Abstract
Objective
Severe malaria with neurologic involvement contributes significantly to the global burden of acquired pediatric epilepsy. We studied quantitative electroencephalographic (EEG) measures in postmalarial epileptogenesis.
Methods
A total of 186 patients, aged 6 months to 11 years, with confirmed central nervous system malaria were enrolled in a prospective observational study conducted in Chipata, Zambia. EEG data were collected during acute illness for 179 patients. Patients were followed for 12 months postdischarge with EEG and clinical reviews to assess postmalarial epilepsy (PME) outcomes. A total of 155 patients were seen for 1‐month, 144 for 6‐month, and 142 for 12‐month follow‐up. Twenty‐six patients were diagnosed with PME. We examined EEG measures (relative power, magnitude‐squared coherence [MSC], and approximate entropy [ApEn]) to identify differences between patients who developed PME and those who did not.
Results
Relative gamma power at admission was significantly greater in the PME group, whereas at 1‐month and 6‐month follow‐up it was greater in the nonepilepsy group. Alpha and beta power increased in both groups over time, suggesting a prolonged process of neurologic recovery, with greater power in both bands observed in the nonepilepsy group at 12‐month follow‐up. ApEn was greater in the nonepilepsy group at admission and 1‐month and 6‐month follow‐up. At admission and 1‐month follow‐up, we observed a significant decrease in beta and gamma MSC in the epilepsy group as compared to the nonepilepsy group. These differences were absent at 6‐month follow‐up but became more prominent again at 12‐month follow‐up.
Significance
These and related EEG measures may provide value for risk stratification of patients with severe malaria.
Keywords: cerebral malaria, epileptogenesis, malaria, postmalarial epilepsy, quantitative EEG
Key points.
EEG measures can help stratify risk for PME in pediatric patients with severe malaria with neurologic involvement.
We observed an initial peak in gamma power acutely in PME patients, followed by a decrease and normalization over 1 year of follow‐up.
Patients with PME exhibit aberrant EEG network structure measured by magnitude‐squared coherence that evolves over 1 year of follow‐up.
1. INTRODUCTION
Epilepsy accounts for approximately 1% of the global disease burden and 26% of the neurological disease burden as measured by disability‐adjusted life years, with at least 45 million people living with epilepsy worldwide. 1 , 2 Approximately 80% of these individuals reside in low‐ and middle‐income countries, with treatment gaps also tending to be disproportionately larger in these regions as compared to high‐income countries. 3 At least 25% of these patients have epilepsy attributable to preventable causes, of which a significant contribution arises from infections affecting the central nervous system (CNS). 4 In endemic regions, cerebral malaria (CM) in children secondary to Plasmodium falciparum infection can present with acute symptomatic seizures in most cases, with approximately 10%–20% of survivors of CM subsequently developing epilepsy within 2 years postinfection. 5 , 6 , 7 Notably, CM is the most severe end of the spectrum of neurologic malaria, strictly defined as P. falciparum parasitemia with coma in the absence of other coma etiology. However, this spectrum also includes features such as prolonged seizures and altered mental status or consciousness (without frank coma), termed severe malaria with neurologic/CNS involvement or CNS malaria. Recent work indicates similar rates of epilepsy and adverse impacts on neurodevelopment in patients with severe malaria with neurologic involvement as CM, pointing toward CNS involvement in general as a risk factor for neurodevelopmental sequelae, including epilepsy. 8 , 9 , 10 , 11 , 12
Few studies have examined electroencephalography (EEG) in large cohorts of patients with severe malaria, and fewer have correlated quantitative findings with neurodevelopmental outcomes, specifically the development of postmalarial epilepsy (PME). Initial analysis from the Blantyre Malaria Project Epilepsy Study (BMPES; 2005–2007) found higher maximum body temperature and acute seizures during admission to be risk factors for the development of PME. 13 Additional results from prior work suggest increased mortality correlated with clinical EEG findings during the acute illness including lower average and maximum voltage, focal slowing, slower dominant frequencies, and lack of reactivity. The presence of >4 seizures identified at admission on EEG corresponded to increased risk of cognitive or neurologic deficits at follow‐up visits. 14 A measure of theta and alpha variability (computed as the variability of the ratio between theta+alpha power and delta power) also showed an association with mortality during acute illness, but examining long‐term outcomes was not within the scope of this study. 15
Most relevant to the present report, past analyses examined EEG collected largely during acute illness and have not prospectively or longitudinally examined factors related to PME development. Previously, our group retrospectively analyzed EEG from a cohort of 70 patients enrolled in the BMPES and found that relative gamma power and maximum body temperature during admission were PME predictors, with higher relative gamma and higher fever during acute illness in patients who went on to develop epilepsy within 2 years of presentation. 16 These observations are congruent with examinations of animal models of acquired epilepsy secondary to traumatic brain injury (TBI), which have suggested an initial peak in cortical gamma activity, mediated by increased glutamatergic drive affecting fast‐spiking, parvalbumin‐positive (FS‐PV+) inhibitory interneurons, followed by a progressive decline in gamma power through epileptogenesis secondary to excitotoxicity and neuronal death. 17 , 18 , 19
Here, we test whether quantitative measures of EEG data collected during acute illness and at 1‐month, 6‐month, and 12‐month follow‐up visits can help to quantify risk of postmalarial epileptogenesis. Our main hypothesis is that we would see in our prospectively collected data a similar initial peak in gamma EEG power that is followed by a decline by the 1‐month follow‐up in those patients who develop PME (but not in others). We additionally examined approximate entropy (ApEn; a measure of signal regularity), and connectivity using magnitude‐squared coherence (MSC).
