вход по аккаунту


Cost-effectiveness of diagnostic strategies prior to carotid endarterectomy.

код для вставкиСкачать
Cost-effectiveness of Diagnostic Strategies
Prior to Carotid Endarterectomy
Jean Marie U-King-Im, MRCS,1 William Hollingworth, PhD,2 Rikin A. Trivedi, MRCS,1
Justin J. Cross, FRCR,1 Nicholas J. Higgins, FRCR,1 Martin J. Graves, MSc,1 Sergei Gutnikov, MD,3
Peter J. Kirkpatrick, FRCS,1 Elizabeth A. Warburton, MRCP,1 Nagui M. Antoun, FRCR,1
Peter M. Rothwell, FRCP,3 and Jonathan H. Gillard, FRCR1
The main objective of this study was to assess the long-term cost-effectiveness of five alternative diagnostic strategies for
identification of severe carotid stenosis in recently symptomatic patients. A decision-analytical model with Markov transition states was constructed. Data sources included a prospective study involving 167 patients who had screening Doppler ultrasound (DUS), confirmatory contrast-enhanced magnetic resonance angiography (CEMRA) and confirmatory
digital subtraction angiography (DSA), individual patient data from the European Carotid Surgery Trial and other
published clinical and cost data. A “selective” strategy, whereby all patients receive DUS and CEMRA (only proceeding
to DSA if the CEMRA is positive and the DUS is negative), was most cost-effective. This was both the cheapest imaging
and treatment strategy ($35,205 per patient) and yielded 6.1590 quality-adjusted life years (QALYs), higher than three
alternative imaging strategies. Probabilistic sensitivity analysis demonstrated that there was less than a 10% probability
that imaging with either DUS or DSA alone are cost-effective at the conventional $50,000/QALY threshold. In conclusion, DSA is not cost-effective in the routine diagnostic workup of most patients. DUS, with additional imaging in the
form of CEMRA, is recommended, with a strategy of “CEMRA and selective DUS review” being shown to be the optimal
imaging strategy.
Ann Neurol 2005;58:506 –515
The pooled analysis of data from the major carotid trials has confirmed the significant benefits of carotid
endarterectomy (CEA) for severe (70 –99%) carotid
stenosis in patients with recent symptoms within a
6-month period.1 In these trials, intraarterial digital
subtraction angiography (DSA) was used to measure
stenosis and, as such, remains the gold standard against
which alternative imaging modalities need to be validated. Concerns about the small but potentially significant risks of neurological complications associated
with DSA have fuelled strong interest in noninvasive
modalities such as Doppler ultrasound (DUS), magnetic resonance angiography (MRA), or computed tomography angiography (CTA) alone or in combination.2– 4
Despite abundant literature, the optimal imaging
strategy remains controversial.5,6 Several factors may
account for this. First, many studies of carotid imaging
may be methodologically limited, and consequently clinicians may be confused by reported accuracy estimates
ranging from perfect to disturbingly poor.7 Second,
clinical studies may not always keep pace with constantly evolving imaging technology, which may have
added to the confusion. Thus, compared with traditional time-of-flight (TOF) MRA with acquisition
times of up to 10 minutes, contrast-enhanced MRA
(CEMRA) can acquire images from the aortic arch to
the circle of Willis in approximately 30 seconds and is
less prone to flow and motion artifacts.8 With technical quality now robust, CEMRA has moved from the
research arena into routine clinical practice.9 –11 Third,
despite numerous studies of diagnostic accuracy, little
consideration has focused on the associated medical repercussions.11,12 To determine the optimal strategy,
the consequences of diagnostic misclassification with
noninvasive tests (ie, inappropriately denying or referring patients for surgery) need to be balanced against
the risks of DSA. Moreover, from a societal perspective, the true costs of using these strategies need to be
The current lack of consensus is reflected by widely
varying practices worldwide.2– 4 Thus, surveys show
From the 1Department of Radiology, Addenbrooke’s Hospital and
the University of Cambridge, Cambridge, United Kingdom; 2Department of Radiology, University of Washington, Seattle, WA; and
Stroke Prevention Research Unit, Department of Clinical Neurology, Radcliffe Infirmary, Oxford, United Kingdom.
Published online Sep 26, 2005 in Wiley InterScience
( DOI: 10.1002/ana.20591
Received Apr 14, 2005, and in revised form Jun 7. Accepted for
publication Jun 21, 2005.
Address correspondence to Dr Gillard, University Department of
Radiology, Addenbrooke’s Hospital, Cambridge, CB2 2QQ, United
Kingdom. E-mail:
© 2005 American Neurological Association
Published by Wiley-Liss, Inc., through Wiley Subscription Services
that up to 50% of centers may still consider DSA as
routinely necessary.2– 4,13,14 Although DUS is portable,
relatively inexpensive, and potentially accurate, it has
significant interobserver variability and many prefer to
use it as an initial screening tool before an additional
confirmatory test.15 Multislice CTA is promising but
remains limited by radiation doses, plaque calcifications that may obscure the actual stenosis, and a relative paucity of published rigorous studies.16 Surveys
also suggest that the combination of screening DUS
with MRA, especially CEMRA, has emerged as the
noninvasive imaging strategy of choice.2– 4,13,14 Although adverse reactions to gadolinium contrast medium are generally less frequent than with iodinated
contrast medium as used in DSA, it is noteworthy that
a small proportion of patients may not be suitable for
MRA because of contraindications such as claustrophobia and metallic foreign bodies. The main aim of this
study was to assess the use of DUS, CEMRA, and two
combinations of these tests in comparison with DSA
and to identify the most cost-effective long-term diagnostic strategy.
