Background and Purpose The value of screening for asymptomatic carotid stenosis has become an important issue with the recently reported beneficial effect of endarterectomy. The purpose of this study is to evaluate the cost-effectiveness of using Doppler ultrasound as a screening tool to select subjects for arteriography and subsequent surgery.
Methods A computer model was developed to simulate the cost-effectiveness of screening a cohort of 1000 men during a 20-year period. The primary outcome measure was incremental present-value dollar expenditures for screening and treatment per incremental present-value quality-adjusted life-year (QALY) saved. Estimates of disease prevalence and arteriographic and surgical complication rates were obtained from the literature. Probabilities of stroke and death with surgical and medical treatment were obtained from published clinical trials. Doppler ultrasound sensitivity and specificity were obtained through review of local experience. Estimates of costs were obtained from local Medicare reimbursement data.
Results A one-time screening program of a population with a high prevalence (20%) of ≥60% stenosis cost $35 130 per incremental QALY gained. Decreased surgical benefit or increased annual discount rate was detrimental, resulting in lost QALYs. Annual screening cost $457 773 per incremental QALY gained. In a low-prevalence (4%) population, one-time screening cost $52 588 per QALY gained, while annual screening was detrimental.
Conclusions The cost-effectiveness of a one-time screening program for an asymptomatic population with a high prevalence of carotid stenosis may be cost-effective. Annual screening is detrimental. The most sensitive variables in this simulation model were long-term stroke risk reduction after surgery and annual discount rate for accumulated costs and QALYs.
With the benefit of endarterectomy for patients with asymptomatic carotid stenosis of ≥60% demonstrated by the recent ACAS, the issue of screening for the presence of clinically occult disease has become important.1 Many factors will affect the usefulness of screening an asymptomatic population. The benefit of screening, in terms of strokes prevented, will depend on the prevalence of disease in the population under study, the sensitivity and specificity of the screening tool, the complication rate of arteriography, the complication rate of surgery, and the benefit that endarterectomy confers. In addition, the costs of diagnosis and treatment must be considered.
The optimal method of identifying patients for endarterectomy is a subject of some controversy. Arteriography is the most accurate technique but is invasive and expensive. MRA is an evolving technology that continues to improve but currently remains less accurate than angiography.2 In part because of its evolving nature, there are no large studies documenting its sensitivity and specificity relative to arteriography. Doppler ultrasound is inexpensive and noninvasive, but its accuracy can vary considerably between institutions.3 In addition, it does not image the carotid beyond its cervical segment. There is currently no proven effective medical or surgical therapy for intracranial disease, however. Therefore, if Doppler ultrasound can reliably exclude surgically amenable extracranial disease, it could function as an effective screening tool by reducing the number of patients undergoing arteriography.
The purpose of this analysis was to evaluate the cost-effectiveness of this approach with the aid of a simulation model to calculate the incremental cost per incremental QALY gained by screening a cohort of 1000 asymptomatic men. The model was designed to represent events occurring in the cohort during a 20-year period, simulating the effect of screening and treating this population. Simulation models allow testing of the effects of the assumed input variables over a range of values as well as the effects of different model structures on the ultimate cost-effectiveness of screening (sensitivity testing).
Analyses such as these may become useful in identifying the most cost-effective approaches to the diagnosis and treatment of specific illnesses. In addition, sensitivity testing is important in identifying the critical variables that have the greatest impact on the outcome of the model.
Subjects and Methods
We used a computer spreadsheet program (Microsoft Excel 4.0, Microsoft Corp) to simulate possible screening programs for a cohort of 1000 men, beginning at age 60 years. Two different patient populations were evaluated: a population with a high prevalence of ≥60% stenosis and risk factors such as prior myocardial infarction, bruit, or peripheral vascular disease, and a population with a low prevalence of ≥60% stenosis, more representative of the general population. The cohort was assumed to be unscreened until the first year of the simulation.
