The temporal distribution of seizures is often as clinically relevant as overall seizure frequency. A patient who experiences one day of seizures per year may be considered fairly well controlled, but if the seizures occur in a flurry or cluster, this episode may result in a yearly visit to an emergency room or even hospital admission.

Epilepsy can be characterized as one of the chronic neurologic disorders with episodic manifestations (CDEM). As with all disorders with episodic manifestations, an attempt to identify patterns and triggers of occurrence opens the possibility of preventive therapy. Seizure clusters are also known as repetitive or serial seizures [1], with return to baseline between events. At the core of the issue of seizure clustering is the question of whether seizure episodes are random, or whether patterns can be predicted, and possibly prevented. Thus, investigation of the clustering phenomenon yields insights into both specific mechanisms of clustering and more general concepts of seizure occurrence.

The timing of seizures: Historical perspective

One of the unfortunate hallmarks of epilepsy is the unpredictability of seizures. The investigation of whether seizures are truly random events or events that follow identifiable patterns is not new. Attempts to comprehensively identify factors influencing seizure occurrence date back more than a century. Menstrual cycle and time of day were features noted by Gowers [2]; others followed by characterizing seizure patterns in large groups of patients with epilepsy, with particular emphasis on external features such as time of day, weather, and season, as well as internal factors such as emotional state and gastrointestinal distress [3] and [4]. While some of these early associations are no longer reported (i.e., constipation), many (i.e., early-morning clustering of seizures) remain widely accepted today.

The advent of continuous EEG monitoring in the 1960s enhanced the evaluation of the temporal occurrence of seizures. Spike and seizure patterns in subjects undergoing continuous monitoring were analyzed for periodicity and pattern ([5] and [6]). The concept of seizures as nonrandom events based on statistical analysis began to emerge; the observation that the occurrence of a seizure appeared to have a relationship with prior and subsequent interictal periods was reported, as was the previously noted influence of sleep–wake cycles [6].

Despite many advances in the diagnosis, classification, and treatment of epilepsy in the ensuing decades, little progress has been made toward addressing perhaps the most essential question in epilepsy: Why does a seizure occur when it does? This investigation is rooted in basic mechanisms of epileptogenicity and seizure termination, combined with accurate clinical data regarding seizure occurrence and patterns.

An ongoing difficulty with conducting clinical studies of seizure patterns is the limitation inherent in the methods of data collection. Information obtained from patient reports in diaries may at times be inaccurate, particularly under reported, as patients may not be aware of all seizures or may be amnestic of their occurrence. Data from continuous EEG monitoring are more accurate, but such monitoring is not performed for prolonged periods, and is often performed in the setting of abrupt medication discontinuation, which may obscure natural patterns. However, many reliable findings have emerged, as discussed below.

Mechanisms of seizure clustering

Of the CDEM, epilepsy is unique in that seizures are typically self-limited. Most seizures do not require acute intervention to terminate, and prolonged seizures, requiring abortive therapies, tend to be the exception. Mechanisms of seizure termination are poorly understood, related to impaired neuronal receptor and ion channel function, and possibly hypoxia-related mitochondrial dysfunction [7]. While a discussion of neuronal excitability is beyond the scope of this review, the concept of clustering implies either impaired termination or increased excitation, possibly due to secondary alterations from an initial seizure that promote a second attack, or an excess of seizure-promoting factors [8].

Seizures occurring within 8 hours of a prior seizure have been reported to be significantly more likely to arise from a concordant focus than seizures more widely separated in time [9]. These data directly suggested that at least in some settings, an ictal focus may be more excitable, or less inhibited, following a first seizure, leading to a seizure cluster. This demonstration supported a commonly used definition of a seizure cluster as three seizures per 24 hours, derived from the fact that seizures within 8 hours are not truly "independent." It was similarly noted that seizures from a secondary focus tended to occur consecutively rather than randomly [10], although a subsequent study did not confirm this finding [11].

Other studies have examined the concept of dependence between seizures, to evaluate whether a seizure may influence the possibility of a second seizure. As extensively explored by Hopkins et al. [12], a Markov model, in which the transitional probability of a subject being in a particular seizure-susceptibility state depends on the state of the previous day, is appropriate for this evaluation. With this method, probability estimates for the expected incidence of subsequent seizure days can be calculated. Although prior studies did not report a correlation between successive interseizure intervals [13], Tauboll et al. [8] computed transitional probabilities between seizures and no seizures and demonstrated this dependence in half of their subjects, potentially associated with the presence of seizure clustering. Negative dependence, or a lower probability of seizure events following a day of high seizure activity, has also been noted [14].