Materials and Methods
Study Design
We constructed a decision analytical model, using DATA
software (version 4.0; Treeage Software, Williamstown, MA)
to evaluate the cost-effectiveness of five imaging strategies for
the diagnostic workup of patients, suspected of having carotid stenosis. The hypothetical cohort of patients was recently symptomatic, with a transient ischemic attack (TIA)
or nondisabling stroke within the previous 6 months.17 Analysis was conducted from the perspective of the health care
provider over a 10.8-year period, representing the length of
follow-up data available from the European Carotid Surgery
Trial (ECST).17
Diagnostic Accuracy
Diagnostic accuracy of the different imaging strategies was
derived from an ethically approved prospective study, performed at a single academic institution between 2000 and
2003. One hundred sixty-seven consecutive recently symptomatic patients (45 women; mean age, 70 years) scheduled
for DSA, after initial DUS had suggested 50% or higher stenosis, were recruited to undergo CEMRA. Full details of this
study have been reported elsewhere.18
In brief, DUS was performed by experienced technologists, using a combination of gray scale, color Doppler, spectral Doppler, and internal carotid artery (ICA) peak systolic
velocity, conforming to the Society of Radiologists in Ultrasound recommendations.19 For DSA, bilateral selective common carotid artery catheterization was performed with a matrix size of 1,024 ⫻ 1,024 and 0.32 ⫻ 0.32mm resolution
(Angioskop; Siemens, Erlangen, Germany). CEMRA was
conducted on a 1.5-tesla machine using a bolus-timed
elliptic-centric phase-encode ordering scheme (resolution,
0.6 ⫻ 0.8mm). All CEMRA and DSA examinations were
reported in a blinded, randomized order by three indepen-
dent attending neuroradiologists according to the North
American Symptomatic Carotid Endarterectomy Trial
(NASCET) method.1 Identical four-view projections were
used for both DSA and CEMRA to avoid measurement bias
secondary to noncircular lumens. Interobserver agreement for
classification into mild, moderate, and severe stenosis were
very good for both DSA (␬ ⫽ 0.80, 0.80, and 0.81) and
CEMRA (␬ ⫽ 0.82, 0.78, and 0.78). Sensitivity and specificity of each diagnostic strategy was calculated with DSA as
reference standard.
Imaging Strategies
We sought to evaluate whether confirmatory tests, such as
DSA or CEMRA, are necessary after initial screening DUS
has shown 50% or higher stenosis. Therefore, only the 200
arteries (in 142 patients) that had 50% or higher stenosis on
DUS were included in our accuracy estimates for DUS and
CEMRA. Indeed, as shown by Kallmes and colleagues, inclusion of the contralateral arteries with less than 50% stenosis on DUS would have artificially inflated our specificity,
and this would not reflect the role of CEMRA and DSA in
clinical practice, that is, as confirmatory tests in a population
prescreened by DUS.20
The diagnostic parameters of the five strategies are illustrated in Table 1. All patients receive optimal medical therapy but are referred for CEA only if test results show a 70 to
99% stenosis.1 As in NASCET and ECST, optimal medical
therapy includes antiplatelet agents (usually aspirin), advice
to stop smoking, and, as indicated, antihypertensive and antilipid therapy.1 For Strategy 1 (DUS alone), no additional
imaging is performed and definitive treatment decision is
based on the results of the initial DUS examination alone.
For the other four strategies, confirmatory tests (DSA or
CEMRA) are performed only if DUS showed 50% or higher
stenosis. For Strategies 2 (CEMRA) and 3 (DSA), definitive
treatment decision is based on the confirmatory test alone,
disregarding the initial DUS results. Strategy 4, termed
“combination DUS and CEMRA” bases treatment decisions
on both tests only if both agree as to the presence or absence
of a severe stenosis. All patients with discordant DUS and
CEMRA results proceed to a definitive DSA examination. In
the fifth strategy, termed “CEMRA and selective DUS review,” patients are treated medically if CEMRA shows 70%
or lower stenosis. If both tests show a severe stenosis, patients
proceed to surgery. However, if CEMRA is positive and initial DUS negative, confirmatory DSA is performed. Compared with Strategy 4, this intermediate strategy has a smaller
discordant rate needing DSA at the expense of a slightly increased misclassification rate (ie, increased false-positives and
false-negatives; see Table 1).18
As DSA is assumed perfectly accurate, all patients imaged
with DSA receive appropriate medical or surgical treatment,
but a proportion (0.85%) incur a procedure-related stroke.
This was calculated by combining data from a large prospective series by Willinsky and colleagues21 and a meta-analysis
by Cloft and colleagues.22 DSA-related TIAs were excluded
because they do not affect either long-term survival or quality of life. Complications were weighted as follows: nondisabling stroke (45%), disabling stroke (33%), and fatal stroke
(22%).11,23 We assumed that patients suffering DSA-related
U-King-Im et al: Imaging of Carotid Stenosis
Table 1. Diagnostic Performance of the Five Diagnostic Strategies Evaluated
Strategy description
DUS alone
Combination DUS
CEMRA and selective
DUS review
Sensitivity (95%
Specificity (95%
(95% CI)
88.1% (80.3–95.8)
93.0% (87.0–98.9)
96.6% (88.3–99.0)
60.7% (52.0–69.3)
80.6% (73.8–87.4)
79.8% (71.1–88.4)
29.6% (23.1–36.1)
15.0% (10.1–19.9)
10.1% (5.8–14.3)
24.9% (18.7–31.0)
91.9% (88.5–95.4)
86.0% (79.6–92.3)
11.6% (7.1–16.2)
6.9% (3.3–10.5)
Discordant Rate
(95% CI)
All five strategies involved initial DUS examination. Diagnostic accuracy refers to identification of severe (70 –99%) stenosis with DSA as
reference standard. Discordant cases need to proceed to DSA. Misclassification rate was defined as the total % of false-positives and falsenegatives.
