Using atypical symptoms and red flags to identify non-demyelinating disease
- Siobhan B Kelly1,
- Elijah Chaila1,
- Katie Kinsella1,
- Marguerite Duggan1,
- Cathal Walsh2,
- Niall Tubridy1,
- Michael Hutchinson1
- 1Department of Neurology, St. Vincent's University Hospital, Dublin, Ireland
- 2Department of Statistics, Trinity College, Dublin, Ireland
- Correspondence to Professor Michael Hutchinson, Consultant Neurologist, St Vincent's University Hospital, Elm Park, Dublin 4, Ireland;
Contributors Michael Hutchinson devised the hypothesis and aim of the study. SB Kelly, E Chaila K Kinsella and M Duggan collected patient data. SB Kelly wrote the initial drafts and N Tubridy and M Hutchinson mentored and edited the drafts. C Walsh provided the statistical advice and analysis.
- Received 13 June 2011
- Accepted 4 July 2011
- Published Online First 17 August 2011
Background Red flags and atypical symptoms have been described as being useful in suggesting alternative diagnoses to multiple sclerosis (MS) and clinically isolated syndrome (CIS); however, their diagnostic utility has not been assessed. The aim of this study was to establish the predictive value of red flags and the typicality/atypicality of symptoms at presentation in relation to the final diagnosis of patients referred with suspected MS.
Methods All patients referred with suspected MS over a 3-year period were assessed by the typicality of the clinical presentation and the occurrence of red flags in relation to the eventual diagnosis. The extent of agreement of trainee and consultant neurologists as to typicality of clinical presentations was determined.
Results Of 244 patients referred, 119 (49%) had MS/CIS and 125 (51%) did not. 41 patients were referred because of an abnormal MRI. Of 203 with clinical symptoms, 96 patients had atypical symptoms of whom, 81 (84%) did not have MS and 15 (16%) had MS/CIS. Typical symptoms occurred in 107 patients; 10% did not have MS/CIS. Atypical symptoms had a sensitivity of 84%, specificity of 90% and positive likelihood ratio (PLR) of 8.4, whereas red flags had a sensitivity of 47%, specificity of 88% and PLR of 3.9 for the exclusion of MS/CIS. Mean percentage agreement between consultants and trainees was 73% with a range of 32–96%.
Conclusions Atypical features at presentation are more sensitive, specific and have a higher PLR than red flags to refute a diagnosis of MS/CIS.
The diagnosis of multiple sclerosis (MS) requires demonstration of dissemination of demyelinating lesions in both time and space using strict clinical, paraclinical and imaging criteria.1 However, many other diseases exhibit dissemination both in time and space; thus, the McDonald criteria emphasise that even if the clinical evidence and paraclinical studies are strongly indicative of MS, there must be “no better explanation for the findings than MS for a secure diagnosis”.1 2 The term ‘red flag’ is used to indicate that something in the history, examination or clinical investigation is not suggestive of MS. The utility of MRI in delineating the distinctive imaging features of diseases considered in the differential diagnosis of MS has been described by the European MAGNIMS (Magnetic Resonance Network in Multiple Sclerosis) group.3 In contrast to imaging criteria, specific guidance for appropriate clinical and laboratory assessment to satisfy the requirement for ‘no better explanation’ has not been available. A number of clinical and MRI red flags were provided by a group of experts and rated as being of major, intermediate or minor significance in suggesting an alternative diagnosis to MS.4 Major red flags (eg, bone lesions or the presence of diabetes insipidus) provide strong evidence of a different diagnosis to MS. Intermediate red flags, such as a prominent family history or seizures as a presenting feature, might suggest an alternative diagnosis; however, it may be entirely consistent with MS. The expert group also described minor red flags, such as the absence of gadolinium enhancement, which also occur with a relatively high frequency within the population of newly diagnosed patients with MS. Red flags may be useful in identifying patients who have another neurological disorder than MS; however, their predictive value has not been formally assessed. In addition to this, they do not address patients who have no determined medical cause for their symptoms—medically unexplained symptoms (MUS). The typicality or atypicality of symptom presentation has been proposed as a discriminator between patients who are or are not eventually diagnosed as having MS.5 Atypical symptoms indicate an alternative diagnosis to MS and include symptoms suggestive of either other neurological disease (OND) or MUS. Atypical symptoms suggesting OND include numbness in a glove and stocking distribution consistent with peripheral sensory neuropathy; those indicating MUS include non-dermatomal facial numbness and quivering, pulsing and or flitting sensations.