Defining seizure clustering

It is difficult to assign a specific definition to seizure clustering. At the simplest level, a seizure cluster is a closely grouped series of seizures. This approach defines clustering clinically, a strategy that is relatively easy to employ in clinical settings. Another approach is to consider clustering to be an increase over the patient's typical seizure frequency. A more interesting question addresses the underlying temporal distribution of seizures: Is seizure occurrence random, or may patterns, periodicity, and deviations from randomness be identified? This approach identifies a cluster as a deviation from randomness, typically addressed statistically.

Clinical definitions of clustering

There is no definitive clinical definition for a cluster or series of seizures. Studies examining clinically defined seizure clustering patterns have used varying empiric definitions, including two to four seizures per <48 hours [15]; 3 seizures per 24 hours ([11] and [16]); or two generalized tonic–clonic or three complex partial seizures in 4 hours [17]. Nonspecific definitions, such as "those having several convulsions within a day or two," have also been described [4]. In a large randomized controlled trial of treatment for acute repetitive seizures, the condition was defined as "multiple seizures occurring with a 24 period for adults or 12 hour period for children, with a pattern distinguishable from the usual seizure pattern" [1].

The strength of applying a clinical definition to identify seizure clustering is that it is easy to administer, with the information available from patient report or examination of diary or inpatient data. Patients can be instructed to institute treatment based on specific criteria they can easily identify, and studies of seizure clustering can stratify subjects into clusterers or nonclusterers based on diary data. Limitations include the possibility that patients with frequent seizures may meet these definitions by chance alone; alternately, for patients with infrequent seizures, two seizures may represent a cluster and be missed.

A strategy to address these potential weaknesses is to relate clustering to the individual's unique seizure pattern, typically by defining clustering as a measurable increase over the patient's typical seizure frequency. Investigators have considered a threefold [18] or fourfold [19] increase over usual seizure frequency within a 3-day period to represent seizure clustering. This approach requires sufficient follow-up so that typical seizure frequency or interseizure interval can be accurately determined, and often involves a sophisticated analysis technique.

Statistical definitions of clustering

An even more complex analytical approach to seizure clustering requires a foray into the concept of seizures as random versus nonrandom events. If seizures are random, then the occurrence of one seizure does not increase or decrease the likelihood of a subsequent one. However, there is ample evidence that many seizures are not "random." As discussed below, cyclical patterns do exist in epilepsy, including circadian and catamenial patterns; furthermore, factors such as medication noncompliance and sleep deprivation are clear precipitants.

Numerous studies have applied tests for Poisson distribution to seizure patterns, to evaluate randomness of seizures ([8], [13], [14], [20-22]). The Poisson process describes a stochastic (random) system in which the numbers of events occurring in disjointed (nonoverlapping) time intervals are independent random variables, and the number of events within each time variable occurs as a random variable with a Poisson distribution. In this memoryless system, the probability of an event is not influenced by the times of occurrences of any past event. This model is expressed in an equation that may be used to calculate the expected seizure frequency per day if seizures occur at random with an average frequency of λ:

where, Pr = probability of n seizures occurring on any day, λ is the average frequency of seizures per day, n! = n factorial, and e = base of natural logarithms.

In the evaluation of seizure patterns, deviations from a Poisson process may reflect clustering, may reflect periodic patterns, or may reflect regularity. Different strategies have been employed to identify true clustering, for example, the definition of clustering as "time periods during which seizure occurrence is significantly increased compared to the rate expected from the individual mean" [20]. In some studies, dependence is considered, where clustering is defined when the "expected seizure rate on a given day depends on the number of seizures during the prior day" [14].

Seizures as periodic events

Numerous data demonstrate the nonrandom periodicity of seizures, on both an individual and a population basis ([4], [8], [13], [14], [20] and [23]), although some diary analyses report that a random seizure model best fits the majority of the data ([13] and [21]).