On day 1 of the first year, all subjects undergo a screening Doppler ultrasound examination. If the Doppler examination is positive for significant stenosis (either false-positive or true-positive), that patient goes on to arteriography on day 2. Patients diagnosed with internal carotid artery occlusion by Doppler ultrasound do not undergo arteriography. The practice of pursuing arteriography to confirm this diagnosis in asymptomatic patients causes more strokes than it prevents.4 The number of patients undergoing arteriography is a function of the prevalence of disease and the sensitivity and specificity of Doppler ultrasound for stenosis and occlusion. Stenoses and occlusions are assumed to be unilateral in this analysis. Therefore, false-positive examinations of the contralateral normal vessel will add to the number of arteriograms to be performed. Not all patients with contralateral Doppler ultrasound false-positive examinations will be subjected to added arteriographic risk because a portion of these will undergo arteriography for the ipsilateral vessel. If the arteriogram indicates ≥60% stenosis, the patient undergoes surgery on day 3. All patients are considered candidates for endarterectomy, and all are assumed to agree to the procedure. All arteriographic and surgical strokes and deaths occur on the day of the procedure. On day 365, the patients who will progress from <60% to ≥60% stenosis during the ensuing year are crossed over. All other strokes and deaths for the year occur on this day as well.
The number of strokes and deaths accumulated over the first year will depend on the number of patients in each stenosis group. The number of patients with ≥60% stenosis will include patients who had false-negative examinations for stenosis and false-positive examinations for occlusion (Doppler diagnosis of occlusion but in reality ≥60% stenosis). The number of patients with <60% stenosis will include postsurgical patients (after their perioperative risk is incurred) and the original group of patients with <60% stenosis.
On the first day of the next year the process repeats for the annual screening model. The number of patients with ≥60% stenosis will include patients who developed ≥60% stenosis the previous year (crossover rate) and those who were missed by the Doppler ultrasound screen. No further screening or surgical treatment occurs in the one-time screening model. Strokes and deaths continue to accumulate, however, and reduce the number of patients in all treatment groups according to their group- and age-specific mortality rates. Postoperative patients undergo annual Doppler examinations and clinic visits in both the annual and one-time screening models.
All costs and QALYs are discounted at a 3% annual rate.5 The incremental present-value costs per incremental present-value QALY were calculated for each screening model by subtracting the cumulative present-value costs and QALYs generated from a natural history simulation.
Data and Assumptions
Prevalence of Disease
Estimates of the disease prevalence of carotid stenosis were obtained through a review of the literature.6 7 8 9 10 11 A 20% prevalence of ≥60% stenosis and a 2.5% prevalence of occlusion were used for the population with a high prevalence of ≥60% asymptomatic stenosis. This value was determined primarily from the study of Hennerici et al.6 A 4% prevalence of ≥60% stenosis was assumed for a population with a low prevalence of ≥60% asymptomatic stenosis on the basis of the data of Hennerici and coworkers and Pujia et al.7 The prevalence of occlusion in this population was assumed to be 0.25%. A crossover rate from <60% to ≥60% stenosis was assumed to be 0.5% for the high-prevalence population and 0.01% for the low-prevalence population on the basis of a longitudinal study of 175 hypertensive asymptomatic patients by Luisiani et al.8 In this study 2.9% (5/175) of patients with <50% stenosis by ultrasound developed >50% stenosis at 3 years.8
Doppler Sensitivity and Specificity
To generate the local institutional Doppler ultrasound values of sensitivity and specificity, we reviewed the examinations of 424 carotid bifurcations in 215 patients evaluated with both Doppler ultrasound and transfemoral selective carotid arteriography at Barnes Hospital during a 2-year period from January 1993 through December 1994. The methodology and results of this review have been previously published.4 The patient population included both symptomatic and asymptomatic patients who could not be reliably separated retrospectively. Patients were excluded if arteriography was performed >1 month before or after the ultrasound examination or if there was interval surgery. Doppler examinations were performed according to accepted methodology. Arteriographic measurement followed NASCET criteria.12 13 The specificity and sensitivity of ultrasound when we used both Doppler ultrasound and color flow imaging for the diagnosis of carotid occlusion were 99.7% (95% confidence limit, ±0.5%) and 97.8% (95% confidence limit, ±4.2%).
