Hyperventilation Revisited: Physiological Effects and Efficacy on Focal Seizure Activation in the Era of Video-EEG Monitoring


Address correspondence and reprint requests to Dr. A. C. Sakamoto at Rua Adolfo Serra 1725, casa 13 Jardim Santa Ângela, Ribeirão Preto-SP CEP:14020-263, Brazil. E-mail: sakamoto@fmrp.usp.br


Summary: Purpose: Hyperventilation is an activation method that provokes physiological slowing of brain rhythms, interictal discharges, and seizures, especially in generalized idiopathic epilepsies. In this study we assessed its effectiveness in inducing focal seizures during video-EEG monitoring.

Methods: We analyzed the effects of hyperventilation (HV) during video-EEG monitoring (video-EEG) of patients with medically intractable focal epilepsies. We excluded children younger than 10 years, mentally retarded patients, and individuals with frequent seizures.

Results: We analyzed 97 patients; 24 had positive seizure activation (PSA), and 73 had negative seizure activation (NSA). No differences were found between groups regarding sex, age, age at epilepsy onset, duration of epilepsy, frequency of seizures, and etiology. Temporal lobe epilepsies were significantly more activated than frontal lobe epilepsies. Spontaneous and activated seizures did not differ in terms of their clinical characteristics, and the activation did not affect the performance of ictal single-photon emission computed tomography (SPECT).

Conclusions: HV is a safe and effective method of seizure activation during monitoring. It does not modify any of the characteristics of the seizures and allows the obtaining of valuable ictal SPECTs. This observation is clinically relevant and suggests the effectiveness and the potential of HV in shortening the presurgical evaluation, especially of temporal lobe epilepsy patients, consequently reducing its costs and increasing the number of candidates for epilepsy surgery.

Hyperventilation (HV) constitutes a classic activation procedure of the electroencephalogram (EEG) that usually provokes physiological slowing of the brain rhythms, more intense and abrupt in children from 8 to 12 years old (1–4). Although this effect also can be observed in normal individuals, it is, however, more prevalent and pronounced in patients with epilepsy (1).

In addition to this physiological response, HV may activate interictal discharges and seizures in epilepsy patients, provoking 3-Hz spike-and-wave complexes (SWCs) in ∼80% of the patients with idiopathic generalized epilepsies (5,6), and slow spike and wave complexes (SSWC) in ≤50% of those patients with symptomatic generalized epilepsies who are able to perform the activation procedure adequately (7–9).

Although the ultimate mechanisms implicated in the effects of HV remain debatable (10), the observation that HV precipitates seizures in individuals with epilepsy was well known long before its introduction in the practice of clinical EEG (11,12). This activation method, although classically quoted as typical of idiopathic generalized epilepsies, also can be effective in other types of epilepsies, including focal epilepsies, where positive activation is obtained in 6–9% of the individuals (13,14). However, objective data addressing its real contribution as a method of activation of focal seizures are still lacking, even though some epilepsy centers routinely apply it during video-EEG monitoring (video-EEG). In this study we assessed the effectiveness of HV in inducing focal seizures in the setting of video-EEG and in the specific scenario of the presurgical evaluation.


We retrospectively analyzed all patients with medically intractable focal epilepsies submitted to video-EEG between August 1988 and March 2001. We excluded from this analysis individuals with history of frequent spontaneous seizures where activation procedures were judged unnecessary, mentally retarded patients, and children younger than 10 years, because of their variable performance during the activation procedure.

Continuous video-EEG recordings were obtained through Grass gold-coated disk electrodes applied to the scalp according to the 10–20 international system plus closed-spaced electrodes, plus additional sphenoidal electrodes in those cases of suspected temporal lobe epilepsy (TLE). All patients or their parents gave informed consent for video-EEG, which routinely included activation by HV. The recordings were acquired in 64-channel Vanguard digital EEG systems equipped with LaMont amplifiers (LaMONT Medical Inc., Madison, WI, USA).

Auras without ictal electrographic patterns on surface EEG were not considered for the purposes of this analysis. The activation was considered positive whenever the patients had a clinical seizure accompanied by ictal EEG changes during the HV or within 5 min after the completion of the test. According to these criteria, patients were divided in two groups: those with positive seizure activation (PSA group) and with negative seizure activation (NSA group).

