Myasthenia Gravis: Subgroup Classification and Therapeutic Strategies


Myasthenia gravis (MG) is a neurological condition characterised by a fatiguing weakness of certain muscle groups, particularly those that control eye opening, eye movements, speech and swallowing When severe the proximal muscles of the limbs and respiratory muscles may be involved. Acquired myasthenia is an autoimmune disease, where antibodies are directed against the post synaptic nicotinic acetyl choline receptors (AChR) of neuromuscular junctions or against other proteins that affect AChR function.

Diagnosis and appropriate management of MG is particularly important because at any time it can transform suddenly from a relatively benign condition of ptosis and diplopia to a crisis with potentially fatal bulbar dysfunction and respiratory failure. These latter symptoms may in turn reverse with prompt emergency supportive care and immunomodulatory treatment.

The discussed review in Lancet Neurology by Gilhus & Vershuuren seeks to provide an insight into the usefulness of the latest antibody assays in predicting in individual patients the clinical course of MG and response to therapy. Evidence was gathered from the literature on the basis of appropriate searches on Medline and the Cochrane library for English language publications from 1995 to 2015.


The review first describes the pathophysiology of the different associated antibodies. AChR antibodies cross link receptors, accelerating their breakdown. Muscle specific kinase (MUSK) and lipoprotein related protein 4 (LRP4) exist as a complex on the post-synaptic membrane. When activated by agrin protein, this complex affects the aggregation of AchR and the morphology of the terminal. Antibodies to MUSK, LRP4 and agrin influence this process and are therefore are likely to be directly pathogenic. Titin and ryanodine receptor antibodies occur in some patients with thymoma related MG, but may be markers of severe disease rather than directly pathogenic.

Comorbidities may be present in MG, and awareness of these is important. Younger onset patients may have other organ specific autoimmune disease , including polymyositis. Thymoma associated MG is associated with increased risk of haematological malignancies and with a severe autoimmune cardiomyopathy.

Classical subtypes include:

  • Early onset MG with ACh antibodies. This often has ocular involvement and has a female preponderance. Thymic hyperplasia may be present and in these cases the condition responds to thymectomy.
  • Late onset MG with AChR antibodies.  This is also often ocular, but there is only rarely thymic hyperplasia.
  • Thymoma-associated MG. These patients usually have generalised disease and AChR antibodies. There are also other paraneoplastic associations, such as pure red cell aplasia and neuromyotonia.
  • MUSK associated MG. These antibodies are present in 1-4% of MG cases. The condition is usually bulbar or generalised rather than ocular and there is no thymic involvement.
  • LRP4 associated MG. This can be ocular or generalised in presentation.
  • Antibody negative MG occurs in 5% and is heterogenous, probably reflecting different undiscovered causative factors.
  • Ocular MG is defined as being restricted to the ocular muscles; if this remains the case for 2 years, 90% of the time it will remain so. Half of such cases have AChR antibodies, but only very rarely do they have MUSK antibodies.

When symptoms are typical, the review considers neurophysiological testing unnecessary in all cases bar those that are seronegative.

Finally the review discusses treatment options. Immunosuppressive treatment is recommended when symptomatic treatments (anticholinesterases such as pyridostigmine) fail alone to control symptoms. (MUSK antibody associated disease often has a poor response to such treatment.) An extensive review of clinical trials reveals disappointing results in many cases when compared with placebo. Nevertheless a clear treatment plan of steroids combined with immunosuppressive drugs is recommended. Other treatment plans may vary from this. The only information regarding treatment in relation to antibody serology is that rituximab in uncontrolled studies may be particularly effective in MUSK associated MG.

The review concludes with a discussion of new treatments, such as other monoclonal antibody therapies targeting autoantibodies, or antigen specific treatments that encourage the development of immune tolerance.



The review provides a welcome revision of management in an important therapeutic area. However it was felt that there was little specific information on serological-clinical correlations that practically affect management. This was the presumed main hypothesis of the review. The lack of ocular and thymic involvement in MUSK associated disease, and its poor symptomatic response to anticholinesterases, were interesting points.

Other points that arose out of the discussion were:

  • The lack of evidence base for treatment compared to the clear benefits observed in practice does point to the limitation of relying solely on evidence based medicine. It was conjectured that in some cases this may reflect patient selection. If for example, all ocular myasthenic patients are started on immunosuppression, in many cases it may be unnecessary and so demonstrating an improved response compared with placebo may prove difficult. Perhaps clinical focus is understandably upon patients with myasthenic crises or who have recently had myasthenic crises, where the response to treatment is more dramatic and clearly in some cases life-saving.
  • The indication in the review that neurophysiology was only necessary in seronegative patients was surprising. In our practice, we often have neurophysiology results before serology becomes available. In patients with ocular symptoms only, the differential includes cranial nerve palsy, sympathetic lesions, myopathic processes and even muscle tension related symptoms. Identification by neurophysiology alerts clinicians to the fact that the patient is at risk of life-threatening myasthenic crisis. Patients with bulbar involvement may have motor neurone disease or myopathy. Finally, there is a significant false positive AChR occurrence; in patients with low positive AChR  titre in whom we feel that myasthenia is actually unlikely, normal neurophysiology on single fibre EMG jitter study helps to confirm this. While not 100% sensitive and specific, neurophysiology does lend valuable diagnostic support.

This paper was presented to our Journal Club by Dr Salman Haider, Specialist Registrar in Neurology, Queens Hospital, Romford, UK.

Posted in Myasthenia | Tagged , , , | Leave a comment

Clinical Features and Pathology of “Parkinson’s Plus” Syndromes

msa jounral clubBackground

It has long been known that there exist variants of Parkinson’s disease (PD), loosely and perhaps inaccurately described as PD plus syndromes, that may carry features of Parkinsonism but which also have other clinical features. Such conditions have distinct pathology at autopsy.

However, it has also long been known that clinicopathological correlations of these conditions are not perfect; in other words, a patient in life may have clinical features indicating one PD plus syndrome but may be found subsequently to bear the pathology of another.

The subject of this journal club, When Dementia with Lewy Bodies, Parkinson’s Disease and Progressive Supranuclear Palsy Masquerade as Multiple Systems Atrophy by Koga et al. (2015) in Neurology, is a retrospective review of Mayo Clinic brain bank cases labelled as having Multiple Systems Atrophy (MSA) in life.

MSA is a neurodegenerative condition that may have one or both of parkinsonism and ataxia features, and may also have autonomic features, pyramidal features and even features of anterior horn cell disease. According to the Second Consensus Statement (2008), the criteria for probable MSA are:

A sporadic neurodegenerative condition of onset >30 with:

  • Urinary incontinence (plus erectile dysfunction in males) or measuredorthostatic hypotension within 3 min of 30mmHg systolic or 15mmHg diastolic and at least one of:
    • Poorly levodopa responsive Parkinsonism (bradykinesia with rigidity, tremor or postural instability)
    • Cerebellar syndrome

However some patients will have pathologically proven MSA without satisfying these criteria, while in others the clinical picture will be confused by coexisting conditions in this age group, such as Alzheimer’s disease (AD) or cerebrovascular disease.


Study Design

The study reviewed the autopsy results of 134 cases that had consecutively been submitted to the brain bank with a clinical label of MSA. Patients came from 37 US states. The pathological assessments were done using a standard protocol. In 125 patients there were useful clinical records, and in some cases further information was gained by questionnaires sent to living relatives.


Study Findings

A pathological diagnosis of MSA was confirmed in 62% of cases. Of the remaining 38% of cases, 37% had Dementia with Lewy body (DLB) pathology, 29% had PSP, and 15% had PD. Two of the 134 total had Corticobasa Degeneration (CBD), two had cerebrovascular disease and five were “miscellaneous”.

On retrospective assessment of clinical features according to the above criteria, only 49 patients had probable MSA and 35 possible. (But incomplete records do not mean that patients did not have particular clinical features). Once this had been done, 71% of probable MSA patients had pathological MSA, and 60% possible MSA patients had MSA pathology.

The paper describes pathological changes in some detail. In the same way that there are, according to Braak, “stages” or at least grades of neurofibrillary tangle involvement in Lewy body disease, there have been described five phases of A beta amyloid deposition in Alzheimer-type disease. These range from phase 1 where deposition is exclusively in neocortex, to phase 5 where there is widespread involvement even in the cerebellum.

In pathological MSA, 8% also had Lewy body pathology. Overall the median Braak stage was I (not 0). A quarter of the MSA brains had phase 1 or worse A beta phase of Alzheimer’s.

With pathological diagnosis as the reference point, the features that were more common in MSA than in DLB were urinary continence, ataxia, nystagmus and pyramidal signs. Cognitive impairment and visual hallucinations were more common in the latter.

Comparing MSA vs PD, incontinence was more frequent and visual hallucinations less frequent.

Comparing MSA vs PSP, urinary incontinence, constipation, orthostatic hypotension and REM sleep behaviour disorder were more frequent, and vertical gaze palsy less frequent.

Levodopa responsiveness and mini-mental state score actually did not distinguish these diagnoses.

The main errors related to assuming that orthostatic hypotension automatically resulted in an MSA diagnosis instead of DLB or PD, and assuming that ataxia resulted in an MSA diagnosis instead of PSP. Severe dysautonomia early in the course of PD should not be considered an exclusion criterion for that diagnosis.

Imaging has poor sensitivity. Only 38% of pathological MSA had imaging changes. The hot-cross bun sign was rare. There were similar rates of abnormality in PSP.



As suggested by the authors, a limitation of the study is that retrospective post mortem analysis suffers from clinical signs being recorded at different stages of disease advancement and there is a selection bias in those that come to autopsy (such as atypical cases).

Our feelings were that for the above reasons the study cannot be used to determine real diagnostic accuracy. The “improvement” in diagnosis from 62% to 71% when a movement disorders specialist applies probable diagnostic criteria carries little meaning, given the limited data available from those who examined the patient in life. A “brain bank” is only as good as the accuracy and detail of clinical label attached to the specimens.

It was pointed out, though, that the very wide geographical distribution of specimens, which included those not from academic centres, does reperesent a cross-section of patients in the US labelled as having MSA.

We wondered if the difference between PD and DLB is essentially quantitative. DLB is rather arbitrarily defined according to dementia changes manifesting before extrapyramidal changes, otherwise it is considered PD dementia. Perhaps “diffuse” Lewy body disease is a better clinical label. Pathologically there is likely to be a borderline state between the localised involvement of PD and the diffuse involvement of DLB, and indeed if the Braak hypothesis is correct, this overlap may apply to all patients at certain stages of disease progression.

Our final point was a philosophical one about what is the gold standard of diagnosis. Is it necessarily always pathology, which presumably accurately reflects the pathophysiological process that led to the observed pathology? What if there is dual pathology, as reflected in a number of specimens in this study? Which supercedes the other? Is it simply relative severity? If one set of clinical features can reflect either one or both of two different pathological appearances, what is actually more important for the patient and clinician? Would we deny a patient a trial of cholinesterase inhibitor for their dementia and hallucinations if we somehow knew that their pathology was MSA or if their Lewy bodies were localised to the brainstem? Would we not treat their autonomic symptoms if their pathology was PD? Would we fail to check a clinical MSA patient for sleep apnoea if their pathology instead revealed Lewy bodies?

While pathology might be the gold standard when conducting clinical trials, in normal clinical practice it is the clinical features guiding practical management and prognostication that are of primary importance. The broad clinical labels of system involvement still help to classify patients according to their present and future clinical needs.

This paper was presented to our Journal Club by Dr Gemma Cummins, Specialist Registrar in Neurology, Queens Hospital, Romford, UK.

Posted in Parkinson's Disease | Tagged , , , , | Leave a comment

Acute Flaccid Myelitis and Enterovirus D68

enterovirus 1Background

Enteroviruses, which may cause gastroenteritis or upper respiratory tract infections, are well-known to have neurotropism – a predilection in a proportion of individuals to spread to certain types of neurones, thereby resulting in characteristic neurological syndromes that occur after initial infection. The most well-known example is of course poliomyelitis which results in acute flaccid paralysis from anterior horn cell involvement. More recently, enterovirus A71, causing hand foot and mouth disease, has resulted in epidemiological clusters of brainstem encephalitis and acute flaccid paralysis.