Angiographic Rate of Stroke and Death
The base case angiographic complication rate used for the simulations was 1.2%, the rate of permanent neurological deficit or death reported in the ACAS.1 This rate is slightly higher than that commonly reported but may be more accurate for several reasons. First, the ACAS rate was observed in patients with asymptomatic carotid artery disease, while most reported series include many patients without carotid atherosclerosis. Second, all patients in the ACAS received neurological examinations after arteriography. Third, the ACAS data reflect a multi-institutional experience.
Strokes and deaths incurred by arteriography and surgery as well as strokes occurring after the perioperative period were accumulated separately in the simulations. The 30-day perioperative rate of stroke or death of 2.3% reported in the ACAS included both preoperative events (including arteriographic complications) and postoperative events. When preoperative and arteriographic strokes and deaths were excluded, the actual 30-day postoperative stroke and death rate observed in the ACAS was 11 of 724 (1.5%). This value was used as the base case variable.
The 5-year projected risk reported in the ACAS for ipsilateral stroke and perioperative stroke or death for asymptomatic patients with ≥60% stenosis after carotid endarterectomy was 5.1%.1 The 5.1% value reported for the surgical cohort incorporated the perioperative stroke and death rate of 2.3%. Therefore, after arteriography and surgery, the 5-year risk in the surgical subgroup was 2.8% (5.1% less 2.3% arteriographic and perioperative events). This becomes an annual rate of stroke of 0.56 after arteriography and surgery.
The benefit of surgery was assumed to be permanent. Patients after endarterectomy were assumed to develop restenosis at the same rate that patients with <60% stenosis develop ≥60% stenosis.
Risk of Stroke With Medical Treatment
The patients with ≥60% stenosis missed by Doppler ultrasound (false-negative for stenosis or false-positive for occlusion) were subject to a 2.2% annual risk of stroke (11%, 5-year projected risk from the ACAS). This value was used in the simulations as the base case value. The base case value for patients with occlusion was a stroke rate of 7.5% per year. This value was calculated from data from the Extracranial to Intracranial Bypass Trial, in which 72 of 276 patients with internal carotid artery occlusion suffered a stroke during the 3.5 mean years of follow-up.14 These patients had presented with ischemic symptoms but were asymptomatic during the course of the study.14
Patients with <60% stenosis in the high-prevalence population were assumed to have a stroke risk of 0.6% per year. This base case value was estimated from data from the ECST.15 The ECST group reported the risk of stroke in the distribution of the asymptomatic carotid artery in 2295 patients who had been randomized to surgery or medical therapy for the contralateral symptomatic carotid artery. A 2.1% incidence of stroke was observed ipsilateral to asymptomatic vessels <70% stenotic (ECST criteria) at 3 years.15 This translates into an annual risk of stroke of 0.7%. A slightly lower value of 0.6% was chosen because many of these ECST patients had >60% stenosis. Seventy percent stenosis by ECST criteria converts to a higher degree of stenosis by NASCET criteria.
In the low-prevalence population model, the rates of surgical and medical stroke and death for patients with ≥60% stenosis were assumed to be the same as for the high-prevalence population. After carotid endarterectomy, patients were assumed to have a stroke risk of 0.56% per year. Patients with <60% stenosis in the general population without cardiovascular risk factors were assumed to have an annual risk of stroke of 0.1%. All patients are assumed to be on aspirin.
The death rate at 5 years for the surgical cohort reported in ACAS was 20.5%. The death rate at 5 years for the medical cohort was 25.5%.1 Patients with carotid occlusion have been reported to have an annual death rate of 5.5%.14 These mortality rates are greater than that which would be expected for patients of these ages and likely reflect the high incidence of cardiovascular disease.
For this analysis, life-table data for competing causes of mortality were used as a baseline mortality rate, and the additional risk observed for these groups of patients was added to the baseline. Patients with <60% stenosis in the high-prevalence population were assumed to have a similarly increased mortality rate due to comorbid conditions such as cardiac and peripheral vascular disease. In the low-prevalence population, the baseline death rate for patients with <60% disease was derived from a life-table of competing causes of mortality by age.16 Twenty percent of patients were assumed to die within 1 year of stroke.13 14 A 10% annual mortality for stroke survivors was assumed.