Entropy measures have been used previously in EEG to quantify depth of sleep and coma among other applications. 20 , 21 , 22 , 23 In particular, there is also evidence that entropy measures show a decrease in neonates with seizures who go on to develop epilepsy versus those who do not, possibly reflecting a reduction in signal regularity secondary to the underlying insult. 24 Aberrant connectivity, measured by MSC early in the disease course, prior to the clinical diagnosis of epilepsy, is also associated with the evolution of an underlying epileptogenic network over time. 25 Thus, our secondary hypotheses were that we would see a reduction of entropy and increased aberrant connectivity (measured by MSC) during acute malarial infection and a subsequent window of epileptogenesis in the children who ultimately developed PME.
2. MATERIALS AND METHODS
2.1. Patient selection and epilepsy diagnosis
Methods related to patient selection were detailed previously BY Patel et al. 7 Briefly, patients examined in this study were enrolled as part of a prospective observational study conducted at Chipata Central Hospital in eastern Zambia. Participants included were those presenting at between 6 months and 11 years of age with severe malaria with neurologic involvement, defined as confirmed P. falciparum malaria (by rapid diagnostic test and confirmatory smear) with neurologic symptoms (i.e., CM or CNS malaria). This age range was chosen based on the observation that severe malaria appears to affect children younger than 5 years most commonly, and is rare in African children aged 10 years and older, possibly due to increased specific acquired immunity. 26 , 27 Neurologic symptoms were defined as one or more of the following: coma (Glasgow Coma Scale [GCS] ≤ 10 or Blantyre Coma Scale [BCS] ≤ 2), complicated seizure (≥15 min in duration, focal seizures, or multiple seizures), and/or impaired consciousness without frank coma (GCS = 11–14 or BCS = 3–4). Patients were excluded if they carried a preexisting epilepsy diagnosis or had a concurrent additional CNS infection, identifiable toxic exposure, or head trauma within 24 h. As part of follow‐up visits, patients underwent screening for epilepsy utilizing validated questionnaires administered by trained research staff, and those with positive screenings were seen and evaluated by a board‐certified pediatric epileptologist (A.A.P.). Epilepsy diagnoses were made on a running basis over the course of the study utilizing International League Against Epilepsy diagnostic criteria. 28 Informed consent was obtained from caregivers for all subjects included in the study, and ethical approvals were obtained from Boston Children's Hospital (where data were hosted), the University of Zambia Biomedical Research Ethics Committee, and the National Health Research Authority of Zambia.
2.2. Data collection
The EEG data used in this study were collected during the initial admission for acute illness and at postdischarge follow‐up visits at 1 month, 6 months, and 12 months. EEG data during acute illness were collected within 48 h of admission. At each time point, a standard, roughly 40‐min EEG was recorded by trained EEG technologists using Natus hardware and XLTEK software with the standard international 10–20 system for lead placement. Data were sampled at 200 or 256 Hz, and EEGs were read by a board‐certified pediatric epileptologist (A.A.P.) for conventional visual interpretation. The average duration of admission EEG segments analyzed was 41.1 min (range = 29.4–81.1 min), of 1‐month follow‐up segments was 39.6 min (range = 28.5–61.8 min), of 6‐month follow‐up was 40.9 min (range = 23.2–65.2 min), and of 12‐month follow‐up was 42.4 min (range = 27.8–67.8 min). These distributions are shown in Figure S1.
2.3. EEG preprocessing and quantitative analysis
EEG data were initially processed through HAPPE (Harvard Automated Pre‐Processing Pipeline for EEG). 29 Steps performed in this process included notch filtering for electrical line noise (50 Hz), lowpass filtering at 80 Hz, denoising using wavelet thresholding, rejection of bad channels based on excessive artifact, and resampling all data to 200 Hz. Data were then processed through MATLAB to compute band power time series, ApEn, and MSC.
We first computed running band power time series in the conventional EEG frequency bands (delta [>.5 and ≤4 Hz], theta [>4 and ≤8 Hz], alpha [>8 and ≤13 Hz], beta [>13 and ≤25 Hz], and gamma [>25 and ≤60 Hz]) at 1‐s resolution. We then computed relative power in each frequency band over the entire epoch on a channel‐by‐channel basis (computed as the ratio between power in a specific frequency band and total signal power between .5 and 80 Hz for the corresponding time segment). Relative power was therefore reported as a ratio between 0 and 1.
