Transient isolated brainstem symptoms preceding posterior circulation stroke: a population-based study

Summary Background Transient isolated brainstem symptoms (eg, isolated vertigo, dysarthria, diplopia) are not consistently classified as transient ischaemic attacks (TIAs) and data for prognosis are limited. If some of these transient neurological attacks (TNAs) are due to vertebrobasilar ischaemia, then they should be common during the days and weeks preceding posterior circulation strokes. We aimed to assess the frequency of TNAs before vertebrobasilar ischaemic stroke. Methods We studied all potential ischaemic events during the 90 days preceding an ischaemic stroke in patients ascertained within a prospective, population-based incidence study in Oxfordshire, UK (Oxford Vascular Study; 2002–2010) and compared rates of TNA preceding vertebrobasilar stroke versus carotid stroke. We classified the brainstem symptoms isolated vertigo, vertigo with non-focal symptoms, isolated double vision, transient generalised weakness, and binocular visual disturbance as TNAs in the vertebrobasilar territory; atypical amaurosis fugax and limb-shaking as TNAs in the carotid territory; and isolated slurred speech, migraine variants, transient confusion, and hemisensory tingling symptoms as TNAs in uncertain territory. Findings Of the 1141 patients with ischaemic stroke, vascular territory was categorisable in 1034 (91%) cases, with 275 vertebrobasilar strokes and 759 carotid strokes. Isolated brainstem TNAs were more frequent before a vertebrobasilar stroke (45 of 275 events) than before a carotid stroke (10 of 759; OR 14·7, 95% CI 7·3–29·5, p<0·0001), particularly during the preceding 2 days (22 of 252 before a vertebrobasilar stroke vs two of 751 before a carotid stroke, OR 35·8, 8·4–153·5, p<0·0001). Of all 59 TNAs preceding (median 4 days, IQR 1–30) vertebrobasilar stroke, only five (8%) fulfilled the National Institute of Neurological Disorders and Stroke (NINDS) criteria for TIA. The other 54 cases were isolated vertigo (n=23), non-NINDS binocular visual disturbance (n=9), vertigo with other non-focal symptoms (n=10), isolated slurred speech, hemisensory tingling, or diplopia (n=8), and non-focal events (n=4). Only 10 (22%) of the 45 patients with isolated brainstem TNAs sought medical attention before the stroke and a vascular cause was suspected by their physician in only one of these cases. Interpretation In patients with definite vertebrobasilar stroke, preceding transient isolated brainstem symptoms are common, but most symptoms do not satisfy traditional definitions of TIA. More studies of the prognosis of transient isolated brainstem symptoms are required. Funding Wellcome Trust, UK Medical Research Council, Dunhill Medical Trust, Stroke Association, National Institute for Health Research (NIHR), Thames Valley Primary Care Research Partnership, and the NIHR Biomedical Research Centre, Oxford.


