The results of this process are summarized in Table 1. Because general internal medicine is so broad, many of the articles addressed in this Update will be discussed in more depth in the other papers in the 1998-1999 Update series.

Cancer Screening

Testing for prostate-specific antigen (PSA) is controversial but common. We still do not fully understand the benefits and risks of this practice. Because no prospective trials have assessed the benefits and harms of PSA testing, we must draw on indirect evidence.

One important component of our understanding is the natural history of prostate cancer. In Johansson and colleagues' 15-year cohort study [1], 223 of 642 men with prostate cancer had localized disease and were given no treatment initially. The 15-year survival rate among these 223 men was 81%; this rate did not differ from the rate among 77 men who did receive initial treatment. About two thirds of the tumors were well differentiated, and about half were diagnosed incidentally in men undergoing transurethral resections. The findings therefore may not be generalizable to men undergoing PSA testing. Nonetheless, Johansson and colleagues' study is the longest study of the natural history of prostate cancer, and it provides evidence that the disease course will be benign for many men.

Twenty-two percent of men with PSA levels of 4.0 to 10 ng/mL and 66% of men with levels greater than 10 ng/mL have biopsy-proven tumors. But is 4.0 ng/mL a reasonable cutoff with which to reassure men that they are cancer-free? In a community study [2], 914 asymptomatic men with PSA levels of 2.6 to 4.0 ng/mL agreed to undergo prostate biopsy. Biopsy findings were positive in 22% of men, and 83% of the tumors were moderately or poorly differentiated or were larger than 0.5 mL. These histologic characteristics contrast markedly with those reported in Johannson and colleagues' study. Therefore, the implications of PSA levels between 2.6 and 4.0 ng/mL seem to be similar to those of levels between 4.0 and 10 ng/mL.

The confusion and controversy over PSA screening prompted the American College of Physicians (ACP) to publish a series of three articles in 1997. These guidelines underscore the beliefs that PSA screening is the most sensitive noninvasive test with which to detect prostate cancer (Table 2) and that tumors detected by this testing are clinically significant [3]. However, both digital rectal examination and PSA testing have poor specificity. Thus, if a one-time PSA test and digital rectal examination were used to screen all men older than 50 years of age, 25% of tested men would require biopsies to follow up positive results on these tests. If the results of either PSA screening or digital rectal examination are abnormal, the positive predictive value ranges from 15% to 21%. If results of both tests are abnormal, the positive predictive value ranges from 38% to 50%. In addition, no trial has shown that PSA screening prolongs life.

The second ACP article on PSA screening was a cost-effectiveness analysis [4]. On the basis of favorable assumptions of the benefits of radical prostatectomy for local disease, this analysis showed that one-time screening with digital rectal examination and PSA may be beneficial in younger men, but only marginally. No benefit could be shown for screening in men 70 years of age and older.

The final ACP article reviewed the screening controversy [5]. Factors that favor screening include the facts that 40 000 men die of prostate cancer each year, that PSA screening detects tumors before they are palpable, and that tumors associated with elevated PSA levels seem to be more histologically aggressive. Critics note that prostate cancer is often indolent and that many men will die with prostate cancer, not of it. Treatment carries great morbidity. Prostate cancer is the second most common cause of cancer-related death in men but ranks 21st in terms of years of life lost. Finally, PSA testing is not highly accurate; it results in many false-positive findings and unnecessary biopsies. Given the controversy about PSA and its relative benefits and risks, the ACP made a strong recommendation that it is inappropriate for physicians to perform routine PSA measurement without frankly discussing these issues with patients [5].

Screening for colon cancer has become less controversial, but many clinical questions remain in day-to-day practice. First, what workup is indicated for patients who, during the review of symptoms, mention rectal bleeding? Second, how should one proceed with patients who are found to have small polyps on routine flexible sigmoidoscopy?

In a cohort study done to answer the first question, patients (mean age, 55 years) visiting a veterans primary care clinic completed an eight-question systems review [6]. Of the 297 patients who mentioned visible blood on toilet paper or stool, 201 underwent double-contrast barium enema and rigid sigmoidoscopy and were followed for up to 10 years (51 patients did not return for testing, 23 had sigmoidoscopy only, 1 had barium enema only, and 21 did not return for a follow-up interview). Serious disease was found in 24% of patients: Six and a half percent had cancer, 13% had polyps, and 4% had inflammatory bowel disease. Pain or other symptoms were not helpful indicators of disease. Although ascertainment bias could have contributed to these strikingly high rates of abnormality, other investigators have found substantial numbers of important lesions even in asymptomatic persons. Such a high likelihood of abnormality should prompt physicians to ask, in some way, about blood in stools and to perform complete colonic investigations when patients mention this symptom.

