If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Sheffield Teaching Hospitals NHS Foundation Trust, Royal Hallamshire Hospital, Glossop Road, Sheffield S10 2JF, United KingdomDepartment of Academic Neurology, University of Sheffield, Beech Hill Road, Sheffield S10 2RX, United Kingdom
Sheffield Teaching Hospitals NHS Foundation Trust, Royal Hallamshire Hospital, Glossop Road, Sheffield S10 2JF, United KingdomDepartment of Academic Neurology, University of Sheffield, Beech Hill Road, Sheffield S10 2RX, United Kingdom
Systematically reviews inter-ictal criteria for differential of transient loss of consciousness.
•
Identifies lack of highly-predictive validated diagnostic criteria.
•
Identifies lack of decision rules validated against gold-standard reference diagnoses.
•
Future research should combine identified criteria in decision rules to support diagnosis.
Abstract
Background
Transient loss of consciousness (TLOC) is a common presentation in primary care. Over 90% of these are due to epileptic seizures (ES), syncope, or psychogenic non-epileptic seizures (PNES). Misdiagnosis rates are as high as 30%.
Methods
Systematic review of inter-ictal clinical criteria to aid differential diagnosis of TLOC. We searched Medline, EMBASE, CINAHL and PsycInfo databases, as well as relevant grey literature depositories and citations of relevant reviews and guidelines for studies giving sensitivity and specificity of inter-ictal clinical characteristics used to differentiate between causes of TLOC. Two independent reviewers selected studies for inclusion and performed critical appraisal of included articles. We performed a narrative synthesis of included studies.
Results
Of 1023 results, 16 papers were included. Two compared syncope, ES, and PNES; all others compared ES and PNES. All were at significant risk of bias in at least one domain. 6 studied patient symptoms, 6 medical and social history, 3 witness reports and 1 examination findings. No individual criterion differentiated between diagnoses with high sensitivity and specificity.
Conclusions
There is a lack of validated diagnostic criteria to help clinicians assessing patients in primary or emergency care settings to discriminate between common causes of TLOC. Performance may be improved by combining sets of criteria in a clinical decision rule, but no such rule has been validated prospectively against gold-standard diagnostic criteria.
]. Accurately distinguishing between these is vital to allow appropriate management and identification of patients at risk of morbidity/mortality from different underlying conditions [
]. Most patients will not be assessed by a health professional during or in the immediate aftermath of a TLOC event, and the post-hoc diagnostic process is complicated by a lack of unique distinguishing clinical features [
Epidemiological characteristics and diagnostic approach in patients admitted to the emergency room for transient loss of consciousness: Group for Syncope Study in the Emergency Room (GESINUR) study.
The emergency or primary care management of many different presentations can be enhanced by clinical decision rules which have been shown to lead to more cost-effective care and improve patient outcomes. [
], but the research studies underpinning these reviews were typically based on observations made during the video-EEG recording of episodes, not on more readily available but much less reliable information from witnesses [
], meanwhile, show that their utility is highly dependent on timely sampling and that they are not sufficiently reliable for diagnostic purposes in unselected patients in primary or emergency care settings. We seek to review the literature on candidate criteria for clinical decision rules for patients first presenting with TLOC, i.e. on features that may help guide the most appropriate further investigation and treatment of patients who were not assessed during or immediately after an episode.
2. Methods
We performed this systematic review according to a pre-specified (but not pre-registered) protocol, available from the authors on request. We report below according to PRISMA guidelines [
]. It was based on primary research studies fulfilling the following eligibility criteria:
2.1 Eligibility criteria
2.1.1 Study type
•
All research studies comparing the scope of clinical features or basic investigations, alone or in combination, to discriminate between at least two of ES, syncope, and PNES.
•
We excluded case reports/series, reviews, guidelines, or other synthesis or non-research articles.
•
We excluded studies not in English.
•
We imposed no limitation on publication date.
2.1.2 Participants
•
Studies involving only patients ≥16 years old.
•
Minimum sample size 5 patients per group.
