Use of serum prolactin in diagnosing epileptic seizures Report of the Therapeutics and Technology Assessment Subcommittee of the American Academy of Neurology

Objective: The purpose of this article is to review the use of serum prolactin assay in epileptic seizure diagnosis. Methods: The authors identified relevant studies in multiple databases and reference lists. Studies that met inclusion criteria were summarized and rated for quality of evidence, and the results were analyzed and pooled where appropriate. Results: Most studies used a serum prolactin of at least twice baseline value as abnormal. For the differentiation of epileptic seizures from psychogenic nonepileptic seizures, one Class I and seven Class II studies showed that elevated serum prolactin was highly predictive of either generalized tonic–clonic or complex partial seizures. Pooled sensitivity was higher for generalized tonic–clonic seizures (60.0%) than for complex partial seizures (46.1%), while the pooled specificity was similar for both (approximately 96%). Data were insufficient to establish validity for simple partial seizures. Two Class II studies were consistent in showing prolactin elevation after tilt-test–induced syncope. Inconclusive data exist regarding the value of serum prolactin following status epilepticus, repetitive seizures, and neonatal seizures. Recommendations: Elevated serum prolactin assay, when measured in the appropriate clinical setting at 10 to 20 minutes after a suspected event, is a useful adjunct for the differentiation of generalized tonic–clonic or complex partial seizure from psychogenic nonepileptic seizure among adults and older children (Level B). Serum prolactin assay does not distinguish epileptic seizures from syncope (Level B). The use of serum PRL assay has not been established in the evaluation of status epilepticus, repetitive seizures, and neonatal seizures (Level U). NEUROLOGY 2005;65:668–675 Prolactin (PRL) release from the pituitary is controlled by the hypothalamus via a PRL inhibitory factor, now believed to be dopamine.1 It has been hypothesized that ictal epileptic activity in the mesial temporal structures may propagate to the hypothalamus, altering the hypothalamic regulation of PRL release.2 Trimble first demonstrated that generalized tonic– clonic seizures, but not nonepileptic seizures (NESs), could raise serum PRL.3 Despite subsequent confirmatory findings, the sensitivity and specificity of serum PRL assay for diagnosis of epileptic seizures (ESs) remain uncertain. Utility of PRL assays for diagnosis of seizures depends upon the study design, standard of seizure classification, and criteria for abnormal PRL elevation. Additional uncertainty arises from the circadian fluctuations of serum PRL, demonstrating surges of 50 to 100% prior to awakening from sleep, although PRL serum levels otherwise are stable during the waking state.4 PRL concentrations usually are higher in females than in males,5 and higher in persons with epilepsy than in healthy individuals.6 In a study that measured baseline PRL every 20 minutes over 24 hours in 20 healthy controls and 17 people with epilepsy,5 no female had baseline PRL exceeding 700 U/mL, and no male exceeded 450 U/mL. However, the study subjects endured disrupted sleep, and thus the effect from nocturnal PRL rises could not be fully established. Psychogenic or physiologic nonepileptic events also can influence serum PRL level. Several studies7-10 have suggested that serum PRL can increase after syncope, a common imitator of epilepsy. Recent interest in the use of PRL has diminished From Stanford University, Palo Alto, CA. Robert S. Fisher, MD, is supported by the Maslah Saul MD Chair, the Susan E. Horngren Fund, and the James and Carrie Anderson Epilepsy Research Fund. Disclosure: The authors report no conflicts of interest. This guideline was approved by the Therapeutics and Technology Assessment Subcommittee on November 19, 2004; by the Practice Committe on April 13, 2005; and by the Board of Directors on June 25, 2005. Received January 6, 2005. Accepted in final form July 5, 2005. Address correspondence and reprint requests to the American Academy of Neurology, 1080 Montreal Avenue, St. Paul, MN 55116. 668 Copyright © 2005 by AAN Enterprises, Inc. with the availability of video-EEG monitoring. However, a serum marker continues to be of potential clinical utility, especially when video-EEG is not readily available. Based on the evidence classification scheme of the American Academy of Neurology (AAN), we propose recommendations for the use of serum PRL assay to differentiate ES from psychogenic NES. We also reviewed the usefulness of serum PRL in other settings, such as following syncope and repetitive seizures and in the neonatal population. Methods. We searched MEDLINE, Science Citation Index, and the Cochrane Database, combining the search term prolactin with the terms seizure(s), pseudoseizure(s), epilepsy, syncope, or status epilepticus. Three hundred ninety-six articles were identified as of March 2005. We reviewed the abstracts of these articles, specifically looking for controlled studies that reported on PRL changes following seizures or seizurelike events. Reviews without original data, letters, meeting abstracts, and case reports/series were excluded. We examined 41 articles in their entirety, along with 5 additional articles identified upon reviewing bibliographies of the retrieved articles. Three articles in German were translated into English. We categorized the articles into two groups: Group 1 consisted of controlled studies investigating the use of postevent PRL to discriminate ES from psychogenic NES. Group 2 consisted of