The American College of Gastroenterology suggests that patients with adenomatous polyps larger than 1 cm on flexible sigmoidoscopy should have colonoscopy. For smaller polyps, it recommends that patients and physicians individualize follow-up decisions. Its cohort study followed 137 asymptomatic patients who were found to have adenomas of 5 mm or less on screening sigmoidoscopy [7]. On colonoscopy, proximal neoplasms were discovered in 29% of these patients. These results imply that patients with diminutive adenomas found on sigmoidoscopy should undergo colonoscopy.

Venous Thromboembolic Disease

Clinically important venous thromboembolic events occur in as many as 4 in 1000 persons each year. In 1997, important questions about venous thromboembolic disease that were addressed included the optimal duration of treatment in patients with a second episode of venous thromboembolism, the use of low-molecular-weight heparin in pulmonary embolism, and the usefulness of D-dimer testing in ruling out pulmonary embolism.

Two large studies of patients with a first episode of deep venous thrombosis had shown that 6 months of oral anticoagulation is better than 4 to 6 weeks of anticoagulation. However, the appropriate duration of anticoagulation after a second event had not been established. In an unblinded randomized, controlled trial, 227 patients with second episodes of acute deep venous thrombosis or pulmonary embolism were initially treated with heparin for at least 5 days, followed by warfarin [8]. Patients were then assigned to receive warfarin for 6 months or indefinitely and were followed for 4 years. Recurrent venous thromboembolism occurred in 21% of patients treated for 6 months and in less than 1% of patients treated indefinitely (number needed to treat [NNT] to prevent one recurrence, 6). There was a trend toward fewer major hemorrhages in the 6-month group (number needed to harm [NNH] to cause one hemorrhage, 17). This study clearly establishes that the benefits of indefinite therapy after a second episode of venous thromboembolism outweigh the risks.

Several studies support the use of low-molecular-weight heparin to treat patients with deep venous thrombosis. Two studies now support the use of this drug in patients with pulmonary embolism as well. In an unblinded trial, 612 inpatients with pulmonary embolism were randomly assigned to receive either a daily fixed dose of the low-molecular-weight heparin tinzaparin, given subcutaneously, or continuous intravenous infusion of unfractionated heparin [9]. All patients then received warfarin for at least 3 months. Recurrence, bleeding, and mortality rates did not differ between the groups. Because the rates of events were low, however, the study may have lacked the power to show a difference.

In a second trial [10], 1021 patients with venous thromboembolic disease were randomly assigned to receive 1) reviparin [another low-molecular-weight heparin], given subcutaneously as a fixed dose twice daily, or another low-molecular-weight heparin or 2) continuous infusion of unfractionated heparin. Both groups then received warfarin for 12 weeks. Most of the patients in the study had deep venous thrombosis. Twenty-five percent of the patients had pulmonary embolism. Once again, clinical outcomes did not differ between groups. A subgroup analysis of the patients with pulmonary embolism showed no difference in outcomes between the two treatment groups. Taken together, these studies suggest that low-molecular-weight heparin is a safe and effective alternative to continuous intravenous heparin in the treatment of pulmonary embolism. It was not superior to standard therapy and has not yet been approved by the U.S. Food and Drug Administration for this indication. However, the potential benefit of reduced hospital costs makes it a promising treatment.

The Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED) [11] established that ventilation-perfusion scanning often cannot rule out the diagnosis of pulmonary embolism. In such cases, if deep venous thrombosis cannot be diagnosed, patients must undergo invasive testing. Testing for D-dimer, a degradation product of cross-linked fibrin, has been proposed as a noninvasive aid in the diagnosis of venous thromboembolic disease. In a prospective study [12], researchers performed D-dimer tests in 671 consecutive patients presenting to a Swiss emergency department who were suspected of having pulmonary embolism. All patients underwent ventilation-perfusion scanning and compression ultrasonography of the legs. Patients with inconclusive test results had pulmonary angiography. Patients thought not to have pulmonary emboli were followed for 3 months. A D-dimer cutoff level of 500 µg/L provided a sensitivity of 99% and a specificity of 41%. Most important, in this group of patients suspected of having pulmonary embolism, the D-dimer assay had a negative predictive value of 99%. These data suggest that if this enzyme-linked immunosorbent D-dimer assay is reproducible and readily available, it may be particularly helpful in ruling out pulmonary embolism. However, its low specificity makes it a poor test for ruling in the diagnosis.