Enteroviruses may also be responsible for acute viral meningoencephalitis, presenting classically with fever, meningism and obtundation, and perhaps with focal cerebral neurological signs or seizures. In fact, in our experience,  enteroviruses, along with herpes simplex virus, are the most commonly identified organisms responsible for sporadic viral meningoencephalitis.

This paper, by Messacar et al. in the Lancet (2015), reports cases of the latest enterovirus identified to have possible neurotropism, namely enterovirus D68. Localised outbreaks occurred between 2008 and 2010, and a large outbreak occurred in the USA in the Autumn of 2014. A cluster of cases of acute flaccid paralysis in children were identified at the same time (Autumn 2014); they presented with weakness in proximal limb, facial and bulbar muscles and radiological changes of either longitudinal grey matter spinal cord lesions and/or brainstem hyperintensities were identified.

Study Design and Findings

Cases were retrospectively identified from records of children admitted to a hospital in Colorado in the Autumn of 2014. Inclusion criteria were acute flaccid paralysis with mainly grey matter spinal cord involvement on imaging and/or acute cranial nerve dysfunction with brainstem lesions on imaging.

In the acquisition period twelve children satisfied the neurological inclusion criteria. All but one had fever and preceding upper respiratory tract symptoms. Ten had meningism. The limb weakness, present in ten cases, was proximal with hyporeflexia and preserved sensation. A similar proportion of children had symmetrical and asymmetrical weakness. Ten had cranial nerve dysfunction, bulbar weakness, diplopia or facial weakness. None had encephalopathy or seizures.

The spinal cord lesions, present in eleven children, always affected the central grey matter, especially the anterior horn cells. The longitudinal spread was from 4 to 20 vertebral levels (we commented that there were only 19 levels from C1 to T12, but younger children’s spinal cords extend a little lower). Brainstem lesions were present in nine children, mainly in the dorsal pontine tegmentum.

Some children had EMG, which showed variable motor denervation, presumably indicating either anterior horn cell or ventral nerve root involvement.

Spinal fluid analysis typically showed a pleocytosis (unlike typical cases of acute motor axonal neuropathy subtype of Guillain Barre syndrome) and normal or mildly elevated protein in all but one case.

In five of eleven cases, nasopharynx specimens were positive for enterovirus D68. Blood and spinal fluid PCR was negative, as is commonly the case in polio and other enterovirus neurotropic infections. Polio also has a delay between initial and neurological symptoms and yet is due to direct viral spread to the anterior horn cells, so the delay after respiratory symptoms and the lack of spinal fluid virus observed in these cases does not necessarily indicate that the condition is post-infectious rather than infectious.

Many of the children were treated with intravenous immunoglobulins or steroids. Some had plasmapheresis. There was no clear benefit from these therapies. Many children had lengthy admissions and significant residual neurological deficits despite treatment.

The paper compared the number of these neurological cases in 2014 with average numbers of similar presentations (from ICD 9 code discharge database) between 2010 and 2014. In any previous 3 month period the maximum number of cases was 4 (significantly lower). One was positive for D68.


The background epidemiological context in the paper describes a 77% increase in children with acute respiratory admissions in 2014 compared with equivalent months in previous years.  It is not clear if all such cases were tested for viruses but a number were tested with a screening nasopharyngeal swab PCR array that does not distinguish enteroviruses from rhinoviruses. There was a “substantial increase” in positive results on this test during the acquisition period. Only 25 cases were actually tested for D68, and these were cases admitted to ITU with severe respiratory disease. These were positive for D68 in 76%.


The paper presents a persuasive argument for a defined neurological syndrome, with the flaccid paralysis, exclusive motor involvement, pleocytosis and rather characteristic neuroimaging features.

Less persuasive is the causative role of enterovirus D68. It was identified in less than half of the neurological cohort during a presumptive D68 outbreak. The background level of enterovirus D68 positive asymptomatic children during the acquisition period is unknown. The increase in neurological cases during that Autumn epidemic compared to baseline levels could be skewed by ascertainment bias and, given the sample size, a coincidental increase is always possible.

Neurological D68 infection is not new. In fact a previous paper (Ayscue et al., 2014) (which this paper mentions at the end of the discussion) reported 23 cases of acute flaccid myelitis in California from 2012 to 2014, two of which were positive for D68. The lower rate of D68 identification could have reflected testing late during the illness, whereas earlier testing as in this paper is more likely to yield positive results.

Nevertheless the paper does contribute to a body of evidence suggesting that enterovirus D68 is one of a group of viruses that have accounted for recent outbreaks of upper respiratory tract infections and may have particular patterns of neurotropism resulting in acute infectious or post-infectious complications. Unfortunately, as the paper identifies, there are as yet no specific and effective preventative or treatment strategies for these neurotropisms.

This paper was presented to our Journal Club by Dr Sian Alexander, Specialist Registrar in Neurology, Queens Hospital, Romford, UK.

Posted in Disease Categories, Infectious Diseases | Tagged , , , , | Leave a comment

Journal Club Review: “Cerebral Amyloid Angiopathy with and without Haemorrhage

BackgroundCAA 1

Sporadic cerebral amyloid angiopathy (CAA) is the most common cause of lobar intracranial haemorrhage, which in itself accounts for about 5-10% of all strokes. Amyloid deposition in small arteries of the cerebrum leads to friability and haemorrhage. There are also rare familial forms of amyloidosis affecting the nervous system that more typically result in early onset dementia or peripheral neuropathy, and amyloid deposition in general constitutes part of the pathological process of other neurological conditions such as Alzheimer’s disease.

There is no specific treatment for CAA other than blood pressure control, as hypertension may be an association. It might nevertheless be useful to identify markers of the condition before major symptomatic haemorrhage occurs, so that one can avoid anticoagulants and perhaps antiplatelets in such susceptible individuals.

Potential markers include the apolipoprotein (APO) E ε2 genotype and imaging markers such as superficial siderosis and centrum semiovale white matter perivascular spaces. Superficial siderosis is not specific to CAA, being the result of any cause of chronic cerebral or spinal subarachnoid leakage of blood. Enlarged perivascular spaces could relate to small haemorrhages in the more distant past.

This paper, by Chridimou et al. (2015)  in Neurology, looks at these associations retrospectively by a database review of brain biopsies or of evacuated haemorrhage material. These are correlated with MRI findings and APO E ε genotype in patients with positive pathological findings.

Study Design and Findings

The brain biopsy and haemorrhage specimen database review found around 100 cases of pathological cerebral amyloid angiopathy (CAA). Roughly half had had a symptomatic lobar intracranial haemorrhage (ICH). The others were presumably identified coincidentally.

There was no difference in pathological CAA severity between those with or without ICH, but neuritic plaques without neurofibrillary tangles were more likely to be found in ICH cases (53% vs 13%); this is because, as is already known, neurofibrillary tangles are associated with a more Alzheimer type amyloid deposition process. Again as expected, the ε2 allele was more associated with ICH cases and the ε4 allele (a risk factor for Alzheimer’s disease) more likely in the others.

Imaging features of white matter changes, enlarged perivascular spaces and microbleeds were the same in both groups but superficial siderosis was more common in ICH cases (52% vs 20 %).

caa2Follow up of the non ICH patients 3 years after brain biopsy revealed that 2 of 51 subsequently had lobar ICH; one of these had superficial siderosis changes.

The study’s main conclusion is that the ICH subgroup of CAA is more likely to have superficial siderosis on imaging.

Strengths and Weaknesses of Study

It is already known that different APO genotypes are associated with Alzheimer’s disease or with ICH, and despite the fact that pathological changes were not worse in the ICH patients it is not surprising that ICH patients are statistically more likely to have the “haemorrhage” genotype than the “dementia” genotype. The presence pathologically of CAA appears to be less specific for ICH as opposed to Alzheimer’s disease than the presence of an APOE ε2 allele or superficial siderosis on imaging.

However, the degree of specificity and sensitivity of these markers is not diagnostically helpful.

The association with superficial siderosis is not surprising as it is a direct marker of haemorrhage more specifically than general markers of small vessel disease that could also reflect atherosclerosis. At least the paper confirms this and raises awareness of superficial siderosis, but anticoagulation would not be given to such patients anyway, even without awareness of the strength of association with ICH. A patient with superficial siderosis would typically be investigated on imaging for a cause of chronic subarachnoid haemorrhage. If no such cause was found, and the patient was relatively elderly, possibly it could be concluded that they were susceptible to CAA and in particular to ICH from such pathology.

Our Journal Club wondered in passing if homozygotes for either ε2 or ε4 had extra susceptibility to haemorrhage or dementia respectively. Certainly, homozygous ε4 carries increased risk of dementia; compared to being homozygous for the “neutral” ε3 allelle, ε4ε4 carries a x15 risk of developing AD, while ε3ε4 carries a x3.2 risk. APOE transports cholesterol and lipoproteins to the neurones by binding to neuronal APOE receptors. Mutations may lead to atherosclerosis because of hyperlipoproteinaemia. The ε2 allele is less efficient at binding so homozygotes may be more susceptible to atherosclerosis. On the other hand one ε2 allele is protective versus the e4 allele in relation to various neurodegenerative conditions including ischaemic stroke!

It is not immediately clear why these polymorphisms of a lipoprotein transporter would influence amyloid deposition into vessels and how friable this would make the vessels become.

A limitation of the study is the selection of patients; it would be heavily skewed to ICH patients because there would be pathology available in many of these. Patients would rarely have a biopsy if superficial siderosis was found on MRI, given its many other causes, and there would presumably have been some other major brain pathology, such as early onset dementia, in the remainder of non ICH cases that would have prompted such an invasive investigation. Hence the increased occurrence of ε4 allele is not surprising.

The true specificity and sensitivity of these markers remains unknown because the biopsy reflects a single snap-shot in time; there may be two subgroups of ICH negative patients, one who have amyloid deposition in blood vessels as part of some other amyloid process, possibly a by-product of dementia-related pathology, and the other who are susceptible to ICH but who simply have not had their haemorrhage yet. Unfortunately 3 years follow up, to see if it is specifically those who also have superficial siderosis who are in the susceptible group, is probably too short to answer this question.

This paper was presented to our Journal Club by Dr Sam Nightingale, Specialist Registrar in Neurology, Queens Hospital, Romford, UK.

Posted in Disease Categories, Stroke | Tagged , , | Leave a comment

Journal Club Review “Certainty of Stroke Diagnosis: Incremental Benefit with CT Perfusion over Non-Contrast CT and CT Angiography”

BackgroundCT Perfusion Journal Review

The accompanying primer,  Thrombolysis for Stroke and role of CT perfusion Imaging, describes the difficulties and potential shortcomings of thrombolysis for acute stroke and the way that CT perfusion may improve patient selection for thrombolysis. This paper, by Hopyat et al. (Radiology 2010) describes a related problem: the risks of thrombolysis, mainly constituting secondary haemorrhage, are greater when reperfusing a large area of infarcted brain. In the second European Cooperative Acute Stroke Study (ECASS II), failure to recognise involvement of more than one-third of the middle cerebral artery territory resulted in a high risk of haemorrhage when such patients received thrombolysis. CT perfusion may allow better identification of this situation and avoidance of thrombolysis. In addition, CT perfusion may aid in identifying the baseline stroke size for prognostication and research purposes, in positive confirmation of ischaemia during a TIA and, as discussed in the primer, in identification of stroke mimics. The study uses an incremental protocol with up to date CT perfusion technology to assess its use in positive identification of stroke.

Study Design

The study took 191 consecutive patients with presumed stroke/ unresolved TIA who were admitted within 3 hours of symptom onset. Unenhanced CT, CT angiogram and CT perfusion were assessed in that order by non-expert reviewers. A final diagnosis of stroke was established about a month later by an experienced clinician with the aid of a subsequently-performed MRI with diffusion weighted imaging (DWI).