Quality of Life Estimates
The stroke-free quality of life factor used in this analysis for all patients was a base case value of 0.90.17 Mark et al17 interviewed patients 1 year after myocardial infarction and found an average time trade-off value of 0.90. Patients were willing to trade 10 years of life at their present state of health for 9 years of excellent health. This value is similar to those found by Fryback et al18 in interviews with 1356 older adults (aged 45 to 85 years) with and without common self-reported health conditions: arthritis, back pain, cataract, hypertension, and sleep disorder. Time trade-off values and quality of well-being scores were obtained and stratified by age. In each age range, an average of 1 to 3 of these chronic health conditions was reported, with mean time trade-off values of 0.89.18
The quality of life estimates for patients after stroke were derived from a study by Gage et al.19 They interviewed 69 patients with atrial fibrillation using a computer-based utility assessment tool. These patients rated the utility of survival with stroke, ranging from mild to severe, using the time trade-off technique. A mean value of 0.6 for all patients after stroke was used as the base case value. This number was multiplied by the 0.90 baseline in this analysis.
The estimated cost of an outpatient Doppler ultrasound examination was $109 (1995 dollars) compared with $2000 for an outpatient arteriogram and $9000 for an endarterectomy. The average values for arteriography and endarterectomy are similar to those reported by other investigators.20 Estimated costs were obtained from 1995 local Medicare reimbursement information (Paul J. Meyer, University of Wisconsin Hospitals and Clinics, written communication, September 15, 1995).
The estimated cost of death was estimated to be $5000. The estimated cost of stroke was assumed to be $20 000 in the first year and $10 000 for each year thereafter.19 21 22 23 These values are based primarily on data from Gage et al19 and represent weighted averages of the estimated costs of acute and chronic major and minor strokes. All future costs were discounted at an annual rate of 3%.
The influence of changes over a wide range for each assumed variable was assessed in one-way analyses. A summary of base case assumptions and the range of values tested in sensitivity analyses is shown in Table 1⇓. The effect of the prevalence of ≥60% stenosis in both a high-prevalence and low-prevalence population was tested. The Doppler ultrasound threshold (the point at which the test is considered positive) was evaluated with local data. The effects of Doppler accuracy were evaluated with reported sensitivities and specificities found in the literature. Angiographic complication rates varied from 0.3% to 3%.24 25 Perioperative stroke and death rates varied from 1% to 3%.26 27 28 29 Changes in baseline mortality and the quality of life factors were evaluated. Costs varied from 50% to 150% of base case values.
Annual screening of a population with a high prevalence of ≥60% stenosis cost $457 773 per incremental QALY gained (Table 2⇓). A one-time screening program gained 30 QALYs at an incremental cost per QALY of $35 130. In the low-prevalence population model, the one-time screening of a population with a low prevalence of ≥60% stenosis yielded an incremental present-value discounted cost per incremental present-value discounted QALY of $52 588. Annual screening of the low-prevalence population was detrimental. More QALYs were lost in the screened population than by natural history.
Sensitivity analyses of one-time screening in the high-prevalence population model demonstrated substantial sensitivity of the model to several variables (Table 3⇓). The most critical were the risk reduction observed with surgery and the annual discount rate. Reducing these variables to the least favorable value for screening led to a loss of QALYs. The model was also sensitive to the prevalence of ≥60% stenosis, the arteriographic complication rate, and the perioperative surgical stroke and death rates. Changes in these variables within the range tested had a strong impact on the cost-effectiveness of screening. The model was less sensitive to the costs of diagnosis and treatment.