In addition to relative power, we also computed average ApEn for each channel. ApEn and related entropy measures provide a quantification of statistical regularity in a signal, with higher entropy values indicative of more irregularity and therefore less predictability. The estimation of ApEn has been described previously by Pincus. 30 , 31 Intuitively, given a time series of length N samples, a vector length m (measured in samples), and a tolerance r, we define consecutive vectors of length m over the time series (total N − m + 1 vectors) and examine the degree to which these remain similar (within a tolerance r) over time. We then perform a similar examination of similarity using vectors of length m + 1, with ApEn defined as the difference between these two, that is, if similarity over time is observed with both vector lengths m and m + 1, the signal is more regular and predictable and ApEn will be low. In our analysis, we used input parameters of 2 samples as the vector length (m) and .2 multiplied by the SD of the time series as the tolerance (r), based on prior applications of ApEn to EEG data. 30 , 31
Finally, we computed MSC in each frequency band for all electrode contact pairs using Welch's overlapped segmentation approach with 3‐min windows and 50% overlap (yielding roughly 13 nonoverlapping windows per epoch). MSC is a measure of functional connectivity that quantifies the degree to which two signals maintain a constant phase relationship at a given frequency and has previously been well characterized in EEG. 32 , 33 , 34
2.4. Statistical analysis
For all time points, we aggregated data for each measure across all patients for each group (epilepsy and nonepilepsy). For each time point, we include data from each electrode contact as individual data points rather than reporting a single aggregate measure for each patient. Because electrode contacts included in the analysis at each time point are dependent on data quality, the specific contacts included may vary for a given patient over the follow‐up visits. Thus, the analysis did not strictly fall under a repeated measures paradigm, and we therefore compared measures between the two groups for each time point using two‐sample t‐tests to assess for differences between the epilepsy and nonepilepsy groups. For relative power measures and MSC, we used a significance level of p < .0102 based on Sidak correction for multiple comparisons. A significance level of p < .05 was used for ApEn.
3. RESULTS
3.1. Clinical results
A total of 184 patients were enrolled as part of this study. Of this cohort, seven died during acute illness, three of whom died before initial EEG was completed, whereas four had admission EEGs performed that were excluded in these analyses. Two additional patients were excluded because they withdrew from the study prior to the admission EEG. This left 175 patients included in the analyses who underwent EEG during the acute admission. Of these 175 patients, 69 were female and 106 were male, with 26 patients who went on to develop epilepsy (diagnosed by 12‐month follow‐up) and 149 who did not (9/26 female in the epilepsy group, 60/149 female in the nonepilepsy group). Age did not significantly differ between groups, with median age of patients in the epilepsy group being 37.6 months (range = 11.4–117.7 months) and in the nonepilepsy group being 36.4 months (range = 6.3–126.5 months, p = .67). These distributions are shown in Figure S2.
Of 26 patients who developed PME, two had an initial positive screen at 1‐month follow‐up, 11 at 6‐month follow‐up, and 13 at 12‐month follow‐up. All patients were seen within 3 months of positive screen for diagnosis confirmation. Once diagnosed, patients were started on antiseizure medications (levetiracetam, sodium valproate, carbamazepine, or phenobarbitone).
A total of 155 patients were seen for 1‐month follow‐up visits, including the 26 who developed PME and 129 who did not. A total of 144 patients were seen for a 6‐month follow‐up visit, of whom 26 patients were in the epilepsy group and 118 were in the nonepilepsy group. One patient in the nonepilepsy group was excluded from the 6‐month analysis due to poor data quality, leaving 117 in the nonepilepsy group and 143 total. At 12‐month follow‐up, 142 patients were seen, with 25 in the epilepsy group and 117 in the nonepilepsy group. One patient was excluded from the nonepilepsy group due to poor EEG data quality, leaving 116 in the nonepilepsy group and 141 total. Patient demographic information is included in Table 1. Qualitative EEG findings, including the presence of seizures, epileptiform activity, and slowing at each time point are presented in Table S1 for reference.
TABLE 1.
Patient demographics including mean age in months (with age range per group), patient sex, and EEG duration in seconds (with range per group).
| Time point | Group | Mean age, months | Sex | Mean EEG duration, s |
|---|---|---|---|---|
| Admission, n = 175 | Epilepsy, n = 26 | 41.4 (11.4–117.7) | 9 F, 17 M | 2725 (1788–4864) |
| Nonepilepsy, n = 149 | 45.1 (6.3–126.5) | 60 F, 89 M | 2424 (1763–3906) | |
| 1 month, n = 155 | Epilepsy, n = 26 | 41.4 (11.4–117.7) | 9 F, 17 M | 2403 (1709–3528) |
| Nonepilepsy, n = 129 | 45.5 (6.3–126.5) | 56 F, 73 M | 2369 (1715–3706) | |
| 6 months, n = 143 | Epilepsy, n = 26 | 41.4 (11.4–117.7) | 9 F, 17 M | 2486 (1797–3909) |
| Nonepilepsy, n = 117 | 44.5 (6.3–126.5) | 48 F, 69 M | 2444 (1391–3896) | |
| 12 months, n = 141 | Epilepsy, n = 25 | 39.8 (11.4–117.7) | 9 F, 16 M | 2445 (1898–3415) |
| Nonepilepsy, n = 116 | 46.4 (6.3–126.5) | 46 F, 70 M | 2565 (1670–4068) |
Abbreviations: EEG, electroencephalography; F, female; M, male.