•
We excluded studies including participants with ES or PNES without disturbance of consciousness e.g. brief motor or purely sensory symptoms.
2.1.3 Reference test
•
‘Gold-standard’ diagnostic criteria.
○
ES, PNES: Expert diagnosis using evidence from video-electro-encephalogram (vEEG) capture of an attack that is confirmed by patient and/or witnesses to be typical of the patient’s usual attacks, with (E) or without (PNES) corresponding epileptiform EEG changes (corresponding to the highest [‘documented’] level of certainty for the clinical diagnosis of PNES according to consensus criteria) [
Syncope: Expert diagnosis supported by pathophysiological evidence e.g. positive tilt-table test findings or syncopal/pre-syncopal symptoms synchronous with ECG or explanatory blood pressure changes.
2.1.4 Index tests
•
Index tests should only involve information and investigations likely to be accessible for patients presenting to primary or emergency care settings post-episodally. If any index tests were identified that were not explicitly covered in the criteria below, two independent raters (AW, EN) assessed their appropriateness for inclusion; in cases of disagreement a third rater (MR) settled the dispute.
•
Included were:
○
Patient descriptions of attacks, peri-episodal and inter-episodal symptoms;
○
General medical history e.g. comorbidities;
○
Witness descriptions of attacks and collateral history;
○
Inter-episodal clinical examination;
○
Simple bedside investigations e.g. ECG.
•
Excluded were:
○
Tests reliant upon direct observation of episodes;
○
Specialist investigations e.g. EEG, tilt-table testing;
○
Psychological inventories where results were not analysed at the individual-item level (such inventories are impractical for use in the primary care clinical setting due both to time required and copyright issues);
○
Tests for which safety concerns may preclude performance in primary care e.g. induction procedures.
○
Laboratory blood tests (these were included in the initial protocol; however, authors subsequently agreed that, given the time-dependence of sampling their use is limited in the contexts addressed by the review question. Other recent reviews address their utility in the diagnosis of TLOC in more immediate post-episodal settings) [
The primary outcome was diagnostic performance of index test compared to reference standard, quantitatively evaluated as sensitivity and specificity (or with sufficient data provided to allow calculation of these e.g. from contingency tables). For index tests comparing more than two populations, we required data sufficient to calculate overall diagnostic accuracy, accuracy for each diagnosis, and sensitivity/specificity for each diagnosis against all others.
2.2 Information sources and search strategy
We searched the Medline, EMBASE, CINAHL and PsycInfo databases to identify relevant papers, using strategies tailored to each database (Appendix 1) drawing on SIGN recommendations [
]. We also performed a free-text search of the OpenGrey grey literature repository, a hand -search of Cochrane database of systematic reviews for all studies under “Heart and Circulation”, “Neurology”, and “Mental Health” tagged as “diagnostic”. We also searched the reference sections of all identified studies, systematic reviews of related topics; and relevant NICE and ESC guidelines for additional relevant primary research studies [
Guidelines for the diagnosis and management of syncope (version 2009): the Task Force for the Diagnosis and Management of Syncope of the European Society of Cardiology (ESC).
A single reviewer (AW) screened the titles and abstracts of all studies initially identified to exclude papers clearly not relevant to the review question (e.g. incorrect article type, not addressing TLOC). Two reviewers (AW, EN) then independently performed more detailed screening of retained studies first by title and abstract, before evaluating full texts of all studies that had passed screening. Reviewers were not blinded to author or publication details. In cases of disagreement, reviewers discussed to reach consensus, with a third author (MR) available to adjudicate in cases of persistent dissensus.
2.4 Data collection and critical appraisal
Two reviewers (AW, EN) independently performed critical appraisal and data extraction, with disagreements resolved by discussion to reach consensus. We used a modified version of the QUADAS-2 tool for studies of diagnostic accuracy [
], to assess risk of bias. We extracted relevant data on a pre-specified data extraction form pilot-tested on three studies. Unless authors provided explicit theoretical or practical motivation for an alternative, we defined statistical significance at α = 0.05; with studies involving multiple comparisons, we used the Bonferroni correction to maintain family-wise error rate (FWER) = 0.05; comparisons with uncorrected p < 0.05 were reported as trending toward significance.
