Tongue biting in epileptic seizures and psychogenic events

Epilepsy & Behavior 25 (2012) 251–255 Contents lists available at SciVerse ScienceDirect Epilepsy & Behavior journal homepage: www.elsevier.com/locate/yebeh Review Tongue biting in epileptic seizures and psychogenic events. An evidence-based perspective Francesco Brigo a, b,⁎, Monica Storti c, Piergiorgio Lochner b, Frediano Tezzon b, Antonio Fiaschi a, Luigi Giuseppe Bongiovanni a, Raffaele Nardone b, d a Department of Neurological, Neuropsychological, Morphological and Movement Sciences, Section of Clinical Neurology, University of Verona, Italy Department of Neurology, Franz Tappeiner Hospital, Merano, Italy Department of Medicine, University of Verona, Italy d Department of Neurology, Christian Doppler Clinic, Paracelsus Medical University, Salzburg, Austria b c a r t i c l e i n f o Article history: Received 13 June 2012 Revised 24 June 2012 Accepted 26 June 2012 Available online 2 October 2012 Keywords: Epileptic seizures Likelihood ratio Meta-analysis Psychogenic non-epileptic events Sensitivity Specificity Tongue biting a b s t r a c t Tongue biting (TB) may occur both in seizures and in psychogenic non-epileptic events (PNEEs). We undertook a systematic review to determine sensitivity, specificity, and likelihood ratios (LR) of TB. Five studies (222 epilepsy patients and 181 subjects with PNEEs) were included. There was a statistically significant higher prevalence of TB (both without further specifications on site of lesions and lateral TB) in patients with seizures. Pooled accuracy measures of TB (no further specifications) were sensitivity 38%, specificity 75%, pLR 1.479 (95% CI 1.117–1.957), and nLR 0.837 (95% CI 0.736–0.951). Pooled measures of lateral TB were sensitivity 22%, specificity 100%, pLR 21.386 (95% CI 1.325–345.169), and nLR 0.785 (95% CI 0.705–0.875). Only a pooled analysis of data demonstrated a statistically significant pLR for lateral TB. Lateral TB but not ‘any’ TB has diagnostic significance in distinguishing seizures from PNEEs, supporting the diagnosis of seizures. Tongue biting without further specifications has, therefore, no value in the differential diagnosis between seizures and PNEEs. © 2012 Elsevier Inc. All rights reserved. 1. Introduction Paroxysmal episodes of loss of consciousness are rarely witnessed by physicians, and the differential diagnosis between epileptic seizures and other episodes is usually based on the history. However, even if witnesses can be given an accurate description of the event, the diagnosis may be difficult and often remains uncertain [1]. The differential diagnosis of paroxysmal episodes of loss of consciousness mainly includes epileptic seizures, syncope, and psychogenic non-epileptic events (PNEEs). Tongue biting (TB) has long been considered a useful clinical feature in the diagnosis of epileptic seizures. However, oral lacerations and TB may occur both in seizures and in PNEEs. Although a lateral TB has been reported to be highly specific for epileptic seizures [1,2], a comprehensive search of the literature to determine the accuracy of this physical finding (with special regard to its positive likelihood ratio) and its prevalence in epileptic seizures and in PNEEs has not yet been performed. ⁎ Corresponding author at: Department of Neurological, Neuropsychological, Morphological and Movement Sciences, Section of Clinical Neurology, University of Verona, Piazzale L.A. Scuro, 10‐37134 Verona, Italy. Fax: +39 0458124873. E-mail address: [email protected] (F. Brigo). 1525-5050/$ – see front matter © 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.yebeh.2012.06.020 In this study, we, therefore, aimed to undertake a systematic review and a meta-analysis of studies evaluating the prevalence of TB in patients with epileptic seizures and PNEEs and to determine sensitivity, specificity, and likelihood ratios (LR) of this physical finding. 