Two specific examples of factors related to seizure periodicity are time of day and menstrual phase. Ultradian patterns for epileptiform discharges and seizures have been clearly demonstrated, highest in sleep ([5],[6],[13] and [24]), and particularly in patients with primary generalized epilepsy. A peak in seizure occurrence at midnight has been reported [4]. Catamenial patterns, while absent or less robust a finding in some studies ([23] and [25]), appear to be present in certain women with epilepsy ([26-29]). Menstruation-related seizures are characterized by a cyclical occurrence of seizures in a perimenstrual, periovulatory, or luteal pattern, and it has been reported that up one-third of women exhibit a twofold increase in seizures during a particular phase of the cycle [28].

Many studies have demonstrated periodicity unrelated to these factors, in proportions ranging from 30% [13] and [14] to 49% [20]. Periodicities described include circadian [3], [4] and [28], circaseptin (6–8 days) [20], circavigintan (21 days) [28], and "near monthly" [14]. Other patterns, such as lunar or seasonal cycles, have not been demonstrated [29].

Prevalence of seizure clustering

The prevalence of clustering varies widely between studies and definitions. Many studies are biased toward patients with intractable epilepsy, who are more likely to maintain prospective long-term diaries (Table 1). This may even be an effect within studies; in one diary study [18], patients who were compliant with diary maintenance had more frequent baseline seizures than the population as a whole.

Table 1.

Prevalence of seizure clustering

Study Prevalence (%) Population
Outpatient studies
Bauer et al. [20] 57 Intractable focal epilepsy
Balish et al. [14] 76a Intractable partial epilepsy
Haut et al. [22] 22 Range of seizure control
Milton et al. [21] 13 Outpatient neurology practice
Newmark and Dubinsky [18] 7b Intractable epilepsy
  47c  
Tauboll et al. [8] 50 Mixed seizure types
Inpatient studies
Haut et al. [11] 61 Undergoing presurgical evaluation
Rose et al. [17] 48d Admitted for epilepsy monitoring
  18e  
Yen et al. [16] 48 Undergoing presurgical evaluation

a Includes negative dependence.
b Repetitive clusters.
c Probable clusters.
d Cluster in 24-hour period.
e Cluster in 4-hour period.

Whether clinically or statistically defined, the prevalence of seizure clustering ranges from very low up to 60%. This range likely reflects varying patient populations and sample size and, most significantly, a lack of uniformity in approaching the definition. Prevalence in the inpatient setting tends to be 50% or higher, possibly reflecting a medication withdrawal effect.

A recent prospective diary study applied both clinical (3 seizures/24 hours) and statistical (deviation from Poisson distribution, with variance estimate greater than the mean) methodology to the analysis of seizure clustering [22] and [30]. Forty-three percent of subjects met either the clinical or statistical definition; all met the clinical definition, while 22% also met the statistical definition. All subjects meeting the statistical definition had experienced at least one episode of three seizures per 24 hours; conversely, the clinical definition falsely identified some patients with high seizure frequencies as clusterers. This study demonstrated that the phenomenon of clustering is common, and that a statistical definition appears more specific than a clinical one.

Risk factors for experiencing seizure clusters

Various studies have addressed risk factors for being a "clusterer." A commonly considered risk factor is epilepsy localization, with extratemporal epilepsy [31], particularly frontal lobe epilepsy ([32] and [33]), demonstrated to be associated with seizure clustering, although this is not present in all studies [34]. In descriptive reports of frontal lobe epilepsy in adults [32] and children [33], seizures were described as "tending to cluster" in 50% of patients. Etiology may play a role; head trauma has been reported to be significantly associated with patient-reported history of seizure clustering, independent of epilepsy localization [31].

Another risk factor for clustering appears to be seizure control. Patients with more intractable epilepsy appear to be at higher risk of experiencing seizure clustering [14] and [31]. However, in this group, the occurrence of seizure clusters defined clinically may be a random phenomenon of high seizure frequency. Milton et al. [21] demonstrated a significant departure from a Poisson process in subjects with more seizures and higher seizure frequencies; they similarly commented that had more seizures been recorded, all diaries might eventually show departures from a Poisson process.

Gender has not been demonstrated to be associated with seizure clustering ([8] and [20]), despite the demonstration of hormonal influence on seizure occurrence. While age does not appear to be an independent risk factor for clustering, longer duration of epilepsy does increase clustering risk ([20] and [35]).

Are there patients who are "true clusterers"?