Stroke

Aspirin is well-established as an agent for the prevention of stroke, but its role in the treatment of ischemic stroke has not been defined. The International Stroke Trial was an unblinded randomized trial of almost 20 000 patients who were randomly assigned to one of six groups: aspirin (300 mg/d), high-dose heparin (12 500 U twice daily), low-dose heparin (5000 U twice daily), two combination groups, or no treatment [13]. Treatment was continued for 14 days or until hospital discharge, and outcomes were evaluated at 14 days and 6 months. Patients treated with aspirin had fewer recurrent ischemic strokes (NNT, 91) and a 1% reduction in the combined end point of death or stroke at 14 days (NNT, 91). The incidence of bleeding (all extracranial episodes) slightly increased (NNH, 200). An important finding was that the number of patients who had died or become dependent (moderately or severely disabled) by 6 months was significantly reduced (NNT, 43). Low-dose heparin reduced the rate of stroke or death at 14 days but not at 6 months, and high-dose heparin increased the mortality rate because of an increase in hemorrhagic strokes. This study, along with a large Chinese study that reported similar results [14], provides adequate evidence for the use of aspirin for secondary, as well as primary, prevention of ischemic strokes.

Preventing stroke is clearly superior to treating it. Although decreasing cholesterol levels is well known to reduce the rate of myocardial infarction, its role in preventing strokes has not been established. In a meta-analysis that addressed this question, Hebert and colleagues [15] reviewed 16 clinical trials (about 29 000 patients) of primary or secondary prevention of coronary disease with 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (the "statin" drugs). Average follow-up was 3.3 years. The researchers found a 29% relative reduction in the rate of all strokes (NNT, 95) and a 22% relative reduction in total mortality (NNT, 45). However, no difference was seen in the rate of fatal strokes. The reduction in mortality rates was due to the occurrence of fewer coronary artery events.

Asthma

Inhaled steroids are now standard therapy for all but the mildest forms of asthma. Doses are generally increased in response to worsening symptoms. Two studies in 1997 examined alternative approaches to steroids. In a study that assessed theophylline's role as an adjunct to inhaled steroids in patients with asthma, Evans and colleagues [16] randomly assigned 62 patients with moderate asthma to receive either budesonide (800 µg twice daily) plus placebo or budesonide (400 µg twice daily) plus theophylline (200 to 375 mg twice daily, based on body weight). After 12 weeks, the theophylline group had greater improvement: The FEV₁ had increased by 8.1% (absolute increase) compared with 4.8% in the control group. Both groups showed similar reductions in symptoms and use of albuterol. Serum cortisol levels were lower in the theophylline group, and the mean theophylline level was only 8.7 µg/mL. Whether the effect would be maintained beyond 12 weeks remains unclear.

Another clinical trial compared inhaled steroids alone with inhaled steroids plus a long-acting ß-agonist [17]. After 3 months of treatment with budesonide (800 µg twice daily), 852 patients were randomly assigned to one of four treatment regimens and followed for 1 year. The low-dose budesonide group was the comparison group for the other three regimens. Results are shown in Table 3. A key clinical question that was not explicitly answered was how high-dose budesonide alone performed in comparison with high-dose budesonide plus formoterol. Adding formoterol did seem to improve outcomes, however, and use of the drug was not associated with any major risks.

Hormone Replacement

Estrogens prevent both heart disease and osteoporosis in postmenopausal women, but some physicians and patients remain concerned about whether this benefit is outweighed by the increased risk for breast cancer. To help resolve this question, Grodstein and colleagues [18] used the cohort of the Nurses' Health Study to assess the effect of hormone replacement therapy (HRT) on mortality in a case–control study that followed the participants from 1976 to 1994. After adjustment for confounding variables, current hormone users had a 37% lower risk for death than women who had never used HRT. The most marked reduction in mortality rates was due to coronary artery disease (relative risk reduction, 53%), but the rate of death from cancer was also lower in the current hormone users (relative risk reduction, 29%). Women with coronary risk factors had the largest reductions in mortality. Benefits of hormone use decreased with long-term use: The relative risk reduction in death from all causes at 10 years was 20% in women who currently used HRT and had done so for more than 10 years compared with women who had never used HRT. This loss of protective effect was due to increasing mortality from breast cancer. Women who had stopped using HRT within the previous 3 years still had a lower risk for death, but no benefit was seen 5 years after cessation of HRT. Thus, even when an increased risk for breast cancer is taken into account, the net benefit of HRT is significant, particularly in women with coronary risk factors.