CI ⫽ confidence interval; DUS ⫽ Doppler ultrasound; NA ⫽ not applicable; CEMRA ⫽ contrast-enhanced MR angiography; DSA ⫽ digital
subtraction angiography.
strokes do not undergo surgery subsequently and remain on
medical therapy. A disabling stroke was defined as a score of
3, 4, or 5 on the modified Rankin scale.1,17
For noninvasive imaging strategies, some patients will be
appropriately treated medically or surgically as a result of
true-negatives (TNs) and true-positives (TPs), respectively.
Others will, however, be inappropriately denied or referred
for surgery as a result of false-negative (FN) or false-positive
(FP) results, respectively. An outline of the decision analysis
model is illustrated in Figure 1.
Markov Model and Transition Probabilities
After diagnosis and initial treatment, a Markov model was
used to track the long-term costs and quality-adjusted life
years (QALYs) for patients, with mild stenosis (0 – 49%),
moderate stenosis (50 – 69%), severe stenosis (70 –99%), or
occlusions, treated either appropriately or inappropriately. In
a Markov model, there is a specified set of mutually exclusive
and exhaustive health states and for which there are transition probabilities of moving from one state to another. We
modeled five possible health states: “healthy,” “nondisabling
stroke,” “disabling stroke,” “fatal stroke,” and “deaths from
other causes.” Based on ECST, we estimated that 58% were
essentially healthy at presentation with a TIA and that 42%
presented with a nondisabling stroke.17 We assumed that patients could either remain in the same health state or move
to a more severe state but not to a less severe state.
Monthly transition probabilities from one state to another
were then modeled and stratified by stenosis grade and diagnostic category (TP, FP, TN, and FN). These probabilities
were directly obtained by analysis of individual patient data
up to 10.8 years from ECST and are summarized in Table
2.17 Full details of the methodology and results of ECST,
reanalyzed according to the NASCET measurement method,
have been published elsewhere.17 ECST data, rather than
population life-tables, were also used to estimate monthly
risks of “death from other causes” because patients with carotid stenosis may harbor excess cardiovascular mortality
compared with the general population. These risks were
stratified according to stenosis but were assumed to be equal
for both surgical and medical arms after the first month (ac-
Annals of Neurology
Vol 58
No 4
October 2005
counting for the initial 30-day operative risk of dying from
other causes), because CEA does not realistically influence
long-term risks of dying from causes other than strokes.
Patients with carotid occlusions were treated medically
and not randomized in ECST. The risks of any stroke for
carotid occlusions were estimated as 5.0% for the first
3 years and then at 2.0% for each subsequent year.24 –26
These strokes then were weighted as follows: nondisabling
stroke (46%), disabling stroke (34%), or fatal stroke
(20%).11,23 Monthly risks of “deaths from other causes” for
patients with carotid occlusions were estimated to equal the
average monthly risk observed in all patients in ECST.
Moreover, if an occlusion was inappropriately treated surgically, an excess stroke risk of 4.2% was applied to the first
month to reflect the excess 30-day operative risks in ECST.17
Utilities and Costs
For the Markov health states, a health-related quality weight
(utility) of 0 was used to represent “death” and of 1 to represent the “healthy” state. Based on a published systematic
review (time tradeoff scores), we assumed that utility for a
year of life was 0.20 after a disabling stroke and 0.70 for a
nondisabling stroke (see Table 4).27
Costs of CEMRA and DSA examinations were derived
from a concurrent activity-based cost analysis in 20 patients
and are shown in Table 3. Full details of this study have
been published elsewhere.28 Because all patients receive
DUS, the cost of DUS in the five diagnostic strategies evaluated is identical and therefore was excluded from our analysis.
The costs associated with CEA (see Table 3) were derived
from a systematic review of studies reporting on costs of
CEA worldwide.29 Costs of long-term medical therapy were
excluded because they would be similar in both medical and
surgical patients. One-off costs for initial hospitalization were
assigned for each patient who suffered a stroke, according to
the severity of stroke (see Table 3). Inpatient care is the major driving force behind total acute stroke costs and was assigned proportions of 61%, 69%, and 91% of total acute
stroke costs for nondisabling, disabling, and fatal strokes, respectively.30 Finally, monthly costs, calculated from the Of-
Fig 1. Outline of decision-analytical model. All nodes marked “M” represent the start of the same Markov process as illustrated in
the TP branch. TP ⫽ true-positive; FP ⫽ false-positive; TN ⫽ true-negative; FN ⫽ false-negative.
fice of Population Censuses and Surveys, were assigned for
stroke survivors for each month spent in the “nondisabling
stroke” and “disabling stroke” states, representing long-term
health care in the community (see Table 3).31
All costs were updated to US $ for financial year 2002/
2003 based on the medical component of the Consumer
Price Index rates and official currency exchange rates at the
time of writing. In the base-case analysis, we discounted both
effects in QALYs and costs at 3.0%.32
For the primary analysis, alternatives that were both
more costly and less effective than another option were eliminated from consideration by simple dominance. Incremental
cost-effectiveness ratios (ICERs) were calculated by dividing
the relative difference in cost by the relative difference in
QALY compared with the next least expensive nondominated strategy.
