190 mg/dL) and early carotid atherosclerosis diagnosed by B-mode ultrasound, lovastatin induced regression of atherosclerosis and reduced total cardiovascular events compared to placebo. 52a Lipid-lowering drug therapy has not been examined specifically for treatment of advanced carotid stenosis, but is generally recommended for patients with high cholesterol and symptomatic vascular disease, based on its ability to reduce coronary heart disease mortality (see Chapter 2). No controlled studies have examined changes in the behavior of patients (e.g., smoking cessation or dietary modification) on learning the results of carotid artery examinations.Recommendations of Other Groups
Although auscultation of the carotid arteries is widely considered a routine component of the physical examination, the Canadian Task Force on the Periodic Health Examination 53 recommended against screening for bruits in asymptomatic persons, based on the poor sensitivity and specificity of cervical bruits as an indicator of significant carotid stenosis. The American Academy of Family Physicians recommends auscultation for carotid bruits in people age 40 and older with risk factors for cerebrovascular or cardiovascular disease, those with neurologic symptoms (e.g., TIA) or those with a history of cardiovascular disease 54 this policy is currently under review. The 1988 guidelines of the American College of Physicians recommend that patients with asymptomatic bruits should not have further diagnostic testing but should be educated about potential symptoms of a TIA in the carotid circulation. 55 In 1988, an ad hoc multidisciplinary consensus panel involved in designing the ACAS study recommended a baseline noninvasive study of the carotid arteries in persons considered at high risk for extracranial carotid arterial disease. 56 In 1992, the Ad Hoc Committee of the Joint Council of the Society for Vascular Surgery and the North American Chapter of the International Society for Vascular Surgery recommended that patients with asymptomatic carotid artery stenosis greater than 75% who are otherwise healthy and have a projected life expectancy more than 5 years should be considered for surgery if the operative morbidity and mortality rates are less than 3%. 57Discussion
The effectiveness of routine screening and intervention to reduce morbidity from asymptomatic carotid artery disease remains uncertain. The most effective interventions to prevent stroke are smoking cessation and the identification and treatment of hypertension. Although screening will detect some patients with asymptomatic high-grade carotid lesions who may benefit from endarterectomy, such patients account for only a small proportion of all strokes. In addition, there are several reasons to be cautious about undertaking widespread screening in asymptomatic persons on the basis of the current evidence: 11,38,58 the risk of major stroke ipsilateral to stenotic lesions is relatively low without surgery (approximately 1% per year) the absolute reduction in major stroke and death due to surgery over 5 years in ACAS was small and not conclusive surgery may result in other nonfatal complications (cranial nerve injury, MI, etc.) and the low complication rate of the ACAS-selected surgeons is not likely to reflect the typical risk of endarterectomy in the community. If complication rates of surgery are higher or underlying risk of stroke lower than reported for the ACAS study, the risks of surgery for asymptomatic carotid artery disease may outweigh the benefits. Routine screening will also subject some patients without significant carotid disease to the risks of angiography (1% risk of stroke), due to occasional false-positive results of carotid ultrasound.
As a result, it is not yet clear whether widespread screening in the primary care setting will be an effective way to reduce morbidity and mortality from stroke. Noninvasive testing for carotid artery stenosis is expensive (over $150 for carotid duplex or Doppler ultrasound) 20 the cost of screening 50% of the population over age 60 in the U.S. has been estimated at over $7 billion. 58 Auscultation for bruits involves little direct expense and may detect a majority of patients with severe stenosis, but the costs of follow-up testing of all patients with asymptomatic bruits could be substantial. Revised cost-effectiveness analyses of various screening and treatment strategies for asymptomatic carotid disease are under way. Patients most likely to benefit from screening are older men (over age 60) who have other risk factors for stroke, no contraindications to major surgery, and access to high-quality vascular surgery centers. Evidence regarding the effectiveness of antiplatelet drugs for asymptomatic persons is not yet sufficient to make a recommendation.
CLINICAL INTERVENTION5. Screening for Peripheral Arterial Disease
Burden of Suffering
Peripheral arterial disease (PAD) becomes increasingly common with age. An estimated 12-17% of the population over age 50 have PAD. 1-4 Increased mortality has been well documented in patients with PAD, a disease that is strongly associated with coronary artery disease and that shares many of the same risk factors. 1,2,5-9 Although only a small proportion of individuals with PAD and intermittent claudication develop skin breakdown or limb loss, pain and associated disability often restrict ambulation and the overall quality of life. 1,5 Persons at increased risk for PAD include cigarette smokers and persons with diabetes mellitus or hypertension. 1,5,9,10 Diabetic PAD is responsible for about 50% of all amputations. 1Accuracy of Screening Tests
There is evidence that a history of intermittent claudication and the palpation of peripheral pulses are unreliable techniques for the detection of PAD. 1,3,8,11 In one study, a battery of noninvasive tests for PAD was administered to 624 hyperlipidemic subjects aged 38-82. 7 In this population, the sensitivity and positive predictive value of a classic history of claudication were only 54% and 9%, respectively, when compared with the results of formal noninvasive testing. The sensitivity of an abnormal posterior tibial pulse was 71%, the positive predictive value was 48%, and the specificity was 91%. An abnormal dorsalis pedis pulse had a sensitivity of only 50% this artery is congenitally absent in 10-15% of the population. 11 The authors concluded that symptoms and abnormal pulses are not pathognomonic for PAD. 7 Greater accuracy has been achieved with noninvasive testing using Doppler ankle-arm pressure ratios, measurement of reactive hyperemia after exercise, pulse reappearance time, ultrasound duplex scanning, and plethysmography. 1,5,12,13 At present, however, additional data on sensitivity, specificity, and positive predictive value of these tests in asymptomatic populations are needed before noninvasive testing can be considered for routine screening.Effectiveness of Early Detection
Because surgery for PAD is offered only to patients with symptomatic disease, the rationale for the early detection of asymptomatic PAD is that risk factor modification following detection might lower subsequent morbidity and mortality from PAD and systemic atherosclerotic disease. By virtue of its strong association with coronary atherosclerosis and coronary events, 5 the early diagnosis of PAD might also lead to the detection of asymptomatic coronary heart disease. Evidence of these benefits is lacking, however. There has been no research to examine whether the detection and treatment of asymptomatic persons with PAD reduces the morbidity or mortality observed in symptomatic patients. It is clear that certain interventions are beneficial in symptomatic persons. There is evidence, for example, that patients who stop smoking have marked improvement in PAD symptoms and reduced overall cardiovascular mortality. 1,14 Certain antithrombotic drugs may also be of benefit. 15 It is unknown whether treatment measures used in symptomatic patients are beneficial in asymptomatic patients. 1,6 Examples include walking programs, control of weight and blood pressure, correction of elevated serum lipids and glucose, proper foot care, and certain drugs.Recommendations of Other Groups
There are no official recommendations for physicians to screen asymptomatic persons for PAD, although inspection of the skin and palpation of peripheral pulses are often included in the physical examination of the extremities. A recent international workshop sponsored by the American Diabetes Association and American Heart Association recommends annual screening for PAD in patients with diabetes. 16Discussion
Evidence is lacking that routine screening for PAD in asymptomatic persons is effective in reducing morbidity or mortality from this disease. Many of the behavioral interventions that might be prescribed after detecting PAD -- smoking cessation (Chapter 54), blood pressure control (Chapter 3), and exercise (Chapter 55) -- can be recommended without screening and are of proven value in the prevention of other atherosclerotic conditions, such as coronary heart and cerebrovascular disease. Screening by physical examination in the general population of asymptomatic adults, where the prevalence of PAD is low, is likely to produce a substantial number of false-positive results. Positive screening results will necessitate expensive noninvasive tests and may lead to potentially hazardous invasive tests such as arteriography. At the same time, it is not known whether the early detection of PAD in asymptomatic patients will result in more effective treatment of risk factors or better outcomes.
CLINICAL INTERVENTION6. Screening for Abdominal Aortic Aneurysm
Burden of Suffering
Approximately 8,700 deaths from abdominal aortic aneurysm (AAA) were reported in the U.S. in 1990, 1 but undiagnosed ruptured aneurysms are probably responsible for many additional cases of sudden death in older persons. Once rupture occurs, massive intraabdominal bleeding is usually fatal unless prompt surgery can be performed. A review of six case-series including 703 cases of ruptured aneurysm estimated that only 18% of all patients with ruptured AAA reached a hospital and survived surgery. 2 The large majority of deaths from AAA occur in older men and women men over 60 and women over 70 accounted for 95% of all deaths from AAA in a recent report. 2 Approximately 0.8% of male deaths and 0.3% of female deaths among persons over 65 years of age in the U.S. were attributed to AAA in 1990. 1
An aneurysm is usually defined as a focal dilation of the aorta at least 150% of the normal aortic diameter. 3 Given a normal aortic diameter in older men of 2 cm (range 1.4-3.0 cm), 4 an aortic diameter above 3 cm usually indicates an aneurysm. The pathogenesis of aneurysms is not completely understood, but well-established risk factors for AAA include increasing age, male gender, and family history of aneurysms. 3 The male to female ratio for death from AAA is 11:1 between ages 60 and 64 and narrows to 3:1 between ages 85 and 90. 5 Other possible risk factors include tobacco use, hypertension, peripheral vascular disease, and presence of peripheral arterial aneurysms. 3,6-8 In populations over age 60, estimates of prevalence range from 2% to 8% and increase with age. 8-10 A recent community study in England screened nearly 9,000 men and women aged 65-80 with ultrasound: 7% of men and 1% of women had an aneurysm at least 3 cm in diameter. 11 Among all patients, only 0.6% had aneurysms 5 cm or larger, and only 0.3% had aneurysms of 6 cm or more. 11 There are only limited data on the incidence of new aneurysms in a previously screened population. In one study, 189 men who had a normal ultrasound at age 65-66 years were rescreened 5 years later only 2 (1%) had an aortic diameter greater than 3 cm. 12
Few aneurysms less than 4 cm in diameter will rupture. 2,13,14 Overall, 3-6% of aneurysms greater than 4 cm in diameter will rupture annually, 14,15 but the rate of rupture is directly related to the size of the aneurysm. The natural history of most aneurysms is one of gradual enlargement growth rates have been estimated to average 0.2 cm/year for aneurysms under 4 cm, and 0.5 cm/year for those over 6 cm. 8Efficacy of Screening Tests
Two tests, palpation of the abdomen during physical examination and abdominal ultrasound, have been seriously advocated as screening tests for AAA. Other tests that can detect aneurysms -- plain radiographs of the abdomen, computed tomography (CT), and magnetic resonance imaging (MRI) -- are either not sensitive enough or are too expensive to be practical for screening in asymptomatic populations.
The accuracy of physical examination in detecting AAA is not completely known. Large aneurysms are easier to detect than small ones, and it is easier to detect aneurysms in thin people. Estimates of the sensitivity of physical examination in detecting AAA range from 22% to 96%. 17,18 The high sensitivity obtained in series of preoperative cases probably represents the preponderance of large aneurysms in this population. Lederle reported a sensitivity of 50% and a positive predictive value of 35% in a high-risk population screened in an internal medicine clinic (9% prevalence of AAA). 19 Four of five aneurysms greater than 5 cm diameter in this series were detected by palpation. In contrast, Allen reported a 22% sensitivity, 94% specificity, and positive predictive value of 17% in a population with a 5% prevalence of aneurysms. 17 No large-scale community-based studies of screening for AAA by physical examination have been reported.