controlled studies assessing serum PRL changes following syncope, repetitive seizures, or neonatal seizures. For Group 1, we selected studies for inclusion into our analysis based on the following criteria: 1) prospective design, 2) implementation of reference standard in the form of continuous EEG monitoring, 3) specification of the threshold for PRL elevation and the postevent lapse time of PRL measure, 4) reporting of the accuracy rates of PRL assay among case and control groups, and 5) publication in a peer-reviewed journal. For Group 2, all published studies that prospectively investigated serum PRL changes following tiltinduced and monitored syncope, repetitive seizures, status epilepticus (SE), or neonatal seizures were included. Wherever a study reported more than one criterion for elevated PRL, we analyzed the data arising from criterion closest to the common criteria chosen by other studies of the same group (i.e., twice baseline level, or 36 ng/mL for Group 1). PRL measures presented in U/mL or g/L were converted for consistency of presentation to ng/mL. We graded each study using the four-tiered classificationof-evidence scheme in Appendix 1. Most laboratories report PRL upper normal limits of 18 to 23 ng/mL.11,18 However, prior literature does not specify a precise and commonly accepted cutoff PRL level as an indicator of epilepsy.12 We accepted the individual investigators’ opinion of abnormal PRL elevation. From the proportion of elevated PRL for each seizure type reported, we calculated sensitivity and specificity, where appropriate. Ninety-five percent CIs for each parameter were calculated using the Wilson score method without continuity correction.13 From all Class I and Class II studies, the sensitivity values were then pooled by calculating the weighted average. The same process was performed for the specificity values. Applying Bayes’ theorem, the positive or negative predictive values of serum PRL assay would depend not only on the sensitivity and specificity parameters, but also on the pretest probability that an event is epileptic.11 We calculated the predictive values for a range of ES pretest probabilities from 99% to 50%, assuming the respective pooled sensitivity and specificity for generalized tonic–clonic seizures (GTCs), complex partial seizures (CPSs), and all ESs combined. We set the requirement that both patient sample size and number of seizures studied must be 50 or greater for a study to be considered “wide spectrum” for the purpose of evidence classification. The varieties of seizure types studied were also weighed in assessing extent of patient spectrum. Analysis of evidence. Question 1: Is serum PRL assay useful in differentiating ES from psychogenic NES? One Class I and nine Class II studies compared serum PRL changes following ES and psychogenic NES (table 1).14-23 Of these 10 studies, all except one study15 ascribed psychogenic etiologies to the NES. The terminology of NES was used in one study that did not elaborate on etiology.15 In one Class I and seven Class II studies, elevated serum PRL measures were positively predictive of a correct diagnosis of GTC or CPS, whereas failure of PRL elevation poorly distinguished between ES and psychogenic NES. Data were insufficient to establish the predictive value of PRL following simple partial seizures. Two Class II studies14,16 showed a small but significant PRL elevation following psychogenic NES and reported the lowest specificity (74% and 66.7%) among this group. The inconsistency may be due to the risk of systematic error inherent from incompleteness of monitoring NES in one study16 and less stringent criterion for abnormal PRL in another14 (see table 1). To increase statistical power, we pooled the available data of Class I and Class II studies in the weighted average analysis. For the diagnosis of GTC or CPS, the pooled specificity ranged from 95.9 to 96.3%, whereas the pooled sensitivities were limited to 46.1 to 60.0% (table 2). Applying Bayes’ theorem, table 3 illustrates how the PRL predictive value depends on the pretest probability of ES in the population studied.11 Assuming an ES pretest probability of 95% in the general epilepsy population, a positive PRL measure is highly predictive ( 99%) of either GTC or CPS. Even with an ES pretest probability as low as 50%, a positive PRL measure is still highly predictive of either GTC or CPS (positive predictive value of about 93%). By contrast, because of the relatively low sensitivity of the test, a negative PRL measure is not predictive of psychogenic NES and hence does not rule out a diagnosis of an ES. Two Class II studies14,19 examined the time course of PRL attenuation following ES. In one study,19 6 patients with ES had attenuation of mean PRL concentration to 17.5 3.6% of mean peak postictal PRL by 2 hours after seizures. The 2-hour postictal PRL levels were similar to the baseline PRL measured at the same time on the subsequent seizurefree day. In another study,14 32 patients with ES had postictal PRL elevation up to 6 hours (p 0.048) when compared with the baseline PRL level. Conclusions. On the basis of one Class I and seven Class II studies, an elevated PRL level when measured 10 to 20 minutes after a suspected event is probably a useful adjunct to differentiate GTC or CPS from psychogenic NES among adults and older children. On the basis of consistent Class I and II studies, a normal serum PRL assay by itself is insufficient to make a diagnosis of psychogenic NES or to exclude the possibility of GTC or CPS because of its low sensitivity and low negative predictive value. On the basis of two Class II studies, serum PRL, when measured more than 6 hours