According to the final diagnosis made retrospectively, 64% of the patients had stroke, 18% had TIA and 17% were stroke mimics.

The sensitivity, averaged over all patients and within and across image reviewers, of correct identification of stroke by unenhanced CT was 52.5%, by unenhanced CT and CT angiography was 58.3% and by unenhanced CT, CT angiography and CT perfusion all together was 70.7%; using all three was significantly better than using one or two modalities (p=0.0003 and p=0.013 respectively).

This was not at the cost of reduced specificity (i.e. false positive errors), which was around 85% for all three conditions. Rather than give an all or none answer, the reviewers scored their confidence levels for diagnosis of stroke, and this allowed calculation of receiver operating characteristic (ROC) curves for unenhanced CT alone, unenhanced CT and CT angiography and all three together.

Receiver operating characteristics plotting sensitivity against false positive rate (i.e. 100-specificity) determined by reviewers scoring their confidence level in diagnosing stroke in various different patients. Unenhanced CT alone is blue, unenhanced CT plus CT angiography is red, and unenhanced CT plus CT angiography plus CT perfusion is orange.

Receiver operating characteristics plotting sensitivity against false positive rate (i.e. 100-specificity) determined by reviewers scoring their confidence level in diagnosing stroke in various different patients. Unenhanced CT alone is the blue trace, unenhanced CT plus CT angiography is in brown, and unenhanced CT plus CT angiography plus CT perfusion is in orange.

For example a reviewer at a very conservative level, requiring a high confidence level for positive diagnosis, may rarely identify stroke; he will have high specificity when he does label a case as having a stroke, but very low sensitivity. A good diagnostic tool will be one where there is a larger area under the curve of sensitivity versus 100% minus specificity. In other words, accepting just slightly less than 100% specificity makes the sensitivity rise very dramatically. It is seen that using all three modalities together improves sensitivity over unenhanced CT alone at all levels of specificity, but at very high levels of specificity, CT perfusion does not improve performance over CT angiography.

Inter-observer agreement (Cohen kappa) was only between 0.28 and 0.44 for unenhanced CT alone, and between 0.68 and 0.78 for all three modalities together. Intra-observer agreement was similarly better using all three modalities together.

Strengths and Weaknesses of Study

The authors attempted a “real-life” situation analysis using an incremental protocol, a realistically early time of imaging, inexperienced reviewers and a range of stroke severities that included mild stroke and TIAs. They demonstrate clear superiority in these circumstances. The circumstances also explain why absolute performance may have been lower than in other studies.

The authors cite the advantages of CT perfusion, namely being done just after standard CT and taking just 1-2 minutes extra to perform and about 5 minutes extra to process. In their hands the radiation dose was only that of another unenhanced CT head. The disadvantages they cited are the confounding effects of chronic internal carotid artery occlusion and chronic ischaemic changes making it hard to determine what is new and what is old.

They consider the addition of CT perfusion well adapted to triage of stroke patients but are cautious about the benefits of identifying penumbra because of the absence of actual evidence that reperfusing penumbra improves outcome.

However, the “real-life” analysis situation might not be without shortcomings in interpretation because real life may be different in different units. Certainly in many stroke units there will be individuals on hand to assess imaging in real-time who may have several years’ experience rather than the one year’s experience of the study’s reviewers. Their lack of skill may have overestimated the extra sensitivity of CT perfusion.

While it is also laudable that they have not selected for their study only patients whose stroke was clinically obvious on admission, it does seem strange that there were so many TIAs when most TIAs do not usually last more than an hour. The mean delay was 117 minutes +/- 59 minutes, so generally there was a 1 to 3 hour window. Some clinicians might delay imaging a little if the patient attended within an hour with as yet not improving symptoms. It would not be a fault of imaging, as such, if a TIA was identified as stroke simply because the scan was performed early enough to detect the ischaemia. A more experienced radiologist might better distinguish ischaemia from established infarction on CT perfusion by the lack of reduction of cerebral blood volume, but this was not specifically examined in this study.

As a tool for ruling out stroke mimics, CT perfusion is clearly and unsurprisingly better than unenhanced CT (which was never intended for this purpose). But with sensitivities around 75% and specificities around 85%, it can hardly be considered a gold standard. Should we thrombolyse on that basis, given the 6% risk of causing harm from intracranial haemorrhage (though the harm engendered on thrombolysing the normal brain of some stroke mimics is likely to be low compared to thrombolysing an extensive established infarct)?

The performance of CT perfusion in positive diagnosis might in fact come up short compared with that of an experienced clinician examining the patient shortly after initial triage, and then one wonders whether that clinician ought not then to rely on his clinical judgement alone or upon an early MRI with DWI, accepting that this might not be universally feasible.

The introduction in the paper started by describing early major middle cerebral artery infarction as a relative contraindication for thrombolysis and how CT perfusion may help identify this. I cannot help but wish this is what the study actually investigated rather than detection of stroke mimics, but it does at least provide a good guide as to how to go about conducting such an investigation rigorously.

Posted in Stroke | Tagged , , , , , , , , | 1 Comment

Primer on Thrombolysis for Stroke and Role of CT Perfusion Imaging

Stroke-Header_FASTStroke, defined as a sudden vascular event resulting in localised brain damage (World Health Organisation, 1978), is without doubt a major challenge in health care, being the third most common cause of mortality in developed countries and the single greatest cause of lasting disability (Mant et al., 2004). In the UK, stroke patients occupy 2.6 million days in hospital beds a year, equivalent to one in five total acute hospital beds and one in four long-term beds (National Audit Office, 2005). Over the last decade, there have been increasing efforts to organise acute stroke care into dedicated stroke units and to raise public awareness that stroke is a medical emergency to be managed in a timely fashion (e.g. the FAST campaign).

The development of thrombolysis has been one of the drivers for management of stroke as an emergency. This “clot buster” treatment may be given intravenously to dissolve the thrombus or embolus in a cerebral artery and allow reperfusion of the territory supplied by the artery before those areas of the brain become irreversibly infarcted. Timing is critical for such treatment to be effective; if given too soon after symptom onset, the thrombolysing agent (tissue plasminogen activator (TPA)) may be unnecessary as the patient may in fact be suffering a transient event that would reverse spontaneously, and if given too late the brain tissue will already be dead.

alteplase-actilyse-for-ischaemic-stroke-500x500The standard European criteria for thrombolysis, developed from the major multicentre study (Safe Implementation of Thrombolysis in Stroke Monitoring Study, abbreviated to SITS-MOST (Lancet. 2007)) that validated its use, originally stipulated a time window of 3 hours after symptom onset and excluded patients whose symptoms were rapidly resolving. In practice, giving the treatment too early is not a major concern as most self-resolving events, called transient ischaemic attacks (TIAs) last less than an hour and it is very rare to be ready logistically to thrombolyse within an hour of symptom onset.

Outcome following Thrombolysis

Unfortunately, thrombolysis is not a panacea even within this narrow time window. A fair comparison is achieved with a randomised double-blind study against placebo, but because of widespread use such studies have not recently been performed. The original positive trial (National Institute of Neurological Disorders (NINDS) Stroke Study Group, 1995) showed no clear clinical differences after 24 hours but what was described as “at least 30% better outcome” at 3 months (global odds ratio 1.7 (95% confidence interval 1.2 to 2.6). By way of example, the percentage of patients achieving 0-1 on the modified Rankin score (meaning no or minimal disability) was 39% vs 26%, in other words 13% more patients had  excellent outcome after thrombolysis than after placebo. There was no improvement in mortality.

However, there have been concerns over the fact that in this study the placebo patients had a worse severity stroke at onset, that other studies have shown unclear benefit, and that some studies have relied upon open label self-reporting by patients to measure outcome.

Underpinning these concerns is the risk of haemorrhage associated with intravenous thrombolysis. Thrombolysis was originally developed for coronary thrombosis in myocardial infarction; the brain is an organ far more sensitive to insult and reperfusing infarcted brain may make it particularly susceptible to haemorrhage, with far worse consequences than a haemorrhage into myocardium. In the NINDS study, 7% of patients had a symptomatic intracerebral haemorrhage (meaning neurological deterioration or other clinical suspicion in presence of haemorrhage on CT not seen pre-treatment) within 36 hours after thrombolysis, versus 1% of patients given placebo. In 3% of thrombolysed patients, the haemorrhage was fatal.

The SITS-MOST study was designed to look at safety of thrombolysis given according to the same protocol and collected data on 6483 patients and found a similar figure of 7.3% patients significantly worsened (<=4 points higher on NIHSS score) within the first 7 days by intracranial haemorrhage.

So when counselling patients on giving thrombolysis, we should say that within the 3 hour window, out of 100 treated patients, around 12 will have a better outcome (more likely to be disability free or minimal disability), 4 will be made worse because of brain haemorrhage, 2 will die from brain haemorrhage and 82 will be unchanged. It does not sound as good as quoting 30% better outcome (taken as the increased proportional percentage gain rather than absolute percentage gain over placebo).

Recent Changes to Prescribing Guidelines

More recent studies have explored widening the window of thrombolysis to 4.5 hours, or even longer in certain circumstances. The third European Cooperative Acute Stroke Study (ECASS III, 2008) randomised patients at 3 to 4.5 hours after stroke onset to thrombolysis or placebo, and found a 52.4% versus 45.2% good outcome; the significance level for this 7% improvement was only 0.04 – the lower 95% CI for the odds ratio of better outcome (according to their chosen criterion on Rankin) was 1.02! If other Rankin criteria were chosen, e.g. 0-2 versus 3-6 instead of 0-1 versus 2-6), no significant improvement would be demonstrated. In fact the chance of being dead or severely disabled at 3 months (modified Rankin scores 5-6) was non-significantly higher if thrombolysed (14.8% versus 13.4%). Concern has also been voiced that, despite randomisation, the placebo group had on average a more severe stroke before thrombolysis (one point worse on NIHSS), and were more likely to have had a previous stroke. The risks of intracerebral haemorrhage were comparable to data from patients thrombolysed within 3 hours.

In counselling a patient within this time window, we would therefore have to add that because the time is more than 3 hours after thrombolysis, the chance of improvement to the state of no or minimal disability increases by around 7% instead of 12%.

Cost effectiveness analysis of patients thrombolysed in this time window show limited favourability but are based on limited evidence; of course fatalities reduce cost compared to disability so I find such analysis morally inappropriate.

In the UK, the National Institute of Clinical Excellence (NICE) guidelines for stroke were updated in 2013 to increase the thrombolysis window from 3 hours to 4.5 hours. There is also now no exclusion of posterior territory infarction and debate over excluding patients over 80 years. I personally have reservations about this, and consider it a situation where we have permission under licence to give it if we feel in our judgement it is clinically appropriate. Despite the trumpeting of trial data, there are ethical reservations about giving a treatment that will help a modest proportion of patients but harm a significant proportion too. The haemorrhage risk is more than those surgeons and anaesthetists typically quote for surgery.

Unsurprisingly, this situation polarises medical opinion. Outcome data on thrombolysis have come under intense scrutiny and been subjected to endless meta-analysis and debate. What is really needed is less spin on statistics and more information on predicting a good outcome of thrombolysis in an individual patient who has just had a stroke.

Current guidelines for selecting patients for thrombolysis depend on a clinical diagnosis of ischaemic stroke, a clinical scale of stroke severity, various exclusion criteria and a CT scan of the head. This CT scan will not demonstrate the stroke; the changes of stroke after 3 hours are too early to be detected on CT, which is simply showing the reduced density of infarcted brain. Instead, the CT excludes a haemorrhagic stroke where thrombolysis would be pointless and dangerous, or an established large stroke where thrombolysis may be too late and also associated with increased risk.

An alternative investigation that positively diagnosed stroke within the thrombolysis time window would be very useful to exclude “stroke mimics”, such as patients with acute unilateral muscular weakness from spondylosis or patients who are imagining that they are having a stroke and reproducing its clinical features (functional stroke). What would be even more useful is an investigation that could positively identify ischaemic but potentially retrievable brain from brain that was already infarcted and might only result in haemorrhage if suddenly reperfused by thrombolysis. This retrievable brain is known as the “penumbra”, alluding to the surrounding region of partial rather than complete shadow cast by an object in front of a non-point light source .