All patients had their antiepileptic drugs (AEDs) tapered and discontinued during the evaluation. They were requested to perform HV by vigorously breathing for 5 min, at a minimum breath frequency (BF) of 20 incursions/min. The activation procedure was repeated approximately every 3 hours, from 6 a.m. to 12 midnight, with a free interval from 12 midnight to 6 a.m., and until clinical seizures were recorded.

The video-EEG was analyzed for the occurrence of spontaneous and activated seizures, and the following additional parameters were registered: duration of EEG seizures, ictal signs and symptoms, and ictal EEG patterns. When focal seizures evolved to secondarily generalized seizures, both partial and generalized periods were considered separately. We additionally studied the following variables: respiratory frequency during HV, degree of slowing in the EEG, and, whenever applicable, the latency between HV and seizure onset.

The degree of slowing was classified in four levels: absent, discrete, moderate, and intense. It was considered absent when no change occurred in the frequency and amplitude of the background rhythms in comparison to the baseline of 5 min before HV; discrete, when there was only posterior slowing and/or 20% increase in the theta activity, with or without occasional delta waves; moderate, when there was an increase of up to 40% of the theta and delta activities; and intense, when delta activity occupied >40% of the recording time.

All patients had high-resolution 1.5-T magnetic resonance imaging (MRI) of the brain. As part of our presurgical evaluation protocol for focal refractory epilepsies, most patients (whenever ictal injections were achieved) additionally had ictal single-photon emission computed tomography (SPECT) through injection of 99Tc-ethyl cysteinate dimer, with images acquired through a dual-head Siemens gamma camera.

For comparison between the two groups, we applied the χ2 and Fisher's exact test for categoric variables and independent samples t test for comparison of means (continuous variable). Significance level was considered at p < 0.05.


We analyzed the effects of HV in 97 patients with medically intractable focal epilepsies submitted to continuous video-EEG during an average of 3 days. The clinical and demographic data of this population are shown in Table 1. The PSA group included 24 patients (13 men and 11 women) who had one to three HV-induced seizures. Their ages ranged from 16 to 50 years (mean, 30 years), and their ages at epilepsy onset varied from 8 months to 24 years old. The NSA group included 73 patients (41 male and 32 female patients) with negative activation, their ages ranging from 11 to 68 years (mean, 30.9 years), and their age at epilepsy onset varying from 5 months to 51 years. No correlation was found between activation of seizures and gender (p = 0.77), age (p = 0.49), duration of epilepsy (p = 0.65), age at onset (p = 0.34), and frequency of seizures (p = 0.38).

Table 1. Clinical and demographic data
All patients VariableN = 97HV seizure activationp Level
(n = 73)
(n = 24)
 Male55 (56.7%)42 (57.5%)13 (54.2%) 
 Female42 (43.3%)31 (42.5%)11 (45.8%) 
Age at surgery (yr)0.49
 Mean (range)30.9 (11–68)30.9 (11–68)30.8(16–50) 
 ≤20 yr22 (22.7%)18 (24.7%)4 (16.7%) 
 >20 yr75 (77.3%)55 (75.3%)20 (83.3%) 
Epilepsy duration (yr)0.65
 Mean (range)19.4 (1–61)18.9 (1–61)21.0 (1–40) 
 ≤20 yr59 (60.8%)45 (61.6%)14 (58.3%) 
 >20 yr38 (39.2%)28 (38.4%)10 (41.7%) 
Age at epilepsy onset (yr)0.34
 Mean (range)11.5 (0.1–51)12.1 (0.5–51)9.7 (0.1–33) 
 Mean (SD)25.8 (44.5)27.5 (49.2)11.5 
 ≤4 seizures39 (40.2%)31 (42.5%)8 (33.3%) 
 5–10 seizures13 (13.4%)9 (12.3%)4 (16.7%) 
 >10 seizures45 (46.4%)33 (45.2%)12 (50.0%) 
Duration of video-EEG (days)0.27
 Mean (range)3.3 (1–9)3.3 (1–7)3.4 (1–9) 
No. of seizures recorded0.93
 Mean (range)7.3 (1–53)7.9 (2–53)5.4 (1–21) 
Generalized seizures0.68
 Yes45 (46.4%)33 (45.2%)12 (50.0%) 
 No52 (53.6%)40 (54.8%)12 (50.0%) 
 Mean (range)1.4 (0–20)1.1 (0–7)2.3 (0–20) 
Focus localization0.15
 Frontal20 (20.6%)19 (26.0%)1 (4.2%) 
 Temporal61 (62.9%)43 (58.9%)18 (75.0%) 
 Parietooccipital13 (13.4%)9 (12.3%)4 (16.7%) 
 Multilobar3 (3.1)2 (2.7%)1 (4.2%) 
 MTS41 (42.3%)28 (38.4%)13 (54.2%) 
 MCD10 (10.3%)9 (12.3%)1 (4.2%) 
 Lesion21 (21.6%)16 (21.9%)5 (20.8%) 
 Undetermined25 (25.8%)20 (27.4%)5 (20.8%) 

