Context Management of patients with acute transient ischemic attack (TIA) varies
widely, with some institutions admitting all patients and others proceeding
with outpatient evaluations. Defining the short-term prognosis and risk factors
for stroke after TIA may provide guidance in determining which patients need
rapid evaluation.
Objective To determine the short-term risk of stroke and other adverse events
after emergency department (ED) diagnosis of TIA.
Design and Setting Cohort study conducted from March 1997 through February 1998 in 16 hospitals
in a health maintenance organization in northern California.
Patients A total of 1707 patients (mean age, 72 years) identified by ED physicians
as having presented with TIA.
Main Outcome Measures Risk of stroke during the 90 days after index TIA; other events, including
death, recurrent TIA, and hospitalization for cardiovascular events.
Results During the 90 days after index TIA, 180 patients (10.5%) returned to
the ED with a stroke, 91 of which occurred in the first 2 days. Five factors
were independently associated with stroke: age greater than 60 years (odds
ratio [OR], 1.8; 95% confidence interval [CI], 1.1-2.7; P = .01), diabetes mellitus (OR, 2.0; 95% CI, 1.4-2.9; P<.001), symptom duration longer than 10 minutes (OR, 2.3; 95% CI,
1.3-4.2; P = .005), weakness (OR, 1.9; 95% CI, 1.4-2.6; P<.001), and speech impairment (OR, 1.5; 95% CI, 1.1-2.1; P = .01). Stroke or other adverse events occurred in 428
patients (25.1%) in the 90 days after the TIA and included 44 hospitalizations
for cardiovascular events (2.6%), 45 deaths (2.6%), and 216 recurrent TIAs
(12.7%).
Conclusions Our results indicate that the short-term risk of stroke and other adverse
events among patients who present to an ED with a TIA is substantial. Characteristics
of the patient and the TIA may be useful for identifying patients who may
benefit from expeditious evaluation and treatment.
Transient ischemic attacks (TIAs) are common1:
approximately 300,000 TIAs occur each year in the United States based on recent
estimates of stroke incidence.2,3
A recent survey suggested a much higher incidence, with 1 in 15 of those older
than age 65 years reporting a history of TIA.4
About 15% of patients experiencing stroke report a history of TIA.5 Effective prevention of subsequent stroke in patients
with TIAs could significantly reduce the overall stroke incidence.6
Evaluation and treatment of patients with stroke is generally rapid,
in part because management of ongoing cerebral ischemia may reduce the final
injury and because patients often have a disability that requires care.7 The need for urgent intervention is less clear for
patients with TIA, who have usually returned to their baseline level of function
by the time they are evaluated. While some interventions are known to be effective
after TIA, such as antiplatelet agents, anticoagulation for atrial fibrillation,
and endarterectomy for symptomatic carotid artery stenosis,8
it is not known whether more urgent therapy is justified. Hospitalization
or rapid evaluation may be warranted only in patients with high short-term
risk because urgent secondary prevention strategies are costly.9-12
Several studies have evaluated stroke risk after TIA,13-24
but they have generally been small, have failed to examine the short-term
risk separately, and have not identified which factors are associated with
an increased risk of stroke. The most reliable data on short-term prognosis
after TIA come from 2 population-based studies,25,26
but their results have limited applicability to current clinical management
because the studies were performed many years ago and included fewer than
200 cases each; 1 study did not include the risk of stroke in the first few
days after the TIA.
It is often impossible to confirm a diagnosis of TIA since symptoms
are transient and may have nonischemic etiologies such as seizure, syncope,
and migraine. In fact, agreement between independent observers on TIA diagnosis
is poor, even among neurologists.27,28
Clinical decisions are based on the final diagnosis of the treating physician,
who is often an emergency-department (ED) physician, internist, or family
practitioner.29 Since those patients given
a diagnosis of TIA by the treating physician form the cohort currently recognized
in practice, it is essential to study these patients regardless of whether
the diagnosis is confirmed by a neurologist.
Lack of knowledge about the acute natural history of patients with TIA
has led to variability in clinical practice.30
Guidelines are generally vague about evaluation and treatment.8,9
To determine the short-term risk of stroke and other adverse events after
TIA and to identify risk factors, we performed a cohort study of patients
diagnosed as having TIA in the EDs of hospitals belonging to a large health
maintenance organization. We used the treating physicians' diagnoses of TIA
to define the cohort so that results would be generalizable to institutions
where neurologists do not routinely evaluate patients with TIA in the ED.