The aim of this study was to establish the predictive value of the absence or presence of red flags and the typicality/atypicality of presentation in relation to the final diagnosis in patients referred with suspected MS.
Patients and methods
All new patients referred with suspected MS to the neuroinflammation clinic at St Vincent's University Hospital, Dublin, were recruited retrospectively from January 2007 to December 2009 and then prospectively to May 2010. The reason for referral, whether clinical symptoms or an abnormal MRI suggestive of demyelination, was documented. The initial clinical appraisal and the presumed diagnosis at the time when the patient was seen were noted—for example, MS, OND or MUS. This was correlated with the final diagnosis after investigation. The diagnostic criteria for MS were that of McDonald's revised criteria.1 Patients with a monophasic demyelinating disease indicating a clinically isolated syndrome (CIS) were included with the MS group as MS/CIS. There were no patients with a diagnosis of neuromyelitis optica or acute disseminated encephalomyelitis. Clinical (not MRI) referrals were assessed by two consultant neurologists on the basis of symptoms at presentation, and unaware of the ultimate diagnosis, into those with symptoms which were typical or atypical for MS/CIS. Each neurologist assessed half of the clinical referrals. In all the patients presenting with clinical symptoms, the presence of one or more of the 79 red flags reported by the MS expert consensus group was recorded (by SK).4 For patients presenting with an abnormal MRI scan, the MAGNIMS red flag criteria were used.3 The diagnostic investigations included MRI brain and spinal cord, cerebrospinal fluid (CSF) examination for cells, protein, glucose, CSF IgG index and matched CSF and serum isoelectric focusing for oligoclonal bands. Other investigations were performed according to clinical indications including C-reactive protein, routine full blood count, biochemistry screen and aquaporin 4 antibodies, among others. Patients not diagnosed with MS/CIS were classified as having either OND or MUS. The basis for the diagnosis of OND was recorded. Where the patient had a final diagnosis of MUS, the reasoning behind this diagnosis was elucidated. A sample of 20 clinical vignettes (13 of patients with OND or MUS and 7 with MS/CIS) was given to 15 consultant neurologists and 11 neurologists in training. They were asked to assign ‘typical’ or ‘atypical’ presentations for MS to each of the vignettes.
Sensitivity, specificity, positive and negative likelihood ratios were calculated for the postulates that (a) typical symptoms at presentation indicated a MS/CIS diagnosis, (b) atypical symptoms at presentation indicated non-MS/CIS diagnoses, (c) the absence of red flags indicated a diagnosis of MS/CIS, (d) the presence of red flags indicated a non-MS/CIS diagnosis. Where p values are quoted, this refers to the χ2 test. SPSS V.16 was used for data analysis. Fleiss's Kappa was calculated as a measure of agreement for multiple raters.6
There were 244 referrals for suspected MS, 119 (49%) were diagnosed as having MS/CIS and 125 (51%) as not having MS/CIS (figure 1). Of the 119 patients diagnosed with MS/CIS, 33 had CIS, 71 had relapsing remitting MS, 2 had secondary progressive MS and 13 had primary progressive MS. Forty-one patients presented because of an abnormality found on a MRI scan; eight (20%) of the 41 were diagnosed with MS/CIS. Two hundred and three patients, referred with clinical symptoms suggesting a diagnosis of MS, were divided almost equally into those presenting with typical symptoms (107/203, 53%) and those with atypical symptoms (96/203, 47%). MS/CIS was diagnosed in 96/107 (90%) of the group presenting with typical symptoms and 15/96 (16%) of those presenting with atypical symptoms (p<0.0001).