CT Perfusion Imaging

Of such investigations, the most promising may be CT perfusion. This requires only the hardware for standard CT, with an intravenous iodinated contrast injection, and may be performed more rapidly and be more easily tolerated than MRI. The limiting factor is likely to be user dependence and the quality of the analysis software.

Technique of CT Perfusion Imaging

After a bolus injection of contrast, a sequence of images are taken that measure the rise and subsequent fall in contrast density as the bolus travels through the cerebral vasculature. Two reference time plots are normally taken; that for the input transit time is the A2 segment of the anterior cerebral artery as it passes perpendicular to the axial plane of imaging, and that for the output transit time is the superior sagittal sinus. For each region of interest voxel, four parameters are then calculated:

  • cerebral blood flow (CBF)
  • cerebral blood volume (CBV)
  • mean transit time (MTT)
  • time to peak contrast enhancement (TTP).

These are then mapped onto an axial slice of the brain to convey visually how the different parameters vary across brain regions, with high values represented as red and low values represented as blue.

Flow dynamics tells us that three of these parameters are interdependent:

  • MTT = CBV / CBF

So if there is a thrombus reducing flow to a region of brain, as the CBF is lowered, the MTT increases in parallel if the CBV stays the same.

Normal grey matter has a higher CBF and CBV, so cortical areas of gyri tend to look more red on both CBF and CBV. Because the increases are similar, the changes cancel out on MTT so the MTT tends to look more blended between white and grey matter. The venous sinuses also look very red on CBF and CBV and similar to background on MTT, presumably because these voxels are purely blood so have relatively high blood volume as well as flow. (One might expect arteries to have higher flow, and therefore blue on MTT, but the resolution of CT perfusion may not be great enough to identify arteries in cross-section.)

In the early hours after an acute stroke, it is considered that an “umbra” of infarcted brain may be surrounded by a “penumbra” of ischaemic brain that will shortly become infarcted but is potentially salvageable on reperfusion. CT perfusion may allow differentiation of the two because the CBV reduces more in infarcted brain. So:

Infarcted Brain ↓↓↓ CBF ↓↓ CBV ↑↑ MTT
Ischaemic Brain ↓↓ CBF slight ↓ CBV ↑↑ MTT
CT perfusion

In a patient 70 minutes after stroke onset (NIHSS score 10), the unenhanced CT, not shown, is normal. The cerebral blood flow (top left) and cerebral blood volume (top right) show reductions in the arrowed area. There is a corresponding increase in mean transit time (bottom left) and a DWI weighted high signal area several days later on MRI (bottom right). (Figures taken from Hopyan et al., 2010.)


Cerebral blood flow is obviously reduced if there is a proximal thrombus, and in infarction there is reduction in cerebral blood volume. This could be because of tissue swelling raising local intracranial pressure and constricting capacitance vessels, because there is a certain elasticity in vessels so that constriction will follow from reduced flow, or because of reflex vasoconstriction of capacitance vessels in damaged brain. The reduced CBV is still less than the dramatically reduced flow, so that MTT is significantly prolonged. The infarcted area appears more blue on CBF and CBV and more red on MTT.

In ischaemic brain, the CBV is relatively preserved, perhaps because the affected brain area is not as swollen or perhaps because of preserved reflex capacitance vessel dilatation in an attempt to improve perfusion of these areas. Cerebral blood flow is reduced (blue), but there is now a mismatch between CBV (relatively normal) and MTT (clearly red).

Problems in interpreting CT perfusion

  • Image processing is complex and user dependent. There may be poor selection of the anterior cerebral artery and superior sagittal sinus reference points
  • If the protocol is poorly designed, the radiation dose may be massive
  • Many protocols do not analyse the whole brain, so clinical knowledge is required to determine if the area of interest is middle cerebral artery territory. Brainstem areas cannot easily be assessed.
  • The resolution of CT perfusion is such that small strokes may not be visualised
  • If there is extracranial vessel occlusion, e.g. carotid artery, the hypoperfused area may give a false impression of acute infarction. The same applies to areas of leukoaraiosis. Thus CT perfusion must be interpreted in the context of unenhanced CT appearances and preferably with CT angiography.

Practical Uses of CT Perfusion

  • Identification of penumbra. If a patient was outside the 3-hour time window for thrombolysis, or the time of onset was unknown, but was otherwise a good candidate, a CT perfusion scan revealing a relatively large penumbra with normal CBV and prolonged MTT would indicate salvageable brain that might benefit from thrombolysis.
  • Positive identification of stroke. CT perfusion reveals changes very early after stroke onset, but there may be poor sensitivity because of lack of clarity over the territory of interest, the possibility of posterior circulation stroke and poor resolution of a small stroke, e.g. lacunar infarction.
  • Measuring cerebrovascular reserve. In the non acute setting, CT perfusion before and after administration of intravenous acetazolamide can help to identify brain areas that are chronically ischaemic. Acetazolamide is a vasodilator, but will have less effect on ischaemic areas because such areas already have ongoing maximal compensatory vasodilation. Thus, after acetazolamide, there will be less increase in CBF in ischaemic areas compared to normal neighbouring areas, less increase in CBV (though this is generally increased throughout the brain), and most clearly an extra prolongation of MTT (more red) in areas that may already have somewhat prolonged times compared to normal areas.
  • Identifying vasospasm. In the situation of subarachnoid haemorrhage, areas of brain suffering reactive vasospasm react like the penumbra of a stroke and indicate that measures taken to reduce vasospasm may reduce the risk of lasting focal neurological deficit after subarachnoid haemorrhage.

An accompanying Journal Club Review looks at a study that investigates the use of CT perfusion in acute stroke primarily in terms of stroke diagnosis.












Posted in Primer Posts for General Readers, Stroke | Tagged , , , , | 1 Comment

Journal Club Review: Cervical Vertigo

bronstein front pageBackground

Cervical pain from spondylosis or muscular problems is a very common symptom in the general adult population, estimated in a recent study to have a point prevalence of 4.9% and a global burden of 33.6 million disability-adjusted life years (Hoy et al., 2014). Symptoms are commonly recurrent within individuals, returning in from 50-85% of cases within 5 years of initial presentation (Haldeman et al., 2008).

The most common aetiologies of cervical pain are joint disease resulting in spondylosis and acute or chronic muscular injury. The muscles of the mobile cervical and lumbar spine tend to develop spasm as a consequence of joint or muscle inflammation and this further exacerbates injury, resulting in a vicious cycle of pain.

Vicious Cycle of Neck pain

Vicious cycle of neck pain

Diagnosis and management of cervical pain is often complicated by a number of associated symptoms, including headache, dizziness, tinnitus and ear discomfort (Baron et al., 2011). While headache of tension type character, often occipital and radiating anteriorly to the frontalis or temporalis areas, is well-established in relation to neck pain with around 20% of cases of chronic headache having a cervical basis, the other associated symptoms are more controversial.

This article focuses on one associated symptom in particular, namely cervical vertigo. The review article by Brandt and Bronstein, published in JNNP in 2001 presents a comprehensive account of the scientific basis underlying the condition, and in the article I will go over this review and more recent studies on the subject.

Cervical Vertigo Definition and Terminology

Cervical vertigo may be defined as:

A  perception of vertigo or imbalance resulting from cervical spondylosis and muscle spasm, normally a chronic tendency to brief attacks and especially brought on by head movement.

Other terms that describe the same thing are cervicogenic vertigo, cervical imbalance and cervical dizziness.

Vertigo arising from the neck presents a particular diagnostic challenge as other potential causes may require alternative management and in some cases will require urgent attention (table 1); the common occurrence of cervical pain means its association with imbalance may be coincidental.

Table 1. Differential diagnosis of vertigo

Table 1. Differential diagnosis of vertigo

Diagnostic Confusion

The two conditions most commonly confused with cervical vertigo are benign peripheral positional vertigo (BPPV) and vertebrobasilar insufficiency. They all typically present with vertigo whenever the head moves in a certain way suddenly, rather than a discrete episode as in Meniere’s syndrome, and there are no other cranial nerve features as may occur with a vestibular Schwannoma. Nowadays the availability of MR imaging means that the latter may have already been excluded before specialist referral.   The distribution of cases of such vertigo between these three apparently nosologically and pathophysiologically distinct entities remains controversial, and indeed some argue that cervical vertigo does not exist at all.

In truth, the conditions themselves may overlap. Patients with cervicogenic vertigo are likely to have cervical spondylosis making them susceptible also to vertebrobasiliar insufficiency, and if the latter is demonstrated clinicians may be tempted to consider it the hierarchically dominant or sole diagnosis even though many of a patient’s attacks may be cervicogenic. Conversely, patients with BPPV are likely to stiffen their neck as a protective mechanism and may consequently develop cervicogenic vertigo even after their BPPV has resolved. Finally, in post-traumatic cases vertigo may result from a combination of dislodgement of otoconia into the lumen of the semi-circular canals producing BPPV, damage to the otolith organs which are vulnerable to mechanical acceleration, and whiplash injury producing cervicogenic vertigo.

Overlapping nosological entities of vertigo

Overlapping nosological entities of vertigo

Clinical Presentation of Cervical Vertigo

When considering the complaint of “dizziness”, it is important to define more closely what the patient actually experiences. When many patients describe dizziness they are actually referring to presyncopal symptoms. True vertigo is a perception that the environment is moving in a rotatory direction or swaying to and fro. Finally, dizziness is sometimes used to describe a spatial disorientation or perception of imbalance. Sometimes it is the perception that appears in itself to be pathological – the patient feels unsteady perhaps more than actually being unsteady; this must be distinguished from patients whose perception of imbalance is a relatively more accurate and objective appraisal of their actual unsteadiness as a result of ataxia or loss of postural reflexes.

The typical patient with cervical vertigo falls into the category of those who describe a perception of spatial disorientation or imbalance. Rather than true vertigo, they report positional unsteadiness, imbalance, giddiness, or a feeling that the ground is sliding underneath them. As is typical for peripheral vertigo, head movements precipitate the symptoms, often neck extension or rising from supine. They may have an excessively cautious gait for their apparent objective level of balance impairment, grabbing hold of walls or “furniture walking”, and this may lead to a mistaken diagnosis of psychogenic vertigo. For such symptoms to be considered cervicogenic, there should obviously be a history of neck pain. However many specialists recommend caution in making a diagnosis of cervicogenic vertigo whenever there is neck pain, especially in cases where the description is of true vertigo. The musculature is obviously not the only structure in the neck that can give neck pain and vertigo and if post traumatic vertebral artery dissection must be excluded. Similarly, vertebrobasilar insufficiency (see below) may result from pinching of the arteries in their course through the cervical vertebrae and, while this would typically also result in other transient brainstem symptoms, a presentation with vertigo alone has been reported.

On examination, patients with cervical vertigo sometimes have their symptoms set off on testing eye movements, and they may have a reluctance or a restriction on testing range of head movement. Instead they may tend to turn the trunk with the neck. Vestibulo-ocular reflex testing or the Hallpike’s test may reproduce symptoms but without nystagmus or with a very slight nystagmus. Instead, to isolate the influence of cervical afferents the patient should be placed on a rotating stool and the head gently fixed by the examiner’s hands while the trunk is rotated back and forth. Cervical nystagmus of immediate onset may result, changing direction with the direction of rotation. However, this sign is unreliable, as discussed below.

Important Alternative Diagnosis: Benign Peripheral Positional Vertigo

Bingin peripheral positional vertigo (BPPV) is thought to relate to otoconia floating freely in the semicircular canals, usually the posterior semicircular canal on one side; head movement in the plane of this canal results in ongoing stimulation and generates vertigo and nystagmus.

Onset of BPPV is usually subacute or chronic, and characterised by brief episodes on making certain head movements. A more severe acute onset of continuous positional vertigo usually points instead to vestibular neuritis, also called labyrinthitis.

The Hallpike’s test is positive in BPPV, with rotatory nystagmus in an extorting direction in the lower eye (top of the eyes jerking towards the floor) and usually adapting after several seconds.