2.5 Synthesis of results
Given the broad scope of the review question, we expected to find a range of different index criteria and a significant degree of clinical and methodological heterogeneity in our results; as such a quantitative, meta-analytic approach would be inappropriate. We instead performed a narrative and tabular synthesis of identified studies via content analysis, grouping them under pre-specified types: symptoms; history and comorbidities; witness reports; clinical examination; and simple investigations. We used the Cochrane Review Manager (RevMan) 5.3 [
] to produce sensitivity and specificity forest plots for all results. Similarly, the broad scope and likely heterogeneity of the index criteria limited the utility of quantitative/graphical assessment of publication bias, but we attempted to identify selective reporting during our synthesis. We appraised quality of evidence for each reported diagnostic criterion using GRADE criteria for diagnostic studies [
Details of the study screening and evaluation process are highlighted in Fig. 1. Database searching was performed using the NICE Healthcare Databases Advanced Search (HDAS) tool, considering all articles published up to 13 December 2017. In addition to the initially-captured studies, we also identified 13 relevant reviews that were included in citation searching. [
Guidelines for the diagnosis and management of syncope (version 2009): the Task Force for the Diagnosis and Management of Syncope of the European Society of Cardiology (ESC).
Use of serum prolactin in diagnosing epileptic seizures: report of the Therapeutics and Technology Assessment Subcommittee of the American Academy of Neurology.
]. The most common reasons for exclusion were: diagnoses not confirmed by gold-standard investigations; combining participant groups (several combined PNES and ES; one combined PNES and syncope); mixed paediatric/adult populations; and insufficient information to evaluate performance of diagnostic criteria. Grey literature and hand-searching identified no further relevant studies. Citation-searching identified a further two papers, as well as one potentially relevant paper that was inaccessible, whose authors did not respond to inquiries [
] all others compared ES with PNES. As some studies involved overlapping patient groups, we cannot state how many participants were captured by the studies overall. All studies were performed in industrialised OECD nations, the majority in the USA and the remainder in Western Europe. Half of the included studies used a retrospective chart review to compare patient groups; others recruited participants prospectively. All participants were recruited from secondary/tertiary care settings, largely from Epilepsy Monitoring Units (EMUs) (studies including a syncope patient group also recruited from individuals referred to a secondary-care syncope service [
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
Table 2 summarises bias/applicability assessments. All included studies were at significant risk of bias in at least one domain. Most commonly, studies did not use separate patient groups for derivation and validation of an index test, [
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
], Both of these are likely to result in model overfitting and significant over-estimation of performance. Frequently individuals involved in the index performance/evaluation were not blinded to results of the reference standard diagnosis [
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
]), while several others provided insufficient information to appraise blinding (Table 1). Applicability concerns arose from the secondary care setting of all included studies. Box 1 summarises potential sources of bias.
Table 2Summary of critical appraisal and assessment of applicability of included studies. Y = Yes. N = No. Unc = Unclear. Robles et al 2015 [
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
Epidemiological characteristics and diagnostic approach in patients admitted to the emergency room for transient loss of consciousness: Group for Syncope Study in the Emergency Room (GESINUR) study.
Guidelines for the diagnosis and management of syncope (version 2009): the Task Force for the Diagnosis and Management of Syncope of the European Society of Cardiology (ESC).
Use of serum prolactin in diagnosing epileptic seizures: report of the Therapeutics and Technology Assessment Subcommittee of the American Academy of Neurology.
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
Recruitment from specialist settings (potentially capturing a different patient group from those presenting with new-onset TLOC disorders).