Referrals with an abnormal MRI scan
An abnormal MRI scan of the brain or spine was the reason for the referral of 41/244 (17%) of the patients; most of these came from other hospital specialists. The most common reasons for performing an MRI were headache (10/41 (24%)), tinnitus, hearing loss or vertigo (8/41 (20%)), transient paraesthesiae and general malaise (4/41 (10%)) and symptoms of transient ischaemic attacks (5%). Only 8 (20%) of these 41 patients received a diagnosis of MS (figure 1); 2 had atypical symptoms (seizures and arthralgia), 6 had symptoms typical of demyelination and all 8 had CSF oligoclonal bands. Thirty-three patients (80%) did not have MS. Three patients were referred with a suspected spinal cord plaque which on further review was an artefact; these three did not have any neurological disorder. Of the remaining 30 patients with non-MS/CIS, the most common clinical diagnoses were small vessel disease 14/30 (47%) and migraine 9/30 (30%). Idiopathic myelopathy, CADASIL, mitochondrial disease, Behcet's disease and a benign cystic lesion with partial seizures were diagnosed in one patient each. Two other patients had small, scattered deep white matter lesions which were felt to be non-specific for any disease process. MRI red flags as defined by the MAGNIMS group3 were present in 28 (93%) of the 30 with a non-MS/CIS diagnosis. These were: absence of corpus callosum or cerebellar lesions in 13 patients, diffuse white matter lesions in 11, one of whom was diagnosed with CADASIL, infarcts in 2 patients, lacunar infarcts in 1 patient and 1 with brainstem lesions consistent with Behcet's disease.
Typical symptoms at referral and their predictive value for MS/CIS
There were 107 patients with typical symptoms of demyelination of whom 96 (90%) were diagnosed as having MS/CIS (figure 1). Only 11 patients (10%) did not have MS; 6 of these had OND and 5 had MUS. Typical symptoms at onset had a sensitivity of 90%, a specificity of 84%, a positive likelihood ratio of 5.63 and a negative likelihood ratio 0.12 for the diagnosis of MS/CIS (table 1).
Atypical symptoms at referral and their predictive value for non-MS/CIS diagnosis
There were 96 patients who presented with symptoms atypical for demyelination; 15 (16%) were diagnosed as having MS/CIS (figure 1). The two other diagnostic categories were 45 (47%) OND and 36 (37%) with MUS. Atypical symptoms at presentation had a sensitivity of 84%, a specificity of 90%, a positive likelihood ratio of 8.4 and a negative likelihood ratio 0.18 for the exclusion of MS/CIS (table 1).
The diagnostic spectrum of non-MS/CIS disorders
The non-MS/CIS diagnoses in 92 (45%) of the 203 patients presenting with clinical symptoms consisted of two groups, OND (51/92 (55%)) and MUS (41/92 (45%)). OND was diagnosed in 51 (25%) patients of the 203 with clinical symptoms at onset; 45/51 (88%) had atypical symptoms at presentation. For a list of the diagnoses and their frequency, see table 2. Thirty-nine of the 51 (76%) patients with OND had brain MRI, 14/51 (27%) had spinal cord MRI, 18/51 (35%) had CSF examination, 8/51 (16%) did not have imaging. In the majority of patients, further investigations, haematological, autoimmune, vascular imaging, EEG, EMG or muscle biopsy, were necessary to confirm the diagnosis.