Based on clinical and MRI data, the following etiologies were encountered in the two groups of patients: mesial temporal sclerosis, tumor, gliosis, malformations of cortical development (cortical heterotopia, Taylor's type focal cortical dysplasia, microdysgenesis), vascular lesions (cavernous angioma and arteriovenous malformation), dual pathology, and “cryptogenic focal epilepsy” (normal MRI studies). No correlation was observed between activation and etiology (p = 0.47).

Regarding seizure origin defined by ictal EEG onset during monitoring, TLE predominated in both groups. The ictal localization was classified as frontal, temporal, parietal, occipital, and multilobar/nonlocalizable. Although no correlation was seen between overall localization and activation (p = 0.47), the comparison between the two larger subgroups showed that patients with temporal focus had a greater probability of having seizure activation during HV (p = 0.03).

Characteristics of activated seizures

In the PSA group, the number of seizures ranged from two to 10 per patient. In total, 113 seizures, including spontaneous and activated ones, were recorded. Of these, 32 (28.3%) were activated by HV, including 20 complex-partial seizures and 12 partial seizures with secondary generalization. The spontaneous seizures occurred during wakefulness and sleep. After the exclusion of 18 spontaneous seizures that occurred during sleep, a total of 95 wakefulness seizures were analyzed; of those, 28 (33.7%) were activated by HV. Among 42 partial seizures with secondary generalization, 12 (28.6%) were activated. No correlation was noted between recording of generalized seizures and activation during HV (p = 0.68).

Considering the 63 patients with TLE, 45 seizures were spontaneous, whereas 18 (28.5%) were activated. Among 41 patients with mesial temporal sclerosis, 13 (31.7%) had seizures activated by HV, whereas this happened in only one of 20 patients with frontal lobe epilepsy.

The latencies for the activation of the seizures by HV are shown in Fig. 1. An ascendant curve was noted until minute 4, when the maximum activation was observed, with an abrupt decrease after this time. Only two patients continued the procedure beyond 5 min and had activation of seizures at that time. No test was prolonged >6 min. After HV, four seizures were recorded as much as 3 min after the end of the procedure.

Figure 1.

Number of activated seizures during and immediately after the hyperventilation test.

The respiratory frequency during HV varied from 13 to 54 incursions per minute (average, 30). Only three patients had breathing frequency of less than 20, two had 18, and the other, 13.

The degree of background slowing showed 11 patients with discrete, 10 patients with moderate, and one patient with intense slowing. Ten patients had no background slowing. Most seizures (21 of 32) occurred during no slowing or discrete background slowing, whereas 11 seizures occurred during moderate (10 seizures) or intense background slowing (only one seizure).

Comparison between spontaneous and activated seizures in the PSA group

Because three patients had only activated seizures, this comparison between spontaneous and activated seizures was possible for only 21 patients. The clinical features of the spontaneous and the activated seizures were similar. The percentage of spontaneous and activated seizures that evolved with secondary generalization was similar, 37.0% (30 of 81 seizures) and 37.5% (12 of 32 seizures), respectively. Very little difference was seen between the spontaneous and activated seizures in terms of their mean duration, 63.5 and 64.7 s, respectively.

In terms of the electrographic characteristics, the ictal EEGs during the spontaneous and activated seizures were similar for the majority of the patients (19 of 21), being focal in two patients, regionalized in 14, hemispheric in two, and nonlateralized in one patient. In only two patients did the activated seizures differ slightly from the spontaneous ones, their EEG patterns appearing more restricted to the sphenoidal electrodes, whereas in the spontaneous seizures, the ictal recordings tended to involve all the temporal contacts.

Eight of the 24 PSA patients had ictal SPECT performed through injections of the radiotracer during the course of the activated seizures. The results, based on nonquantitative interpretation of the images, are shown in Table 2. All five patients with structural lesions in the temporal lobe had a focal increase of the cerebral blood flow (CBF) concordant with the localization of the lesion and the ictal recording. The fact that the seizure was induced by HV did not affect the analysis of the SPECT, although the hyperperfusion extended to the ipsilateral caudate and thalamus (Fig. 2).