plausible ranges. This refers to a simulation technique that
allows combining a large number of factors with probabilistic
outcomes to characterize the distribution of an end result.33
The uncertainty of sensitivity, specificity, and the probability
of discordant test results requiring DSA confirmation were
estimated using 〉 distributions based on the diagnostic accuracy study (see Table 1).18 Costs of diagnostic tests and
treatment were assumed to follow an approximately normal
distribution as described in Table 3. Uncertainty surrounding prevalence and remaining probability estimates in the
model was also assumed to follow a 〉 distribution (see Table
4). Finally, utility estimates were assumed to be uniformly
distributed over a plausible range of values (see Table 4). We
computed cost-effectiveness results for each of 1,000 iterations for each diagnostic strategy.
Sensitivity Analyses
Base-Case Analysis
Results of the base-case analysis are illustrated in Table
5. “Combination DUS and CEMRA” (6.1591) resulted in the greatest quality-adjusted life expectancy
(QALE) of the five options considered, followed in order by “CEMRA and selective DUS review” (6.1590),
CEMRA (6.1585), “DUS alone” (6.1572), and DSA
(6.1256). Strategies involving CEMRA resulted in similar long-term effectiveness but differed significantly
To examine the robustness of our results to parameter variations, critical parameters were varied according to clinically
plausible ranges or 95% confidence interval (CI), in one-way
sensitivity analyses. These included costs (see Table 3), utilities, prevalence of severe stenosis in study population, and
risks of DSA (Table 4). Annual discount rates for both costs
and utilities were also varied from 0 to 6%.
We also conducted a probabilistic sensitivity analysis in
which all model parameters were simultaneously varied using
a Monte-Carlo simulation on the basis of 95% CIs or other
U-King-Im et al: Imaging of Carotid Stenosis
Table 2. Summary of Risks of Nondisabling Stroke, Disabling Stroke, and Fatal Stroke for Each Stenosis Category for Medical and
Surgical Therapy, Derived from Analysis of Individual Patient Data from European Carotid Surgery Trial
(reanalyzed according to North American Symptomatic Carotid Endarterectomy measurement method)
0–49% Stenosis Treated Medically
(n ⫽ 722)
30 days
1 year
3 years
5 years
10 years
0–49% Stenosis Treated Surgically
(n ⫽ 1081)
4 (0.6%)
16 (2.2%)
35 (4.8%)
47 (6.5%)
56 (7.8%)
1 (0.1%)
6 (0.8%)
16 (2.2%)
26 (3.6%)
31 (4.3%)
1 (0.1%)
3 (0.4%)
8 (1.1%)
15 (2.1%)
18 (2.5%)
25 (2.3%)
51 (4.7%)
74 (6.8%)
86 (8.0%)
110 (10.2%)
16 (1.4%)
34 (3.1%)
48 (4.4%)
60 (5.6%)
74 (6.8%)
7 (0.6%)
14 (1.3%)
21 (1.9%)
26 (2.4%)
38 (3.5%)
50–69% Stenosis Treated Medically
(n ⫽ 266)
30 days
1 year
3 years
5 years
10 years
50–69% Stenosis Treated Surgically
(n ⫽ 378)
2 (0.8%)
7 (2.6%)
17 (6.3%)
21 (7.9%)
25 (9.3%)
1 (0.4%)
8 (3.0%)
14 (5.3%)
21 (7.9%)
26 (9.8%)
0 (0%)
3 (1.1%)
9 (3.4%)
12 (4.5%)
15 (5.6%)
14 (3.7%)
21 (5.5%)
26 (6.9%)
28 (7.4%)
33 (8.7%)
8 (2.1%)
14 (3.7%)
16 (4.2%)
19 (5.0%)
22 (5.8%)
4 (1.1%)
6 (1.6%)
8 (2.1%)
8 (2.1%)
10 (2.6%)
70–99% Stenosis Treated Medically
(n ⫽ 219)
30 days
1 year
3 years
5 years
10 years
70–99% Stenosis Treated Surgically
(n ⫽ 335)
1 (0.5%)
14 (6.4%)
29 (13.2%)
33 (15.1%)
37 (16.9%)
0 (0%)
12 (5.5%)
16 (7.3%)
18 (8.2%)
22 (10.0%)
1 (0.5%)
4 (1.8%)
8 (3.7%)
9 (4.1%)
12 (5.5%)
4 (1.2%)
11 (3.25)
15 (4.5%)
19 (5.7%)
27 (8.1%)
10 (3.0%)
13 (3.9%)
16 (4.8%)
18 (5.4%)
24 (7.1%)
1 (0.3%)
3 (0.9%)
6 (1.8%)
9 (2.7%)
10 (3.0%)
n ⫽ number of patients in group. Cumulative absolute numbers are presented, with cumulative risks presented within parentheses.
when costs were incorporated. Thus, CEMRA, DSA,
or “DUS alone” were dominated strategies compared
with the optimal strategy “CEMRA and selective DUS
review” (see Table 5). Although “combination DUS
and CEMRA” was slightly more effective, it was also
more expensive, resulting in an unfavorable ICER in
excess of $4 million/QALY, far higher than the
$50,000/QALY generally accepted in health care.32
Table 3. Summary of Costs Used in the Decision-Analytical Model, with Plausible Ranges Used for
Both Univariate and Probabilistic Sensitivity Analyses
95% CI
Cost of DSA
Cost of CEMRA
Cost of CEA
Cost of acute hospital admission for a nondisabling stroke
Cost of acute hospital admission for a disabling stroke
Cost of acute hospital admissions for a fatal stroke
Monthly costs of community health care for nondisabling stroke survivors
Monthly costs of community health care for disabling stroke survivors
SD ⫽ standard deviation; CI ⫽ confidence interval; DSA ⫽ digital subtraction angiography; CEMRA ⫽ contrast-enhanced MR angiography;
CEA ⫽ carotid endarterectomy.