Ultrasound is an extremely sensitive and specific test for AAA of all sizes, at least in cases where the diagnosis and size of the aneurysm can be confirmed at surgery. Reported sensitivities range from 82% to 99%, with sensitivity approaching 100% in some series of patients with a pulsatile mass. 16 In a small proportion of patients, visualization of the aorta will be inadequate due to obesity, bowel gas, or periaortic disease. Although ultrasound screening is noninvasive and relatively simple, compliance with invitations to be screened has been variable (50-64% attendance) in community screening trials. 7,11 Diagnostic abdominal ultrasound is currently expensive in the U.S. ($100-$175 per examination), but screening for AAA alone could probably be performed much more quickly and cheaply. 2Effectiveness of Early Detection
No prospective or retrospective controlled trials of screening for AAA that include outcome data have yet been reported. A pilot trial in England that offered screening at random to older subjects has enrolled 15,000 men and women, but it may not have sufficient power to prove a benefit in mortality. 8 The difficulty of identifying all deaths caused by AAA, combined with varying compliance with screening, may make it difficult to conduct definitive controlled trials of AAA screening. 8,20
Surgical resection and repair with an artificial graft is a very effective treatment for AAA. Among 13 large case-series of surgery for nonruptured aneurysms published since 1980, overall surgical mortality was 4% (range 1.4-6.5%) mortality during emergency surgery for rupture is much higher, averaging 49% (range 23-69%). 3 Mortality after elective surgery is often due to underlying cardiovascular disease in patients with AAA. If the patient survives the immediate postoperative period, long-term survival is comparable to similar persons without aneurysms, but late postoperative complications (graft infection, graft occlusion, and aortoenteric fistula) may result in additional deaths and morbidity. 3 The high prevalence of cardiovascular disease in patients with AAA and competing causes of morbidity and mortality in older patients may diminish the benefit of detecting asymptomatic aneurysms in older populations. Of 124 patients aged 65-80 who had large aneurysms detected in a community screening program, 27% were deemed unfit for surgery or died of other causes before surgery. 11 In recent series, up to 40% of patients undergoing surgery for nonruptured aneurysms had died within 6 years after surgery, primarily due to coronary heart disease or stroke. 3,21
Risk of elective surgery must be balanced against the risk of rupture of an untreated aneurysm, which is directly related to aneurysm size. Most vascular surgeons currently recommend surgery for asymptomatic aneurysms 5 cm or larger, since the risk of rupture (25-41% over 5 years) is substantially higher than risks from surgery. 3 While more aggressive management of smaller aneurysms (4-5 cm) has been recommended by some, 22 others have suggested that asymptomatic, slow-growing aneurysms under 6 cm can be successfully followed by serial ultrasound. 2,11 A large community-based screening program, which employed this conservative strategy over 8 years, observed two cases of rupture among 29 subjects with aneurysms 5-5.9 cm, for a rate of 1.5%/year. 11,23 A model fitted to data from 13 studies of untreated aneurysms supports a relatively low risk of rupture in aneurysms less than 6 cm estimated annual rates of rupture for aneurysms 4-4.9, 5-5.9, 6-6.9, and over 7 cm were 1%, 3%, 9%, and 25%, respectively. 2 These data, which are based largely on incidentally detected cases, may not reflect accurately the prognosis of asymptomatic aneurysms discovered by routine ultrasound screening. Furthermore, decisions to forgo surgery in patients with larger aneurysms were likely to have been influenced by factors (e.g., age, comorbidity, lack of symptoms) that may have independently influenced the risk of rupture. Trials are currently ongoing to determine the optimal management of patients with AAA that are 4-5.4 cm in size. 24Recommendations of Other Groups
The Canadian Task Force on the Periodic Health Examination 7 concluded that there was poor evidence to include or exclude screening for AAA in the periodic health examination of asymptomatic individuals. They noted that targeted physical examination may be considered prudent for men over 60, however, and that ultrasound screening could be considered in selected high-risk men over 60: smokers with other risk factors for AAA (hypertension, claudication, family history, or other vascular disease).Discussion
No prospective or retrospective controlled trials of screening for AAA have yet been reported that include data on mortality or other clinical outcomes. At present, the only effective intervention available for patients with aneurysms is major abdominal surgery. Until further data are available from population-based screening trials, it is uncertain whether the projected benefit from preventing ruptured aneurysms is sufficient to justify the costs of widespread screening and the potential risks from increased surgery. While there is general consensus that resection is indicated for incidentally discovered, large aneurysms (6 cm or larger), these are relatively uncommon in the general population the appropriate management of smaller (4-5 cm) aneurysms remains controversial. Data from older case series may not be a reliable guide to the natural history of asymptomatic aneurysms discovered by ultrasound. For many older patients with small aneurysms, the risk of dying from coronary heart disease or stroke is much higher than the risk from ruptured AAA. 21
The benefits of routine screening will depend on other parameters that merit further research: the proportion of clinically important aneurysms that are detected without screening the sensitivity and specificity of abdominal palpation for detecting AAA in the primary care setting risk factors for rapid growth or rupture of AAA and long-term morbidity of patients undergoing elective surgery. Patient compliance with recommendations for follow-up or surgery will also directly influence the ability of screening to prevent ruptured aneurysms.
A recent cost-effectiveness analysis compared different screening protocols in a high-risk population of men between 60 and 80 years of age. 25 The authors concluded that a single screen for AAA by abdominal palpation might be considered cost-effective, but it would be of small clinical benefit (average increase in life expectancy of 0.002 year). A single screen with ultrasound was at the high end of the cost range that might be considered cost-effective ($41,550/year of life gained), and repeat screening was not cost-effective. They noted that, due to the variable quality of the available data, screening for AAA could prove to be very cost-effective or could actually cause a net harm. If low-cost screening ultrasound were available (vs. $150 average charge for diagnostic ultrasound in the U.S.), ultrasound screening would be much more cost-effective, and preferable to physical examination. 25
CLINICAL INTERVENTION7. Screening for Breast Cancer
Burden of Suffering
In the U.S. in 1995, there were an estimated 182,000 new cases of breast cancer diagnosed and 46,000 deaths from this disease in women. 1 Approximately 32% of all newly diagnosed cancers in women are cancers of the breast, the most common cancer diagnosed in women. 1 The annual incidence of breast cancer increased 55% between 1950 and 1991. 2 The incidence in women during the period 1987-1991 was 110/100,000. 2 In 1992, the annual age-adjusted mortality from breast cancer was 22/100,000 women. 3 The age-adjusted mortality rate from breast cancer has been relatively stable over the period from 1930 to the present. 1,2 For women, the estimated lifetime risk of dying from breast cancer is 3.6%. 2 Breast cancer resulted in 2.2 years of potential life lost before age 65 per 1,000 women under age 65 in the U.S. during 1986-1988. 4 This rate was surpassed only by deaths resulting from motor vehicle injury and infections. Breast cancer is the leading contributor to cancer mortality in women aged 15-54, 1 although 48% of new breast cancer cases and 56% of breast cancer deaths occur in women age 65 and over. 2 As the large number of women in the "baby boom" generation age, the number of breast cancer cases and deaths will increase substantially unless age-specific incidence and mortality rates decline.
Important risk factors for breast cancer include female gender, residence in North America or northern Europe, and older age. 5 In American women, the annual incidence of breast cancer increases with age: 127 cases/100,000 for women aged 40-44 229/100,000 for women aged 50-54 348/100,000 for women aged 60-64 and 450/100,000 for women aged 70-74. 2 The risk for a woman with a family history of breast cancer in a first-degree relative is increased about 2-3-fold, and for women under 50 it is highest when the relative had premenopausally diagnosed breast cancer. 6-9 Women with previous breast cancer or carcinoma in situ and women with atypical hyperplasia on breast biopsy are also at significantly increased risk. 6,7,10-12 Other factors associated with increased breast cancer risk include a history of proliferative breast lesions without atypia on breast biopsy, late age at first pregnancy, nulliparity, high socioeconomic status, and a history of exposure to high-dose radiation. 6,7,10-12 Associations between breast cancer and oral contraceptives, long-term estrogen replacement therapy, obesity, and a diet high in fat have been suggested, but causal relationships have not been established. 6,7,13,14Accuracy of Screening Tests
The three screening tests usually considered for early detection of breast cancer are clinical breast examination (CBE), x-ray mammography, and breast self-examination (BSE). Estimates of the sensitivity and specificity of these maneuvers depend on a number of factors, including the size of the lesion, the characteristics of the breast being examined, the age of the patient, the extent of follow-up to identify false negatives, the skill and experience of the examiner or radiographic interpreter, and (in the case of mammography) the quality of the mammogram. Because multiple clinical trials have demonstrated the effectiveness of screening, measures of screening test performance (such as sensitivity and specificity) are primarily helpful in comparing trials, screening programs, and community practice. Uniform definitions, however, are necessary for such comparisons. For example, different studies may use similar definitions of sensitivity, such as the number of screen-detected cancers compared to the total of screen-detected cancers plus interval cancers, but one may use a fixed interval (e.g., 12 months) 15 and another a variable interval (e.g., time to next screen), 16 making direct comparisons difficult. The ability to detect interval cancers may also vary and will affect such estimates.
A review 17 of the current clinical trial data, published and unpublished, summarized screening test performance for mammography using uniform definitions. Sensitivity of mammography did not dramatically differ across the trials. Estimates from three Swedish trials using mammography alone averaged about 75%, while estimates for mammography combined with CBE ranged from 75% in the Health Insurance Plan of Greater New York (HIP) to 88% in the Edinburgh trial and the Canadian National Breast Cancer Screening Study in women aged 50-59 (NBSS 2). Specificity estimates ranged from 98.5% in the HIP trial to 83% in the Canadian NBSS 2. Sensitivity estimates for mammography alone and for combined screening with CBE have generally been 10-15% lower for women aged 40-49 compared with women greater than age 50. 15,17-19 Preliminary results from two North American demonstration projects suggest improved sensitivity of mammography, especially for women in their forties, with current mammographic techniques. 20 Significant variations in interpreter performance have also been observed. 21-23 In the Canadian trials, agreement was about 50% beyond that attributable to chance between radiologists at five screening centers and a single reference radiologist. 21
The effectiveness of CBE alone has not been evaluated directly, but comparisons of the sensitivity and specificity of this maneuver to that of mammography can be considered. The Canadian NBSS 2 was designed to assess the incremental value of mammography above a careful, thorough (5-10 minutes) CBE. 24,25 Preliminary results showing no incremental benefit highlighted the fact that higher sensitivity (88% for mammography plus CBE vs. 63% for CBE alone) 17 may not guarantee improved effectiveness. Specificity was comparable or slightly better for CBE alone. Sensitivity of CBE for women aged 40-49 (Canadian NBSS 1) was about 10% lower at initial screen compared to the estimate for women aged 50-59 (Canadian NBSS 2). 26 Specificity estimates were similarly lower for younger women.
Data regarding the accuracy of BSE are extremely limited. One report calculated an upper limit of sensitivity ranging from 12 to 25% by assuming all interval cases in the clinical trials were detected by BSE. 17 Using a similar approach, the overall sensitivity of BSE alone was estimated to be 26% in women also screened by mammography and CBE in the Breast Cancer Detection Demonstration Project (BCDDP). 27 Estimated BSE sensitivity decreased with age, from 41% for women aged 35-39 to 21% for women aged 60-74. 27 Thus, as currently practiced, BSE appears to be a less sensitive form of screening than is CBE or mammography, and its specificity remains uncertain. The sensitivity of BSE can be improved by training, as measured by the proportion of benign lumps 28 detected on human models and artificial lumps 29 on silicone breast models, although whether this improved detection on models translates into improved personal BSE performance is unknown.