Hallpike's Test and Epley Manoeuvre (Fife et al., 2008)

Hallpike’s Test and Epley Manoeuvre (Fife et al., 2008). Otoconia are loose in the right posterior semicircular canal (arrowed fig. 1). The patient’s head is turned 45 degrees to the right so that the posterior canal is in the plane of motion (and the effect of the posterior canal on the other side is negated) when the patient is lain flat (fig. 2). The consequent nystagmus is extorting in the right eye.


Vertebrobasilar Insufficiency

This is an oft cited but rarely demonstrated syndrome thought to relate to pinching of the ipsilateral vertebral artery when a patient with cervical spondylosis turns their neck. The tortuous course of the artery through the transverse foramina of cervical vertebrae C6 to C1 and then across the posterior arch of C1 makes the artery particularly susceptible to such compression.

Course of the vertebral artery

Course of the vertebral artery

In the related Barré Liéou syndrome (1926), the artery is not directly pinched but irritation of the sympathetic plexus around vertebral arteries causes reflex vasoconstriction. However, the existence of this sympathetically mediated phenomenon remains doubtful.

On examination, it is suggested that if the head is moved to one extreme, compared to cervical vertigo, the nystagmus of vertebrobasilar insufficiency would start only after a delay of several seconds to minutes. However, given it is assumed that a perhaps already atherosclerotic artery is being badly pinched, I cannot help but think that such a test should only be performed in a catheter lab! Indeed, it would only be during formal arterial angiography demonstrating occlusion with the head turned to one side but not the other coul done really make a confident diagnosis of this condition.

It is considered by some that the phenomeon of disruption of cervical afferents mediating the cervico-ocular reflex does not exist and “cervicogenic vertigo” always results from vertrbrobasilar insufficiency. However, a review of the anatomy of the blood supply to the brain via the circle of Willis reveals many collaterals and means that ischaemia will result only when there is already occlusion of the contralateral artery, and probably additional significant atheromatous narrowing of the anterior circulation.

Circle of Willis

Circle of Willis

Such a situation must be rare and is different of course from the aetiology of a vertebrobasilar territory transient ischaemic attack, where an embolus from these arteries passes up and lodges into a smaller artery without collaterals. This is why vertebral or carotid artery occlusion in the neck carries a much lower risk of stroke than does a stenosis where emboli may still pass up through the narrowed lumen.

In addition, if there was transient ischaemia from hypoperfusion, why would it selectively result in vertigo and no other brainstem features such as ataxia, dysarthria, collapse, hemianopia or loss of consciousness? Nevertheless, some cases of likely vertebrobasilar insufficiency have been reported to present with vertigo without other brainstem features (Dvorak & Dvorak, 1990). Conceivably if the blood supply to an already stenosed anterior inferior cerebellar artery, which comes off the basilar artery not the vertebral artery, was critically dependent on one remaining patent vertebral artery, pinching of the latter could result in transient ischamia only of the inner ear structures, resulting in a peripheral nystagmus with or without deafness and tinnitus.

The main point is that while vertebrobasilar insufficiencty may indeed exist, the circumstances required for it to occur seem too rare to account for all cases of presumed cervical vertigo.

Scientific Basis of Cervicogenic Vertigo

Signals important in balance control, including vision, eye position, vestibular signals and processed postural information and perceptual information, are integrated in the vestibular nuclei located in the pons. These nuclei in turn output to postural control centres, to the eye movement apparatus to control compensatory eye movements and to perceptual processes.

Inputs to the Vestibular Nuclei

Inputs to the Vestibular Nuclei

Vertigo is a false perception of movement, and typically results not from a deficit but a mismatch of balance signals. This mismatch may be between defective and normal vestibular canal signals on either side of the head, or between vestibular and visual signals. For cervical vertigo to exist, there would therefore have to be a physiological basis not only in functionally important cervical signals inputting head position on the trunk to the vestibular nuclei but also in a process that compared these signals with vestibular or visual signals at a perceptual level so that a mismatch could lead to vertigo.

Why have cervical signalling for balance?

Visual inputs signal movement and position relative to the retina, while vestibular inputs signal movement and position relative to the head; however, the balance system needs information primarily on the centre of mass which mainly reflects the trunk. In essence, afferents from muscle spindles and joint receptors in the neck would allow determination of centre of mass by correction of the vestibular signal for head position with respect to the trunk.

Balance Responses. 1) Whole trunk movement to left may leave head behind from inertia. Stretch of left neck muscles signals head on trunk movement to right (red). Lateral semicircular canal signals partial head movement to left (blue), as does retinal slip signal (green). The cervico-ocular reflex (COR) will result in slow phase eye movement to left. This acts to compensate for relative head movement to fix gaze, or shift gaze to overall body facing. Any actual left head movement will result in vestibule-ocular reflex (VOR) slow phase to right and optokinetic reflex (OKR) slow phase to right. Overall eye movement will be integrated in the vestibular nuclei as the difference between COR and an amalgam of VOR and OKR.  The separate head on trunk, head in space and retina in space movements may reach level of perception. For an overall perception of trunk motion, leftward head perception, must be added to rightward head on neck perception (which reflects a leftward trunk under head movement). Other reflexes in action include direct cervico-collic stretch reflexes that will turn the head left in response to head on trunk movement, and from the vestibular nuclei an integrated vestibulo-collic reflex that will stabilise the head on the trunk and integrated postural reflexes that will stabilise trunk positioning. 2) Experimental blocking of afferents of right neck will lead to unopposed stretch signalling on left, simulating right head on trunk motion. This will generate an unopposed COR signal slow phase eye movement to left, so fast phase of spontaneous nystagmus is on the same side as the block. 3 Vibration applied to the neck muscle stimulates stretch reflexes without any vestibular or ocular involvement (unless the stretch actually secondarily moves the head).

Balance Responses.
1) Whole trunk movement to left may leave head behind from inertia. Stretch of left neck muscles signals head on trunk movement to right (red). Lateral semicircular canal signals partial head movement to left (blue), as does retinal slip signal (green). The cervico-ocular reflex (COR) will result in slow phase eye movement to left. This acts to compensate for relative head movement to fix gaze, or shift gaze to overall body facing. Any actual left head movement will result in vestibule-ocular reflex (VOR) slow phase to right and optokinetic reflex (OKR) slow phase to right. Overall eye movement will be integrated in the vestibular nuclei as the difference between COR and an amalgam of VOR and OKR. The separate head on trunk, head in space and retina in space movements may reach level of perception. For an overall perception of trunk motion, leftward head perception, must be added to rightward head on neck perception (which reflects a leftward trunk under head movement). Other reflexes in action include direct cervico-collic stretch reflexes that will turn the head left in response to head on trunk movement, and from the vestibular nuclei an integrated vestibulo-collic reflex that will stabilise the head on the trunk and integrated postural reflexes that will stabilise trunk positioning.
2) Experimental blocking of afferents of right neck will lead to unopposed stretch signalling on left, simulating right head on trunk motion. This will generate an unopposed COR signal slow phase eye movement to left, so fast phase of spontaneous nystagmus is on the same side as the block.
3) Vibration applied to the neck muscle stimulates stretch reflexes without any vestibular or ocular involvement (unless the stretch actually secondarily moves the head).

Cervico-Ocular Reflex Nystagmus

In the same way as vestibular signals are responsible for vestibulo-ocular reflexes and vestibular vertigo, a functionally important cervical balance pathway that could result in cervicogenic vertigo might be expected to be associated with a demonstrable cervico-ocular reflex, where stimulation of spindle afferents results in reflexive compensatory eye movements. In other words, neck proprioception, if input to vestibular nuclei, may result not only in perception of motion, but a compensatory eye movement that may be recorded by electronystagmography, infrared or video systems.

Trunk rotation, e.g. to left, under a fixed head in the dark would be interpreted by neck proprioceptors as head movement to right and would generate a compensatory slow phase to left. Fast phase of nystagmus would therefore be to the right, the opposite side to trunk rotation.

More physiologically, if the trunk turned and the head was not fixed but lagged behind by inertia, the reflex would make the direction of gaze follow the direction of trunk movement even if the head did not.

Evidence for Cervical Balance Signals: Anatomical Connections

The deep short intervertebral neck muscles are rich in muscle spindle afferents that are able to provide a signal of head on trunk position or head on trunk movement (Cooper & Daniel, 1963) and there is anatomical demonstration of connectivity to the vestibular nuclei and neighbouring brainstem reticular formation areas (Ciriani et al., 1992).

Evidence for Cervical Balance Signals: Vibration Induced Responses

Selective stimulation of cervical afferents by vibration over the neck muscles simulates a stretch reflex; unilateral stimulation is indeed found to result in postural responses, apparent movement of a visual target and a weak deviation of perception of subjective vertical to give the illusion of ipsilateral head tilt.

There is also an associated cervico-ocular reflex of low amplitude. Furthermore it is found that this response is increased after a unilateral vestibular lesion, building up over several weeks as a presumed compensatory enhancement. The automatic postural responses are greater than the perception of motion, unlike the major perception of motion that results from caloric testing of vestibular function, and thus fits with cervicogenic vertigo constituting more a sense of imbalance than actual vertigo.

Evidence for Cervical Balance Signals: Disruption of Cervico-Ocular Reflexes

Interference with cervical afferents in an attempt to mimic the situation in cervicogenic vertigo also yields unclear results. Local anaesthesia of the deep neck muscles in humans results in gait deviation, a tendency to fall with a positive Romberg test to the injected side, a perception of altered position and an unsteadiness on sudden head movement that lasts for several hours after the injection. These findings are confirmed on therapeutic C2 level anaesthetic block to treat patients with cervicogenic headache.

However there is no associated nystagmus (ie cervico-ocular reflex), nor any actual vertigo. Some of the effects could reflect an imbalance in muscle tone as a result of cervicocollic reflexes rather than the cervico-ocular reflex. Nevertheless, the pattern fits with the perception of imbalance or “quasi-vertigo” on head movement rather than the true vertigo of vestibular dysfunction.

Evidence for Cervical Balance Signals: Physiological Responses

As described above clinically, on testing trunk rotation under fixed head in the dark, there is sometimes a weak cervico-ocular reflex. However, if the head is not fixed, there may be head movement, limited by inertia but brought on by tissue elasticity and by cervico-collic reflexes.  Any actual movement will result in vestibular ocular reflexes and vestibulo-collic reflexes that will secondarily stabilise the head. If tthe head strongly fixed, perceptual processes or pressure detection on the side of the head may also suppress any illusion of head rotation. —

It is therefore not surprising that physiological stimulation of putative cervico-ocular reflexes in normal human adults using trunk movements with a stabilised head produces less clear perception of motion than does muscle vibration and unreliable cervico-ocular reflexes. Nevertheless, under carefully controlled conditions, such as sinusoidal trunk movements with the head fixed by a bite bar, a reliable cervico-ocular response can be recorded and compared with analoguos vestibular and optokinetic responses.

Infra-red recordings of cervico-ocular reflex, vestibulo-ocular reflex and optokinetic reflex resulting from sinusoidal movements (0.04 Hz, ± 5° amp). This isolates the slow phase component as there is no need for resetting saccades of nystagmus when tracking a back and forth sinusoid.

Infra-red recordings of cervico-ocular reflex, vestibulo-ocular reflex and optokinetic reflex resulting from sinusoidal movements (0.04 Hz, ± 5° amp). This isolates the slow phase component as there is no need for resetting saccades of nystagmus when tracking a back and forth sinusoid. (Kelders et al., 2003)

Mean amplitudes of reflex responses at different sinusoid stimulus frequencies. Gain of COR is lowest (VOR low at slow frequencies but increases with higher frequencies). Phase of VOR and COR are more variable and COR lags behind trunk rotation at higher frequencies.  With old age, VOR and OKN gain decrease; there is a compensatory increase in COR gain, as there is after vestibular  dysfunction.