2. Methods Our aim was to critically and systematically evaluate the literature to determine (A) the prevalence of TB in patients with epileptic seizures and PNEEs as reported in the literature and (B) the sensitivity, specificity, positive LR (pLR), and negative LR (nLR) of this physical finding. We included prospective and retrospective studies comparing the prevalence of TB between patients with epileptic seizures (all types) and patients with PNEEs. Only data on tongue lesions (not lacerations to the cheek, to the lip, or in other sites) were considered. No age, race, or gender restrictions were applied. Studies could rely on historical reports of TB from patients, on direct examination of patients who presented to the emergency department after a seizure, or on videoEEG monitoring evaluation. Studies providing data on the TB prevalence without reporting the number of patients were excluded. The MEDLINE (accessed by Pubmed; 1966–April 2011) electronic database was searched using the following medical subject headings (MeSH): “Epilepsy”, “Seizures”, and ‘Tongue”, as well as the following 252 F. Brigo et al. / Epilepsy & Behavior 25 (2012) 251–255 free terms, combined in multiple search strategies with Boolean operators in order to find relevant articles: “tongue”, “epileps*”, “epilept*”, “seizur*”, “bit*”, “biting”, “bite” (see Appendix). Furthermore, all reference lists in identified trials were scrutinized for studies not indexed in the electronic database. Only full-length papers and articles already published were considered eligible for inclusion, in order to ensure maximum transparency of the results and to enable the readers to reproduce the methodological approaches we adopted; it should also be considered that in abstracts many methodological aspects are not declared, and results are often summarized. The methodological quality of each study was evaluated. Quality assessment of included studies focused on following criteria: 1. presence or absence of the target disorder (epileptic seizures/PNEEs) confirmed by means of a valid test (“gold” or reference diagnostic standard); 2. evaluation of the physical sign on an appropriate spectrum of patients; 3. application of both the physical finding being evaluated (TB) and the reference diagnostic standard to all patients; and 4. comparison of the physical sign independent from and blind to the study test results. Provided that we thought it to be clinically appropriate and that no important clinical and methodological heterogeneity was found, we summarized the results in a meta-analysis. Prevalence of TB (dichotomous data) as reported in included studies was analyzed by calculating odds ratio (OR) for each study, with the uncertainty in each trial being expressed using 95% confidence intervals (CIs). A weighted effect across studies was also calculated. In case there were sufficient data available, we planned to undertake subgroup analyses to assess the presence of oral lacerations involving the side of the tongue (lateral TB) and to present results on the same forest plot to give an overall impression. Homogeneity among study results was evaluated using a standard Chi‐squared test, combined with the I 2 statistics, and the hypothesis of homogeneity was rejected if the p value was less than 0.10. Prevalence was combined to obtain a summary estimate of value (and the corresponding CIs) using a random-effect model. A random-effect model is considered more conservative than a fixed-effect model, since it takes into account the variability between studies, thus leading to wider CIs. The meta-analysis was undertaken with the Review Manager software developed by the Cochrane Collaboration (5.1). Sensitivity, specificity, pLR, and nLR with 95% CIs were determined for each included study and for the summary estimate of pooled analysis using equations reported in the Appendix [3–5]. The sensitivity measures the proportion of positives that are correctly identified, whereas the specificity measures the proportion of negatives that are correctly identified. In this review, sensitivity represents the proportion of patients with epileptic seizures who have TB, while specificity refers to the proportion of patients without seizures (but with PNEE) who lack TB. The LR of a physical sign is defined as the proportion of patients with epileptic seizure who have TB divided by the proportion of subjects without seizures (but with PNEE) who also have the same finding [5]. A pLR refers to the presence of the physical sign, whereas a nLR refers to the absence of that physical sign. The interpretation of LRs is straightforward: (1) values greater than 1 increase the probability of disease (epileptic seizure), and the greater the LR, the more compelling the argument for disease; (2) values between 0 and 1 decrease the probability of disease, and the closer the LR is to zero, the more the finding argues against the diagnose of disease; and (3) values equal to zero have no diagnostic values, as they do not change pre-test probability [5]. A pLR describes, therefore, how probability changes when the finding is present; nLR describes how probability changes when the finding is absent. SPSS 16.0 was used to calculate accuracy measures. The random‐ effect model, which considers both within study and between study variance to calculate a pooled LR, was used to summarize the LRs from the included studies [6]. 