Introduction
The diagnosis of transient ischaemic attack (TIA) is often based purely on the description of symptoms, particularly in patients with vertebrobasilar events in whom positive diff usion weighted imaging is less common 1,2 and interobserver agreement in diagnosis by experienced neurologists is only moderate. 3,4 Rapid access to secondary prevention after TIA reduces the risk of early recurrent stroke 5,6 and is recommended in clinical guidelines. [7][8][9][10][11][12][13] However, a false-positive diagnosis of TIA can adversely aff ect patient confi dence, employment, lifestyle, and insurance costs and can result in unnecessary preventative treatments. Neverthe less, missing the diagnosis means that the opportunity to prevent a disabling stroke is lost. Indeed, evidence exists that apparently non-TIA transient neurological attacks (TNAs) are associated with an increased risk of stroke, coronary events, and dementia. [14][15][16] The inconsistency in clinical diagnosis of vertebrobasilar TIA is due mainly to uncertainty about the nature of several transient isolated brainstem symptoms. Most clinical guide lines do not specify which of these symptoms should be deemed as TIA, [7][8][9][10][11][12][13] although some cite the National Institute of Neurological Disorders and Stroke (NINDS) criteria. 17,18 The NINDS criteria state that most transient isolated brainstem symptoms, such as vertigo, dysarthria, dizziness or wooziness, focal symptoms suggestive of migraine, confusion, and amnesia, should not be defi ned as TIAs, 19 although many clinicians take a more fl exible view in routine practice. Previous studies of the generality of TIAs or TNAs have not looked specifi cally at isolated transient brainstem symptoms and there have been no prospective unselected population-based studies of incidence and prognosis on which to base clinical decisions. Indeed, such studies would be diffi cult because many (probably most) patients with transient neurological symptoms do not report them to medical attention. However, one useful epidemiological approach, which was helpful in confi rming the high risk of stroke after clinically defi nite TIA, 20 is to determine the frequency of events during the days and weeks immediately preceding a stroke. This temporal relation can be estimated fairly reliably because most patients do present to medical attention after a stroke, 20 and it also focuses on the most clinically important subset of TNAs-ie, those that are associated with impending stroke, irrespective of whether the individual sought medical attention after the warning symptoms. We therefore aimed to assess the frequency of isolated brainstem TNAs preceding vertebrobasilar stroke and to compare the frequency with the background rate of such symptoms in patients with stroke in the carotid territory.

Study design and patients
The Oxford Vascular Study (OXVASC) is a prospective, population-based study of all stroke and TIA in 91 105 individuals of all ages registered with 63 primary care physicians in Oxfordshire, UK. We restricted analysis to patients with a fi rst stroke presenting to medical attention from April 1, 2002, to March 31, 2010.
The study methods have been described elsewhere. 21,22 Briefl y, various overlapping methods of hot and cold pursuit were used to achieve near complete ascertainment of all individuals presenting to medical attention with TIA or stroke. 21,22 These included: (1) An open-access TIA service available 7 days per week to which participating general practitioners and the local accident and emergency department send all individuals with suspected TIA or minor stroke whom they would not usually admit to hospital, with emergency provision at weekends supplementing a weekday clinic; (2) daily searches of admissions to the stroke unit, general medical, neurology, and other relevant wards; (3) daily searches of the local accident and emergency department and eye hospital attendance register; (4) monthly computerised searches of general practitioner diagnostic coding and hospital discharge codes; (5) monthly searches of all cranial and carotid imaging studies done in local hospitals; and (6) monthly reviews of all death certifi cates and coroners' reports.
OXVASC was approved by our local ethics committee (Oxfordshire Ethics Committee A: 05/Q1604/70). All patients provided written informed consent and were seen by study physicians as soon as possible after their initial presentation. Relatives pro vided written assent and clinical details for those unable to provide written consent.

Procedures
The study senior neurologist (PMR) reviewed all cases and classifi ed them as stroke or other disorder using standard defi nitions. 21 Severity of event was assessed with the National Institutes of Health stroke scale (NIHSS) 23 and clinical features. We classifi ed events as minor stroke if there was a focal neurological defi cit lasting more than 24 h and an NIHSS score of 5 or lower 5 at time of assessment by a study physician. We classifi ed vascular territory (carotid or vertebrobasilar) of stroke on the basis of brain imaging and by use of the NINDS criteria applied to clinical features of the event. 19 We excluded strokes of uncertain vascular territory. Vascular assessment was done with CT, MRI, or catheter angiography as clinically appropriate.
A detailed history about preceding events was obtained from every patient with a standardised questionnaire. [21][22] This questionnaire included date of symptom onset, duration and type of symptoms, baseline characteristics, and time of fi rst seeking medical attention. All patients were asked about TIA or TNA symptoms during the previous 90 days. Additionally, we searched hospital notes and family doctors' records for TIA or TNA symptoms in the previous 90 days for all patients. We prospectively classifi ed TNAs using a predefi ned, diagnostic classifi cation system. We classi fi ed isolated vertigo, vertigo with non-focal symptoms, isolated double vision, transient generalised weakness, and binocular visual disturbance as TNAs in the vertebrobasilar territory. Binocular visual disturbance includes patients with partial or complete visual fi eld loss that is not suffi ciently clear cut to be a defi nite TIA; this category would not include a homonymous hemianopia or quadrantinopia but does include patients with lone bilateral blindness, lone bilateral visual blurring, visual scrolling, or other unusual visual perceptions that are not clearly related to migraine or other non-vascular cause. These cases diff er from migraine in that so-called positive  symptoms (such as fortifi cation spectra or photopsia) will not be prominent. Atypical amaurosis fugax (atypical monocular visual disturbance, such as isolated visual blurring, a sensation of looking through water) and limbshaking episodes (presumed to be due to low fl ow in the ipsilateral carotid artery) were classifi ed as TNAs in the carotid territory. Isolated slurred speech, migraine variants, transient confusion, and hemisensory tingling symp toms were classifi ed as TNAs in uncertain territory. All TNAs were further classifi ed according to whether they fulfi lled NINDS TIA criteria. 19