Mean amplitudes of reflex responses at different sinusoid stimulus frequencies. Gain of COR is lowest (VOR low at slow frequencies but increases with higher frequencies).
Phase of VOR and COR are more variable and COR lags behind trunk rotation at higher frequencies. With old age, VOR and OKN gain decrease; there is a compensatory increase in COR gain, as there is after vestibular dysfunction.

Sinusoidal movements of slow frequency and small amplitude generate cervico-ocular reflex (COR) of lower gain than VOR and OKN, and a tendency to lag behind the movement at higher frequencies. With old age, VOR and OKN gain decrease but there is a compensatory increase in COR gain, as there is after vestibular dysfunction. It is tempting to speculate that the same might apply to patients with clincal cervical vertigo.

Studies on patients with cervical vertigo

Having experimentally demonstrated the functioning of cervical balance signals in normal subjects, the next step is to demonstrate disordered signalling in patients with presumed cervical vertigo.

Such patients do have myofascial trigger points for their pain that exhibit spontaneous EMG activity compatible with hyperactive muscle spindles (Hubbard & Berkoff, 1993). However, no correlation is found between the magnitude of physiological cervico-ocular reflexes and the severity of clinical cervicogenic vertigo. Perhaps, if there is an abnormality of cervico-ocular reflexes associated with cervical vertigo, it does not simply relate to the gain (amplitude) of what is after all a physiological rather than pathological phenomenon, but to a mismatching of signalling from either side or to a failure to calibrate such signals with vestibular and visual balance information.

—What has been reported in patients with cervical pain (with or without vertigo) is that they tend to have poorer postural control based on vibration or galvanically induced body sway and when such patients are treated with physiotherapy there is improvement in their dizziness and imbalance as well as in their cervical pain (Karlberg et al., 1996).

Does Cervical Vertigo Exist?

—There appears to be a scientific basis for the notion that stretch of neck muscles influences balance mechanisms, and a physiological cervico-ocular reflex, especially in controlled conditions, has been demonstrated. However, there has as yet been no demonstrated abnormality in this reflex in patients with cervical vertigo. This lack of a reliable diagnostic test, unlike the clear abnormality of vestibulo-ocular reflexes in vestibular vertigo, hampers study of the condition because of the consequent problem with defining a patient population.

Given the lack of any diagnostic abnormality in such patients other their associated neck pain, the question may then be asked why all patients with neck pain do not get vertigo. This has been felt to signify that the condition does not actually exist. However, there are innumerable examples in medicine where patients do not have to have the “full house” of clinical features to have a syndrome. There may be additional factors that trigger vertigo, such as the nature and asymmetry of muscle spasm, previous vestibular problems or a constitutional tendency to heightened vertiginous perceptions as is found in visual vertigo. Rather than proof the condition does not exist, it might be more constructive to consider patients with neck pain without vertigo a good control population. Using healthy subjects as controls runs the risk of identifying abnormalities that are more the direct result of pain than a manifestation of cervical vertigo.

—Since physiotherapy appears to help cervical vertigo as well as pain, it might be regarded that the diagnosis is somewhat pointless since management is the same as if a patient presented with pain alone. However, there is an important differential diagnosis of vertigo that could be occurring coincidental to cervical spondylosis. And management may be the same precisely because the condition is poorly defined. There could be future refinements of physiotherapy techniques if the subset of patients with neck pain who also have vertigo were better understood.

As mentioned above, the contrast with vestibular vertigo where there is an obvious abnormality of vestibule-ocular reflexes, is one factor that has thrown doubt upon the entitiy of cervical vertigo. However, this contrast should in fact be expected given that patients with cervical vertigo actually complain of imbalance and vague giddiness more than true vertigo. Part of the problem with recognising cervical vertigo as a nosological entity may be that the term itself is a misnomer: cervical imbalance may be a more accurate name, serving to remind clinicians that spasm of neck muscles may result in imbalance through cervico-collic as well as cervico-vestibular pathways and that true vertigo in the context of neck pain bears further investigation of other causes of vertigo where the spondylosis is coincidental.

Other Symptoms Associated with Cervical Spondylosis

—I have commonly found in clinical practice that certain symptoms often cluster together in the same individuals. In the presence of cervical spondylosis, there seems more than a chance occurrence not only of vertigo, but headache and tinnitus. While it could be argued that their vertigo is always benign peripheral positional vertigo, their headache is migraine and their tinnitus is from cochlear degeneration or “functional”, Occam’s razor and common sense encourages us to look for a single unifying cause.

I tend to call these associations with cervical spondylosis CHIT syndrome (Cervical Headache, Imbalance and Tinnitus).

Headache is well-described in association with cervical spondylosis, and this review has discussed the likely association with vertigo. Tinnitus seems a very unlikely association, but has in fact previously been cited as being linked with cervical vertigo (Brown, 1992) or abnormalities in the neck muscles (Reisshauer et al., 2006).

It initially seems bizarre how an auditory symptom could be associated with spondylosis. However, there is an interesting physiological phenomenon of muscle contraction called the Piper rhythm. When muscles contract steadily and strongly, their individual motor units tend to fire synchronously in a tuned rhythm at around 40 Hz so that the whole muscle vibrates at this frequency. This frequency, and most likely harmonics thereof, can actually be heard by placing a stethoscope over the belly of the contracting muscle. It can be demonstrated using frequency analysis of electromyogram, tremor and electronic stethoscope signals that what is heard does indeed relate to this motor unit activity.

Power speectral estimates and coherence analysis of 50% MVC of 1DI against an elastic resustance. Peaks at 10, 22 and 41 Hz in accelerometer tremor record and rectified surface EMG power spectra. Coherence analysis reveals strong coherence especially at these peaks. Upper horizontal line is the 95% confidence interval for significantly greater coherence compared to the whole spectrum – only lower 100 Hz of spectrum shown. Lower horiz line is 95% confidence interval for non-zero coherence. There is a constant linear phase lag of tremor behind EMG at all frequencies, indicating a value of 6.5 ms lag.

Power speectral estimates and coherence analysis of 50% maximum voluntary contraction of first dorsal interosseous muscle against an elastic resistance. There are peaks at 10, 22 and 41 Hz in accelerometer tremor record and rectified surface EMG power spectra. Coherence analysis reveals strong coherence especially at these peaks. Upper horizontal line is the 95% confidence interval for significantly greater coherence compared to the whole spectrum – only lower 100 Hz of spectrum shown. Lower horizontal line is 95% confidence interval for non-zero coherence. There is a constant linear phase lag of tremor behind EMG at all frequencies, indicating a value of 6.5 ms lag.


Same subject as in above figure under identical conditions. The microphone is not as sensitive at 10 Hz as 40 Hz, hence the larger 40 Hz peak compared with tremor and EMG.

Same subject as in above figure under identical conditions. The acoustomyogram (AMG) is recorded by an electronic heart sounds monitor placed over the belly of first dorsal interosseous during contraction against elastic resistance. The “sounds” are generated directly by muscle activity as seen on the EMG power spectrum in previous slide. The microphone is not as sensitive at 10 Hz as 40 Hz, hence the larger 40 Hz peak compared with tremor and EMG.

It is temping therefore to speculate that the tinnitus of cervical spondylosis relates to overactivity of the sternomastoid muscles that originate just behind the external ear. Rather than the tinnitus sounds being imaginary or related to cochlear damage, patients are actually hearing their own muscles contracting. Certainly this interesting notion bears further investigation.

Posted in Vertigo | Tagged , , , | Leave a comment

Journal Club Review: Driving after a Single Seizure

BMJ seizuresBackground

One of the main issues facing a patient diagnosed as having had a first epileptic seizure without any sinister underlying lesion – often a young adult and otherwise well – is the driving ban. One can only be sympathetic to the impact that it may have for some on travelling to work or actually performing their job. Some react with understanding, while others have the attitude that they will never expose themselves or others to harm even if the risk is tiny and they later become legally entitled to drive. A few react with incredulity: “I totally lost consciousness without warning, may do so again at any time, and you are ruining my career or social life by preventing me from driving for several months?!”

This can be a difficult conversation for clinicians, but at least one can remind oneself that the conversation might have been more difficult if the cause of their seizure was a brain tumour rather than cryptogenic, in which case they might only be alive for several months.

Two other points can help. First, in the European Union and in most other countries the rules are standardised and set by government authorities. The physician is only explaining the law of the land. In the US, some states have similar standard rules while others, perhaps unfortunately, do leave it to the doctor or to a medical review panel. Second, these rules were developed and modified after extensive review and consultation. Briefly communicating this process may help the patient to appreciate that they are designed to protect, not to punish. The paper reviewed here describes statistical data on risk of seizure recurrence that were used to help develop a consistent European Union Guideline, which informs the UK’s Driving and Vehicle Licencing Agency (DVLA) guideline (2013) and could be used to help doctors who must form their own guidelines.

The paper was published in the good old British Medical Journal (2010) and reanalyses data from the MESS (Multicentre Early Epilepsy and Single Seizure) study (2005), specifically on patients over 16 years of age who had a single unprovoked seizure and looks at the 12-month risk of recurrence at certain time points after the index seizure. In other words, if a patient has already gone some months following an initial seizure without a subsequent seizure, how likely are they to remain seizure-free for another 12 months?

This website had an accompanying commentary that discusses the original MESS study in more detail and the wider issues around prognosis and management after a single seizure. Clearly, the data in this paper are helpful for prognosis, but only in patients who have already gone a certain period seizure-free after their initial event.

Study design

The original MESS study’s inclusion criterion was that both patient and physician were uncertain about whether or not to start antiepileptic medication. Exclusion criteria included previous treatment with antiepileptic drugs or the presence of a progressive neurological disease. Out of around 1800 patients meeting the criteria, 1400 were enrolled; the others refused on the basis that they did not want to be randomised. Demographics showed no particular bias in these patients.

Patients were randomised to immediate treatment – the drug of the physician’s choice as early as possible after seizure (usually carbamazepine or sodium valproate) – or to deferred treatment, generally if the patient had a second seizure.

Where there were around 720 with single seizures in each arm in MESS, in the BMJ reanalysis there were around 320 in each arm who were 16+ and who had had only one seizure at the time of randomisation and whose date of seizure, as opposed to date of randomisation in the MESS study, was known.


The main finding of the BMJ reanalysis was that in the immediate treatment group the risk of recurrence in the next 12 months, having already gone 6 months without a seizure after the first seizure, was 14% (95% Confidence Interval (CI) 10-18%). In the deferred treatment group the risk rose to 18% (95% CI 13-23%). In the deferred treatment group, if the patient had already gone 12 months without a second seizure, their chance of recurrence dropped to 10% (95% CI  6% to 15%).

The overall general principle regarding driving has been arbitrarily set that if the risk of a seizure is less than 20% over the next year, then it is permissible for the patient to drive a private vehicle and if the risk is less than 2% they may drive a public or heavy goods vehicle. This is not a medical but a policy decision, presumably taking into account the proportion of time that the average person spends driving and the likelihood of risk to self and others should an accident occur as a result. The role of clinicians is simply to provide guidance on which patients have a 20% or greater risk.

It can be clearly seen from the data in this review that if a patient starts treatment, their 12 month risk 6 months after a seizure is lower than 20%. Therefore they may be allowed to drive at 6 months. The same applies to patients not on treatment – if one takes the mean estimate of risk of 18%. However, if a clinician were to be asked, “At what time would you be confident that the risk of recurrence in the next 12 months would be less than 20%”, he or she should use the upper confidence limit for the risk and so the 23% figure for patients not on treatment is too high. Only if patients not on medication have already gone a year without seizures is the upper confidence interval of 15% acceptable.

Strengths and weaknesses of the study

As mentioned in the paper, a potential weakness is that the data were taken from a randomised controlled trial (MESS) of patients having immediate vs. delayed treatment. From looking at the inclusion and exclusion criteria, one might suspect a selection bias that clustered patients of intermediate severity – those who definitely wanted medication or definitely didn’t want medication were excluded. So the risks in the low-risk subcategories might be overestimated and those in the high risk subcategories underestimated.