Table 2. SPECT during activated seizures
PatientHyperperfusion locationHyperperfusion degreeIctal EEG locationStructural lesionEEG slowing degreeSeizure generalization
  1. R, right; L, left; MTS, mesial temporal sclerosis; SPECT, single-photon emission computed tomography.

1R temporal lobeModerateR temporalR MTSAbsent
2Cerebral cortex diffuselyDiscreteDiffuseAbsentAbsent+
3L temporal lobe, anterior, mesial and lateral partsModerateL temporalL MTSDiscrete+
4L temporal lobe, inferior, mesial and anterior partsModerateL mesial temporalL parahippocampal tumorModerate
5L temporal lobe, lateral, mesial and anterior partsModerateL temporalL MTSDiscrete
6L temporal lobe, lateral, mesial and anterior partsModerateL temporalL MTSDiscrete
7L temporal lobe, anterior neocortical regionDiscreteR hemisphericExtensive cortical bilateral parietooccipital heterotopiaModerate+
8L temporal lobe, mesial neocortical regionModerateL hemisphericAbsentAbsent
Figure 2.

Ictal single-photon emission computed tomography during an activated seizure.

The degree of slowing in the EEG sample preceding the activated seizures was absent or discrete in six patients and moderate in another two patients, suggesting that no correlation exists between this parameter and the activation of the seizures. The degree of hyperperfusion on the ictal SPECT varied from moderate to intense, being discrete in only two cases.


In this study we gathered evidence suggesting that HV is a useful activation method in the Video-EEG Unit, inducing habitual seizures in 24.7% of the patients with medically intractable focal epilepsies, activating 28.3% of all focal seizures during monitoring, and provoking 33.7% of all seizures recorded during wakefulness in the PSA group.

If we consider the specific type of epilepsy, 29.5% of those patients with TLE and 31.7% of those with mesial temporal sclerosis were activated by HV. In this series, the frontal lobe epilepsies were not as susceptible to activation, whereas for the posterior cortex epilepsies, the number was too small to allow any conclusion. It is important to emphasize here that TLE is the most frequent type of focal epilepsy, and, moreover, is the most frequent type of medically intractable epilepsy submitted to video-EEG and presurgical evaluation.

Some decades ago, anecdotal reports were made of focal seizures precipitated by HV (11,12,15), and a few studies showed an increase in the frequency of focal discharges (14). Others also observed activation of seizures in a minority of their patients, already emphasizing the semiologic similarities between the spontaneous and the induced seizures (16,17). Since the 1970s, attention had already been drawn to the fact that the activation of discharges and of focal clinical seizures was not as rare as previously thought (18). In their series, 11% of the patients with complex partial seizures had striking activation of spikes (6.6%) and clinical seizures (4.4%). Only one recent study did not confirm the activation effects of HV. However, this conclusion was based on a single 5-min test per patient, on an outpatient basis, during a routine EEG examination (19).

Twenty-three percent of our patients had positive activation of seizures, but we must consider that our subjects, for the purpose of video-EEG, had their AEDs routinely tapered off. This finding is similar to that of Schüller et al. (17), who found HV-positive seizure activation in 20% (two of 10) of their patients with drug-resistant focal epilepsies.

The characteristics of the activated seizures were similar to those of the spontaneous ones, both clinically and electrographically, and duration and ictal EEG localization also were not different. HV induced mainly partial seizures, contrary to what would be expected. This observation is of utmost importance, considering that video-EEG is an expensive method and frequently limits the number of patients that can potentially benefit from epilepsy surgery. Activation methods that can shorten the presurgical evaluation without interfering qualitatively in the type and duration of seizures have, therefore, major clinical impact.

The activation effect occurred in an ascendant fashion, from the beginning of HV until minute 4, the most efficient time for seizure activation. This result shows that HV in video-EEG units should last ≥4 min. Even though few patients tolerated >5 min of HV, no significant increase was seen in the number of seizures after minute 5. Conversely, seizures can occur even ≤3 min after the end of this proof.

In the literature, it is suggested that HV should be performed for ≥3 min (18); others, however, consider 5 min as the minimum necessary time for activation of focal seizures (20,21).

It also has been suggested that frontotemporal seizures would be more easily activated than focal seizures originating in other cortical areas (22). In this retrospective study, we did not emphasize the effect of HV on interictal discharges. Regarding this specific point, reports have mentioned the influence of HV usually increasing the interictal discharges, but this aspect was not analyzed in our study, which mainly addressed the better understanding of its influence over epileptic seizures. We observed that most of the activated seizures were originated from the temporal lobe, and clearly, temporal lobe seizures were more frequently activated than frontal lobe seizures.

