What does this guideline say?

Bob: If you thought you knew what a transient ischemic attack (TIA) is, and how to evaluate for it, think again. This scientific statement snuck past a lot of us and its implications are far-reaching.

The guideline includes three game changers for all physicians:

• It revises the definition of TIA

• It recommends that magnetic resonance imaging (MRI) be performed within 24 hours in patients presenting with TIA

• It recommends that the ABCD² (age, blood pressure, clinical features, duration, diabetes mellitus) score be used in the evaluation of TIA

Let's start with the change in definition. Many of us learned the classic 24-hour definition: if neurologic symptoms resolve within 24 hours of onset, it is considered a TIA. In 2002, the TIA Working Group recognized that an ischemic episode that improves at 23 hours is not biologically different from one that improves at 25 hours. Simply put, there was nothing magical about the 24-hour cutoff. The TIA Working Group recommended changing the definition to the following: a brief episode of neurologic dysfunction caused by focal brain or retinal ischemia, with clinical symptoms typically lasting less than one hour and without evidence of acute infarction.¹

The big change in the current guideline is the elimination of reference to duration of symptoms. In other words, the “less than one hour” is gone. This is because one-third of patients whose symptoms abate within one hour have evidence of acute stroke on diffusion-weighted MRI.

The new definition of TIA also states that there must be no evidence of acute infarction. This leads into the second recommendation, that patients with suspected TIA should preferably undergo diffusion-weighted MRI as a means of ruling out infarction within 24 hours of symptom onset.

Lastly, the guideline recommends that the ABCD² score (Table 1² ) be used to determine which patients should be hospitalized: patients with a score of 3 or greater and patients with lower scores who may not follow up within two days.

What should we make of this guideline?

Bob: I don't have a major problem with the change in definition. I understand that the authors of the guideline want to focus more on the pathophysiology of the disease than on the temporal course of symptoms. This is analogous to a patient who presents with chest pain—a biomarker (typically, troponin) is used to distinguish angina from myocardial infarction. In this guideline, the authors want to use MRI to distinguish between a TIA and a cerebrovascular accident. This sounds good in theory, but MRI is different than a blood test. And there is a timing issue—how many institutions can get immediate MRI, or even MRI within 24 hours of symptom onset? How many insurance companies allow this? Even if you can get an MRI within the time-frame recommended, does it make any difference in the long-term outcome of the patient who has returned to baseline?

Mark: The authors like MRI because it detects more lesions (i.e., infarcts) than computed tomography. But just because a small cerebrovascular accident was uncovered, will it lead to a change in care (better or worse)? The authors of this guideline note that patients with a stroke diagnosed with MRI have a greater risk of having a more disabling subsequent stroke than those with a normal MRI result. Again, although this may be interesting prognostic information, there is no evidence that an infarct identified on MRI will lead to more aggressive medical/surgical treatment or to better outcomes. This is another recommendation based on disease-oriented evidence, and not on patient-oriented evidence that matters.

Andrea: Expanding the use of MRI also will invariably lead to the documentation of more strokes, thereby accelerating cerebrovascular accident rates to much higher levels than previously reported. This is the dilemma of having sensitive radiologic studies—they identify more lesions, but does this lead to improved outcomes? Case in point: from 1993 to 1998, the rate of pulmonary embolism diagnosis remained stable, but this rate nearly doubled in the eight years following the introduction of computed tomography diagnosis for pulmonary embolism. However, mortality rates did not improve with computed tomography diagnosis, and the rates of anticoagulation-related complications increased by 71 percent.³

I believe we are going to have to face up to this coming challenge: when increasingly sophisticated imaging technology enables us to see inside vessels millimeters in diameter and detect a clot or a few ischemic cells, what are we going to do? Will this actually represent disease, or will it represent incidentalomas or normal aging? Will treatment provide any benefit or just potential harm?

Bob: That brings us to the second recommendation in the guideline: using the ABCD² score to decide on hospitalization. The ABCD² scoring system was developed to assess the short-term risk of TIA progressing to a complete cerebrovascular accident. It is a combination of two earlier scoring systems: the California score, created by the Kaiser Group in 2000, and the ABCD score developed in Great Britain in 2005. The composite ABCD² score was derived and validated in 2007 (Table 1).²

Mark: Be careful here. The authors of the ABCD² study say they “validated” their scoring system. How did they do that? They retrospectively applied this new score to a previous population of patients they themselves studied. In a previous AFP Journal Club on the San Francisco Syncope Rule, we made the point that before a clinical decision rule can be used, it needs to be externally validated.⁴ In other words, it needs to be “road tested” by other investigators in a completely new population before widespread use.

Bob: Interestingly, many other researchers have subsequently tried to assess the prognostic accuracy of the ABCD² score. Some find it accurate, whereas others don't.⁵ The problem with all of these follow-up studies is their heterogeneity—some include emergency department patients only, some include outpatients, some are retrospective, and some are prospective. I don't think there is a definitive answer as to the prognostic accuracy of the score.

Andrea: I am even more troubled by the recommendation that the ABCD² score be used to decide whether a patient should be hospitalized. This recommendation has far-reaching consequences. If you were to hospitalize patients with a score of at least 3 points in the original Kaiser study, 92 percent of all patients with TIA would be hospitalized.⁶ This recommendation is frustrating because there has never been a study showing that admission decisions based on ABCD² scores result in better outcomes. The recommendation is based entirely on hypothesis and no evidence.

Bob: That is the problem you have every time level C evidence is reported. There is no evidence—it is just the opinion of the authors of the guideline.

Unfortunately, level C evidence recommendations are commonplace. A review of 16 current American Heart Association/American College of Cardiology guidelines produced from 1984 to 2008 revealed that only 11 percent of recommendations were based on level A evidence (i.e., derived from multiple randomized trials or meta-analyses), whereas 48 percent were level C (expert opinion).⁷

What should the family physician do?

Mark: Like Bob, I don't have a major problem with the change in the definition of TIA. It is just unfortunate that we are being held to recommendations to obtain studies (in this case, MRI) and use scoring systems (ABCD²) without any evidence that their use improves patient outcomes.

Andrea: Although the promise of guidelines is to deliver well-proven strategies of medical care, we have all too often seen these well-intentioned documents run amok. Here's an example: give antibiotics within four hours of arrival for patients with pneumonia. Any evidence? No. Subsequently this recommendation was withdrawn after years of driving ourselves crazy to meet this standard.

Bob: How organizations develop clinical guidelines has come under recent scrutiny. In an upcoming AFP Journal Club, we will take a look at how two organizations looking at the same question can publish guidelines with differing answers.