Adverse effects of screening tests are an important consideration. False-positive tests, resulting from the effort to maximize disease detection, may have negative consequences including unnecessary diagnostic tests. In the Canadian trials there were 7-10% false positives from combined screening with mammography and CBE among women aged 40-49 and 4.5-8% among those aged 50-59. 24,30 In a study of the yield of a first mammographic screening among women, half as many cancers per 1,000 first screening mammograms were diagnosed in women aged 40-49 (3/1,000) compared to women aged 50-59 (6/1,000). 31 Yet, women aged 40-49 underwent twice as many diagnostic tests per cancer detected compared to women aged 50-59 (43.9 vs. 21.9). Women aged 60-69 had a higher yield from screening, with 13 breast cancers diagnosed per 1,000 first screening mammograms and 10.2 diagnostic tests performed per cancer detected.
Mammographic screening may also adversely affect psychological well-being. Increased anxiety about breast cancer after a false-positive mammogram has been reported both at short- and long-term follow-up in studies surveying groups of screened women. 32,33 No impact on compliance in obtaining future screening examinations was observed, however. Women who underwent a surgical biopsy as a result of a false-positive screening mammogram were more likely to report their workup as a stressful experience than were those who did not have a biopsy. 32
Excess breast cancers in populations that received doses of ionizing radiation significantly greater than currently delivered by mammography, such as survivors from atomic bombing in Japan 34 and patients with benign breast disease, 35 have raised concerns about the potential radiation risk from screening mammograms. There is no direct evidence of an increased risk of breast cancer from mammographic screening, however. Assuming a mean breast dose of 0.1 rad from a mammogram and extrapolating from higher doses of radiation, modeling suggests that in a group of 100,000 women receiving annual screening from ages 50 to 75, 12.9 years would be lost due to radiogenic cancers but 12,623 years would be gained through a 20% reduction in breast cancer mortality as a result of that screening. 34
Fewer data are available regarding adverse effects associated with CBE and BSE. A dramatic increase in false-positives was observed after instruction in BSE in a nonrandomized controlled trial evaluating performance on human models, 28 although no increase was found in a randomized controlled trial evaluating performance on silicone breast models. 29 The latter study also assessed the impact of training on variables other than detection performance on models. Adverse effects, such as unnecessary physician visits, heightened anxiety levels, or increased radiographic and surgical procedures, were not observed. 29Effectiveness of Early Detection
Seven randomized controlled trials 16,30,36-40 have evaluated the effectiveness of screening for breast cancer in women by either mammography alone or combined with CBE compared to no periodic screening. The age of participants at date of first invitation ranged from 40 to 74. The six trials 16,36-40 that included women aged >=50 showed a reduction in breast cancer mortality of 20-30% in the intervention group. The reduction was statistically significant in the Health Insurance Plan of Greater New York (HIP), 37 the Swedish two-county trials, 16 an overview of the Swedish trials, 40 and two meta-analyses of the trials. 41,42
The results of these six trials including women aged >=50 have convincingly demonstrated the effectiveness of mammographic screening (with or without CBE) for breast cancer in women aged 50-69. The HIP trial screened women aged 40-64 with annual CBE and two-view mammography. 37 For women who were over age 50 at the time of entry into the study, mortality from breast cancer in the intervention group was more than 50% lower than in the control group at 5 years, decreasing to a 21% difference after 18 years of follow-up. The Edinburgh trial 36 screened women aged 45-64 from 84 general medicine practices with two-view mammography and CBE on the initial screen followed by annual CBE and biennial single-view mammography. Preliminary results at seven years found a relative risk of 0.80 (95% confidence interval [CI], 0.54 to 1.17) for women aged 50 and older at entry. The results from 10-year follow-up showed little change. 17 An overview pooled the data through 1989 from the four Swedish randomized controlled trials of breast cancer screening with mammography alone. 40 All women diagnosed with breast cancer before randomization were excluded and endpoints were independently reviewed. Breast cancer mortality was reduced by about 30% for women aged 50-69 at entry using an endpoint of breast cancer as the underlying cause of death. A meta-analysis that included the most recently published results of these trials reported a 23% reduction in breast cancer mortality for women aged 50 and older. 42 A meta-analysis of European case-control studies done within screening mammography programs also reported significantly reduced breast cancer mortality among women aged 50 and older. 42
There are few data regarding the optimal periodicity of screening in this age group. Although an annual interval has been recommended by many groups, an analysis of data from the Swedish two-county study found little evidence that an annual interval would confer greater benefit than screening every 2 years for women over the age of 50. 19 This trial used mammography alone, but the reduction in breast cancer mortality was similar to that seen in the trials combining CBE with mammography. 36,37 The similar mortality reductions found in screening trials using periodicities ranging from 12 to 33 months in women aged >=50 suggests that biennial screening intervals are as effective as annual intervals. In a meta-analysis of the trials evaluating screening mammography, 42 the estimated reduction in breast cancer mortality was the same (23%) for screening intervals of 12 months and 18-33 months in women aged 50-74.
There is limited and conflicting evidence regarding the benefit of screening women aged 70-74. The Swedish two-county trial and BCDDP time series included women up to age 74 at entry, and each found a reduction of breast cancer mortality for the intervention group as a whole. 16,44 The Swedish overview, however, reported a relative risk of 0.98 (95% CI, 0.63 to 1.53) at 12-year follow-up for the age subgroup 70-74. 40 The wide confidence interval, due to small numbers in this subgroup analysis, does not preclude the possibility of a substantial benefit from screening in this age group. No clinical trials have evaluated screening in women over 74 years of age at enrollment.
Although all six trials found a benefit of screening among the total group of enrolled women who were 40-74 years at entry, 16,36-40 there is uncertainty about the effectiveness of screening women between the ages of 40 and 49. The Canadian NBSS 1 was specifically designed to address this uncertainty. 30 This trial compared combined annual mammography and CBE to an initial CBE among women aged 40-49 at entry. At 7-year follow-up, no benefit of annual screening was observed (RR = 1.36 95% CI, 0.84 to 2.21). This Canadian trial has been the subject of much criticism. 45-47 Possible irregularities with randomization have been refuted by its investigators. 48 An independent review is planned by the National Cancer Institute of Canada to determine whether the randomization was compromised. Although mammography quality issues have also been a concern, there is little evidence to suggest that the practices were inconsistent with the standards of the other clinical trials or community practice at the time of the study. 48 In addition, improvement in mammographic quality over the course of the study period was noted by both inside and outside observers. 48 The proportion of controls receiving mammography, 26%, can be compared to available estimates of 13% in the two-county trial, 24% in the Malmo trial (35% for women 45-49), and 24% in the Stockholm trial. 38,39,49 This contamination may nevertheless have reduced the trial's ability to detect a benefit from the screening intervention. An excess of node-positive cancers detected in the intervention group raised concerns about subject randomization. 30 While this may have been the result of chance, other contributing factors suggested by the investigators include under-ascertainment secondary to lower surgical dissection rates in the control group and incomplete breast cancer ascertainment at preliminary follow-up (although these possibilities are unlikely to account for all of the observed excess). 48 Although the effect of these factors should diminish with long-term follow-up, the results are unlikely to achieve statistical significance because sample size calculations were based on an estimated 40% reduction in breast cancer mortality, which is greater than the typical reduction in mortality observed in the other six trials that included women in this age group. 30
Subgroup analyses of the other trials that included women under 50 have yielded conflicting evidence regarding the benefit of screening women aged 40-49. No benefit was observed in the Stockholm trial or in one arm of the two-county trial, while the remaining trials reported nonsignificant benefits of about 22% or more. 17,50 One meta-analysis, which pooled the results from 7-year follow-up of six published clinical trials without adjustment for variation in screening method or interval, found no reduction in breast cancer mortality for women in their forties (RR = 1.08 95% CI, 0.85 to 1.39). 41 When the Canadian trial was excluded from the analysis, the estimate changed little (RR= 0.99 95% CI, 0.74 to 1.32). The overview of the Swedish trials found a nonsignificant 13% reduction (RR= 0.87 95% CI, 0.63 to 1.20) only after 8-12 years of follow-up in this age group. 40 More recent meta-analyses of published mammography trial data reported nonsignificant 8-10% reductions in breast cancer mortality in women aged 40-49. 42,43 One meta-analysis reported a significant benefit for women in this age group when unpublished data were included and the Canadian trial was excluded. 43 Longer duration of follow-up was associated with a greater reduction in mortality, although this finding may have been due to chance. 42 Thus, there is conflicting evidence from clinical trials and meta-analyses, primarily based on subgroup analyses, regarding the benefit of screening women aged 40-49. An ongoing British trial is evaluating the effectiveness of annual mammography screening in women enrolled at age 40 or 41.
A recent analysis of data by tumor size, nodal status, and stage from the BCDDP, a U.S. screening project using annual two-view mammography and CBE, suggests comparable 14-year survival rates for women 40-49 and women 50-59. 51 A similar analysis of breast cancers detected in the Swedish two-county trial confirms this finding. 52 Based on these data, time series comparisons of survival, frequency of interval cancers in the two-county trial, and subgroup analysis of available clinical trial data, some experts have suggested that annual screening intervals may be necessary to achieve a reduction in breast cancer mortality from screening for women aged 40-49. 19,20,52 In a meta-analysis of published trial results, however, the estimated mortality reduction from screening women in this age group was similar for 12- and 18-33-month screening intervals (1% and 12%, respectively). 42
There is no direct evidence that assesses the effectiveness of CBE alone compared to no screening. Modeling studies of the HIP trial estimated that two thirds of the effectiveness of the combined screening may have been a result of CBE. 53-56 The Canadian NBSS 2 was designed to test the incremental value of annual mammography over a careful annual CBE among women aged 50-59 at study entry. 24 At 7-year follow-up, there was no difference in breast cancer mortality for the group receiving combined screening compared to CBE alone (RR = 0.97 95% CI, 0.62 to 1.52). This result suggests that thorough CBE may be as effective as mammography for screening in this age group. The confidence interval is wide, however, and substantial benefit or harm from screening is not excluded by the preliminary data. Concerns regarding the early quality of mammography of Canadian NBSS 1 also apply to this trial. 48 Long-term follow-up and additional studies are needed to confirm this apparent lack of an incremental benefit of mammography above a careful, thorough annual CBE. It is also unclear whether CBE adds benefit to screening with mammography. A meta-analysis of mammography trial results reported similar reductions in breast cancer mortality with and without the addition of CBE. 42
Evidence for effectiveness of BSE alone is also limited. In the United Kingdom Trial of Early Detection of Breast Cancer, a nonrandomized community trial, 40-50% of women living in two districts participated in BSE instruction that included a short film and a lecture by a specially trained health provider. 57 At 7-year follow-up, there was no reduction in breast cancer mortality in the BSE communities compared with the control districts. Baseline comparability of intervention and control districts, treatment variation by community, and contamination by other screening modalities were not assessed, however. A World Health Organization (WHO) population-based randomized controlled trial in Leningrad comparing formal BSE instruction to no intervention in women aged 40-64 has reported increases in physician visits, referrals for further screening tests, and excisional biopsies in the intervention group at 5-year follow-up. 58 Breast cancer patients in the two groups did not differ in number, size, or nodal status of their tumors. Completeness of endpoint assessment is a concern in this study, given the lack of a national tumor registry. Follow-up through 1999 is planned for reporting mortality results. In a case-control study of women who had been diagnosed with advanced stage (TNM III or IV) breast cancer, there was no association between disease status and self-reported BSE. 59 Proficiency in practicing BSE, however, was reported as poor by both cases and controls. For the small group of women reporting thorough BSE compared to all others, the relative risk was 0.54 (95% CI, 0.30 to 0.98). A meta-analysis of pooled data from 12 descriptive studies found that women who practiced BSE before their illness were less likely to have a tumor of 2.0 cm or more in diameter or to have evidence of extension to lymph nodes. 60 The studies from which these data were obtained, however, suffer from important design limitations and provide little information on clinical outcome (i.e., breast cancer mortality). Retrospective studies of the effectiveness of BSE have produced mixed results. 27,61-63Recommendations of Other Groups
The American Cancer Society (ACS), 64 American College of Radiology, 65 American Medical Association, 66 American College of Obstetricians and Gynecologists (ACOG), 67 and a number of other organizations 65 recommend screening with mammography every 1-2 years and annual CBE beginning at the age of 40, and annual mammography and CBE beginning at age 50.