○
Recruitment of patients with longer-term disorders
○
Recruitment of patients with disorders characterised by sufficiently frequent TLOC events for events to have been captured during monitoring of physiological functions
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
] as ‘symptom-based’ criteria. Two studies reported results from the same patient sample; they were the only two studies to compare patients with all three common causes of TLOC (ES, syncope, and PNES)..440 None featured independent validation samples.
] found that an 86-item peri-episodal symptom questionnaire, together with basic demographic and historical details, predicted 78.4% of diagnoses accurately (syncope 91%, ES 66%, PNES 78%) via multinomial logistic regression. Most classification errors arose from labelling ES as PNES or vice versa. They did not provide results for individual symptoms, as the predictive models were based on a five-factor model onto which different symptom scores were weighted. In a post-hoc interpretation they described these as: ‘feeling overpowered’; ‘sensory experience’; ‘amnesia’; ‘mind/body/world disconnection’; and ‘catastrophic experience’.
Rawlings et al. focused on panic symptoms in the same dataset, constructing an ‘ictal panic score’ from 7 panic-related symptoms in the questionnaire. They found that a receiver operating curve (ROC) statistic yielded an area under curve (AUC) of 0.74 for diagnosing PNES (against ES or syncope); a post-hoc threshold ictal panic score ≥ 12.5 identified PNES with a sensitivity of 71.1% and a specificity of 71.2%. They were less successful in discriminating ES from syncope, with AUCs for ES v (syncope or PNES) and syncope v (ES or PNES) of 0.44 and 0.32 respectively. These results are qualitatively consistent with Hendrickson et al’s retrospective chart review of DSM-IV-TR panic symptoms present peri-episodally (AUC 0.782 for diagnosing PNES).
A recent systematic review suggests that observed ictal eye closure is less predictive of PNES than previously thought [
]. In their prospective study Syed et al. found that self-report does not differ significantly from chance (sensitivity = 53.5%, specificity = 50.7%). Ettinger et al. retrospectively examined post-ictal headache, lethargy, and confusion; while confusion did not differ significantly between ES and PNES (p = 0.960), both post-ictal headache (p = 0.008, sensitivity=37.5%, specificity=95.7%) and lethargy did (p = 0.004, sensitivity=56.3%, specificity=87.0%).
Two studies used ROS questionnaires to distinguish ES from PNES, hypothesising that PNES patients would endorse more complaints than those with ES. The 79-item questionnaire used by Robles et al [
]. (AUC = 0.845) performed better than the ten items proposed by Asadi-Pooya et al’s46 (AUC = 0.67). Both studies derived post-hoc threshold percentages of positive complaints to diagnose PNES (Robles et al: sensitivity = 78.3%. specificity = 85.7% for ≥17% positive symptoms; Asadi-Pooya et al: sensitivity = 40.0%, specificity = 90.0% for >25% positive).
Fig. 2 summarises the findings of studies based on symptoms. As the vast majority of studies focused on ES and PNES populations, we present only sensitivity and specificity of criteria for the differentiation between ES and PNES. Given concerns regarding study design, indirectness of reported results as surrogates for patient-important outcomes, and non-representativeness of study samples, reported results constitute very low-quality evidence according to GRADE criteria [[
],]. This was also the case for all further outcomes reported below and as such we do not comment further on quality of evidence.
Fig. 2Summary outcomes for symptom-based criteria for detecting PNES. aQuoted figure compares PNES group v (E or S) bSensitivity and specificity for pairwise logistic regression comparing PNES v E.
Historical criteria focus on aspects of the patient’s background other than symptoms. Six studies used historical criteria, all involving ES and PNES groups: three focused on comorbidities; [
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
] and psychiatric disorders including post-traumatic stress disorder (PTSD), depression, and panic or anxiety disorder, individually or in combination [
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
Data at individual diagnosis level is available for some comorbidities. Benbadis et al. find that a diagnosis of CP or FM is highly specific (99%) for PNES, but with low sensitivity (9%); [
Diagnoses of neurobehavioral paroxysms in veterans of operation enduring freedom/ operation iraqi freedom (OEF/OIF) - Experiences from a VA epilepsy center.