3. Results The search strategy described above yielded 74 results (71 MEDLINE, 3 in reference lists). After reading the abstracts, we provisionally selected fifteen studies. After reading the full text of the retrieved articles, we included 5 studies. Thus, 5 studies, comprising 218 epilepsy patients and 228 subjects with PNEEs, contributed to this review [1,2,7–9]. 3.1. Assessment of methodological quality of included studies (Table 1) A video-EEG recording of the paroxysmal event ideally represents the reference (“gold”) standard in the differential diagnosis between epileptic seizures and PNEEs. In all included studies, a clinical evaluation was performed by epileptologists working in tertiary epilepsy centers and applied both to epilepsy patients and to PNEE patients. Although not all studies used an ictal video-EEG recording in patients with epileptic seizures, in all subjects with PNEEs, a video-EEG recording of the paroxysmal events was obtained. In all studies except from that of Brown [7] and Oliva [2], it was not specified whether the presence of TB was evaluated independently from and blind to the definite diagnosis. Three studies were prospective [1,2,7], whereas two studies obtained data retrospectively from hospital records and postal/telephone questionnaires [8,9]. Retrospective studies determined the presence of TB by patient history only, so that their results might be less accurate than those directly determining TB by means of physical examination. Based on the information provided, it is difficult to evaluate whether the spectrum of patients with epileptic seizures was sufficiently large to include both patients with and without motor phenomena, although it is possible that patients with motor phenomena were selectively/ predominantly included, as explicitly made in the study of Benbadis [1]. The choice of an appropriate spectrum of patients may have great influence on accuracy measures. For instance, when considering that TB occurs in patients with motor seizures, the adoption of less strict inclusion criteria (i.e., including also non-motor epileptic seizures) would decrease sensitivity of TB, without affecting specificity. More detailed characteristics of included studies are reported in Table 1. 3.2. Quantitative synthesis 3.2.1. TB (no further specifications on site of laceration) 3.2.1.1. Prevalence of TB (Fig. 1a). There were 5 studies with 446 participants. Significant statistical heterogeneity among trials was not detected. There was a statistically significant difference in the prevalence of TB between epilepsy and PNEE groups, with higher prevalence in the epilepsy group (82/218 vs. 58/228 participants; OR 3.07; 95% CI 1.66–5.68). 3.2.1.2. Sensitivity, specificity, pLR, and nLR of TB for the diagnosis of epileptic seizures. Sensitivity, specificity, pLE, and nLR for each included study are reported in Table 2. Pooled accuracy measures were sensitivity 38%, specificity 75%, pLR 1.479 (95% CI 1.117–1.957), and nLR 0.837 (95% CI 0.736–0.951). 3.2.2. Lateral TB Two studies (147 participants) reported enough data to allow a subgroup analysis on lateral TB. 3.2.2.1. Prevalence of lateral TB (Fig. 1b). There were 2 studies with 147 participants. Significant statistical heterogeneity among trials was not detected. There was a statistically significant difference in prevalence of TB between the epilepsy and PNEE groups, with higher 253 F. Brigo et al. / Epilepsy & Behavior 25 (2012) 251–255 Table 1 Description of included studies. Study Group Inclusion criteria Exclusion criteria Number of subjects, male/female Age Type of seizures/ PNEE Diagnostic reference used Type of study, information on TB Brown et al. [7] ES – – 25a, 9/16 (Range 18–46) 12 primary generalized tonic-clonic seizures; 12 complex partial seizures; 1 simple partial seizure. Epileptic EEG abnormalities documented by two electroencephalographers on at least two EEGs. Prospective, data obtained from interview data and video recording. Examiners were blind to subjects diagnoses. PNEE Attacks similar to those reported by history, in the absence of ictal interictal, or postictal EEG abnormalities. Bilateral motor (stiffening and/or shaking) phenomena, loss of consciousness, or both. Equivocal interictal EEG recordings. 