3.2. EEG power and ApEn results
During the acute admission, relative gamma power was significantly higher in the PME group as compared to the nonepilepsy group (p < .01). ApEn was significantly higher in the nonepilepsy group (p < .05). We did not observe a significant difference in delta, theta, alpha, or beta bands. At 1‐month follow‐up, interestingly, we observed a significant difference in the theta, alpha, beta, and gamma bands, as well as ApEn. Relative theta power was higher in the PME group, whereas alpha, beta, and gamma power and ApEn were greater in the nonepilepsy group (p < .01 for all measures). At 6‐month follow‐up, relative gamma power and ApEn remained lower in the epilepsy group (p < .01), with no difference observed in other measures. Finally, at 12‐month follow‐up, significant differences were only seen in relative alpha and beta power (p < .01), with the nonepilepsy group being higher. Notably, no difference was seen in relative gamma power at 12‐month follow‐up. Relative alpha and beta power showed a general upward trend over time in both groups. Relative gamma power showed an initial peak at time of admission, followed by a decrease at 1‐month follow‐up, with steady increase at 6‐month and 12‐month follow‐up. These results are found in Figure 1 and Table 2.
FIGURE 1.
Mean relative delta, theta, alpha, beta, and gamma power and approximate entropy (ApEn) at admission and 1‐month, 6‐month, and 12‐month follow‐up visits compared between the postmalarial epilepsy and nonepilepsy groups. Error bars represent 95% confidence intervals. Statistical significance is denoted by an asterisk (p < .01, two‐sample t‐test with Sidak correction). Notable differences are present in relative gamma power at admission and 1‐month and 6‐month follow‐up. Alpha and beta power also show longer term differences at 12 months.
TABLE 2.
Admission, 1‐month, 6‐month, and 12‐month mean electroencephalographic measures and results of statistical testing.
| Measure | Epilepsy, n = 26 | Nonepilepsy, n = 149 | p |
|---|---|---|---|
| Admission | |||
| Relative delta | .896 ± .011 | .907 ± .003 | .065 |
| Relative theta | .058 ± .004 | .057 ± .002 | .709 |
| Relative alpha | .012 ± .001 | .014 ± .001 | .047 |
| Relative beta | .010 ± .002 | .011 ± .001 | .711 |
| Relative gamma | .018 ± .005 | .011 ± .001 | <.01 |
| ApEn | .441 ± .018 | .47 ± .008 | <.01 |
| Delta MSC | .183 ± .006 | .188 ± .002 | .118 |
| Theta MSC | .197 ± .006 | .202 ± .003 | .188 |
| Alpha MSC | .197 ± .006 | .197 ± .002 | .926 |
| Beta MSC | .153 ± .005 | .174 ± .002 | <.01 |
| Gamma MSC | .154 ± .005 | .173 ± .002 | <.01 |
| 1 month | |||
| Relative delta | .799 ± .010 | .802 ± .005 | .477 |
| Relative theta | .145 ± .007 | .128 ± .003 | <.01 |
| Relative alpha | .032 ± .003 | .037 ± .001 | <.01 |
| Relative beta | .015 ± .001 | .018 ± .001 | <.01 |
| Relative gamma | .008 ± .001 | .013 ± .001 | <.01 |
| ApEn | .521 ± .012 | .554 ± .007 | <.01 |
| Delta MSC | .171 ± .005 | .162 ± .002 | <.01 |
| Theta MSC | .204 ± .006 | .206 ± .003 | .569 |
| Alpha MSC | .184 ± .005 | .195 ± .003 | <.01 |
| Beta MSC | .141 ± .005 | .150 ± .002 | <.01 |
| Gamma MSC | .121 ± .003 | .146 ± .002 | <.01 |
| 6 months | |||
| Relative delta | .798 ± .010 | .799 ± .004 | .788 |
| Relative theta | .127 ± .006 | .120 ± .003 | .042 |
| Relative alpha | .044 ± .003 | .044 ± .001 | .784 |
| Relative beta | .020 ± .001 | .021 ± .001 | .098 |
| Relative gamma | .010 ± .001 | .014 ± .001 | <.01 |
| ApEn | .550 ± .013 | .573 ± .006 | <.01 |
| Delta MSC | .170 ± .005 | .166 ± .002 | .143 |
| Theta MSC | .205 ± .006 | .209 ± .003 | .297 |
| Alpha MSC | .196 ± .005 | .195 ± .002 | .697 |
| Beta MSC | .146 ± .005 | .136 ± .002 | <.01 |
| Gamma MSC | .120 ± .004 | .116 ± .002 | .086 |
| 12 months | |||
| Relative delta | .783 ± .009 | .777 ± .005 | .216 |
| Relative theta | .138 ± .006 | .132 ± .003 | .049 |
| Relative alpha | .045 ± .003 | .054 ± .002 | <.01 |
| Relative beta | .021 ± .001 | .024 ± .001 | <.01 |
| Relative gamma | .012 ± .001 | .013 ± .001 | .437 |
| ApEn | .575 ± .012 | .573 ± .007 | .757 |
| Delta MSC | .154 ± .005 | .162 ± .002 | <.01 |
| Theta MSC | .201 ± .005 | .202 ± .003 | .677 |
| Alpha MSC | .184 ± .005 | .199 ± .002 | <.01 |
| Beta MSC | .125 ± .004 | .132 ± .002 | <.01 |
| Gamma MSC | .100 ± .003 | .107 ± .001 | <.01 |
Note: All values are reported as mean ± 95% confidence interval. Bolded p‐values represent significance.