Analysis of the Doppler ultrasound threshold revealed moderate sensitivity of the model to the point at which a Doppler examination was considered positive (Figure⇓). The point with the greatest cost-effectiveness in the high-prevalence model was a peak systolic velocity of 230 cm/s. At thresholds below this, more true-positive patients were identified but more patients underwent arteriography (both Doppler false-positive and true-positive). Increasing the Doppler threshold reduced the number of arteriograms but also reduced the detection of surgically significant disease. We assessed the effects of Doppler ultrasound accuracy by using published values of sensitivity and specificity from other laboratories (Table 4⇓).
The sensitivity analysis of the model for the population with a high prevalence of ≥60% stenosis demonstrates critical sensitivity of the model to the rate of stroke observed in patients with ≥60% stenosis after surgery or with medical therapy. The base case values used for these two variables came entirely from data from the ACAS. The limitations of this study, therefore, require discussion. To participate, all centers were required to document excellent rates of surgical morbidity and mortality. The observed benefit applied only to men. Statistical significance was achieved for the reduction of all strokes but not for major ones.
In addition, the 95% confidence levels for the aggregate risk reduction of 53% observed in the trial were 22% to 72%.1 In this simulation model, QALYs were lost when we used an annual stroke rate of 1.2% after surgery. This annual stroke rate converts to an aggregate risk reduction of 25%, which is still within the 95% confidence level observed in the ACAS.
The Veterans Affairs trial was another well-designed study investigating surgery for asymptomatic patients with stenoses.29 The investigators failed to demonstrate a reduction in stroke risk when transient ischemic attacks were not included as an end point. The surgical morbidity and mortality were higher than those observed in ACAS (4.3% compared with 2.3%). The rate of ipsilateral stroke after the perioperative period was similar (3.3% risk for ipsilateral stroke at 5 years).
In summary, the benefit demonstrated by ACAS was small, although statistically significant, and it applied only to men. Sensitivity analyses of surgical risk reduction values still within the 95% confidence level reported in the ACAS revealed lost QALYs. A clinical trial of surgery for asymptomatic patients is currently under way in Europe and may shed further light on this subject.
These results are in stark contrast to those of NASCET for symptomatic patients with ≥70% carotid stenosis. Statistically significant reductions were observed for virtually all end points, and the absolute annual risk reductions with surgery were larger.17
The other variable that had a profound effect on the outcome of the model was the annual discount rate for accumulated costs and QALYs. Increasing the discount rate to 7% resulted in lost QALYs. A discount rate for future costs or life-years is routinely used in cost-effectiveness analyses.5 In this simulation, the long-term stroke risk reduction conferred by surgery must overcome the initial arteriographic and perioperative strokes and deaths. Higher annual discount rates reduce the beneficial effect of the long-term stroke risk reduction.
Can a high-prevalence population be reliably identified? A population with a 10% prevalence of ≥60% stenosis led to a marked increase in the cost per QALY gained. Prevalence data from neurologically asymptomatic populations are based on Doppler studies. Some of these reports document validation of Doppler performance against angiography, while others do not. The assumed values for a high-prevalence population of ≥60% asymptomatic stenosis used in this study were based primarily on the studies of Hennerici et al6 because of their rigorous documentation of angiographic validation. Hennerici et al used continuous-wave Doppler to evaluate 2009 neurologically asymptomatic patients. Of the 375 who were admitted for peripheral vascular disease, 123 (32.8%) had stenoses >50% or occlusion. Alexandrova et al9 reported a lower prevalence of stenosis >70% in 17% of 348 consecutive patients with peripheral vascular disease but did not mention validation of their Doppler measurements. Chambers and Norris10 evaluated 336 patients with an asymptomatic cervical bruit with Doppler. Sixty-one percent of patients had extracranial internal carotid artery stenoses >35%. Berens et al11 studied 1184 consecutive patients aged 65 years or older presenting for cardiac surgery. The prevalence of ≥50% carotid stenosis by Doppler ultrasound was 17%. Increased risk for ≥80% carotid stenosis in the 951 neurologically asymptomatic patients by univariate analysis was found in patients with peripheral vascular disease (P=.007) or a prior myocardial infarction (P=.023). It therefore appears that the presence of peripheral vascular disease, prior myocardial infarction, or a cervical bruit may be associated with asymptomatic carotid stenosis.