23, 5/18 (Range 19–59) Ictal video-EEG 26 (range 3–57) 11 generalized epilepsy; 23 localization-related epilepsy. Video-EEG with evaluation of both interictal and ictal data. Benbadis ES et al. [1] PNEE Peguero et al. [8] Reuber et al. [9] Oliva et al. [2] Typical complex 34, 13/21 partial seizures, with altered awareness but no loss of consciousness. – 29, 10/19 ES – – PNEE Attacks recorded with video-EEG and considered typical by relatives who had witnessed the events. ES – PNEE Documentation of spontaneous psychogenic events with video-EEG, EEG, seizure observation and ictal examinations, clinical assessment of an experienced epileptologist, or provocation of a typical event by intravenous injection of 0.9% saline unde video-EEG. Multiple admissions to hospital. Occurrence of at least one convulsive event, defined clinically as one that involved simultaneous shaking of the body including all limbs. 73, 17/56 Events characterized only by a subjective experience, a subtle motor activity, or behavioral change in infants or children; epileptic seizures of mesio-frontal origin. Evidence of 64, 40/24 concurrent PNEE. Evidence of 85, 15/70 concurrent epilepsy; epileptiform potentials in interictal EEGs. ES PNEE 30, 10/20 32 (range 1–57) 27 psychogenic seizures; 2 preverbal children with behavioral posturing. Ictal video-EEG 29 (range 7–56) 27 partial epilepsy (complex partial with/without generalization); 3 generalized epilepsy (2 myoclonic and tonic-clonic seizures, 1 tonic and atypical absence seizures) 32 (range 9–52) 38.8 (SD 10.1) 37.1 (SD 15.8) – 66, 35/31 37.4 (SD 1.7) – 18, 7/11 40.4 (SD2.7) Prospective, direct documentation of oral lesions. Not reported whether TB assessment was made independently and blinded to the diagnosis. Retrospective, data obtained using a telephone interview. Not reported whether TB assessment was made independently and blinded to the diagnosis. Ictal EEG or video-EEG; clinical assessment of an 33 history of seizures lasting over experienced epileptologist. 30 min leading to more than one hospital admission; 52 subjects without history of seizures lasting over 30 min. Retrospective, data extracted from hospital records and a postal questionnaire. Not reported whether TB assessment was made independently and blinded to the diagnosis. Ictal video-EEG, clinical and investigational findings. Prospective, direct documentation of oral lesions. Information regarding TB was gathered independently and blinded to the diagnosis. – 36 temporal lobe epilepsy; 15 extratemporal lobe epilepsy; 15 primary generalized epilepsy. – ES: epileptic seizures; PNEE: psychogenic non-epileptic events; SD: standard deviations; TB: tongue biting; –: not reported. a Data from one patient missing. prevalence in the epilepsy group (22/100 vs. 0/47 participants; OR 13.86; 95% CI 1.80–106.53). Pooled accuracy measures were sensitivity 22%, specificity 100%, pLR 21.386 (95% CI 1.325–345.169), and nLR 0.785 (95% CI 0.705–0.875). 4. Discussion 3.2.2.2. Sensitivity, specificity, pLR, and nLR of lateral TB for the diagnosis of epileptic seizures. Sensitivity, specificity, pLE, and nLR for each included study are reported in Table 2. The diagnosis of epileptic seizures is primarily clinical and relies on the patient's history and an accurate witness description of the attacks 254 F. Brigo et al. / Epilepsy & Behavior 25 (2012) 251–255 Fig. 1. Prevalence of TB. a. no further specifications on site of laceration; b. lateral TB. in the event of a loss of awareness, consciousness, or recall of the events. Sometimes, the diagnosis of seizures can be supported by clinical findings, such as TB. However, TB may occur also in patients with PNEE, so that the diagnostic utility of this finding should be evaluated in the clinical context. A modern and evidence-based approach to the clinical diagnosis of seizures should take into account the