Abbreviations: ApEn, approximate entropy; MSC, magnitude‐squared coherence.
A comparison of averaged power spectra at each time point shows similar findings to those above, notably that power diverges at higher frequencies between the epilepsy and nonepilepsy groups at admission, with the epilepsy group having higher power in the gamma range. The opposite is seen at 1‐month and 6‐month follow‐up, with very little difference seen at 12‐month follow‐up. Averaged spectra are shown in Figure 2.
FIGURE 2.
Averaged power spectra for postmalarial epilepsy and nonepilepsy groups at each time point. Curves are depicted as mean power, with shaded regions representing 95% confidence intervals. An initial increase in gamma range power in the epilepsy group is reflected in admission spectra, following a decrease at 1 month and 6 months and normalization at 12‐months.
3.3. MSC results
During the acute admission, beta and gamma MSC were both significantly higher in the nonepilepsy group (p < .01), with no difference observed between the two groups in delta, theta, and alpha bands. At 1‐month follow up, beta and gamma MSC remained significantly higher in the nonepilepsy group, along with increased delta MSC in the epilepsy group and alpha MSC in the nonepilepsy group (p < .01). No difference was observed in the theta band. Only beta MSC showed significance at 6‐month follow‐up, with the epilepsy group being significantly higher (p < .01). Finally, delta, alpha, beta, and gamma MSC were all significantly higher in the nonepilepsy group at 12‐month follow‐up (p < .01), with no difference observed in the theta band. Delta, beta, and gamma MSC show a general downward trend over time in both groups. These results are summarized in Figure 3 and Table 2.
FIGURE 3.
Mean delta, theta, alpha, beta, and gamma magnitude‐squared coherence (MSC) at each time point compared between postmalarial epilepsy and nonepilepsy groups. Error bars represent 95% confidence interval. Statistical significance is denoted by an asterisk (p < .01, two‐sample t‐test with Sidak correction). Differences are most apparent in beta and gamma MSC, with the nonepilepsy group showing stronger connectivity at admission and 1‐month follow‐up, convergence or reversal between groups at 6 months, and divergence at 12 months.
4. DISCUSSION
In this study, we examined quantitative EEG measures in a cohort of children with severe malaria seen at a referral hospital in Chipata, Zambia with the goal of identifying differences in patients who go on to develop epilepsy versus those who do not. Our results suggest that there is a quantifiable difference between these groups in multiple measures including relative band power, ApEn, and MSC that are present during the acute admission and at 1‐month, 6‐month, and 12‐month follow‐up. We believe these results also begin to shed light on neurophysiological underpinnings of epileptogenesis in severe malaria, with potential relevance in other forms of acquired epilepsy as well.
We found that the trajectory of relative gamma power over time partially matched our hypothesis, with an initial peak in the epilepsy group during the acute admission higher than in the nonepilepsy group, followed by a relative decrease as compared to the nonepilepsy group thereafter. Interestingly, this difference was quite pronounced at 1‐month and 6‐month follow‐up, but not apparent at 12‐month follow‐up, suggesting a possible component of slow recovery but with residual changes that are more lasting. There is also a possibility that this normalization at 12 months is at least partially related to the initiation of antiseizure medication following diagnosis of epilepsy, as half (13/26) of the epilepsy diagnoses were made between 6‐ and 12‐month follow‐up. In gamma MSC, we observe stronger connectivity in the nonepilepsy group at all time points except 6‐month follow‐up. The reason for this reversal at 6 months is not entirely clear, although there is a possible contribution of the medication effect described above. Although not examined directly in this study, we speculate whether this parallels observations from animal models cited previously, which showed increased initial gamma power from acute or subacute glutamatergic drive of FS‐PV+ inhibitory interneurons, with subsequent gradual loss of these interneuron populations through epileptogenesis, followed by network‐level remodeling. Supporting the potential role of glutamatergic excitotoxicity in this process, prior work in a mouse model of P. berghei CM demonstrated attenuation of behavioral and cognitive sequelae with administration of a noncompetitive N‐methyl‐D‐aspartate receptor antagonist during acute illness. 35
Although the direction of gamma MSC findings appears at first glance counterintuitive (i.e., one could expect increased MSC in the epilepsy group as an epileptogenic network arises), it may be that decreased MSC in the epilepsy group represents increased network fragmentation and/or more network‐level resilience in the nonepilepsy group.
Looking more generally at the temporal progression of quantitative measures over the 12 months studied, we found that the greatest divergence between the epilepsy and nonepilepsy groups was observed at 1‐month follow‐up, with many measures showing significant difference. Broadly, we posit that this may be indicative of two possibilities, which may not necessarily be mutually exclusive. First, given that higher frequency activity (alpha, beta, gamma) is more prominent in the nonepilepsy group, there may be some component of prolonged recovery and/or subtle residual injury in the epilepsy group that may not be apparent with conventional clinical EEG interpretation. Second, it is also possible that active mechanisms for remodeling and recovery are more robust in the nonepilepsy group, leading to more favorable outcomes. The physiologic basis for this remains unclear and warrants further study. Recent work also suggests a role for changes in delta and theta power in distinguishing patients with and without epilepsy following a first unprovoked seizure. 36 ApEn may be a putative measure for encapsulating this capacity for flexibility, with slightly but significantly higher average ApEn observed at admission and 1‐month and 6‐month follow‐up in the nonepilepsy group, perhaps suggesting developing dysfunction in the epilepsy group.