]. They also found between-group differences in rates of PTSD (clinical diagnosis by psychiatrist; sensitivity = 63%/specificity = 81.3%) and both PTSD and mTBI (sensitivity = 41.3%/specificity = 87.5%), though Arnold and Privitera found lower PTSD rates in a general EMU population (PTSD diagnosis [based on clinical assessment by a psychiatrist familiar with the structured questionnaire] yielding sensitivity = 36% / specificity = 85% for PNES) [
]. Rates of depression and panic disorder (diagnosed by a psychiatrist as above) did not differ significantly between their ES and PNES groups, consistent with the findings of Schramke et al., who found depression, panic disorder, and anxiety disorder to be more prevalent in PNES, though only the latter (sensitivity = 50%, specificity = 78%) differentiated it significantly from ES (p < 0.002, FWER = 0.05) [
]. In the only prospective evaluation of comorbidity scores, Arnold and Privitera found no significant association between current or lifetime Axis I or Axis II psychiatric diagnoses and PNES (uncorrected p > 0.05) [
]. Dixit et al.’s retrospective study used a pre-specified criterion of ≥1 diagnosis of a functional somatic syndrome (CFS, FM, CP, IBS) or chronic physical health condition with paroxysmal symptoms (headache, asthma, GORD) as an indicator of PNES (sensitivity = 65.6% / specificity = 73.0%).
Three studies evaluated historical criteria other than comorbidities. Arnold and Privitera found that a history of traumatic experience (especially sexual or physical abuse) predicted PNES (sensitivity = 86%, specificity = 67%) [
]. Schramke et al. reviewed a range of historical criteria (determined by retrospective review of clinical interviews conducted by a psychologist at EMU admission) [
]. They also found that childhood abuse or neglect significantly (p < 0.002, FWER=0.05) predicted PNES (sensitivity=57%, specificity=88%). Other criteria significantly associated with PNES were marital instability, a family history of seizure disorder or alcohol abuse, and psychotropic medication use. Features that tended toward significance in predicting PNES (uncorrected p ≤ 0.05) included a history of sexual abuse, female gender, a history of psychiatric hospitalisation or drug/alcohol abuse, a family history of CFS, fibromyalgia or psychiatric disorder, and an unstable work history. Features not predictive of a diagnosis included pending litigation or disability claims, a healthcare background, and a history of antisocial behaviour or head injury. Combining five highly-predictive features chosen for minimal covariance (age at first spell; psychiatric diagnosis other than depression or anxiety; marital instability; anxiety disorder; years of education) via logistic regression accounted for 44% of variance in classification outcome, with an accuracy of 87%.
The remaining study described a historical score incorporating age and social stressors at TLOC onset, comorbid psychiatric or chronic pain diagnoses, number of reported allergies, unusual seizure triggers, and a history of TLOC in healthcare setting or involving serious injury. The authors claimed 89.5% sensitivity and 88.5% specificity for PNES in the derivation sample using a score threshold derived post-hoc; no validation figures or AUC are given, indicating that these results are associated with a very high risk of bias [
]. used two independent epileptologists’ evaluations of vEEG recordings to identify a set of three best semiological predictors of ES (eye-opening or widening at onset; abrupt onset; post-ictal confusion or sleep) and PNES (apparent preserved awareness; eye flutter; intensification/alleviation of attack by others), and then assessed whether witness-reports of these features predicted diagnosis in a validation sample. None of the reported features emerged as a statistically significant predictor in logistic regression. The same group also found in an earlier study that witness report of another commonly-cited potentially diagnostic semiological feature of PNES, ictal eye closure, is a poor diagnostic criterion [
] who used a prospective blinded questionnaire design in a similar patient population to examine witness-reported semiology, both as single features and as a combined score derived post-hoc via logistic regression. Of twelve features described, they classified six as ‘stronger’ predictors (accuracy≥0.7): ictal eye opening, side-to-side head movements, and duration; and post-ictal deep, loud, or snoring breathing. Three were ‘intermediate-strength’ (0.7 > accuracy≥0.6) – ictal mouth opening, and post-ictal irregular or prolonged abnormal breathing. The remaining three (continuous motor activity, limb synchrony, and post-ictal confusion) were poor predictors. In logistic regression, only ictal eye-opening and duration, and post-ictal deep, loud, or snoring breathing were statistically significant predictors. They claim that their post-hoc score could distinguish ES from PNES (sensitivity = 84.2%, specificity = 84.6%) but provide insufficient information to support this statement.