Statistical analysis
We compared baseline characteristics of patients with stroke in the vertebrobasilar territory (vertebrobasilar stroke) versus those with stroke in the carotid territory (carotid stroke) using the χ² test for categorical variables and t test for age. Where signifi cant diff erences were noted, appropriately stratifi ed analyses were done. We used the χ² test to compare the frequency and type of TNAs reported during the 90 days preceding a vertebrobasilar stroke versus a carotid stroke and to compare the frequency of events preceding stroke with or without symptomatic arterial stenosis. Given that some patients experience multiple TNAs before stroke, to avoid double counting, analysis was based on the fi rst TNA or TIA during the 90 days preceding stroke, rather than the most recent TNA or TIA. We assumed an arbitrary level of 5% signifi cance.

Role of the funding source
The funding sources had no role in study design, collection, analysis, and interpretation of data, writing of the report, or the decision to submit for publication. The corresponding author had full access to all of the data in the study and had fi nal responsibility for the decision to submit for publication.

Results
Of 1141 consecutive patients with defi nite stroke, vascular territory could be identifi ed in 1034 (91%). Of these, 742 (72%) patients had a minor stroke and 292 (28%) had a major stroke. 759 (73%) patients had a stroke in the carotid territory and 275 (27%) in the vertebrobasilar territory (table 1). Patients with carotid stroke were older than those with vertebrobasilar stroke, but other baseline characteristics were similar (table 1). Events (TIAs and other TNAs) were more common in the 90 days preceding a vertebrobasilar stroke than before carotid stroke (in 59 of 275 patients before a vertebrobasilar stroke vs 65 of 759 patients before a carotid stroke; OR 2·92, 95% CI 1·99-4·28, p<0·0001). Of 275 patients with vertebrobasilar stroke, 59 (22%, 95% CI 16·6-26·4) had at least one event within the preceding 90 days (median 4 days, IQR 1-30) of which only fi ve (8%) had events that fulfi lled the NINDS criteria for TIA (fi gure). In the remaining 54 patients, 45 had isolated brainstem TNAs (23 patients with isolated vertigo, nine with non-NINDS binocular visual disturbance, ten with vertigo with other non-NINDS symptoms, two with isolated diplopia, and one with transient generalised weakness), and nine had TNAs of uncertain territory (four patients with isolated slurred speech, two with isolated hemisensory symptoms, two with transient confusion, and one with a migraine variant). In seven (13%) of these 54 patients, multiple TNAs occurred, with isolated vertigo being the most common (four of seven). In all four cases, the vertigo occurred suddenly and without provocation by movement. The temporal relation between isolated vertebrobasilar TNA and vertebrobasilar stroke was very similar to that of defi nite carotid TIA and carotid stroke (fi gure), both being concentrated during the 2 days preceding the stroke. The rate of defi nite carotid TIA in the 2 days preceding carotid stroke was about 60 times greater than that during the preceding 3-90 days (9·5 events per day, 19 TIAs in 0-2 days, vs 0·15 events per day, 13 TIAs in 3-90 days; relative rate 64·3, 95% CI 31·76-130·21). This relative rate was similar for isolated brainstem TNAs preceding vertebrobasilar stroke (11·0 events per day, 22 TNAs in 0-2 days vs 0·26 events per day, 23 TNAs in 3-90 days).
TNAs thought to be of carotid territory origin, such as atypical amaurosis fugax and limb shaking episodes, occurred only before defi nite carotid stroke (six of 759 patients, three with atypical amaurosis fugax and three with limb-shaking). Of the 759 patients with defi nite carotid territory stroke, 65 (9%) had at least one TNA or TIA within the preceding 90 days (median 7·0 days, IQR 1-19·5) of which 32 (49%) fulfi lled the NINDS criteria for TIA (fi gure, table 2).
Only ten (22%) of the 45 patients with isolated brainstem TNAs sought medical attention before the stroke and a vascular cause was suspected by their physician in one of these cases. In the remainder of cases, the initial diagnosis was documented as postural hypotension (three cases), peripheral vestibular disturbance (two cases), sepsis (two cases), migraine (one case), and cranial nerve palsy (one case). Vascular risk factors were more common in those with isolated vertigo, compared with other vertebrobasilar TNAs (table 3).
In patients with stroke in whom 50% or higher symptomatic stenosis was documented, TNAs were also more frequent preceding vertebrobasilar stroke than preceding carotid stroke (     (table 4). Among the 16 patients with vertebrobasilar TIA and 50% or higher symptomatic vertebrobasilar stenosis, preceding TNAs were more commonly binocular visual disturbance (n=7) than isolated vertigo (n=2).