In addition to relative power measures and ApEn, we also found differences in functional connectivity with MSC, most prominent at 1‐month follow‐up, when all bands but theta showed significance. Specifically, MSC was elevated in all bands except delta in the nonepilepsy group, which may point to broad network aberrance in the nonepilepsy group. With multiple measures showing robust changes at 1‐month follow‐up, this may be a critical window in the epileptogenic process.
Despite providing additional insight into changes in neurophysiology in patients with severe malaria with neurologic involvement over time, the overall pathogenic basis for postmalarial epileptogenesis remains poorly understood and is likely multifactorial. Published work points toward a complex interplay of microvascular sequestration and resultant ischemic injury, neuroinflammation and blood–brain barrier breakdown, and excitotoxicity. 13 , 37 , 38 In a previous examination of an animal model of PME, a measure of brain–heart coupling was proposed as a possible biomarker for epileptogenesis, with study mice showing both focal and primary generalized seizures. 39 , 40 It is likely that the pathogenesis of epilepsy in severe malaria shows commonality with other acquired epilepsies (e.g., poststroke, meningitis/encephalitis, or TBI), thus opening the door to examining the generalizability of measures identified in our analyses.
There are several limitations to our study to acknowledge. First, our analyses did not incorporate patient sleep/wake state, which may be of relevance for the 1‐month and 6‐month follow‐up recordings more than acute admission (i.e., due to the presence of coma or encephalopathy in most patients acutely). State‐specific analysis may accentuate differences between groups and increase predictive yield of the measures examined. Another route for future study we aim to pursue is examining disturbances in sleep structure and features (e.g., sleep spindles) in more detail. It is also possible that age‐related changes in EEG composition could impact our results, although this was felt to be less likely because our patients were predominantly clustered between approximately 1 and 6 years old (10th percentile and 90th percentile, respectively). Related to this, we focused our network analysis on MSC in the traditional EEG frequency bands only, and in future work we hope to examine features such as cross‐frequency coupling (e.g., state‐dependent coupling between slow oscillations and spindles), which might be expected to show aberrance in the epilepsy group. 41 , 42 Finally, most studies examining PME outcomes assess up to 2 years follow‐up, as this has been estimated as the highest risk period for epilepsy emergence. Thus, we continue to follow our cohort with a plan to assess the evolution of our findings as more of our cohort is diagnosed, as well as to better understand longer term EEG changes and more rigorously assess trends over time, including on a patient‐by‐patient basis in addition to the group‐level observations reported here.
Because the magnitude of difference between groups with some of our measures is quite small, prospective prediction is difficult with our current cohort size. We nonetheless hope that these measures, along with additional clinical data, will prove to have utility in building predictive models to risk stratify patients during acute illness and identify those who would benefit from adjunctive therapies to reduce the risk of epileptogenesis. 43 This is particularly salient in resource‐constrained settings where close and regular follow‐up visits for all patients may not be feasible.
FUNDING INFORMATION
Research reported in this publication was supported by the National Institute of Neurological Disorders and Stroke (NINDS) of the National Institutes of Health (NIH) under award number K23NS118051 (A.A.P.), with additional support from the National Institute of Environmental Health Sciences pilot grant P30 ES 000002 (A.A.P.), and NIH/NINDS grants R01NS102176, R01NS111057, and R35NS122265 (G.L.B.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
CONFLICT OF INTEREST STATEMENT
A.A.P. received research grants from the National Institutes of Health (NIH)/National Institute of Neurological Disorders and Stroke (NINDS) and National Institute of Environmental Health Sciences for the proposed work, has served as a paid consultant for the World Health Organization and Noviguide/Global Strategies, and serves on the advisory board of the ROW Foundation, which supplies antiseizure medications through a donation grant to Zambia. G.L.B. has received research grants from NIH/NINDS to support the proposed and other work, is on the editorial board of Lancet Neurology and the Lundbeck Foundation's Neurotorium, and is a consultant for Blue Spark Technologies. A.R. is cofounder of Neuromotion, PrevEp, and Galibra Neuroscience; has consulted for, served on advisory boards for, or received research support from AbbVie, Autifony, BioMarin, CRE Medical, Encoded, Epihunter, Neuroelectrics, Neural Dynamics, NeuroRex, Ovid, Roche, and Takeda; and is listed as inventor on patents pertaining to brain stimulation, drug delivery, and gene therapy that are unrelated to the present work. The remaining authors have no conflicts of interest to disclose and declare that the research was conducted in the absence of any relevant commercial or financial relationships. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.
Supporting information
Figure S1.
ACKNOWLEDGMENTS
We thank the participating families in Chipata, Zambia as well as Dr. Mbinga Mbinga and the Pediatrics Department at Chipata Central Hospital, where the research was conducted.
Contributor Information
Rasesh B. Joshi, Email: rasesh.joshi@childrens.harvard.edu.
Archana A. Patel, Email: archana.patel@seattlechildrens.org.
DATA AVAILABILITY STATEMENT
The full dataset is available upon review of reasonable written request and establishment of an approved data‐sharing agreement from relevant institutions.