The utility of red flags in the diagnostic process
Red flags were found in 84/244 (43%) of all referrals. In the MRI group, 28/30 (93%) of the patients with OND had red flags by MAGNIMS criteria.3 In the clinically referred patients, 56/203 (28%) had red flags, 16 (8%) patients had more than one red flag with 76 red flags in total. Red flags were found in 43/92 (47%) of the not MS/CIS group (41/51 (80%) patients with OND, 2/41 (5%) of patients with MUS) and 13/111 (12%) of patients with MS/CIS (p<0.0001). Red flags predicted that patients who were referred with clinical symptoms did not have MS/CIS with a sensitivity of 47%, a specificity of 88%, a positive likelihood ratio of 3.9 and a negative likelihood ratio 0.6 (table 1). The absence of red flags predicted a diagnosis of MS/CIS with a sensitivity of 88%, a specificity of 47%, a positive likelihood ratio of 1.7 and a negative likelihood ratio 0.26 (table 1).
The spectrum of red flags
In 56 clinically referred patients, there were a total of 76 red flags, 61 in 43/92 (47%) of the not MS/CIS group and 15 in 13/111 (12%) of those with MS/CIS. These were grouped by importance as major, intermediate and minor (tables 3 and 4).4 Of the 79 red flags suggested by the consensus group,4 we found at least one of 20 red flags (table 3). Seven red flags were common to both the MS/CIS and non-MS/CIS patients but more prevalent in not-MS/CIS patients (table 3). Eleven red flags were found only in the non-MS/CIS group and two red flags only in the MS/CIS group (table 4). In the 81 patients with atypical symptoms who were not diagnosed with MS/CIS, 52 red flags were found in 37 (46%) patients. Major red flags were more common than intermediate or minor red flags in this group. The presence of major and intermediate red flags rather than minor red flags was more likely to result in a diagnosis of OND (table 4). Ten of 15 patients with atypical symptoms and a diagnosis of MS had 11 false positive red flags (table 4). Four false positive red flags were found in three (3%) patients of 96 with typical symptoms and a diagnosis of MS/CIS; these were headache (major), myelopathy alone (minor), uveitis and a raised serum ACE level (both intermediate red flags) (table 4). Of the 11 patients who had typical symptoms but did not have MS/CIS, 6 patients had nine red flags consisting of two major, two intermediate and five minor red flags.
Medically unexplained symptoms
MUS was the ultimate diagnosis in 41 (20%) of the 203 patients presenting with clinical symptoms or signs. Thirty-six (85%) of these 41 patients had presented with atypical symptoms most commonly characterised by (a) non-dermatomal sensory disturbance, often transient lasting minutes or hours, flitting between body parts (65%), (b) fatigue, muscle twitching, ill-defined muscle weakness and pressure sensations and (c) transient visual symptoms (supplementary table e-1). All had a normal neurological examination. Despite the clinical diagnosis at presentation, all of these atypically presenting patients with MUS had an MRI brain scan, 19/36 (39%) had MRI of the spinal cord and 8/36 (22%) had CSF examination. MUS was also the diagnosis in 5 (6%) of the 113 patients with typical symptoms of demyelination; all these patients had a normal clinical examination, normal or small vessel disease on brain MRI, normal spinal cord MRI scans and normal CSF findings.
Agreement by clinicians as to typicality of presentation
For the 20 clinical vignettes, agreement among all 26 clinicians varied from 32% to 96% with a mean agreement of 73%. Overall, the κ statistic was 0.40 indicating modest agreement. There was one vignette of a patient with OND which produced poor agreement. Ten of the 20 cases received an agreement in excess of 80%. The mean agreement among consultants was 76% (range 60–100%) and that for trainee neurologists was slightly lower at 68% (range 50–100%). The κ statistic for consultants alone was 0.40 and for trainees alone was 0.33.
In this study, approximately half the patients referred with suspected demyelinating disease did not have MS/CIS. In a similar study from a USA university hospital of 281 referrals with suspected MS, only 94 (33%) of the patients were ultimately diagnosed as having MS.5 In that population, the frequencies of MS diagnoses were: 11% of 103 patients presenting with an abnormal MRI scan, 66 (88%) of 75 patients with typical symptoms of demyelination, 17 (27%) of 75 patients with possible symptoms and none of 41 patients with atypical symptoms. In the USA study, 103 (37%) of referrals were because of an abnormal MRI scan,5 whereas only 17% of our patients presented in this manner. This may reflect differences in referral pattern in the two health systems.