Annals of Neurology
Vol 58
No 4
October 2005
Table 4. Summary of Key Variables Used in the Probabilistic Sensitivity Analysis
Annual utility of living with a nondisabling stroke
Annual utility of living with a disabling stroke
Prevalence of severe stenosis in study population
Proportion of patients presenting with TIA
Risk of stroke after DSA
Proportion of fatal strokes after DSA-related stroke
Proportion of nondisabling stroke in survivors
after DSA-related stroke
(11, 21, 22)
(21, 22)
(21, 22)
TIA ⫽ transient ischemic attack; DSA ⫽ digital subtraction angiography.
Sensitivity Analyses
In one-way sensitivity analyses, the base-case results
were insensitive to plausible variations for (1) costs of
acute hospital admission for a nondisabling, disabling,
or fatal stroke; (2) costs of CEA; (3) costs of CEMRA
and DSA; (4) monthly incremental costs of living with
a nondisabling or disabling stroke; (5) discount rates
for costs and utilities; (6) prevalence of severe stenosis
in the study population; and (7) utility values for nondisabling and disabling strokes. Compared with the optimal strategy, DSA remained a dominated option even
when procedural risk was assumed to be as low as
Results of the probabilistic sensitivity analysis also
supported the base-case results (Fig 2). If society always chooses the cheapest strategy (ie, willingness to
pay for a QALY was zero), the probability that the
designated strategy was most cost-effective was 66.5%
for “CEMRA and selective DUS review,” 15.2%
for CEMRA, 13.5% for “combination DUS and
CEMRA,” 4.1% for DSA, and 0.7% for DUS alone.
Moreover, even as society’s willingness to pay for a
QALY increases, the probability that DSA becomes
cost-effective remains extremely low (0.6% at a
threshold of $50,000/QALY). However, actual differences in the cost-effectiveness profile of CEMRA or
the two combination strategies tended to decrease because of uncertainty in our model. Thus, at the
$50,000/QALY threshold, the probability that
“CEMRA and selective DUS review” would be most
cost-effective was 39.2% compared with 26.5% for
“combination DUS and CEMRA” and 24.5% for
CEMRA. The probability that “DUS alone” was most
cost-effective was only 8.7%; in fact, threshold analysis indicates that for “DUS alone” to become optimal the specificity of DUS needs to be 75% or
higher, assuming a constant sensitivity.
Our results may help resolve the ongoing controversy
regarding whether DSA is still routinely necessary for
carotid imaging.5 In our study, all four noninvasive
strategies resulted in better QALE compared with
DSA. This demonstrates that in the long-term, the
clinical benefit of avoiding DSA outweighs the consequences of misclassification due to noninvasive tests.
Moreover, if costs are incorporated, DSA was a dominated strategy compared with all three noninvasive
strategies involving CEMRA. These results were highly
robust to plausible parameter variations in both the
univariate and probabilistic sensitivity analyses. Compared with DSA, “DUS alone” was more effective
(6.1572 vs 6.1256 QALY) but also more expensive
($35,970 vs $35,632). “DUS alone” was, however,
cost-effective resulting in an ICER of only $10,696.
These conclusions are in strong agreement with two
previously published cost-effectiveness studies, confirming that DSA is no longer routinely necessary in the
Table 5. Results of Base-Case Cost-effectiveness Analysis
Cost, $
⌬ Cost, $
⌬ Cost/⌬ QALY
CEMRA and selective DUS review
Combination DUS and CEMRA
DUS alone
Incremental cost-effectiveness ratios (⌬ Cost/⌬ QALY) were calculated by dividing the relative difference in cost by the relative difference in
QALY compared with the next least expensive alternative nondominated strategy.
QALY ⫽ quality-adjusted life years; CEMRA ⫽ contrast-enhanced magnetic resonance angiography; DUS ⫽ Doppler ultrasound; DSA ⫽
digital subtraction angiography.
U-King-Im et al: Imaging of Carotid Stenosis
Fig 2. Graph of probability that a strategy is most cost-effective versus willingness to pay for a QALY. QALY ⫽ quality-adjusted
life years. CEMRA ⫽ contrast-enhanced magnetic resonance angiography; DUS ⫽ Doppler ultrasound; DSA ⫽ digital subtraction
majority of patients.12,34 DSA does, however, remain
the gold standard in terms of diagnostic accuracy and
has still an important role as the definitive test in cases
of disagreement.
We limited our analysis to the identification of only
severe stenosis. Although in clinical practice, selected
patients with moderate stenosis may benefit from CEA,
surgery is not routinely indicated in all such patients
and treatment decisions in this subgroup need to be
individualized, taking into account other risk factors
such as severity and type of symptoms or plaque morphological characteristics such as ulceration.35 Indeed,
patients with moderate stenosis, as a whole, only derive
a marginal benefit (absolute risk reduction of 4.6%)
from CEA.1 The fact that patients with moderate stenosis who are misclassified as having severe stenosis (ie,
FP) derive some, albeit minor, benefit from CEA, also
explain our findings that, purely from an effectiveness
perspective, there was little difference between DUS,
CEMRA, and the two combination strategies tested.