One interesting point to consider further is the possibility of reflex influences originating from the respiratory tract during HV in TLE. It was previously demonstrated in frogs and turtles that air insufflation into the nasal cavities triggers discharges in experimentally created foci of forebrain structures (23). In clinical trials involving patients with unilateral temporal EEG abnormalities, nasal instead of oral HV was much more effective in activating epileptiform discharges. Moreover, when patients had one nasal cavity occluded, unilateral nasal HV provoked activation 2.5 times greater when the nasal cavity ipsilateral to the epileptogenic focus was used (24). Another clinical study revealed that the same activating effect could be obtained in TLE patients through air insufflation into the superior nasal meati and without pulmonary HV. Conversely, local anesthesia of the mucosal membrane of the superior nasal meatus led to temporary suppression of the activation in 91.3% of the patients. All this evidence associated with the very short latency (30–60 s after the start of deep breathing) for the activating effect to take place favors the possibility of reflex mechanisms triggered by mechanical stimulation of the olfactory epithelium in opposition to the metabolic hypothesis in the group of TLE (25).

The general mechanisms of the activation effect of HV are still poorly understood. The hypoxia theory, suggesting that the EEG slowing is due to vasoconstriction and diminution of oxygen and dextrose supply to the cerebral cortex (26), has many arguments against it: (a) differences exist in the qEEG changes due to HV and hypoxia (27); and (b) EEG changes are independent of the concentration of inspired oxygen (28) and the reduction of cerebral blood flow (29).

The hypocapnia theory, implying that the low levels of carbon dioxide would lead to the predominance of the nonspecific thalamic projection system over the activating reticular ascending system (30,31), also has been questioned by observations that the slow waves on the EEGs of patients and normal individuals appear in a wide range of arterial and expired air CO2 tension (32,33). Moreover, indomethacin produces hypoperfusion similar to that caused by hypocapnia without provoking any slowing on the EEG (34).

Some evidence indicates that activation of EEG and slowing have independent mechanisms. This is supported by the observation that intravenous diazepam (DZP) prevents activation of epileptiform discharges without affecting the slow waves (35). Conversely, the slowing also did not modify after the control of absence seizures by medication (36). Our data clearly showed that occurrence of seizures was independent of the slowing, thus favoring the hypothesis of the existence of independent mechanisms.

Some investigators disagree with the unification of the two different effects, believing that independent mechanisms probably play a role. This is supported by the observation that intravenous DZP prevents activation of epileptiform discharges without altering the slow-waves component (35). Conversely, the slowing also did not modify after the control of absence seizures by medication (36). In this study, despite the variable degree of slowing, the occurrence of seizures was independent of it, thus favoring the hypothesis of the existence of independent mechanisms.

HV may reduce CBF by 30% and possibly impair the results of ictal SPECTs. However, we observed that scans obtained during HV-activated seizures (Table 1) were similar to those from spontaneous ones (Fig. 2). This observation also carries great importance for the presurgical evaluation of medically intractable epilepsies, showing that HV can optimize the clinical use of ictal SPECT without affecting the interpretation of the test. Katayama et al. (37) previously suggested that an impaired vasoconstrictive response to HV in the epileptogenic area could contribute to the relative hyperperfusion observed there, but this seems not to be the case; the interictal SPECT during HV revealed that only the patients that had increase of the spiking rate had hyperperfusion, thus suggesting that the epileptiform activity would be the main factor contributing to the regional hyperperfusion observed during HV (38).

In conclusion, in the evaluation of patients with focal epilepsy refractory to clinical treatment, the time spent in video-EEG should be as short as possible. Tapering off the medication is one of the methods for cutting down the duration of monitoring. HV has revealed an important and simple procedure for seizure activation, especially in cases of TLE. Four minutes is the minimum duration of HV, and the degree of slowing in the EEG does not appear to be the determining factor of seizure induction. Ictal SPECTs obtained during induced seizures did not differ from those obtained during spontaneous seizures, suggesting that HV can be a helpful method to optimize the obtaining of ictal SPECTs in the monitoring unit. The method also proved to be safe; it did not increase the duration of the seizure or the risk of generalization.


Acknowledgment:  We thank Dr. Roger Walz for his assistance in statistical analysis.