Although in clinical practice it often appears that certain patients tend to be true clusterers, long-term diary studies reveal very few patients who experience all or most of their seizures in clusters ([18] and [22]). It is likely that precipitants associated with episodes of clustering may be more relevant than risk factors for being a "clusterer."

Precipitants for episodes of clustering

Most studies have not identified specific precipitants for seizure clustering. As noted, high prevalence rates of clustering in the inpatient setting imply a medication withdrawal effect. A pediatric condition known as benign convulsions with mild gastroenteritis is noted to be associated with seizure clusters [36], generally but not entirely confined to Asia [37]. Gastroenteritis is not otherwise a recognized precipitant of seizure clusters.

Risks and implications of seizure clustering

Seizure clusters, while not as life threatening as status epilepticus, have a significant impact on patient health and well being [38]. Clusters frequently result in emergency room visits and, if left untreated, have been reported to evolve into status epilepticus [39]. Seizure clusters are often treated at home with benzodiazepines, which are safe but not without adverse effects ([1] and [40]). The socioeconomic effects of seizure clustering include missed school and work, as well as greater utilization of health care resources. Furthermore, the treating physician should be aware that the presence of clusters may indicate a poor prognosis of epilepsy [41] and a higher risk of status epilepticus [42].

The clinical implications of clustering are not confined to merely the impact of a flurry of seizures. As noted, seizure clustering occurs commonly in the epilepsy monitoring unit, and the role of clusters in the localization of the epileptogenic zone has potential impact on the presurgical evaluation of patients with refractory epilepsy. It has been estimated that five seizures should be captured for identification of the primary epileptogenic focus [10]; it is unclear whether clustered seizures should count as single seizures in this analysis [43]. Available data suggest that seizures in a cluster are not independent [9], which should be a consideration during presurgical evaluation. As an outpatient history of seizure clustering indicates a high risk for clustering during epilepsy monitoring [11], this information may be useful in managing the medication schedule for inpatients with this history, to avoid clustering as much as possible.

Another implication of potential clustering of seizures is in regard to outcome of drug trials [12]. Seizure frequency reduction in drug trials implies seizure randomness. If seizures are not occurring randomly, the presumed responses to trial agents may not reflect true response; additionally, it may be appropriate to include consideration for changes in seizure patterns as well as frequency in drug trial outcome measures.

Therapeutic and future considerations

Other than the use of benzodiazepines for clusters ([1], [38-40]) and occasionally acetozolamide for catamenial patterns, no specific therapy for seizure clustering has been reported, and no preventive measures have been clearly identified. While a rare "true" clusterer may be instructed on the immediate use of rectal diazepam following a seizure, infrequent clusterers cannot identify when a first seizure will be single or the beginning of a cluster. Probability estimates based on past seizure history are not feasible in the typical clinical setting, and using a clinical definition to identify and treat clusterers will result in false positives.

Prospective data are needed to clarify these issues, and a uniform definition of clustering would provide a firm basis for future studies. As episode triggers are identified in other CDEM, these potential precipitants need to be evaluated in epilepsy. The emergence of seizure prediction modeling may both enhance our understanding of the pathophysiology of clustering and present opportunities for potential interventions to prevent seizure clusters.

Summary

The phenomenon of seizure clustering is clinically recognized and particularly prevalent in certain settings, although patients who experience all their seizures in clusters appear to be rare. Clustering may be defined clinically as a closely grouped series of seizures or, statistically, as the occurrence of seizures deviating from an expected random distribution. Seizures in clusters have demonstrated dependence, in that the presence of a seizure event influences the probability of a subsequent one. Patients with extratemporal epilepsy, history of head trauma, intractable epilepsy, and long epilepsy duration may be at particular risk for seizure clusters. Among the implications of seizure clustering are concerns for patient safety, impact on presurgical evaluation of intractable epilepsy, and effect on study design of drug trials. Prospective data are needed to better identify risks and precipitants of clustering, in an effort to identify preventive measures and enhance our understanding of the pathophysiology of seizure patterns and occurrence.

References

[1] F.E. Dreifuss, N.P. Rosman and J.C. Cloyd et al., A comparison of rectal diazepam gel and placebo for acute repetitive seizures, N Engl J Med 338 (1998), pp. 1869–1875.

