Timing of the Evaluation of TIAs

Over the past 10 years, a number of studies have established a one-week risk for CVA of up to 10% after a TIA, with up to half of those CVAs occurring in the first 2 days.  Given this high risk of completed CVA, we sought to examine 3 questions:  is there an accurate and simple clinical scoring system which helps identify patients at high risk of early CVA after TIA?  Does urgent evaluation and treatment of TIA improve clinical outcomes and decrease the risk of CVA after TIA?  Is an ED observation unit accelerated diagnostic TIA protocol feasible, efficient, and cost-effective?
1.  Asimos AW et al. A Multicenter Evaluation of the ABCD2 Score’s Accuracy for Predicting Early Ischemic Stroke in Admitted Patients With Transient Ischemic Attack. Annals of Emergency Medicine. 2010;55:201-210.
This large multicenter study of 1667 patients evaluated the ability of the ABCD2 score (age, BP, clinical features, duration, diabetes) to predict 7 day risk of CVA in patients admitted to the hospital within 24 hours after TIA.  This was a convenience sample, and although billed as a prospective study, actually retrospectively examined the patients’ charts.  Therefore ABCD2 scores were unavailable for almost 35% of patients.  Although they attempted to impute (estimate and fill in) missing data based on other observed variables, this still limits the study’s internal validity.  Twenty-three percent of patients were diagnosed with CVA within 7 days,  in part reflecting that the study only included admitted patients (many minor TIAs sent home).  Most CVAs occurred in the first 2 days.  The c statistic (=area under the ROC curve) was 0.59 for the risk of any ischemic CVA in 7 days, and 0.71 for disabling CVA within 7 days.  In other words, the ABCD2 score poorly predicted the risk of ischemic CVA, and was better but still not great at predicting disabling CVA.  The sensitivity of a low ABCD2 score (</= 3) was only 87% for identifying patients at low risk for CVA in 7 days (sensitivity was 96% identifying low risk for disabling CVA, CI 88-99%).  
Their definition of “disabling CVA” was Modified Rankin Scale score of greater than 2, which implies a degree of dependence.  Point was made that a MRS score of 2, although “mild,” still means that a pt is unable to do everything they were capable of before the CVA, eg maybe an EP couldn’t practice emergency medicine.  For us, that would be a pretty devastating CVA, so most people in the room wanted a score that is really good at predicting all CVAs, not just “disabling” ones.  Other points of conversation:  tremendous variability in the admission rates for TIA in the participating hospitals (35-100%) and no standardized work-up or treatment plans-really heterogeneous study.  They chose patient oriented outcomes (clinical TIA/CVA) rather than disease oriented outcomes (MRI findings) which is laudable, but there is an inherent difficulty in defining TIA vs. CVA early in the presentation (are some early TIAs really CVAs…probably yes).  We also don’t know when patients received MRIs and this is a problem since the natural history of MRI changes very early in TIA/CVA has yet to be well defined. 
Bottom line:  ABCD2 score not sufficiently accurate in this study to use it to predict short-term risk of CVA.  Also not sensitive enough to use a low ABCD2 score to identify patients who can be sent home (would miss about 10% of early CVAs in this study).  However, it is definitely true that as the ABCD2 score goes up, the pt’s risk of CVA increases,  so calculating the ABCD2 score is useful to more accurately counsel high scoring patients about their risk of CVA, encourage lifestyle modification, and these patients should be relatively easy to admit after discussion with PMD.
2.  Rothwell PM et al. Effect of urgent treatment of transient ischaemic attack and minor stroke on early recurrent stroke (EXPRESS study): a prospective population-based sequential comparison. Lancet. 2007;370:1432-1142.