It could have been a problem that there was an inconvenient delay between seizure and randomisation in MESS of around 3 months. This would rule out patients who had a second seizure in that time. But 3 months is half of the six month seizure free period in which we are interested! Fortunately, in this paper the investigators back-tracked to get the actual seizure time rather than randomisation time; this means that the six month free period is an accurate reflection.

But if one wanted to generalise the findings to prognostication of seizure risk, surely something that the patients will want to know about, if one is making this prognostication on a patient just after their seizure (which should typically be the case as all patients having a seizure should be promptly reviewed by a specialist), then we cannot use the figures from MESS (which included children) or those reviewed here. All we can do is wait three months, say in a subsequent clinic, and if they have not had a seizure in that time, the figures reviewed here can be used. A more full discussion on prediction of risk and decisions on treatment is in the accompanying commentary on management after a first seizure.

Finally, there is the issue of validating seizures in the outpatient department, as was done in the study. Clinicians more inexperienced than those used in MESS might make more mistakes in correctly identifying seizures, or patients might deny or forget seizure occurrences. This is likely to be more of a problem in real life than in a trial. So we cannot say that MESS is overestimating risk, but we can say that MESS does not simulate the real life underestimation of risk that may occur in daily practice.

Different risks in different patients

If the policymakers wanted to finesse the guidelines to take into account other factors, there are adjustments that could be made. In a univariate factor model, it was found that remote symptomatic seizures (seizures occurring as a result of a brain insult e.g. head injury, encephalitis, neurosurgery, that occurred some time before the seizure) were associated with significantly higher risk, as were presence of neurological deficit, seizures while asleep, abnormal electroencephalogram (EEG), and lack of brain imaging information.

Calculating the risks for these subcategories reveals that, if taking the upper confidence intervals, remote symptomatic seizures, neurological deficit, sleeping seizures and abnormal EEG all shift the risk above the 20% threshold after 6 months seizure freedom, and the first two are above the threshold even after 12 months seizure freedom. However, the data numbers are getting small and estimations more inaccurate.

A multivariate analysis of various combinations of factors, much in the same way as risk of osteoporosis can be calculated, is a better way of addressing this issue.  This is shown in table 5 of the paper (below), noting that they have excluded patients with a first degree relative with epilepsy and sleep seizures. The latter are a special case; while recurrent seizures are more likely (because they may reflect particular epileptic syndromes) they are also more likely to recur in sleep and so be less relevant for driving risk. The UK DVLA rules now in fact permit driving with continuing sleep seizures provided a pattern of seizures only while asleep has been established for at least 1 year.

multivariate seizure risk factors

One can see, for example, that a non-remote symptomatic seizure with an abnormal EEG has an upper confidence interval of risk of 23% at 6 months even if imaging is normal. One might argue that the current blanket rule of 6 months is rather lenient for patients with an abnormal EEG or with a remote symptomatic seizure, especially if the patient is not on antiepileptic medication.

A careful view of the wording of the current UK Driving and Vehicle Licensing Agency guidelines in fact includes a clause “provided no risk factors indicate a more than 20% risk of a recurrence over the next 12 months”. If this is interpreted as being confident that the risk not more than 20%, then all the above-mentioned categories would entail a 12 month not 6 month ban, and we would be needing EEGs on everyone to inform this decision. If it is interpreted as being most likely risk level, then abnormal EEG still entails too high a risk if not on medication (23%), as does abnormal imaging if remote symptomatic and not on medication (22%). Only if it is interpreted as being possible that the risk is as low as 20% and the patient was started on medication and the seizure type was non-remote symptomatic is an EEG not necessary because it is only in this circumstance where the lowest confidence interval of risk is not above 20% whether or not the EEG is abnormal.

Data from other studies

A population rather than outpatient based study on 252 patients who had a single seizure as their index seizure (National General Practice Study of Epilepsy (1990)) found a 37% risk of a second seizure within 12 months, and an 18% risk if the patient had already been seizure free for 6 months. This shows just how much the risk level reduces if the patient has already undergone a modest seizure-free period. Factors increasing the risk  of recurrence were symptomatic seizures, neurological deficit, and no antiepileptic drug treatment. The findings are therefore comparable to the reviewed data.


This paper clearly does what it intends; to ascertain whether, after 6 or 12 months seizure free following a first seizure, the level of risk of a seizure over the ensuing 12 months is greater or lower than the policy threshold of safe private vehicle driving of 20%.

The analysis provides a rationale for the duration of the driving ban that might help some patients better come to terms with what may seem a punitive measure.

Partly as a result of this study, a number of changes have been made to the UK’s DVLA regulations (2013) regarding epilepsy:

  • The ban following a single seizure is reduced to 6 months from 12 months.
  • If a pattern of sleeping-only seizures is established for 1 year (formerly 3 years) the individual is allowed to drive.
  • If a patient was seizure free on medication, and then a seizure occurred as a result of a medication change, the patient can return to driving after only 6 months if they go back on the original medication.
  • If a patient has only ever had seizures that do not affect conscious level or ability to drive, they can drive a year after this diagnosis is established even if they continue to have these seizures.

However, the multivariate analysis of risk factors does raise some issues about higher risk categories, and draws attention to the clause in the DVLA guidelines “provided no risk factors indicate a more than 20% risk of a recurrence over the next 12 months”. I am not sure how many clinicians actually apply this rule.

Could they be sued if a patient had a single remote symptomatic seizure, was started on medication, and had a second seizure 11 months later resulting in a fatal road accident if the clinician had not performed an EEG, or if the EEG was performed and found to be abnormal?

Could they be similarly sued if they had any kind of seizure, but had not started medication and the EEG was not performed or was abnormal, or if both EEG and MRI were abnormal?

Or is there a “get-out argument” that one would have to use the lower confidence estimate of risk to prove that the risk was greater than 20%? In some categories even the lower confidence estimates are above 20%. Happy days for lawyers, if not for everyone else…

Posted in Epilepsy | Tagged , , | 1 Comment

Journal Club Commentary: Management of Single Seizures

MESS studyIntroduction

For this edition of the Neurology Online Journal Club I wanted to review not one but a series of papers to address a specific issue, namely predicting the risk of seizure recurrence after a single seizure and predicting how much this is reduced by starting anti-epileptic medication. I started with the Multi-centre study of early Epilepsy and Single Seizures (MESS) study, but there is more than one report on the same data set, and its main points prompted a more detailed look at other literature on the subject and my personal views. Hence I have described this as a commentary.

There is an accompanying Journal Club review that deals specifically with risk of seizure recurrence in relation to driving.


Epilepsy is certainly one of the more common conditions managed in neurology and indeed in general medical practice. The lifetime prevalence of seizures (% of people who will have a non-febrile seizure at some point in their lives) is 2-5%, and the prevalence of active epilepsy is around 0.5%. A first seizure often presents as a sudden, shocking event in a previously well person, and often leaves the patient in a similarly well state with the expectation of returning to a reasonably normal life – and yet bewildered and worried. As a result, it is a condition where in my view counselling of the patient regarding management options and involving the patient in decision-making is particularly important.

A specific issue with epilepsy management is that typically there are no ongoing symptoms or abnormal clinical signs. The patient may be starting treatment, exposing them to potential side effects, without making them feel better at all. We may have no idea whether or not the drug is working until it manifestly fails much later in the form of a recurrent seizure, and even then we are not sure what would have happened if we had not started treatment, or had started a different treatment. In this respect epilepsy management is more akin to management of episodic headache or TIA than of Parkinson’s disease or chronic pain.

When management revolves around predicting and minimising risk, statistics inevitably play a part. Clinicians need to have the communication skills to explain clearly to patients in broad terms the likely risks of seizure recurrence in different circumstances, and of course that means knowing those risks and understanding basic statistics themselves. Knowledge of risks is covered in this review, but communicating them remains a challenge. (For example, in the UK a survey revealed that the majority of adults did not appreciate that it was equally likely for one to roll a 6 on a die as any other number, or that a previous coin toss does not affect the result of a subsequent one.)

The key questions to which patients and clinicians need answers are:

  1. What is a specific patient’s risk of a further seizure over a certain time period? This estimate should factor in whether or not this was their first seizure, the seizure type and aetiology, the time they have already gone without a seizure and other factors that determine risk such as EEG, imaging abnormalities and family history of epilepsy.
  2. How much is this risk reduced if the patient goes on antiepileptic medication?
  3. If starting medication, and there are no further seizures, when should this medication be stopped again?

Risk after a first seizure

The FIRST study (First Seizure Trial Group Study) in 1993 reported recurrence risks of 18%, 28%, 41%, and 51% at 3, 6, 12, and 24 months if not given medication, and 7%, 8%, 17%, and 25% if given medication.  Randomisation on or off medication was done within 7 days of their seizure, so this is nicely applicable to an “early clinic” or inpatient decision. The odds ratio of reduction of risk by medication was 0.4 (i.e. seizures were only 40% as likely on medication as off medication).

The largest single study on risk of seizure recurrence with randomisation for initial treatment was that conducted by the Multi-centre study of early epilepsy and Single Seizures (MESS) study group; here the risk of recurrence in the 404 randomized to immediate treatment was somewhat lower at 18%, 32%, 42%, and 46% at 6 months, 2, 5, and 8 years after randomization versus 26%, 39%, 51%, and 52% in the deferred treatment group.

Cumulative risk of recurrence years after a seizure. Note that it is the top figure that specifically refers to a first seizure.

Cumulative risk of recurrence years after a seizure. Note that it is the top figure that specifically refers to a first seizure.

A key difference between the studies is that in the MESS study patients were randomised generally 3 months after their initial seizure. The six month figure is therefore the risk from 3 to 9 months after a seizure, having already gone about 3 months without a seizure.

Further analysis published in the BMJ (2010) of a subgroup of MESS study patients  looked specifically at implications for driving and is the subject of a complementary journal club review. This subgroup naturally consisted of those over 16 years of age and those who could have their seizure-free period dated back to their first seizure rather than to time of randomisation; it was found that the 12-month risk of a seizure, having already gone 6 months without a second seizure, was 18% off medication and 14% on medication and this difference did not reach statistical significance.

The lower risk found in the MESS study than in the FIRST study is supported by a prospective study without treatment randomisation (Hauser et al., 1998) and largely on adults; the risk of a first recurrence was 21%, 27%, and 33% at 1, 2, and 5 years after the initial seizure. In those who recurred, the risk of a second recurrence was 57%, 61%, and 73% at 1, 2, and 5 years after the first recurrence. The risk of a second recurrence approached 90% after remote symptomatic seizures (those that are secondary to a brain insult at a previous time and therefore indicating an ongoing risk) and was 60% following cryptogenic/idiopathic seizures.

A problem with comparison and interpretation of study data is in patient selection. While there were 1443 patients randomised in the MESS study, another 404 did not consent to randomisation. Those where the risk might be considered lowest might not want to consider taking medication, while those at high risk might not want to take the chance not to have medication. Furthermore, an actual selection criterion was that for ethical reasons both patient and clinician had to be unsure about whether or not to start medication to be invited to participate.

It is likely that low risk groups in such a study will have overestimated risk, while high risk groups might have underestimated risk and underestimated treatment effect. This possible shortcoming is important in guiding actual practice. If there is a policy from opinion leaders that treatment is not warranted for first seizures, this might get interpreted rigidly by others as a blanket rule and those patients at high risk after a first seizure – the very patients who might not have enrolled on the study – might not even get counselling about the possibility of taking medication.

Finally, different studies may have differing proportions of different seizure type. The MESS study took anyone over the age of 1 year, and there may have been a relatively high proportion presenting with a single minor complex partial seizure.

Decision to treat

Most epileptologists do not treat a single seizure. In fact they define epilepsy as two or more seizures, to try to exclude the significant proportion of individuals who have a single seizure and no further attacks.

Perhaps this conservative strategy is because of the side effects of antiepileptic drugs. These include potential teratogenicity if falling pregnant while on the drug, long-term effects contributing to osteoporosis, possible long-term effects on fertility and possible long-term effects on cognition (mainly mooted in children).