We have shown that the clinical impression at the first consultation in a specialty clinic as to typicality or atypicality of symptoms or signs is a more sensitive predictor as to the eventual diagnosis than the presence or absence of red flags. In particular, atypical symptoms suggested a diagnosis other than MS with greater positive and negative likelihood ratios (8.4, 0.18) than the presence of red flags (3.9, 0.6). Patients with typical symptoms have a high probability of a diagnosis of MS/CIS and have a low prevalence of red flags. Patients with atypical symptoms have a high likelihood of having either OND or MUS. Although red flags are useful in alerting the neurologist to OND (present in 80% OND), they are much less useful in suggesting a non-MS/CIS diagnosis (OND and MUS combined). This is because patients with MUS, who formed 20% of clinical referrals, had a low (5%) prevalence of red flags.
The entity, MUS, can be regarded as a form of MS mimic which, despite its prevalence, does not appear in many reviews of the differential diagnosis of MS/CIS.7–9 These are patients with no identifiable disease who present with atypical symptoms, have no objectively abnormal motor or sensory signs and have normal investigations. These patients formed 23% of the 281 patients referred to a USA clinic with suspected MS5 and 20% of patients in our study. Patients with MUS are also seen in general neurology clinics; 90/300 (30%) new general outpatient neurology referrals in Scotland had MUS.10 There is evidence that these patients when followed have persisting symptoms and do not develop any identifiable disease.11 12 The majority (92%) of the 63 patients with atypical and possible symptoms seen in the USA study5 were considered to have a primary psychiatric disorder by DSM-IV-TR (Diagnostic and Statistical Manual of Mental Disorders-IV-text revision) criteria, mainly somatoform disorder, mood disorder and anxiety disorder.13
Patients with OND form the other main diagnostic group in patients referred with suspected MS. They formed 30/41 (73%) of MRI scan referrals, 25% of our referrals with clinical symptoms and 88/281 (31%) in the similar US study.5 It is this group, often regarded as true MS mimics, in which red flags are most useful. Red flags were found in 41/51 (80%) of our patients with OND and in 13/111 (12%) of the MS/CIS patients. It is clear that the presence of a red flag in the setting of a referral with suspected MS suggests that the diagnosis is not MS, and extra effort should be made to establish or deny the diagnosis. One limitation of our study in delineating red flags is the small study population. Given the large number of diseases considered as MS mimics, there are several hundred red flags, varying considerably in their prevalence. Thus, a study of the validity of individual red flags might only be achieved in a large multicentre study recruiting several thousand patients.
Another limitation of the study is the reproducibility of the essentially subjective judgement as to typicality or atypicality of symptom presentation. We attempted to assess this in a small validation study including experienced consultants and trainee neurologists. This is a rather imperfect experiment because a written clinical description will not convey all the nuances of the clinical setting and is affected by heuristics, which will vary between the clinicians.14 The advantage of using atypical presentations as a marker of non-demyelinating disease is that it includes the entity of MUS, which is not uncommon in the clinic. Red flags have a useful secondary role since their presence is indicative of OND. Thus, we advocate a sequential diagnostic process, using first the typicality of the presentation and then noting the presence of red flags in order to determine whether there is any better explanation than MS for the clinical presentation.
Competing interests N Tubridy receives a grant support from Bayer Schering and the Health Research Board of Ireland and MS Ireland. M Hutchinson serves on a medical advisory board [BG00012] for Biogen-Idec; serves on the editorial boards of the Multiple Sclerosis journal and the International MS journal and receives research support from Dystonia Ireland and the Health Research Board of Ireland.
Ethics approval The study was approved by the Medical Research Ethics Committee of St. Vincent's University Hospital, Dublin.
Provenance and peer review Not commissioned; externally peer reviewed.