Therefore, the overall clinical benefit from decreasing
the FP rate with more accurate strategies was quite limited. However, when costs were incorporated, the varying diagnostic accuracy became increasingly important
because more accurate strategies resulted in fewer expensive and unnecessary CEA.
Thus, “DUS alone” was not cost-effective when
compared with strategies involving CEMRA. Even at
the $50,000/QALY threshold, the probability that
“DUS alone” was most cost-effective remained as low
as 8.7%. This was a direct result of the tendency of
Annals of Neurology
Vol 58
No 4
October 2005
DUS to overestimate stenosis, resulting in many unnecessary CEA. We used a base-case specificity of
60.7% for detection of severe stenosis for DUS. This
value may appear to be at the lower extreme of published ranges but this reflects the actual specificity in a
prescreened population in whom DUS has shown a
50% or higher stenosis. If we had included the contralateral normal or near-normal arteries, our adjusted
specificity for DUS would have been 76%, which
would be in keeping with previously quoted figures reflecting an unscreened population.36 Assuming a constant sensitivity, threshold analysis indicates that, for
“DUS alone” to become optimal, specificity for severe
stenosis needs to be at least 75% in a prescreened population. In an unscreened population, this specificity is
likely to be approximately equivalent to 85% or
more.37 Although our findings are in agreement with
the conclusions reached by Kent and colleagues, it conflicts with those of Buskens and colleagues who found
that a single-test strategy with DUS was optimal.12,34
Several reasons may account for this. First, we used a
state-of-the-art CEMRA technique in our study as opposed to an older generation TOF technique, which is
reflected in our increased MRA specificity (80.6 vs
76%) compared with Buskens and colleagues.34 Second, Buskens and colleagues found a higher specificity
of 75% for DUS compared with 60.6% in our study.
Such interobserver variability unfortunately remains
one of the most well-recognized drawbacks of DUS.36
Third, our model used individual patient data from
ECST to model the monthly risks of strokes and
deaths. This is more accurate than cruder annual estimates of these risks derived from combinations of literature review and expert opinions.
“Combination DUS and CEMRA” was more effective than CEMRA alone (6.1591 vs 6.1585 QALY)
but also more expensive ($35,476 vs $35,436), resulting in an unfavorable ICER of $66,667/QALY. In our
study, up to 24.9% of patients would have needed to
proceed to DSA in cases where CEMRA and DUS
were discordant. In this study, we propose a novel selective strategy in which DSA is necessary only if
CEMRA were to indicate a severe stenosis and DUS
does not. Compared to the combination strategy, this
resulted in 18% fewer patients needing DSA at the expense of a 1.5% increase in misclassification rates (Table 1). The selective strategy dominated the CEMRA
strategy, being both more effective (6.1590 vs 6.1585
QALY) and less expensive ($35,205 vs $35,436). Compared with the selective strategy, “combination
CEMRA and DUS” was marginally more effective
(6.1591 vs 6.1590 QALY) but also more expensive
($35,476 vs $35,205), resulting in an unfavorable
ICER in excess of $4 million.
Although the base-case results were robust to univariate sensitivity analysis, probabilistic sensitivity analysis
suggests that at higher thresholds of willingness to pay
for a QALY, it is not possible to completely exclude
the fact that CEMRA or “combination DUS and
CEMRA” may be most cost-effective because of model
uncertainty. In practice, which of these CEMRA strategies to adopt therefore may ultimately be dictated by
the expertise and local availability of DSA. It appears
likely that, if DSA is not readily available, basing treatment decisions on the results of CEMRA alone may
also be acceptable.
The strengths of our study are that it is based on
prospectively acquired cost data, as well as individual
patient data from the ECST trial. The study of diagnostic accuracy is moreover currently the largest published prospective study with CEMRA and was adequately powered to detect significant differences
between imaging modalities.18 There are, however, several potential limitations. Because we based all our
monthly risk probabilities of strokes and deaths on actual source data from ECST, we did not perform a
complex sensitivity analysis to analyze variations in
these probabilities. However, this is an acceptable
simplification as ECST remains, to date, one of the
largest randomized trials of CEA, which still guide
both clinical and research practice. Furthermore, reanalysis of ECST data according to NASCET confirmed that both trials are largely consistent.1,17. Similarly, although there may be intercenter variations in
the 30-day operative risks after CEA, there has been
no evidence of a systematic reduction in such complication rates, confirming that the surgical data from
these large trials are still applicable to current practice.38,39
We assumed that patients are appropriately treated
medically if initial DUS suggests 50% or lower stenosis. There may be an element of verification bias, because it would not be ethical for such patients to be
subjected to the risks of confirmatory DSA. However,
analysis of contralateral arteries in our series indicates
that a 50% cutoff on DUS was 100% accurate to exclude severe stenosis on DSA. Also, our model is only
valid for patients in whom MR imaging is practically
feasible and in whom there is genuine uncertainty as to
which diagnostic pathway should be undertaken. In
our study, 16 patients did not complete the CEMRA
examination, although only nine of these patients
(5.3%) had an absolute contraindication to MR imaging. Moreover, anxiety in some patents may tend to
cause motion artifacts and reduce image quality. Such
patients may still need confirmatory DSA or may be
better suited for CTA, depending on local expertise.