Changes in Doppler ultrasound accuracy had a moderate effect on the outcome of the model. The Doppler ultrasound data used in this analysis came from review of local experience. The performance of Doppler ultrasound can vary tremendously between laboratories.3 In a study of ACAS centers, Howard et al3 reported that several Doppler ultrasound units could not achieve any correlation between peak systolic velocity and angiographic stenosis. They concluded that the high reported sensitivity and specificity in the literature were due to publication bias, and they recommended caution in extrapolating results between laboratories. These data and the results of our simulation emphasize the importance of good quality control in Doppler ultrasound laboratories.
A retrospective review of Doppler ultrasound performance may be affected by several different sources of bias: workup bias, threshold bias, and selection bias. As long as the distribution and spectrum of stenoses in the source population and the target population are similar, the effect of these biases is reduced. Applying the retrospectively derived values of Doppler ultrasound sensitivity and specificity to the low-prevalence population will be less valid than with the higher-prevalence population.
This model considers only Doppler ultrasound and intra-arterial arteriography for the detection and measurement of extracranial carotid stenosis. In the future, however, other imaging techniques such as MRA may come to replace either or both of these modalities. Alternatively, some authors have advocated the use of Doppler ultrasound alone or in conjunction with MRA to preoperatively define the degree of stenosis and anatomy of the extracranial carotid.33 This approach sacrifices accuracy in diagnosis in return for reduced arteriographic complications. The benefit of this method of diagnosis remains unproved and controversial. The NASCET and ACAS trials were based on the degree of stenosis measured by arteriographic criteria.1 17 As with Doppler ultrasound, MRA accuracy will also have to be validated in comparison to arteriography on a center-by-center basis.
One-time screening of an asymptomatic population with a high prevalence of >60% stenosis with Doppler ultrasound followed by arteriography and surgery, if indicated, is cost-effective compared with other medical interventions. Goldman et al34 and Kupersmith and coworkers35 have developed the following cost-effectiveness categories for different treatments: highly cost-effective (<$20 000 per QALY or years of life saved), cost-effective (<$40 000), borderline ($40 000 to $60 000), expensive (>$60 000 to $100 000), and very expensive (>$100 000).34 35
King et al36 calculated the incremental cost per QALY gained in the surgical treatment of unruptured cerebral aneurysms to be $24 200. Treatment of hypertension ranges from $16 900 to $111 600 per year of life saved, depending on the agent chosen.37 The cost of oral cholestyramine for hyperlipidemia ranges from $99 500 to $1 380 000 per year of life saved.37 The cost-effectiveness analyses for many of these medical interventions assumed a quality-adjusted life factor of 1.0 for disease-free survival, which as we have discussed may not be valid. Correcting for this would increase the cost per incremental QALY.
In conclusion, this analysis demonstrates that one-time screening of an asymptomatic population with a high prevalence of ≥60% stenosis with Doppler ultrasound followed by arteriography and endarterectomy, if indicated, may be cost-effective. Annual screening of this population was very expensive. There does not appear to be a role for Doppler ultrasound screening in low-prevalence populations. One-time screening of a low-prevalence population was borderline cost-effective, and annual screening was detrimental, producing negative QALYs.
The most critical variables in this simulation model were the risk reduction conferred by surgery and the annual discount rate. The prevalence of ≥60% stenosis in the screening population also had a strong impact on the cost-effectiveness of screening. Risk factors for a high prevalence of ≥60% stenosis appear to be prior myocardial infarction, peripheral vascular disease, or a carotid bruit. The most important locally definable variables affecting the cost-effectiveness of this practice were the perioperative and angiographic stroke and death rates and the accuracy of Doppler ultrasound. Before applying this information to local practice, therefore, clinicians must first be aware of this information.