The American Academy of Family Physicians (AAFP) recommends CBE every 1-3 years for women aged 30-39 and annually for those aged 40 and older, and mammography annually beginning at age 50 68 these recommendations are currently under review. The American College of Physicians (ACP) recommends screening mammography every 2 years for women aged 50-74 and recommends against mammograms for women under 50 or over 75 years and baseline mammograms. 69 The ACP makes the same recommendations for high-risk women, unless the woman expresses great anxiety about breast cancer or insists on more intensive screening. The Canadian Task Force on the Periodic Health Examination recommends annual CBE and mammography for women aged 50-69 and recommends against mammograms in women under 50. 70 The National Cancer Institute states there is a general consensus among experts that routine mammography and CBE every 1-2 years in women aged 50 and over can reduce breast cancer mortality, and that randomized clinical trials have not shown a statistically significant reduction in mortality for women under the age of 50. 71Discussion
At this time, there is little doubt that breast cancer screening by mammography with or without CBE has the potential of reducing mortality from breast cancer for women aged 50 through about 70. The benefit derived from biennial screening appears to be quite similar to the benefit derived from annual screening. Given this similarity in effectiveness, biennial screening is likely to have the added benefit of increased cost-effectiveness. The incremental value of CBE above mammography or vice versa is uncertain, although the Canadian NBSS 2 24 suggests that careful CBE may be as effective as mammography.
Evidence does not establish a clear benefit from screening in women aged 40-49. Only the Canadian NBSS 1 30 was designed to test the effectiveness of screening in this age group, however, and none of the trials had adequate power for subgroup analysis. If screening is in fact ineffective in younger women, one possible explanation is a lower sensitivity of mammography in younger women (see Accuracy of Screening Tests ). Other possibilities include suboptimal screening intervals, differential (less aggressive) treatment offered to women with mammographically detected cancer, and varying biologic characteristics of breast tumors. 52,72 The Swedish overview, 40 HIP, 37 and Edinburgh 36 trials suggest some benefit in women aged 40-49 after 8-12 years of follow-up, but it is possible that the delayed benefit is due to screening women in their fifties who entered the trials in their middle to late forties. 72a The final results of the Canadian NBSS 1 may provide important information. An ongoing British trial and a proposed trial in Europe which will enroll women only in their early forties and compare mammography to no screening could also clarify this issue. 73 Until further information is available, it is unclear whether any potential improvement in breast cancer mortality achieved by screening women aged 40-49 is of sufficient magnitude to justify the potential adverse effects that may occur as a result of widespread screening.
Because breast cancer incidence increases with age, the burden of suffering due to breast cancer in elderly women is substantial. In addition, there is no evidence (as there is in younger women) that sensitivity of mammography in older women is not comparable to that in women aged 50-69. This is an age group, moreover, in which underutilization of breast cancer screening is common. 74-76 In a decision analysis of the utility of screening women over 65 for breast cancer, screening saved lives at all ages, but the savings decreased substantially with increasing age and comorbidities. 77 In the oldest women, those aged >=85, short-term morbidity such as anxiety or discomfort from the screening may have outweighed the small benefits. Until more definitive data become available for elderly women, it is reasonable to concentrate the large effort and expense associated with screening mammography on women in the age group for which benefit has been most clearly demonstrated: those aged 50-69. Screening women aged 70 and older might be considered on an individual basis, depending on general health and other considerations (e.g., personal preference of the patient).
The age range of 50-69 years, for which mammography has been proven effective, is to a large extent based on artificial cutpoints chosen for study purposes rather than on biologic cutpoints above or below which the ratio of benefits to risks sharply decreases. Both the incidence of breast cancer and the sensitivity of mammography increase with age. Thus, it is logical that women in their seventies, for whom only limited clinical trial experience is available, benefit from breast cancer screening. For women aged <50 years, evidence has been insufficient to establish a clear benefit from breast cancer screening. This age cutpoint may be a marker for biologic changes that occur with age, especially menopause. It is therefore plausible that women in their late forties, particularly postmenopausal women, might derive some intermediate benefit from screening. The risks and benefits of mammography and CBE may be best considered as changing on an age continuum rather than at a specific chronologic age. Guidelines for breast cancer screening should be interpreted with this in mind.
No large study has quantitated the effectiveness of breast cancer screening by either CBE or mammography for women who are at higher risk of developing breast cancer than the general population. The increased incidence of disease in high-risk women increases the positive predictive value (PPV) of screening tests used in this group. For example, in a community screening program, the PPV of mammography was increased 2-3-fold for women with a family history of breast cancer. 31 This is an especially important consideration for women under 50, in whom the benefit of screening has not been established for the general population. There may be a benefit from screening younger women in high-risk groups, but studies confirming this effect are lacking. Nevertheless, given their increased burden of suffering, screening high-risk women under age 50 may be considered on an individual basis for women who express a strong preference for such screening.
Data regarding the effectiveness of BSE are extremely limited, and the accuracy of BSE as currently practiced appears to be considerably inferior to that of CBE and mammography. False-positive BSE, especially among younger women in whom breast cancer is uncommon, could lead to unnecessary anxiety and diagnostic evaluation, although a small randomized clinical trial 29 did not find such adverse effects. A point also worth consideration is that time devoted to teaching BSE may reduce time available for prevention efforts with proven effectiveness. Given the present state of knowledge and the potential adverse effects and opportunity cost, a recommendation for or against inclusion of teaching BSE during the periodic health examination cannot be made.
CLINICAL INTERVENTION8. Screening for Colorectal Cancer
Burden of Suffering
Colorectal cancer is the second most common form of cancer in the U.S. and has the second highest mortality rate, accounting for about 140,000 new cases and about 55,000 deaths each year. 1 An individual's lifetime risk of dying of colorectal cancer in the U.S. has been estimated to be 2.6%. 2 About 60% of patients with colorectal cancer have regional or distant metastases at the time of diagnosis. 1 Estimated 5-year survival is 91% in persons with localized disease, 60% in persons with regional spread, and only 6% in those with distant metastases. 2 The average patient dying of colorectal cancer loses 13 years of life. 2 In addition to the mortality associated with colorectal cancer, this disease and its treatment -- surgical resection, colostomies, chemotherapy, and radiotherapy -- can produce significant morbidity. Persons at highest risk of colorectal cancer include those with uncommon familial syndromes (i.e., hereditary polyposis and hereditary nonpolyposis colorectal cancer [HNPCC]) and persons with longstanding ulcerative colitis. 3,4 Familial syndromes are estimated to account for 6% of all colorectal cancers, 3 and various genetic mutations associated with these syndromes have been identified. 4a Other principal risk factors include a history of colorectal cancer or adenomas in a first-degree relative, a personal history of large adenomatous polyps or colorectal cancer, and a prior diagnosis of endometrial, ovarian, or breast cancer. In an analysis of two large cohorts involving over 840,000 patient-years of follow-up, a family history of colorectal cancer was associated with a significant increase in risk in younger persons (1.7-4-fold increase between ages 40 and 60), but was not associated with a significantly increased risk in persons after age 60 4b risk was higher in persons with more than one affected relative. The absolute increase in lifetime risk in persons with a family history was modest, however: an estimated cumulative incidence of colorectal cancer by age 65 of 4% vs. 3% in persons without a family history. 4b Diets high in fat or low in fiber may also increase the risk of colorectal cancer. 3Accuracy of Screening Tests
The principal screening tests for detecting colorectal cancer in asymp-tomatic persons are the digital rectal examination, FOBT, and sigmoidoscopy. Less frequently mentioned screening tests include barium enema and colonoscopy, which have been advocated primarily for high-risk groups. The digital rectal examination is of limited value as a screening test for colorectal cancer. The examining finger, which is only 7-8 cm long, has limited access even to the rectal mucosa, which is 11 cm in length. A negative digital rectal examination provides little reassurance that the patient is free of colorectal cancer because fewer than 10% of colorectal cancers can be palpated by the examining finger. 3
A second screening maneuver is FOBT. The reported sensitivity and specificity of FOBT for detecting colorectal cancer in asymptomatic persons are 26-92% and 90-99%, respectively (usually based on two samples from three different stool specimens), with the widely varying estimates reflecting differences in study designs. 5-10 Positive reactions on guaiac impregnated cards, the most common form of testing, can signal the presence of bleeding from premalignant adenomas and early-stage colorectal cancers. The guaiac test can also produce false-positive results, however. The ingestion of foods containing peroxidases, 11 and gastric irritants such as salicylates and other antiinflammatory agents, 12 can produce false-positive test results for neoplasia. Nonneoplastic conditions, such as hemorrhoids, diverticulosis, and peptic ulcers, can also cause gastrointestinal bleeding. FOBT can also miss small adenomas and colorectal malignancies that bleed intermittently or not at all. 13,14 Other causes of false-negative results include heterogeneous distribution of blood in feces, 15 ascorbic acid and other antioxidants that interfere with test reagents, 16 and extended delay before testing stool samples. 17
As a result, when FOBT is performed on asymptomatic persons, the majority of positive reactions are falsely positive for neoplasia. The reported positive predictive value among asymptomatic persons over age 50 is only about 2-11% for carcinoma and 20-30% for adenomas. 6,5,9,18-20 Assuming a false-positive rate of 1-4%, a person who receives annual FOBT from age 50 to age 75 has an estimated 45% probability of receiving a false-positive result. 21 This large proportion of false-positive results is an important concern because of the discomfort, cost, and occasional complications associated with follow-up diagnostic tests, such as barium enema and colonoscopy. 22,23 Rehydration of stored slides can improve sensitivity, but this occurs at the expense of specificity. 24 In one study, rehydration improved sensitivity from 81% to 92%, but it decreased specificity from 98% to 90% and lowered positive predictive value from 6% to 2%. Due to the high false-positive rate, about one third of the entire screened population of asymptomatic patients underwent colonoscopy for abnormal FOBT results within a 13-year period. 5
Other tests have been proposed to improve the accuracy of screening for fecal occult blood. Current evidence is equivocal as to whether HemoQuant (SmithKline Diagnostics, Sunnyvale, CA), a quantitative measurement of hemoglobin in the stool, has better sensitivity or specificity than does qualitative FOBT. 9,10,25-29 Recently developed hemoglobin immunoassays offer the promise of improved sensitivity and specificity but require further evaluation before being considered for routine screening. 30,31
The third screening test for colorectal cancer is sigmoidoscopy. Sigmoidoscopic screening in asymptomatic persons detects 1-4 cancers per 1,000 examinations. 32,33 However, the sensitivity and diagnostic yield of sigmoidoscopy screening varies with the type of instrument: the rigid (25 cm) sigmoidoscope, the short (35 cm) flexible fiberoptic sigmoidoscope, or the long (60 cm) flexible fiberoptic sigmoidoscope. Since only 30% of colorectal cancers occur in the distal 20 cm of bowel, and less than half occur in or distal to the sigmoid colon, 34-37 the length of the sigmoidoscope has a direct effect on case detection. The rigid sigmoidoscope, which has an average depth of insertion of about 20 cm 38-44 and allows examination to just above the rectosigmoid junction, 45 can detect only about 25-30% of colorectal cancers. The 35-cm flexible sigmoidoscope, however, can visualize about 50-75% of the sigmoid colon and can detect about 50-55% of polyps. Longer 60-cm instruments have an average depth of insertion of 40-50 cm, reaching the proximal end of the sigmoid colon in 80% of examinations 46,47 with the capability of detecting 65-75% of polyps and 40-65% of colorectal cancers 48-52 Researchers have examined the feasibility of introducing a 105-cm flexible sigmoidoscope in the family practice setting, 53 but it is unclear whether the added length substantially increases the rate of detection of premalignant or malignant lesions. Barium enema studies have confirmed that some neoplasms within reach of the sigmoidoscope may not be seen on endoscopy. 54
Sigmoidoscopy can also produce false-positive results, primarily by detecting polyps that are unlikely to become malignant during the patient's lifetime. Autopsy studies have shown that 10-33% of older adults have colonic polyps at death, 55 but only 2-3% have colorectal cancer. 56-58 Depending on the type of adenomatous polyp, an estimated 5-40% eventually become malignant, 59 a process that takes an average of 10-15 years. 60,61 It follows that the majority of asymptomatic persons with colonic polyps discovered on routine sigmoidoscopic examination will not develop clinically significant malignancy during their lifetime. For these persons, interventions that typically follow such a discovery (i.e., biopsy, polypectomy, frequent colonoscopy), procedures that are costly, anxiety provoking, and potentially harmful, are unlikely to be of significant clinical benefit.