Furthermore, it was outside the scope of our study,
which focused purely on extracranial carotid stenosis,
to incorporate the potential findings of tandem lesions
or incidental aneurysms on decision algorithms given
that the implications of such lesions remain highly debatable in clinical practice.5,40
Moreover, the final results of ECST and NASCET
were published nearly 6 or 7 years ago, and it is currently suggested that advances in medical therapy such
as aggressive lipid lowering with high-dose statins,
combination antiplatelet therapy, folate supplements,
or targeted blood pressure reduction may reduce the
risk benefit of CEA compared with medical therapy.
However, this still remains to be conclusively proved,
and results of trials such as the Intensive Carotid Artery Stenosis Treatment Trial, which aim to evaluate
such advances against CEA in selected patients, are
currently awaited.41 Similarly, our cost-effectiveness
analysis evaluates carotid endarterectomy only and not
emerging therapies such as carotid angioplasty and
stenting. It is undeniable that with advances in technology and experience, the risks associated with such
procedures are approaching those reported with surgery. However, until further data become available,
CEA currently remains the gold standard treatment
It is noteworthy that our study was conducted at an
academic center with expertise in MRA and was reported by experienced neuroradiologists. Although
DSA is generally performed by specialist vascular radiologists or neuroradiologists, techniques such as
CEMRA are increasingly becoming available outside of
specialist centers and may be interpreted by general radiologists or physicians with limited experience in carotid imaging. It is important that such centers do im-
U-King-Im et al: Imaging of Carotid Stenosis
plement internal quality control or audit to validate
their local expertise.
A final point of note is that experience and technology are constantly improving. Whereas we were able to
achieve submillimeter resolution with our CEMRA technique, the true in-plane resolution (0.60 ⫻0.80mm) is
still two to three times less than that of DSA (0.32 ⫻
0.32mm), which in part explains why DSA was still the
gold standard method. However, with advances in technology, it is likely that the quality of CEMRA may soon
become rival DSA.43 Such improvements, coupled with
three-dimensional visualization, may further increase the
accuracy of CEMRA and hence also its cost-effectiveness
This work was supported by the United Kingdom National Health
Service Research and Development Programme (HSR 0500/13;RC
33206, J.J.C. N.J.H, M.J.G., R.J.K., N.M.A., J.H.G.).
1. Rothwell PM, Eliasziw M, Gutnikov SA, et al. Analysis of
pooled data from the randomised controlled trials of endarterectomy for symptomatic carotid stenosis. Lancet 2003;361:
2. Long A, Lepoutre A, Corbillon E, et al. Modalities of preoperative imaging of the internal carotid artery used in France. Ann
Vasc Surg 2002;16:261–265.
3. Dawson DL, Roseberry CA, Fujitani RM. Preoperative testing
before carotid endarterectomy: a survey of vascular surgeons’
attitudes. Ann Vasc Surg 1997;11:264 –272.
4. Athanasoulis CA, Plomaritoglou A. Preoperative imaging of the
carotid bifurcation. Current trends. Int Angiol 2000;19:1–7.
5. Davis SM, Donnan GA. Is carotid angiography necessary? Editors disagree. Stroke 2003;34:1819.
6. Norris JW, Rothwell PM. Noninvasive carotid imaging to select
patients for endarterectomy: is it really safer than conventional
angiography? Neurology 2001;56:990 –991.
7. Rothwell PM, Pendlebury ST, Wardlaw J, Warlow CP. Critical
appraisal of the design and reporting of studies of imaging and
measurement of carotid stenosis. Stroke 2000;31:1444 –1450.
8. Carr JC, Shaibani A, Russell E, Finn JP. Contrast-enhanced
magnetic resonance angiography of the carotid circulation. Top
Magn Reson Imaging 2001;12:349 –357.
9. Huston J, 3rd. MRA of carotids moves from novelty to definitive method. Diagn Imaging 2000;22:129 –131.
10. Cosottini M, Pingitore A, Puglioli M, et al. Contrast-enhanced
three-dimensional magnetic resonance angiography of atherosclerotic internal carotid stenosis as the noninvasive imaging
modality in revascularization decision making. Stroke 2003;34:
660 – 664.
11. Berry E, Kelly S, Westwood ME, et al. The cost-effectiveness of
magnetic resonance angiography for carotid artery stenosis and
peripheral vascular disease: a systematic review. Health Technol
Assess 2002;6:1–155.
12. Kent KC, Kuntz KM, Patel MR, et al. Perioperative imaging
strategies for carotid endarterectomy. An analysis of morbidity
and cost-effectiveness in symptomatic patients. JAMA 1995;
274:888 – 893.
13. Norris JW, Morriello F, Rowed DW, Maggisano R. Vascular
imaging before carotid endarterectomy. Stroke 2003;34:E16.
Annals of Neurology
Vol 58
No 4
October 2005
14. Robless P, Halliday A. Vascular Surgical Society of Great Britain and Ireland: carotid angiography is used more selectively in
the Asymptomatic Carotid Surgery Trial. Br J Surg 1999;86:
690 – 691.
15. Perkins JM, Galland RB, Simmons MJ, Magee TR. Carotid
duplex imaging: variation and validation. Br J Surg 2000;87:
320 –322.
16. Koelemay MJ, Nederkoorn PJ, Reitsma JB, Majoie CB. Systematic review of computed tomographic angiography for assessment of carotid artery disease. Stroke 2004;35:2306 –2312.
17. Rothwell PM, Gutnikov SA, Warlow CP. Reanalysis of the final results of the European Carotid Surgery Trial. Stroke 2003;
34:514 –523.