This conclusion was supported by another study that assessed the risk for breast cancer in women taking HRT who had first-degree relatives with breast cancer [19]. Investigators identified almost 42 000 such women who had been receiving HRT for at least 5 years and followed them for 8 years. They found no increased risk for breast cancer in either former or current users of HRT with a family history of breast cancer compared with women who had never used HRT, although the CIs were wide. No increased risk for breast cancer was found in former or current users of HRT with no family history of breast cancer, and the accompanying CIs were narrower. The decreased overall mortality rate in the women who used HRT supports the results of the Nurses' Health Study.

Coronary Artery Disease

Many general internists are consulted for preoperative care and often see patients who have or are at risk for coronary artery disease. Cardiac-related postoperative death usually occurs in the first week after surgery, but patients can be at risk for as long as 2 years after surgery. Mangano and colleagues [20] sought to determine whether the use of ß-blockers would decrease mortality rates in such circumstances. In a clinical trial, they assigned 200 patients to receive atenolol (50 to 100 mg/d) or placebo before surgery and during the hospital stay. Patients were veterans who had coronary artery disease or at least two risk factors for coronary artery disease. All were undergoing elective noncardiac surgery. The findings show a striking mortality benefit with the use of atenolol: The 2-year mortality rate was 9% in the treatment group and 21% in the placebo group (NNT, 8). The primary benefit was due to decreased cardiac mortality rates in the first 6 to 8 months after surgery. Benefits were greatest in diabetic patients.

Homocysteine is associated with venous and arterial thrombosis. The specific association between elevated homocysteine levels and coronary artery disease has been controversial. In a cohort study, Norwegian researchers measured plasma homocysteine levels in 587 patients with known coronary artery disease and followed the patients for 4.6 years [21]. The association between homocysteine level and death is summarized in Table 4. Because of this strong association, clinical trials should be done to assess the efficacy of drug therapy (for example, folic acid treatment) in decreasing homocysteine levels. Whether a physician should recommend folate supplementation is unclear, but doing so may become common practice.

Hypertension

Most of the numerous studies on hypertension published in 1997 confirmed conventional beliefs. For example, the sixth report of the Joint National Commission reiterated the results of clinical trials showing that for most patients, thiazide diuretics and ß-blockers are the optimal drugs for prolonging life [22]. Unfortunately, the report also indicated that blood pressure control in the general U.S. population has slipped in the past few years. More data from the Systolic Hypertension in the Elderly (SHEP) trial showed that treating isolated systolic hypertension in the elderly decreases mortality from congestive heart failure; these findings augment existing knowledge that treating hypertension can prevent stroke and reduce overall mortality [23, 24].

Ambulatory blood pressure monitoring allows frequent measurement of blood pressure levels by use of a cuff and a portable device similar to an audiotape recorder. It is not used very often because we have not known how to interpret the results. In a community-based clinical trial of 419 patients labeled hypertensive, Belgian researchers sought to determine the effects on outpatient practice of measuring blood conventionally or with an ambulatory monitor [25]. After 6 months, 26% of the patients in the ambulatory monitoring group had discontinued therapy with at least one medication compared with 7% of patients in the conventional group. In addition, patients with ambulatory monitors required fewer drugs but had superior blood pressure control. Left ventricular mass and symptoms were similar in the two groups. The cost of medications and physicians' fees were greater in the conventional blood pressure monitoring group, but the cost of ambulatory blood pressure monitoring counterbalanced those costs. This study suggests that if the cost of ambulatory blood pressure monitoring can be reduced, such monitoring may become the preferred method of measurement.

Congestive Heart Failure

Like theophylline, digoxin is an old drug whose perceived efficacy has waxed and waned. In a clinical trial of 6800 patients, the Digitalis Investigation Group [26] evaluated the effects of digoxin on mortality and hospitalization for heart failure. All patients had chronic heart failure (left ventricular ejection fraction 0.45) and were in sinus rhythm. Patients were followed for 37 months. Neither all-cause nor cardiovascular mortality differed between patients taking digoxin and those taking placebo. However, digoxin recipients were hospitalized less frequently (NNT, 36 for all hospitalizations and 13 for hospitalization related to heart failure). Digoxin had more benefit in patients with cardiomegaly, lower ejection fraction, or New York Heart Association class III or IV heart failure. Although this study supports the use of digoxin, it does not support its use over that of agents known to reduce mortality, such as angiotensin-converting enzyme inhibitors.