Other potential screening tests for colorectal cancer include colonoscopy and barium enema, which appear to have comparable accuracy. About 95% of colorectal cancers are within reach of the colonoscope, and the examination has an estimated 75-95% sensitivity in detecting lesions within its reach. 20,21 Colonoscopy, which requires sedation and often involves the use of a hospital suite, is more expensive than other screening tests and has a higher risk of anesthetic and procedural complications. The estimated sensitivity and specificity of air-contrast barium enema in detecting lesions within its reach are about 80-95% and 90%, respectively, using subsequent diagnosis as a reference standard. 21 Some retrospective studies have reported a higher sensitivity of barium enema for detecting colorectal cancer (about 90-96%), 62,63 with pathologic diagnosis as the reference standard, but these estimates generally do not account for the selection bias introduced by the case-selection methods.Effectiveness of Early Detection
Persons with early-stage colorectal cancer at the time of diagnosis appear to have longer survival than do persons with advanced disease. 2 Since there is little information on the extent to which lead-time and length biases (see Chapter ii) account for these differences, researchers in the U.S. and Europe launched large clinical trials in the late 1970s to collect prospective data on the effects of screening on co-lorectal cancer mortality.
Two of these trials 5,6 examined the effect of routine FOBT on colorectal cancer mortality. A randomized controlled trial involving over 46,000 volunteers over age 50 found that the 13-year cumulative mortality from colorectal cancer was 33% lower among persons advised to undergo annual FOBT (5.9 deaths per 1,000) than among a control group that was not offered screening (8.8 deaths per 1,000). 5 The report provided insufficient data, however, to determine to what extent observed differences in outcome were attributable to FOBT or to the large number of colonoscopies that were performed due to frequent false-positive FOBT. An analysis of the study data by other authors suggested that one third to one half of the mortality reduction was due to "chance" selection of persons for colonoscopy, 64 but the assumptions in the analysis have been disputed by the authors. 65 Another controlled trial, 6 which was not randomized, assigned over 21,000 patients to a control group that received a standard periodic health examination or to a study group that was also offered FOBT both groups received sigmoidoscopy screening. Among new patients (first visit to the preventive medicine clinic), colorectal cancer mortality was 43% lower in the study group than in controls, a difference of borderline statistical significance (p = 0.053, one-tail), and there was no difference in outcomes among patients seen previously at the clinic. Recent case-control studies have also reported a 31-57% reduction in risk among persons receiving FOBT. 66,67 Three large clinical trials of FOBT screening, currently under way in Europe, are expected to report their results in the coming years. 7,68,69
Recent case-control studies have provided important information on the effectiveness of sigmoidoscopy screening. The largest study found that 9% of persons who died of colorectal cancer occurring within 20 cm of the anus had previously undergone a rigid sigmoidoscopic examination, whereas 24% of persons in the control group had received the test. 70 The adjusted odds ratio of 0.41 (95% confidence interval, 0.25-0.69) suggested that sigmoidoscopy screening reduced the risk of death by 59% for cancers within reach of the sigmoidoscope. The investigators noted that the adjusted odds ratio for patients who died of more proximal colon cancers was 0.96. This finding added support to the hypothesis that the reduced risk of death from cancers within reach of the rigid sigmoidoscope was due to screening rather than to confounding factors. Another case-control study reported that the odds ratio for dying of colorectal cancer was 0.21 in screened subjects, and the benefit appeared to be limited to cancers within the reach of the sigmoidoscope. 71
Older evidence of the effectiveness of sigmoidoscopy screening suffered from important design limitations. A randomized controlled trial of multiphasic health examinations, which included rigid sigmoidoscopy, reported that the study group had significantly lower incidence and mortality rates from colorectal cancer. 72-74 A subsequent analysis of the data, however, revealed that the proportion of subjects receiving sigmoidoscopy and the rate of detection or removal of polyps were similar in both the study and control groups, thus suggesting little benefit from sigmoidoscopy. 75 Two large screening programs found that persons receiving periodic rigid sigmoidoscopy had less advanced disease and better survival from colon cancer than was typical of the general population. 76-78 However, both studies lacked internal controls and used nonrandomized methods to select participants other methodologic problems with these investigations are outlined in other reviews. 75,79
An important consideration in assessing the effectiveness of sigmoidoscopic screening is the potential iatrogenic risk associated with the procedure. Complications from sigmoidoscopy are relatively rare in asymptomatic persons but can be potentially serious. Perforations are reported to occur in approximately 1 of 1,000-10,000 rigid sigmoidoscopic examinations. 20,21,32,80 Although there are fewer data available on flexible sigmoidoscopy, the complication rate appears to be less than or equal to that observed for rigid sigmoidoscopy. The reported risk of perforation from colonoscopy is about one in 500-3,000 examinations, 5,81 and the risk of serious bleeding is 1 in 1,000. 5 The estimated risk of perforation during barium enema is 1 in 5,000-10,000 examinations. 82
There is little useful evidence regarding the effectiveness of colonoscopy or barium enema screening in asymptomatic persons. Several recent studies describe colonoscopy screening of asymptomatic persons, but they report only the anatomic distribution of polyps and do not address clinical outcomes. 48,49,83 A prospective study demonstrated a significantly lower incidence of subsequent colorectal cancer in patients with previously diagnosed adenomas who received periodic colonoscopy and polypectomy, but potential biases in the control groups (historical controls and population incidence rates) prevent definitive conclusions. 84 No studies have directly examined the effectiveness of routine barium enema screening in decreasing colorectal cancer mortality in asymptomatic persons. Modeling studies suggest its effectiveness might be comparable to a screening strategy of periodic sigmoidoscopy. 21
There is limited information on the optimal age to begin or end screening and the frequency with which it should be performed. The age groups in which screening has been shown to decrease mortality are ages 50-80 for FOBT 5 and over age 45 for sigmoidoscopy. 70 Theoretically, the potential yield from screening should increase beyond age 50 since the incidence of colorectal cancer after this age doubles every 7 years. 2 Modeling studies suggest that beginning screening at age 40 rather than at age 50 offers no improvement in life expectancy. 21 There is little evidence from which to determine the proper age for discontinuing screening. The optimal interval for screening is less certain for sigmoidoscopy than for FOBT, for which there is good evidence of benefit from annual screening. A modeling study of sigmoidoscopy screening estimated that an interval of 10 years would preserve 90% of the effectiveness of annual screening this model assumes that adenomatous polyps take 10-14 years to become invasive cancers. 21 Another model suggested that an interval of 2-4 years would allow detection of 95% of all polyps greater than 13 mm in diameter. 85 In a case-control study, the risk reduction associated with sigmoidoscopy screening did not diminish during the first 9-10 years after sigmoidoscopy. 70 Other studies suggest that a single sigmoidoscopic screening examination may be adequate for low-risk individuals, 86 an approach being investigated in the United Kingdom. 87
Primary preventive measures to prevent colorectal cancer are currently under investigation. An association between colorectal cancer and dietary intake of fat and fiber has been demonstrated in a series of epidemiologic studies (see Chapter 56). Case-control and cohort studies also suggest that aspirin use may decrease the risk of colon cancer. 88-90Recommendations of Other Groups
The American Cancer Society recommends annual digital rectal examination for all adults beginning at age 40, annual FOBT beginning at age 50, and sigmoidoscopy every 3-5 years beginning at age 50. 91 Similar recommendations have been issued by the American Gastroenterological Association, 92 the American Society for Gastrointestinal Endoscopy, 92 and the American College of Obstetricians and Gynecologists. 93 The American College of Physicians' (ACP) guidelines, revised in 1995, recommend offering a variety of screening options to persons from age 50 to 70, depending on local resources and patient preferences: flexible sigmoidoscopy, colonoscopy, or air-contrast barium enema, repeated at 10-year intervals. The ACP recommends that annual FOBT be offered to persons who decline these screening tests, but concluded that there was relatively little benefit of continuing endoscopic screening beyond age 70 in individuals who had been adequately screened up to that age. 21 The American College of Radiology recommends screening with barium enema every 3-5 years as an equivalent alternative to periodic sigmoidoscopy. 94 The recommendations of the American Academy of Family Physicians are currently under review. 95 Most organizations recommend more intensive screening of those in high-risk groups (e.g., familial polyposis, inflammatory bowel disease) with periodic colonoscopy or barium enema. The Canadian Task Force on the Periodic Health Examination concluded that there was insufficient evidence to support screening of asymptomatic individuals over age 40 but that persons with a history of cancer family syndrome should be screened with colonoscopy. 96 An expert panel convened by the Agency for Health Care Policy and Research is expected to issue guidelines for colorectal cancer screening and surveillance in 1996.Discussion
In summary, recent studies have provided compelling evidence of the effectiveness of FOBT and sigmoidoscopy screening, but the evidence is not definitive. At least one randomized controlled trial and several observational studies have shown that annual FOBT in persons over age 50 can reduce colorectal cancer mortality. This evidence does not, however, clarify whether the observed benefits were due to FOBT or to the effect of performing colonoscopy on a large proportion of the screened population. For sigmoidoscopy, a case-control study supports a strong association between regular screening and reduced colorectal cancer mortality from cancers within reach of the sigmoidoscope. This study was limited, however, by its small number of cases, potential selection biases, and inability to provide prospective evidence of benefit. There are additional concerns about the adverse effects, costs, and optimal frequency of screening. Studies that will help resolve these uncertainties are currently in progress the final results of ongoing European FOBT trials will be unavailable for several years, however, and a large United States study 97 of FOBT and sigmoidoscopy screening will not be completed until the turn of the century.