18. U-King-Im J, Trivedi RA, Graves MJ, et al. Contrast-enhanced
MR angiography for carotid disease: diagnostic and potential
clinical impact. Neurology 2004;62:1282–1290.
19. Grant EG, Benson CB, Moneta GL, et al. Carotid artery
stenosis: gray-scale and Doppler US diagnosis—Society of Radiologists in Ultrasound Consensus Conference. Radiology
2003;229:340 –346.
20. Kallmes DF, Omary RA, Dix JE, et al. Specificity of MR angiography as a confirmatory test of carotid artery stenosis. Am J
Neuroradiol 1996;17:1501–1506.
21. Willinsky RA, Taylor SM, TerBrugge K, et al. Neurologic complications of cerebral angiography: prospective analysis of 2,899
procedures and review of the literature. Radiology 2003;227:
22. Cloft HJ, Joseph GJ, Dion JE. Risk of cerebral angiography in
patients with subarachnoid hemorrhage, cerebral aneurysm, and
arteriovenous malformation: a meta-analysis. Stroke 1999;30:
23. Hankey GJ, Warlow CP. Treatment and secondary prevention
of stroke: evidence, costs, and effects on individuals and populations. Lancet 1999;354:1457–1463.
24. Verlato F, Camporese G, Bernardi E, et al. Clinical outcome of
patients with internal carotid artery occlusion: a prospective
follow-up study. J Vasc Surg 2000;32:293–298.
25. Faught WE, van Bemmelen PS, Mattos MA, et al. Presentation
and natural history of internal carotid artery occlusion. J Vasc
Surg 1993;18:512–523; discussion, 524.
26. Powers WJ, Derdeyn CP, Fritsch SM, et al. Benign prognosis
of never-symptomatic carotid occlusion. Neurology 2000;54:
878 – 882.
27. Post PN, Stiggelbout AM, Wakker PP. The utility of health
states after stroke: a systematic review of the literature. Stroke
28. U-King-Im J, Hollingworth W, Trivedi RA, et al. Contrastenhanced MR angiography vs intra-arterial digital subtraction
angiography for carotid imaging: activity-based cost analysis.
Eur Radiol 2004;14:730 –735.
29. Benade MM, Warlow CP. Costs and benefits of carotid endarterectomy and associated preoperative arterial imaging: a systematic review of health economic literature. Stroke 2002;33:
629 – 638.
30. Caro JJ, Huybrechts KF, Duchesne I. Management patterns
and costs of acute ischemic stroke: an international study. For
the Stroke Economic Analysis Group. Stroke 2000;31:582–590.
31. Kavanagh S, Knapp M, Patel A. Costs and disability among
stroke patients. J Public Health Med 1999;21:385–394.
32. Siegel JE, Torrance GW, Russell LB, et al. Guidelines for pharmacoeconomic studies. Recommendations from the panel on
cost effectiveness in health and medicine. Panel on Cost Effectiveness in Health and Medicine. Pharmacoeconomics 1997;11:
159 –168.
33. Doubilet P, Begg CB, Weinstein MC, et al. Probabilistic sensitivity analysis using Monte Carlo simulation. A practical approach. Med Decis Making 1985;5:157–177.
34. Buskens E, Nederkoorn PJ, Der Woude TB, et al. Imaging of
carotid arteries in symptomatic patients: cost-effectiveness of diagnostic strategies. Radiology 2004;233:101–112.
35. Rothwell PM, Eliasziw M, Gutnikov SA, et al. Endarterectomy
for symptomatic carotid stenosis in relation to clinical subgroups and timing of surgery. Lancet 2004;363:915–924.
36. Johnston DC, Goldstein LB. Clinical carotid endarterectomy
decision making: noninvasive vascular imaging versus angiography. Neurology 2001;56:1009 –1015.
37. Nederkoorn PJ, Mali WP, Eikelboom BC, et al. Preoperative
diagnosis of carotid artery stenosis: accuracy of noninvasive testing. Stroke 2002;33:2003–2008.
38. Bond R, Rerkasem K, Shearman CP, Rothwell PM. Time
trends in the published risks of stroke and death due to endarterectomy for symptomatic carotid stenosis. Cerebrovasc Dis
2004;18:37– 46.
39. Bond R, Rerkasem K, Rothwell PM. Systematic review of the
risks of carotid endarterectomy in relation to the clinical indication for and timing of surgery. Stroke 2003;34:2290 –2301.
40. Rouleau PA, Huston J III, Gilbertson J, et al. Carotid artery tandem lesions: frequency of angiographic detection and consequences
for endarterectomy. AJNR Am J Neuroradiol 1999;20:621– 625.
41. Chaturvedi S, Clagett P, Cote R, et al. Intensive Carotid Artery
Stenosis Treatment Trial. Proceedings of the 29th International
Stroke Conference 2004. Stroke 2004;35:253–340.
42. Coward LJ, Featherstone RL, Brown MM. Safety and efficacy
of endovascular treatment of carotid artery stenosis compared
with carotid endarterectomy: a Cochrane systematic review of
the randomized evidence. Stroke 2005;36:905–911.
43. Hu HH, Madhuranthakam AJ, Kruger DG, et al. Improved
venous suppression and spatial resolution with SENSE in elliptical centric 3D contrast-enhanced MR angiography. Magn Reson Med 2004;52:761–765.
U-King-Im et al: Imaging of Carotid Stenosis
Без категории
Размер файла
169 Кб
cost, carotid, prior, strategia, effectiveness, endarterectomy, diagnostika
Пожаловаться на содержимое документа