REFERENCES
- 1. GBD 2016 Epilepsy Collaborators . Global, regional, and national burden of epilepsy, 1990–2016: a systematic analysis for the global burden of disease study 2016. Lancet Neurol. 2019;18(4):357–375. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. de Boer HM, Mula M, Sander JW. The global burden and stigma of epilepsy. Epilepsy Behav. 2008;12(4):540–546. [DOI] [PubMed] [Google Scholar]
- 3. Espinosa‐Jovel C, Toledano R, Aledo‐Serrano Á, García‐Morales I, Gil‐Nagel A. Epidemiological profile of epilepsy in low income populations. Seizure. 2018;56:67–72. [DOI] [PubMed] [Google Scholar]
- 4. Thurman DJ, Begley CE, Carpio A, Helmers S, Hesdorffer DC, Mu J, et al. The primary prevention of epilepsy: a report of the prevention task force of the international league against epilepsy. Epilepsia. 2018;59(5):905–914. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Alizadeh Khatir A, Sepidarkish M, Daryabari Y, Taghipour A, Mollalo A, Aghapour S, et al. Malaria infection and the risk of epilepsy: a meta‐analysis. Parasitology. 2023;150(4):382–390. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Ngoungou EB, Preux P‐M. Cerebral malaria and epilepsy. Epilepsia. 2008;49(Suppl 6):19–24. [DOI] [PubMed] [Google Scholar]
- 7. Patel AA, Birbeck GL, Mazumdar M, Mwanza S, Nyirongo R, Berejena D, et al. Identifying biomarkers for epilepsy after cerebral malaria in Zambian children: rationale and design of a prospective observational study. BMJ Open. 2022;12(7):e062948. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Carter JA, Mung'ala‐Odera V, Neville BGR, Murira G, Mturi N, Musumba C, et al. Persistent neurocognitive impairments associated with severe falciparum malaria in Kenyan children. J Neurol Neurosurg Psychiatry. 2005;76(4):476–481. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Carter JA, Ross AJ, Neville BGR, Obiero E, Katana K, Mung'ala‐Odera V, et al. Developmental impairments following severe falciparum malaria in children. Trop Med Int Health. 2005;10(1):3–10. [DOI] [PubMed] [Google Scholar]
- 10. Bangirana P, Musisi S, Boivin MJ, Ehnvall A, John CC, Bergemann TL, et al. Malaria with neurological involvement in Ugandan children: effect on cognitive ability, academic achievement and behaviour. Malar J. 2011;10:334. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Carter JA, Neville BGR, White S, Ross AJ, Otieno G, Mturi N, et al. Increased prevalence of epilepsy associated with severe falciparum malaria in children. Epilepsia. 2004;45(8):978–981. [DOI] [PubMed] [Google Scholar]
- 12. Kariuki SM, Abubakar A, Newton CRJC, Kihara M. Impairment of executive function in Kenyan children exposed to severe falciparum malaria with neurological involvement. Malar J. 2014;13:365. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Birbeck GL, Molyneux ME, Kaplan PW, Seydel KB, Chimalizeni YF, Kawaza K, et al. Blantyre malaria project epilepsy study (BMPES) of neurological outcomes in retinopathy‐positive paediatric cerebral malaria survivors: a prospective cohort study. Lancet Neurol. 2010;9(12):1173–1181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Clark DJ, Bond C, Andrews A, Muller DJ, Sarkisian A, Opoka RO, et al. Admission clinical and EEG features associated with mortality and long‐term neurologic and cognitive outcomes in pediatric cerebral malaria. Neurology. 2023;101(13):e1307–e1318. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Andrews A, Zelleke T, Harrar D, Izem R, Gai J, Postels D. Theta‐alpha variability on admission EEG is associated with outcome in pediatric cerebral malaria. J Clin Neurophysiol. 2023;40(2):136–143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Patel AA, Jannati A, Dhamne SC, Sapuwa M, Kalanga E, Mazumdar M, et al. EEG markers predictive of epilepsy risk in pediatric cerebral malaria – a feasibility study. Epilepsy Behav. 2020;113:107536. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Pavlov I, Huusko N, Drexel M, Kirchmair E, Sperk G, Pitkänen A, et al. Progressive loss of phasic, but not tonic, GABAA receptor‐mediated inhibition in dentate granule cells in a model of post‐traumatic epilepsy in rats. Neuroscience. 2011;194:208–219. [DOI] [PubMed] [Google Scholar]
- 18. Nichols J, Bjorklund GR, Newbern J, Anderson T. Parvalbumin fast‐spiking interneurons are selectively altered by paediatric traumatic brain injury. J Physiol. 2018;596(7):1277–1293. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. MacMullin P, Hodgson N, Damar U, Lee HHC, Hameed MQ, Dhamne SC, et al. Increase in seizure susceptibility after repetitive concussion results from oxidative stress, parvalbumin‐positive interneuron dysfunction and biphasic increases in glutamate/GABA ratio. Cereb Cortex. 2020;30(12):6108–6120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Burioka N, Miyata M, Cornélissen G, Halberg F, Takeshima T, Kaplan DT, et al. Approximate entropy in the electroencephalogram during wake and sleep. Clin EEG Neurosci. 2005;36(1):21–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Frohlich J, Chiang JN, Mediano PAM, Nespeca M, Saravanapandian V, Toker D, et al. Neural complexity is a common denominator of human consciousness across diverse regimes of cortical dynamics. Commun Biol. 2022;5(1):1374. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Amorim E, van der Stoel M, Nagaraj SB, Ghassemi MM, Jing J, O'Reilly U‐M, et al. Quantitative EEG reactivity and machine learning for prognostication in hypoxic‐ischemic brain injury. Clin Neurophysiol. 2019;130(10):1908–1916. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Joshi RB, Gaspard N, Goncharova II, Pavlova M, Duckrow RB, Gerrard JL, et al. Progressive change in sleep over multiple nights of intracranial EEG monitoring. Clin Neurophysiol. 2016;127(5):2302–2307. [DOI] [PubMed] [Google Scholar]