]. Oliva et al retrospectively reviewed the presence of intra-oral lacerations on examination of EMU patients, finding the presence of oral lacerations (tongue, cheek, or lip) was specific (100%) but not sensitive (26%) for ES (see Fig. 5).
Fig. 5Summary outcomes for examination criteria for detecting PNES.
The review process highlighted several studies that did not meet the original inclusion criteria but which the authors feel worthy of comment in that they go further than any of the included studies in developing and validating CDRs for differential diagnosis of TLOC at first presentation, and additionally include syncope patient groups.
We identified two candidate CDRs for discriminating between syncope and ‘seizures’ at first presentation in primary or emergency care. The most comprehensively evaluated of these is the 9-point witness/symptom score developed by Sheldon et al. [
] which predicts syncope versus seizures or on the basis of six criteria positively correlated with seizures (waking with cut tongue, witness-reported abnormal behaviour [amnesia, unresponsiveness, unusual posturing or limb jerking], association with emotional stress, post-ictal confusion, unilateral head-turning, and prodromal deja/jamais vu), and three correlated with syncope (presyncope, loss of consciousness with prolonged standing or sitting, and pre-episodal diaphoresis). The original study quoted sensitivity and specificity of 94% for seizures in a separate validation sample of patients; independent prospective validation claims sensitivity and specificity for syncope of 86.54% and 92.13% [
]. However, these studies do not distinguish ES from PNES, and did not use gold-standard vEEG-based diagnostic criteria for defining the seizure group; Sheldon et al state that the ‘seizure’ group contained patients with generalised seizures and focal with secondary generalisation, determined by ‘positive’ EEG alone. A briefer, four-point score proposed by Hoefnagels et al predicted ES versus syncope “or other causes” on the basis of: post-episode disorientation; lack of pre-episodal diaphoresis; age ≤ 45y; and tongue-biting. They report only that expected frequencies “agree well” with observed frequencies and did not validate their model on a separate sample. Furthermore, diagnosis of ES was based purely on semiological criteria (clonic movement, automatism, or aura) and no clear criteria for diagnosis of syncope were given, limiting applicability of their results.
Two further studies provided CDRs for distinguishing ES from PNES. Kerr et al. propose an 11-point history-based score (using information on comorbidities, gender, and medication history) to distinguish ES (all forms except temporal-lobe epilepsy) and PNES in a combined paediatric and adult patient group. They found that a history of migraines, asthma, and chronic pain, as well as overall number of comorbidities predicted PNES, while diabetes mellitus and non-metastatic neoplasia suggested ES; female gender and overall number of non-anti-epileptic or psychiatric medications were also predictive of PNES, while the number of current and previously tried antiepileptic drugs predicted ES. Their regression-based score identified PNES with 90% sensitivity and 55% specificity in a validation sample [
]. Syed et al. adopted a different approach from conventional regression-based CDR development, developing a classifier to predict PNES using machine-learning methods able to exploit non-linear interactions between predictors [
]. Their classifier predicted PNES on the basis of 53 patient-reported questionnaire responses covering a range of psychosocial variables with 85% sensitivity and specificity, though patients with comorbid ES were also present in the PNES group and no syncope group was included.