Discussion
Our study shows that there is a temporal relation between transient isolated brainstem symptoms and vertebrobasilar stroke (panel). Such events occurred in the days and weeks immediately preceding 16% of vertebrobasilar strokes. Of all events occurring during the 90 days preceding a vertebrobasilar stroke, fewer than 10% satisfi ed the NINDS defi nition of TIA. We also noted that most patients with such symptoms failed to present to medical attention at the time but presented only after a subsequent stroke, potentially leading to major bias in determining risk prospectively even in a population-based study of TNAs.
Results from large cohort studies have shown that 12-18% of people without a previous diagnosis of stroke reported TIA-like symptoms. [24][25][26][27] However, none of these studies determined the risk of stroke during the subsequent days and weeks and there are very few prospectively collected data of the prognosis after TNA. Prospective studies can underestimate stroke risk by missing those who do not seek medical attention at the time of the TNA. Questionnaire-based surveys of patients looking at the frequency of previous TNAs have various recall biases and exclude those who had a major or fatal stroke after the TNA and are therefore unable to respond. [24][25][26][27] Given these diffi culties, large prospective population-based studies with complete ascertainment are probably impractical. However, it is possible to identify and study the most important subset of TNAsie, those that are associated with a subsequent stroke and which should therefore not be missed or misdiagnosed. Since half of all recurrent strokes that occur in the 5 years after a defi nite TIA occur in the fi rst 90 days, 28 with a similar front-loaded risk after vertebrobasilar TIA, 29,30 if we assume a similar time course for atypical TIAs (or TNAs) then useful information on the nature of highrisk TNAs can be gleaned from studying those events that occur in the 90 days preceding a defi nite stroke.
Isolated vertigo was the most common transient symptom preceding vertebrobasilar stroke in our study. Results from the Atherosclerosis Risk in Communities Study (ARIC) study 24 showed that dizziness was the most frequently reported atypical symptom (36%) in the underlying population, although only 1·2% were classifi ed as having a TIA or stroke. 24 In elderly people, the prevalence of vertigo in the year before stroke has been reported to be 15%. 31 Results from a prospective study of patients admitted to hospital with isolated vertigo showed a three-fold increase in stroke risk compared with controls up to 4 years after hospital discharge. 32 Results from a retrospective study of patients with vertigo identifi ed from MRI requests made after emergency department attendance showed 9·2% had acute stroke. Gait instability and subtle neurological examination fi ndings were associated with increased stroke risk. 33 Studies using Doppler sonography, magnetic resonance angiography, or catheter angiography have shown that isolated vertigo can be the only manifestation of vertebrobasilar ischaemia. [33][34][35][36] In a study of 29 patients admitted to hospital with possible vertebrobasilar TIA, isolated vertigo was present in 21%. 34 In our study, the close temporal relation between vertigo and vertebrobasilar stroke suggests that these TNAs are often secondary to vertebrobasilar ischaemia.
Our fi ndings have implications for public education to improve recognition and awareness of the signifi cance of symptoms of TIA. In our study, of those patients who did present to medical attention after the initial event, most were not managed as having TIAs, suggesting that broader diagnostic criteria are needed. Although we do not know the rate of TNAs in the population, in patients with isolated vertigo plus vascular risk factors, referral for specialist opinion would appear to be reasonable.
The main strengths of our study were the populationbased design, the rigorous case-ascertainment, and the prospective defi nition and classifi cation of the various types of TNA. However, our study has limitations. First, the prevalence of TNAs before stroke events might be inaccurate if patients failed to report them during direct questioning or if recall bias increased recollection and reporting of TNAs. However, the very rapid fall in rate of preceding TNAs over the few days before the stroke is