In this prospective before-and-after English study of 1278 patients presenting with TIA or minor stroke, the authors compared their local  practice of having PMDs refer patients to a daily TIA clinic, which was appointment based, with no treatment immediately started (Phase I), to treatment at a TIA clinic where no appointments were necessary, and at which treated was immediately initiated if the diagnosis of TIA or small stroke was confirmed (Phase II).  All patients received neuroimaging and carotid US, some patients received Echo.  Patients were followed up for 24 months (100% followup), and the primary outcome was CVA within 90 days.  Risk of CVA at 90 days fell from 10.3% in Phase I to 2.1% in Phase 2 (huge).  Early treatment did not increase the risk of ICH or other bleeding.  Although this is an extremely positive study, it’s important to look at the access to care timing.  Median delay in assessment in the TIA study clinic fell from 3 days in Phase I to less than 1 day in Phase II.  Also, median delay to first prescription of treatment (ASA, clopidogrel, BP meds) fell from 20 days to 1 day.  So, it’s difficult to extrapolate this to the United States, where patients are seen/have a CT/and are started on ASA or other meds at the time of their presentation to the ED (not 20 days later).  Maybe some of the great outcomes in this study were because their Phase I was so slow.  The authors acknowledge that starting meds early likely was the largest factor in their positive results, although patients in Phase II requiring CEA received surgery earlier than in Phase I. 
Also, there is the possibility of the Hawthorne effect coming into play.  As an example, patients in Phase II were more likely to be on statins at the time of initial presentation than patients in Phase I, and it’s possible that other minor subtle changes in care were taking place during the second phase which influenced the overall outcomes of these patients and added to the positive results. 
Bottom line:  Early diagnosis and treatment of TIA is associated with improved clinical outcome, but it's unknown how much of their huge benefit was a reflection of the inefficiencies of their system at baseline.
3.  Ross MA et al. An Emergency Department Diagnostic Protocol for Patients With Transient Ischemic Attack: A Randomized Controlled Trial. Annals of Emergency Medicine. 2007;50:109-119.
Finally, this study examines the efficiency and potential cost savings of an ED Observation Unit based accelerated diagnostic protocol (ADP) for TIA.  The study randomized 149 patients with TIA either to an ADP which included cardiac monitoring, carotid dopplers, echo, neuro checks and a neuro consult, or to hospital admission.  The same order set was used for ADP patients and admitted patients (although some admitted patients never receive all of the tests).   Primary outcome was the index visit length of stay.  Secondary outcomes were 90 day cost, and 90 day clinical outcomes (which included CVA).  Ninety day followup occurred with all patients.  There were a large number of inclusion/exclusion criteria to be enrolled, resulting in many patients with TIA not being enrolled in the study.  However, when comparing the 2 groups, patients in the ADP had a significantly shorter length of stay than admitted patient (30 hours shorter-basically saved a day), and 90 day costs were $890 versus $1,547.  Approximately 85% of ADP patients were discharged.  Clinical outcomes were similar, with comparable rates of return visits, CVA, and major clinical events.  All ADP admissions were for clinical events detected on serial clinical exams; no admissions were primarily because of carotid stenosis, arrhythmia, or echo findings.  That being said, it’s accepted that echo/carotid US findings which lead to emergent interventions are unusual (eg only 3% echo findings of cardioembolic source of CVA/TIA in the absence of clinical suspicion of cardiac etiology-2009 AHA Stroke Guidelines), and this study only enrolled 174 relatively low risk patients.  Not really powered to find significant clinical outcomes differences.  If significant carotid stenosis is identified, early surgical intervention is associated with improved outcomes.  Carotid US and Echo were performed more frequently and more quickly in the ADP group-need a larger study to see if this would make a clinical outcome difference.
A couple of other philosophical points.  Assigning hospital resources to a TIA protocol and prioritizing these patients to receive rapid diagnostic evaluations and neurology consults may mean diversion of resources from other patients in the ED/hospital.  On the other hand, rapid throughput/discharge of ADP patients allows backfill of open beds upstairs with patients needing different/more specialized resources.  There is a potential for overuse of this protocol-enrolling such atypical patients that it becomes a “no-risk” rather than “low-risk” protocol.   Point made that there are fewer TIA presentations than chest pain presentations to the ED, so decent chance this wouldn’t be a problem.
Bottom line:  an ED accelerated diagnostic protocol appears to be feasible and when enrolling low risk patients saves a day of hospital admission and significant money, with similar clinical outcomes.