[2] W.R. Gowers, Epilepsy and other convulsive diseases (2nd ed), Blackiston, Philadelphia (1901).

[3] M. Langdon-Down and W.R. Brain, Time of day in relation to convulsion in epilepsy, Lancet 1 (1929), pp. 1029–1032.

[4] F.L. Patry, The relation of time of day, sleep and other factors to the incidence of epileptic seizures, Am J Psychiatry 87 (1931), pp. 789–813.

[5] J.R. Stevens, B. Lonsbury, H. Kodama and L. Mills, Statistical characteristics of spontaneous seizure discharges recorded by radiotelemetry over 24 hour periods in man, Electroencephalogr Clin Neurophysiol 27 (1969), p. 691.

[6] J.F. Stevens, H. Kodama, B. Lonsbury and L. Mills, Ultradian characteristics of spontaneous seizure discharges recorded by radio telemetry in man, Electroencephalogr Clin Neurophysiol 31 (1971), pp. 313–325.

[7] G. Doman and R. Pelligra, A unifying concept of seizure onset and termination, Med Hypoth 62 (2004), pp. 740–745.

[8] R. Tauboll, A. Lundervold and L. Gjerstad, Temporal distribution of seizures in epilepsy, Epilepsy Res 8 (1991), pp. 153–165.

[9] S. Haut, A. Legatt, C. O'Dell, S. Moshé and S. Shinnar, Seizure lateralization during EEG monitoring in patients with bilateral foci: the cluster effect, Epilepsia 38 (1997), pp. 937–940.

[10] D. Blum, Prevalence of bilateral partial seizure foci and implications for electroencephalographic telemetry monitoring and epilepsy surgery, Electroencephalogr Clin Neurophysiol 91 (1994), pp. 329–336.

[11] S. Haut, C. Swick and S. Spencer, Seizure clustering during epilepsy monitoring, Epilepsia 43 (2002), pp. 711–715.

[12] A. Hopkins, P. Davies and C. Dobson, Mathematical models of patterns of seizures: their use in the evaluation of drugs, Arch Neurol42 (1985), pp. 463–467.

[13] C.D. Binnie, J.H.P. Aarts and M.A. Houtkooper et al., Temporal characteristics of seizures and epileptiform discharges, Electroencephalogr Clin Neurophysiol 58 (1984), pp. 498–505.

[14] M. Balish, P. Albert and W.H. Theodore, Seizure frequency in intractable partial epilepsy: a statistical analysis, Epilepsia 32 (1991), pp. 642–649.

[15] R.H. Caraballo, R.O. Cersosimo and N. Fejerman, Benign focal seizures of adolescence: a prospective study, Epilepsia 45 (2004), pp. 1600–1603.

[16] D.J. Yen, Y.H. Shih and Y.C. Guo et al., Antiepileptic drug withdrawal in patients with temporal lobe epilepsy undergoing presurgical video-EEG monitoring, Epilepsia 42 (2001), pp. 251–255.

[17] A.B. Rose, P.H. McCabe and F.G. Gilliam et al., Occurrence of seizure clusters and status epilepticus during inpatient video-EEG monitoring, Neurology 60 (2003), pp. 975–978.

[18] M.E. Newmark and S. Dubinsky, The significance of seizure clustering: a review of 343 outpatients in an epilepsy clinic. In: F.E. Dreifuss, Editor, Chronopharmacology in therapy of the epilepsies, Raven Press, New York (1990), pp. 89–103.

[19] Newmark ME, Penry JK. Catamenial epilepsy, In: Dam, Gram L, Penry JK, editors. Advances in epileptology: XII Epilepsy International Symposium. p. 433–9.

[20] J. Bauer and W. Burr, Course of chronic focal epilepsy resistant to anticonvulsant treatment, Seizure 10 (2001), pp. 239–246.

[21] J.G. Milton, J Gotman and G.M. Remillard et al., Timing of seizure recurrence in adult epileptic patients: A statistical analysis, Epilepsia 28 (1987), pp. 471–478.

[22] Haut SR, Lipton RB, LeValley AJ, Hall CB, Shinnar S. Identifying seizure clusters in patients with epilepsy. Neurology 2006: in press.

[23] R. Almquist, The rhythm of epileptic attacks and its relationship to the menstrual cycle, Acta Psychiatry Neurol Scand 30 (1955) (S105), pp. 1–116.