Selected Abbreviations and Acronyms
|ACAS||=||Asymptomatic Carotid Atherosclerosis Study|
|ECST||=||European Carotid Surgery Trial|
|MRA||=||magnetic resonance angiography|
|NASCET||=||North American Asymptomatic Carotid Endarterectomy Trial|
This study was supported in part by a Siemens Medical Systems/RSNA Research and Education Fund fellowship (Dr Derdeyn), the Charles A. Dana Foundation through the Dana Consortium on Neuroscience: Neuroimaging Leadership Training (Drs Derdeyn and Powers), and National Institutes of Health, National Institute of Neurological Disorders and Stroke grant NS28947 (Dr Powers). The authors are grateful for useful advice and guidance from Dennis G. Fryback, PhD.
- Received April 24, 1996.
- Revision received July 15, 1996.
- Accepted July 15, 1996.
- Copyright © 1996 by American Heart Association
Lipscomb J, Weinstein MC, Torrance GW, and Members of the Panel on Cost-effectiveness in Health and Medicine. Time preference. In: Gold M, Siegel J, Weinstein MC, Russell L, eds. Cost-effectiveness in Health and Medicine. New York, NY: Oxford University Press; 1996:chap 7.
Hennerici M, Aulich A, Sandmann W, Freund H-J. Incidence of asymptomatic extracranial arterial disease. Stroke. 1981;12:750-758.
Pujia A, Rubba P, Spencer MP. Prevalence of extracranial carotid artery disease detectable by echo-Doppler in an elderly population. Stroke. 1992;23:818-822.
Luisiani L, Visona A, Pagnan A. Noninvasive study of arterial hypertension and carotid atherosclerosis. Stroke. 1990;21:410-414.
Alexandrova NA, Gibson WC, Maggisano R. Carotid artery disease and peripheral vascular disease. Stroke. 1995;26:175. Abstract.
Chambers BR, Norris JW. Clinical significance of asymptomatic neck bruits. Neurology. 1985;35:742-745.
Mark DB, Hlatky MA, Califf RM, Naylor CD, Lee KL, Armstrong PW, Barbash G, White H, Simoons ML, Nelson CL. Cost effectiveness of thrombolytic therapy with tissue plasminogen activator as compared with streptokinase for acute myocardial infarction. N Engl J Med. 1995;332:1418-1424.
Fryback DG, Dasbach EJ, Klein R, Klein BE, Dorn W, Peterson K, Martin PA. The Beaver Dam Health Outcomes Study: Initial catalog of health-state quality factors. Med Decis Making. 1993;13:89-102.
Gage BF, Cardinalli AB, Albers GW, Owens DK. Cost effectiveness of warfarin and aspirin for prophylaxis of stroke in patients with nonvalvular atrial fibrillation. JAMA. 1995;275:1839-1845.
Smurawaska LT, Alexandrov AV, Bladin CF, Norris JW. Cost of acute stroke care in Toronto, Canada. Stroke. 1994;25:1628-1631.
Waugh JR, Sacharias N. Arteriographic complications in the DSA era. Radiology. 1992;182:245-246.
Heiserman JE, Dean BL, Hodak JA, Flom RA, Bird CR, Drayer BP, Fram EK. Neurological complications of cerebral angiography. AJNR Am J Neuroradiol. 1994;15:1408-1411.
Brott T, Thalinger K. The practice of carotid endarterectomy in a large metropolitan area. Stroke. 1984;15:950-955.
Easton JD, Sherman DG. Stroke and mortality rate in carotid endarterectomy: 228 consecutive operations. Stroke. 1977;8:565-570.
Mittl RL, Broderick M, Carpenter JP, Goldberg HI, Listerud J, Mishkin MM, Berkowitz HD, Atlas SW. Blinded-reader comparisons of magnetic resonance angiography and duplex sonography for carotid artery bifurcation stenosis. Stroke. 1994;25:4-10.
Eliasziw M, Rankin RN, Fox AJ, Haynes RB, Barnett HJ. Accuracy and prognostic consequences of ultrasonography in identifying severe carotid artery stenosis: NASCET Group. Stroke. 1995;26:1747-1752.
Goldman L, Gordon DJ, Rifkind BM, Hulley SB, Detsky AS, Goodman DW, Kinosian B, Weinstein MC. Cost and health implications of cholesterol lowering. Circulation. 1992;85:1960-1968.