Kaiser-Permanente Northern California is a health maintenance organization
with 16 hospitals in the San Francisco Bay Area, Sacramento, and Santa Rosa,
Calif. There are 2.9 million enrollees, with demographics similar to the regional
population.31 Beginning in March 1997, ED physicians
at all 16 facilities were asked to select a primary diagnosis (including TIA)
for all patient visits. These diagnoses were entered into a database.
Patients (n = 1797) seen in the ED and diagnosed as having TIA from
March 1997 through February 1998 were identified. Those who did not have ED
records available (n = 30), were not members of the health plan (n = 27),
had coded diagnoses other than TIA (n = 25), or had a prior TIA treated in
the ED during the study period (n = 8) were excluded. Medical records were
reviewed by trained analysts in conjunction with a neurologist. A neurologist
blinded to follow-up events reviewed the medical records of all patients in
whom the diagnosis of TIA was unclear based on physician documentation or
analyst review. Definite TIA was defined based on World Health Organization
criteria32 as rapidly developed clinical signs
of focal or global disturbance of cerebral function lasting fewer than 24
hours, with no apparent nonvascular cause. Analyses of definite TIA cases
are reported separately from the primary analysis, which is based on the ED
diagnosis.
Patient characteristics, medical history, TIA symptom details, medications,
examination findings attributed to the TIA, and treatment plans were abstracted
using predefined criteria. Emergency department physician notes were considered
the primary source, supplemented with nursing notes when necessary. Subtypes
of speech impairment and vision loss were not identified, given the difficulty
of distinguishing dysarthria from aphasia and monocular blindness from a homonymous
visual field deficit based on patient recollection. Heart rate and blood pressure
were recorded from the initial ED measurement. A murmur was coded as absent
if not mentioned in notation of cardiac auscultation. A bruit was coded as
absent only if specifically noted in the record.
Patients were followed up for 90 days after presentation. Strokes, TIAs,
deaths, and hospitalizations for cardiovascular events were identified from
computerized databases and review of medical records. Hospitalizations outside
the system were recorded in a separate database, so it was possible to obtain
very complete follow-up information.33,34
The primary outcome was stroke occurring within 90 days of TIA presentation.
Stroke was defined as a rapidly developed focal or global disturbance of cerebral
function, with no apparent nonvascular cause, lasting more than 24 hours or
until death,32 and distinguishable from the
event leading to the initial TIA diagnosis. A final stroke diagnosis required
independent confirmation by 2 neurologists, who also determined whether the
stroke led to hospitalization or was disabling (defined as a Modified Rankin
Score of ≥2).35 There were initial disagreements
in 2% of stroke diagnoses, 4% of hospitalization determinations, and 12% of
disability assessments; all disagreements were resolved through discussion.
Diagnosis of recurrent TIA required confirmation by a reviewing neurologist
and a written diagnosis in the medical record. Cardiovascular events requiring
hospitalization included myocardial infarction, unstable angina, congestive
heart failure, and ventricular arrhythmias. Deaths were identified from medical
records, enrollment files, and the Social Security Death Index.
The risks of stroke, recurrent TIA, and cardiovascular events were determined
as the percentage of patients with these events during the 90-day period after
resolution of all TIA symptoms. For patients without documentation of symptom
resolution, follow-up began at the time of ED discharge.
In univariate analysis, we identified clinical signs and symptoms that
were significant risk factors for stroke using the Fisher exact test. Variables
that were associated with stroke risk (at P<.20)
were included in multivariable logistic regression models, and those no longer
associated with risk (at P>.10) were removed in a
stepwise fashion. To simplify the models, continuous variables were dichotomized
at prespecified cutpoints (eg, diastolic blood pressure = 90 mm Hg, systolic
blood pressure = 140 mm Hg). Because specific neurologic signs and symptoms
were often correlated, neurologic abnormalities present either by history
or on examination were combined (eg, weakness by history or on examination).
Variables representing treatments initiated after ED evaluation were not included
in the model since they were likely to be associated with perceived prognosis
and could alter prognosis; these complex interactions would limit their interpretation.
First-order interaction terms with combinations of all independent predictors
were reintroduced into the multivariable model 1 at a time.
