In Reply:

We appreciate the thoughtful comments of Dr. Kohn on our recent study describing the test properties of the “HINTS” decision rule compared to ABCD2 and magnetic resonance imaging (MRI).[1] Dr. Kohn makes several important points worthy of further discussion. First, he suggests ABCD2 was not designed to diagnose posterior circulation stroke, so our primary results showing its lack of diagnostic accuracy are not surprising. We concur that our result was “predictable”—in fact, it was our prespecified hypothesis. It was not, however, a mere straw man, since a large, published study of cerebrovascular causes of emergency department (ED) dizziness suggested the ABCD2 rule be used for this purpose.[2]

Dr. Kohn notes that our acute vestibular syndrome (AVS) population is not “typical” of ED dizziness. This is absolutely true and crucial to correct clinical use of HINTS for diagnosis and correct interpretation of our study results. Dr. Kohn suggests the need for “validation for use by emergency physicians in a more typical population of ED patients with dizziness.” Unfortunately, the key message for clinicians is obscured by conflating two different types of nonrepresentativeness: 1) AVS patients are not typical of all ED patients presenting dizziness and 2) our particular AVS patients were at especially high risk for stroke, so not representative of all patients with AVS. The former of these relates to how HINTS should be used clinically (i.e., only in AVS), while the latter relates to how our study results should be interpreted when applied in AVS.

Acute vestibular syndrome is a well-defined clinical syndrome of new, continuous, persistent dizziness or vertigo with associated vestibular features, lasting days to weeks.[3] AVS accounts for only about 10% to 20% of the now 4.1 million[4] ED dizziness presentations each year in the United States, but the vast majority of “dizzy” strokes present in this fashion.[3] The AVS clinical presentation, typical of vestibular neuritis, labyrinthitis, and stroke, differs substantially from the majority of ED dizziness presentations that have brief, repetitive episodes of dizziness, typical of benign paroxysmal positional vertigo, orthostatic hypotension, and cardiac arrhythmia.[5] Applying the HINTS clinical decision rule to this latter group with transient dizziness would be totally inappropriate and conceptually similar to using an electrocardiogram to rule out cardiac angina in a patient who had intermittent chest pain in the prior week, but was presently asymptomatic. Dr. Kohn states that the AVS presentation alone is “generally sufficient to justify hospitalization from the ED.” We concur that admitting all AVS patients would reduce missed stroke in ED dizziness. Unfortunately, such an approach would not be cost-effective,[6] likely increasing the costs of care for ED dizziness by more than $2 billion annually in the United States through greater hospital admission for those with inner ear disease presenting as AVS. By contrast, the correct application of HINTS to guide subsequent decision-making in AVS would be highly cost-effective.[6]

Dr. Kohn rightly points out that our AVS patients were at especially high risk for stroke (with at least one vascular risk factor required). We took this approach because in high-risk patients, we could justify inpatient admission, stroke MRI, and repeat delayed MRIs on clinical grounds, but we did not have funds to apply the same rigorous testing to a lower-risk AVS population. This raises an important question about the generalizability of our findings to AVS patients without vascular risk factors. The inclusion of a large number of stroke patients with a wide clinical and pathologic illness spectrum (large infarcts to small, old to young, neurologic signs present to absent) makes it unlikely that our estimates of stroke sensitivity (99.1%) are artificially inflated.[7] By contrast, because some low-risk patients were excluded, the measured specificity for peripheral causes (97.0%) might be influenced by an overly narrow illness spectrum.[7] Although sensitivity and specificity can also vary purely as a function of disease prevalence,[8] this difference is small as long as the measurement error (intraindividual variation in test results due to examiner or patient) is small, particularly when considering variation in prevalence over the range relevant here (~60% strokes in our AVS population vs. ~25% in the broader group of AVS patients to whom we might wish to generalize the HINTS rule).[8] This measurement reliability issue suggests that examiner skill is important, especially when the HINTS approach is applied to a general AVS population. Nevertheless, it would take a highly error-prone HINTS measure for its stroke sensitivity to fall below that of MRI (86%[1]) even in an AVS population with stroke prevalence as low as 25%.[8]

Finally, Dr. Kohn suggests that most patients with dizziness do not have stroke and that most stroke patients can be readily detected by routine neurologic assessments, concluding that the HINTS approach should be reserved for only the “the most difficult to diagnose stroke patients.” It is true that only about one in 20 ED dizziness patients has a cerebrovascular cause, although roughly one in four presenting with AVS has a posterior circulation stroke.[3] Contrary to conventional clinical wisdom, obvious focal neurologic signs (e.g., hemiparesis, gaze palsy) are present in fewer than 20% of strokes.[3] In an AVS patient without hearing loss (pretest probability of stroke ~25%), the posttest probability of stroke would only be reduced to ~21% after a nonfocal neurologic exam, ~20% after a negative CT, and ~7% after a negative early MRI. In contrast, the residual risk of stroke would be ~0.3% for a peripheral-appearing HINTS exam in the same patient.

Emergency physicians should seek to gain competence performing the HINTS exam in AVS. Such exams may soon be facilitated by the use of portable eye movement recording devices.[9] Absent skills or equipment, AVS patients should routinely undergo MRI (not CT) and be cautioned to return if symptoms worsen or evolve, with a low threshold for repeat MRI or hospital admission for stroke workup, given the high false-negative rate for MRI in the first 48 hours after AVS onset.

Case definition is critical. HINTS will work best in AVS with spontaneous or gaze-evoked nystagmus. HINTS should not be applied to ED patients with transient or episodic dizziness, for whom different bedside tests (e.g., Dix-Hallpike, orthostatic vital signs) are appropriate.


  1. Top of page
  2. References