concept of refining probability, an evidence-based technique which refines probability of an epileptic event, thus modifying the estimate of the likelihood of a disease (seizure) through the application of a diagnostic test (EEG) or the evaluation of a physical finding (such as TB) [10,11]. Refining probability represents a clinical way of stating the Bayes' theorem: the probability of an event depends on new information applied to what is previously known about that event. Such a concept may be simplified to the following equation: what we thought before Table 2 Accuracy measurements for each study and for pooled results. Study Sensitivity (95% CIs) Specificity (95% CIs) pLR (95% CIs) 1. Tongue biting (no further specifications on site of laceration) Brown et al. [7] 21% 100% 10.56 (0.617–180.808) Benbadis et al. [1] 24% 100% 14.571 (0.877–242.071) 60% 56% 1.369 Peguero et al. [8] (0.926–2.023) 56% 69% 1.839 Reuber et al. [9] (1.25–2.706) Oliva et al. [2] 23% 100% 8.791 (0.551–140.251) Pooled results 38% 75% 1.479 (1.117–1.957) 2. Lateral tongue biting Benbadis et al. [1] 24% 100% Oliva et al. [2] 21% 100% Pooled results 22% 100% 14.571 (0.877–242.071) 8.224 (0.514–131.608) 21.386 (1.325–345.169) nLR (95% CIs) 0.797 (0.642–0.989) 0.77 (0.635–0.934) 0.712 (0.439–1.154) 0.63 (0.462–0.861) 0.789 (0.679–0.918) 0.837 (0.736–0.951) 0.77 (0.635–0.934) 0.805 0.695–0.931 0.785 (0.705–0.875) When calculating LR, if any cell of the 2 × 2 table contained the value of zero, 0.5 was added to all cells, to avoid creating the unlikely LRs of 0 or infinity (McGee [3]). (pre-test probability) + test information (likelihood ratios) = what we think after (post-test probability) [13]. In other terms, we start with a certain pre-test probability, and after the application of a diagnostic test (EEG) or evaluation of a certain clinical sign, we finish with a post-test probability of disease [12]). Estimates of the likelihood of a disease range in probability scale from 0% (disease ruled out) to 100% (disease ruled in) [12]. Pre-test probability is the probability of disease before application of the results of a physical finding (i.e., prevalence). Likelihood ratios (i.e., information given by a test or by evaluation of a clinical sign) assess the discriminatory power of a test or of a physical finding, calculated from the sensitivity and specificity of the test/sign to determine whether or not and how much test results or a certain physical sign change the likelihood of a condition. Therefore, LR represents a very useful and accurate measure to interpret test results for an individual patient. In this systematic review, we used systematic and explicit methods to identify, select, and critically appraise studies and to extract data, analyzing them with a meta-analysis. A meta-analysis is the statistical combination of results from two or more separate studies (pair-wise comparisons of interventions), allowing an increase in statistical power, an improvement in precision, and, sometimes, to answer questions not posed by individual studies as well as to settle controversies arising from conflicting claims. Meta-analysis showed a statistically significant difference in TB prevalence (both without further specifications on site of lesions and lateral TB) between the epilepsy and PNEE groups, with higher prevalence in the epilepsy group (more marked in lateral TB prevalence). Pooled accuracy measures for TB (both without further specifications on site of lesions and lateral TB) showed a statistically significant pLR, but only that of lateral TB was clinically relevant. In fact, if the probability of epileptic seizures is estimated by means of a nomogram describing how pre-test probability relates to post-test probability given the LR for such a physical finding [13], the possibility that the patient may have an epileptic seizure instead of a PNEE appears to be only slightly increased by the presence of TB when no further specifications on site of laceration are provided. Although greater than 1 (1.479), the pLR is still too close to 1; therefore, the diagnostic value of this finding is limited, as it does not significantly change pre-test probability. Conversely, when in doubt between the diagnosis of seizures and PNEE, the high value of pLR for lateral TB is clinically relevant, F. Brigo et al. / Epilepsy & Behavior 25 (2012) 251–255 255 seizures and PNEE, showing a statistically significant pLR (95% CI 1.325–345.169). 