Many patients cannot tolerate angiotensin-converting enzyme inhibitors, usually because of cough (which occurs in about 15% of patients). Angiotensin II receptor antagonists, such as losartan, are a new class of drugs that rarely cause cough but do not have an established role in the treatment of heart failure. In a study that compared the safety and efficacy of losartan with those of captopril in patients older than 65 years of age with heart failure, Pitt and colleagues [27] randomly assigned 722 patients with ejection fractions less than 0.40 to receive losartan (50 mg/d) or captopril (50 mg three times daily). After 48 weeks, the groups did not differ for the primary end point, which was an increase in creatinine level of at least 0.3 mg/dL. All-cause mortality dramatically decreased with losartan (4.8%) compared with captopril (8.7%); the 95% CIs were very wide, however, indicating that this finding lacks precision and certainty. We therefore believe it is premature to change the standard approach to heart failure on the basis of this study alone. However, losartan is an alternative for patients who cannot tolerate angiotensin-converting enzyme inhibitors because of cough.

Alzheimer Disease

As the general population ages, Alzheimer disease is becoming more common; it now affects 24% of persons older than 85 years of age. Unfortunately, drug treatment has not been effective in altering the course of the disease. In 1997, three studies evaluated new treatment options.

In the first study [28], 341 community-dwelling patients with moderate Alzheimer disease were randomly assigned to receive the monoamine oxidase inhibitor selegiline (10 mg/d), vitamin E (1000 IU twice daily), both, or placebo. The primary outcomes were times to death, institutionalization, loss of the ability to perform two or three basic activities of daily living, or development of severe dementia. After 2 years, times to these end points did not differ in unadjusted comparisons. When the results were adjusted for differences in baseline Mini-Mental Status Examination score, significant reductions were seen in several outcome events (Table 5). Falls and syncope both occurred more often in all treatment groups. Because the groups differed somewhat after randomization and because selegiline requires careful monitoring, we would not recommend using selegiline. Vitamin E offered modest clinical benefits, however, and there may be little harm in trying it as long as patient and family expectations for its efficacy are not too high.

Donepezil and Ginkgo biloba have also been investigated for Alzheimer disease. Donepezil was recently approved by the U.S. Food and Drug Administration for use in Alzheimer disease, partly on the basis of a 12-week clinical trial [29]. In another study of ginkgo biloba, a 52-week clinical trial, more than 50% of the patients dropped out [30]. A primary end point in both trials was the cognitive assessment subscale of the Alzheimer disease assessment scale. Donepezil caused a mean improvement of 9% on this scale, and Ginkgo biloba caused a mean improvement of 7.5%. We believe that the clinical significance of these changes is negligible, and we would not recommend using these drugs on the basis of these reports alone.

Osteoporosis

Much has been written about how calcium, estrogen, and bisphosphonate drugs affect bone mineral density, but many questions remain. One is whether calcium and vitamin D actually help healthy elderly persons. In a clinical trial, 400 healthy men and women at least 65 years of age were randomly assigned to receive either calcium (500 mg/d) and vitamin D₃ (700 IU/d) or placebo [31]. The participants were followed for 3 years. After 1 year, the treatment group had less bone loss than the placebo group. For example, femoral bone mineral density increased by 0.5% in the treatment group and decreased by 0.7% in the placebo group. In the second and third years of the study, this difference was not as marked. However, nonvertebral fractures occurred in 12.9% of the placebo group compared with 5.9% of the treatment group (NNT to prevent one fracture, 14). Thus, calcium and vitamin D seem to be beneficial in healthy older persons.