An important limitation to the effectiveness of screening for colorectal cancer is the ability of patients and clinicians to comply with testing. Patients may not comply with FOBT for a variety of reasons, 68,98 but compliance rates are generally higher than for sigmoidoscopy. Recent clinical trials report compliance rates of 50-80% for FOBT among volunteers, 5-7,68,69 but lower rates (about 15-30%) have been reported in community screening programs. 99-101 Although the introduction of flexible fiberoptic instruments has made sigmoidoscopy more acceptable to patients, 102 the procedure remains uncomfortable, embarrassing, and expensive, and therefore many patients may be reluctant to agree to this test. A survey of patients over age 50 found that only 13% wanted to receive a sigmoidoscopy examination after being advised that they should receive the test the most common reasons cited for declining the test were cost (31%), discomfort (12%), and fear (9%). 103 In a study in which sigmoid-oscopy was recommended repeatedly, only 31% of participants consented to the procedure, 72-74 but this study was performed during years when rigid sigmoidoscopy was common. Compliance rates as low as 6-12% have been reported. Studies suggest that physician motivation is a major determinant of patient compliance, 104,105 and physicians may be reluctant to perform screening sigmoidoscopy on asymptomatic persons. It has been estimated that a typical family physician with 3,000 active patients (one third aged 50 or older) would have to perform five sigmoidoscopies daily to initially screen the population and two daily procedures for subsequent screening. 33 In addition, examinations using 60-cm sigmoidoscopes are more time-consuming 35-39 and require more extensive training 106-108 than do those using shorter instruments.
Another limitation to screening is its cost. Although a formal cost-effectiveness analysis of screening for colorectal cancer is beyond the scope of this chapter, the economic implications associated with the widespread performance of FOBT and sigmoidoscopy are clearly significant. A single flexible sigmoidoscopic examination costs between $100 and $200. 22,109 A policy of routine FOBT and sigmoidoscopic screening of all persons in the United States over age 50 (about 63 million persons) would cost over $1 billion per year in direct charges. 109 Others have calculated that FOBT screening alone could cost the United States and Canada between $500 million and $1.2 billion each year. 110,111 Another model predicted that performing annual FOBT on persons over age 65 would cost about $35,000 per year of life saved adding flexible sigmoidoscopy would increase the cost to about $42,000 to $45,000 per year of life saved. 112 Mathematical models suggest that barium enema screening every 3-5 years might have comparable or superior cost-effectiveness when compared with sigmoidoscopy screening, but neither the clinical effectiveness nor acceptability of barium enema screening has been demonstrated directly in clinical studies.
The downstream effects of screening are also of concern. The logistical difficulties and costs of performing FOBT and sigmoidoscopy on a large proportion of the U.S. population are significant, due to the limited acceptability of the tests and the expense of performing screening and follow-up on a large proportion of the population. Moreover, the tests have potential adverse effects that must be considered, such as false-positive results that lead to expensive and potentially harmful diagnostic procedures. Studies that have reported reduced mortality from FOBT used rehydrated slides to increase sensitivity, thereby producing a higher proportion of false-positive results than with nonrehydrated slides 32% of the annually screened population underwent colonoscopy during a 13-year follow-up period. 65 If this rate is extrapolated to the 63 million Americans over age 50 who would receive annual FOBT, it can be predicted that about 20 million persons would require colonoscopy.
The full implications of this "screening cascade" need to be considered, along with the scientific evidence of clinical benefits, before reaching conclusions about appropriate public policy. For example, using nonrehydrated slides rather than rehydrated slides could substantially reduce the adverse effects and costs of a national screening program. As noted earlier, data from a major screening trial suggest that using nonrehydrated slides rather than rehydrated slides could increase the positive predictive value of FOBT from 2% to 6%, subjecting far fewer screened persons to unnecessary colonoscopy. This improvement in specificity, however, comes at the expense of sensitivity, which decreased from 92% with rehydration to 81% in nonrehydrated slides. The use of nonrehydrated slides would therefore allow a much larger proportion of persons with cancer to escape detection.
The special considerations that apply to persons at increased risk of colorectal cancer are complicated by inadequate epidemiologic and effectiveness data and inconsistent disease classifications. Having a single family member with colorectal cancer does not carry the high risk associated with hereditary cancer syndromes (e.g., familial polyposis, HNPCC). 4,113 A family history that is suggestive of the latter includes a pattern of diagnoses consistent with autosomal dominant inheritance of a highly penetrant disorder. Characteristic features include a family history of colorectal cancer being diagnosed at an early age, frequent cases of multiple primary cancers, or florid adenomatous colonic polyps. Performing periodic colonoscopy to screen for cancer in these groups may be justified in light of the high risk of disease and the incidence of proximal colonic lesions, but there is no direct evidence to determine the optimal strategy in this population.
CLINICAL INTERVENTION9. Screening for Cervical Cancer
Burden of Suffering
Approximately 16,000 new cases of cervical cancer are diagnosed each year, and about 4,800 women die from this disease annually. 1 The lifetime risk of dying from cervical cancer in the U.S. is 0.3%. 1a Although the 5-year survival rate is about 90% for persons with localized cervical cancer, it is considerably lower (about 14%) for persons with advanced (Stage IV) disease. The incidence of invasive cervical cancer has decreased significantly over the last 40 years, due in large part to organized early detection programs. Although all sexually active women are at risk for cervical cancer, the disease is more common among women of low socioeconomic status, those with a history of multiple sex partners or early onset of sexual intercourse, and smokers. The incidence of invasive cervical cancer among young white women has increased recently in the United States. Infection with human immunodeficiency virus (HIV) and certain types of human papilloma virus (HPV) also increases the risk of cervical cancer. 2Accuracy of Screening Tests
The principal screening test for cervical cancer is the Pap smear. Although the Pap smear can sometimes detect endometrial, vaginal, and other cancers, 3,4 its use as a screening test is intended for the early detection of cervical dysplasia and cancer. Other proposed cervical screening tests include cervicography, colposcopy, and testing for HPV infection. The role of pelvic examination, which usually accompanies the collection of the cervical specimen, is discussed in Chapter 14 in relation to ovarian cancer screening.
Precise data on the sensitivity and specificity of the Pap smear in detecting cancer and dysplasia are lacking due to methodologic problems. Depending on study design, false-negative rates of 1-80% have been reported a range of 20-45% has been quoted most frequently, primarily in studies comparing normal test results with subsequent smears. 5-11 Studies using cone biopsy results as the reference standard have reported false-negative rates as low as 10%. 12 Although reliable data are lacking, specificity is probably greater than 90% 13 and may be as high as 99%. 6,11 The detection of precursor cervical intraepithelial neoplasia (CIN) by Pap smears may have poor specificity for cervical carcinoma, however, because a substantial proportion of CIN-1 lesions do not progress to invasive disease or may regress spontaneously. The test-retest reliability of Pap smears is influenced to some extent by variations in the expertise and procedures of different cytopathology laboratories.
A large proportion of diagnostic errors may be attributable to laboratory error. In one study of over 300 laboratories given slides with known cytologic diagnoses, false-negative diagnoses were made in 7.5% of smears with moderate dysplasia or frank malignancy, and false-positive diagnoses were made in 8.9% of smears with no more than benign atypia. 14 A survey of 73 laboratories in one state revealed a false-negative rate of 4.4% and a false-positive rate of 2.7%. 15 These data were reported in 1990, before the introduction of federal legislation designed to improve the accuracy of cytopathologic laboratory interpretation. 16 With the adoption of the Bethesda system for classification of cervical diagnoses, 17 a large proportion of benign smears are interpreted as "atypical," a finding that poses little premalignant potential but that often generates intensive follow-up testing.
Another cause of false-negative Pap smears is poor specimen collection technique. A 1991 survey of 600 laboratories found that 1-5% of specimens received were either unsatisfactory or suboptimal, generally because endocervical cells were absent from the smear. 18 Another study found that poor sampling technique accounted for 64% of false-negative results. 19 The Pap smear has traditionally been obtained with a spatula, to sample the ectocervix, and a cotton swab, to obtain endocervical cells. A 1990 survey found that about half of physicians used a spatula and cotton swab to collect Pap smears. 20 In recent years, new devices have been introduced to improve sampling of the squamocolumnar junction. Controlled studies have shown that using an endocervical brush in combination with a spatula is more likely to collect endocervical cells than using a spatula or cotton swab. 21-30 There is conflicting evidence, however, that the endocervical brush increases the detection rate for abnormal smears or affects clinical outcomes. 31-33 There is also conflicting evidence regarding the importance of collecting endocervical cells. Although some large series have reported that CIN is detected over 2 times more frequently when endocervical cells are present, 34,35 other series 36,37 have shown no association between the presence of endocervical cells and the detection rate for dysplasia. The brush is more expensive than the cotton swab, but studies suggest that this cost is easily recovered by the reduced need for repeat testing. 38 Other methods for improving the sensitivity of cervical cancer screening, such as acetic acid washes to improve the visibility of lesions, remain investigational. 39,40
There are important potential adverse effects associated with inaccurate interpretation of Pap smears. False-negative results are significant because CIN or more invasive lesions may escape detection and progress to more advanced disease during the period between tests. The potential adverse effects of false-positive results include patient anxiety regarding the risk of cervical cancer, 41,42 as well as the unnecessary inconvenience, discomfort, and expense of follow-up diagnostic procedures. Studies have shown that the distribution of patient education materials that explain the meaning of abnormal results is associated with a reduction in patient anxiety and stress and a better patient understanding of test results. 43-45
Other tests, such as cervicography and colposcopy, have been proposed to help improve the sensitivity of screening, 46 but their accuracy and technical requirements are suboptimal. Cervicography, in which a photograph of the cervix is examined for atypical lesions, has a sensitivity that is comparable to the Pap smear (approximately 60%) but a much lower specificity (approximately 50%) the reported positive predictive value in most studies is only 1-26%, and about 10-15% of cervigrams are unsatisfactory. 47-51 Colposcopy, in which the cervix is examined under magnification with acetic acid washing and suspicious lesions are biopsied, is widely performed on women with abnormal Pap smears but has poor sensitivity (34-43%), specificity (68%), and positive predictive value (4-13%) when used as a screening test for cervical neoplasia in asymptomatic women. 52-54 Other disadvantages of colposcopy screening include its cost, the limited availability of the equipment, the time and skills required to perform the procedure, and patient discomfort. Using a 10-point score for assessing pain, one study reported that women who underwent colposcopy gave the procedure a range of scores from 3 to 4.6. 55
Another proposed screening strategy is testing for HPV infection, a known risk factor for cervical cancer. Of the more than 70 types of HPV that have been identified, several oncogenic forms (e.g., types 16 and 18) have a strong epidemiologic association with cervical cancer. However, the natural history of how HPV infection progresses to cancer is poorly understood. 56 One study of women infected with either HPV type 16 or 18 found that 67% of the lesions remained unchanged or regressed after a mean of 5 years, 29% progressed to a more advanced stage of dysplasia, and 3% recurred. 57 The high prevalence of HPV infection in young women also limits its predictive value. In one study, nearly half of female college students had evidence of HPV when tested by polymerase chain reaction technology. 58 The reported positive predictive value of this HPV test for CIN-2 or CIN-3 lesions and carcinoma is less than 10%. 59 HPV typing to identify women with oncogenic strains may improve the future accuracy of the test and its role in directing follow-up, but its current suitability for routine screening in asymptomatic women is limited by its poor predictive value, uncertain natural history, and, due to the absence of an effective treatment, the lack of evidence that screening affects clinical outcomes. 60Effectiveness of Early Detection
Early detection of cervical neoplasia provides an opportunity to prevent or delay progression to invasive cancer by performing clinical interventions such as colposcopy, conization, cryocautery, laser vaporization, loop electrosurgical excision, and, when necessary, hysterectomy. 61 There is evidence that early detection through routine Pap testing and treatment of precursor CIN can lower mortality from cervical cancer. Correlational studies in the United States, Canada, and several European countries comparing cervical cancer data over time have shown dramatic reductions in the incidence of invasive disease and a 20-60% reduction in cervical cancer mortality rates following the implementation of cervical screening programs. 62-70 Case-control studies have shown a strong negative association between screening and invasive disease, also suggesting that screening is protective. 71-75 These observational studies do not constitute direct evidence that screening was responsible for the findings, 76 and randomized controlled trials to provide such evidence have not been performed. Nonetheless, the large body of supportive evidence accumulated to date has prompted the adoption of routine cervical cancer screening in many countries and makes performance of a controlled trial of Pap smears unlikely for ethical reasons.