Kaplan-Meier life-table analysis was used to illustrate the timing of
follow-up events. Since a stroke occurring after TIA but before evaluation
would not have been included, resulting in an underestimation of stroke risk,
we excluded person-time prior to evaluation in those patients whose symptoms
resolved prior to ED arrival. Follow-up was terminated at death or endarterectomy.
The protocol was approved by the institutional review board of Kaiser-Permanente
Northern California. The risk of study participation was considered minimal,
so informed consent was not required.
Of the 1707 patients with a diagnosis of TIA, more than 99% (n = 1697)
arrived within 1 day of symptom onset. Mean age was 72 years and mean symptom
duration was 207 minutes. The mean heart rate and blood pressure were 78/min
and 161/83 mm Hg, respectively. Risk factors for vascular disease were common,
and included diabetes mellitus, hypertension, and known history of vascular
disease (Table 1). Symptoms were
present on arrival in half of the patients (n = 854).
Strokes occurred in 180 patients (10.5%) within 90 days of TIA presentation,
91 of which occurred during the first 2 days
(Figure 1A). Strokes were fatal in 38 patients (21%), and disabling
in another 115 (64%). Seventy-six percent of patients required hospitalization.
Only 7% of strokes were not disabling and did not require hospitalization.
Age older than 60 years, diabetes mellitus, duration of the TIA, symptoms
on arrival, and signs or symptoms of weakness, speech impairment, and gait
disturbance were associated with increased risk of stroke in univariate analysis,
while symptoms of numbness were associated with reduced risk (Table 1). Medications taken prior to evaluation did not influence
prognosis.
A variety of treatments were initiated in the ED or at discharge (Table 2). Patients prescribed anticoagulation
at discharge from the ED were more likely to have a stroke, but a perception
of greater stroke risk may have prompted more aggressive therapy in these
patients. When analysis was limited to the 918 patients who were not previously
taking an anticoagulant or antiplatelet agent, those initiating an antiplatelet
agent (n = 775) tended to have lower stroke risk than those receiving no prophylactic
medication (n = 143) (9% vs 13%; P = .12). The same
trend was apparent in evaluating any prophylactic therapy vs none (9% vs 13%; P = .17). In multivariable models adjusting for prognostic
factors, neither of these variables was significant (new antiplatelet therapy
vs none: odds ratio [OR], 0.63, 95% confidence interval [CI], 0.37-1.05, P = .10; any new prophylactic therapy vs none: OR, 0.66,
95% CI, 0.38-1.15, P = .14).
Five factors were independently associated with 90-day stroke risk (Table 3). A simple index (1 point for each
risk factor) was useful in estimating risk, which varied from 0% in patients
with no factors to 34% in those with all 5
(Figure 1B), though observations were sparse at the extremes (Table 4).
Models Including First-Order Interactions
The only significant interaction term was one that represented speech
impairment in patients with symptoms lasting more than 10 minutes (OR, 6.0;
95% CI, 1.9-6.5; P = .003). Symptom duration was
no longer a significant predictor of stroke in that model, and speech impairment
alone was associated with a lower risk of stroke (OR, 0.29; 95% CI, 0.09-0.95; P = .04). The magnitude of other predictors was unchanged
in the model with the interaction term.
Hospitalization for cardiovascular events occurred in 44 patients (2.6%)
within 90 days of the TIA. Reasons for admission included congestive heart
failure (n = 22), myocardial infarction (n = 13), ventricular arrhythmia (n
= 5), and unstable angina (n = 4). Forty-five patients (2.6%) died within
90 days of the TIA; causes of death were strokes (n = 20), cardiovascular
events (n = 9), infections (n = 4), cancer (n = 4), pancreatitis (n = 1),
and unknown (n = 7). Recurrent TIA occurred in 216 patients (12.7%), and was
associated with hospitalization in 106 patients (6.2%). An adverse event,
including stroke, cardiovascular hospitalization, death, or recurrent TIA,
occurred in 428 patients (25.1%) in the 90 days after the TIA. More than 50%
of adverse events occurred within the first 4 days.