Calcium is a beneficial but often insufficient treatment for persons who already have osteoporosis. Bisphosphonates have been proposed for such patients. A randomized trial reported a 25% relative risk reduction in clinically important fractures over 3 years among postmenopausal women taking alendronate (absolute risk reduction, 4.6%; NNT, 22) [32]. The next logical question is whether these drugs can prevent osteoporosis among persons at high risk for developing it. One such group is patients receiving long-term corticosteroid therapy. In a clinical trial, 141 patients receiving high-dose corticosteroid therapy for various chronic illnesses were randomly assigned to receive etidronate or placebo for 14 days, followed by calcium supplementation for 76 days [33]. This cycle was repeated three times during the 1-year study. After 1 year, bone mineral density was assessed. Patients receiving treatment had a 0.6% increase in lumbar bone mineral density; placebo recipients had a 3.2% decrease. New vertebral fractures were less common in the treatment group (NNT, 5), and all vertebral fractures occurred in postmenopausal women who had lower density at baseline than the men. Although this study did not examine long-term outcomes, it does provide evidence that using etidronate and calcium to prevent corticosteroid-induced osteoporosis is useful, especially in postmenopausal women.

Raloxifene, a selective estrogen receptor modulator, is marketed as a drug that can prevent osteoporosis without causing the potential breast and uterine cancer associated with estrogen use. The effects of raloxifene were reported in a clinical trial of 601 postmenopausal women who were randomly assigned to receive one of three doses of raloxifene or placebo and were followed for 2 years [34]. Raloxifene was associated with a 6% decrease in cholesterol levels and a 1.5% increase in bone mineral density; it was not associated with endometrial hyperplasia, as assessed by ultrasonography. Raloxifene does not alleviate hot flashes, and its long-term clinical effects, including its impact on heart disease and fracture prevention, have not been fully assessed. As a result, we believe that for most patients, it is too early to prescribe raloxifene instead of estrogen.

HIV Disease

Because knowledge and treatment of HIV infection continue to advance rapidly, publication delays are a major problem for nonexperts. However, a few advances in 1997 are important even for physicians who see HIV-infected patients only occasionally.

First, a 10-year multicenter study of disease markers [35] reinforced the idea that plasma viral load is the single best predictor of the progression of HIV disease, although other measures, such as CD4 counts, are also independent predictors of outcome. In fact, many experts believe that future treatment trials will use viral load, not mortality rates, as the primary outcome measure of effectiveness. Second, several studies have underscored the point that the optimal therapeutic approach against HIV infection currently comprises two nucleoside analogues plus one protease inhibitor [26-37]. Third, despite early hopes that regimens using protease inhibitors might result in a "cure" of HIV disease, follow-up studies have shown that special techniques can detect latent virus in almost all patients, even those with undetectable plasma RNA levels [38, 39].

Chemical Dependency and Violence

Alcohol dependence, smoking, and domestic violence are all common in primary care practice, and each presents considerable problems in both diagnosis and treatment. One important study [40] highlighted the prevalence of alcohol abuse in a primary care population and addresses steps the general internist can take to reduce alcohol use. This large clinical trial screened more than 17 000 primary care patients. After all participants had completed a health screening survey, 14% were found to meet criteria for excessive drinking (>14 drinks/week for men and >7 drinks/week for women). These patients were randomly assigned to one of two groups. Patients in both groups received brochures about general health, and those in the experimental group also had two 15-minute visits with their own physicians to discuss ways to reduce alcohol intake. After 12 months, self-reported results showed that 30% of the experimental group decreased drinking compared with 15% of the control group.

The antidepressant bupropion is now marketed for use as an aid to smoking cessation. In a clinical trial of 615 smokers, bupropion at three dosing regimens was compared with placebo [41]. Brief counseling was also provided to all patients weekly for 8 weeks, then three more times during this 1-year study. After 7 weeks of treatment, the proportion of patients who had quit smoking, as measured by carbon monoxide levels, was 10% in the placebo group and 29% to 44% in the bupropion groups; higher doses of bupropion produced greater effects. At 1 year, the rates were 12% and 20% to 23%, respectively. This study indicates that bupropion may be a helpful adjunct to counseling patients who are motivated to quit smoking.

Finally, domestic violence is another prevalent problem, affecting 8% to 12% of women each year. Detection is difficult. This prospective study, conducted in an emergency department, compared the accuracy of three screening questions for detecting domestic violence [42]. According to previously validated surveys, the prevalence of domestic violence was 30%. Domestic violence was the cause of the visit in 14% of women. The most accurate question was "Have you ever been hit, kicked, punched, or otherwise hurt by someone in the past year? If so, by whom?" The sensitivity of this question was 68%, and the specificity was 95%. Because domestic violence is so common in the general population, we believe that this question should be used as a screening question in the office setting.

Dr. Roberts (Series Editor): Madrona Medical Group, 3199 Steller Court, Bellingham, WA 98226-7805.