Observational data suggest that the effectiveness of cervical cancer screening increases when Pap testing is performed more frequently. 72 Aggressive dysplastic and premalignant lesions are less likely to escape detection when the interval between smears is short. There are, however, diminishing returns as frequency is increased. 71,77 Although studies have shown that reducing the interval between Pap smears from 10 years to 5 years is likely to achieve a significant reduction in the risk of invasive cervical cancer, case-control studies and mathematical modeling have demonstrated that increasing to a 2-3-year interval offers only slight added benefit. 71,78-80 There is little evidence that women who receive annual screening are at significantly lower risk for invasive cervical cancer than are women who are tested every 3-5 years. These findings were confirmed in a major study of eight cervical cancer screening programs in Europe and Canada involving over 1.8 million women. 81 According to this report, the cumulative incidence of invasive cervical cancer was reduced 64.1% when the interval between Pap tests was 10 years, 83.6% at 5 years, 90.8% at 3 years, 92.5% at 2 years, and 93.5% at 1 year. These estimates were for women aged 35-64 who had at least one screening before age 35, and they are based on the assumption of 100% compliance.Recommendations of Other Groups
A consensus recommendation that all women who are or have been sexually active, or who have reached age 18, should have annual Pap smears has been adopted by the American Cancer Society, National Cancer Institute, American College of Obstetricians and Gynecologists (ACOG), American Medical Association, American Academy of Family Physicians (AAFP), and others. 82 The recommendation permits Pap testing less frequently after three or more annual smears have been normal, at the discretion of the physician. Guidelines for determining frequency based on risk factors have been issued by ACOG. 83 The consensus did not recommend an age to discontinue Pap testing. The AAFP recommends that screening can be discontinued at age 65 if there is documented evidence of previously negative smears, but its recommendations are currently under review. 84 The American College of Physicians (ACP) recommends Pap smears every 3 years for women aged 20-65, and every 2 years for women at high risk. 85 The ACP also recommends screening women aged 66-75 every 3 years if not screened in the 10 years before age 66. The Canadian Task Force on the Periodic Health Examination recommends screening for cervical cancer with annual Pap smears in women following initiation of sexual activity or age 18, and after two normal smears, screening every 3 years to age 69. 86 The Canadian Task Force recommends considering more frequent screening for women at increased risk. In their guidelines for adolescent preventive services (GAPS), the American Medical Association recommends annual screening with a Pap test for female adolescents who are sexually active or age 18 or older. 87 Bright Futures also recommends annual Pap testing for sexually active adolescent females. 88 Similar recommendations have been endorsed by the American Academy of Pediatrics. 89Discussion
It has been estimated that screening women aged 20-64 every 3 years with Pap testing reduces cumulative incidence of invasive cervical cancer by 91%, requires about 15 tests per woman, and yields 96 cases for every 100,000 Pap smears. Annual screening reduces incidence by 93%, but requires 45 tests and yields only 33 cases for every 100,000 tests. 81 Empirical data also support the effectiveness of a 3-year interval. A study of 25,000 Dutch women found that screening a stable population every 3 years reduced the incidence of squamous cell carcinoma of the cervix from 0.38 per 1000 to zero within 12 years. 67 There are, in addition, important economic considerations to performing Pap tests every 2-3 years, since annual testing could double or triple the total number of smears taken on over 92 million American women at risk, 90 yet provide only limited added benefit in lowering mortality. 81
Annual testing, however, has been common. In the mid-1980s, a survey of recently trained gynecologists found that 97% recommend a Pap test at least once a year. 91 The preference of many clinicians for performing annual Pap smears is based on concerns that less frequent testing may result in more harm than good, but reliable scientific data to support these opinions are lacking. Specifically, advocates of annual testing have expressed concerns that data demonstrating little added value to annual testing are based on retrospective studies and mathematical models that are subject to biases and invalid assumptions that an interval longer than 1 year may permit aggressive, rapidly growing cancers to escape early detection that the public may obtain Pap smears at a lower frequency than that publicized in recommendations that a longer interval might affect compliance among high-risk women, a group with poor coverage even with an annual testing policy that repeated testing may offset the false-negative rate of the Pap smear that the test is inexpensive and safe and that a large proportion of women believe it is important to have an annual Pap test and, while visiting the clinician, may receive other preventive interventions. Definitive evidence to support these concerns is lacking.
Women who have never engaged in sexual intercourse are not at risk for cervical cancer and therefore do not require screening. 92-94 In addition, screening of women who have only recently become sexually active (e.g., adolescents) is likely to have low yield. The incidence of invasive cancer in women under age 25 is only about 1-3 per 100,000, a rate that is much lower than that of older age groups. 11 One study found that most women with CIN who had become sexually active at age 18 were not diagnosed with severe dysplasia or carcinoma in situ until age 30. 93
Although invasive cervical cancer is uncommon at young ages, authorities have recommended since the early 1980s that screening should begin with the onset of sexual activity. 82,92,94 This policy is based in part on the concern that a proportion of young women with CIN may have an aggressive cell type that can progress rapidly and go undetected if screening is delayed to a later age. There is some evidence that adenocarcinomas are accounting for a growing proportion of new cervical cancer cases in young women, 95,96 but the exact incidence and natural history of aggressive disease in young women remain uncertain. The Pap smear is also a poor screening test for adenocarcinoma, compared with squamous cell carcinoma. Another reason given for early screening is the concern that the incidence of cervical dysplasia occurring in young women appears to be on the rise, coincident with the increasing sexual activity of adolescents. On these grounds, testing should begin by age 18, since many American teenagers are sexually active by this age. Screening in the absence of a history of sexual intercourse may be justified if the credibility of the sexual history is in question.
When screening is initiated, it is frequently recommended that the first two to three smears be obtained 1 year apart as a means of detecting aggressive tumors at a young age. There is little evidence to suggest, however, that young women whose first two tests are separated by 2 or 3 years, rather than 1 year, have a greater mortality or person-years of life lost. 78 Recommendations to perform these first tests annually are based primarily on expert opinion.
Elderly women do not appear to benefit from Pap testing if repeated cervical smears have consistently been normal. 97,78 Modeling data suggest that continued testing of previously screened women reduces the risk of dying from cervical cancer by only 0.18% at age 65 and 0.06% at age 74. 80 Many older women have had incomplete screening, however. A reported 17% of women over age 65 and 32% of poor women in this age group have never received a Pap test. 98 In a study of elderly minority women with an average age of 75 years, the mean reported number of prior Pap smears received since age 65 was 1.7. 99 Further screening in this group of older women is important 78,100 and some studies suggest that it is cost-effective. 101 Women who have undergone a hysterectomy in which the cervix was removed do not benefit from Pap testing, unless it was performed because of cervical cancer. Post-hysterectomy screening has the potential to detect vaginal cancer, but the yield and predictive value are likely to be very low. Women who had hysterectomies performed in which the cervix was left behind probably still require screening.
The effectiveness of cervical cancer screening is more likely to be improved by extending testing to women who are not currently being screened and by improving the accuracy of Pap smears than by efforts to increase the frequency of testing. Studies suggest that those at greatest risk for cervical cancer are the very women least likely to have access to testing. 102,103 Incomplete Pap testing is most common among blacks, the poor, uninsured persons, the elderly, and persons living in rural areas. 98,104-106 In addition, many women who are tested receive inaccurate results due to interpretative or reporting errors by cytopathology laboratories or specimen collection errors by clinicians. The failure of some physicians to provide adequate follow-up for abnormal Pap smears is another source of delay in the management of cervical dysplasia. 107 Finally, a large proportion of patients with abnormal smears (30% in studies of poor, elderly black women 108 ) do not return for further evaluation. Various techniques may enhance physician and patient compliance with screening, follow-up of abnormal results, and patient compliance with rescreening. 109-112
CLINICAL INTERVENTION10. Screening for Prostate Cancer
Burden of Suffering
Prostate cancer is the most common noncutaneous cancer in American men. 1 After lung cancer, it accounts for more cancer deaths in men than any other single cancer site. Prostate cancer accounted for an estimated 244,000 new cases and 40,400 deaths in the U.S. in 1995. 1 Risk increases with age, beginning at age 50, and is also higher among African American men. Because it is more common in older men, prostate cancer ranks 21st among cancers in years of potential life lost. 2 The age-adjusted death rate from prostate cancer increased by over 20% between 1973 and 1991. 3 The lifetime risk of dying from prostate cancer is 3.4% for American men. 3 The reported incidence of prostate cancer has increased in recent years by 6% per year, a trend attributed to increased early detection efforts. 4 Because local extension beyond the capsule of the prostate rarely produces symptoms, about one to two thirds of patients already have local extracapsular extension or distant metastases at the time of diagnosis. 5 Ten-year survival rates are 75% when the cancer is confined to the prostate, 55% for those with regional extension, and 15% for those with distant metastases. 6 The potential morbidity associated with progression of prostate cancer is also substantial, including urinary tract obstruction, bone pain, and other sequelae of metastatic disease.Accuracy of Screening Tests
The principal screening tests for prostate cancer are the digital rectal examination (DRE), serum tumor markers (e.g., prostate-specific antigen [PSA]), and transrectal ultrasound (TRUS). The reference standard for these tests is pathologic confirmation of malignant disease in tissue obtained by biopsy or surgical resection. The sensitivity and specificity of screening tests for prostate cancer cannot be determined with certainty, however, because biopsies are generally not performed on patients with negative screening test results. False-negative results are unrecognized unless biopsies are performed for other reasons (e.g., abnormal results on another screening test, tissue obtained from transurethral prostatic resection). The resulting incomplete information about the number of true- and false-negative results makes it impossible to properly calculate sensitivity and specificity. Only the positive predictive value (PPV) -- the probability of cancer when the test is positive -- can be calculated with any confidence.
Even the PPV is subject to uncertainty because of the inaccuracies of the usual reference standard. Needle biopsy, the typical reference standard used for calculating sensitivity and specificity, has limited sensitivity. One study suggested that as many as 19% of patients with an initially negative needle biopsy (but abnormal screening test results) had evidence of cancer on a second biopsy. 7 Moreover, studies vary in the extent to which the gland is sampled during needle biopsy. Recent studies, in which larger numbers of samples are obtained from multiple sections of the gland, provide a different reference standard than the more limited needle biopsies performed in older studies. These methodologic problems account for the large variation in the reported sensitivity, specificity, and PPV of prostate cancer screening tests and the current controversy over their true values.