4. Discussion
This review highlights that there is a lack of validated diagnostic criteria to help clinicians assessing patients in primary or emergency care settings to discriminate between the common causes of TLOC; ES, syncope and PNES. All included studies focused on patients in secondary/tertiary-care settings, who may differ from patients at first presentation in important respects. Only two studies aiming to identify potential diagnostic criteria included patients with syncope [
], although the most pressing concern in the initial diagnostic assessment by a non-expert clinician may be the differentiation between syncope on the one hand and seizures (ES or PNES) on the other. Many patients with syncope may be adequately managed in emergency or primary care settings. Only some will require further investigation (typically by experts in cardiology or internal medicine). The most appropriate management of patients presenting with TLOC in the context of a new seizure disorder (ES or PNES) is likely to require referral to a clinician with expertise in the diagnosis and treatment of such disorders (typically a neurologist).
Candidate diagnostic criteria were generally limited by poor sensitivity or specificity. Prediction rates were improved by combining individual features in a number of studies, [
] but none of these combination scores have so far been validated prospectively against gold-standard diagnoses in settings in which unselected patients present and many of the proposed scores do not discriminate between all common causes of TLOC.
The review protocol introduced two important sources of potential bias. The requirement of a gold-standard diagnosis as reference standard necessarily excludes participants for whom such a diagnosis is not reached. Exclusion of these ‘difficult to diagnose’ patients overestimates performance of index tests [
]. Furthermore, the requirement that patients had gold-standard diagnoses meant that in all included studies participants were recruited from secondary/tertiary care settings, leading to significant applicability concerns. Patients referred to specialist services (EMUs, syncope services) are likely to differ from those with a first presentation of TLOC in duration of symptoms, complexity of symptoms, difficulty of diagnosis, and other dimensions. Settings may also differ in the relative frequency of different diagnoses (for instance, PNES may be over-represented and epilepsy underrepresented in EMU compared to primary care settings). This is notable as the diagnostic value of particular observations will in part depend on the frequency distribution of the different diagnoses in the population studied. Furthermore, by requiring pathophysiological evidence to support expert diagnosis in our ‘gold-standard’ criterion for syncope diagnosis, we would potentially exclude patients for whom a clear diagnosis of vasovagal syncope could, according to current guidelines, be made on the basis of history alone [
]. However, on reviewing excluded studies none was excluded for this reason alone.
It is also important to note that all included studies were conducted in industrialised OECD nations in the Western hemisphere. It is feasible that some of the factors identified would not have the same diagnostic potential in less-industrialised or non-Western nations (for instance, gender distribution of PNES and association with sexual abuse may differ between USA/Western Europe and Iran [
This review demonstrates the need for development and validation of diagnostic tools to aid differential diagnosis of TLOC. Machine-learning classifiers have the potential to exploit non-linear interactions between predictors,57 although they may be challenging to implement in primary care. A predictive tool would not need to classify patients perfectly; even if imperfect, it could be used to guide initial investigation and referral pathways. Furthermore, such a quantitative measure could provide a numeric pre-test probability of particular diagnoses that would help with the interpretation of test results [
]. Ideally a diagnostic tool used in this setting would also identify TLOC presentations of patients at particular risk – for instance of sudden cardiac death or Sudden Unexpected Death in Epilepsy (SUDEP).
Funding
Part of this work was undertaken during a Health Education England Academic Foundation Research programme at the University of Sheffield. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Conflict of interest statement
We have no Conflicts of interest to declare.
Appendix A. Supplementary data
The following is Supplementary data to this article:
Epidemiological characteristics and diagnostic approach in patients admitted to the emergency room for transient loss of consciousness: Group for Syncope Study in the Emergency Room (GESINUR) study.
Guidelines for the diagnosis and management of syncope (version 2009): the Task Force for the Diagnosis and Management of Syncope of the European Society of Cardiology (ESC).
Use of serum prolactin in diagnosing epileptic seizures: report of the Therapeutics and Technology Assessment Subcommittee of the American Academy of Neurology.
30165-1&id=gr1.jpg){kind=link}
30165-1&id=gr2.jpg){kind=link}
30165-1&id=gr3.jpg){kind=link}
30165-1&id=gr4.jpg){kind=link}
30165-1&id=gr5.jpg){kind=link}