- 24. Lu W‐Y, Chen J‐Y, Chang C‐F, Weng W‐C, Lee W‐T, Shieh J‐S. Multiscale entropy of electroencephalogram as a potential predictor for the prognosis of neonatal seizures. PLoS One. 2015;10(12):e0144732. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Spencer SS. Neural networks in human epilepsy: evidence of and implications for treatment. Epilepsia. 2002;43(3):219–227. [DOI] [PubMed] [Google Scholar]
- 26. Severe malaria. Trop Med Int Health. 2014;19(Suppl 1):7–131. [DOI] [PubMed] [Google Scholar]
- 27. Farrar J. Neurological aspects of tropical disease. J Neurol Neurosurg Psychiatry. 2000;68(2):135–136. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Scheffer IE, Berkovic S, Capovilla G, Connolly MB, French J, Guilhoto L, et al. ILAE classification of the epilepsies: position paper of the ILAE Commission for Classification and Terminology. Epilepsia. 2017;58(4):512–521. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Gabard‐Durnam LJ, Mendez Leal AS, Wilkinson CL, Levin AR. The Harvard automated processing pipeline for electroencephalography (HAPPE): standardized processing software for developmental and high‐artifact data. Front Neurosci. 2018;12:97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Pincus SM. Approximate entropy as a measure of system complexity. Proc Natl Acad Sci. 1991;88(6):2297–2301. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Pincus S. Approximate entropy (ApEn) as a complexity measure. Chaos. 1995;5(1):110–117. [DOI] [PubMed] [Google Scholar]
- 32. Carter G, Knapp C, Nuttall A. Estimation of the magnitude‐squared coherence function via overlapped fast Fourier transform processing. IEEE Trans Audio Electroacoust. 1973;21(4):337–344. [Google Scholar]
- 33. Carter GC. Coherence and time delay estimation. Proc IEEE. 1987;75(2):236–255. [Google Scholar]
- 34. Zaveri HP, Williams WJ, Sackellares JC, Beydoun A, Duckrow RB, Spencer SS. Measuring the coherence of intracranial electroencephalograms. Clin Neurophysiol. 1999;110(10):1717–1725. [DOI] [PubMed] [Google Scholar]
- 35. De Miranda AS, Brant F, Vieira LB, Rocha NP, Vieira ÉLM, Rezende GHS, et al. A neuroprotective effect of the glutamate receptor antagonist MK801 on long‐term cognitive and behavioral outcomes secondary to experimental cerebral malaria. Mol Neurobiol. 2017;54(9):7063–7082. [DOI] [PubMed] [Google Scholar]
- 36. Segovia‐Oropeza M, Rauf EHU, Heide E, Focke NK. Quantitative EEG signatures in patients with and without epilepsy development after a first seizure. Epilepsia Open. 2025;10:epi4.13128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Idro R, Marsh K, John CC, Newton CRJ. Cerebral malaria: mechanisms of brain injury and strategies for improved neurocognitive outcome. Pediatr Res. 2010;68(4):267–274. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Nishanth G, Schlüter D. Blood–brain barrier in cerebral malaria: pathogenesis and therapeutic intervention. Trends Parasitol. 2019;35(7):516–528. [DOI] [PubMed] [Google Scholar]
- 39. Bahari F, Ssentongo P, Schiff SJ, Gluckman BJ. A Brain–Heart Biomarker for Epileptogenesis. J Neurosci. 2018;38(39):8473–8483. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Ssentongo P, Robuccio AE, Thuku G, Sim DG, Nabi A, Bahari F, et al. A murine model to study epilepsy and SUDEP induced by malaria infection. Sci Rep. 2017;7(1):43652. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Staresina BP, Bergmann TO, Bonnefond M, van der Meij R, Jensen O, Deuker L, et al. Hierarchical nesting of slow oscillations, spindles and ripples in the human hippocampus during sleep. Nat Neurosci. 2015;18(11):1679–1686. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Staresina BP, Niediek J, Borger V, Surges R, Mormann F. How coupled slow oscillations, spindles and ripples coordinate neuronal processing and communication during human sleep. Nat Neurosci. 2023;26(8):1429–1437. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Klein P, Friedman A, Hameed MQ, Kaminski RM, Bar‐Klein G, Klitgaard H, et al. Repurposed molecules for antiepileptogenesis: missing an opportunity to prevent epilepsy? Epilepsia. 2020;61(3):359–386. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S1.
Data Availability Statement
The full dataset is available upon review of reasonable written request and establishment of an approved data‐sharing agreement from relevant institutions.