Validation of the ED Diagnosis
The TIA diagnosis was thought improbable by a reviewing neurologist
in 96 patients (5.6%), in whom symptoms were attributed to syncope (n = 24),
peripheral vestibulopathy (n = 12), anxiety (n = 11), migraine (n = 10), seizure
(n = 6), medication effects (n = 6), neuropathy (n = 4), transient global
amnesia (n = 4), and other etiologies (n = 19). Three of these patients had
a stroke during 90-day follow-up. In an additional 182 patients (10.6%), neurologic
symptoms improved but complete resolution within 24 hours was not documented;
19 of these patients (10.4%) had a stroke during follow-up. The stroke risk
was not significantly different in the full cohort of patients with TIA and
in the subgroup with TIAs confirmed by neurologist review (10.5% vs 11.1%; P = .64 by the Fisher exact test). Compared with the full
cohort, independent risk factors for stroke were similar in those with TIA
diagnoses after neurologist review (age >60 years: OR, 1.8, 95% CI, 1.1-2.8;
diabetes: OR, 2.4, 95% CI, 1.6-3.4; duration >10 min: OR, 2.5, 95% CI, 1.3-4.6;
weakness: OR, 2.2, 95% CI, 1.6-3.2; speech impairment: OR, 1.3, 95% CI, 1.0-1.9).
Transient ischemic attacks are ominous, carrying a substantial short-term
risk of stroke, hospitalization for cardiovascular events, and death. The
90-day stroke risk was 10.5%, over 50 times that expected in a cohort of similar
age.2,36 Half of the strokes occurred
within 2 days of the TIA. Short-term risks of cardiovascular events, death,
and recurrent TIA were also high, with a combined 25.1% risk of an adverse
event during the 3 months after a TIA.
Diagnosis of TIA is often problematic. It may be difficult to determine
whether focal neurologic symptoms are due to ischemia from impaired cerebral
circulation or due to seizure, migraine, or even syncope. Prior studies have
required a neurologist's evaluation and strict entry criteria to establish
the diagnosis of TIA,25,26 but
this does not solve the problem. Even among neurologists there is low interobserver
agreement in the diagnosis of TIA.27,28
Furthermore, evaluation by a neurologist is often not part of routine practice,
in which ED and primary care physicians are primarily responsible for TIA
diagnosis and management. In these settings, the diagnosis of TIA made by
physicians caring for a patient is more useful for assessing natural history
and defining stroke risk factors.29 Therefore,
we focused on the ED diagnosis of TIA. Based on neurologist review, these
diagnoses appeared to be accurate in 94% of patients, although complete resolution
of symptoms was not documented in an additional 11%. The prognosis of patients
whose diagnoses were confirmed by neurologist review was not different from
that of the entire cohort, suggesting that the ED diagnosis of TIA is reasonably
reliable.
The short-term risk that we observed was similar to that found in 2
smaller population-based studies of patients with TIA. Among 198 patients
treated from 1955 to 1969 in Minnesota, the 3-month stroke risk was 10%,25 similar to what we observed. Another study reported
a 4% stroke risk in the first month among 184 patients enrolled a median of
3 days after the index TIA26; failure to include
strokes that occurred during the first 3 days may account for the lower stroke
risk in that study.
We studied patients who were evaluated in EDs of the hospitals of a
health maintenance organization. It is possible that disincentives to using
the ED led to selection of cases perceived to be at greater risk. Patients
had longer duration of symptoms than previously reported for a sample that
included patients identified in outpatient clinics.37
The timing of presentation may also be important. Since more than half of
strokes occurred within 2 days of the TIA, a patient arriving for evaluation
3 days after a TIA has already passed through the period of highest risk.
Further study of patients diagnosed as having TIA in outpatient clinics is
required to extend our findings to that group.
Currently available interventions for patients with TIA, including hospitalization
and urgent carotid artery ultrasound, are expensive10-12
and may not be cost-effective if used in all such patients. Stratification
of stroke risk allows targeting expensive interventions to those at greatest
short-term risk. We found 5 independent risk factors for stroke within 90
days after TIA: age older than 60 years, diabetes mellitus, duration of episode
greater than 10 minutes, and weakness and speech impairment with the episode.
These risk factors may identify patients whose symptoms are more likely due
to cerebral ischemia or may indicate pathophysiologic conditions associated
with greater risk. Combining these risk factors, we identified subgroups with
minimal (0%) and very high (34%) short-term risk of stroke.