after an ES, is probably representative of the baseline PRL level. September (1 of 2) 2005 NEUROLOGY 65 669 Question 2: Does serum PRL measure change following other neurologic conditions? Syncope. Two Class II studies investigated serum PRL changes during 60° head-up tilt-table test in subjects at risk of syncope (table 4). In one study of subjects with a mean age of 70 years,7 11 syncopal subjects showed an elevated mean PRL level of 44 ng/mL (95% CI 27 to 61) when tested within 5 minutes of syncope, compared with a mean baseline PRL level of 10 ng/mL (95% CI 7 to 14). Ten nonsyncopal subjects showed no significant PRL change following head-up tilt. In another study of younger subjects (mean age approximately 30 to 40 years),8 14 syncopal subjects showed a mean PRL level of 18.1 ng/mL at 5 to 10 minutes after syncope. This was more than twice their mean baseline PRL level of 7.7 ng/mL. This relative increase was significant compared with the relatively unchanged baseline and postevent PRL mean measures of 22 nonsyncopal subjects (p 0.004). A mean PRL level of 18.1 ng/mL is high, but not clearly abnormal. The reasons for the difference of effect size are unclear and may reflect differences in age and PRL assay of the two studies. Conclusion. On the basis of limited Class II studies, serum PRL possibly increases from baseline level when measured within 10 minutes after syncope in adults. An elevated PRL level cannot be used to differentiate between seizure and syncope. Repetitive seizures. One Class II study24 measured serum PRL level within 20 minutes following termination of monitored SE among 15 patients (table 5). All post-SE PRL measures were within normal range and were in fact lower than baseline measures in most of the 15 patients. Two Class II studies prospectively investigated serum PRL measures following repetitive, discrete seizures (not SE), using video-EEG as a reference standard25,26 (see table 5). Postictal PRL rise was reduced when seizures occurred after short seizure-free intervals of less than 25 hours, compared with those occurring after longer seizure-free intervals.25 In contrast, another study26 showed that following repetitive seizures (mean 3 hours and 32 minutes apart), postictal PRL measures were markedly and consistently increased in 5 of 14 patients studied, regardless of the time interval between seizures. None of the 14 patients showed a decrease in postictal PRL measure. The reasons for the inconsistency of the data are unclear, although sample sizes were small in both studies, reflecting low statistical power. Conclusions. On the basis of inconsistent studies, no conclusion can be established regarding seTable 1 Prospective controlled studies investigating PRL changes following either ES or psychogenic NES Reference Patients, n Age range, y Reference standard Lapse time of PRL measure, min Criterion for elevated PRL Class Willert et al.14 44 18–62 Video-EEG 20* 16.5 ng/mL (male)† 23 ng/mL (female) II Shah et al.15 89 Not provided Video-EEG Average 15–20 2 baseline level II‡ Alving16 58 13–68 Video and/or ambulatory cassette EEG 15 2 baseline level at 2 h post event II§ Ehsan et al.17 50 6–61 Video-EEG 15 2 baseline level at 1 h post event II¶ Fisher et al.18 20 18 Video-EEG 10–20 36 ng/mL II Rao et al.19 12 13–47 Video-EEG Immediately, then every 15 min 2 h At least 2 baseline level II Wroe et al.20 33 15–73 Video-EEG 10 45 ng/mL II Laxer et al.21 70 9–54 Video-EEG 20 Discriminant function I Pritchard et al.22 12 Not provided Continuous EEG ( video) 15 2 baseline level II Oxley et al.23 18 Not provided Continuous EEG (Medilog 4-cassette recorder) 20 36 ng/mL II * Serum prolactin (PRL) levels were measured at 10, 20, 30, and 60 minutes and at 6, 12, and 24 hours post event. For consistency of presentation, data collected at 20 minutes post event were analyzed in this review. † Lowest cutoff criterion for abnormal PRL among the studies in this group. ‡ Data collected prospectively but analyzed retrospectively. Despite wide spectrum of patients, we downgraded the study to Class II given risk of expectation bias. § EEG corroborated diagnosis reported for 17 of 20 nonepileptic seizure (NES) patients (85%). Twelve NES patients had multiple events studied, but the proportion of monitored patients was not provided. One NES patient was excluded from validity analysis because of absence of baseline PRL level. We downgraded the study to Class II because of the unclear number of unmonitored NES. ¶ Diagnostic accuracy of capillary blood PRL levels measured by a modified immunoradiometric assay.17 Serum and capillary blood PRL values correlated with a Pearson coefficient of 0.92. We downgraded the study to Class II given additional risk of systemic error (instrument bias) and restricted availability of the immunoradiometric assay. Not explicitly specified by author, and based upon interpretation of figures in the article. ES epileptic seizure; Lapse time of PRL measure time elapsed from start of event until blood draw for PRL assay. 670 NEUROLOGY 65 September (1 of 2) 2005 rum PRL changes following termination of SE. On the basis of conflicting Class II studies, no conclusion can be established regarding serum PRL changes following repetitive seizures (not SE). Neonatal population. Two Class II studies prospectively assessed serum PRL changes following EEG-corroborated seizures among neonates27,28 (table 6). In one study,27 neonates with electrographic Table 2 Validity measures of serum PRL assay in differentiating ES from NES Reference No. 1PRL/no. seizures studied by type Sensitivity, % (95% CI) Specificity, % (95% CI) Willert et al.14 All ES—28/32 87.5 (71.9–95.0) 66.7 (39.1–86.2)