[24] P. Kellaway, J.D. Frost and J.W. Crawley, Time modulation of spike and wave activity in generalized epilepsy, Ann Neurol 8 (1980), pp. 491–500.

[25] B. Bandler, I.C. Kaufman, J.W. Dykens, M. Schleifer and L.N. Shapiro, Seizures and the menstrual cycle, Am J Psychol 113 (1957), pp. 704–708.

[26] T. Backstrom, Epileptic seizures in women related to plasma estrogen and progesterone during the menstrual cycle, Acta Neurol Scand 54 (1976), pp. 321–347.

[27] M.E. Newmark and J. Kiffin Penry, Catamenial epilepsy: a review, Epilepsia 21 (1980), pp. 281–300.

[28] T. Bowman, I. Leppik and E. Haus, Periodicity of seizure activity in persons with complex partial seizures, Epilepsia 25 (1984), p. 658.

[29] M. Pastrnak, The influence of lunar and seasonal periodicity on epileptic seizures, Cesk Neurol 4 (1967), pp. 268–276.

[30] Boots BN, Getis A. Point pattern analysis, 1988 Sage p. 8–92.

[31] S.R. Haut, S. Shinnar and S.L. Moshe, Seizure clustering: Risks and outcomes, Epilepsia46 (2005), pp. 146–149.

[32] D.T. Laskowitz, M.R. Sperling, J.A. French and M.J. O'Connor, The syndrome of frontal lobe epilepsy: characteristics and surgical management, Neurology 45 (1995), pp. 780–787.

[33] A. Fogarisi, J. Janszky, E. Faveret, T. Pieper and I. Tuxhorn, A detailed analysis of frontal lobe seizure semiology in children younger than 7 years, Epilepsia 42 (2001), pp. 80–85.

[34] M. Manford, D.R. Fish and S.D. Shorvon, An analysis of clinical seizure patterns and their localizing value in frontal and temporal lobe epilepsies, Brain 119 (1996), pp. 17–40.

[35] J. Bauer, Y. Ghane and D. Flugel et al., Etiology, follow up and therapy of seizure clusters in temporal lobe epilepsy and catamenial epileptic seizures, Schweiz Arch Neurol Psychiatr 143 (1992), pp. 117–134.

[36] N. Uemura, A. Okumura, T. Negoro and K. Watanabe, Clinical features of benign convulsions with mild gastroenteritis, Brain Dev 24 (2002), pp. 745–749.

[37] H. Narchi, Benign afebrile cluster convulsions with gastroenteritis: an observational study, BMC Pediatr 4 (2004), p. 2.

[38] C.T. Lombroso, Intermittent home treatment of status and clusters of seizures, Epilepsia 30 (1989), pp. S11–S14.

[39] W.G. Mitchell, Status epilepticus and acute repetitive seizures in children, adolescents and young adults: etiology, outcome and treatment, Epilepsia 37 (1996), pp. S74–S80.

[40] W.G. Mitchell, J.A. Conry and P.K. Crumrine et al., For the North American Diastat Group. An open-label study of repeated use of diazepam rectal gel (Diastat) for episodes of acute breakthrough seizures and clusters: safety, efficacy and tolerance, Epilepsia 40 (1999), pp. 1610–1617.

[41] E.A. Rodin, The prognosis of patients with epilepsy, Charles C Thomas, Springfield, IL (1969).

[42] S. Haut, S. Shinnar, S.L. Moshe, C. O'Dell and A.D.L. Legatt, The association between seizure clustering and status epilepticus in patients with intractable complex partial seizures, Epilepsia 40 (1999), pp. 1832–1834.

[43] A.B. Todorov, R.P. Lesser, S.S. Uematsu, Y.A. Yankov and A.A. Todorov Jr., Distribution in time of seizures during presurgical EEG monitoring, Neurology 44 (1994), pp. 1060–1064.

This section is an adaptation of an article written by Sheryl R. Haut, Comprehensive Epilepsy Management Center and Department of Neurology, Montefiore Medical Center and the Albert Einstein College of Medicine, Bronx, NY, originally published in Epilepsy & Behavior, Volume 8, Issue 1, February 2006, Pages 50-55

Reviewed by: Steven C. Schachter MD on 5/2008
ADVERTISEMENT
ADVERTISEMENT