The Role of the Clinical Examination in the Subclassification of Stroke

It has been recognised for many years that there is a need to subclassify patients with stroke, for the following reasons: to improve prognostication of survival, recurrence and functional outcome, to permit the cost effective and timely search for aetiological factors, to provide a measure of case mix for audit of management, to assist with the planning of services and to allow stratification within clinical trials. Although studies reporting the accuracy (or rather lack of accuracy) of clinical scoring systems to distinguish infarction and haemorrhage still appear in the literature, for practical purposes there is general agreement that CT or MR scanning is required to make this distinction. Therefore, this discussion will concentrate on the subclassification of ischaemic stroke.It can certainly be argued that a subclassification of ischaemic stroke according to aetiological mechanism is the most logical since it is only with a clear understanding of the underlying mechanisms of stroke that rational acute and secondary preventative therapies can be deployed. Not surprisingly, therefore, classifications such as those used in the Stroke Data Bank [1]and the TOAST study [2]have been widely adopted. One argument against using this type of division as a primary method of subclassification is the difficulty of applying the necessary technological investigations to all patients (e.g. because of age, pre-morbid disability or availability of state-of-the-art equipment). However, it is worth remembering that there are aspects of the clinical examination which contribute to the final allocation of aetiological mechanism, e.g. the presence of atrial fibrillation. When these are combined with the information from an acute CT scan and baseline investigations (emergency blood examination and ECG), in the setting of an academic specialist stroke service, an accuracy of 70% when compared with the final aetiological diagnosis has been reported, with a further 24% finally being classified as ‘Infarct of Unknown Origin’, the original diagnosis being that of large artery disease in 16% and cardioembolism in 8%. However, in the TOAST study the accuracy was only 62% [3]. Perhaps one of the most telling things about the drawbacks of such a system of classification is that there are very few good natural history studies of ischaemic stroke subdivided in this way.Another method of subclassification that is used is that of infarct topography. This can be approached in two ways. In an era of increasingly sophisticated imaging many people hoped that this would become a relatively straightforward matter of matching the imaged infarct onto standard maps of arterial territories. Following on from this, one would be able to gain an idea of the relative likelihood of different aetiological mechanisms causing infarcts in any particular arterial territory. Unfortunately, even though worldwide access to neuroimaging has improved dramatically, several factors have conspired to frustrate this method of localisation. Perhaps most importantly, the work of van der Zwan et al. [4, 5]highlighted the great inter- and intra-individual variation of the arterial distribution within the human brain. Additionally, the emergence of hyper-acute stroke treatments has meant that, as far as CT is concerned, initial scans are often negative – indeed, for patients being contemplated for thrombolysis, lack of change on the initial CT scan is the desired scenario. However, this does not help the problem of subclassification of infarct type!Another approach to topography is the more traditional one of clinical localisation, a method that has its origins in classical neurological teaching. This was the basis for the development of the Oxfordshire Community Stroke Project classification [6]. Although the fact that there are only four categories, total anterior circulation infarcts (TACI), partial anterior circulation infarcts (PACI), lacunar infarcts (LACI) and posterior circulation infarcts (POCI), is sometimes viewed as being too simplistic, the scheme implicitly acknowledges the relatively poor inter- and intra-rater reliability of many clinical symptoms and signs [7, 8]and therefore reflects the inevitable trade off between sensitivity and specificity. It is worth remembering, however, that the accuracy of syndromic diagnosis has been shown to be better if clinicians have a specific interest in cerebrovascular disease [9]. The classification itself has been shown to have very reasonable reliability [8, 10, 11]. As a primary method of subclassification it has obvious attractions in that the information is quick, easy and inexpensive to collect on virtually all patients and the syndromes occur frequently enough to promote pattern recognition particularly amongst junior staff.However, what has been shown is that the clinical subtypes also predict the underlying vascular pathology (tables 1, 2) i.e. that TACI/PACI patients have much higher rates of large artery disease and cardioembolism than LACI patients, a fact that may be of value when trying to assess the likely value of further investigations and the speed at which they should be obtained.One point that needs to be remembered is that the OCSP classification was derived from a dataset where patients had been assessed at a mean of 4 days after stroke. This is important because the syndromic classification is based on the pattern at the time of maximum deficit. Clearly, there must be some concern that assessment in the hyper-acute phase may not be as reliable. Certainly Toni et al. [14]have suggested that particular care needs to be taken with the clinical diagnosis of LACI in patients seen at a mean of 6 h after onset. However, a large study reported by Tei et al. [15], where all patients were seen within 24 h after onset, suggests that the OCSP classification remains reasonably robust (table 3) and provides further support for the correlation of clinical subtype with final diagnosis of mechanism of infarction (table 4).In conclusion, the clinical history and examination continues to have a very important part to play in the initial subclassification of patients with cerebral infarction. Topographic and aetiological classifications are not mutually exclusive but rather complimentary and provide the clinician with a degree of flexibility that allows a useful degree of subclassification in virtually all patients, whatever their circumstances.