Prospective validation with an independent cohort will be required before
these risk factors can be combined into a prediction model.38
Validation also will be needed to test the possibility that the effects of
symptom duration and speech impairment may be interdependent, as we observed
in analyses that included interaction terms. Those analyses suggested that
speech impairment may be a predictor only in those with symptoms lasting more
than 10 minutes, or that symptom duration may be relevant only in those with
speech impairment.
This observational study cannot provide reliable data on efficacy of
therapies for TIA. It is likely that those perceived to be at highest risk
of stroke were more often treated aggressively, and this would bias assessment
of treatment benefit. Also, patients presented with TIAs despite any prophylactic
antiplatelet or anticoagulant medications they were taking, so it is not surprising
that prior medications did not reduce subsequent vascular events. Among patients
who had not previously received an anticoagulant or antiplatelet medication,
there was a nonsignificant trend toward reduced stroke risk in those who then
received a prophylactic agent, but the size of the untreated group was small
(n = 143).
The short-term risk of stroke and other adverse vascular events is high
among patients who present to the ED with a TIA. Urgent intervention in this
group may be warranted, and the efficacy of such intervention should be studied.
Stratification of short-term stroke risk appears practical, and may allow
more cost-effective intervention and more efficient study of prevention strategies.
1.Brown Jr RD, Petty GW, O'Fallon WM.
et al. Incidence of transient ischemic attack in Rochester, Minnesota, 1985-1989.
Stroke.1998;29:2109-2113.Google Scholar 2.Broderick J, Brott T, Kothari R.
et al. The Greater Cincinnati/Northern Kentucky Stroke Study.
Stroke.1998;29:415-421.Google Scholar 3.Williams GR, Jiang JG, Matchar DB, Samsa GP. Incidence and occurrence of total (first-ever and recurrent) stroke.
Stroke.1999;30:2523-2528.Google Scholar 4.National Stroke Association. TIA/ mini strokes: public knowledge and experience–Roper Starch
worldwide survey. Roper Starch Worldwide; 2000.
5.Brainin M, McShane LM, Steiner M.
et al. Silent brain infarcts and transient ischemic attacks.
Stroke.1995;26:1348-1352.Google Scholar 6.Wolf PA, Clagett GP, Easton JD.
et al. Preventing ischemic stroke in patients with prior stroke and transient
ischemic attack.
Stroke.1999;30:1991-1994.Google Scholar 7.Adams Jr HP, Brott TG, Crowell RM.
et al. Guidelines for the management of patients with acute ischemic stroke.
Stroke.1994;25:1901-1914.Google Scholar 8.Albers GW, Hart RG, Lutsep HL.
et al. Supplement to the guidelines for the management of transient ischemic
attacks.
Stroke.1999;30:2502-2511.Google Scholar 9.Feinberg WM, Albers GW, Barnett HJ.
et al. Guidelines for the management of transient ischemic attacks.
Circulation.1994;89:2950-2965.Google Scholar 10.Health Care Financing Administration. Medicare program; changes to the hospital inpatient prospective payment
systems and fiscal year 1997 rates—HCFA. Final rule.
61 Federal Register.46166 (1996).Google Scholar 11.Oster G, Huse DM, Lacey MJ, Epstein AM. Cost-effectiveness of ticlopidine in preventing stroke in high-risk
patients.
Stroke.1994;25:1149-1156.Google Scholar 12.Hankey GJ, Warlow CP. Cost-effective investigation of patients with suspected transient ischaemic
attacks [editorial].
J Neurol Neurosurg Psychiatry.1992;55:171-176.Google Scholar 13.Kernan WN, Horwitz RI, Brass LM.
et al. A prognostic system for transient ischemia or minor stroke.
Ann Intern Med.1991;114:552-557.Google Scholar 14.Hankey GJ, Slattery JM, Warlow CP. The prognosis of hospital-referred transient ischaemic attacks.
J Neurol Neurosurg Psychiatry.1991;54:793-802.Google Scholar 15.Muuronen A, Kaste M. Outcome of 314 patients with transient ischemic attacks.
Stroke.1982;13:24-31.Google Scholar 16.Sorensen PS, Marquardsen J, Pedersen H.
et al. Long-term prognosis and quality of life after reversible cerebral ischemic
attacks.
Acta Neurol Scand.1989;79:204-213.Google Scholar 17.Calandre L, Bermejo F, Balseiro J. Long-term outcome of TIAs, RINDs and infarctions with minimum residuum.