5. Conclusions In conclusion, a pooled analysis of data from the literature shows that TB without further specifications on site of oral lacerations has no value in the differential diagnosis between epileptic seizures and PNEEs. Conversely, due to its high specificity and pLR, the presence of lateral TB suggests the diagnosis of epileptic seizures. Systematic reviews with pooled analyses (meta-analyses) of data from the literature allow an increased statistical power and an improved precision and represent a useful tool for determining the accuracy of physical examination findings in the differential diagnosis between seizures and other paroxysmal events. Despite the useful information provided by an evidence-based approach to the evaluation of a physical sign, the diagnosis of epileptic seizure, syncope, or paroxysmal non-epileptic event requires careful integration of history, ictal signs, and other clinical and investigational information and should never be driven by any single clinical sign alone. Appendix A. Supplementary data Supplementary data to this article can be found online at http:// dx.doi.org/10.1016/j.yebeh.2012.06.020. References Fig. 2. The probability of epileptic seizures is estimated by means of a nomogram describing how pre-test probability relates to post-test probability given the LR for TB. When in doubt between the diagnosis of seizure and that of PNEE and given the same pre-test probability of seizures (i.e. prevalence) of 50%, the presence of TB (without further specifications on site of oral lacerations) does not increase the chance that the patient had an epileptic seizure (dashed line) (pLR=close to 1). Conversely, given the same pre-test probability, the presence of lateral TB greatly increases the chance that the patient had an epileptic seizure (continuous line) instead of a PNEE (pLR=21.386). as it can greatly change pre-test probability, strongly supporting the diagnosis of epileptic seizure (Fig. 2). Both studies evaluating the diagnostic accuracy of lateral TB showed a high specificity to epileptic seizures [1,2]. The validity of this physical sign was, however, limited by a pLR that is statistically not significant (the 95% CI of the pooled value included 1, and the LR value of 1 has no discriminatory value), due to a relatively small sample size of each study. Only a pooled analysis of data from these two studies definitively proved the value of lateral TB in the differential diagnosis between [1] Benbadis S, Wolgamuth BR, Goren H, Brener S, Fouad-Tarazi F. Value of tongue biting in the diagnosis of seizures. Arch Intern Med 1995;155:2346–9. [2] Oliva M, Pattison C, Carino J, Roten A, Matkovic Z, O'Brien TJ. The diagnostic value of oral lacerations and incontinence during convulsive “seizures”. Epilepsia 2008;49:962–7. [3] Simel DL, Samsa GP, Matchar DB. Likelihood ratios with confidence: sample size estimation for diagnostic test studies. J Clin Epidemiol 1991;44:763–70. [4] Newcombe RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998;17:857–72. [5] McGee SR. Diagnostic accuracy of physical findings. In: McGee SR, editor. Evidence-based physical diagnosis. Saunders Elsevier; 2007. p. 3–19. [6] DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7: 177–88. [7] Brown MC, Levin BE, Ramsay RE, Debra AK, Duchowny MS. Characteristics of patients with non-epileptic seizures. J Epilepsy 1991;4:225–9. [8] Peguero E, Abou-Khalil B, Fakhoury T, Mathews G. Self-injury and incontinence in psychogenic seizures. Epilepsia 1995;36:586–951. [9] Reuber M, Pukrop R, Mitchell AJ, Bauer J, Elger CE. Clinical significance of recurrent psychogenic nonepileptic seizure status. J Neurol 2003;250:1355–62. [10] Brigo F. Refining probability: a simple concept to be considered when interpreting interictal epileptiform discharges in EEG. J Clin Neurophysiol 2011;28:334. [11] Brigo F. An evidence-based approach to proper diagnostic use of the electroencephalogram for suspected seizures. Epilepsy Behav 2011;21:219–22. [12] Go AS. Refining probability: an introduction to the use of diagnostic tests. In: Friedland DJ, editor. Evidence-based medicine: a framework for clinical practice. Stamford, CT: Appleton & Lange; 1998. p. 11–34. [13] Fagan TJ. Nomogram for Bayes theorem. N Engl J Med 1975;293:257.