Interpretation
We studied all potential ischaemic events during the 90 days preceding an ischaemic stroke in patients ascertained within a prospective, population-based incidence study in Oxfordshire, UK (the Oxford Vascular Study). Isolated brainstem TNAs were more frequent before a vertebrobasilar stroke than before a carotid stroke, particularly during the preceding 2 days. Our study shows a close temporal relation between vertebrobasilar stroke and preceding isolated brainstem transient neurological attacks that do not satisfy the NINDS defi nition of TIA. More studies of the prognosis of the diff erent transient isolated brainstem symptoms are needed.
not what would be expected on the basis of recall biaspeople would recall events for the preceding few weeks and not just the preceding few hours. Most importantly, in previous studies of defi nite TIA before stroke, this same time course of preceding TIA was noted 20 and was subsequently confi rmed in prospective studies. 21 We might also have missed TNAs in patients who were unable to give a clear history after stroke because of aphasia or reduced consciousness. Second, although our attribution of the probable vascular territory of TNAs was predefi ned, it might not be accurate in all cases. Finally, we do not have data of the total number of patients with TNAs in our population who sought medical attention but were not considered to have TNAs and were not referred to secondary care during the period of the study. Without these data on the denominator we cannot identify the absolute risk of vertebrobasilar stroke after an isolated brainstem TNA reported to medical attention. Moreover, we do not, of course, know how many patients had such symptoms and did not seek medical attention. Furthermore, a previous study of patients with TNAs who were referred to hospital suggested that family doctors tend to refer those with vascular risk factors, potentially leading to a referral bias. 37 Nevertheless, in those studies that have attempted to determine the incidence of TNAs that are reported to medical attention, the numbers of events have been similar to the number of defi nite TIAs ascertained over the same period, 38 suggesting that the number of denominator events (ie, the total number of TNAs in the population) might not be so large as to substantially dilute the risk of stroke.
In conclusion, our study shows a close temporal relation between vertebrobasilar stroke and preceding isolated brainstem TNAs which do not satisfy the NINDS defi nition of TIA. A high index of suspicion is particularly important in view of the short-time window available for prevention of stroke. However, because of the adverse eff ects of a false-positive diagnosis of TIA, more studies of the prognosis of the diff erent transient isolated brainstem syndromes are needed.

Contributors
NLMP contributed to acquisition of data, draft and revision of the report, statistical analysis, and interpretation of data. MS contributed to the revision of the report content and acquisition of data. PMR contributed to study concept and design, draft and revision of the report, analysis and interpretation of data, and study supervision.

Confl icts of interest
We declare that we have no confl icts of interest.