Acta Neurol Scand.1990;82:104-108.Google Scholar 18.Canadian Cooperative Study Group. A randomized trial of aspirin and sulfinpyrazone in threatened stroke.
N Engl J Med.1978;299:53-59.Google Scholar 19.Dutch TIA Trial Study Group. A comparison of two doses of aspirin (30 mg vs. 283 mg a day) in patients
after a transient ischemic attack or minor ischemic stroke.
N Engl J Med.1991;325:1261-1266.Google Scholar 20.EAFT Study Group. Secondary prevention in non-rheumatic atrial fibrillation after transient
ischaemic attack or minor stroke.
Lancet.1993;342:1255-1262.Google Scholar 21.UK-TIA Study Group. The United Kingdom transient ischaemic attack (UK-TIA) aspirin trial:
final results.
J Neurol Neurosurg Psychiatry.1991;54:1044-1054.Google Scholar 22.Heyman A, Burch JG, Rosati R.
et al. Use of a computerized information system in the management of patients
with transient cerebral ischemia.
Neurology.1979;29:214-221.Google Scholar 23.Heyman A, Wilkinson WE, Hurwitz BJ.
et al. Risk of ischemic heart disease in patients with TIA.
Neurology.1984;34:626-630.Google Scholar 24.Hankey GJ, Slattery JM, Warlow CP. Transient ischaemic attacks: which patients are at high (and low) risk
of serious vascular events?
J Neurol Neurosurg Psychiatry.1992;55:640-652.Google Scholar 25.Whisnant JP, Matsumoto N, Elveback LR. Transient cerebral ischemic attacks in a community: Rochester, Minnesota,
1955 through 1969.
Mayo Clin Proc.1973;48:194-198.Google Scholar 26.Dennis M, Bamford J, Sandercock P, Warlow C. Prognosis of transient ischemic attacks in the Oxfordshire Community
Stroke Project.
Stroke.1990;21:848-853.Google Scholar 27.Koudstaal PJ, Gerritsma JG, van Gijn J. Clinical disagreement on the diagnosis of transient ischemic attack.
Stroke.1989;20:300-301.Google Scholar 28.Kraaijeveld CL, van Gijn J, Schouten HJ, Staal A. Interobserver agreement for the diagnosis of transient ischemic attacks.
Stroke.1984;15:723-725.Google Scholar 29.Kothari RU, Brott T, Broderick JP, Hamilton CA. Emergency physicians: accuracy in the diagnosis of stroke.
Stroke.1995;26:2238-2241.Google Scholar 30.Johnston SC, Smith WS. Practice variability in management of transient ischemic attacks.
Eur Neurol.1999;42:105-108.Google Scholar 31.Krieger N. Overcoming the absence of socioeconomic data in medical records.
Am J Public Health.1992;82:703-710.Google Scholar 32.World Health Organization MONICA Project Principal Investigators. The World Health Organization MONICA Project (monitoring trends and
determinants in cardiovascular disease).
J Clin Epidemiol.1988;41:105-114.Google Scholar 33.Selby JV, Fireman BH, Lundstrom RJ.
et al. Variation among hospitals in coronary-angiography practices and outcomes
after myocardial infarction in a large health maintenance organization.
N Engl J Med.1996;335:1888-1896.Google Scholar 34.Selby JV, Fireman BH, Swain BE. Effect of a copayment on use of the emergency department in a health
maintenance organization.
N Engl J Med.1996;334:635-641.Google Scholar 35.van Swieten JC, Koudstaal PJ, Visser MC.
et al. Interobserver agreement for the assessment of handicap in stroke patients.
Stroke.1988;19:604-607.Google Scholar 36.Brown RD, Whisnant JP, Sicks JD.
et al. Stroke incidence, prevalence, and survival: secular trends in Rochester,
Minnesota, through 1989.
Stroke.1996;27:373-380.Google Scholar 37.Dyken ML, Conneally M, Haerer AF.
et al. Cooperative study of hospital frequency and character of transient
ischemic attacks.
JAMA.1977;237:882-886.Google Scholar 38.Laupacis A, Sekar N, Stiell IG. Clinical prediction rules: a review and suggested modifications of
methodological standards.
JAMA.1997;277:488-494.Google Scholar