DRE is the oldest screening test for prostate cancer. Its sensitivity is limited, however, because the examining finger can palpate only the posterior and lateral aspects of the gland. Studies suggest that 25-35% of tumors occur in portions of the prostate not accessible to the examining finger. 8 In addition, Stage A tumors, by definition, are nonpalpable. Most recent studies report that DRE has a sensitivity of 55-68% in detecting prostate cancer in asymptomatic men, 9,10 but values as low as 18-22% have also been reported in studies using different screening protocols. 11,12 The DRE also has limited specificity, producing a large proportion of false-positive results. The reported PPV in asymptomatic men is 6-33% 10,13-15 but appears to be somewhat higher when performed by urologists rather than by general practitioners. 16
Elevations in certain serum tumor markers (e.g., PSA and prostatic acid phosphatase) provide another means of screening for prostate cancer. In screening studies, a PSA value greater than 4 ng/dL has a reported sensitivity of over 80% in detecting prostate cancer in asymptomatic men, 10 although a sensitivity as low as 29% has also been reported in studies using different screening protocols. 11 Prostatic acid phosphatase has a much lower sensitivity (12-20% for Stage A and B disease) and PPV (below 5%) than PSA, 17 and its role in screening has largely been replaced by PSA. PSA elevations are not specific for prostate cancer. Benign prostatic conditions such as hypertrophy and prostatitis can produce false-positive results about 25% of men with benign prostatic hypertrophy (BPH) and no malignancy have an elevated PSA level. 18
In most screening studies involving asymptomatic men, the reported PPV of PSA in detecting prostate cancer is 28-35%. 10,19-21 In many instances, however, other screening tests (e.g., DRE) are also positive. The PPV of PSA when DRE is negative appears to be about 20%. 22 It is unclear whether the same PPV applies when screening is performed in the general population. Participants in most screening studies are either patients seen in urology clinics or volunteers recruited from the community through advertising. Studies suggest that such volunteers have different characteristics than the general population. 23 For example, in one screening study, 53% of the volunteers had one or more symptoms of prostatism. 10 Since PPV is a function of the prevalence of disease, routine PSA testing of the general population, if it had a lower prevalence of prostate cancer than volunteers, would generate a higher proportion of false-positive results than has been reported in the literature. A significant difference in prevalence in the two populations has not, however, been demonstrated.
Several techniques have been proposed to enhance the specificity and PPV of the PSA test. The serum concentration of PSA appears to be influenced by tumor volume, and some investigators have suggested that PSA density (the PSA concentration divided by the gland volume as measured by TRUS) may help differentiate benign from malignant disease. 24-26 According to these studies, a PSA density greater than 0.15 ng/mL may be more predictive of cancer. Other studies suggest that the rate of change (PSA velocity), rather than the actual PSA level, is a better predictor of the presence of prostate cancer. An increase of 0.75 ng/mL or higher per year has a reported specificity of 90% and 100% in distinguishing prostate cancer from BPH and normal glands, respectively. 27 PSA values tend to increase with age, and investigators have therefore proposed age-adjusted PSA reference ranges. 28,29 Current evidence is inadequate to determine the relative superiority of any of these measures or to prove conclusively that any is superior to absolute values of PSA. 30 The most effective method to increase the PPV of PSA screening is to combine it with other screening tests. In a large screening study, the combination of an elevated PSA and abnormal DRE achieved a PPV of 49%. Even with this improved accuracy, however, combined DRE and PSA screening led to the performance of needle biopsies on 18% of the screened population, 10 raising important public policy issues (see below).
A large proportion of cancers detected by PSA screening may be latent cancers, indolent tumors that are unlikely to produce clinical symptoms or affect survival. Autopsy studies indicate that histologic evidence of prostate cancer is present in about 30% of men over age 50. The reported prevalence of prostate cancer in men without previously known prostate cancer during their lifetimes is 10-42% at age 50-59, 17-38% at age 60-69, 25-66% at age 70-79, and 18-100% at age 80 and older. 31-37 Recent autopsy studies have even found evidence of carcinoma in 30% of men aged 30-49. 38 Although patients who undergo autopsy may not be entirely representative of the general population, these prevalence rates, combined with census data, 39 suggest that millions of American men have prostate cancer. Fewer than 40,000 men in the U.S. die each year from prostate cancer, however, suggesting that only a subset of cancers in the population are clinically significant. Natural history studies indicate that most prostate cancers grow slowly over a period of many years. 40 Thus, many men with early prostate cancer (especially older men) will die of other causes (e.g., coronary artery disease) before their cancer becomes clinically apparent. Because a means of distinguishing definitively between indolent and progressive cancers is not yet available, widespread screening is likely to detect a large proportion of cancers whose effect on future morbidity and mortality is uncertain.
Recent screening studies have suggested, however, that cancers detected by PSA screening may be of greater clinical importance than latent cancers found on autopsy. Studies of asymptomatic patients with nonpalpable cancers detected through PSA screening have reported extracapsular extension, poorly differentiated cell types, tumor volumes exceeding 3 mL, and metastases in 31-38% of cancers that were pathologically staged. 20,41-43 In a retrospective review of radical prostatectomies performed on patients with nonpalpable prostate cancer detected by PSA screening, 65% had a volume greater than 1 mL, and surgical margins were positive in 26% of cases. 44 In a similar series, the mean tumor volume was 7.4 mL and 30% of the tumors had penetrated the capsule. 45
The sensitivity of PSA for clinically important cancers was examined in a recent nested case-control study among 22,000 healthy physicians participating in a long-term clinical trial. 46 Archived blood samples collected at enrollment were compared for 366 men who were diagnosed clinically with prostate cancer during a 10-year follow-up period and 1,098 matched controls without cancer. PSA was elevated (>4 ng/mL) in 46% of the men who subsequently developed prostate cancer and 9% of the control group (i.e., sensitivity 46%, specificity of 91%). For cancers diagnosed within the first 4 years of follow-up, the sensitivity of PSA was 87% for aggressive cancers but only 53% for nonaggressive cancers (i.e., small, well-differentiated tumors), suggesting that PSA is more sensitive for clinically important disease. Given the low incidence of aggressive prostate cancer in this study (1% over 10 years), the reported specificity of 91% would generate a PPV (10-15%) that is lower than that reported from studies using routine biopsies (28-35%). 10 Furthermore, this study could not address the central question of whether PSA would have identified aggressive cancers at a potentially curable stage.
TRUS is a third means of screening for prostate cancer, but its performance characteristics limit its usefulness as a screening test. In most studies, TRUS has a reported sensitivity of 57-68% in detecting prostate cancer in asymptomatic men. 9,10 Because TRUS cannot distinguish between benign and malignant nodules, its PPV is lower than PSA. Although a PPV as high as 31% has been reported for TRUS, 47 its reported PPV when other screening tests are normal is only 5-9%. 15,19 Even when cancers are detected, the size of tumors is often underestimated by TRUS. The discomfort and cost of the procedure further limit its role in screening.Effectiveness of Early Detection
There is currently no evidence that screening for prostate cancer results in reduced morbidity or mortality, in part because few studies have prospectively examined the health outcomes of screening. A case-control study found little evidence that DRE screening prevents metastatic disease the relative risk of metastatic prostate cancer for men with one or more screening DREs compared with men with none was 0.9 (95% confidence interval, 0.5-1.7). 48 A cohort study also reported little benefit from DRE screening, 49 but its methodologic design has been criticized. Randomized controlled trials of DRE and PSA screening, which are expected to provide more meaningful evidence than is currently available, are currently under way in the U.S. and Europe. 50 The results of these studies, however, will not be available for over a decade. Therefore, recommendations for the next 10 years will depend on indirect evidence for or against effectiveness.
Indirect evidence that early detection of prostate cancer improves outcome is limited. Survival appears to be longer for persons with early-stage disease 5-year survival is 87% for Stage A (nonpalpable) tumors, 81% for Stage B (palpable, organ-confined cancer), 64% for Stage C (local extracapsular penetration), and 30% for Stage D (metastatic). 5 Due to recent screening efforts, prostate cancer is now increasingly diagnosed at a less advanced stage. As with survival advantages observed with other cancers, however, it is not known to what extent lead-time and length biases account for differences in observed survival rates (see Chapter ii). The frequently indolent nature of prostate cancer makes length bias a particular problem in interpreting stage-specific survival data. Successful treatment of indolent tumors may give a false impression that "cure" was due to treatment. Prostate cancers detected through screening are more likely to be organ-confined than cancers detected by other means. 20 Proponents of radical prostatectomy often argue that such cancers are potentially curable by removing the gland. As already noted, however, current evidence is inadequate to determine with certainty whether these organ-confined tumors are destined to progress or affect longevity thus the need for treatment is often unclear.
Even if the need for treatment is accepted, the effectiveness of available treatments is unproven. Stage C and Stage D disease are often incurable, and the efficacy of treatment for Stage B prostate cancer is uncertain. Currently available evidence about the effectiveness of radical prostatectomy, radiation therapy, and hormonal treatment derives largely from case-series reports without internal controls, usually involving carefully selected patients and surrogate outcome measures for monitoring progression (e.g., PSA levels). 51-55 Although men treated for organ-confined prostate cancer have a normal life expectancy, it is not clear how much their prognosis owes to treatment. The only randomized controlled trial of prostate cancer treatment, which compared radical prostatectomy with expectant management, reported no difference in cumulative survival rates over 15 years, but the study was conducted in the 1970s and suffered from several design flaws. 56,57 Randomized controlled trials to evaluate the effectiveness of current therapies for early disease are being launched in the U.S. and Europe, but results are not expected for 10-15 years. 58,59
Some observational studies suggest that survival for early-stage prostate cancer may be good even without treatment. A Swedish population-based cohort study of men with early-stage, initially untreated prostate cancer found that, after 12.5 years, 10% had died of prostate cancer and 56% had died of other causes. The 10-year disease-specific survival rate (adjusted for deaths from other causes) for the study population was 85%. Cancer-related morbidity was significant, however. Over one third of the cancers progressed through regional extension, and 17% metastasized. The patient's age and the tumor stage did not significantly influence survival rates, but tumor grade (degree of differentiation) did affect survival the 5-year survival rate was only 29% for poorly differentiated tumors. 59-61 Critics of the study have argued that the high survival rates were due to the relatively large proportion of older men and of tumors detected incidentally during transurethral prostatic resection, and that Swedish data are not generalizable to the U.S. 22,62 Other studies have reported similar results in one series of selected men with well- and moderately differentiated cancer and extracapsular (nonmetastatic) extension, 5- and 9-year survival rates were 88% and 70%, respectively, without treatment. 63 Reported 10-year disease-specific survival for expectant management of palpable but clinically localized prostate cancer is 84-96%. 64-66 Finally, it is unclear whether reported survival rates in these studies, in which many cancers were detected without screening, are generalizable to screen-detected cancers.
Reviewers have attempted to compare the efficacy of treatment and watchful waiting by pooling the results of uncontrolled studies. An analysis of six studies concluded that conservative management of clinically localized prostate cancer (delayed hormone therapy but no surgical or radiation therapy) was associated with a 10-year disease-specific survival rate of 87% for men with well- or moderately differentiated tumors and 34% for poorly differentiated tumors. 67 The assumptions used in the model are not universally accepted, however. 68,69 A structured literature review concluded that the median annual rates of metastatic disease and prostate cancer mortality were 1.7% and 0.9%, respectively, without treatment. 70 This study was criticized for including a large proportion of patients with well-differentiated tumors and those receiving early androgen deprivation therapy. 71 Another review concluded that the annual rates for metastasis and mortality were higher (2.5% and 1.7%, respectively), but the review was limited to patients with palpable clinically localized cancers and excluded studies of cancers found incidentally at prostatectomy. In this population, disease-specific survival was estimated to be 83% for deferred treatment, 93% for radical prostatectomy, and 74% for external radiation therapy. 72 Thus, the effectiveness of treatment when compared with watchful waiting remains uncertain.
Uncertainties about the effectiveness of treatment are important because of its potentially serious complications. Needle biopsy, the diagnostic procedure performed on about 20% of men screened with DRE and PSA, 10 is generally safe but results in infection in 0.3-5% of patients, septicemia in 0.6% of patients, and significant bleeding in 0.1% of patients. 19,73-75 The potential adverse effects of radical prostatectomy are more substantial. Although urologists at specialized centers report operative mortality rates of 0.2-0.3%, 55,76 published rates in clinical studies and national databases range between 0.7% and 2%. 6,70,77-79 An examination of Medicare claims files estimated that the 30-day mortality rate was 0.5%.