However, there are now many antiepileptic drugs from which to choose, increasing the chance of finding one to suit, and modern drugs may minimise many of these risks. If one looks at the side effects of most drugs taken for any length of time, the list looks at least as scary as that for modern antiepileptics. For example, most anti-migraine drugs also have potential teratogenicity.

If a cardiologist said to a patient who had just had a heart attack, “Well you could have secondary prevention to reduce your risk of a subsequent myocardial infarction (MI) over the next year from 41% to 17% (using the FIRST trial data), but we won’t bother because we don’t really say you have heart disease until you get your second MI”, they would be dialling up for a second opinion before he or she had finished the sentence! And secondary preventatives such as beta-blockers, antiplatelets and statins, and certainly coronary stenting procedures and coronary artery bypass grafts, are not without their risks either.

While the mortality associated with a generalised tonic clonic seizure is lower than that for an MI, it is not insignificant. Quite apart from the circumstances of the attack potentially posing a risk, there is a small but well-documented risk of sudden unexpected death in epilepsy, thought to relate to a number of factors including the extreme autonomic disturbance that occurs during the attack. The event may occur in a young completely healthy person out of the blue, reflects a total loss of self-control, may be potentially embarrassing and stigmatising, and may leave the patient exhausted or potentially even in a psychotic state for days afterwards. I think any trivialisation of a seizure in comparison with an MI can only reflect an age-old prejudice against neurological disease that it is “difficult”, “untreatable” and not suffered by “normal” people.

But other data presented here show that if for some reason an adult patient only saw someone in a position to advise on antiepiletic treatment about six months after their first seizure (the BMJ trial went back to the seizure date not the recruitment date), and they had not had a second seizure in that time, the 12-month seizure risk figures are only 18% vs 14%. This presents a completely different picture of risk of treatment side effects versus reduction of risk of seizures.

Stratification of Risk

Another follow-up to the MESS study (2006) stratified risk of seizure recurrence according to a scoring system (below).

Scoring system for sratification of risk of recurrence after a single seizure according to the MESS study data.

Scoring system for stratification of risk of recurrence after a single seizure according to the MESS study data.

Half of the patients in the MESS study were used to investigate these risk factors to develop the scoring system, and the other half were used to see if subgroups divided post-hoc according to this risk stratification would have differing benefits from medication. It was found that all but the lowest risk subgroup would benefit from medication (see below); in fact it bizarrely seems in the lowest risk category that avoiding treatment is non-significantly protective (p=0.2).

Kaplan-Meier derived estimates of probabilities of seizure recurrence divided according to different risk groups

Kaplan-Meier derived estimates of probabilities of seizure recurrence divided according to different risk groups. Start and delayed treatment refers to treatment started at randomisation or delayed until subsequent seizures.

This information could therefore provide a basis for individualising risk assessment and individualising decisions to treat on that basis, or at least providing a default strategy. However, it would be applicable only to patients seen in a clinic fully three months after their seizure who had not already started medication or had another seizure in the meantime.

When to stop treatment

If one is to embark on treatment, perhaps controversially so after a first seizure, when does one stop?

Antiepileptic drugs are probably only protective while being taken. This is indirectly illustrated by long-term remission figures in the MESS trial. Initial treatment decisions did not affect the overall figure of 92% of patients being at least 2 years seizure free 5 years after enrolment. In other words if treatment was deferred until a second seizure, they were as likely eventually to go into remission, but had obviously had more than one seizure while getting to that point and might still be on medication at that point.

One rationale would be to treat for as long as the drug appears from population studies to be significantly reducing the risk of a subsequent seizure.

The longer the patient is seizure-free, the closer data taken from patients with single seizures recruited 3 months late will correspond to those taken immediately, so the more accurate the original MESS data become. We see that from this study’s long term follow-up, the 2 year risks were 32% vs 39%, 5 year risks were 42% vs 51%, and the 8 year risks 46% vs 52%. There is probably a diminishing return over time, but it is difficult to draw a firm conclusion as to significance of this reduced risk at different times.

Most studies specifically looking at timing of antiepileptic withdrawal are on patients who had had more than one seizure, precisely because most clinicians do not start treatment for a single seizure in the first place! Obviously the findings cannot be applied to those who had a single unprovoked seizure, because the overall risk is lower in this group.

One study (JNNP 2002) on patients who mainly had multiple seizures but which at least selected patients on monotherapy, and so tended to reflect patients more easily controlled, found that after 2 years the 12 month recurrence risk was 9% continuing on medication vs 26% stopping medication; on a multivariate analysis, the hazard ratio was 2.6 (CI 1.5 -4.8), and the hazard ratio dropped to 1.6 (1.0 – 2.6) if 3 to five years seizure free and to 1.0 if >5 years seizure free. So after multiple seizures there is clearly no excess risk from stopping medication only if seizure free for >5 years.



We have conflicting risks, conflicting risk reductions from medication and data that apply only in specific circumstances.

What we need is a large multi-centre study that:

  • Randomises patients immediately, so we can make an informed treatment decision at an appropriate time when the recurrence risk is highest
  • Subdivides into age groups, as the paediatric population and geriatric population may have different seizure aetiologies from young adults, and even different clinicians.
  • Subdivides according to generalised tonic clonic versus complex partial seizures. The latter are by no means as severe and dangerous, and one might imagine that if the first seizure is complex partial, there may be a higher chance of a subsequent one being of the same type.
  • Stratifies risk as in the MESS study, taking account of EEG, MRI, neurological deficit and cognitive impairment.
  • Uses more modern drugs – nowadays lamotrigine and levetiracetam are common first-line agents, as opposed to carbamazepine and valproate which were the drugs that mainly featured in the MESS study. While these are admittedly not clearly more efficacious, they are better tolerated.
  • Includes an analysis of the side effects of drugs in those randomised to treatment, and the quality of life impact of these side effects and of the “inconvenience” factor of taking regular medication.

Given the current lack of clear data, we are left with clinical judgement and patient preference.

My practice with regard to a patient who has just had a generalised tonic clonic seizure is largely to ignore the data from MESS indicating that treating a first seizure non-significantly increases risk when the EEG and neurological examination are normal. How much were the data distorted by being randomised 3 months after the seizure? How many in this category had a complex partial seizure? A particular problem is that often I am not going to get an EEG within a week of the seizure, so a major risk stratification factor is unknown at the most important time to start treatment. I quote the FIRST trial as a “worst case scenario”, something like:

The risks of recurrence could be as high as 41% over the next year and medication could reduce this to 17%. However, given your neurological examination and imaging (and possibly EEG) are normal, and there is no particular evidence of a recurrent epileptic syndrome (e.g. clear family history, developmental delay, juvenile myoclonic epilepsy), the risk may be appreciably lower and the benefit of medication therefore appreciably less. The risk, which includes a slight risk of sudden death as a result of a second seizure, must be balanced against the risk of side effects of taking medication.

Particular factors relevant for you might be the further 12-month driving ban after a subsequent seizure, and teratogenic risk of drugs if you fall pregnant while taking them. (Though lamotrigine and levetiracetam have rather favourable teratogenic risk profiles.)

Then, when it comes to stopping medication, as this should really be addressed before starting:

Since you have only had one seizure, we would empirically consider you in the generally accepted “best category in whom one would initially treat” and advise at least 2 years treatment assuming no further seizures. This 2 year figure is somewhat arbitrary, reflecting that FIRST demonstrated continued risk reduction two years after starting medication but did not investigate a longer period.

If the patient has had a single complex partial seizure and no risk factors, I would explain:

For this relatively minor seizure type there is a lack of evidence for treatment and most patients are not treated. Only if you are very keen on treatment, e.g. regarding driving, would I offer it to you after counselling on potential drug side effects.

If the patient is in the medium or high risk category according to the stratification of the MESS data, in other words neurological deficit, developmental delay, cognitive impairment, features of an epileptic syndrome, or if I have an EEG already and it is abnormal, or perhaps an epileptogenic lesion on an MRI scan to boot, I will tend to use the MESS data:

A potentially risky time for seizure recurrence is in the next 3 months. Even if you do get to three months without a seizure a major study has shown that the risk of a second seizure by one year is 35% and medication may reduce this to 24% (or for the high risk category 59% to 36%). Given these risks, and the slight possibility of death from a seizure, I would advise treatment despite the potential risks of drug side effects unless you had any particular issues.

And for stopping medication again:

The long-term 5+ year follow-up in the MESS study indicated that many patients go into seizure remission at this time after their first seizure, whether or not they started on medication initially or had seizures during this time, but those who were initially treated were less likely to have seizures in getting to that 5-year milestone. Furthermore, another study (though on patients who had had more than one seizure) showed that antiepileptic drugs may still reduce the risk of a recurrence over the subsequent 12 months if you have gone up to, but not beyond, 5 years without a seizure. Even if you remain seizure-free, I therefore generally recommend 5 years of treatment before slow medication withdrawal.

If the first presentation was with status, the risk of recurrence is not much greater but the risk of recurrent status is greater and so I would advise at least a 5 year seizure-free period before withdrawal even if no risk factors. And, moving away from the single seizure scenario, if the patient has had many seizures before the seizure-free interval, or evidence of an ongoing epileptic syndrome, even beyond 5 years seizure-free I counsel that there is always a risk of recurrence and being on epileptics may reduce this risk though they have not been proven to do so.

If one happens to see the patient for the first time at around 3 months after the event, and one has an EEG, then I think one might directly apply the MESS reanalysis of stratification of risk, namely to recommend treatment only if the patient is not in the lowest risk category. However, if the seizure was generalised tonic clonic, I am still uncertain about the applicability of that study, and I counsel the patient that while there is no clear evidence for treatment from clinical trials there are still arguments for as well as against treatment.

Of course, all these recommendations would only be a basis for discussion. Some patients may be focussed on taking a medication for any possible benefit to minimise risk of extended driving bans or sudden unexpected death in epilepsy. Others may not want to risk drug side effects unless they are of proven benefit or there is any possible risk of teratogenicity (despite the risk that a generalised tonic-clonic seizure in a mother poses to her unborn baby). I do counsel strongly that if one does embark on medication treatment for an unprovoked seizure there is little point in taking it for a period of less than 2 years. I also counsel them up front about the UK’s 3-month recommended period off driving at a future time of withdrawal of medication. Even this relatively short time off driving during this potentially risky period immediately after drug withdrawal could have important connotations for some patients who have been back driving again for 18 months.

Posted in Epilepsy | Tagged , , | 1 Comment

Primer on Statistics for Non-Statisticians

Many of the journals discussed assume a knowledge of statistics. In fact, it is often the statistics that are the crucial issue in a critical review of a research study. And paradoxically it seems that the further we move from the more scientific field of basic science and towards the more “accessible” field of clinical medicine, the statistics becomes more not less complicated.

“Hard” science might involve testing a complex hypothesis with a single complex experiment in a controlled, perhaps in vitro, environment. The experiment might have a few runs, or a few test subjects or perhaps only one. Statistics are all about estimation and sampling, so little if any statistics may be involved after the result is obtained – especially if there is only one result!

Oon the other hand, a clinical medicine study might involve a relatively easy to conceptualise hypothesis and easy measurements but tested on a real life subject where there are myriad other variables over which the investigator has no control. As a consequence, the test may have to be repeated in many different subjects in order to minimise the “noise” of random variabilities and maximise the “signal” of the variable under investigation. With repetition, the “signal” is amplified in an additive fashion, while the “noise” cancels out. Furthermore, in clinical medicine the hypothesis may be more vague; the investigation might involve an empirical study of a number of different factors which might interact with one another. Often the more vague the hypothesis, the more advanced the statistics required to make any sense of the data.

So I am really simply warning the reader, in a rather long-winded manner, that one may find the most advanced statistics lurking behind the abstracts of the most seemingly accessible research, and that probing the authors’ statistical interpretation of their data is sometimes the key to deciding how seriously to take their findings.

With this in mind I have attempted a statistical primer for the non-statistician, perhaps to dip into as a statistical topic comes up in a journal review, or perhaps to peruse in a more thorough manner. The contents link is below:

Primer on Statistics for Non-Statisticians: Introduction and Contents

Posted in Primer Posts for General Readers | Tagged , , | Leave a comment